optimisation of biofuels production from microalgal biomass

Optimisation of Biofuels Production from

Microalgal Biomass

A thesis submitted to The University of Manchester for the degree of

Doctor of Philosophy

in the Faculty of Science and Engineering

2018

Gonzalo M. Figueroa Torres

School of Chemical Engineering and Analytical Science

3

List of Contents

List of Contents ........................................................................................... 3

List of Tables ............................................................................................... 6

List of Figures .............................................................................................. 7

Abbreviations ............................................................................................... 9

Nomenclature ............................................................................................. 10

Abstract ...................................................................................................... 13

Declaration ................................................................................................. 14

Copyright Statement ................................................................................. 15

Acknowledgements .................................................................................... 17

Chapter 1. Introduction and Research Contribution ......................... 19

1.1. Introduction. ....................................................................................................... 19

1.2. Biofuels – An alternative energy source. ........................................................... 19

1.2.1. Butanol – A promising biofuel. ....................................................................... 22

1.2.2 Biofuel feedstocks – The generational gap. ..................................................... 23

1.3. Microalgae – A third-generation biofuel feedstock............................................ 24

1.3.1. Microalgal cultivation – The bottleneck. ........................................................ 26

1.4. Mathematical modelling – A bioprocess optimisation tool. ............................... 30

1.4.1. Microscopic models. ....................................................................................... 31

1.4.2. Macroscopic models. ...................................................................................... 32

1.5. Research objective. ............................................................................................. 33

1.5.1. Research contributions and thesis structure. ................................................... 33

Chapter 2. Literature Review ............................................................... 37

2.1. Introduction. ....................................................................................................... 37

4

2.2. Cultivation considerations for biofuels production. ........................................... 37

2.2.1. Selection of microalgae................................................................................... 37

2.2.2. Selection of cultivation systems...................................................................... 41

2.3. Cultivation strategies targeting starch and lipid formation. ............................... 43

2.3.1. Strategies based on light and temperature. ..................................................... 43

2.3.2. Strategies based on nutrient stress. ................................................................. 44

2.3.3. Strategies based on carbon fixation mechanism. ............................................ 49

2.3.4. Strategies based on operating mode. ............................................................... 52

2.4. Mathematical modelling of microalgae cultivation. .......................................... 56

2.4.1. Modelling algal growth dynamics. ................................................................. 57

2.4.2. Modelling starch and lipid dynamics. ............................................................. 66

2.5. Concluding remarks. .......................................................................................... 71

Chapter 3. Kinetic Modelling of Starch and Lipid Formation during

Mixotrophic, Nutrient-limited Miroalgal Growth ................................. 73

3.1. Introduction. ....................................................................................................... 73

3.2. Contribution 1. ................................................................................................... 75

3.3. Supplementary Information 1........................................................................... 107

Chapter 4. Optimisation of Microalgal Starch and Lipid Formation

via Nitrogen and Phosphorous Co-limitation ....................................... 137

4.1. Introduction. ..................................................................................................... 137

4.2. Contribution 2. ................................................................................................. 141


Chapter 5. An Experimental and Model-based Evaluation of Fed-

Batch Microalgal Cultivation for Biofuels Production ....................... 185

5.1. Introduction. ..................................................................................................... 185

5.2. Contribution 3. ................................................................................................. 189


5

Chapter 6. Microalgal Biomass as a Biorefinery Platform for

Biobutanol and Biodiesel Production: A Case Study .......................... 223

6.1. Introduction. ..................................................................................................... 223

6.2. Contribution 4. A biofuels production case study. ........................................... 227

6.3. Supplementary Information 4. .......................................................................... 259

Chapter 7. Conclusions and Recommendations ................................ 267

7.1. Conclusions. ..................................................................................................... 267

7.2. Recommendations. ........................................................................................... 273

References ................................................................................................ 275

APPENDIX A .......................................................................................... 287

A.1 Preparation of TAP medium. ............................................................................... 287

FINAL WORD COUNT: 55,070

(Not including references)

6

List of Tables

Table 1.1. Performance properties of butanol, ethanol, and gasoline (Bankar et al., 2013;

Harvey and Meylemans, 2011). ...................................................................................... 22

Table 1.2. Comparison of major environmental impacts between different generation

biofuel feedstocks (Groom et al., 2008; Ribeiro and Silva, 2013). ................................ 25

Table 2.1. Cellular composition of several microalgae species (in a dry matter basis).

Adapted from Zhu (2013). .............................................................................................. 39

Table 2.2. List of studies implementing nutrient-stressed cultivation strategies targeting

increased starch and lipid formation ............................................................................... 46

Table 2.3. List of studies implementing two-stage or fed-batch cultivation strategies for

increased starch and lipid accumulation. ........................................................................ 54

Table 2.4. List of microalgae-based kinetic models incorporating starch and/or lipid

dynamics ......................................................................................................................... 69

Table A-1. Components of stock solutions used for 1 L of TAP medium. .................. 287

7

List of Figures

Figure 1.1. Shares of: a) renewable energy sources (RES) in transport energy consumption

across the EU (Eurostat, 2016), and b) projected shares in transport energy demand by fuel

(Capros et al., 2016). ....................................................................................................... 21

Figure 1.2. Schematic diagram of the microalgae-to-fuel production route. ................. 27

Figure 1.3. Visual representation of the responses in a bioprocess model and their

application for simulation and/or optimisation. .............................................................. 31

Figure 2.1. Schematic representation of the major starch and lipid synthetic pathways in

C. reinhardtii. [ACCase, Acetyl-CoA carboxylase; ACP, acyl carrier protein; ADP,

adenosine diphosphate; AGPase, ADP-Glucose Pyrophosphorylase; CoA, coenzyme A; P,

phosphate; PtdOH, phosphatidic acid; WSP, Water Soluble Polysaccharides]. Simplified

from (Ball and Deschamps, 2009; Johnson and Alric, 2013; Riekhof and Benning, 2009).

......................................................................................................................................... 40

Figure 2.2. Extracellular and intracellular elements employed in growth kinetic models for

microalgae. ...................................................................................................................... 58

Figure 2.3. Visual comparison of the μ vs S curves predicted by the growth kinetic models

of Monod, Andrews, and Molina-Grima. ....................................................................... 60

Figure 2.4. Results of simulated double-substrate growth kinetics, as predicted by Eq.

2.12: a) biomass, substrate 1, and substrate 2; b) specific growth rate; and c) weighing

functions. Feasible kinetic parameters and initial values (as shown in table) were randomly

selected for simulation purposes. .................................................................................... 64

8

Figure 6.1. Market sizes and volumes for conventional microalgal products. Adapted from

Zhu (2015). ................................................................................................................... 223

Figure 6.2. A schematic representation of the various microalgal conversion routes

suitable for the co-production of liquid biofuels (highlighted) and other value-added

chemicals. Adapted from Suganya et al. (2016). .......................................................... 224

9

Abbreviations

ANOVA Analysis of variance

CCAP Culture Collection of Algae and Protozoa

DSMZ Deutsche Sammlung von Mikroorganismen und Zellkulturen

DCW Dry cell weight

FAME Fatty Acid Methyl Esther

MB-F Fermented microalgal biomass

GC-FID Gas Chromatography - Flame Ionization Detection

GC-MS Gas Chromatography - Mass Spectrometry

HPLC High Performance/Pressure Liquid Chromatography

MB-H Hydrolysed microalgal biomass

ICP Inductively Coupled Plasma

MB Microalgal biomass

OD Optical density

OES Optical Emission Spectroscopy

ODE Ordinary differential equation

RCM Reinforced Clostridial Medium

rpm revolutions per minute

SE Standard deviation

SQP Succesive Quadratic Programming

TAG Triacylglyceride

TAP Tris-Acetate-Phosphate medium

UV Ultraviolet index

10

Nomenclature

Chapters 3, 4 and 5

ki,A Acetate inhibition constant, gC L-1

Ks,A Acetate saturation constant, gC L-1

YX/A Acetate yield coefficient, gC gC-1

A Acetic acid concentration, gC L-1

P1 Acetic acid concentration from 1st pulse, gC L-1

P2 Acetic acid concentration from 2nd pulse, gC L-1

P3 Acetic acid concentration from 3rd pulse, gC L-1

x* Active biomass concentration, gC L-1

X Biomass concentration, gC L-1

z Culture depth (m)

µH Heterotrophic growth rate

Io Incident light intensity, µmol m-2s-1

ki,S Inhibition constant (R1), gN L-1

ki,L Inhibition constant (R3), gN L-1

Aint Intracellular acetic acid concentration, gC L-1

Nint Intracellular nitrogen concentration, gN L-1

Ϭ Light attenuation coefficient, L gC-1 m-1

ki,I Light inhibition constant, µmol m-2s-1

I Light intensity throughout the culture, µmol m-2s-1

Ks,I Light saturation constant, µmol m-2s-1

L Lipid concentration, gC L-1

R4 Lipid degradation rate, gC L-1h-1

r4 Lipid degradation rate (R4), gN gC-1h-1

r3 Lipid formation rate (R3), gN gC-1h-1

ksat,L Lipid saturation constant (R4)

R3 Lipid synthetic rate, gC L-1h-1

ρN,max Maximum nitrogen uptake rate, gN gC-1h-1

ρP,max Maximum phosphorous uptake rate, gPO4 gC-1h-1

11

µmax Maximum specific growth rate, h-1

qN,0 Minimum nitrogen quota, gN gC-1

qP,0 Minimum phosphorus quota, gPO4 gC-1

ФN N Uptake regulation coefficient, L gC-1

N Nitrogen concentration, gN L-1

qN Nitrogen quota, gN gC-1

ρN Nitrogen uptake rate, gN gC-1h-1

λ Optical depth

H pH

KH pH coefficient, L gC-1 h-1

P Phosphorous concentration, gPO4 L-1

qP Phosphorous quota, PO4 gC-1

KP Phosphorous quota supporting N uptake, PO4 gC-1

ρP Phosphorous uptake rate, gPO4 gC-1h-1

µI Phototrophic growth rate

ФS Regulation coefficient (R1), L gC-1

ФL Regulation coefficient (R3), L gC-1

k1 Regulation constant (R1)

k2 Regulation constant (R3)

Ks,S Saturation constant (R1), gN L-1

Ks,L Saturation constant (R3), gN L-1

K* Saturation constant, No, gN L-1

nS Shape parameter (R1)

nL Shape parameter (R3)

n Shape-controlling parameter

S Starch concentration, gC L-1

R2 Starch degradation rate, gC L-1h-1

r2 Starch degradation rate (R2), gC gC-1

r1 Starch formation rate (R1), gC gC-1

ksat,S Starch saturation constant (R2)

R1 Starch synthetic rate, gC L-1h-1

ki,A:N Uptake inhibition constant, A:N, gC L-1

ki,N Uptake inhibition constant, N, gN L-1

12

ki,P Uptake inhibition constant, P, PO4 L-1

Ks,A:N Uptake saturation constant, A:N, gC L-1

Ks,N Uptake saturation constant, N, gN L-1

Ks,P Uptake saturation constant, P, gPO4 L-1

Chapter 6

AA Acetic acid concentration, g L-1

A Acetone concentration, g L-1

B Butanol concentration, g L-1

BA Butyric acid concentration, g L-1

E Ethanol concentration, g L-1

y Independent variable, g L-1

α5 Interactive regression coefficient, L g-1

α0 Interception coefficient, g L-1

α1 Linear regression coefficient

α2 Linear regression coefficient

N Nitrogen, g L-1

H pH

α3 Quadratic regression coefficient, L g-1

α4 Quadratic regression coefficient, L g-1

Xi Response variable, g L-1

S Substrate (glucose) concentration, g L-1

ABE Total acetone, butanol, and ethanol, g L-1

YS/ABE Yield of ABE on glucose, %

YS/B Yield of biobutanol on glucose, %

13

Abstract

Optimisation of Biofuels Production from Microalgal Biomass

The University of Manchester, 2018

Microalgae are positioned as a promising platform for sustainable biofuels production due

to their ability to synthesise starch and lipid molecules, which can be directed towards the

production of bioethanol and biobutanol via fermentation, or biodiesel via

transesterification. The commercialisation of microalgal biofuels, however, is unlikely to

become a reality unless large-scale algal cultivation systems can efficiently generate high-

density algal cultures rich in starch and lipids. Numerous metabolic studies have revealed

the ability of cells to counteract nutrient-stressed conditions by inducing starch and lipid

accumulation, allowing the exploration of tailor-made biofuel-oriented cultivation

strategies. Nevertheless, it has been demonstrated that those conditions that favour starch

and lipid formation do not typically favour biomass growth, complicating the identification

of cultivation strategies fit for biofuels production.

In this research, the challenging identification of optimal cultivation strategies maximising

starch and lipid formation is approached by developing a predictive kinetic model

supported by experimental observations and suitable for the simulation and optimisation

of algal mixotrophic growth dynamics co-limited by nitrogen and phosphorous. The model

uses a compartmentalised approach in which cells are comprised of an active biomass

fraction and storage molecule fractions, allowing the identification of the individual starch

and lipid concentration profiles. To construct and validate the model, laboratory-scale

batch experiments were carried out with the green model species Chlamydomonas

reinhardtii under various acetic acid (i.e. carbon substrate), nitrogen, and phosphorous

concentration regimes. The model was then built in line with experimental data and

existing modelling approaches, and the associated kinetic parameters were quantified via

an optimisation-based fitting methodology. The validated model was subsequently

exploited as an optimisation tool by identifying the required nutrient compositions

maximising starch and lipid formation. These optimised scenarios yielded significant

increases in starch (+ 270 %) and lipids (+ 74 %) compared to the non-optimised strategy.

The model’s predictive capacity for fed-batch cultivation dynamics was additionally

assessed via the evaluation of a nutrient feeding strategy consisting of intermittent pulses

of acetic acid. Such a strategy was found to significantly increase biomass formation (+

126 %) against standard batch cultivation. Finally, a case study was carried out to quantify

the production of biobutanol and biodiesel within the framework of a microalgal

biorefinery. Results showed biofuel yields (g fuel per g of dry algae) of 0.103 biobutanol

via the ABE fermentation of microalgal starch, and 0.038 biodiesel via the

transesterification of microalgal lipids.

In summary, this research presents an optimisation framework combining both modelling

and experimental tools which can be systematically applied for the establishment of

optimal biofuel-oriented microalgal cultivation systems and additionally reaffirms the

exploitative value of microalgae as a promising biorefinery platform for biofuels

production.

14

Declaration

No portion of the work referred to in the thesis has been submitted in support of an

application for another degree or qualification of this or any other university or other

institute of learning

Gonzalo M. Figueroa Torres

15

Copyright Statement

i. The author of this thesis (including any appendices and/or schedules to this thesis)

owns certain copyright or related rights in it (the “Copyright”) and s/he has given

The University of Manchester certain rights to use such Copyright, including for

administrative purposes.

ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic

copy, may be made only in accordance with the Copyright, Designs and Patents

Act 1988 (as amended) and regulations issued under it or, where appropriate, in

accordance with licensing agreements which the University has from time to time.

This page must form part of any such copies made.

iii. The ownership of certain Copyright, patents, designs, trademarks and other

intellectual property (the “Intellectual Property”) and any reproductions of

copyright works in the thesis, for example graphs and tables (“Reproductions”),

which may be described in this thesis, may not be owned by the author and may be

owned by third parties. Such Intellectual Property and Reproductions cannot and

must not be made available for use without the prior written permission of the

owner(s) of the relevant Intellectual Property and/or Reproductions.

iv. Further information on the conditions under which disclosure, publication and

commercialisation of this thesis, the Copyright and any Intellectual Property and/or

Reproductions described in it may take place is available in the University IP Policy

(see http://documents.manchester.ac.uk/DocuInfo.aspx?DocID=2442 0), in any

relevant Thesis restriction declarations deposited in the University Library, The

University Library’s regulations (see

http://www.library.manchester.ac.uk/about/regulations/) and in The University’s

policy on Presentation of Theses.

17

Acknowledgements

Immense thanks to my family and above all to my parents, Karina and Gonzalo, and to my

sister, Cesiah. Your love, support, and words of encouragement from afar made this

journey easier to navigate.

I would like to express my gratitude to my supervisor, Prof. Constantinos Theodoropoulos,

for his guidance and fruitful discussions had throughout the course of this research, and to

my co-supervisor, Dr. Jon Pittman, for his disposition and his insightful comments and

suggestions.

My appreciation to Dr. Mesut Bekirogullari for his valuable help and advice both in and

outside the lab, and to Wan M. Asyraf for his assistance with part of the laboratory work

included in this research. I also recognise the support of the technical staff in the School

of Chemical Engineering and Analytical Science.

The ups and downs of this journey were shared with some wonderful people I met along

the way. They have a place in here. Special thanks to Michelle, Merve, and Fernando for

making my stay in Manchester a gratifying and rewarding experience.

Finally, I kindly acknowledge the financial support provided by the Mexican National

Council of Science and Technology (CONACyT).

19

Chapter 1

Introduction and Research Contribution

1.1. Introduction.

The research presented in this thesis is driven by the global need to develop optimal

feedstock-to-biofuel production technologies capable of competing against the giant fossil-

based fuel industries, major contributors to climate change (Scaife et al., 2015). Although

biofuel production processes have slowly been capable of reaching commercial

implementation, they still suffer from several technological drawbacks among which the

most notorious is the use of unsustainable biomass feedstocks. The screening and

evaluation of alternative feedstocks, however, have led to the recognition of microalgal

biomass as a promising substrate for biofuels production (Shuba and Kifle, 2018).

Therefore, the focus of this research is to establish methods of designing optimal

microalgae cultivation strategies suitable for the large-scale production of biofuels. In

order to lay down the supporting background behind this research, this Chapter will first

establish the importance of biofuels and their current stage of development, followed by

the role of microalgae in satisfying current biofuel production requirements. Finally, an

overview of bioprocess modelling tools will be provided to highlight how their adequate

application can solve the existing challenges that motivate this research.

1.2. Biofuels – An alternative energy source.

According to the latest statistics, energy demands across the European Union (EU) are

primarily satisfied by petroleum products, estimated to account for almost 40 % of the

annual energy consumption (Eurostat, 2017). Our over-dependence on petroleum products,

however, raises two major environmental concerns: i) the excessive generation of

greenhouse gases (GHGs) which aggravate global warming, and ii) the overexploitation of

Chapter 1 – Introduction and Research Contribution

20

finite crude oil reserves which shrink faster than their natural formation rate (Scaife et al.,

2015; Shuba and Kifle, 2018).

The largest consumer of petroleum products is, by far, the transportation sector, where

fossil fuels like gasoline (petrol) and diesel currently make up around 80 % of the total

transport energy demands (Eurostat, 2017). The high contribution of fossil fuels to GHGs

emissions, worsened by the continuous growth of the transport sector, have positioned

liquid biofuels as one of the most promising alternative transport energies capable of

alleviating global warming whilst favouring the shift towards a sustainable economy

(Azadi et al., 2017; Brennan and Owende, 2013; Nigam and Singh, 2011).

In an effort to ensure the widespread use of liquid biofuels in transport, the European

Council established two ambitious environmental policies: i) the Renewable Energy

Sources (RES) directive, which imposes a legally binding target mandating all Member

States to reach a 10 % minimum share of renewable fuels in transport by 2020, and ii) the

Fuel Quality Directive (FQD), which requires similarly by 2020, a minimum 6 % reduction

in the GHGs emissions from transportation (European Commission, 2009a; European

Commission, 2009b).

The current increasing trend in the share of renewables in transport energy consumption,

shown in Figure 1.1 (a), suggests that the 10 % target (mandated by the RES directive)

will be met. However, even though this implies an optimistic outlook for biofuels usage,

it is important to note that the EU shares of renewables do not account exclusively for

biofuels, but also for other alternative energies such as renewable electricity, biogases, or

solar energy. In fact, it has been projected that based on current technological progress and

existing legislation, the share of biofuels in transport, shown in Figure 1.1 (b) and which

currently stand at the 5 % level, will remain virtually unchanged by the year 2050 (Capros

et al., 2016).

With regards to the FQD, a recent report from the European Commission anticipates that

the 6% target for GHGs reduction in transport will most likely be met in a timely manner,

with biofuels being responsible for most of the current GHGs savings. The report points

out, however, that assessing the progress made by the FQD throughout all Member States

is challenging due to the lack of an appropriate monitoring system that does not formally


21

measure the impact of other non-renewable fuels (e.g. electricity, liquefied petroleum gas,

liquid natural gas) on GHGs reductions (European Commission, 2017a).

Figure 1.1. Shares of: a) renewable energy sources (RES) in transport energy

consumption across the EU (Eurostat, 2016), and b) projected shares in transport

energy demand by fuel (Capros et al., 2016).

Despite the increased and positive presence of liquid biofuels in transport, excessive

pressure is being faced by biofuel-producing industries to not only maintain, but also to

further increase their role as major contributors towards the European energy targets. Fuel

suppliers have raised concerns over the strict and complex regulations restricting the

replacement of fossil fuels with biofuels as per the maximum blending limits that ensure

compliance with fuel quality standards (e.g. vapour pressure, sulphur content, distillation

point, etc.) without excessive engine alterations (European Commission, 2017a; Küüt et

al., 2017).

Greater shares of biofuel usage in transport and consequently further reductions in GHGs

emissions can potentially be achieved if biofuels could be blended beyond the established

limits, which thus far allow a 7 % (v/v) of Fatty Acid Methyl Ester (FAME) in diesel (so-

called B7 limit), and a 10 % (v/v) ethanol in gasoline (so-called E10 limit) (European

Commission, 2017b).


22

1.2.1. Butanol – A promising biofuel.

The current blending ratios are driven by differences in physical properties between fuels

and biofuels (Bankar et al., 2013). Therefore, research has aimed to identify biofuels that

could be blended at much higher ratios or replace fossil fuels completely. One such

alternative biofuel is butanol, a sugar-based liquid biofuel with the potential to displace

ethanol as the current leader in gasoline blends.

Ethanol, which dominates the biofuel market in countries such as Brazil and the United

States (Chen et al., 2013; Kumar and Gayen, 2011), suffers from three major drawbacks:

i) a Net Heat of Combustion (NHOC) 38 % lower than that of gasoline, ii) a much higher

vapour pressure than gasoline which complicates its secure transportation in pipelines, and

iii) a higher hygroscopicity (i.e. a measure of a compound’s miscibility in water) than

gasoline which increases the chances for undesired water-fuel mixtures (Bankar et al.,

2013; Harvey and Meylemans, 2011). Unlike ethanol, butanol has a higher NHOC closer

to that of gasoline, a lower vapour pressure, and a lower hygroscopicity (Table 1.1),

increasing its overall compatibility with existing infrastructures.

Table 1.1. Performance properties of butanol, ethanol, and gasoline (Bankar et al.,

2013; Harvey and Meylemans, 2011).

Performance properties Gasoline Butanol Ethanol

Net heat of combustion, NHOC [MJ/L] 32.3 26.8 21.1

NHOC relative to gasoline 1 0.84 0.66

Vapour pressure at 20°C [kPa] 0.7-207 0.53 7.58

Hygroscopicity Low Low High

Compatibility with infrastructure - High Low

Butanol can be biochemically produced through the ABE fermentation, a well-known

fermentative processes whereby Clostridia strains metabolise carbohydrates into acetone,

butanol and ethanol (Green, 2011). Butanol production via fermentation is one of the oldest

known biochemical processes in the world (the first Clostridium strain was isolated

between 1912 and 1914 by Chaim Weizmann, from Manchester University), but ABE


23

fermentation still suffers from technological drawbacks that prevents its widespread

commercialisation (Jones and Woods, 1986; Moon et al., 2016).

The major problems of the ABE fermentation are the low butanol yields (a typical

fermentation yields 2 % w/v butanol) which require energy-intensive recovery processes

(Xue et al., 2014), and the adequate selection of simple and renewable fermentation

feedstocks which are estimated to account for up to 75 % of total processing costs (Jiang

et al., 2015). Whilst the improvement of butanol yield is crucial if ABE fermentation

technologies are to attain commercial success, the selection of appropriate feedstocks has

been a fundamental challenge faced by all biofuel production technologies to date.

1.2.2 Biofuel feedstocks – The generational gap.

In order for any liquid biofuel (e.g. biobutanol, bioethanol, or biodiesel) to be considered

an appropriate large-scale fossil-fuel replacement, its production route must avoid a

number of environmental, social, and economic risks by complying with various

sustainability criteria such as minimal GHG emissions over their production cycle (i.e. low

carbon footprint), low capital and operational costs, a low indirect land use change (ILUC)

certification, and low water usage and pollution (RAEng, 2017).

Biofuels currently used in transport have reached advanced production stages, but their

widespread commercialisation is limited due the environmental burdens and controversy

surrounding the food-based feedstocks (e.g. bioethanol from corn, and biodiesel from

rapeseed oil) from which they have been conventionally produced (Suganya et al., 2016).

Food-based biofuels, so-called first-generation (1G) biofuels, have proven to be both

economically and socially unsustainable due to their intrinsic competition for arable land

and food grown for human needs (Nigam and Singh, 2011). They additionally possess a

high global warming potential (GWP) due to their large contribution to nitrous oxide (N2O)

emissions, a GHG generated through nitrogen-based fertilisers and organic matter decay

(RAEng, 2017).

In order to prevent feedstocks competition for food, research and development (R&D)

schemes gradually shifted their efforts towards the use of lignocellulosic materials (e.g.

non-food crops such as agricultural and forestry residues, energy crops and wood wastes)

for biofuel production (Nigam and Singh, 2011; Suganya et al., 2016). Biofuels produced


24

from lignocellulosic materials, so called second generation (2G) biofuels, have lower

carbon footprint (crops generally require less fertiliser) and exhibit larger availability and

cheap harvesting costs than food-based biomass (RAEng, 2017; Suganya et al., 2016).

Commercialisation of second generation biofuels, however, is largely restricted by the

complex lignin-based crystalline structure of lignocellulosic biomass, which complicates

the extraction of the biofuel substrate molecules unless costly and time-consuming pre-

treatment steps are implemented (Chundawat et al., 2011; Scaife et al., 2015). Second

generation biofuels are in an early developmental stage and still require intensive research

to improve their profitability and ensure their commercial success. However, they

represent thus far a more sustainable and long-term feedstock for biofuels than traditional

food-based biomass (Oh et al., 2018; Suganya et al., 2016).

Therefore, the European Commission has recently adopted the term “advanced biofuels”

to refer to those biofuels produced from non-food biomass and has additionally announced

plans to establish target policies for advanced biofuels shares in transport (starting with a

0.5 % share as a reference value) across Member States (European Commission, 2017b;

European Environment Agency, 2017).

Although such plans unquestionably favour the shift to a greener and more sustainable

economy beyond the 2020 European targets, their successful implementation will require

the development of mature feedstock production technologies suitable for biofuels

production.

1.3. Microalgae – A third-generation biofuel feedstock.

In recent years, microalgae have emerged as one of the most promising and long-term

feedstocks for the production of advanced biofuels (Kim et al., 2013; Lee et al., 2015b;

Leong et al., 2018). Microalgae, recognised as a third-generation (3G) biofuel feedstock

(among other oleaginous organisms like bacteria and yeast), comprise a large group of

aquatic photosynthetic organisms capable of synthesising carbohydrates (mainly in the

form of starch granules) and lipids, the two major precursors of both sugar- and lipid-based

fuels (Leong et al., 2018; Nigam and Singh, 2011; Oh et al., 2018).


25

As shown in Table 1.2, third-generation biofuels (3G) are estimated to have lower

environmental impacts than those exhibited by 1G or 2G biofuels, which has helped

establish microalgae as one of the most sustainable biofuels feedstock to date.

Table 1.2. Comparison of major environmental impacts between different

generation biofuel feedstocks (Groom et al., 2008; Ribeiro and Silva, 2013).

Biofuel Source Land

use a

Water

use b

Fertiliser

use

Energy

usec

GHG

emissionsd

1st Gen

(1G)

Corn High High High High (+) 81 - 85

Sugar cane High High High Med (+) 4 -12

Soybeans High High Low-Med Low-Med (+) 49

Rapeseed/Canola Med High Med Low-Med (+) 37

2nd Gen

(2G)

Wood residues High Med Low Low -

Switchgrass High Low-Med Low Low (-) 24

3rd Gen

(3G) Microalgae Low Med Low High (-) 183

a in terms of the land area needed to meet 50 % of the U.S. transport fuel demands.

b includes water used to grow feedstock and biofuel refining.

c includes energy inputs for mechanical equipment and biofuel transport and refining.

d kgCO2e/MJ fuel over full production cycle, compared against either gasoline (94 kgCO2e/MJ) or diesel

(83 kgCO2e/MJ) fuels.

When compared against food-based and lignocellulosic biomass, microalgae exhibit other

preferred advantages such as: faster growth rates, higher photosynthetic efficiency, and

simpler growing conditions (Brennan and Owende, 2010; Salama et al., 2018).

Furthermore, microalgae’s simple cellular structure is lignin-free (unlike lignocellulosic

biomass), which facilitates saccharification (i.e. the process by which carbohydrates are

hydrolysed to soluble sugars) and therefore avoids the necessity of carrying out costly and

often inefficient pre-treatment steps (Chen et al., 2013; Gouveia, 2011; Markou et al.,

2012a).


26

Algal fuel technologies, however, are not yet sufficiently developed nor economically

competitive to satisfy global biofuel production demands (Pragya et al., 2013; RAEng,

2017). In this regard, one of the major technological challenges that is yet to be addressed

to ensure the widespread use of algal biofuels is the establishment of optimal algal

cultivation systems capable of yielding high starch and lipid productivities (Markou et al.,

2012a; Oh et al., 2018; Rashid et al., 2014; Rodolfi et al., 2009; Shuba and Kifle, 2018).

1.3.1. Microalgal cultivation – The bottleneck.

The algae-to-fuel production route can be depicted as a multi-stage bioprocess comprising

numerous upstream and downstream operations (Figure 1.2), among which the major steps

are strain selection, cultivation, harvesting, extraction, conversion, and recovery (Kim et

al., 2013; Pragya et al., 2013; Rashid et al., 2014; Shuba and Kifle, 2018). Each of these

operations plays an important role in securing the viability and success of microalgae-

based fuels such as bioethanol and biobutanol, produced by means of fermentative

processes (e.g. the ABE fermentation), or biodiesel, produced by means of the

transesterification reaction (Figure 1.2). Although these biofuel conversion processes have

been well studied, their optimal implementation remains a crucial challenge to favour high

biofuel production yields that require low energy recovery operations, which in turn

translate into a low cost and competitive biofuel production process (Suganya et al., 2016).

Increasing the efficiency of the conversion process itself is thus of critical importance to

guarantee the commercialisation of biofuels, regardless of whether algal biomass is

employed. However, if microalgal starch and lipids are to be used as the biofuel substrates

because of their promising and renewable nature (Table 1.2), upstream operations need to

ensure the generation of mass-scale algal cultures. This represents a major challenge that

makes algal cultivation one of the most crucial processing stages of third-generation

biofuels (Günerken et al., 2015; Kim et al., 2013; Lee et al., 2015b). Cultivation

technologies are usually species-specific and dependent upon the target end-product, but

they must satisfy algae’s basic growing requirements (e.g. light, carbon, and nutrients) to

allow the mass production of high-density algal cultures, which in turn regulates the

performance of the harvesting and other subsequent stages (Adeniyi et al., 2018; Su et al.,

2017). For the purposes of large-scale biofuel production, microalgae cultivation must


27

additionally be able to favour increased starch and lipid production to guarantee the

profitability and effectiveness of the biofuel conversion process.

Figure 1.2. Schematic diagram of the microalgae-to-fuel production route.

It has been demonstrated that microalgae’s cultivation environment, which influences their

intracellular composition, can be artificially manipulated to induce starch and lipid

accumulation (Markou et al., 2012a; Vitova et al., 2015). For example, it is well known

that algal growth, as well as starch and lipid accumulation, are favoured when the light

intensity of the cultivation system is increased (Singh and Singh, 2015a). Manipulation of

light is thus considered a suitable cultivation strategy for biofuels production, but the

optimal light conditions are strain-specific and should be carefully established to avoid

photoinhibition or self-shading effects (see Chapter 2 for additional details).


28

Other tailor-made cultivation strategies that have been commonly explored in the literature

include not only the manipulation of light but also the photoperiod, the cultivation

temperature, the composition of culture media, or the cultivation mode (e.g. batch, fed-

batch, continuous). Among these cultivation strategies, perhaps the most widely

acknowledged for their potential to significantly increase starch and lipid contents involves

nutrient limitation (Bajhaiya et al., 2016; Ball et al., 1990; Behrens et al., 1989; Fernandes

et al., 2012; Morales-Sánchez et al., 2016), which consists of reducing nutrient availability.

However this strategy can reduce biomass growth, and in consequence, starch and lipid

productivities (Markou et al., 2012a; Oh et al., 2018).

The low biomass concentrations attained by cultivation systems represent a major

drawback for microalgae-to-biofuel technologies (Shuba and Kifle, 2018), but this

problem can be addressed by: i) the use of mixotrophic strains (i.e. those that assimilate

organic carbon sources in addition to CO2, see Chapter 2) which show higher growth rates

than those grown phototrophically (Chapman et al., 2015), or by ii) implementing fed-

batch systems maintaining high biomass densities by means of appropriate nutrient feeding

strategies (Fields et al., 2018; Jeffryes et al., 2013).

Nutrient-limited mixotrophic cultivation combined with fed-batch operation thus offers

great potential for the purposes of biofuel production, but the adequate integration of such

strategies requires the identification of the optimal media composition and/or operating

parameters. Such tasks, however, are complicated and may involve labour-intensive

experimentation due to the various nutritional and cultivation parameters that regulate

cellular growth (Bernard, 2011; Gouveia, 2011; Markou et al., 2012a; Rashid et al., 2014).

Optimisation of all the parameters affecting the productivity of the cultivation stage is thus

necessary to speed-up the commercialisation of microalgal biofuel technologies, which

will be ultimately driven by competitive biomass production and processing costs (Singh

and Gu, 2010; Slade and Bauen, 2013). Available techno-economic analyses targeting

biofuel production estimate that algal biofuels (microalgal biodiesel price is estimated as

8 US dollars per gallon), are not yet competitive against standard fossil fuels (petroleum

diesel price averages 3.17 US dollars per gallon) (Delrue et al., 2012; GlobalPetrolPrices,

2018; Zhu, 2015). The high costs associated to algal biofuels are mainly attributed to the


29

cultivation stage, which requires an adequate yet high input of carbon, energy, and other

nutrients (Kim et al., 2013; Shuba and Kifle, 2018).

Microalgae have long been praised for their ability to grow phototrophically whilst fixating

inorganic CO2, making them suitable to grow on atmospheric CO2 emissions and therefore

lowering costs. However, and as mentioned before, phototrophic cultures are hindered by

having low biomass productivities, thus requiring large CO2 inputs (their CO2 fixation

efficiency may be as low as 10%) which may only be satisfied by means of (gaseous)

industrial waste streams (Salama et al., 2018; Slade and Bauen, 2013). Although

mixotrophic cultures are more attractive since they yield high biomass productivities and

require lower carbon inputs (Zhan et al., 2017), these cultures require additional organic

carbon substrates that are generally expensive, which may end up increasing further the

price of biofuels unless these substrates are optimally supplied (Adeniyi et al., 2018;

Salama et al., 2018).

On the other hand, the costs of supplying other essential fertilisers for algal cultivation

(e.g. nitrogen, phosphorus, and potassium) must also be taken into account when

evaluating the economics of the microalgae-to-fuel process. It is estimated, for example,

that the EU annual capacity for fertilizer production (accounting for nitrogen and

phosphorous) would need to double if fossil fuels were completely substituted with algal

fuels (Slade Bauen 2013), which would raise fertiliser prices and consequently algal

cultivation costs. Increasing the competitiveness of microalgal biofuels prices to ultimately

guarantee their commercialisation is thus a significant challenge that requires the adequate

management of carbon and nutrient inputs in cultivation systems, yielding maximal

productivities whilst also reducing any potential waste.

The development of optimal microalgae cultivation systems targeting biofuels production

would benefit by optimisation frameworks where both experimental and computational

approaches are efficiently integrated. Predictive models, particularly, have been shown to

be adequate tools for the evaluation and optimisation of bioprocesses subject to multiple

growth-limiting factors such as those involved in microalgal cultivation (Lee et al., 2015a).

The value of modelling as a bioprocess optimisation tool will thus be discussed next.


30

1.4. Mathematical modelling – A bioprocess optimisation tool.

A common goal of processing industries, regardless of their end-product or scale, is to

improve product yields or reduce energy consumption to increase process profitability and

performance (Jiménez-González and Woodley, 2010; Wang et al., 2009). Process

optimisation has thus become an essential requirement to: i) identify those variables with

the greatest influence on process performance, and ii) to subsequently establish the most

beneficial processing strategies (Edgar et al., 2001).

Process optimisation requires not only the identification of an objective function (e.g.

maximise yields, minimise waste, etc.), but also the formulation of an adequate process

model (i.e. a mathematical representation of the input-output responses of a given system)

that ensures a reliable analysis and evaluation of any potential process-enhancing strategy

prior to its implementation (Edgar et al., 2001). The formulation of highly representative

process models thus constitutes one of the most important optimisation tasks.

Large-scale chemical industries involving well-established operations (e.g. distillation,

heat exchange, etc.), where slight improvements may yield substantial cost or energy

savings, have relied on process modelling and simulation tools to identify optimal

operating strategies (Jiménez-González and Woodley, 2010; Velayudhan, 2014). Although

these optimisation tools can be applied to biochemical industries, the formulation of

accurate models representative of biological systems is more challenging due to the highly

complex and dynamic nature of living cells, which often act as the major catalysts for

product formation (Kiparissides et al., 2011; Olivier and Isabelle, 2010).

The development of modelling frameworks for biological processes is often an iterative

process which requires the proposed input-output model formulations to be tested for their

capacity to describe a specific system. Therefore, the construction of bioprocess models

requires biological and engineering concepts to be integrated to evaluate and describe the

dynamics of cellular systems, and additionally generate reliable experimental data against

which the model’s performance can be assessed (Kiparissides et al., 2011; Shuler and

Kargi, 1992).

Bioprocess models can be readjusted or reformulated based on performance, but once the

input-output responses are validated the model can be used with confidence to: i) identify


31

an unknown output based on a known input (simulation), or ii) to identify an unknown

input based on a known or targeted output (optimisation) (Figure 1.3).

Figure 1.3. Visual representation of the responses in a bioprocess model and their

application for simulation and/or optimisation.

Model formulations for bioprocesses have various degrees of complexity and

mathematical structures, but existing modelling approaches are generally differentiated by

whether the specific purpose for which they were developed (which is a crucial modelling

consideration) employs a microscopic or macroscopic approach (Bailey, 1998).

1.4.1. Microscopic models.

In microscopic-oriented models (also called white-box models), cell dynamics are

portrayed at the metabolic level by simulating the intricate network of intracellular

reactions governing growth-related processes (e.g. nutrient uptake, product synthesis).

These models facilitate the identification of rate-limiting metabolic reactions and have

therefore found application within novel strain-enhancing biotechnological strategies such

as metabolic engineering, whereby key cellular pathways are blocked or favoured via

genetic manipulation (Wang et al., 2009).

Microscopic modelling approaches can thus be applied as robust optimisation tools

suitable for the development and isolation of highly productive and resistant strains

suitable for industrial applications. It should be mentioned that the level of complexity of

these models is wide and their implementation involves various mathematical techniques


32

(e.g. stoichiometric matrixes, flux balance analysis) which broadens their classification (de

Prada et al., 2018; Schwartz and Soons, 2019). A limitation of these types of models,

however, is that their construction demands a thorough understanding of the strain-specific

biochemical reaction networks (which may or may not be available), and their validation

will be evidently restricted by the need to carry out appropriate but complicated

experimental measurements of intracellular metabolites (Bailey, 1998; Baroukh et al.,

2014).

1.4.2. Macroscopic models.

Macroscopic models (also called grey-box models) simulate the overall performance of

biological processes by establishing appropriate relationships between major cellular

targets (i.e. product titre and/or cell yield) and their corresponding responses to the input

variables of cellular systems, such as substrate concentration or environmental conditions

(Velayudhan, 2014). Although the development of macroscopic models generally relies on

empirical mathematical formulations, this approach has proven to be ideal for the

identification and optimisation of critical bioprocess parameters (e.g. growth-limiting

factors, environmental conditions, culture media composition, operating mode) whilst

reducing time-consuming and/or expensive experimentation trials (Kiparissides et al.,

2011; Velayudhan, 2014).

The essential yet challenging task of establishing optimal microalgal cultivation strategies

suitable for biofuel production, as highlighted by the preceding sections, can therefore be

accomplished by means of macroscopic modelling approaches capable of portraying the

antagonistic responses between algal biomass growth and starch and lipid formation (i.e.

model outputs) when subject to nutrient limiting conditions (i.e. model inputs). However,

although plenty of macroscopic-based models have accurately described algal growth

dynamics and their response to nutrient or environmental stress (Lee et al., 2015a), models

incorporating the simultaneous formation of starch and lipids are scarce or limited.


33

1.5. Research objective.

Microalgae cultivation for biofuel production purposes is thus far limited by a

characteristic trade-off between biomass formation and starch and lipid accumulation (e.g.

the main carbon sinks from microalgal metabolism). Such a trade-off can be adequately

balanced by implementing optimal nutrient-limited cultivation strategies within

mixotrophic systems. Although the analysis and identification of optimal bioprocessing

strategies are highly complex and time-consuming tasks, they can be accomplished via

optimisation frameworks integrating both experimental and modelling tools. Therefore, to

optimise microalgal biofuels production, this thesis aims to identify optimal cultivation

strategies for maximal starch and lipid formation by developing a macroscopic kinetic

model suitable for the analysis and simulation of microalgal growth dynamics responsive

to nutrient-limited mixotrophic conditions.

1.5.1. Research contributions and thesis structure.

The contributions of this thesis are presented in a “journal format” as a series of academic

papers published or submitted for publication in scientific journals. Motivated by the need

to address the current gaps in literature (which will be further discussed in Chapter 2) the

research contributions of this thesis are summarised as follows:

Contribution 1:

A predictive and experimentally validated multi-parametric kinetic model was

developed to portray microalgal growth coupled with starch and lipid formation

during mixotrophic, nitrogen-limited cultivation. Model fitting and validation was

carried out against experimental cultivation datasets generated from cultures of

Chlamydomonas reinhardtii CCAP 11/32C grown under various nitrogen and

acetic acid (i.e. organic carbon source) concentration regimes. The validated model

was subsequently exploited to establish optimal “starch-enhancing” and “lipid-

enhancing” algal cultivation strategies. Such optimal strategies yielded significant

increases in starch and lipids against the base case and were successfully validated


34

via experimental analysis. All findings associated with this work are presented in

Chapter 3.

Contribution 2:

The model developed in Contribution 1 was further enhanced by incorporating the

responses of algal growth and starch and lipid formation to phosphorus limitation.

Additional modelling considerations were made to refine the predictive features of

the model and avoid unfeasible accumulation scenarios when extrapolated to cases

of extreme nutrient limitation. The resulting model, which was experimentally

validated, was then employed to identify optimal “starch-enhancing” and “lipid-

enhancing” algal cultivation strategies subject to mixotrophic dynamics, co-limited

by nitrogen and phosphorus. Such optimal strategies were successfully validated

via the experimental analysis of C. reinhardtii CCAP 11/32C grown under the

model-based optimal initial concentrations of nitrogen, phosphorus, and acetic

acid, and yielded higher starch and lipids than those presented in Contribution 1.

All findings associated with this work are presented in Chapter 4.

Contribution 3:

A fed-batch cultivation strategy for sustained biomass formation during nutrient-

limited mixotrophic conditions was evaluated by means of a nutrient feeding

strategy consisting of an intermittent pulse of the carbon substrate. The fed-batch

strategy was evaluated in laboratory-scale cultures of C. reinhardtii CCAP 11/32C

using pulses with various concentrations of acetic acid (i.e. the mixotrophic carbon

substrate). Microalgal cultures subject to three consecutive pulses of acetic acid

attained biomass concentrations significantly higher than that obtained by cultures

grown in batch. Such increase in biomass translated into significant increases of

the starch and lipid concentrations. To simulate the fed-batch strategy, the

previously developed model was adapted to portray the dynamics of the pulse-

assisted cultivation. The predictive capacity of the adapted model is enhanced with


35

respect to typical batch scenarios, but it is limited since it can portray the outcome

of a single pulse. All findings associated with this work are presented in Chapter

5.

Contribution 4 – A biofuels production case study:

The use of microalgae as a feedstock for biofuels was evaluated within a

biorefinery context by quantifying the production of microalgal biobutanol via the

ABE fermentation and microalgal biodiesel via transesterification. The assessment

of butanol production involved: i) the initial identification and optimisation of key

fermentation parameters (assisted by glucose-based fermentation experiments),

and ii) the subsequent evaluation of biobutanol production by non-hydrolysed and

hydrolysed microalgal biomass. Meanwhile, the assessment of biodiesel

production involved the analysis of the lipid content and fatty acid composition of

both unfermented and fermented microalgal biomass. Biofuel conversion yields

obtained from the case study are within a range comparable to existing studies and

reinforce the potential of microalgae as a platform for biorefineries. All findings

associated with this case-study are presented in Chapter 6.

Maintaining the journal format employed in this thesis, Chapters 3 – 6 include: i) a brief

introduction to the contributions presented within, ii) the corresponding academic paper

presented in manuscript format, and iii) supplementary information associated to each

contribution.

Finally, a summary of all the major findings from this research and recommendations for

future work are provided in Chapter 7.

37

Chapter 2

Literature Review

2.1. Introduction.

Due to their acknowledged potential to provide a sustainable energy alternative for the

transport sector, microalgae-based biofuels have become a key target for both research and

industry around the globe. Specifically, biofuels production from microalgal biomass

requires the identification and optimisation of cultivation strategies maximising starch and

lipid formation (i.e. the biofuels precursors). This Chapter will thus provide a summary of

technological advances, and associated drawbacks, of biofuel-oriented microalgae

cultivation. In order to highlight the contributions of this thesis, as defined in Chapter 1,

focus will be given to existing experimental and modelling approaches targeting starch and

lipid formation.

2.2. Cultivation considerations for biofuels production.

2.2.1. Selection of microalgae.

Algae comprise a wide range of aquatic organisms capable of performing photosynthesis,

a process whereby sunlight energy is used to convert water and atmospheric carbon dioxide

(CO2) into the chemical energy necessary to sustain cellular growth (Nigam and Singh,

2011). Of the estimated 72,500 algae species (conservative figure) most are unicellular

microalgae, a primitive type of plant, but devoid of roots, stems or leaves (Guiry, 2012;

Rasul et al., 2017).

Microalgae have already been positioned as a suitable precursor for a variety of

pharmaceutical and nutraceutical products (e.g. vitamins B1, B6, B12, β-carotene, lutein,

astaxanthin) due to a composition generally rich in proteins, pigments, vitamins,

Chapter 2 – Literature Review

38

antioxidants, and other bioactive compounds (Enamala et al., 2018; Suganya et al., 2016).

Due to their ability to additionally synthesise carbohydrate and lipid molecules (raw

precursors for biofuels), microalgae are also acknowledged as a promising feedstock for

advanced biofuels (Chen et al., 2013).

However, since microalgae are diverse and do not all share the same characteristics, a

careful selection process must be undertaken to identify those species and strains most

suited for a particular commercial application (Scaife et al., 2015). For biofuels production,

microalgae should commonly satisfy the following cultivation criteria (Shuba and Kifle,

2018): i) allow minimum fouling and easy harvesting, ii) maintain stable growth rates, and

most importantly, iii) high formation of carbohydrate and lipids (Brennan and Owende,

2010).

Among all species, including prokaryotic (e.g. cyanobacteria) or eukaryotic (e.g. green

algae, red algae, and diatoms), green microalgae are regarded as the most productive for

biofuel production due to a richer composition in carbohydrate and lipid (Su et al., 2017).

As observed in Table 2.1, the carbohydrate:lipid ratios between green species may differ

greatly and should be taken into consideration depending on the targeted biofuel.

Microalgae species with high lipid contents are more suited for the production of biodiesel

(FAME, fatty acid methyl ester) given that this process relies on the transesterification of

oils (Chisti, 2007). Meanwhile, microalgae species with high carbohydrate contents are

more suited for the production of bioethanol or biobutanol since these alcohols are the

major products of carbohydrate-driven microbial fermentations (Kim et al., 2014; Wang et

al., 2016).

Several studies have already demonstrated the production of biofuels from various

microalgae species. To name a few, biodiesel production has been evaluated from biomass

residues of Chlorella protothecoides (Xu et al., 2006), Oscillatoria sp. and Cyclotella sp.

(Velasquez-Orta et al., 2014), and Chlorella zofingiensis (Liu et al., 2011), and bioethanol

or biobutanol production from Chlorella vulgaris (Gao et al., 2016; Ho et al., 2013; Kim

et al., 2014), Chlorella sorokiniana (Cheng et al., 2015), and even wastewater microalgae

populations (Castro et al., 2015).


39

Whilst the selection of microalgae species for commercial biofuels production is largely

driven by their intracellular composition, attention should also be given to their cultivation

requirements since many species are known to modify their carbon composition in

response to the growing environment (Shuba and Kifle, 2018).

Table 2.1. Cellular composition of several microalgae species (in a dry matter basis).

Adapted from Zhu (2013).

Storage molecules

Microalgae Carbohydrate (%) Lipids (%)

Anabaena cylindrica 25 - 30 4 - 7

Aphanizomenon flosaque 23 3

Arthrospira maxima 13 - 16 6 - 7

Chlamydomonas reinhardtii 17 21

Chlorella pyrenoidosa 26 2

Chlorella vulgaris 12 - 17 14 - 22

Chlorella zofingiensis 25 - 28 26 - 46

Dunaliella salina 32 6

Euglena gracilis 14 - 18 14 - 20

Porphyridium cruentum 40 - 57 9 - 14

Scenedesmus dimorphus 21 - 52 16 - 40

Scenedesmus obliquus 10 - 17 12 - 14

Scenedesmus quadricauda - 1.9

Spirogyra sp. 33 - 64 11 - 21

Spirulina maxima 13 - 16 6 - 7

Spirulina platensis 8 - 14 4 - 9

In particular, microalgal cells are known to counteract stressed growing conditions by

synthesising higher carbohydrate and lipid molecules through a shift in their carbon

metabolism (Markou et al., 2012a). This particular trait has widened microalgae’s potential

and positioned them as perfect candidates for the exploration of tailor-made cultivation

strategies where biofuel-specific requirements are targeted via an artificial manipulation

of their metabolism (Markou et al., 2012a).

It should be mentioned that although other species are shown to accumulate higher

carbohydrate and lipid contents (Table 2.1), the research presented in this thesis employs


40

the green microalgae Chlamydomonas reinhardtii, a model organism widely studied in the

literature for its central carbon metabolism and its distribution into storage molecules,

which are the major focus of this work. A brief overview of this metabolism is presented

next, but more comprehensive reviews are available in: Ball and Deschamps (2009),

Johnson and Alric (2013), and Riekhof and Benning (2009).

Starch and lipid metabolism.

Despite its complexity, the elucidation of the full genome sequence in C. reinhardtii has

provided a clearer understanding of the major mechanisms regulating the assimilation of

carbon and its partitioning into carbohydrate and lipid reserves (Johnson and Alric, 2013).

A simplified yet visual representation of these pathways is shown in Figure 2.1.

Figure 2.1. Schematic representation of the major starch and lipid synthetic

pathways in C. reinhardtii. [ACCase, Acetyl-CoA carboxylase; ACP, acyl carrier

protein; ADP, adenosine diphosphate; AGPase, ADP-Glucose Pyrophosphorylase;

CoA, coenzyme A; P, phosphate; PtdOH, phosphatidic acid; WSP, Water Soluble

Polysaccharides]. Simplified from (Ball and Deschamps, 2009; Johnson and Alric,

2013; Riekhof and Benning, 2009).


41

The carbohydrate composition in microalgae is species-dependent, but most green species

accumulate carbohydrates in the form of starch granules within the chloroplasts (Vitova et

al., 2015). These granules are made up of amylopectin (a branched polysaccharide chain

composed of α-1,6 glycosidic linkages) and amylose (composed of α-1,4 linkages), with

the former accounting for the largest fraction and providing granules with a semi-

crystalline structure (Ball and Deschamps, 2009). Being a polymeric carbohydrate, starch

granules can be directed towards the production of sugar-based biofuels such as ethanol or

butanol (Markou et al., 2012a).

Microalgal lipids are found either in cellular membranes acting as structural components

(e.g. polar lipids such as phospholipids, sulfolipids, and galactolipids), or in the cytoplasm

in the form of oil bodies acting as energy storage reserves. Cytosolic oil bodies, which are

predominantly produced during stressed conditions, consist of neutral tricacylgliceride

(TAG) lipids (Riekhof and Benning, 2009). Unlike structural polar lipids, TAGs stand out

due to a composition richer in fatty acids, making them better suited for the production of

biodiesel via transesterification (Liu et al., 2011; McNichol et al., 2012).

Briefly, the rate controlling step for starch synthesis (Figure 2.1) is the conversion of

Glucose-1-P into ADP-Glucose, catalysed by the enzyme ADP-Glucose

Pyrophosphorylase (AGPase). Glucose units from ADP-Glucose are then transferred to

pre-existing water soluble polysaccharides (WSP), forming an elongating chain of

amylopectin and amylose by means of starch synthases and branching enzymes (Ball and

Deschamps, 2009). Lipid formation, on the other hand, is predominantly controlled by the

enzyme Acetyl-CoA carboxylase (ACCase), responsible for the production of malonyl-

CoA from which the biosynthesis of fatty acids initiates (Riekhof and Benning, 2009).

2.2.2. Selection of cultivation systems.

Cultivation technologies directed towards biofuel production must allow the mass-scale

production of microalgal biomass by providing those suitable species with adequate

conditions for growth, including a nutrient-rich aquatic environment, and adequate light,

temperature, and pH (Enamala et al., 2018). Currently, the two major types of commercial

algal cultivation systems are open or raceway ponds, and photobioreactors (Brennan and

Owende, 2010).


42

Open ponds are natural or artificial systems (e.g. lakes, lagoons, shallow tanks).

Meanwhile, raceway ponds are artificial systems made up of circle-shaped circuits coupled

with recirculation units (Brennan and Owende, 2010). Both systems are considered to be

the most cost-effective alternative for commercial algal growth since their operation

requires low energy inputs and maintenance. However, their open structure is prone to

contamination and the light and temperature fluctuations inherent to diurnal cycles are

difficult to control. Additionally, they might suffer from large water evaporation rates or

insufficient mixing (Lam and Lee, 2012).

Photobioreactors (PBRs), on the other hand, are closed systems (e.g. tubular, flat plate, or

column) designed to increase the control of cultivation parameters. They are typically built

from clear materials and light can be provided either naturally or artificially. PBRs increase

the surface area of the culture exposed to light and their closed structure prevents

contamination, making them appropriate for the cultivation of pure cultures (Brennan and

Owende, 2010). However, the lower volumes of PBRs and their high maintenance and

operating costs limits their large-scale implementation (Raslavičius et al., 2014; Su et al.,

2017).

Each of the above systems exhibit preferred advantages over the other, but the scalability

of the selected setup should predominantly be measured in terms of biomass productivity

given its impact on the energy efficiency of the subsequent stages. Following cultivation,

algal cells undergo a harvesting (dewatering) stage to separate biomass for further

treatment (Brennan and Owende, 2010). Common harvesting technologies (e.g.

centrifugation, filtration, and flocculation), however, are energy-intensive and become

unsuitable at large-scales due to the low biomass densities (usually lower than 1 g/L)

attained during cultivation (Barros et al., 2015; Rashid et al., 2014).

Microalgal cultivation systems should thus target high biomass productivity to ensure

reduced operational and processing costs. Nevertheless, achieving such a target can be

challenging for biofuel-oriented cultivation since those strategies that favour biomass

growth do not generally favour starch and lipids (Markou et al., 2012a). Successful

strategies should instead be capable of balancing the existing trade-off between growth and

starch and lipid formation.


43

2.3. Cultivation strategies targeting starch and lipid formation.

The starch and lipid contents of microalgae can increase in response to specific changes to

the cultivation environment, particularly under conditions of stress. Therefore, approaches

used to induce a favourable response in starch and lipid synthesis have relied on the

manipulation of major growth-limiting factors such as the nutrient concentration (Bajhaiya

et al., 2016), carbon source (Xu et al., 2006), and environmental factors such as light

intensity and temperature (Singh and Singh, 2015b). These strategies will be discussed

below.

2.3.1. Strategies based on light and temperature.

Light is the major energy supply for photosynthetic organisms, which thrive under natural

or artificial light whilst assimilating CO2. The intensity and type of light supplied to cells,

along with the photoperiod regime (i.e. the light/dark cycle) have been shown to play an

important role in microalgae growth as well as in starch and lipid accumulation (Blair et

al., 2014; Brányiková et al., 2010; Friedman et al., 1991; Ho et al., 2010; Rodolfi et al.,

2009). The effects of light intensity on microalgae are species-specific, but growth rates

typically increase with increasing light up to a maximum value, usually in a range of 200

- 400 μmol/m2s, after which photoinhibition effects take place (Markou et al., 2012a; Singh

and Singh, 2015b). Such inhibitory effects are thought to occur due to a disruption of the

membrane in chloroplasts and the inactivation of photosynthetic enzymes (Juneja et al.,

2013).

Starch and lipid accumulation may follow a similar response pattern with respect to

variations in light. Both the biomass dry weight and the starch content of the green species

Porphyridium aerugineum were shown to increase proportionally with increasing

illumination. Specifically, starch contents increased from 16.17 to 23.01 μg/106 cells as

the light intensity increased from 75 to 300 μmol/m2s (Friedman et al., 1991). Meanwhile,

the lipid contents in Nannochloropsis sp. increased from 15 % to 20 % when the light

intensity increased from 115 to 230 μmol/m2s. A further increase in lipids of up to 35 %

was attained by maintaining a light intensity of 230 μmol/m2s but changing from a one-

side illumination to a two-side illumination setup (Rodolfi et al., 2009).


44

The cultivation temperature can similarly regulate the growth of biomass and its

composition by affecting metabolic processes such as nutrient uptake and protein synthesis

(Juneja et al., 2013; Singh and Singh, 2015b). The temperature range appropriate for

microalgal growth is usually within 5 to 35°C (Rashid et al., 2014). However, similar to

light intensity, the optimal temperature suitable for cultivation systems is species-specific

(Singh and Singh, 2015b), and its appropriate manipulation can induce starch and lipid

formation. The effect of temperature on algal growth and storage molecule accumulation

are attributed to the denaturalisation of ribulose biphosphate carboxylase/oxygenase

(known as Rubisco), a major enzyme which participates in the intracellular mechanisms

governing carbon fixation (Zhan et al., 2017).

Identifying optimal light and temperature conditions is an important yet complicated

optimisation task which not only depends on the selected strain, but also on the depth,

shape, and mixing pattern of the cultivation system (Rashid et al., 2014). Although

establishing optimal strategies dependent on light and temperature are not within the scope

of this thesis, it is worth noting that relevant research aiming to address such a challenge

has been recently carried out in-house by Bekirogullari et al. (2018).

2.3.2. Strategies based on nutrient stress.

The aquatic environment in which microalgal growth takes place requires not only

appropriate light and temperature conditions, but also a balanced concentration of

nutrients. The lack of an adequate nutrient supply creates a nutrient-stressed environment

that leads cells to modify their metabolism and alter their composition. As previously

explained, this compositional change typically leads to higher accumulation of starch and

lipids. Nutrient stress is therefore recognised as one of the most cost-effective cultivation

strategies for biofuels production (Markou et al., 2012a).

The concept of nutrient stress has been equally and extensively referred to as nutrient

limitation to portray any restriction in growth arising from reduced nutrient supply. For the

sake of consistency and clarity, this thesis will consider both nutrient stress and nutrient

limitation to be equivalent concepts. However, it should be noted that nutrient stress may

also be a consequence of nutrient starvation, which specifically refers to the complete


45

removal of a nutrient from the culture medium, or to the specific period of growth right

after a nutrient that was initially present becomes exhausted (MacIntyre and Cullen, 2005).

The effects of nutrient limitation on starch and lipid accumulation are widely reported in

the literature, and in addition to being species-specific, they are also dependent upon the

limiting nutrient and its degree of limitation. A (non-exhaustive) list of studies

implementing nutrient-stressed strategies for increased starch or lipid accumulation is

presented in Table 2.2. In these studies, microalgal growth occurs via phototrophic,

heterotrophic, or mixotrophic carbon assimilation. Such mechanisms will be addressed in

the following section. Although nutrient limitation has been approached by various

nutrients (Table 2.2), this research focuses on the two macronutrients with the most

evaluated and validated starch and lipid increased responses: nitrogen and phosphorus.

Limitation by nitrogen has led to increased starch contents in C. vulgaris (Brányiková et

al., 2010; Dragone et al., 2011), Tetraselmis subcordiformis (Yao et al., 2012), and C.

zofigiensis (Zhu et al., 2014). Meanwhile, Scenedesmus obliquus (Ho et al., 2010),

Issochrysis aff. galbana (Ra et al., 2015), and C. protothecoides (Wang et al., 2017) have

similarly shown increased lipid contents when grown under nitrogen-limited conditions.

In C. reinhardtii, both starch and lipid contents have been shown to increase as a result of

nitrogen limitation (Bajhaiya et al., 2016; Ball et al., 1990; Wang et al., 2015), which has

helped position this species as a model organism suitable for the evaluation of starch and

lipid nutrient-limited responses (Moseley and Grossman, 2009).

Increases in starch and lipid caused by phosphorus limitation have similarly been reported

for C. reinhardtii (Bajhaiya et al., 2016; Ball et al., 1990), Nannochloropsis sp. (Rodolfi

et al., 2009), Scenedesmus sp. (Xin et al., 2010), and C. vulgaris (Brányiková et al., 2010).

It has been suggested that phosphorus limitation can predominantly induce the synthesis

of starch, rather than lipids (Wang et al., 2017). As shown in Table 2.2, for example, the

lipid contents of phosphorus-limited Arthrospira platensis decreased at the expense of

carbohydrate accumulation (Markou et al., 2012b).


46

(Ball et al., 1990; Brányiková et al., 2010; Dragone et al., 2011; Liu et al., 2011; Xin et

al., 2010)

Mic

roalg

ae

stra

in a

Gro

wth

condit

ions

b

Lim

itin

g

nutr

ient

Sto

rage m

ole

cule

resp

onse

c

Bio

mass

resp

onse

cR

efe

rence

Gro

wth

: P

hoto

trophic

Lig

ht:

4000 lx

Te

mp

: 25 °

C

Phosp

horu

s (P

)

Nitro

gen (

N)

Sulp

hur

(S)

Starch increase

d from 2-5 μg/10

6 c

ells

to:

20-60 μg/10

6 c

ells

(P

-sta

rved),

15-40 μg/10

6 c

ells

(N

-sta

rved),

50 μg/10

6 c

ells

(S

-sta

rved).

Bio

mass

dro

pped f

rom

2-5

x 1

06 c

ells

/mL

to:

0.2

-0.5

x 1

06 c

ells

/mL

(P

-sta

rved),

0.1

5-0

.4 x

10

6 c

ells

/mL

(N

-sta

rved)

0.5

x 1

06 c

ells

/mL

(S

-sta

rved).

Gro

wth

: H

ete

rotr

ophic

Lig

ht:

in

th

e d

ark

Te

mp

: 25 °

C

Phosp

horu

s (P

)

Nitro

gen (

N)

Sulp

hur

(S)

Starch increase

d from 4 μg/10

6 c

ells

to:

56 μg/10

6 c

ells

(P

-sta

rved),

40-70 μg/10

6 c

ells

(N

-sta

rved),

18 μg/10

6 c

ells

(S

-sta

rved).

Bio

mass

dro

pped f

rom

5 x

10

6 c

ells

/mL

to:

0.6

4 x

10

6 c

ells

/mL

(P

-sta

rved),

0.1

-0.4

x 1

06 c

ells

/mL

(N

-sta

rved)

0.4

x 1

06 c

ells

/mL

(S

-sta

rved).

Gro

wth

: M

ixotr

ophic

Lig

ht:

4000 lx

Te

mp

: 25 °

C

Phosp

horu

s (P

)

Nitro

gen (

N)

Sulp

hur

(S)

Starch increase

d from 0.4-4 μg/10

6 c

ells

to:

30-100 μg/10

6 c

ells

(P

-sta

rved),

20-100 μg/10

6 c

ells

(N

-sta

rved),

62 μg/10

6 c

ells

(S

-sta

rved).

Bio

mass

dro

pped f

rom

5-1

1 x

10

6 c

ells

/mL

to:

0.7

- 3

x 1

06 c

ells

/mL

(P

-sta

rved),

0.1

5 -

0.4

x 1

06 c

ells

/mL

(N

-sta

rved),

0.3

x 1

06 c

ells

/mL

(S

-sta

rved).

Scen

ed

em

us

sp.

LX

1

Gro

wth

: P

hoto

trophic

Lig

ht:

~60 µ

mol/m

2s

(14:1

0 h

, lig

ht:dark

)

Te

mp

: 25 °

C

Nitro

gen (

N)

Phosp

horu

s (P

)

Lip

id c

onte

nts

incre

ase

d f

rom

~ 2

5%

to:

~ 3

0 %

(N

-lim

ited),

~ 5

5 %

(P

-lim

ited).

Bio

mass

dro

pped f

rom

~0.5

g/L

to:

~ 0

.17 g

/L (

N-l

imited),

~ 0

.4 g

/L (

P-l

imited).

Xin

et

al. (

2010)

Ch

lore

lla

vu

lga

ris

CC

AL

A 9

24 (

P12)

Gro

wth

: P

hoto

trophic

Lig

ht:

780 µ

mol/m

2s

Te

mp

: 30 °

C

Nitro

gen (

N)

Phosp

horu

s (P

)

Sulp

hur

(S)

Sta

rch c

onte

nts

of

incre

ase

d f

rom

40 %

to:

55 %

(P

-sta

rved),

38 %

(N

-sta

rved),

60 %

(S

-sta

rved).

Bio

mass

dro

pped f

rom

~3 g

/L t

o:

~ 0

.6 g

/L (

P-s

tarv

ed),

~ 0

.2

g/L

(N

-sta

rved),

~ 1

g/L

(S

-sta

rved).

Brá

nyik

ová e

t al.

(2010)

Ch

lore

lla

vu

lga

ris

P12

Gro

wth

: P

hoto

trophic

Lig

ht:

70 µ

mol/m

2s

Te

mp

: 30 °

C

Nitro

gen (

N)

Iron (

Fe)

Sta

rch c

onte

nt

incre

ase

d f

rom

~5 %

to:

41 %

(N

-sta

rved),

and 2

3 %

(F

e-s

tarv

ed).

Bio

mass

dro

pped f

rom

~1.5

x 1

08 c

ell/

mL

to ~

5 x

10

7 c

ell/

mL

in

nutr

ient-

starv

ed c

ulture

s.

Dra

gone e

t al.

(2011)

Gro

wth

: P

hoto

trophic

Lig

ht:

30 µ

mol/m

2s

Te

mp

: 25 °

C

NA

Lip

ids

conte

nt: 1

0 %

of

cell

dry

weig

ht

Bio

mass

densi

ty: 1.9

g/L

Grw

oth

rate

: 0.2

35 d

-1

Gro

wth

: H

ete

rotr

ophic

Lig

ht:

In

th

e d

ark

Te

mp

: 25 °

C

NA

Lip

ids

conte

nt: 4

0%

of

cell

dry

weig

ht

Bio

mass

densi

ty: 9.7

g/L

Gro

wth

rate

of

0.7

69 d

-1

Ch

lam

yd

om

on

as

rein

ha

rdti

i

CC

126, C

C43,

CC

155

Ball

et

al. (

1990)

Ch

lore

lla

zofi

ng

ien

sis

AT

CC

30412

Liu

et

al. (

2011)

Tab

le 2

.2. L

ist

of

stu

die

s im

ple

men

tin

g n

utr

ien

t-st

res

sed

cu

ltiv

ati

on

str

ate

gie

s ta

rgeti

ng

incr

ease

d s

tarch

an

d l

ipid

form

ati

on

a I

f a

vail

ab

le,

the

com

ple

te s

pec

ies’

na

me

an

d s

tra

in I

D i

s p

rovi

ded

. b I

llu

min

ati

on i

s p

rovi

ded

co

nti

nuo

usl

y un

less

oth

erw

ise

spec

ifie

d.

c

Sto

rag

e m

ole

cule

s a

nd

bio

ma

ss r

esp

on

ses

are

co

mpa

red

ag

ain

st n

utr

ien

t-re

ple

te c

ond

itio

ns

un

less

oth

erw

ise

spec

ifie

d.


47

(Bajhaiya et al., 2016; Markou et al., 2012b; Mirzaie et al., 2016; Nzayisenga et al.,

2018; Yao et al., 2012; Zhu et al., 2014)

Art

hro

spir

a

pla

ten

sis

SA

G 2

1.9

9

Gro

wth

: P

hoto

trophic

Lig

ht:

120 µ

mol/m

2s

Te

mp

: 30 ±

1.5

°C

Phosp

horu

s (P

)

Lip

id c

onte

nts

dro

pped f

rom

8.2

0 %

to 3

.78 %

(P

-lim

ited).

Carb

ohydra

te c

onte

nts

incre

ase

d f

rom

10.9

9 %

to 6

6.6

0 %

(P

-lim

ited).

Bio

mass

dro

pped f

rom

:

1,9

31 ±

126 m

g/L

to

865 ±

42 m

g/L

(P

-lim

ited).

Mark

ou e

t al. (

2012)

Tetr

ase

lmis

sub

co

rdif

orm

is

Gro

wth

: M

ixotr

ophic

Lig

ht:

200 µ

mol/m

2s

Te

mp

: 25.2

± 2

°C

Nitro

gen (

N)

Sulp

hur

(S)

Sta

rch c

onte

nt

incre

ase

d f

rom

47.8

% t

o:

54 %

(N

-lim

ited),

62.1

% (

S-l

imited).

Sta

rch c

oncentr

ation d

ropped f

rom

2.7

g/L

to:

0.7

g/L

(N

-lim

ited),

1.2

g/L

(S

-lim

ited).

Bio

mass

dro

pped f

rom

5.7

g/L

to:

1.3

g/L

(N

-lim

ited),

2 g

/L (

S-l

imited).

Yao e

t al., (2

012)

Ch

lore

lla

zofi

ng

ien

sis

Gro

wth

: P

hoto

trophic

Lig

ht:

150 µ

mol/m

2s

Te

mp

: 25 °

C

Nitro

gen (

N)

Sta

rch c

onte

nt

incre

ase

d f

rom

9.7

% t

o:

43.4

% (

N-s

tarv

ed).

Bio

mass

dro

ppped f

rom

~3 g

/L t

o:

~0.7

g/L

(N

-sta

rved).

Zhu e

t al. (

2014)

Gro

wth

: P

hoto

trophic

Lig

ht:

65 µ

mol/m

2s

Te

mp

: N

A

NA

Lip

id c

onte

nt: 2

9 %

of

cell

dry

weig

ht

Lip

id c

oncentr

ation: 0.3

1 g

/L

Bio

mas

densi

ty: 1.0

8 g

/L

Gro

wth

rate

: 0.0

85 d

ay

-1.

Gro

wth

: M

ixotr

ophic

Lig

ht:

65 µ

mol/m

2s

Te

mp

: N

A

NA

Lip

id c

onte

nt: 3

3 %

of

cell

dry

weig

ht

Lip

id c

oncentr

ation: 0.8

6 g

/L

The b

iom

as

densi

ty r

eached 2

.62 g

/L

Gro

wth

rate

: 0.1

2 d

ay

-1

Ch

lam

yd

om

on

as

rein

ha

rdti

i

CC

125

Gro

wth

: M

ixotr

ophic

Lig

ht:

150 µ

mol/m

2s

(16:8

h, lig

ht:dark

)

Te

mp

: 22 °

C

Nitro

gen (

N)

Phosp

horu

s (P

)

Carb

ohydra

te a

nd lip

ids

incre

ase

d:

48.1

-fold

and 2

4.2

-fold

(N

-lim

ited),

and 1

4.5

-fold

and 5

.2-f

old

(P

-lim

ited),

resp

ectively

.

Wet

bio

mass

weig

ht

(~6.5

g/L

) dro

pped:

92 %

(N

-lim

ited),

and 8

7 %

(P

-lim

ited).

Bajh

aiy

a e

t al.

(2016)

Gro

wth

: P

hoto

trophic

Lig

ht:

120 µ

mol/m

2s

(16/8

h, lig

ht/

dark

)

Te

mp

: 25 °

C

NA

Carb

ohydra

te c

onte

nt: 5

3.1

2 %

Lip

id c

onte

nt: 1

3.4

%B

iom

ass

densi

ty: ~

1.1

g/L

Gro

wth

: M

ixotr

ophic

Lig

ht:

120 µ

mol/m

2s

(16/8

h, lig

ht/

dark

)

Te

mp

: 25 °

C

Glu

cose

(G

lu)

Gly

cero

l (G

ly)

Carb

ohydra

te c

onte

nt

(Gly

): 5

0.3

%

Carb

ohydra

te c

onte

nt

(Glu

): 4

2.9

%

Lip

id c

onte

nt

(Gly

): 1

0.5

%

Lip

id c

onte

nt

(Glu

: 14.6

%

Bio

mass

densi

ty (

Gly

): 1

.29 g

/L

Bio

mass

densi

ty (

Glu

): 1

.17 g

/L

Ch

lore

lla

sp.

Nza

yis

enga e

t al.

(2018)

Ch

lore

lla

vu

lga

ris

CC

AP

211/1

1B

Mir

zaie

et

al. (

2016)

Tab

le 2

.2. (c

on

t.)

Lis

t of

stu

die

s im

ple

men

tin

g n

utr

ien

t-st

res

sed

cu

ltiv

ati

on

str

ate

gie

s

targ

etin

g i

ncr

ease

d s

tarch

an

d l

ipid

form

ati

on

.

a I

f a

vail

ab

le,

the

com

ple

te s

pec

ies’

na

me

an

d s

tra

in I

D i

s p

rovi

ded

. b I

llu

min

ati

on i

s p

rovi

ded

co

nti

nuo

usl

y un

less

oth

erw

ise

spec

ifie

d.

c

Sto

rag

e m

ole

cule

s a

nd

bio

ma

ss r

esp

on

ses

are

co

mpa

red

ag

ain

st n

utr

ien

t-re

ple

te c

ond

itio

ns

un

less

oth

erw

ise

spec

ifie

d.


48

Nitrogen, found in proteins or nucleic acids, accounts for up to 20 % of the cell’s dry

weight. Meanwhile, phosphorus is estimated to make up about 1% of the cell dry weight,

and is found in biomolecules such as nucleic acids, membrane lipids (e.g. phospholipids)

or ATP molecules (Juneja et al., 2013). Nitrogen stress is thought to affect the

photosynthetic pathways responsible for protein and pigment synthesis by switching them

instead towards storage molecule accumulation. On the other hand, the link between

phosphorus limitation and increased starch accumulation has been associated to the

enzyme that regulates starch synthesis, ADP-Glucose Pyrophosphorylase (Figure 2.1),

which is inhibited in the presence of inorganic phosphorus (Markou et al., 2012a).

Despite the positive effects of nitrogen and phosphorus limitation on starch and lipid

synthesis, it is observed in Table 2.2 that both of these strategies are in most cases not

favourable for biomass growth. The drop in biomass growth, which reduces starch and

lipid overall productivity, is considered to be a consequence of the reduced concentration

of proteins responsible for photosynthetic mechanisms, evidenced by the typically high

carbohydrate:protein or lipid:protein ratios of nitrogen-limited and phosphorus-limited

cultures (Cade-Menun and Paytan, 2010; Dean et al., 2008; Juneja et al., 2013).

As explained before, nutrient-limited strategies may simply rely on the complete removal

of a nutrient from the cultivation media (i.e. nutrient starvation). Although such studies

clearly provide insights into the species-specific responses of starch and lipid synthesis

(Ball et al., 1990; Brányiková et al., 2010; Dragone et al., 2011), studies in which different

degrees of limitation are evaluated suggest an optimal limiting-nutrient concentration

suitable to avoid drastic reductions in biomass growth whilst still inducing starch and lipid

accumulation (Bajhaiya et al., 2016; Markou et al., 2012b).

The optimal trade-off between biomass growth and starch and lipid formation represents

one of the key challenges of nutrient-limited cultivation strategies, since their

implementation requires not only the adequate selection of growth-limiting nutrients, but

also the optimisation of media composition allowing for a well-balanced environment

where increases in starch and lipid contents are not ultimately overshadowed by low

biomass densities (Markou et al., 2012a).


49

2.3.3. Strategies based on carbon fixation mechanism.

Based on their carbon fixation route, microalgae can grow phototrophically,

heterotrophically, or mixotrophically (Brennan and Owende, 2010). The fixation

mechanism by which microalgae assimilate carbon can have an effect on biomass

productivities, and can therefore be exploited to overcome challenges faced by nutrient-

limited strategies. Each of these growth modes will be described next.

2.3.3.1. Phototrophic cultivation.

Phototrophic cultivation (also referred to as photoautotrophic), is the oldest and most

common cultivation method given that it relies on microalgae’s natural photosynthetic

ability to fixate CO2 using sunlight as an energy source (Lowrey et al., 2015; Zhan et al.,

2017). During photosynthesis, which takes place in the chloroplasts, CO2 is first fixated

and metabolised into 3 molecules of phosphoglycerate (3PGA), and subsequently into

glucose in a series of metabolic reactions using energy obtained from light (Venkata

Mohan et al., 2015).

Phototrophic cultivation is greatly acknowledged by research and commercial ventures due

to its contribution towards the reduction of anthropogenic CO2 waste emissions whilst

simultaneously co-producing valuable fuels and chemical products (Colling Klein et al.,

2018; Shuba and Kifle, 2018; Su et al., 2017). The scalability of outdoor phototrophic

cultivation, however, is restricted by: i) the atmospheric CO2 concentration levels which

are not high enough to solely sustain dense microalgal cultures (as desired for biofuel

production), and ii) the low solubility of CO2 in water (Scaife et al., 2015).

Cultivation systems are thus commonly supplemented with additional carbon streams such

as industrial flue gases or soluble carbonates (Colling Klein et al., 2018), needing constant

mixing to ensure accessibility of CO2 (Markou et al., 2012a; Rashid et al., 2014). In

addition, it is estimated that chlorophyll pigments (responsible for photosynthesis) can

only absorb up to 30 - 40% of the sunlight radiation, restricting the implementation of

cultivation systems to geographical locations with sufficient light and appropriate

temperatures (Scaife et al., 2015) and requiring systems designs with large surface area

and shallow depths (Venkata Mohan et al., 2015).


50

The concentration of CO2 may have significant effects on photosynthetic processes and

consequently on the growth of biomass and the assimilation of carbon within the cells

(Spalding, 2009). The microalgae Chlorella kessleri C-531, for instance, has been shown

to accumulate more starch (between 3 - 7 pg/cell) when grown in 0.04 % CO2, than when

grown in 3 % CO2 (between 0.5 – 3 pg/cell). In volumetric concentrations, however, starch

differences were not noticeable since the cell concentration obtained in low CO2 conditions

was lower than that obtained by high CO2 conditions (Izumo et al., 2007).

The effects of the CO2 concentration (ranging from 0.03 to 15 %) on lipid production was

evaluated in C. vulgaris CCTCC-M-209256 by Zheng et al. (2012). However, no

antagonistic responses between lipid and biomass were observed, as the lipid contents

remained relatively constant (~41 % of the cell dry weight) when CO2 concentrations

ranged between 0.03 – 10 %, and only dropped (~30 % of the cell dry weight) when CO2

was supplied at 15%. Because of these small changes, the highest lipid and biomass

productivities were both obtained at the same CO2 concentration of 5 %.

2.3.3.2. Heterotrophic cultivation.

Unlike phototrophic organisms who use light as their sole energy source, heterotrophs can

grow in the absence of light by using organic carbon substrates of low molecular weight

(e.g. pentoses, hexoses, acetic acid, or glycerol) as their energy source (Colling Klein et

al., 2018; Venkata Mohan et al., 2015). A downside of this cultivation approach is the most

expensive nature of organic carbon compounds than atmospheric or industrially generated

CO2. However, heterotrophic cultures generally attain much higher growth rates than

phototrophic ones (Colling Klein et al., 2018), a desired trait for biofuel production

purposes.

An additional benefit of heterotrophic cultivation is that the need to provide constant

illumination, be it natural or artificial, is avoided (Enamala et al., 2018), which simplifies

the design and operation of cultivation systems and consequently lowers production costs

(Zhan et al., 2017). Heterotrophic cultivation has also been suggested as a promising clean

alternative for waste water treatment plants, whereby microalgae can act as the primary

force behind the removal of organic carbon-based matter (thus overcoming costs resulting


51

from supplying expensive carbon substrates) and also as the precursor for biofuels or other

value-added chemicals (Adeniyi et al., 2018; Zhan et al., 2017).

In addition to the increased biomass densities, heterotrophic-based strategies have also

been shown to favour starch and lipid accumulation. For instance, when grown

photrotrophically C. protothecoides accumulates up to 10.62 % and 14.57 % of its cell dry

weight as starch and lipid molecules, respectively. However, when the same cells were

grown under heterotrophic conditions (using hydrolysates of corn as carbon source), the

intracellular contents of starch and lipids showed increases of 1.5-fold and 3.7-fold,

respectively (Xu et al., 2006).

An increase in lipid contents, from 10 to 40 %, has also been reported for C. zofingiensis

ATCC 30412 when cells are grown under phototrophic or heterotrophic conditions,

respectively (Table 2.2). The carbon fixation mechanism was additionally found to have

an effect on the lipid composition of C. zofingiensis, with heterotrophic cells synthesising

more neutral lipids (80.9 % of total lipids, of which 88.7 % is TAGs) than phototrophic

cultures (29.4 % of total lipids, of which 65.9 % is TAGs) (Liu et al., 2011). As mentioned

previously, neutral lipids (particularly TAGs) are the preferred choice for biodiesel

production due to their higher content of saponifiable fatty acids, making heterotrophic

cultures more advantageous.

A limitation of heterotrophic cultivation is the higher chances for culture contamination

due to the presence of organic carbon sources, usually preferred by other microorganisms

such as bacteria. Avoiding contamination requires either careful sterilisation and aseptic

methods which may prove expensive, or the optimisation of the carbon substrate

concentrations to fit the specific requirements of microalgae (Scaife et al., 2015).

2.3.3.3. Mixotrophic cultivation.

Mixotrophic cultivation employs species capable of growing both phototrophically and

heterotrophically, so that its internal carbon pool is maintained by the fixation of either

CO2 or organic compounds (Colling Klein et al., 2018; Venkata Mohan et al., 2015).

Mixotrophs are not as restricted by light conditions as phototrophs, so they exhibit a greater

resistance to photoinhibition effects and additionally yield higher biomass productivities


52

than phototrophic or heterotrophic cultures (Chojnacka and Noworyta, 2004; Enamala et

al., 2018).

Cultures grown mixotrophically may suffer from the limitations exhibited by phototrophic

and heterotrophic cultures, such as the need to optimise and control light intensity or the

greater chances for contamination (Zhan et al., 2017). In addition, the use of organic

substrates to support growth instead of exploiting waste CO2 streams is economically

questionable. However, mixotrophic cultivation is regarded a better suited strategy that, if

implemented adequately, could combine the reduction of atmospheric CO2 waste

emissions during phototrophic growth and the re-valorisation of organic matter present in

wastewaters during heterotrophic growth (Lowrey et al., 2015).

Although the starch and lipid contents of mixotrophic cells have been found not to differ

greatly from phototrophically grown cells (Ball et al., 1990; Mirzaie et al., 2016;

Nzayisenga et al., 2018), the enhanced growth rates and higher biomass densities attained

by mixotrophic cultivation (Chapman et al., 2015; Enamala et al., 2018; Mirzaie et al.,

2016) can ultimately lead to higher starch and lipid productivities, making mixotrophic

cultivation a suitable strategy for large-scale production of microalgae-based fuels.

The potential of mixotrophic cultivation for biodiesel production is particularly evidenced

in the works of Bekirogullari et al. (2017) and Mirzaie et al. (2016), where even though

the presence of an additional organic carbon source did not seem to affect the lipid contents

of C. reinhardtii 11/32C and C. vulgaris CCAP 211/11B, respectively, the higher cell

densities translated into higher lipid concentrations. Implementing mixotrophic strategies,

however, requires careful evaluation and further optimisation since both the type of carbon

source and its concentration can affect cell densities (Nzayisenga et al., 2018; Zhan et al.,

2017).

2.3.4. Strategies based on operating mode.

It has thus far been established that a major challenge for biofuel-oriented microalgal

cultivation is establishing strategies balancing the trade-off between biomass growth and

starch and lipid accumulation. Although nutrient-limited strategies increase the

intracellular contents of starch and lipids, the associated reduction of biomass hinders


53

overall productivity. Low biomass densities may be approached by mixotrophic cultivation

systems since they generally attain high biomass productivities.

The combined implementation of nutrient limitation and mixotrophic growth, however,

requires both cultivation strategies to be adequately integrated to attain optimal starch and

lipid yields whilst still sustaining biomass growth. The integration of these strategies can

be properly addressed by implementing two-stage or fed-batch operating systems rather

than standard batch systems. As shown in Table 2.3, cultivation strategies based on two-

stage or fed-batch systems can successfully maintain, or increase, biomass growth subject

to nutrient stress.

The principle behind two-stage cultivation is to initially grow cells in nutrient-sufficient

medium to attain high biomass densities. Once a suitable cell density is attained, cultures

are then transferred into a nutrient-stressed stage favouring starch and lipid accumulation.

Although this strategy has been validated successfully for increased lipid production in

phototrophic cultures (Table 2.3), the large-scale applicability of two-stage systems is

restricted by the large energy required to harvest cells: first to transfer them between each

stage, and then to separate them for final processing (Markou et al., 2012a).

Fed-batch operation, on the other hand, is a dynamic system in which nutrients are

intermittently, semi-continuously, or continuously supplied to cells. Since the effluents are

either removed discontinuously, or not withdrawn at all, fed-batch operation is also

referred to as a variable-volume culture system (Shuler and Kargi, 1992; Volesky and

Votruba, 1992). Fed-batch strategies have been widely used in biological processes to

extend the life of cultured cells and consequently product concentrations, or to tackle

substrate inhibition effects (Kiparissides et al., 2011; Shuler and Kargi, 1992).

Unlike two-stage cultivation strategies in which cells are simply transferred from a

nutrient-replete to a nutrient-limited system, fed-batch operation relies on the

implementation of optimal nutrient feeding strategies which can, on one hand, sustain high

biomass densities, and on the other, favour starch and lipid accumulation. Fed-batch

strategies have been proven successful for increasing starch and/or lipid yields against

standard batch systems in C. reinhardtii, Cyclotella sp., C. protothecoides, and

Desmodesmus sp. (Table 2.3).


54

(Ho et al., 2010; Jeffryes et al., 2013; Ji et al., 2015; Ra et al., 2015; Su et al., 2011)

Mic

roa

lga

e

stra

in a

Gro

wth

condit

ions

b

Ma

nip

ula

ted

va

ria

ble

(s)

Sto

rag

e m

ole

cule

resp

onse

Bio

ma

ss

resp

onse

Refe

rence

Scen

ed

esm

us

ob

liq

uu

s C

NW

-N

Gro

wth

: P

hoto

trophic

Mo

de

: T

wo-s

tage

Lig

ht:

60 µ

mol/m

2s

(continuous,

14:1

0 h

, lig

ht:dark

,

10:1

4 h

, lig

ht:dark

)

Te

mp

: 28 °

C

Fir

st

sta

ge

:

Nitro

gen (

N)

CO

2 (

C)

Lig

ht

(I)

Se

co

nd

sta

ge

:

Nitro

gen (

N)

Fir

st

sta

ge

(N

-re

ple

te):

Lip

id c

onte

nt

in n

utr

ient-

rich m

ediu

m (

at

10%

CO

2):

12.3

% o

f th

e c

ell

dry

weig

ht.

Se

co

nd

sta

ge

(N

-lim

ite

d):

Lip

id c

onte

nts

incre

ase

d u

p t

o 3

8.9

% a

fter

bein

g t

ransf

err

ed t

o N

-sta

rved m

ediu

m.

Fir

st

sta

ge

(N

-re

ple

te):

The h

ighest

concentr

ation (

3.5

g/L

) w

as

att

ain

ed

at

10 %

CO

2 a

nd

continuous

illim

ination.

Se

co

nd

sta

ge

(N

-lim

ite

d):

Bio

mass

densi

ty c

ontinued t

o incre

ase

aft

er

transf

er,

but

pro

ductivity d

ropped g

radually

.

Ho e

t al. (

2010)

Na

nn

och

loro

psi

s

ocu

lata

Gro

wth

: P

hoto

trophic

Mo

de

: T

wo-s

tage

Lig

ht:

300 µ

mol/m

2s

(continuous)

Te

mp

: 25 °

C

Sin

gle

sta

ge

:

None

Tw

o-s

tag

e:

Nitro

gen (

N)

Sin

gle

sta

ge

cu

ltiv

ati

on

:

Lip

id c

onte

nts

: 15.8

%.

Tw

o-s

tag

e c

ult

ivati

on

:

Lip

id c

onte

nts

: 48.2

%.

Sin

gle

sta

ge

cu

ltiv

ati

on

:

Bio

mass

densi

ty: 0.6

5 g

/L.

Tw

o-s

tag

e c

ult

ivati

on

:

Bio

mas

densi

ty: 0.8

6 g

/L.

Su e

t al. (

2010)

Cyclo

tell

a s

p.

Gro

wth

: P

hoto

trophic

Mo

de

: F

ed-b

atc

h

Lig

ht:

150 µ

mol/m

2s

(14:1

0 h

, lig

ht:dark

)

Te

mp

: 22 °

C

Sili

con (

Si)

Feed r

ate

(F

)

Feedin

g s

trate

gie

s consi

sted o

f an initia

l batc

h

phase

, fo

llow

ed b

y: i)

Si puls

e, ii)

Si perf

usi

on f

or

48 h

, iii

) 72 h

, or

iv)

96 h

, each y

ield

ing:

i)

~ 0

.7 g

/L lip

ids,

ii)

~ 1

.1 g

/L lip

ids,

iii)

~ 0

.9 g

/L lip

ids,

and iv)

~ 0

.9 g

/L lip

ids.

Feedin

g s

trate

gie

s consi

sted o

f an initia

l batc

h

phase

, fo

llow

ed b

y: i)

Si puls

e, ii)

Si perf

usi

on f

or

48 h

, iii

) 72 h

, or

iv)

96 h

, each y

ield

ing:

i)

~ 1

.8 g

/L b

iom

ass

, ii)

~ 2

.5 g

/L b

iom

ass

,

iii)

~ 2

.1 g

/L b

iom

ass

, and iv)

~ 3

g/L

bio

mass

.

Jeff

ryes

et

al. (

2013)

Desm

od

esm

us

sp.

EJ 1

5-2

Gro

wth

: P

hoto

trophic

Mo

de

: F

ed-b

atc

h

Lig

ht:

98 µ

mol/m

2s

Te

mp

: 30 ±

1 °

C

Feed r

ate

(F

)

Batc

h f

ind

ing

s:

Lip

id c

onte

nts

: ~

25.7

%.

Lip

id c

oncentr

ation: ~

83 m

g/L

.

Fe

d-b

atc

h f

ind

ing

s:

Lip

id c

onte

nts

: ~

25%

.

Lip

id c

oncentr

ation: ~

261.8

mg/L

.

Batc

h f

ind

ing

s:

Bio

mass

densi

ty: 0.3

24 g

/L, gro

wn f

or

10 d

ays

in a

naero

bic

dig

est

ion w

ast

ew

ate

r (A

DW

)

at

a 5

% v

/v d

ilution.

Fe

d-b

atc

h f

ind

ing

s:

AD

W w

as

loaded e

very

2 d

ays

for

40 d

ays.

Bio

mass

densi

ty: 1.0

39 g

/L.

Ji e

t al. (

2015)

Iso

ch

rysi

s g

alb

an

a

& *

oth

er

stra

ins

Gro

wth

: P

hoto

trophic

Mo

de

: T

wo-s

tage

Lig

ht:

109 µ

mol/m

2s

(12:1

2 h

, lig

ht:dark

)

Te

mp

: 20 ±

1 °

C

Nitro

gen (

N)

Fir

st

sta

ge

(N

-re

ple

te):

Lip

ids

conte

nts

: 24 %

of

the c

ell

dry

weig

ht.

Se

co

nd

sta

ge

(N

-lim

ite

d):

Fin

al lip

id c

onte

nts

: in

cre

ase

d t

o 4

7 %

*In

cre

ase

s in

lip

ids

were

obse

rved in a

ll oth

er

stra

ins

test

ed.

Fir

st

sta

ge

(N

-re

ple

te):

Bio

mass

densi

ty r

eached 0

.8 g

/L

Se

co

nd

sta

ge

(N

-lim

ite

d):

Fin

al bio

mass

densi

ty w

as

0.7

8 g

/L

Ra e

t al. (

2015)

Tab

le 2

.3. L

ist

of

stu

die

s im

ple

men

tin

g t

wo

-sta

ge

or

fed

-batc

h c

ult

ivati

on

stra

tegie

s fo

r in

crea

sed

sta

rch

an

d l

ipid

acc

um

ula

tion

.

a I

f a

vail

ab

le,

the

com

ple

te s

pec

ies’

na

me

an

d s

tra

in i

s p

rovi

ded

. b I

llu

min

ati

on i

s p

rovi

ded

co

nti

nuo

usl

y un

less

oth

erw

ise

spec

ifie

d.


55

(Fields et al., 2018; Wang et al., 2015; Wang et al., 2017)

Ch

lam

yd

om

on

as

rein

ha

rdti

i C

C1

25

Gro

wth

: P

hoto

trophic

Mo

de

:

Tw

o-s

tage, F

ed-b

atc

h

Lig

ht:

67,1

35

µm

ol/m

2s

Te

mp

: 25 °

C

Nitro

gen (

N)

CO

2 (

C)

Feed r

ate

(F

)

Batc

h o

pti

mis

ati

on

:

Da

ta n

ot

ava

ila

ble

Fir

st

sta

ge

(re

pe

ate

d f

ed

-batc

h):

Carb

ohydra

te c

onte

nts

(in

5 %

CO

2):

7.8

9 %

.

Se

co

nd

sta

ge

(N

-lim

ite

d):

Carb

ohydra

te incre

ase

d u

p t

o 5

0 %

(in

5%

CO

2)

or

70 %

(in

0.0

4 %

CO

2).

Batc

h o

pti

mis

ati

on

:

The h

ighest

bio

mass

pro

ductivity (

0.2

6 g

/L-d

)

was

att

ain

ed a

t 5%

CO

2 a

nd 1

35 µ

mol/m

2s

Fir

st

sta

ge

(re

pe

ate

d f

ed

-batc

h):

A r

epeate

d f

ed-b

atc

h (

repla

cin

g 4

5 %

of

culture

mediu

m)

yie

lded a

bio

mass

densi

ty o

f 0

.85 g

/L.

Se

co

nd

sta

ge

(N

-lim

ite

d):

Bio

mass

reached 0

.9 g

/L (

in 5

% C

O2)

or

1 g

/L

(in 0

.04 %

CO

2)

Wang e

t al. (

2015)

Ch

lore

lla

pro

toth

eco

ides

IOC

AS

03

8F

Gro

wth

: H

ete

rotr

ophic

Mo

de

: F

ed-b

atc

h

Lig

ht:

in

th

e d

ark

Te

mp

: 30 °

C

Batc

h:

Phosp

horu

s (P

)

Nitro

gen (

N)

Glu

cose

(G

)

Fe

d-b

atc

h:

Feed r

ate

(F

)

Batc

h f

ind

ing

s:

The h

ighest

lip

id c

onte

nt

(33.3

%)

was

obta

ined

under

N-l

imited &

hig

h g

lucose

conditio

ns.

Lip

id

conte

nts

did

not

change s

ignif

icantly in P

-lim

ited

conditio

ns.

Fe

d-b

atc

h f

ind

ing

s:

Syst

em

was

separa

ted into

phase

1 (

nutr

ient-

rich, continuous

feed)

and p

hase

2 (

stre

ssed,

feed s

topped):

betw

een e

ach p

hase

, lip

id

conte

nts

incre

ase

d f

rom

20.4

to 3

9.2

%.

Batc

h f

ind

ing

s:

Bio

mass

decre

ase

d in N

and P

-lim

ited c

onditio

ns.

The h

ighest

bio

mass

concentr

ation

(38.4

7 g

/L)

was

obta

ined a

t hig

h g

lucose

concentr

ations

and

N-P

-suff

icie

nt

conditio

ns.

Fe

d-b

atc

h f

ind

ing

s:

Betw

een

each p

hase

, bio

mass

changed f

rom

82.6

to 8

1.4

g/L

.

Wang e

t al. (

2017)

Ch

lam

yd

om

on

as

rein

ha

rdti

i C

C-

29

37

Gro

wth

: M

ixotr

ophic

Mo

de

: F

ed-b

atc

h

Lig

ht:

150 µ

mol/m

2s

Te

mp

: 30 °

C

Feed r

ate

(F

)

Batc

h f

ind

ing

s:

Fro

m t

he t

ota

l lip

ids

extr

acte

d (

no

t q

ua

nti

fied

),

27.3

5 %

are

non-p

ola

r lip

ids

(dia

cylg

lycero

ls a

nd

tria

cylg

lycero

ls).

Fe

d-b

atc

h f

ind

ing

s:

Fro

m t

he t

ota

l lip

ids

extr

acte

d (

not

quantifi

ed),

8.3

4 %

are

non-p

ola

r lip

ids

(dia

cylg

lycero

ls a

nd

tria

cylg

lycero

ls).

Batc

h f

ind

ing

s:

Bio

mass

densi

ty incre

ase

d f

rom

0.3

1±

0.0

2 t

o

2.2

9±

0.1

2 g

/L.

Fe

d-b

atc

h f

ind

ing

s:

Feed w

as

continuous

as

long a

s m

ediu

m p

H

rem

ain

ed a

bove 7

(via

pH

mete

r contr

ol)

.

Bio

mass

densi

ty incre

ase

d f

rom

0.4

5±0

.03 t

o

23.6

9±

0.5

g/L

.

Fie

lds

et

al. (

2018)

Tab

le 2

.3. (c

on

t.)

Lis

t of

stu

die

s im

ple

men

tin

g t

wo

-sta

ge

or

fed

-batc

h c

ult

ivati

on

stra

tegie

s fo

r in

crea

sed

sta

rch

an

d l

ipid

acc

um

ula

tion

.

a I

f a

vail

ab

le,

the

com

ple

te s

pec

ies’

na

me

an

d s

tra

in i

s p

rovi

ded

. b I

llu

min

ati

on i

s p

rovi

ded

co

nti

nuo

usl

y un

less

oth

erw

ise

spec

ifie

d.


56

A comparison of the biomass responses shown in Table 2.3 also indicates that the

integration of fed-batch operation with heterotrophic (Wang et al., 2017) or mixotrophic

(Fields et al., 2018) cultures has the potential to attain much higher biomass densities than

phototrophic-based fed-batch systems. Nevertheless, the success of fed-batch operation is

ultimately dictated by the complex task of establishing the most appropriate feeding

strategy, which often relies on multiple experimental and statistical analyses.

It is worth noting that the identification of optimal fed-batch feeding strategies can lead to

the development of more complex, yet enhanced and often desired, continuous cultivation

processes where nutrients are constantly and adequately supplied (e.g. maintaining stressed

conditions whilst avoiding nutrient exhaustion). Such a continuous strategy, provided it is

optimally applied, would allow the continuous production of biomass and therefore

increase the commercial viability of algal biofuels by greatly facilitating the scale up of

optimal biofuel-oriented cultivation systems (Colling Klein et al., 2018; Sforza et al.,

2014). Although continuous operation can offer plenty of benefits, the evaluation of such

a system is outside the scope of this thesis. However, and as will be shown in a subsequent

chapter, the work presented in this thesis provides useful experimental and modelling tools

that, if extrapolated correctly, may similarly be employed for the evaluation of continuous

cultivation systems (see Chapter 5).

2.4. Mathematical modelling of microalgae cultivation.

The preceding sections described various tailor-made cultivation strategies suitable for

biofuels production. The implementation of such strategies typically requires the

identification of optimal media composition (e.g. degree of nutrient limitation or optimal

carbon source and concentration), or optimal nutrient feeding regimes. Identifying such a

large number of growth-limiting and operating factors, however, can entail a detailed and

time-consuming experimental methodology.

A good design of experiments (DoE) can simplify the analysis of potential cultivation

strategies, but such tasks could be greatly facilitated by optimisation frameworks

employing predictive models describing the macroscopic input-output dynamics of

cultivation. Model-based optimisation does not only speed-up the implementation of


57

bioprocessing strategies, but also saves considerably more time and resources

(Kiparissides et al., 2011). For the purposes of biofuels production, modelling approaches

should portray the dynamics of algal growth and the formation of starch and lipids.

Existing models targeting such dynamics are discussed below.

2.4.1. Modelling algal growth dynamics.

Cellular growth is a process that results from cells being exposed to appropriate

physiochemical and nutrient-sufficient conditions. Nutrient uptake and cell replication

processes can be expressed simply as (Shuler and Kargi, 1992):

Substrates + Cells → Products + Cells

𝑆 + 𝑋 → 𝑃 + 𝑛𝑋 Eq. 2.1

where X, S, and P represent the concentration of cellular mass, substrates, and cellular

products, respectively. The rate at which microorganisms grow (so-called growth kinetics)

over a period of time is measured by the specific growth rate, μ, a powerful concept that

establishes, through adequate mathematical relationships, how cell growth changes in

response to its environment (Shuler and Kargi, 1992):

𝜇 =1

𝑋∙

𝑑𝑋

𝑑𝑡 Eq. 2.2

In microalgae, the specific growth rate refers to the rate of change of biomass through

photosynthetic and respiration processes (MacIntyre and Cullen, 2005). Such processes

are dependent on the nutrient availability and the environmental conditions, and the most

suitable expression of the specific growth rate should thus simulate these relationships.

2.4.1.1. Single-factor growth kinetic models.

In some cases, the specific growth rate expression is defined by a function of the single

most growth-limiting factor for each species, be it an extracellular (e.g. nutrient

concentration, light, temperature) or an intracellular (e.g. nutrient quotas) element (Figure

2.2). If the cultivation setup is assumed to provide optimal and controlled light and

temperature conditions to well-mixed cultures, the specific growth rate can simply be

expressed in terms of substrate or nutrients concentration (Lee et al., 2015a).


58

Figure 2.2. Extracellular and intracellular elements employed in growth kinetic

models for microalgae.

The relationship between microalgal growth and nutrient availability has been majorly

portrayed by two famous mathematical structures: the classic Monod model for

extracellular nutrients, and the Droop model for intracellular nutrients. These models,

along with other derived formulations, are explained below.

Models dependent on extracellular factors.

In Monod’s model (Eq. 2.3), the specific growth rate of cells limited by any given growth-

limiting substrate, S, is considered to increase hyperbolically with increasing substrate

concentration until a saturation point is reached and the growth rate attains a constant value

(Monod, 1949):

𝜇 = 𝜇𝑚𝑎𝑥 ∙𝑆

𝑆 + 𝐾𝑆 Eq. 2.3

Here, μmax is the maximum specific growth rate, and Ks is Monod’s half-saturation

constant. Monod’s model has been used to predict the effects of nitrogen, phosphorus, or

CO2 medium concentration in the growth of species such as C. reinhardtii, Chlorella sp.,

and Scenedesmus sp., among many others (Eriksen et al., 2007; Lee et al., 2015a; Xin et

al., 2010). However, the model’s simple structure is unable to capture complex processes


59

associated to microalgal growth, such as multiple-substrate limitation, substrate inhibition,

or photoinhibition.

To account for substrate inhibitory effects, growth kinetic models have employed instead

the formulation proposed by Andrews (1968) (Eq. 2.4), which only differs from the Monod

model by the incorporation of an additional term including an inhibition constant, Ki, that

allows the specific growth rate to decrease at high substrate concentrations:

𝜇 = 𝜇𝑚𝑎𝑥 ∙𝑆

𝑆 + 𝐾𝑆 + 𝑆2 𝐾𝑖⁄ Eq. 2.4

When compared against Monod’s classic model, Andrew’s formulation has been proven

to be much better suited, for example, to portray the inhibitory effects of high acetic acid

(i.e. carbon source) concentrations on the heterotrophic growth of C. reinhardtii, as

observed in both batch and fed-batch cultivation systems (Chen and Johns, 1994; Zhang et

al., 1999).

It should be mentioned that the models of Monod (Eq. 2.3) and Andrews (Eq. 2.4) can also

be used to predict the light-dependent growth of microalgal populations by simply

replacing the substrate concentration, S, by the light intensity, I, received by the culture

(Béchet et al., 2013). With regards to light-limited models, Molina-Grima et al. (1994)

proposed a Monod-type equation to portray both a lag-phase and sudden increase in the

microalgae’s growth rate as the intensity of illumination is increased.

𝜇 = 𝜇𝑚𝑎𝑥 ∙𝐼𝑛

𝐼𝑛 + 𝐾𝑆,𝐼𝑛 Eq. 2.5

In Eq. 2.5, 𝐾𝑆,𝐼 is a half-saturation constant associated to light-limited growth, and n is a

parameter that regulates the “shape” of the 𝜇 𝑣𝑠 𝐼 curve (if 𝑛 = 1, Eq. 2.5 is equivalent to

Monod’s model). A visual comparison between the model responses by Monod (Eq. 2.3),

Andrews (Eq. 2.4), and Molina-Grima (Eq. 2.5), based on the effects of substrate

concentration on growth rate, is presented in Figure 2.3.

Other available (yet less employed) growth models for substrate-limited microbial kinetics

include those of Contois, Blackman, Tessier, and Martinez-Sancho (Contois, 1959; Dabes

et al., 1973; Martínez Sancho et al., 1997).


60

Figure 2.3. Visual comparison of the μ vs S curves predicted by the growth kinetic

models of Monod, Andrews, and Molina-Grima.

Models dependent on intracellular factors.

Droop’s model (Eq. 2.6), is employed to describe microalgal growth as a function of the

intracellular nutrient availability represented by the cellular nutrient “quota”, q. The

nutrient quota is defined as the internal concentration of the growth-limiting nutrient with

respect to the cell concentration (Droop, 1968):

𝜇 = ��𝑚𝑎𝑥 ∙ (1 −𝑞𝑚𝑖𝑛

𝑞) Eq. 2.6

Here, ��𝑚𝑎𝑥 is Droop’s maximum specific growth rate, and 𝑞𝑚𝑖𝑛 is the minimum cell quota

(also known as subsistence quota) below which growth stops (𝜇 = 0 if 𝑞 < 𝑞𝑚𝑖𝑛 ). It

should be mentioned that Droop’s formulation of the specific growth rate defined the

nutrient uptake rate, 𝜌, as a function of the extracellular nutrient concentration by means

of the following expression:

𝜌 = 𝜌𝑚𝑎𝑥 ∙𝑆

𝑆 + 𝐾𝑆 Eq. 2.7

where 𝜌𝑚𝑎𝑥 is the maximum uptake rate of the growth-limiting nutrient, and 𝐾𝑆 is the half-

saturation constant for nutrient uptake.


61

Both Eq. 2.6 and Eq. 2.7 were originally employed by Droop to model the growth of the

marine species Monochrysis Lutheri as a function of its internal vitamin B12 quota (Droop,

1968). Since then, Droop’s equations have been used to model the growth and nutrient

uptake kinetics of I. galbana under nitrogen limitation (Mairet et al., 2011), Nitzchia sp.,

Tetraselmis subcordiformis, Uva pertusa, Scenedesmus sp. and Chlorella sp. under

phosphorus limitation (Grover, 1991; Nan and Dong, 2004), and Achnanthes sp., Amphora

sp., Navicula sp., and Nitzchia sp. under either nitrogen or phosphorus limitation (Kwon

et al., 2013).

An advantage of Droop’s model is that it can predict growth even after a nutrient becomes

depleted extracellularly (Lee et al., 2015a), as observed, for example, in phosphorus-

depleted cultures of C. protothecoides (Wang et al., 2017) and T. subcordiformes (Yao et

al., 2013). The nature of Droop’s equation (Eq. 2.6), however, introduces a maximum

specific growth rate which can only be attained at a hypothetical infinite cell quota (i.e.

𝜇 → ��𝑚𝑎𝑥 as 𝑞 → ∞). To improve the interpretation of quota-dependent growth, Caperon

and Meyer (1972) proposed to combine Droop’s quota-dependent kinetics with saturation-

type kinetics, as in:

𝜇 = 𝜇𝑚𝑎𝑥 ∙(𝑞 − 𝑞min )

𝐾𝑞 + (𝑞 − 𝑞𝑚𝑖𝑛) Eq. 2.8

where 𝐾𝑞 is equivalent to Monod’s half-saturation constant. Although the hyperbolic

nature of Caperon-Meyer’s model avoids the interpretation of an infinite cell quota, Droop

argued that the level of fit offered by Eq. 2.8 (with 3 parameters) is not superior to that of

Eq. 2.6 (with 2 parameters), making Droop’s model the preferred formulation (Droop,

1983).

2.4.1.2. Multiple-factor growth kinetic models.

Single-factor models predict well algal growth processes, but more accurate model

representations must evidently account for all those factors that simultaneously affect

microalgal growth. Although the increased level of accuracy of multiple-factor growth

models is inevitably achieved at the expense of increased mathematical complexity, such

models are more applicable for the evaluation and scale-up of cultivation systems given

their capacity to predict a wider range of cultivation scenarios. The interactions between


62

multiple growth-limiting factors and their effect on microalgal growth have thus far been

approached by three different mathematical structures: non-interactive, additive, and

interactive.

Non-Interactive models:

In non-interactive models (also referred to as threshold models), growth processes are

assumed to be affected by the most growth-limiting factor at any given time, for which the

following structure is adopted:

𝜇 = 𝜇𝑚𝑎𝑥 ∙ min [𝜇1(𝑆1), 𝜇2(𝑆2), … , 𝜇𝑛(𝑆𝑛)] Eq. 2.9

Here, min [𝜇1(𝑆1), 𝜇2(𝑆2), … , 𝜇𝑛(𝑆𝑛)] is a function commonly known as Liebig’s law of

the minimum (Cherif and Loreau, 2010), which establishes that the specific growth rate of

microalgae is equivalent to the lowest growth rate as determined by the most growth-

limiting substrate. The individual growth rate expressions for each limiting substrate (i.e.

𝜇1(𝑆1), 𝜇2(𝑆2), … , 𝜇𝑛(𝑆𝑛)) can adopt any of the model structures previously discussed in

Section 2.4.1.1. Therefore, non-interactive models ultimately take the form of single-factor

growth kinetic models.

The growth of I. galbana and Scenedesmus sp. co-limited by nitrogen and phosphorus has

been portrayed by non-interactive models integrating Droop-type kinetics within the

individual growth rates (as in Eq. 2.10), limited by either the nitrogen quota, 𝑞𝑁, or the

phosphorus quota, 𝑞𝑃 (Bougaran et al., 2010; Klausmeier et al., 2004).

𝜇 = 𝜇𝑚𝑎𝑥 ∙ min [1 −𝑞𝑚𝑖𝑛,𝑁

𝑞𝑁, 1 −

𝑞𝑚𝑖𝑛,𝑃

𝑞𝑃] Eq. 2.10

Here, 𝑞𝑚𝑖𝑛,𝑁 and 𝑞𝑚𝑖𝑛,𝑃 are the minimum nitrogen and phosphorous quotas sustaining

algal growth, respectively. In a similar fashion, Monod-type kinetics were used by

Spijkerman et al. (2011) to model the growth of Chlamydomonas acidophila subject to

CO2 or phosphorus limitation. Other models employing non-interactive formulations

include that of Packer et al. (2011), where the growth of Pseudochlorococcum sp. was

assumed to be limited by either the nitrogen quota or the light intensity. A more complex

approach, incorporating both Droop and Monod kinetics, was developed by Litchman et


63

al. (2006) to model the effects of nitrogen, phosphorus, iron, silicon, and light on the

growth of phytoplankton communities.

Additive models:

In additive models, the specific growth rate of microalgae is equivalent to the weighted

sum of the individual growth rate expressions of all growth-limiting factors (Shuler and

Kargi, 1992):

𝜇 = 𝜇𝑚𝑎𝑥 ∙ [𝑤1 ∙ 𝜇1(𝑆1) + 𝑤2 ∙ 𝜇2(𝑆2) + ⋯ + 𝑤𝑛 ∙ 𝜇𝑛(𝑆𝑛)] Eq. 2.11

Here, 𝑤𝑖=1,2,…𝑛 are weighing functions which determine the extent of limitation exerted by

each individual substrate on the maximum specific growth rate. The formulation of

adequate weighing functions is a rather empirical task, but in those cellular systems where

each growth rate expression, 𝜇(𝑆) , is described by simple Monod-type kinetics, the

weighing functions can potentially be expressed in terms of the concentration of each

substrate, 𝑆𝑖, at any given time, and their corresponding half-saturation constants, 𝐾𝑆,𝑖, as

follows (Shuler and Kargi, 1992):

𝑤𝑖(𝑆𝑖) 𝑖=1,2,…,𝑛 =𝐾𝑆,𝑖 𝑆𝑖⁄

𝐾𝑆,1 𝑆1⁄ + 𝐾𝑆,2 𝑆2⁄ + ⋯ + 𝐾𝑆,𝑖 𝑆𝑖⁄ Eq. 2.12

The above formulation was shown to be applicable to the growth of Saccharomyces

cerevisiae (a yeast) limited by various glucose and nitrogen concentrations, but it was

pointed out that providing a physiological interpretation for this model is challenging

(Mankad and Bungay, 1988). It should be emphasised, additionally, that Eq. 2.12 is purely

empirical and might not be applicable to all Monod-based cellular systems.

In order to showcase and provide a visual representation of the dynamics exhibited by the

weighing function formulation described above, a cellular growth system following

double-substrate limitation (i.e. Eq. 2.12, with n = 2) was simulated using simple Monod-

type kinetics for each substrate. The dynamics of this simulated system are shown in

Figure 2.4, where it is observed that cell biomass growth, and substrate consumption

follow typical microbial kinetics. The results of the simulation shown in Figure 2.4.c

additionally allow to observe the magnitude of the time-dependent weighing functions for

each limiting substrate as they are consumed, and their effect on overall cell growth.


64

Figure 2.4. Results of simulated double-substrate growth kinetics, as predicted by

Eq. 2.12: a) biomass, substrate 1, and substrate 2; b) specific growth rate; and c)

weighing functions. Feasible kinetic parameters and initial values (as shown in

table) were randomly selected for simulation purposes.

In addition to the challenging interpretation of Eq. 2.12, the appearance of the substrate,

𝑆𝑖, in the denominator of each saturating term (i.e. 𝐾𝑆,𝑖 𝑆𝑖⁄ ) may prevent the numerical

convergence of this equation when one or multiple substrates are completely exhausted

(i.e. 𝑆𝑖 = 0), requiring additional constraints to be taken into account. Such a problem

might be avoided by inverting the aforementioned term, so that 𝐾𝑆,𝑖 𝑆𝑖⁄ 𝑆𝑖 𝐾𝑆,𝑖⁄ .

Other growth model additive expressions may simply assume that the maximum specific

growth rate is unique to each individual substrate, such as (Yoon et al., 1977):

𝜇 = 𝜇𝑚𝑎𝑥,1 ∙ 𝜇1(𝑆1) + 𝜇𝑚𝑎𝑥,2 ∙ 𝜇2(𝑆2) + ⋯ + 𝜇𝑚𝑎𝑥,𝑛 ∙ 𝜇𝑛(𝑆𝑛) Eq. 2.13

Similar to non-interactive models, the individual growth rate expressions employed in

additive models can adopt single-factor model structures. Additive models for microalgal

growth include that proposed by Turon et al. (2014, where the combined growth-limiting


65

effects of acetic acid, 𝑆𝑎 , and butyric acid, 𝑆𝑏 , on the heterotrophic growth of C.

sorokiniana and C. protothecoides were described by the following expression:

𝜇 = 𝜇𝑎_𝑚𝑎𝑥 ∙𝑆𝑎

𝐾𝑆𝑎+𝑆𝑎+…

… + 𝜇max _𝑏 ∙𝐾𝐷

𝐾𝐷 + 𝑆𝑎∙

𝑆𝑏

𝑆𝑏 +𝜇𝑏_𝑚𝑎𝑥

𝛼∙ (

𝑆𝑏𝑆𝑏,𝑜𝑝𝑡

− 1)2

Eq. 2.14

Where 𝜇𝑎_𝑚𝑎𝑥 and 𝜇𝑏_𝑚𝑎𝑥 are the maximum specific growth rates by acetic and butyric

acid, respectively, 𝐾𝑆𝑎 is a half-saturation constant associated to acetic acid, 𝐾𝐷 is a

constant accounting for inhibitory effects of acetic acid on the growth driven by butyric

acid, and 𝑆𝑏,𝑜𝑝𝑡 is the optimal butyric acid concentration. Meanwhile, Adesanya et al.

(2014) assumed the mixotrophic growth rate of C. vulgaris to be equivalent to the sum

(𝜇𝑚 = 𝜇𝐴 + 𝜇𝐻) of the light-driven phototrophic growth rate, 𝜇𝐴, and the heterotrophic

growth rate, 𝜇𝐻.

Interactive models:

Interactive models incorporate the simultaneous effects of all potential growth-limiting

factors on microalgal growth through a multiplicative approach:

𝜇 = 𝜇𝑚𝑎𝑥 ∙ 𝜇1(𝑆1) ∙ 𝜇2(𝑆2) ∙ … ∙ 𝜇𝑛(𝑆𝑛) Eq. 2.15

The simple multiplicative nature of Eq. 2.15 has made interactive model formulations the

most common and widespread in the literature (Lee et al., 2015a), which can adopt either

the same kinetic expression (e.g. Monod, Droop, etc.) for each growth-limiting substrate,

or a combination of them, depending on the species or the cultivation scenarios.

For example, an integrated Monod multiplicative model proposed by Al Ketife et al. (2016)

was found to accurately predict the growth of C. vulgaris CCAP 211/11B limited by

nitrogen, N, phosphorus, P, CO2, C, and the average light intensity, Iave, received by the

culture:

𝜇 = 𝜇𝑚𝑎𝑥 ∙𝑆𝑁

𝐾𝑁 + 𝑆𝑁∙

𝑆𝑃

𝐾𝑃 + 𝑆𝑃∙

𝑆𝐶

𝐾𝐶 + 𝑆𝐶∙

𝐼𝑎𝑣𝑒𝑛

𝐾𝐼𝑛 + 𝐼𝑎𝑣𝑒

𝑛 Eq. 2.16


66

where 𝐾𝑁 , 𝐾𝑃, 𝐾𝐶 , and 𝐾𝐼 are the half-saturation constants for nitrogen, phosphorus,

carbon, and light, respectively; and n is a shape-controlling parameter such as that

employed by Molina-Grima et al. (1994).

Meanwhile, Yoo et al. (2014) proposed a growth kinetic model for C. protothecoides

UTEX B25 that integrated Droop kinetics for nitrogen-limited growth, Monod kinetics for

carbon-limited growth, and Andrews kinetics for light-limited growth, resulting in the

following expression:

𝜇 = 𝜇𝑚𝑎𝑥 ∙ (1 −𝑞0

𝑞) ∙

𝑆𝐶

𝐾𝐶 + 𝑆𝐶∙

𝐼

𝐾𝐼 + 𝐼 + 𝐼2 𝐾𝐼,𝑖⁄ Eq. 2.17

Where q0 is the minimum nitrogen quota required to sustain growth, 𝐾𝐶 is the half-

saturation constant for carbon-limited growth, and 𝐾𝐼 and 𝐾𝐼,𝑖 are the half-saturation and

inhibition constants associated to light-limited growth, respectively.

As shown in the comprehensive reviews of Béchet et al. (2013) and Lee et al. (2015),

interactive models have been widely employed to describe co-limitation effects of nutrients

and/or environmental conditions (e.g. light and temperature) on microalgal growth. Such

models are undoubtedly useful for simulation and optimisation purposes, but their

application is inherently restricted to the identification of biomass-enhancing strategies. If

models are to be used as optimisation tools targeting biofuels production, they must also

portray the formation of starch and lipids and their responses to various cultivation

conditions.

2.4.2. Modelling starch and lipid dynamics.

Predictive mathematical models capable of reflecting the dynamics of carbon assimilation

towards biomass growth, and also its partitioning between the internal starch and lipid

pools can become powerful tools for the optimal design of biofuel-oriented microalgal

cultivation plants. The microscopic-based modelling of starch and lipid formation requires

a detailed understanding of the reaction networks governing carbon uptake and its internal

distribution, which as previously established, are numerous and highly complex (Johnson

and Alric, 2013). However, starch and lipid dynamics can be established via macroscopic

models considering simplified reaction mechanisms and substrate-to-product interactions

similar to those already employed in microbial kinetics.


67

The formation of microbial products is usually expressed by the specific product formation

rate, 𝑞𝑃𝑟𝑜𝑑, as in:

𝑑𝑃

𝑑𝑡= 𝑞𝑃𝑟𝑜𝑑 ∙ 𝑋 Eq. 2.18

The specific product formation rate is generally dependent on whether the product is

(Shuler and Kargi, 1992): i) growth-associated, if their production rate is proportional to

the cell growth rate (i.e. 𝑞𝑃𝑟𝑜𝑑 = 𝛼 ∙ 𝜇), ii) non-growth-associated if they are produced at

a constant rate during the stationary phase (i.e. 𝑞𝑃𝑟𝑜𝑑 = 𝛽), or iii) mixed-growth associated

if production occurs during both linear and stationary growth phases, in which case the

rate of product formation is described by the famous Luedeking-Piret equation:

𝑑𝑃

𝑑𝑡= 𝛼 ∙ 𝜇 ∙ 𝑋 + 𝛽 ∙ 𝑋 Eq. 2.19

where 𝜇 is the specific growth rate limited by either single or multiple factors, and 𝛼 and

𝛽 are growth-related constants, so that if 𝛼 = 0 , product formation is non-growth

associated, and if 𝛽 = 0, product formation is growth-associated.

Eq. 2.19 has been employed to simulate lipid production in Chlorella salina and

Nannochloropsis oculata (Surendhiran et al., 2014). However, the nature of Luedeking-

Piret’s equation restricts its use only to those cases in which the formation of cellular

products can be described by the same kinetic relationships that simulate growth. Whilst

cell growth and starch and lipid formation may in fact be affected by the same factors (e.g.

nutrient availability, light, temperature), their responses are usually antagonistic (Table

2.2). Therefore, the accurate simulation of starch and lipid dynamics may require that

product formation kinetics be adequately uncoupled from cell growth kinetics.

A list of existing models accounting for the kinetics of starch and/or lipid formation is

presented in Table 2.4, which additionally includes: i) the cultivation variables predicted

by the model, ii) the type of operating mode the model is employed for (i.e. batch or fed-

batch), and ii) the type of kinetics used to simulate microalgal growth.

Since starch and lipids are both intracellular products, modelling works often exclude these

components from the total biomass and consider them as separate elements. Packer et al.

(2011) and Bekirogullari et al. (2017), for example, proposed macroscopic kinetic models


68

for the simulation of nitrogen-limited microalgal growth and lipid dynamics by separating

total biomass into a fat-free biomass fraction, and a lipid fraction. The model of Packer et

al. (2011), which also considered limitation by light, was developed to predict the

phototrophic batch growth of Pseudochlorococcum sp. This model was built under the

basis that lipid synthesis results from an excess of the photosynthetically assimilated

carbon with respect to the minimal carbon (subsistence) quota required for cellular growth,

i.e. once the cell quota reaches its minimum value, any increase in biomass weight

originates from lipid formation.

The model of Bekirogullari et al. (2017) followed a similar logic to account for nitrogen

and light limitations, but also incorporated the effects of acetic acid (using inhibition-type

kinetics) on the production of lipids during the mixotrophic batch cultivation of C.

reinhardti. The accurate simulation of lipid production by microalgae is of great interest

for biodiesel production. However, starch is often the preferred carbon sink of microalgal

cells (Chen et al., 2013; Fan et al., 2012). Thus, models capable of accounting for the

simultaneous dynamics of starch and lipid formation further improve the assumption that

any excess of assimilated carbon is only directed towards lipid formation.

A model accounting for both starch (sugars) and lipid dynamics was proposed by Mairet

et al. (2011), where the cell was compartmentalised into three carbon-based pools (e.g.

sugar, lipid, and functional biomass). The formation of each of these compartments was

adequately portrayed via Droop-based kinetic expressions which simulated the carbon

metabolism of I. galbana during nitrogen-limited phototrophic growth.

Following the cellular compartmentalisation proposed by Mairet et al. (2011), an

alternative quota-model accounting for sugar and lipid dynamics was later developed by

Kumar et al. (2016), but also considering phosphorus and temperature effects. However,

given the biodiesel-oriented nature of this work, the model was only evaluated for its

capacity to predict biomass and lipid formation (against data obtained from phototrophic

cultures of Dunaliella tertiolecta). Therefore, unlike the work of Mairet et al. (2011) where

all state variables were validated experimentally, the model of Kumar et al., (2016) was

not validated for its capacity to predict nutrient consumption dynamics nor the sugar

concentration profile.


69

(Adesanya et al., 2014; Bekirogullari et al., 2017; Jeffryes et al., 2013; Kumar et al.,

2016; Mairet et al., 2011; Packer et al., 2011; Surisetty et al., 2010)

Op

era

tin

g

mo

de

Bio

mass

Carb

on

Su

bstr

ate

Nu

trie

nt

Sta

rch

Lip

idL

igh

tT

em

pp

HL

imit

ing

nu

trie

nt(

s)

Ad

dit

ion

al

rem

ark

sR

efe

ren

ce

Fed

-bat

chR

RR

R

RR

RN

itro

gen

Str

ain

: A

uxen

och

lore

lla

pro

toth

ecoid

es

Gro

wth

: H

eter

otr

ophi

c

Gro

wth

kin

eti

cs:

Dro

op-t

ype

kin

etic

s

No

. of

kin

eti

c pa

ram

ete

rs:

12

Sur

iset

ty e

t al

.

(2010)

Bat

chR

R

R

R

N

itro

gen

Str

ain

: P

seudoch

loro

cocc

um

sp.

Gro

wth

: P

hoto

trop

hic

Gro

wth

kin

eti

cs:

Dro

op-t

ype

No

. of

kin

eti

c pa

ram

ete

rs:

12

Pac

ker

et al

.

(2011)

Bat

chR

R

RR

N

itro

gen

Str

ain

: Is

och

rysi

s aff

inis

galb

ana

Gro

wth

: P

hoto

trop

hic

Gro

wth

kin

eti

cs:

Dro

op-t

ype

No

. of

kin

eti

c pa

ram

ete

rs:

9

Mai

ret et

al.

(2011)

Fed

-bat

chR

R

R

R

S

ilico

n

Str

ain

: C

yclo

tella

sp.

Gro

wth

: P

hoto

trop

hic

Gro

wth

kin

eti

cs:

Dro

op-t

ype

No

. of

kin

eti

c pa

ram

ete

rs:

11

Jeffry

es e

t al

.

(2013)

Bat

chR

RR

R

N

itro

gen

Str

ain

: C

hlo

rella

vulg

ari

s

Gro

wth

: M

ixotr

ophi

c

Gro

wth

kin

eti

cs:

Dro

op-t

ype

No

. of

kin

eti

c pa

ram

ete

rs:

12

Ades

anya

et al

.

(2014)

Bat

chR

R

RR

RR

N

itro

gen

Pho

spho

rus

Str

ain

: D

unaliel

la t

erti

ole

cta

Gro

wth

: P

hoto

trop

hic

Gro

wth

kin

eti

cs:

Dro

op-t

ype

No

. of

kin

eti

c pa

ram

ete

rs:

14

Kum

ar e

t al

.

(2016)

Bat

chR

RR

R

RR

RN

itro

gen

Str

ain

: C

hla

myd

om

onas

rein

hard

tii

Gro

wth

: M

ixotr

ophi

c

Gro

wth

kin

eti

cs:

Mono

d-t

ype

No

. of

kin

eti

c pa

ram

ete

rs:

25

Bek

irogu

llari e

t al

.

(2017)

Vari

ab

les i

nclu

de

d i

n t

he

mo

de

l

R

Tab

le 2

.4. L

ist

of

mic

roalg

ae-

base

d k

inet

ic m

od

els

inco

rpora

tin

g s

tarc

h a

nd

/or

lip

id d

yn

am

ics.


70

The kinetic model proposed by Adesanya et al. (2014), which as explained in Section

2.4.1.2 incorporated an additive model structure to simulate the mixotrophic growth rate

of Chlorella vulgaris, additionally aimed to portray the formation of sugars and lipids

during nitrogen-limited growth. Although model predictions were validated

experimentally, the formulation employed by Adesanya et al. (2014) considered the algal

cell to be divided only into a storage compartment (made up of both starch and lipid

molecules) and a functional compartment, which prevented the identification of the

individual starch and lipid profiles.

As indicated by the above works, efforts are being made towards the development of

microalgae-based models addressing the formation of intracellular storage molecules.

Once models’ predictability is successfully validated, model-based optimal cultivation

conditions can be reliably identified. For example, based on the model developed by

Kumar et al. (2016), an optimisation study was recently carried out by Sinha et al. (2017)

to identify model-based conditions maximising biomass and lipid productivity, or

minimising cultivation costs. Optimal model-based cultivation conditions for maximal

lipid productivity were also identified, and further validated experimentally, in the work

of Bekirogullari et al. (2017). Such conditions were shown to yield up to 33 % increase in

lipid with respect to non-optimised conditions.

Although not many modelling works take on the task of further identifying optimal

cultivation scenarios, the works of Sinha et al. (2017) and Bekirogullari et al. (2017)

highlight the applicability of macroscopic models as powerful optimisation or scaling-up

tools. If models are going to be exploited as such, it is preferred that they are also assessed

for their capacity to predict fed-batch cultivation dynamics given their potential to sustain

both biomass and starch and lipid yields. However, a limited number of microalgal models

have been specifically developed (or validated) to describe fed-batch systems (some of

which are shown in Table 2.4), and those that have successfully done so have accounted

for only lipid formation (Jeffryes et al., 2013; Surisetty et al., 2010) rather than both starch

and lipids.


71

2.5. Concluding remarks.

Microalgae are positioned as a promising feedstock for biofuels given their ability to

intracellularly synthesise two major carbon-based elements: i) starch: a polymeric

carbohydrate which could be directed towards sugar-based fuels such as bioethanol of

biobutanol, and ii) lipids: oily bodies with the potential to be directed towards biodiesel

production via transesterification. If microalgae are to become commercially viable

feedstocks for biofuels production, optimal cultivation strategies maximising starch and

lipid production must be carefully identified.

The review of literature presented in this Chapter aimed to provide a clearer understanding

of the current state of developments, both experimental and computational, on the

optimisation of biofuel-oriented microalgal cultivation. The major findings identified in

this review are summarised below:

Nitrogen-limited and/or phosphorus-limited cultivation (i.e. where cellular stress is

artificially induced by reducing nitrogen or phosphorus availability) has been, thus

far, the most widely acknowledged strategy for increased starch and lipid

accumulation. Nutrient limitation can reduce biomass growth, but such an

undesirable outcome can be approached by employing mixotrophically grown

cultures, rather than phototrophic ones, since the former generally attain higher

biomass densities (Table 2.2).

Nutrient limitation and mixotrophic cultivation have been shown to be efficiently

integrated within fed-batch operating systems, whereby both high biomass

densities and increased starch and lipid accumulation can be maintained via

appropriate nutrient feeding strategies (Table 2.3). However, the implementation

of optimal fed-batch strategies is a challenging task that requires the optimisation

of both the media composition and the feeding strategy.

Predictive models capable of simulating the dynamics of algal growth and starch

and lipid formation, subject to single or multiple growth-limiting factors, can

facilitate the identification of optimal media composition and/or nutrient feeding


72

strategies. In this regard, a large number of models have been developed to predict

microalgae growth, but few of them can similarly address the more complex, yet

necessary, dynamics of starch and lipid formation during nutrient-limited growth

(Table 2.4).

The above points are the main drivers behind the research Contributions of this thesis

which will be presented in the following Chapters.

73

Chapter 3

Kinetic Modelling of Starch and Lipid Formation during

Mixotrophic, Nutrient-Limited

Microalgal Growth

3.1. Introduction.

As mentioned in Chapter 1, third-generation biofuels produced from microalgal biomass

are promising and long-term transport energy alternatives to fossil fuels. Microalgae-to-

biofuel technologies, however, are not yet commercially viable due in part to the crucial

but challenging task of establishing cultivation systems highly productive for not only algal

biomass, but also for the two major biofuel precursors: starch and lipids.

The Literature Review presented in Chapter 2 validated nitrogen limitation as a suitable

microalgal cultivation strategy for increased starch and lipid formation. Nitrogen-limited

strategies, however, can become problematic since they can lead to low biomass densities

and ultimately reduce the volumetric production yields of starch and lipids. The

undesirable reduction of microalgal biomass densities can be addressed by employing

mixotrophic cultures which, when compared to phototrophic cultures, generally attain

higher biomass densities.

Therefore, the implementation of a nitrogen-limited mixotrophic cultivation system is a

suitable strategy for the production of microalgal biofuels. In order for this strategy to yield

maximal starch and lipid formation, media composition should be optimised in terms of

the nitrogen and carbon concentrations to provide cells with: i) an optimal degree of

nitrogen limitation inducing starch and lipid accumulation, and ii) sufficient organic

carbon to maintain mixotrophic cell growth.

Chapter 3 – Kinetic Modelling of Starch and Lipid Formation during Mixotrophic,

Nitrogen-Limited Microalgal growth

74

Optimisation of media composition is often time-consuming and expensive, but it can be

facilitated by means of predictive kinetic models capable of effectively simulating the

simultaneous interactions of multiple growth-limiting nutrients and their corresponding

effects on biomass growth and starch and lipid formation. In this regard, numerous models

have been developed to predict microalgal growth by employing well-known kinetic

expressions such as those proposed by Monod, Droop, or Andrews which exhibit a wide

variety of simple yet powerful predictive traits (see Chapter 2).

Fewer models, however, have aimed to simulate the dynamics of starch and lipid

molecules. From the existing models accounting for starch and/or lipid formation and their

responses to a changing cultivation environment, none has yet simulated nitrogen-limited

mixotrophic cultivation dynamics. The paper that follows addresses this gap by presenting

an experimentally validated multi-parametric kinetic model capable of simulating

mixotrophic algal growth and starch and lipid formation, responsive to the initial nitrogen

and organic carbon medium concentrations.

The kinetic expressions employed in the model to represent the relationships between the

evaluated nutrients (inputs) and their effects on biomass, starch, and lipids (outputs) were

developed based on the evaluation of data collected from in-house cultivation

experiments* and the careful integration of modelling expressions available in literature.

As will be shown throughout the text, the model’s value as an optimisation tool was

successfully exploited by identifying starch and lipid enhancing cultivation strategies.

* Note: The experimentation included in this research was carried out with the green

microalgae species Chlamydomonas reinhardtii (strain CCAP 11/32C), a model organism

whose central carbon metabolism has been widely studied (see Chapter 2). The selected

strain grows mixotrophically in Tris-Acetate-Phosphate (TAP) medium, containing acetic

acid as organic carbon source and ammonium chloride (NH4Cl) as main nitrogen source.

The experimental data presented in this work was thus obtained by growing the strain under

various ammonium chloride and acetic acid initial concentration regimes. The detailed

preparation of TAP medium and the concentration of all components are presented in

Appendix A.



75

3.2. Contribution 1.

Figueroa-Torres GM, Pittman JK, Theodoropoulos C. (2017). Kinetic modelling of

starch and lipid formation during mixotrophic, nutrient-limited microalgal growth.

Bioresource Technology. 241:868–878.

DOI: 10.1016/j.biortech.2017.05.177

Authors’ contribution:

Gonzalo M. Figueroa-Torres performed the experimental and computational tasks

associated to this work, analysed data, and wrote the manuscript.

Jon K. Pittman co-supervised the research, and revised the manuscript.

Constantinos Theodoropoulos contemplated and supervised the research, reviewed, and

edited the manuscript.



76

Kinetic Modelling of Starch and Lipid Formation

during Mixotrophic, Nutrient-limited Microalgal

Growth

Gonzalo M. Figueroa-Torres a, Jon K. Pittman b, Constantinos Theodoropoulos a,*

a School of Chemical Engineering and Analytical Science, Biochemical and Bioprocess

Engineering Group, The University of Manchester, Manchester, M13 9PL, UK.

b School of Earth and Environmental Sciences, The University of Manchester,

Manchester, M13 9PT, UK.

* Corresponding author:

Prof. Constantinos Theodoropoulos

E-mail: [email protected]

mailto:[email protected]



77

ABSTRACT

Microalgal starch and lipids, carbon-based storage molecules, are useful as potential

biofuel feedstocks. In this work, cultivation strategies maximising starch and lipid

formation were established by developing a multi-parameter kinetic model describing

microalgal growth as well as starch and lipid formation, in conjunction with laboratory-

scale experiments. Growth dynamics are driven by nitrogen-limited mixotrophic

conditions, known to increase cellular starch and lipid contents whilst enhancing biomass

growth. Model parameters were computed by fitting model outputs to a range of

experimental datasets from batch cultures of Chlamydomonas reinhardtii. Predictive

capabilities of the model were established against different experimental data. The model

was subsequently used to compute optimal nutrient-based cultivation strategies in terms of

initial nitrogen and carbon concentrations. Model-based optimal strategies yielded a

significant increase of 261% for starch (0.065 gC L-1) and 66% for lipid (0.08 gC L-1)

production compared to base-case conditions (0.018 gC L-1 starch, 0.048 gC L-1 lipids).

Keywords: Biofuels, microalgal dynamics, kinetic modelling, starch and lipids

optimisation, Chlamydomonas reinhardtii



78

1. Introduction.

Our current dependence on fossil fuels raises two major concerns: the overexploitation of

finite crude oil resources and the associated emissions of greenhouse gases (GHG) leading

to global warming (Scaife et al., 2015). About 60 % of the fossil fuels directed annually

towards primary energy consumption are taken up by the transportation sector (Escobar et

al., 2009). Although biofuels have emerged as a suitable and renewable replacement for

transport-associated fuels such as gasoline and diesel, sustainable and cost-effective

biofuel production systems must first be developed (Escobar et al., 2009; Scaife et al.,

2015).

Due to exhibiting faster growth rates than terrestrial plants, microalgal biomass has

recently been considered as a potential biofuel feedstock (Brennan and Owende, 2010), as

opposed to traditional food-based or lignocellulosic substrates which compete for food or

arable land (Scaife et al., 2015). Microalgae are photosynthetic organisms which

synthesize biologically important compounds such as carbohydrates, lipids, proteins and

nucleic acids, acting as storage or functional elements (Brennan and Owende, 2010; Choix

et al., 2012). In particular, their internal pool of carbohydrates and lipids has directed the

attention towards the use of microalgae as a renewable feedstock for sugar and lipid-based

fuels.

In green microalgae, the main storage carbohydrate synthesized by cells is starch, which

is located within the chloroplast in the form of granules (Choix et al., 2012; Markou et al.,

2012). Meanwhile, oil bodies are found in the cytosol and chloroplast (Ball and

Deschamps, 2009; Goodson et al., 2011). Cellular contents of starch and lipids

(triacylglycerol, TAG) have been shown to increase under nitrogen-starved conditions

(Bajhaiya et al., 2016). Nevertheless, enhanced starch and lipid accumulation rates come

hand-in-hand with a decrease in biomass growth (Markou et al., 2012), suggesting the need

to identify an optimal balance between microalgal growth and starch and lipid

accumulation.

Microalgal cells can be photototrophic or heterotrophic depending on whether their carbon

fixation route requires the presence of an inorganic or an organic carbon source (Brennan

and Owende, 2010). Some strains are able to grow mixotrophically by utilizing both

inorganic and organic carbon sources. In most cases, mixotrophic cultivation leads to



79

improved growth rates against standard phototrophic conditions (Chapman et al., 2015;

Johnson and Alric, 2013), causing substantial increases in biomass productivities (Moon

et al., 2013). Consequently, quantifying starch and lipid accumulation during nitrogen-

limited mixotrophic growth is of great relevance for microalgae-based biofuels.

Chlamydomonas reinhardtii, the chosen microalgal strain in this work, has been widely

studied (e.g. Goodenough et al., 2014; Bajhaiya et al., 2016). Its carbon metabolism is well

known (Johnson and Alric, 2013), making it suitable for the analysis of the carbon

assimilation and its distribution between starch and lipid reserves. In particular, a clear

increase in starch and lipid accumulation for mixotrophic growth under nutrient stress has

been observed for this strain (Bajhaiya et al., 2016; Bekirogullari et al., 2017).

Nevertheless, despite nutrient stress being regarded as a simple and cost-effective strategy

to enhance starch and lipid formation, optimisation of the process is required for viable

large-scale cultivation.

Robust kinetic models capable of simultaneously predicting starch and lipid formation can

significantly aid in the establishment of optimal cultivation strategies. Models that take

into account the structured and segregated (i.e. each cell behaves as an individual unit with

dynamic composition) nature of cells, and/or even the stochastic nature of cell growth

(Alonso et al., 2014) can realistically predict the formation of multiple intracellular

components as well as cells’ response to cultivation conditions. Such models can provide

useful insights about the algal metabolic networks and intracellular fluxes (Chapman et al.,

2015; Rügen et al., 2012), but are usually highly complex and computationally expensive

(Shuler and Kargi, 1992). Unstructured non-segregated models, on the other hand, have a

simpler formulation (assume all cells in culture are identical), but have been shown to be

applicable to practical algal cultivation systems.

Most of the existing unstructured models for microalgae, however, have focused solely on

the simulation of lipid production (Bekirogullari et al., 2017; Packer et al., 2011). These

models assume that lipid formation is a consequence of excess carbon (between the amount

fixated and the amount required for cell growth) directed towards synthesis of lipids rather

than other carbon-based elements, such as cellular organelles or proteins. Although this

assumption is in agreement with the carbon pathways of microalgae, any excess of

assimilated carbon is also directed towards formation of starch reserves (Johnson and



80

Alric, 2013). Models for nutrient-limited algal growth considering both sugar and lipids

dynamics have been proposed by Mairet et al., 2011b and Kumar et al., 2016, but with a

focus on lipid production during phototrophic growth. Mixotrophic dynamics were

proposed by Adesanya et al., 2014, but encompassing sugar and lipids into one single

storage molecule, preventing the identification of each individual profile.

Thus, the aim of this work is to develop a predictive multi-parameter model for the

simultaneous optimization of starch and lipid formation during nitrogen-limited

mixotrophic microalgal growth. Our proposed model couples both carbon (C) and nitrogen

(N) substrates. The model is fitted and validated against datasets obtained from lab-scale

culture experiments under different N and C regimes. The validated model can then be

used with confidence for the identification of optimal cultivation conditions for maximum

starch and lipid production.

2. Materials and Methods.

2.1. Strain and cultivation.

All experiments were carried out with the wild-type strain C. reinhardtii CCAP11/32C,

obtained from the Culture Collection of Algae and Protozoa, UK. The strain was

maintained under batch mixotrophic conditions in Tris-Acetate-Phosphate (TAP) medium

(Harris, 1989) at a temperature of 25°C. Prior to lab-scale experimentation, an initial algal

inoculum was propagated in 150 mL of TAP medium up to the late exponential phase (5-

7 days). This inoculum was placed in an orbital shaker at 150 rpm and constant illumination

of 125 µmol m-2s-1 (from above) in a light/dark cycle of 16/8 hours. All further lab-scale

experimental tests were carried out at the same environmental growth conditions in vessels

containing 500 mL of sterile culture medium and 1 mL of algal inoculum.

2.2. Lab-scale culture experiments.

In order to evaluate nitrogen and carbon effects on microalgal growth as well as starch and

lipid accumulation, lab-scale cultures of C. reinhardtii were grown under different nitrogen

and acetate concentrations. Given that TAP medium is the most routinely used growth

medium for this strain and allows results to be compared with other studies, a “control”

culture was grown in TAP medium, in which No = 0.3824 gN L-1 and Ao = 0.42 gC L-1.



81

Subsequently, and whilst keeping constant the concentration of all the remaining TAP

components, acetate-dependent cultures were grown in: A0 = 0.21 gC L-1 (A-), 0.75 gC L-

1 (A+), 1.26 gC L-1 (A++), and 2.52 gC L-1 (High A). Similarly, nitrogen-dependent

cultures were grown in: N0 = 0.3350 gN L-1 (N--), 0.3568 gN L-1 (N-), and 0.7430 gN L-1

(High N). Two additional cultures were grown by simultaneously changing the initial

concentrations of nitrogen and acetate: one in 2.52 gC L-1 and 0.7430 gN L-1 (High A-N),

and another in 1.16 gC L-1 and 0.3151 gN L-1 (A’-N’). Samples were taken daily during the

cultivation period, until cells attained the stationary phase after 8 days. Sufficient identical

culture vessels were prepared to allow for duplicate samples to be fully harvested

(sacrificed) at each sampling time. For the nitrogen-dependent cultures, media was

prepared by modifying exclusively the initial concentration of ammonium chloride

(NH4Cl) in the TAP medium, which contains two other nitrogen sources: i) Tris-base

buffer (H2NC(CH2OH)3), and ii) ammonium molybdate tetrahydrate, a smaller trace

element ((NH4)6Mo7O24·4H2O). The total concentration of these two nitrogen sources

amounted to 0.2844 gN L-1 and was kept constant in all experiments. For acetate-dependent

cultures, media was prepared simply by increasing or decreasing accordingly the volume

of acetic acid. When necessary, the pH was adjusted to a starting value of 7, using

potassium hydroxide (KOH) 3M or hydrochloric acid (HCl) 3M. Experimental data was

statistically analysed by the two-way ANOVA test in GraphPad Prism 7 (version 7.02).

2.3. Analytical Methods.

2.3.1. Cell growth.

Microalgal growth was measured in terms of dry cell weight (DCW), quantified by

harvesting all 500 mL cultures for 3.5 min at 3,000 g in an Eppendorf Centrifuge 5424.

The residual pellet was separated from the supernatant and allowed to dry for 24 hr at

70°C. The DCW was then measured gravimetrically in an analytical balance (Sartorius M-

Pact AX124, Germany). Samples of the supernatant were stored in 50 mL Falcon tubes

and frozen at -20°C for further analysis of the acetate and nitrogen concentrations.

2.3.2. Starch and lipid quantification.

The starch content of cells was quantified according to a Total Starch Assay kit (Megazyme

International, Ireland). Briefly, this assay consists of a high temperature, two-stage (α-



82

amylase and β-amyloglucosidase) enzymatic hydrolysis which solubilises starch and

releases free D-glucose. The concentration of free D-glucose was determined

colourimetrically by measuring sample absorbance values at 508 nm against a D-glucose

standard curve. Total starch concentration was then calculated by multiplying D-glucose

concentration by 0.9 (162/180, a factor adjusting free D-glucose to anhydrous D-glucose).

Quantification of the lipid content was determined by solvent extraction in a SOXTEC

Unit 1043 over a triple-stage procedure involving: extraction, rinsing, and solvent recovery

(Bekirogullari et al., 2017). Hexane (ACS spectrophotometric grade, ≥ 98.5 %, Sigma

Aldrich, UK) was used as extracting solvent since it has shown to perform well as an

extracting agent of neutral lipids (TAGs) induced under nitrogen-deprived conditions

(McNichol et al., 2012). Prior to extraction, dried cell pellets were pulverised by a double-

cycle of liquid nitrogen immersion and manual grinding with mortar and pestle. Pulverized

cells were then placed in cellulose extraction thimbles (26 x 60 mm, thickness 1.5 mm,

Whatman, UK) and positioned in the SOXTEC unit. Extracted lipids were then measured

gravimetrically. Starch and lipid concentration is reported in volumetric terms (g L-1),

calculated by relating the storage content (%) of each sample with the corresponding total

DCW medium concentration.

2.3.3. Acetate concentration.

The residual acetate concentration was measured by High Pressure Liquid

Chromatography (HPLC) in a Hi-Plex 8 µm 300x7.7mm column using sulphuric acid

(H2SO4) 5 mM as mobile phase at a flow rate of 0.6 mL min-1 and a temperature of 50 °C.

Acetate was identified by a UV detector at a wavelength of 210 nm. Prior to analysis, all

supernatant and calibration samples were filtered in 0.45 µm nitrocellulose membranes

(Millipore Ltd.) and diluted appropriately in HPLC grade water.

2.3.4. Total nitrogen and nitrogen quota.

The residual concentration of total nitrogen was measured in a Total Organic Carbon/Total

Nitrogen measuring unit (TOC-VCSH/TNM-1 Shimadzu). Prior to analysis, calibration

standards were prepared with ammonium chloride as the sole nitrogen source. All

experimental samples were diluted appropriately in distilled water. The nitrogen quota, qN,

at each sampling time point was calculated according to Eq. (, which is equivalent to the



83

one employed by Bougaran, Bernard, & Sciandra (2010) to quantify phosphorus cell

quotas:

𝑞𝑁 =𝑁𝑜 − 𝑁

𝑋 (1)

Here No is the initial concentration of total nitrogen in the medium, and N and X are the

residual concentrations of total nitrogen and biomass (DCW), respectively.

2.3.5. Active biomass and carbon equivalent concentrations.

The active biomass (or starch- and lipid-free biomass) concentration was determined by

subtracting the concentration of storage molecules from the total biomass (DCW)

concentration. The elemental composition of the active biomass fraction was assumed to

be constant in all experiments regardless of the nutrient regime in which the cultures were

grown, taken as CH1.75O0.56N0.08, reported for C. reinhardtii by Eriksen et al. (2007). In all

computations, the concentration of each carbon-based compound was expressed in terms

of their specific carbon content, for which the following conversion factors were employed

(gC g-1): 0.444 for starch, 0.77 for lipids (C55H98O6), 0.40 for acetate, and 0.504 for the

active biomass fraction.

3. Model construction.

A multi-parameter kinetic model was developed to predict C. reinhardtii growth and

formation of starch and lipid under mixotrophic conditions. The model includes 8 state

variables: total biomass (X, gC L-1), total nitrogen (N, gN L1), nitrogen quota (qN, gN gC-

1), acetate (A, gC L-1), starch (S, gC L-1), lipids (L, gC L-1), active biomass (x*, gC L-1),

and pH (H). Total biomass is equivalent to the sum of the two major carbon-based

compartments: the storage pool made up of starch and lipids, and the active biomass.

Microalgal growth and the formation of each cellular component are regulated by the flows

shown in Figure 1.

Carbon flows were based on those presented by Mairet et al. (2011) for the microalgal

strain Isochrysis aff. Galbana, in which the carbon source was assumed to be directed

initially towards sugar synthesis. In the present model carbon assimilation is initially

directed towards the formation of active biomass so as to follow more closely the central



84

carbon metabolism of C. reinhardtii. A detailed diagrammatic representation of this

metabolism is provided by Johnson and Alric (2013), which shows that assimilation of

acetate is not only used for starch formation, but also for other important functions such as

cellular respiration, flagellar motion, and formation of acetyl-CoA, a precursor of

numerous biochemical reactions.

Figure 1. Schematic representation of the cellular compartments and flows used in

the kinetic model. X, total biomass; μ, specific growth rate; ρN, nitrogen uptake

rate; R1, starch synthetic rate; R3, lipid synthetic rate; R2, starch degradation rate;

R4, lipid degradation rate.

The cellular flows for carbon assimilation as well as for nitrogen uptake, as depicted in

Figure, are regulated by six governing equations: the specific growth rate, μ, the nitrogen

uptake rate, ρN, and the intracellular reaction rates R1, R2, R3, and R4. These equations are

described below. All definitions and corresponding units of the kinetic parameters used in

the model are listed in Table 1.

3.1. Specific growth rate, μ

One of the most widespread equations for microalgal growth is the Droop model (Eq.(2)),

where the growth rate, μ, is linked to the internal quota, q, of a limiting nutrient (qN, for

nitrogen-limited growth) rather than to its external concentration.



85

𝜇 = ��𝑚𝑎𝑥 ∙ (1 −𝑞𝑁,0

𝑞𝑁) (2)

In Eq. (2), ��𝑚𝑎𝑥 is the hypothetical maximum growth rate and qN,0 is the minimum nitrogen

quota required for growth (Droop, 1968). This simple yet effective model has been used

successfully to predict microalgae growth with additional terms accounting for multiple-

nutrient limitation or the self-shading effects observed at high cell densities (Adesanya et

al., 2014; Bernard, 2011; Bougaran et al., 2010; Mairet et al., 2011b; Packer et al., 2011).

Microalgal cells can be autotrophic, heterotrophic, or mixotrophic, but most kinetic models

describe solely autotrophic or heterotrophic growth. Adesanya et al. (2014) described the

kinetics of a mixotrophically growing culture by expressing the hypothetical maximum

growth rate as the sum of the autotrophic and heterotrophic growth rates. This approach

was adapted into the present model with the inclusion of weighting functions controlling

the extent of each rate on overall growth. The specific growth rate was thus expressed as:

𝜇 = ��𝑀,𝑚𝑎𝑥(𝐴, 𝐼) ∙ (1 −𝑞𝑁,0

𝑞𝑁) (3)

Here ��𝑀,𝑚𝑎𝑥(𝐴, 𝐼) is the maximum specific growth rate under mixotrophic conditions, as

shown in Eq. (4), and is proportional to the sum of the heterotrophic and phototrophic rates,

μH (A) and μI (I), respectively.

��𝑀,𝑚𝑎𝑥(𝐴, 𝐼) = 𝜇𝑚𝑎𝑥 ∙ [𝑤𝐻 ∙ 𝜇𝐻(𝐴) + 𝑤𝐼 ∙ 𝜇𝐼(𝐼)] (4)

In order to account for photoinhibition and substrate inhibition, μH (A) and μI (I) were

expressed as Andrews functions (Andrews, 1968) as shown in Eq. (5):

��𝑀,𝑚𝑎𝑥(𝐴, 𝐼) = 𝜇𝑚𝑎𝑥 ∙ [𝑤𝐻 ∙𝐴

𝐴 + 𝐾𝑠,𝐴 + 𝐴2

𝑘𝑖,𝐴⁄

+ 𝑤𝐼 ∙𝐼

𝐼 + 𝐾𝑠,𝐼 + 𝐼2

𝑘𝑖,𝐼⁄

] (5)

Here, 𝐾𝑠,𝐴 and 𝑘𝑖,𝐴 are the acetate-associated saturation and inhibition constants for

growth, whereas 𝐾𝑠,𝐼 and 𝑘𝑖,𝐼 are the light-associated saturation and inhibition constants,

respectively. The weighting functions, wH and wI, shown in Eq. (6), were defined in terms

of the saturation constants in a similar fashion to those presented in Shuler and Kargi

(1992):



86

𝑤𝐻 =𝐴 𝐾𝑠,𝐴⁄

𝐴 𝐾𝑠,𝐴⁄ +𝐼 𝐾𝑠,𝐼⁄ ; 𝑤𝐼 =

𝐼 𝐾𝑠,𝐼⁄

𝐴 𝐾𝑠,𝐴⁄ +𝐼 𝐾𝑠,𝐼⁄ (6)

Light distribution (I) throughout the culture vessel was represented by the Beer-Lambert

law shown in Eq. (7), where I0 (µmol m-2s-1) is the incident light intensity, σ is the light

attenuation coefficient, and z (m) is the culture depth within the vessel.

𝐼 = 𝐼0 ∙ 𝑒−𝜎∙𝑋∙𝑧 (7)

3.2. Nitrogen uptake rate, ρN

The expression for nitrogen uptake rate is a crucial element of the model, since nitrogen

entering the cells is directly linked to the nitrogen quota, which regulates cell growth (Eq.

(3)). The uptake rate of nitrogen, shown in Eq. (8), was expressed as Andrews-type kinetics

to account for the growth inhibition of C. reinhardtii observed at high external nitrogen

concentrations. Since analogous observations were made in the cultures subject to high

acetate treatments, acetate inhibition was similarly considered.

𝜌𝑁 = ��𝑁,𝑚𝑎𝑥(𝑁𝑜, 𝑋) ∙𝑁

𝑁 + 𝐾𝑠,𝑁 + 𝑁2

𝑘𝑖,𝑁⁄

∙𝐴

𝐴 + 𝐾𝑠,𝐴:𝑁 + 𝐴2

𝑘𝑖,𝐴:𝑁⁄

(8)

Here, 𝐾𝑠,𝑁 and 𝑘𝑖,𝑁 are the nitrogen-associated saturation and inhibition constants for

nitrogen uptake. Similarly, 𝐾𝑠,𝐴:𝑁 and 𝑘𝑖,𝐴:𝑁 are the acetate-associated saturation and

inhibition constants for nitrogen uptake. The maximum nitrogen uptake rate,

��𝑁,𝑚𝑎𝑥(𝑁𝑜, 𝑋), depends on the initial nitrogen concentration (N0) under which cultures

were grown and on the current biomass concentration (X):

��𝑁,𝑚𝑎𝑥(𝑁0, 𝑋) = 𝜌𝑁,𝑚𝑎𝑥 ∙𝑁𝑜

𝑛

𝑁𝑜𝑛 + 𝐾∗

𝑛 ∙ 𝑒−𝜙𝑁∙𝑋 (9)

The maximum nitrogen uptake rate was built under the concept of “luxury consumption”,

used to describe the abrupt uptake of a nutrient from the cultivation medium (Droop, 1983).

K* is a saturation constant, n is a shape-controlling parameter, and 𝜙 is a regulation

coefficient. A detailed explanation is included in section 4.2.



87

3.3. Rates of formation of cellular compartments (R1, R2, R3, R4).

The synthesis rates for starch and lipids, R1 and R3, respectively, were assumed to be

controlled by the specific growth rate and the active biomass, as shown in Eq. (10) and Eq.

(11). Both sets of reactions include: i) an Andrews-type term dependent on the internal

nitrogen concentration, Ni = qN·X, and ii) an exponential term dependent on the internal

carbon concentration, Aint = Ao-A. The exponential term accounts for the higher formation

of storage molecules in the cultures grown in high acetate media, which was observed to

take place even after biomass had reached stationary phase (Figure 3). This increase was

thus assumed to be uncoupled from cellular growth and only a consequence of excess in

the internal carbon pool.

𝑅1 = 𝑟1 ∙𝑁𝑖

𝑛𝑠

𝑁𝑖𝑛𝑠 + 𝐾𝑠,𝑆

𝑛𝑠 + (𝑁𝑖2 𝑘𝑖,𝑆⁄ )

𝑛𝑠∙

𝑘1

𝑘1 + 𝑁 𝑁𝑜⁄∙ [1 +

1

𝜇∙ 𝑒𝜙𝑆∙𝐴𝑖𝑛𝑡] ∙ 𝜇 ∙ 𝑥∗ (10)

𝑅3 = 𝑟3 ∙𝑁𝑖

𝑛𝐿

𝑁𝑖𝑛𝐿 + 𝐾𝑠,𝐿

𝑛𝐿 + (𝑁𝑖2 𝑘𝑖,𝐿⁄ )

𝑛𝐿∙

𝑘2

𝑘2 + 𝑁 𝑁0⁄∙ [1 +

1

𝜇∙ 𝑒𝜙𝐿∙𝐴𝑖𝑛𝑡] ∙ 𝜇 ∙ 𝑥∗ (11)

Here, Ks,S and Ks,L are saturation constants, ki,S and ki,L are inhibition constants, ns and nL

are shape-controlling exponents, ΦS and ΦL are regulation coefficients, and k1 and k2

regulate synthesis rates with respect to nitrogen consumption. Active biomass formation

was also linked with starch and lipid degradation (R2 and R4). These rates, shown in Eq.

(12), were defined as functions of the nitrogen quota since cellular components such as

proteins or nucleic acids, depend on nitrogen availability.

𝑅2 =𝑟2

𝑞𝑁∙ 𝑋; 𝑅4 =

𝑟4

𝑞𝑁∙ 𝑋 (12)

3.4. Time-dependent kinetic expressions.

The dynamics for active biomass (x*), starch (S), lipids (L), and total biomass (X), were

obtained from the corresponding mass conservation equations as follows:

𝑑𝑥∗

𝑑𝑡= 𝜇 ∙ 𝑋 + 𝑅2 + 𝑅4 − (𝑅1 + 𝑅3) (13)



88

𝑑𝑆

𝑑𝑡= 𝑅1 − 𝑅2 (14)

𝑑𝐿

𝑑𝑡= 𝑅3 − 𝑅4 (15)

Total biomass (i.e. X = x* + S + L) conservation simplifies thus to:

𝑑𝑋

𝑑𝑡= 𝜇 ∙ 𝑋 (16)

Acetate consumption was expressed by means of the acetate to biomass yield coefficient,

YX/A, multiplied by a time-varying fraction accounting for the carbon used

heterotrophically:

𝑑𝐴

𝑑𝑡= −

1

𝑌𝑋/𝐴∙

𝜇𝐻(𝐴)

𝜇𝐻(𝐴) + 𝜇𝐼(𝐼)∙

𝑑𝑋

𝑑𝑡 (17)

By considering that the removal of H+ ions from the medium is a direct consequence of

microalgal growth (e.g. acetate consumption), pH evolution was represented as:

𝑑𝐻

𝑑𝑡= 𝐾𝐻 ∙

𝑑𝑥∗

𝑑𝑡 (18)

Here, 𝐾𝐻 is a pH coefficient. The rate of nitrogen consumption was expressed as:

𝑑𝑁

𝑑𝑡= −𝜌𝑁 ∙ 𝑋 (19)

Differentiation of the nitrogen quota with respect to time yields:

𝑑𝑞𝑁

𝑑𝑡=

𝑑(𝑁𝑖 𝑋⁄ )

𝑑𝑡=

𝑑𝑁𝑖𝑑𝑡

∙𝑋−𝑁𝑖∙𝑑𝑋𝑑𝑡

𝑋2=

1

𝑋∙

𝑑𝑁𝑖

𝑑𝑡−

𝑁𝑖

𝑋∙ (

1

𝑥∙

𝑑𝑋

𝑑𝑡) (20)

where Ni is the internal nitrogen concentration, and its accumulation rate is given by:

𝑑𝑁𝑖

𝑑𝑡= −

𝑑𝑁

𝑑𝑡= 𝜌𝑁 ∙ 𝑋 (21)



89

By substituting Eq. (16) and Eq. (21) in Eq. (20), the time-dependent equation for the

nitrogen quota simplifies thus to:

𝑑𝑞𝑁

𝑑𝑡= 𝜌𝑁 − 𝜇 ∙ 𝑞𝑁 (22)

3.5. Parameter estimation.

The proposed model, given by Eq. (13) - Eq. (19) and Eq. (22), consists of 8 state variables

and 31 kinetic parameters (Table 1). Sensitivity analysis was carried out by estimating

sensitivities (gradients of each state variable with respect to each of the parameters)

numerically using central finite differences for a 10% change in each parameter. The

results can be found in the Supplementary material. We noticed that sensitivities above

value of 0.02 denoted that the corresponding variable was sensitive to changes in the

parameter. Through this sensitivity analysis, 4 parameters were deemed insensitive, Ϭ, Ks,I,

Ks,S and ФL. Ks,S and ФL were neglected from the final model as it was noticed that setting

them to zero did not affect results. Ks,I was set to 1.4 as in the literature (Mairet et al.,

2011b) and Ϭ was set equal to 1. Hence, Eq. (10) and Eq. (11) become:

𝑅1 = 𝑟1 ∙𝑁𝑖

𝑛𝑠

𝑁𝑖𝑛𝑠 + (𝑁𝑖

2 𝑘𝑖,𝑆⁄ )𝑛𝑠

∙𝑘1

𝑘1 + 𝑁 𝑁𝑜⁄∙ [1 +

1

𝜇∙ 𝑒𝜙𝑆∙𝐴𝑖𝑛𝑡] ∙ 𝜇 ∙ 𝑥∗ (23)

𝑅3 = 𝑟3 ∙𝑁𝑖

𝑛𝐿



𝑛𝐿∙

𝑘2

𝑘2 + 𝑁 𝑁0⁄∙ [𝜇 + 1] ∙ 𝑥∗ (24)

Estimation of the remaining 27 kinetic parameters was carried out by minimizing an

objective function defined as the sum of the squared relative error between the model

predictions and the experimental data, as shown in (Vlysidis et al., 2011):

min 𝐺(𝑃) = ∑ ∑ ∑ (𝑍ℎ𝑖𝑘

𝑃𝑟𝑒𝑑(𝑃) − 𝑍ℎ𝑖𝑘𝐸𝑥𝑝

𝑍ℎ𝑖𝑘𝐸𝑥𝑝 )

2𝑛𝑘

𝑘=1

𝑛𝑖

𝑖=1

𝑛ℎ

ℎ=1

(25)

where G(P) is the objective function dependent on a vector P containing all kinetic

parameters and Z is a vector containing all state variables, nk is the number of experimental

datasets used for parameter fitting, ni is the number of state variables (ni = 8), and nh is

the number of data points in time (nh = 7). Minimization of the objective function was

performed by first employing Simulated Annealing (SA), a stochastic optimisation



90

algorithm which is capable of approximating the solution set around a global minimum.

Then, using the solution obtained by SA as initial guess, a refined and final solution set of

parameters were computed by using Successive Quadratic Programming (SQP) (Vlysidis

et al., 2011). Both techniques were coded in-house and implemented in MatLab®.

The value of each kinetic parameter was restricted to specified bounds according to data

found in literature (for those cases in which available data existed) or relevant experimental

analysis. Initial values for the model ODEs were equal to those implemented in each of the

five experimental datasets used for parameter fitting (nk = 5): TAP, N--, A++, High N, and

High A-N. These datasets were selected so as to cover scenarios representative of the

cultivation stage under both low and high concentrations of nitrogen and acetate. The

remaining datasets (N-, A+, and A’-N’) were used for model validation.



91

Table 1. Kinetic parameters used in our proposed model.

Parameter description Value Units Strain Reference

µmax Maximum specific growth

rate

0.106

0.227

0.084

h-1

C. reinhardtii

C. reinhardtii

C. reinhardtii

This work

Fouchard et al., 2009

Chen and Johns, 1994

qN,0 Minimum nitrogen quota

0.876

0.25

0.0975

gN gC-1

gN gC-1

gN gDW-1

C. reinhardtii

I. galbana

C. vulgaris

This work

Bernard, 2011

Adesanya et al., 2014

Ks,A Saturation constant, A

1.789

0.014a

1.04a

gC L-1

gC L-1

gC L-1

C. reinhardti

C. reinhardti

C. reinhardti

This work


Zhang et al., 1999

ki,A Inhibition constant, A

0.109

0.708a

0.042a

gC L-1

gC L-1

gC L-1

C. reinhardti

C. reinhardti

C. reinhardti

This work


Zhang et al., 1999

Ks,I Saturation constant, I 1.4 µmol m-2s-1 I. galbana Mairet et al., 2011a

ki,I Inhibition constant, I 186.52

295.00 µmol m-2s-1

C. reinhardti

I. galbana

This work

Mairet et al., 2011a

YX/A Yield coefficient

0.059

0.84

0.50

gC gC-1

gC gC-1

g g-1

C. reinhardti

C. sorokiniana

C. reinhardtii

This work

Turon et al., 2014


Ϭ Light attenuation coefficient 1 L gC-1 m-1 C. reinhardtii This work

ρN,max Maximum N uptake rate 40.445 gN gC-1 h-1 C. reinhardtii This work

K* Saturation constant, No 0.3125 gN L-1 C. reinhardtii This work

n Shape parameter 18.183 - C. reinhardtii This work

ФN Uptake regulation coefficient 137.455 L gC-1 C. reinhardtii This work

Ks,N Uptake saturation constant, N 0.162 gN L-1 C. reinhardtii This work

ki,N Uptake inhibition constant, N 0.113 gN L-1 C. reinhardtii This work

Ks,A:N Uptake saturation constant, A 1.004 gC L-1 C. reinhardtii This work

ki,A:N Uptake inhibition constant, A 1.098 gC L-1 C. reinhardtii This work

r1 Rate of reaction, R1 0.0420 gC gC-1 C. reinhardtii This work

r2 Rate of reaction, R2 0.1620 gN gC-1h-1 C. reinhardtii This work

r3 Rate of reaction, R3 0.0041 gC gC-1 C. reinhardtii This work

r4 Rate of reaction, R4 0.0049 gN gC-1h-1 C. reinhardtii This work

Ks,S Saturation constant for R1 0 gN L-1 C. reinhardtii This work

ki,S Inhibition constant for R1 0.2079 gN L-1 C. reinhardtii This work

nS Shape parameter for R1 3.6205 - C. reinhardtii This work

k1 Regulation constant for R1 0.1771 - C. reinhardtii This work

ФS Regulation coefficient for R1 0.6675 L gC-1 C. reinhardtii This work

Ks,L Saturation constant for R3 0.0227 gN L-1 C. reinhardtii This work

ki,L Inhibition constant for R3 0.0861 gN L-1 C. reinhardtii This work

nL Shape parameter for R3 1.8117 - C. reinhardtii This work

k2 Regulation constant for R3 0.2135 - C. reinhardtii This work

ФL Regulation coefficient for R3 0 L gC-1 C. reinhardtii This work

KH pH coefficient 4.653 L gC-1 h-1 C. reinhardtii This work a Reported values have been converted to gC L-1



92

4. Results and discussion.

4.1. Effect of nitrogen and acetate in biomass, starch, and lipid formation.

All of the cultures analysed under the conditions established in section 2.2 reached early

stationary phase after 150 h, but cultures were allowed to grow for a further period of 48 h

to ensure they had all reached stationary phase and were accumulating carbon storage

products. Experimental results for biomass growth as well as for starch and lipid formation

are shown in Figure 2, which are representative of the cultures during the stationary stage

(192 h).

Figure 2. Biomass production and corresponding distribution of carbon

compartments at t=192h (8th day of cultivation) for: a) b) N-dependent cultures

(starting Ao = 0.42 gC L-1), c) d) A–dependent cultures (starting No = 0.3824 gN L-1),

and e) d) a high A-N culture. Treatments that do not share uppercase letters are

significantly different (p < 0.05), as determined by two-way ANOVA.



93

Results showed (Figure 2.b) that the cellular contents of starch and lipids increased

significantly (p < 0.0001, two-way ANOVA) as the initial nitrogen concentration in the

culture medium was reduced from 0.3824 gN L-1 (TAP), to both 0.3568 gN L-1 and 0.335

gN L-1. Specifically, starch concentration increased from 6% (at No=0.3824 gN L-1) to 17%

(at No=0.3350 gN L-1), whereas lipid increased from 14% to 21%, respectively. This

enhanced accumulation observed under nitrogen limitation is in agreement with previous

analysis of C. reinhardtii (Bajhaiya et al., 2016) and with findings reported for other

microalgae strains, such as Chlorella vulgaris P12 (Brányiková et al., 2010) or Tetraselmis

subcordiformis (Yao et al., 2012). However, this increase is at the expense of biomass

growth (Figure 2.a), which was observed to decrease significantly under nitrogen-limited

conditions (p = 0.0006 between 0.3824 gN L-1 and 0.335 gN L-1, two-way ANOVA). The

magnitude of this negative trade-off in biomass growth ultimately controls starch and lipid

formation in terms of volumetric yields, and should be considered in any nutrient-based

cultivation strategy.

As per the ANOVA test, increases in starch and lipid contents in the culture grown at a

high nitrogen concentration (No=0.7426 gN L-1) were not statistically significant with

respect to the culture grown under standard TAP concentrations. However, biomass

concentration decreased significantly (p < 0.0001, two-way ANOVA), indicating that a

high nitrogen concentration inhibited biomass growth. Nitrogen has been widely reported

as a limiting nutrient suitable for increased accumulation of lipid (Cakmak et al., 2012;

Rodolfi et al., 2009; Xin et al., 2010) and carbohydrate (Behrens et al., 1989; Dragone et

al., 2011). Nitrogen is a vital component of important biomolecules like proteins and DNA,

and it is estimated to represent 7-20% of the cellular mass. When cells are exposed to a

nitrogen depleted environment, the protein synthesis pathway is negatively affected, which

results in the carbon fixation mechanism being instead re-directed towards the production

of carbohydrates or lipids (Juneja et al., 2013; Markou et al., 2012).

In addition, experimental results (Figure 2.c) showed that when compared to the culture

grown in TAP (Ao=0.42 gC L-1), an increase in the initial acetate concentration had a

significant effect on C. reinhardtii growth (p < 0.0001 for Ao=0.75 gC L-1 and Ao=1.26 gC

L-1; p = 0.0024 for Ao=0.21 gC L-1). Specifically, the biomass concentration (192 h) rose

from 0.25 gC L-1 to 0.41 gC L-1 as the initial acetate concentration increased from Ao=0.21



94

gC L-1 to Ao=1.26 gC L-1. The presence of an additional organic carbon source (such as

acetate) has been shown to: i) boost microalgal biomass growth (Chapman et al., 2015),

and ii) increase starch and lipid accumulation, caused possibly by either the greater cell

sizes of acetate-enhanced cultures (Goodson et al., 2011) or the larger availability of the

carbon pool which shifts or lengthens the biosynthetic pathways (Fan et al., 2012;

Goodenough et al., 2014). Although the cellular contents of the storage molecules

increased slightly as a result of acetate addition (Figure 2.d), the extent of this

accumulation was less noticeable (p > 0.05, between 0.42 gC L-1 and all carbon treatments)

than the nitrogen-driven accumulation.

Similar to nitrogen-limited growth observations, biomass concentration decreased at a high

acetate concentration of Ao=2.52 gC L-1. The combined inhibitory effects posed by high

nitrogen and acetate concentrations were further verified experimentally in the HIGH A-

N culture (Ao=2.52 gC L-1 and No=0.7430 gN L-1), which attained a biomass concentration

of 0.21 gC L-1 (Figure 2.e). Thus, the expected increase in costs for such a high-nutrient

strategy, coupled with the growth inhibition, undermines its potential use for C. reinhardtii

cultivation.

4.2. Predictive performance of the kinetic model.

The microalgae-based model developed in this work consists of 8 ODEs and 31 kinetic

parameters. The estimated values of each kinetic parameter computed by the methodology

described in Section 3.5 are presented in Table 1, which also provides reference values

available in the open literature. The resulting concentration profiles of each state variable,

as predicted by the model, are shown in Figure 3 against their corresponding experimental

values. The model was capable of predicting accurately all 8 state variables, as shown by

the good agreement obtained between predicted and experimental values in both datasets

that were used in the fitting process (e.g. TAP and N-) and datasets obtained at different

conditions (e.g. A++ and N’-A’) for validation. Additional details are available as

Supplementary Information.



95

Figure 3. Comparison between the predicted time-profile (lines) and experimental

data (points) for the cultures grown in: TAP (Ao=0.42 gC L-1, No=0.3824 gN L-1), N-

(Ao=0.42 gC L-1, No=0.356 gN L-1), A++ (Ao=1.26 gC L-1, No=0.3824 gN L-1), and N’-

A’ (Ao=1.16 gC L-1, No=0.3151 gN L-1). Fitting datasets: TAP and A++; Validating

datasets: N- and N’-A’. Data and standard deviation are the mean of 2 experimental

replicates.

The model was able to compute accurate dynamic concentration profiles for all species

involved under different nitrogen and carbon concentration regimes (Figure 3), including

total biomass, X, indicating that the Droop-based expression used for the specific growth

rate (Eq. (3)) can adequately describe microalgal growth dynamics under nitrogen-limited

mixotrophic conditions. Although both the classic Monod and Droop’s model have been

widely used to model microalgal growth, an added advantage of the Droop’s model is its

dependence on internal nutrient availability, which allows to capture the observed ability

of microalgae to grow even after complete exhaustion of a limiting nutrient (Lee et al.,

2015).

Small disagreements between predictions and experimental data can be seen for nitrogen

and pH dynamics (Figure 3.b, Figure 3.h). The variation in pH predictions might be the

result of using a rather simple expression (Eq.()) that does not take into account the



96

formation of other organic acids produced in small quantities by C. reinhardtii, such as

formic acid or glycolic acid (Bekirogullari et al., 2017). Another potential cause for these

disagreements could be related to the presence of tris-base in the cultivation medium. Tris-

base acts as a biochemical buffer, but its concentration profile is not predicted by the

kinetic model. Instead, tris-base is only implicitly included within nitrogen dynamics due

to its high contribution to total nitrogen concentration: almost 70% of the total nitrogen

present in standard TAP medium originates from tris-base. The latter might also explain

the slight discrepancies observed between the predicted and experimentally obtained

dynamics of nitrogen uptake. However, rather than incorporating individual uptake

expressions for each nitrogen source, which would increase complexity and computational

time, the kinetic model was built with one single expression for nitrogen uptake (Eq. (8)).

Models in which microalgal growth is limited by an internal nutrient pool generally assume

Michaelis-Menten (MM) uptake kinetics, as in Droop’s original approach (Droop, 1983).

Under this assumption, nutrient uptake is assumed to be dependent on a single enzyme

system that controls the uptake of extracellular substrates (Shuler and Kargi, 1992). In the

current model, however, Andrews-type kinetics were employed to effectively predict C.

reinhardtii’s growth inhibition at high nitrogen concentrations. Inhibited uptake dynamics

then cause less nitrogen to enter the cells, which translates into smaller nitrogen quota and,

consequently, decreased growth.

Although MM-type kinetics have been successfully implemented in microalgae-oriented

models (Adesanya et al., 2014; Mairet et al., 2011b; Packer et al., 2011), it was suggested

(Bonachela et al., 2011) that the use of this rather static model is not capable of capturing

the ability of microalgae to adapt their “uptake machinery” to a changing environment. It

follows from the same logic that inhibited-kinetics (as employed in this model) might

suffer from the same weakness. This flaw is potentially a result of treating the maximum

uptake rate, ρN,Max, as a constant rather than as a dynamic variable (Morel, 1987),

preventing an organism’s uptake kinetics to respond to environmental changes (Bonachela

et al., 2011). In microalgae, an abrupt increase in nutrient availability might lead to the

phenomenon of luxury consumption, which refers to the sudden drop (uptake) of a nutrient

from the surrounding medium (Droop, 1983). This phenomenon was observed in all our

lab-scale experiments, where nitrogen concentration decreased rapidly in the first 48 hours



97

following inoculation (Figure 3.b). Since the degree of luxury consumption was expected

to be dependent on the current cell density and the nutrient concentration of the fresh

medium (Droop, 1983), the maximum uptake rate (Eq. (9)) was expressed as a decreasing

function of biomass (Zhang et al., 2008) coupled with a Monod-like function of the initial

nitrogen concentration.

Moreover, the model was able to predict the higher nitrogen consumption yields observed

in those cultures grown under high acetate concentrations. From the datasets shown in

Figure 3.b, for example, it was estimated that the culture grown in Ao=1.26 gC L-1 (A++)

consumed about 93% of the total nitrogen supplied, whereas the culture grown in Ao=0.42

gC L-1 (TAP) consumed 77% (both cultures grown in No=0.3824 gN L-1). This is because

cells require a large supply of nitrogen to compensate for acetate-enhanced growth rates,

as observed in this work and that of Chapman et al. (2015), where C. reinhardtii cells

showed a higher growth rate under mixotrophic rather than phototrophic conditions.

As observed in Figure 3.e, Figure 3.f, the proposed model was also able to predict

adequately the simultaneous concentration profile of starch and lipids under a wide range

of initial nitrogen and acetate concentrations. The high predictive behaviour shown by the

model proposed in this study, particularly for starch and lipid formation during nitrogen-

limited mixotrophic growth conditions, confirms its potential as a robust tool in the

development of optimal nutrient-based microalgal cultivation strategies. An optimisation

study was thus undertaken and is presented next.

4.3. Optimal nutrient-based strategies for starch and lipid formation.

In order to profit from the accurate predictions obtained, the model was subsequently used

to establish the optimal initial conditions to attain maximum concentrations of the two

valuable biofuel feedstocks: starch and lipids. This procedure was carried out by

identifying the maximum in a contour plot of each variable, as computed by the validated

model, at a time equivalent to the point of highest storage molecule formation (t=192 h).

The resulting contour plots can be seen in Figure 4.

The maxima of the contour plots shown in Figure 4.b and Figure 4.c allow to identify

optimal nitrogen and acetate concentrations that maximise lipid (lipid-enhanced scenario)

or starch (starch-enhanced scenario). These optimal sets are: i) OPTStarch = (Ao=1.06 gC L-



98

1 and No=0.336 gN L-1) for the starch-enhanced scenario, producing a starch concentration

of 0.065 gC L-1 with corresponding lipid concentration of 0.069 gC L-1; and ii) OPTLipids =

(Ao=1.15 gC L-1 and No=0.378 gN L-1) for the lipid-enhanced scenario producing a lipid

concentrations of 0.08 gC L-1 with corresponding starch concentration of 0.043 gC L-1,

respectively.

Figure 4. Contour plots generated from model predictions for: a) biomass, b)

starch, and c) lipid formation (at t=190h) during C. reinhardtii cultivation.

When compared to the base case, TAP = (Ao=0.42 gC L-1 and No=0.3824 gN L-1), which

predicts a starch concentration of 0.018 gC L-1 and a lipid concentration of 0.048 gC L-1,

the starch-enhanced scenario accounts for a drastic increase in starch of 261% (and

corresponding 44% increase in lipids), whereas the lipid-enhanced scenario accounts for

an increase in lipids of 66% (and 139% increase in starch). In each optimised case both

starch and lipid concentrations are maximised with respect to the base case, due in part by

the acetate boost (enhanced mixotrophic conditions), which was previously shown to

increase biomass growth. Indeed, the model predicted that the highest microalgal

concentration (Figure 4.a) could be achieved at Ao = 1.1 gC L-1 and No = 0.415 gN L-1.

The starch and lipid enhanced-scenarios were further validated experimentally by growing

two additional microalgal cultures under the initial concentrations OPTStarch and OPTLipids,



99

respectively. Predicted concentration profiles and data obtained experimentally from these

enhanced scenarios are presented in Figure 5. For comparison, microalgal dynamics

obtained by standard TAP concentrations are also plotted.

Figure 5. Comparison between the predicted time-profile (lines) and experimental

data (points) for the cultures grown in: TAP (Ao=0.42 gC L-1, No=0.3824 gN L-1),

OPTStarch (Ao=1.06 gC L-1 and No=0.336 gN L-1), and OPTLipids (Ao=1.15 gC L-1 and

No=0.378 gN L-1). Data and standard deviation are the mean of 2 experimental

replicates.

As observed, the model performance proved once more its ability to capture adequately

the trade-off between starch and lipid formation under nitrogen-limited mixotrophic

growth. It could be argued that the magnitude of the predicted increases in biomass, starch,

and lipids would not justify the required increase in acetate inputs. The good predictive

performance of the model, however, allows carrying out alternate optimization strategies

in which other factors are taken into account as per cultivation requirements (e.g.

productivity, nutrient consumption yields, etc.). The outcome of such model-based

optimised scenarios will undoubtedly aid in the development of microalgae as a biofuel



100

feedstock by tackling challenges faced during the cultivation stage, such as reducing the

nutrient-associated costs whilst simultaneously increasing starch and lipid productivities.

5. Conclusions.

A multi-parameter kinetic model was developed to predict nitrogen-limited mixotrophic

microalgal growth coupled with simultaneous starch and lipid formation. All kinetic

parameters were accurately computed by minimising the squared relative error between

experimental values and model predictions. The predicted time-profiles of the model’s

state variables were then validated against additional experimental datasets obtained under

different nutrient concentration regimes. Model-based optimised cultivation strategies,

maximising starch (261 % increase with respect to base case) and lipid (66 % increase with

respect to base case) production, were subsequently computed, and further experimentally

validated.

Acknowledgements

Gonzalo M. Figueroa-Torres kindly acknowledges the Mexican National Council of

Science and Technology (CONACyT) for its financial support.



101

References.

Adesanya, V.O., Davey, M.P., Scott, S.A., Smith, A.G., 2014. Kinetic modelling of

growth and storage molecule production in microalgae under mixotrophic and

autotrophic conditions. Bioresour. Technol. 157, 293–304.

Alonso, A.A., Molina, I., Theodoropoulos, C., 2014. Modeling Bacterial Population

Growth from Stochastic Single-Cell Dynamics. Appl. Environ. Microbiol. 80,

5241–5253.

Andrews, J.F., 1968. A mathematical model for the continuous culture of

microorganisms utilizing inhibitory substrates. Biotechnol. Bioeng. 10, 707–723.

Bajhaiya, A.K., Dean, A.P., Driver, T., Trivedi, D.K., Rattray, N.J.W., Allwood, J.W.,

Goodacre, R., Pittman, J.K., 2016. High-throughput metabolic screening of

microalgae genetic variation in response to nutrient limitation. Metabolomics 12, 9.

Ball, S.G., Deschamps, P., 2009. The Chlamydomonas Sourcebook: Starch Metabolism,

The Chlamydomonas Sourcebook. Elsevier.

Behrens, P.W., Bingham, S.E., Hoeksema, S.D., Cohoon, D.L., Cox, J.C., 1989. Studies

on the incorporation of CO2 into starch by Chlorella vulgaris. J. Appl. Phycol. 1,

123–130.

Bekirogullari, M., Fragkopoulos, I.S., Pittman, J.K., Theodoropoulos, C., 2017.

Production of lipid-based fuels and chemicals from microalgae: An integrated

experimental and model-based optimization study. Algal Res. 23, 78–87.

Bernard, O., 2011. Hurdles and challenges for modelling and control of microalgae for

CO2 mitigation and biofuel production. J. Process Control 21, 1378–1389.

Bonachela, J.A., Raghib, M., Levin, S.A., 2011. Dynamic model of flexible

phytoplankton nutrient uptake. Proc. Natl. Acad. Sci. 108, 20633–20638.

Bougaran, G., Bernard, O., Sciandra, A., 2010. Modeling continuous cultures of

microalgae colimited by nitrogen and phosphorus. J. Theor. Biol. 265, 443–54.

Brányiková, I., Maršálková, B., Doucha, J., Brányik, T., Bišová, K., Zachleder, V.,



102

Vítová, M., 2010. Microalgae—novel highly efficient starch producers. Biotechnol.

Bioeng. 108, 766–776.

Brennan, L., Owende, P., 2010. Biofuels from microalgae—A review of technologies for

production, processing, and extractions of biofuels and co-products. Renew. Sustain.

Energy Rev. 14, 557–577.

Cakmak, T., Angun, P., Demiray, Y.E., Ozkan, A.D., Elibol, Z., Tekinay, T., 2012.

Differential effects of nitrogen and sulfur deprivation on growth and biodiesel

feedstock production of Chlamydomonas reinhardtii. Biotechnol. Bioeng. 109,

1947–1957.

Chapman, S.P., Paget, C.M., Johnson, G.N., Schwartz, J.-M., 2015. Flux balance analysis

reveals acetate metabolism modulates cyclic electron flow and alternative glycolytic

pathways in Chlamydomonas reinhardtii. Front. Plant Sci. 6, 474.

Chen, F., Johns, M.R., 1994. Substrate inhibition of Chlamydomonas reinhardtii by

acetate in heterotrophic culture. Process Biochem. 29, 245–252.

Choix, F.J., De-Bashan, L.E., Bashan, Y., 2012. Enhanced accumulation of starch and

total carbohydrates in alginate-immobilized Chlorella spp. induced by Azospirillum

brasilense: I. Autotrophic conditions. Enzyme Microb. Technol. 51, 294–9.

Dragone, G., Fernandes, B.D., Abreu, A.P., Vicente, A. a., Teixeira, J. a., 2011. Nutrient

limitation as a strategy for increasing starch accumulation in microalgae. Appl.

Energy 88, 3331–3335.

Droop, M.R., 1983. 25 Years of Algal Growth Kinetics A Personal View. Bot. Mar. 26,

99.

Droop, M.R., 1968. Vitamin B12 and Marine Ecology. IV. The Kinetics of Uptake,

Growth and Inhibition in Monochrysis Lutheri. J. Mar. Biol. Assoc. United

Kingdom 48, 689–733.

Eriksen, N., Riisgård, F., Gunther, W., Lønsmann Iversen, J., 2007. On-line estimation of

O2 production, CO2 uptake, and growth kinetics of microalgal cultures in a gas-

tight photobioreactor. J. Appl. Phycol. 19, 161–174.



103

Escobar, J.C., Lora, E.S., Venturini, O.J., Yáñez, E.E., Castillo, E.F., Almazan, O., 2009.

Biofuels: Environment, technology and food security. Renew. Sustain. Energy Rev.

13, 1275–1287.

Fan, J., Yan, C., Andre, C., Shanklin, J., Schwender, J., Xu, C., 2012. Oil accumulation is

controlled by carbon precursor supply for fatty acid synthesis in Chlamydomonas

reinhardtii. Plant Cell Physiol. 53, 1380–1390.

Fouchard, S., Pruvost, J., Degrenne, B., Titica, M., Legrand, J., 2009. Kinetic modeling

of light limitation and sulfur deprivation effects in the induction of hydrogen

production with Chlamydomonas reinhardtii: Part I. Model development and

parameter identification. Biotechnol. Bioeng. 102, 232–45.

Goodenough, U., Blaby, I., Casero, D., Gallaher, S.D., Goodson, C., Johnson, S., Lee, J.-

H., Merchant, S.S., Pellegrini, M., Roth, R., Rusch, J., Singh, M., Umen, J.G.,

Weiss, T.L., Wulan, T., 2014. The Path to Triacylglyceride Obesity in the sta6

Strain of Chlamydomonas reinhardtii. Eukaryot. Cell 13, 591–613.

Goodson, C., Roth, R., Wang, Z.T., Goodenough, U., 2011. Structural Correlates of

Cytoplasmic and Chloroplast Lipid Body Synthesis in Chlamydomonas reinhardtii

and Stimulation of Lipid Body Production with Acetate Boost. Eukaryot. Cell 10,

1592–1606.

Harris, E.H., 1989. The Chlamydomonas sourcebook : a comprehensive guide to biology

and laboratory use. Academic Press.

Johnson, X., Alric, J., 2013. Central Carbon Metabolism and Electron Transport in

Chlamydomonas reinhardtii: Metabolic Constraints for Carbon Partitioning between

Oil and Starch. Eukaryot. Cell 12, 776–793.

Juneja, A., Ceballos, R., Murthy, G., 2013. Effects of Environmental Factors and

Nutrient Availability on the Biochemical Composition of Algae for Biofuels

Production: A Review. Energies 6, 4607–4638.

Kumar, A., Guria, C., Chitres, G., Chakraborty, A., Pathak, A.K., 2016. Modelling of

microalgal growth and lipid production in Dunaliella tertiolecta using nitrogen-



104

phosphorus-potassium fertilizer medium in sintered disk chromatographic glass

bubble column. Bioresour. Technol. 218, 1021–1036.

Lee, E., Jalalizadeh, M., Zhang, Q., 2015. Growth kinetic models for microalgae

cultivation: A review. Algal Res. 12, 497–512.

Mairet, F., Bernard, O., Lacour, T., Sciandra, A., 2011a. Modelling microalgae growth in

nitrogen limited photobiorector for estimating biomass , carbohydrate and neutral

lipid, in: Preprints of the 18th IFAC World Congress. Milano (Italy), pp. 10591–

10596.

Mairet, F., Bernard, O., Masci, P., Lacour, T., Sciandra, A., 2011b. Modelling neutral

lipid production by the microalga Isochrysis aff. galbana under nitrogen limitation.

Bioresour. Technol. 102, 142–9.

Markou, G., Angelidaki, I., Georgakakis, D., 2012. Microalgal carbohydrates: an

overview of the factors influencing carbohydrates production, and of main

bioconversion technologies for production of biofuels. Appl. Microbiol. Biotechnol.

96, 631–645.

Maršálková, B., Širmerová, M., Kuřec, M., Brányik, T., 2010. Microalgae Chlorella sp .

as an Alternative Source of Fermentable Sugars. Chem. Eng. Trans. 21, 1279–1284.

McNichol, J., MacDougall, K.M., Melanson, J.E., McGinn, P.J., 2012. Suitability of

Soxhlet Extraction to Quantify Microalgal Fatty Acids as Determined by

Comparison with In Situ Transesterification. Lipids 47, 195–207.

Moon, M., Kim, C.W., Park, W.-K., Yoo, G., Choi, Y.-E., Yang, J.-W., 2013.

Mixotrophic growth with acetate or volatile fatty acids maximizes growth and lipid

production in Chlamydomonas reinhardtii. Algal Res. 2, 352–357.

Morel, F.M.M., 1987. Kinetics of Nutrient Uptake and Growth in Phytoplankton. J.

Phycol. 23, 137–150.

Packer, A., Li, Y., Andersen, T., Hu, Q., Kuang, Y., Sommerfeld, M., 2011. Growth and

neutral lipid synthesis in green microalgae: a mathematical model. Bioresour.

Technol. 102, 111–7.



105

Rodolfi, L., Chini Zittelli, G., Bassi, N., Padovani, G., Biondi, N., Bonini, G., Tredici,

M.R., 2009. Microalgae for oil: strain selection, induction of lipid synthesis and

outdoor mass cultivation in a low-cost photobioreactor. Biotechnol. Bioeng. 102,

100–12.

Rügen, M., Bockmayr, A., Legrand, J., Cogne, G., 2012. Network reduction in metabolic

pathway analysis: Elucidation of the key pathways involved in the photoautotrophic

growth of the green alga Chlamydomonas reinhardtii. Metab. Eng. 14, 458–467.

Scaife, M.A., Merkx-Jacques, A., Woodhall, D.L., Armenta, R.E., 2015. Algal biofuels

in Canada: Status and potential. Renew. Sustain. Energy Rev. 44, 620–642.

Shuler, M.L., Kargi, F., 1992. Bioprocess Engineering: Basic Concepts. Prentice Hall.

Turon, V., Baroukh, C., Trably, E., Latrille, E., Fouilland, E., Steyer, J.-P., 2014. Use of

fermentative metabolites for heterotrophic microalgae growth: Yields and kinetics.

Bioresour. Technol. 175C, 342–349.

Vlysidis, A., Binns, M., Webb, C., Theodoropoulos, C., 2011. Glycerol utilisation for the

production of chemicals: Conversion to succinic acid, a combined experimental and

computational study. Biochem. Eng. J. 58–59, 1–11.

Xin, L., Hu, H., Ke, G., Sun, Y., 2010. Effects of different nitrogen and phosphorus

concentrations on the growth, nutrient uptake, and lipid accumulation of a

freshwater microalga Scenedesmus sp. Bioresour. Technol. 101, 5494–500.

Yao, C., Ai, J., Cao, X., Xue, S., Zhang, W., 2012. Enhancing starch production of a

marine green microalga Tetraselmis subcordiformis through nutrient limitation.


Zhang, K., Burns, I.G., Turner, M.K., 2008. Derivation of a Dynamic Model of the

Kinetics of Nitrogen Uptake Throughout the Growth of Lettuce: Calibration and

Validation. J. Plant Nutr. 31, 1440–1460.

Zhang, X.-W., Chen, F., Johns, M.R., 1999. Kinetic models for heterotrophic growth of

Chlamydomonas reinhardtii in batch and fed-batch cultures. Process Biochem. 35,

385–389.



106



107

3.3. Supplementary Information 1.

Additional information supporting and/or expanding the findings shown previously is

presented next.



108

SUPPLEMENTARY INFORMATION

Associated to:

Kinetic Modelling of Starch and Lipid Formation during Mixotrophix, Nutrient-

limited Microalgal Growth

Gonzalo M. Figueroa-Torresa, Jon K. Pittmanb, Constantinos Theodoropoulosa,*


Engineering Group, The University of Manchester, Manchester, M13 9PL

b School of Earth and Environmental Sciences, The University of Manchester,

Manchester, M13 9PL



Phone number: (+44) 161 306 4386





109

Appendix A. Experimental datasets.

Experimental datasets used for parameter fitting, as obtained from lab-scale cultures.

Data is the average of two biological replicates.

Table A.1. Dataset from culture grown in: TAP (No = 0.3824 gN L-1 and Ao = 0.42

gC L-1). Time Biomass (DCW) Nitrogen N. quota Acetate Starch Lipids Active biomass pH

h gC L-1 gN L-1 gN gC-1 gC L-1 gC L-1 gC L-1 gC L-1

0 0.00105 0.38240 1.00000 0.42000 0.00010 0.00015 0.00079 7.00

48 0.01406 0.09204 20.65586 0.34612 0.00160 0.00208 0.01037 7.14

68 0.07645 0.08901 3.83754 0.24167 0.00432 0.00472 0.06741 7.35

90 0.14126 0.09818 2.01203 0.09319 0.00874 0.01161 0.12091 7.79

115 0.24427 0.09916 1.15955 0.02056 0.01244 0.02738 0.20445 7.88

144 0.30059 0.08772 0.98033 0.00400 0.01593 0.03755 0.24711 8.15

168 0.31664 0.08766 0.93085 0.00000 0.01710 0.04351 0.25603 8.30

192 0.31832 0.08777 0.92557 0.00000 0.01790 0.04483 0.25559 8.28

Table A.2. Dataset from culture grown in: N- - (No = 0.3350 gN L-1 and Ao = 0.42 gC

L-1). Time Biomass (DCW) Nitrogen N. quota Acetate Starch Lipids Active biomass pH


0 0.00105 0.33500 1.00000 0.42000 0.00010 0.00015 0.00079 7.00

48 0.02777 0.12473 7.57137 0.33987 0.00109 0.00407 0.02262 7.00

72 0.08538 0.12030 2.51476 0.19412 0.00477 0.00717 0.07344 7.45

98 0.17233 0.11991 1.24808 0.08062 0.01458 0.01888 0.13888 7.73

122 0.22121 0.11099 1.01266 0.02647 0.02541 0.03186 0.16394 7.88

144 0.27014 0.10991 0.83323 0.00000 0.03869 0.05516 0.17629 8.14

168 0.27552 0.10199 0.84571 0.00000 0.04296 0.05731 0.17524 8.13

192 0.28094 0.10991 0.80119 0.00000 0.04727 0.05961 0.17406 8.12

Table A.3. Dataset from culture grown in: A++ (No = 0.3824 gN L-1 and Ao = 1.26 gC

L-1). Time Biomass (DCW) Nitrogen N. quota Acetate Starch Lipids Active biomass pH

h gC L-1 gN L-1 gN gC-1 gC L-1 gC gC-1 gC gC-1 gC gC-1

0 0.00105 0.38240 1.00000 1.26000 0.00010 0.00015 0.00079 7.00

48 0.01778 0.05863 18.20639 1.14502 0.00290 0.00265 0.01224 7.06

72 0.09754 0.03515 3.56021 1.01186 0.00548 0.01046 0.08160 7.39

96 0.20409 0.03200 1.71692 0.85857 0.01004 0.02274 0.17131 7.63

120 0.33426 0.03900 1.02734 0.66372 0.02031 0.04214 0.27181 7.97

144 0.39395 0.03000 0.89453 0.45977 0.03034 0.06350 0.30011 8.29

168 0.40917 0.02800 0.86613 0.38052 0.03620 0.07576 0.29722 8.52

192 0.41425 0.02500 0.86276 0.34577 0.03800 0.07582 0.30043 8.42

Table A.4. Dataset from culture grown in: High N (No = 0.7426 gN L-1 and Ao = 0.42

gC L-1). Time Biomass (DCW) Nitrogen Nit quota Acetate Starch Lipids Active biomass pH


0 0.00105 0.74260 1.00000 0.42000 0.00010 0.00015 0.00079 7.00

48 0.01155 0.67820 5.57363 0.40756 0.00127 0.00171 0.00857 7.11

74 0.08516 0.63225 1.29585 0.39513 0.00194 0.01535 0.06787 7.28

96 0.11523 0.61580 1.10039 0.28687 0.00608 0.01740 0.09176 7.37

120 0.14512 0.55780 1.27345 0.17860 0.01028 0.01893 0.11591 7.46

170 0.16764 0.52875 1.27569 0.06886 0.01405 0.02184 0.13174 7.57

192 0.16770 0.54195 1.19651 0.06886 0.01413 0.02421 0.12936 7.58



110

Table A.5. Dataset obtained from culture grown in: High A-N (No = 0.7426 gN L-1

and Ao = 2.52 gC L-1). Time Biomass (DCW) Nitrogen N. quota Acetate Starch Lipids Active biomass pH


0 0.00105 0.74260 1.00000 2.52000 0.00010 0.00015 0.00079 7.00

48 0.01664 0.63240 6.62351 2.58000 0.00204 0.00246 0.01214 6.94

72 0.05503 0.57222 3.09626 2.47228 0.00390 0.00608 0.04505 6.99

96 0.14179 0.56222 1.27216 2.18427 0.00736 0.01375 0.12068 7.32

144 0.19781 0.54160 1.01614 2.06362 0.01175 0.02647 0.15959 7.71

168 0.20553 0.55008 0.93670 1.97515 0.01581 0.03183 0.15788 7.65

192 0.20965 0.54419 0.94638 1.98139 0.01423 0.02958 0.16584 7.68

Experimental datasets used for model validation, as obtained from lab-scale culture

cultures. Data is the average of two biological replicates.

Table A.6. Dataset from culture grown in: N- (No = 0.354 gN L-1 and Ao = 0.42 gC L-

1). Time Biomass (DCW) Nitrogen Nit quota Acetate Starch Lipids Active biomass pH


0 0.00105 0.35430 1.00000 0.42000 0.00010 0.00015 0.00079 7.00

48 0.01896 0.11002 12.88144 0.34951 0.00141 0.00279 0.01477 6.98

75 0.08189 0.09705 3.14149 0.21067 0.00597 0.00924 0.06668 7.31

98 0.15951 0.08919 1.66205 0.06025 0.01315 0.02390 0.12246 7.59

122 0.25562 0.08919 1.03715 0.02132 0.02223 0.03915 0.19424 7.70

144 0.29031 0.08919 0.91320 0.00800 0.02786 0.05431 0.20815 7.88

168 0.30282 0.08919 0.87550 0.00000 0.02965 0.05502 0.21815 8.11

192 0.30488 0.08919 0.86958 0.00000 0.03090 0.05659 0.21739 8.08

Table A.7. Dataset from culture grown in: A+ (No = 0.3824 gN L-1 and Ao = 0.75 gC

L-1). Time Biomass (DCW) Nitrogen Nit quota Acetate Starch Lipids Active biomass pH


0 0.00105 0.38240 1.00000 0.75000 0.00010 0.00015 0.00079 7.00

48 0.01723 0.08111 17.49073 0.69855 0.00167 0.00401 0.01154 7.15

72 0.09965 0.06300 3.20530 0.50837 0.00323 0.01147 0.08495 7.47

96 0.19677 0.05400 1.66898 0.42188 0.01096 0.02125 0.16455 7.86

120 0.30352 0.05300 1.08527 0.25726 0.01652 0.04613 0.24088 8.07

144 0.36536 0.04518 0.92300 0.19122 0.01863 0.06382 0.28290 8.32

168 0.37990 0.04509 0.88789 0.15703 0.02198 0.06853 0.28938 8.36

192 0.39019 0.04502 0.86467 0.08694 0.02197 0.06661 0.30161 8.37

Table A.8. Dataset from culture grown in: A’-N’ (No = 0.3151 gN L-1 and Ao = 1.16

gC L-1). Time Biomass (DCW) Nitrogen Nit quota Acetate Starch Lipids Active biomass pH


0 0.00105 0.31510 1.00000 1.16000 0.00010 0.00015 0.00079 7.00

48 0.02754 0.13420 6.56761 1.12266 0.00201 0.00405 0.02148 7.02

72 0.09013 0.12032 2.16115 0.91109 0.00482 0.00875 0.07656 7.27

96 0.16787 0.12352 1.14122 0.77674 0.01778 0.01989 0.13020 7.58

144 0.24133 0.12352 0.79386 0.58752 0.04245 0.04517 0.15372 7.83

168 0.24391 0.12352 0.78546 0.57600 0.04937 0.04667 0.14787 7.90

192 0.23367 0.12352 0.81988 0.58326 0.05355 0.04787 0.13225 7.74



111

Figure A.1. Comparison between model predictions (lines) and experimental data

(points) for lab-scale cultures grown under all different nutrient regimes.

Appendix B. Sensitivity analysis.

The model proposed in this work originally consisted of 31 kinetic parameters instead of

the 29 presented in the manuscript. The additional parameters, 𝐾𝑠,𝑆 and 𝜙𝐿, were part of

Eq. (10) and Eq. (11) in the text, which were initially expressed as follows:

𝑅1 = 𝑟1 ∙𝑁𝑖

𝑛𝑠



𝑛𝑠∙

𝑘1

𝑘1 + 𝑁 𝑁𝑜⁄∙ [1 +

1

𝜇∙ 𝑒𝜙𝑆∙𝐴𝑖𝑛𝑡] ∙ 𝜇 ∙ 𝑥∗

𝑅3 = 𝑟3 ∙𝑁𝑖

𝑛𝐿



𝑛𝐿∙

𝑘2

𝑘2 + 𝑁 𝑁0⁄∙ [1 +

1

𝜇∙ 𝑒𝜙𝐿∙𝐴𝑖𝑛𝑡] ∙ 𝜇 ∙ 𝑥∗

All remaining equations and parameters remain unchanged, as shown in manuscript. A

sensitivity analysis was performed for all 31 parameters. This was carried out by

calculating the sensitivity (Eq. (B.1)), for all 8 state variables with respect to each

parameter at four different cultivation times (t= 60, 90, 125, and 190 h).



112

𝑆𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 = 𝑎𝑏𝑠 (

𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑖𝑜𝑛(𝑡, 𝑃 + 𝛥𝑃) − 𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑖𝑜𝑛(𝑡, 𝑃 − ∆𝑃)

2 ∗ ∆𝑃) Eq. B.1

A 10 % change (ΔP) in parameter values was used in all calculations. The results of the

sensitivity analysis for all 31 kinetic parameters are shown in Table B.1. The threshold for

sensitivity was 0.02, i.e. parameters with sensitivities lower than 0.02 were deemed not-

sensitive Sensitivities greater than 0.02 are highlighted.

Table B.1. Sensitivity analysis of the model kinetic parameters.

Sensitivity

Parameter Value Variable 60 h 90 h 125 h 190 h

µmax 0.106 X 1.0442 3.3231 0.1298 3.0151

N 2.2789 2.7440 2.7440 2.7440

qN 302.7901 70.2186 10.7041 0.2853

A 2.6658 4.7339 0.0332 0.4044

S 0.0565 0.2514 0.3166 0.4609

L 0.0956 0.3048 0.0243 0.5402

x* 0.8921 2.7669 0.1626 2.9358

pH 4.1483 12.8659 0.7560 13.6514

qN,0 0.7893 X 0.0098 0.0845 0.2720 0.3373

N 0.0171 0.0210 0.0210 0.0210

qN 2.6372 1.2683 1.0298 1.0000

A 0.0250 0.1161 0.0607 0.0442

S 0.0005 0.0032 0.0011 0.0191

L 0.0009 0.0047 0.0027 0.0589

x* 0.0084 0.0766 0.2758 0.4153

pH 0.0391 0.3564 1.2824 1.9312

Ks,A 1.789 X 0.0001 0.0005 0.0005 0.0003

N 0.0002 0.0002 0.0002 0.0002

qN 0.0268 0.0054 0.0008 0.0000

A 0.0149 0.0783 0.0454 0.0297

S 0.0000 0.0003 0.0018 0.0033

L 0.0000 0.0001 0.0004 0.0014

x* 0.0001 0.0007 0.0019 0.0021

pH 0.0006 0.0033 0.0088 0.0099

ki,A 0.11 X 0.0001 0.0054 0.0236 0.0303

N 0.0253 0.0267 0.0267 0.0267

qN 0.6891 0.1125 0.0121 0.0002

A 0.2510 0.3287 0.0538 0.0308

S 0.0003 0.0029 0.0070 0.0141

L 0.0001 0.0006 0.0035 0.0084

x* 0.0005 0.0077 0.0271 0.0361

pH 0.0022 0.0360 0.1260 0.1679

Ks,I OK to 1.4 1.078 X 0.0007 0.0022 0.0002 0.0020

N 0.0015 0.0018 0.0018 0.0018

qN 0.1945 0.0418 0.0058 0.0001

A 0.0014 0.0022 0.0003 0.0005

S 0.0000 0.0002 0.0002 0.0005



113

L 0.0001 0.0002 0.0000 0.0004

x* 0.0006 0.0018 0.0003 0.0022

pH 0.0027 0.0086 0.0016 0.0100

ki,I 186.52 X 0.0002 0.0008 0.0000 0.0007

N 0.0005 0.0006 0.0006 0.0006

qN 0.0681 0.0152 0.0022 0.0001

A 0.0004 0.0007 0.0001 0.0002

S 0.0000 0.0001 0.0001 0.0001

L 0.0000 0.0001 0.0000 0.0001

x* 0.0002 0.0007 0.0001 0.0007

pH 0.0009 0.0031 0.0003 0.0032

Ϭ 0.65 X 0.0001 0.0008 0.0011 0.0001

N 0.0001 0.0001 0.0001 0.0001

qN 0.0177 0.0123 0.0046 0.0002

A 0.0001 0.0004 0.0002 0.0004

S 0.0000 0.0000 0.0000 0.0000

L 0.0000 0.0000 0.0000 0.0001

x* 0.0001 0.0008 0.0012 0.0000

pH 0.0003 0.0036 0.0054 0.0001

ρN,max OK 40.445 X 0.0002 0.0020 0.0065 0.0081

N 0.0072 0.0072 0.0072 0.0072

qN 0.1587 0.0235 0.0025 0.0000

A 0.0005 0.0027 0.0015 0.0011

S 0.0000 0.0002 0.0005 0.0013

L 0.0000 0.0002 0.0007 0.0014

x* 0.0003 0.0020 0.0063 0.0080

pH 0.0012 0.0091 0.0293 0.0371

ФN 137.455 X 0.0000 0.0006 0.0022 0.0028

N 0.0024 0.0024 0.0024 0.0024

qN 0.0592 0.0092 0.0010 0.0000

A 0.0001 0.0008 0.0005 0.0004

S 0.0000 0.0000 0.0001 0.0004

L 0.0000 0.0001 0.0002 0.0005

x* 0.0000 0.0006 0.0021 0.0026

pH 0.0002 0.0027 0.0096 0.0122

Ks,N 0.163 X 0.0069 0.0888 0.3189 0.4004

N 0.3511 0.3537 0.3537 0.3537

qN 8.4105 1.2646 0.1346 0.0038

A 0.0178 0.1224 0.0719 0.0529

S 0.0017 0.0088 0.0330 0.1095

L 0.0007 0.0069 0.0340 0.0770

x* 0.0094 0.0907 0.3179 0.4328

pH 0.0435 0.4218 1.4782 2.0126

ki,N 0.113 X 0.0475 0.4074 1.3001 1.6073

N 1.4376 1.4195 1.4195 1.4195

qN 30.3029 4.4303 0.4712 0.0108

A 0.1217 0.5643 0.3016 0.2207

S 0.0068 0.0334 0.1308 0.3197

L 0.0032 0.0309 0.1358 0.2853

x* 0.0575 0.4098 1.2952 1.6418

pH 0.2675 1.9056 6.0225 7.6342

Ks,A:N 1.004 X 0.0053 0.0514 0.1714 0.2135

N 0.1896 0.1885 0.1885 0.1885

qN 4.1888 0.6185 0.0661 0.0014

A 0.0137 0.0708 0.0387 0.0283

S 0.0009 0.0044 0.0164 0.0356

L 0.0004 0.0042 0.0189 0.0382

x* 0.0066 0.0516 0.1689 0.2109



114

pH 0.0308 0.2402 0.7855 0.9806

YX/A 0.059 X 0.0014 0.0318 0.1335 0.1691

N 0.1383 0.1470 0.1470 0.1470

qN 3.6035 0.5963 0.0600 0.0005

A 1.4961 3.2952 1.0379 0.6857

S 0.0016 0.0210 0.0639 0.1246

L 0.0002 0.0039 0.0229 0.0624

x* 0.0032 0.0490 0.1745 0.2313

pH 0.0151 0.2276 0.8112 1.0758

K* 0.313 X 0.0184 0.1733 0.5582 0.6889

N 0.6112 0.6080 0.6080 0.6080

qN 13.5426 2.0028 0.2125 0.0041

A 0.0472 0.2478 0.1516 0.1155

S 0.0030 0.0161 0.0597 0.1913

L 0.0020 0.0089 0.0447 0.1005

x* 0.0233 0.1805 0.5732 0.7797

pH 0.1085 0.8393 2.6655 3.6256

n 18.183 X 0.0000 0.0004 0.0014 0.0017

N 0.0015 0.0015 0.0015 0.0015

qN 0.0338 0.0050 0.0005 0.0000

A 0.0001 0.0006 0.0003 0.0002

S 0.0000 0.0000 0.0001 0.0005

L 0.0000 0.0000 0.0001 0.0003

x* 0.0001 0.0004 0.0014 0.0019

pH 0.0003 0.0020 0.0064 0.0088

ki,A:N 1.098 X 0.0008 0.0070 0.0231 0.0287

N 0.0256 0.0254 0.0254 0.0254

qN 0.5564 0.0815 0.0086 0.0001

A 0.0020 0.0097 0.0052 0.0038

S 0.0001 0.0006 0.0024 0.0082

L 0.0001 0.0005 0.0024 0.0055

x* 0.0009 0.0071 0.0230 0.0314

pH 0.0044 0.0330 0.1071 0.1460

kH 4.65 X 0.0000 0.0000 0.0000 0.0000

N 0.0000 0.0000 0.0000 0.0000

qN 0.0000 0.0000 0.0000 0.0000

A 0.0000 0.0000 0.0000 0.0000

S 0.0000 0.0000 0.0000 0.0000

L 0.0000 0.0000 0.0000 0.0000

x* 0.0000 0.0000 0.0000 0.0000

pH 0.0280 0.1190 0.2283 0.2484

r1 0.0486 X 0.0003 0.0117 0.0316 0.0229

N 0.0009 0.0004 0.0004 0.0004

qN 0.0362 0.1054 0.0523 0.0101

A 0.0007 0.0237 0.0111 0.0055

S 0.0421 0.1877 0.6247 1.2180

L 0.0057 0.0275 0.1204 0.4629

x* 0.0361 0.1486 0.4727 0.7321

pH 0.1679 0.6910 2.1979 3.4043

Ks,S 0.0004 X 0.0000 0.0000 0.0000 0.0000

N 0.0000 0.0000 0.0000 0.0000

qN 0.0004 0.0001 0.0000 0.0000

A 0.0000 0.0000 0.0000 0.0000

S 0.0001 0.0001 0.0001 0.0001

L 0.0000 0.0001 0.0001 0.0001

x* 0.0000 0.0000 0.0000 0.0000

pH 0.0002 0.0001 0.0002 0.0002

ki,S 0.2137 X 0.0002 0.0029 0.0057 0.0029



115

N 0.0008 0.0014 0.0014 0.0014

qN 0.0790 0.0428 0.0139 0.0025

A 0.0005 0.0058 0.0022 0.0010

S 0.0192 0.0884 0.3204 0.7368

L 0.0015 0.0115 0.0571 0.2533

x* 0.0175 0.0739 0.2575 0.4806

pH 0.0814 0.3437 1.1976 2.2346

ФS 0.6751 X 0.0000 0.0000 0.0000 0.0000

N 0.0000 0.0000 0.0000 0.0000

qN 0.0003 0.0001 0.0000 0.0001

A 0.0000 0.0001 0.0000 0.0000

S 0.0001 0.0025 0.0204 0.0630

L 0.0000 0.0002 0.0031 0.0126

x* 0.0001 0.0023 0.0174 0.0503

pH 0.0004 0.0108 0.0807 0.2340

k1 0.1103 X 0.0001 0.0003 0.0037 0.0029

N 0.0003 0.0001 0.0001 0.0001

qN 0.0275 0.0077 0.0095 0.0041

A 0.0004 0.0004 0.0009 0.0005

S 0.0146 0.0669 0.2256 0.4099

L 0.0021 0.0091 0.0416 0.1579

x* 0.0123 0.0575 0.1803 0.2491

pH 0.0574 0.2675 0.8383 1.1584

r2 0.0033 X 0.0000 0.0026 0.0017 0.0569

N 0.0013 0.0015 0.0015 0.0015

qN 0.0505 0.0105 0.0041 0.1223

A 0.0002 0.0050 0.0004 0.0071

S 0.0449 0.8425 6.7797 16.1974

L 0.0048 0.0666 0.8481 5.3395

x* 0.0401 0.7733 5.9299 10.8011

pH 0.1863 3.5957 27.5741 50.2250

nS 4.144 X 0.0000 0.0000 0.0000 0.0000

N 0.0000 0.0000 0.0000 0.0000

qN 0.0032 0.0008 0.0001 0.0001

A 0.0000 0.0001 0.0000 0.0000

S 0.0003 0.0031 0.0131 0.0339

L 0.0000 0.0004 0.0024 0.0068

x* 0.0003 0.0027 0.0106 0.0271

pH 0.0014 0.0127 0.0494 0.1259

r3 0.162 X 0.0014 0.0032 0.0007 0.0023

N 0.0023 0.0019 0.0019 0.0019

qN 0.3820 0.0583 0.0076 0.0019

A 0.0037 0.0045 0.0001 0.0003

S 0.0020 0.0091 0.0427 0.1744

L 0.0230 0.1088 0.3577 0.8541

x* 0.0224 0.1029 0.3144 0.6774

pH 0.1043 0.4787 1.4618 3.1500

Ks,L 0.0227 X 0.0060 0.0140 0.0003 0.0106

N 0.0092 0.0077 0.0077 0.0077

qN 1.4988 0.2306 0.0337 0.0044

A 0.0154 0.0197 0.0008 0.0014

S 0.0023 0.0043 0.0077 0.0158

L 0.0056 0.0040 0.0054 0.0100

x* 0.0093 0.0137 0.0026 0.0164

pH 0.0432 0.0635 0.0121 0.0761

ki,L 0.0861 X 0.0023 0.0048 0.0014 0.0035

N 0.0038 0.0032 0.0032 0.0032

qN 0.6052 0.0950 0.0121 0.0042



116

A 0.0058 0.0068 0.0001 0.0005

S 0.0039 0.0241 0.1241 0.4367

L 0.0606 0.3231 1.0794 2.4101

x* 0.0590 0.3039 0.9539 1.9699

pH 0.2743 1.4129 4.4356 9.1602

ФL 1.00E-05 X 0.0000 0.0000 0.0000 0.0000

N 0.0000 0.0000 0.0000 0.0000

qN 0.0004 0.0000 0.0000 0.0000

A 0.0000 0.0000 0.0000 0.0000

S 0.0000 0.0001 0.0017 0.0112

L 0.0002 0.0033 0.0192 0.0534

x* 0.0002 0.0032 0.0175 0.0423

pH 0.0007 0.0147 0.0814 0.1966

k2 0.2135 X 0.0010 0.0023 0.0004 0.0016

N 0.0016 0.0014 0.0014 0.0014

qN 0.2671 0.0418 0.0052 0.0011

A 0.0026 0.0032 0.0000 0.0002

S 0.0011 0.0044 0.0192 0.0842

L 0.0114 0.0480 0.1550 0.3691

x* 0.0113 0.0458 0.1353 0.2833

pH 0.0525 0.2131 0.6293 1.3173

r4 0.0049 X 0.0000 0.0000 0.0002 0.0133

N 0.0002 0.0002 0.0002 0.0002

qN 0.0111 0.0017 0.0002 0.0363

A 0.0000 0.0000 0.0000 0.0018

S 0.0023 0.0360 0.5080 4.1835

L 0.0425 0.8125 6.4409 22.3625

x* 0.0402 0.7765 5.9331 18.1923

pH 0.1869 3.6107 27.5891 84.5942

nL 1.8117 X 0.0001 0.0002 0.0000 0.0001

N 0.0001 0.0001 0.0001 0.0001

qN 0.0169 0.0027 0.0003 0.0002

A 0.0002 0.0002 0.0000 0.0000

S 0.0001 0.0011 0.0065 0.0221

L 0.0027 0.0174 0.0594 0.1215

x* 0.0025 0.0161 0.0529 0.0995

pH 0.0117 0.0751 0.2458 0.4625

As observed in Table B.1, two parameters had sensitivities lower than 0.02: Ϭ and Ks,S.

The sensitivity of four parameters (ФN, ki,I , Ks,A and Ks,I ) was further evaluated at a

greater number of time points to improve the sensitivity assessment. This additional test

is shown in Table B.2.



117

Table B.2. Sensitivity analysis for ФN, ki,I , Ks,A, and Ks,I at a 8 different time points.

Sensitivity

Variable 25 h 50 h 75 h 85 h 125 h 150 h 175 h 190 h

ФN X 0.0000 0.0000 0.0003 0.0011 0.0043 0.0053 0.0055 0.0055

N 0.0005 0.0039 0.0049 0.0049 0.0049 0.0049 0.0049 0.0049

qN 0.1124 0.1814 0.0476 0.0185 0.0020 0.0004 0.0001 0.0000

A 0.0000 0.0001 0.0007 0.0016 0.0010 0.0008 0.0007 0.0007

S 0.0000 0.0000 0.0001 0.0001 0.0003 0.0005 0.0007 0.0007

L 0.0000 0.0000 0.0000 0.0001 0.0005 0.0007 0.0009 0.0010

x* 0.0000 0.0000 0.0004 0.0011 0.0041 0.0050 0.0052 0.0052

pH 0.0000 0.0002 0.0017 0.0053 0.0192 0.0233 0.0243 0.0244

ki,I X 0.0000 0.0002 0.0011 0.0015 0.0003 0.0007 0.0010 0.0010

N 0.0002 0.0003 0.0009 0.0009 0.0009 0.0009 0.0009 0.0009

qN 0.0851 0.1721 0.0639 0.0381 0.0043 0.0011 0.0003 0.0001

A 0.0000 0.0003 0.0017 0.0023 0.0019 0.0047 0.0056 0.0057

S 0.0000 0.0000 0.0001 0.0001 0.0001 0.0001 0.0003 0.0003

L 0.0000 0.0000 0.0001 0.0002 0.0000 0.0002 0.0002 0.0002

x* 0.0000 0.0002 0.0009 0.0012 0.0002 0.0004 0.0005 0.0005

pH 0.0001 0.0009 0.0042 0.0058 0.0010 0.0017 0.0022 0.0021

Ks,A X 0.0000 0.0003 0.0013 0.0017 0.0000 0.0009 0.0012 0.0013

N 0.0003 0.0004 0.0011 0.0011 0.0011 0.0011 0.0011 0.0011

qN 0.1054 0.2097 0.0751 0.0439 0.0042 0.0009 0.0002 0.0001

A 0.0007 0.0039 0.0183 0.0328 0.1451 0.1774 0.1850 0.1863

S 0.0000 0.0000 0.0000 0.0000 0.0027 0.0065 0.0101 0.0116

L 0.0000 0.0000 0.0001 0.0002 0.0003 0.0010 0.0010 0.0011

x* 0.0000 0.0002 0.0011 0.0016 0.0024 0.0045 0.0079 0.0114

pH 0.0001 0.0011 0.0051 0.0072 0.0111 0.0211 0.0369 0.0530

Ks,I X 0.0000 0.0000 0.0015 0.0021 0.0002 0.0015 0.0019 0.0020

N 0.0003 0.0000 0.0018 0.0018 0.0018 0.0018 0.0018 0.0018

qN 0.1222 0.0029 0.0939 0.0549 0.0058 0.0013 0.0003 0.0001

A 0.0001 0.0000 0.0028 0.0027 0.0003 0.0004 0.0004 0.0005

S 0.0000 0.0000 0.0001 0.0001 0.0002 0.0003 0.0004 0.0005

L 0.0000 0.0000 0.0001 0.0002 0.0000 0.0002 0.0003 0.0004

x* 0.0000 0.0000 0.0013 0.0017 0.0003 0.0016 0.0021 0.0022

pH 0.0002 0.0000 0.0060 0.0081 0.0016 0.0076 0.0096 0.0100

From the sensitivity analysis, four parameters were identified as not-sensitive: Ϭ, Ks,S, and

Ks,I. and ФL, from which two (Ks,S and ФL) were neglected given that results did not change

when set to zero. Thus, these four parameters were adjusted as follows:

Parameter Final value

Ϭ 1

Ks,I 1.4 (Mairet et al., 2011)

Ks,S 0

ФL 0



118

Eq. (10) and Eq. (11) were thus modified accordingly to account for the removal of Ks,S

and ФL, resulting in Eq. (23) and Eq. (24):

𝑅1 = 𝑟1 ∙𝑁𝑖

𝑛𝑠

𝑁𝑖𝑛𝑠 + (𝑁𝑖

2 𝑘𝑖,𝑆⁄ )𝑛𝑠

∙𝑘1

𝑘1 + 𝑁 𝑁𝑜⁄∙ [1 +

1


𝑅3 = 𝑟3 ∙𝑁𝑖

𝑛𝐿



𝑛𝐿∙

𝑘2

𝑘2 + 𝑁 𝑁0⁄∙ [𝜇 + 1] ∙ 𝑥∗

References

Mairet, F., Bernard, O., Masci, P., Lacour, T., Sciandra, A., 2011. Modelling neutral lipid

production by the microalga Isochrysis aff. galbana under nitrogen limitation.

Bioresour. Technol. 102, 142–9



119

Appendix C. COMPLEMENTARY INFORMATION

The following material complements the computational results presented in

(Figueroa-Torres et. al., 2017), but is not available in the electronic version of the

publication.

C.1. Clarification of modelling equations.

Specific growth rate:

The specific growth rate, µ, is dependent on the medium concentration of acetic acid, A,

the light received by the culture, I, and the nitrogen quota, 𝑞𝑁, as follows:

𝜇 = ��𝑀,𝑚𝑎𝑥(𝐴, 𝐼) ∙ (1 −𝑞𝑁,0

𝑞𝑁)

��𝑀,𝑚𝑎𝑥(𝐴, 𝐼) = 𝜇𝑚𝑎𝑥 ∙ [𝑤𝐻 ∙ 𝜇𝐻(𝐴) + 𝑤𝐼 ∙ 𝜇𝐼(𝐼)]

As explained in text, the weighting functions above, wH and wI, are defined in terms of the

saturation constants of the acetic acid and the light distribution by using the formulation of

Eq. 2.12 presented in Chapter 2 (Literature Review, Section 2.4.1.2) of this thesis. Such

formulation, despite its empirical nature, was found to be applicable for additive-type

growth kinetics as the one employed here.

As per the original formulation of Eq. 2.12, the weighing functions would be expressed as:

𝑤𝐻 =𝐾𝑠,𝐴/𝐴

𝐾𝑠,𝐴/𝐴+𝐾𝑠,𝐼/𝐼 ; 𝑤𝐼 =

𝐾𝑠,𝐼/𝐼

𝐾𝑠,𝐴/𝐴+𝐾𝑠,𝐼/𝐼

The saturating terms (i.e. 𝐾𝑠,𝐴/𝐴 and 𝐾𝑠,𝐼/𝐼), however, were instead inverted to avoid

convergence problems during the optimisation-based fitting methodology used in this

work, which arose when either acetate was exhausted (i.e. A = 0), or under complete light

attenuation (i.e. I = 0) conditions. Weighing functions were thus simply expressed as in

Eq. 6 of the main text (Contribution 1):



120


𝐴 𝐾𝑠,𝐴⁄ +𝐼 𝐾𝑠,𝐼⁄ ; 𝑤𝐼 =


𝐴 𝐾𝑠,𝐴⁄ +𝐼 𝐾𝑠,𝐼⁄

The light, I, received by the culture throughout the vessel is attenuated as the algal density

(denoted by the biomass concentration) increases. This attenuation was portrayed by the

Beer-Lambert law, which assumes an exponential decrease in light as biomass increases.

This function depends on the incident light, I0, the residual biomass concentration, X, and

the depth of the culture, z, as follows:

𝐼 = 𝐼0 ∙ 𝑒−𝜎∙𝑋∙𝑧

The culture vessels used for experimentation consisted of 500 mL clear glass bottles with

plastic caps (Duran®). As the incident light was supplied from above, the culture depth, z,

was measured from the bottom of the vessel up to the surface of the algal culture within

the bottle. Other light considerations dependent on the geometry of the vessel were

neglected.

Nitrogen uptake rate:

The nitrogen uptake rate was expressed by employing double-substrate inhibited-type

kinetics, dependent on the residual nitrogen, N, and acetic, A, concentration, as follows:

𝜌𝑁 = ��𝑁,𝑚𝑎𝑥(𝑁𝑜, 𝑋) ∙𝑁

𝑁 + 𝐾𝑠,𝑁 + 𝑁2

𝑘𝑖,𝑁⁄

∙𝐴

𝐴 + 𝐾𝑠,𝐴:𝑁 + 𝐴2

𝑘𝑖,𝐴:𝑁⁄

Whilst most models employ simple Monod-type kinetics to simulate nutrient consumption

(Bougaran et al., 2010; Droop, 1968; Mairet et al., 2011), the experimental data obtained

in this work (see Figure 2 of the main text) indicated that high concentrations of both

nitrogen, N, and acetic, A, were inhibitory for biomass growth, and for nitrogen uptake

itself (Figure A.1, shown above). Therefore, the nitrogen uptake rate employed Andrews

functions, which account for inhibition due to the incorporation of the inhibition constants

𝑘𝑖,𝑁 and 𝑘𝑖,𝐴:𝑁.

In all experiments, the uptake rate of nitrogen was observed to occur initially fast, followed

by a sudden stop after two days of cultivation (residual nitrogen concentration remained

relatively constant after this point). Additionally, the initial nitrogen concentration of the



121

culture media, which ranged from 0.3151 gN L-1 to 0.7430 gN L-1, was observed to regulate

the magnitude of nitrogen consumption. Therefore, the maximum nitrogen uptake rate,

��𝑁,𝑚𝑎𝑥(𝑁𝑜 , 𝑋), was expressed as a saturating function dependent on the initial nitrogen

concentration, N0, and on the residual biomass concentration, X:


𝑛


𝑛 ∙ 𝑒−𝜙𝑁∙𝑋

In line with the term 𝑒−𝜙𝑁∙𝑋 , nitrogen uptake decreases exponentially with increasing

biomass concentration, which allows nitrogen consumption to stop appropriately. This

term was found to fit rather well model predictions to experimental datasets, albeit by

requiring one additional kinetic parameter (i.e. 𝜙𝑁). It should be mentioned, however, that

one alternative expression that could be explored is the Contois model, a Monod-type

kinetic function in which the half-saturation constant is regulated by residual biomass

(Contois, 1959):

Contois model: 𝜇 =𝑆

𝐾𝑆∙𝑋+𝑆

Rates of formation of cellular compartments (R1, R2, R3, R4).

The starch and lipid synthetic rates, R1 and R3, are expressed as follows:

𝑅1 = 𝑟1 ∙𝑁𝑖

𝑛𝑠



𝑛𝑠∙

𝑘1

𝑘1 + 𝑁 𝑁𝑜⁄∙ [1 +

1


𝑅3 = 𝑟3 ∙𝑁𝑖

𝑛𝐿



𝑛𝐿∙

𝑘2

𝑘2 + 𝑁 𝑁0⁄∙ [1 +

1

𝜇∙ 𝑒𝜙𝐿∙𝐴𝑖𝑛𝑡] ∙ 𝜇 ∙ 𝑥∗

As observed, both expressions share the same structure:

The first term in both synthetic rates is dependent on the internal nitrogen

concentration (i.e. Ni = qN·X) and employs inhibited-type kinetics to portray the

following: as the concentration of internal nitrogen increases (during nitrogen-

replete conditions), storage molecule formation decreases; on the contrary, when

the concentration of internal nitrogen decreases (during nitrogen-limited

conditions), storage molecule formation increases. This was in line with

experimental observations. The shape-controlling coefficients, nS and nL, were



122

found to improve the fitting of the model to experimental data, and derive from the

model proposed by Molina-Grima (see Chapter 2, section 2.4.1.1).

𝑁𝑖𝑛𝑠

𝑁𝑖𝑛𝑠+𝐾𝑠,𝑆

𝑛𝑠+(𝑁𝑖2 𝑘𝑖,𝑆⁄ )

𝑛𝑠 and 𝑁𝑖

𝑛𝐿

𝑁𝑖𝑛𝐿+𝐾𝑠,𝐿

𝑛𝐿+(𝑁𝑖2 𝑘𝑖,𝐿⁄ )

𝑛𝐿

The second term is a regulating function dependent on the ratio of residual nitrogen

to initial nitrogen supplied (i.e. N/N0) to microalgal cultures, so that starch and lipid

formation is greater as the fraction (scaled to each culture) of residual nitrogen

decreases (i.e. as (𝑁 𝑁0)⁄ 0, [𝑘/(𝑘 + 𝑁 𝑁0⁄ )] 1). This term has been also

employed by Bekirogullari et al. (2017) to portray the increased formation of lipids

as external nitrogen concentration decreases, but without accounting for the initial

nitrogen supplied.

𝑘1

𝑘1+𝑁 𝑁𝑜⁄ and

𝑘2

𝑘2+𝑁 𝑁0⁄

The third term is an exponential term dependent on the internal concentration of

acetic acid (i.e. Aint = Ao-A), so that formation of storage molecules increased as the

internal acetic acid concentration increases, which is in line with experimental

observations: high-acetate treatment yielded higher starch and lipid concentrations.

In addition, the storage molecules (particularly starch, see Figure 3.e in main text)

were observed to increase even after biomass reached stationary phase. Therefore,

the exponential term was divided by the specific growth rate, µ, so as to uncouple

the acetate-induced storage formation from cellular growth. It should be

acknowledged that whilst this formulation is purely empirical, its implementation

in the expressions above was found to simulate adequately the cultivation dynamics

observed experimentally.

[1 +1

𝜇∙ 𝑒𝜙𝑆∙𝐴𝑖𝑛𝑡] and [1 +

1

𝜇∙ 𝑒𝜙𝐿∙𝐴𝑖𝑛𝑡]

The starch and lipid degradation rates were expressed as functions inversely proportional

to the nitrogen quota, as follows:

𝑅2 =𝑟2

𝑞𝑁∙ 𝑋; 𝑅4 =

𝑟4

𝑞𝑁∙ 𝑋

Although a low nitrogen quota is representative of nitrogen-limited conditions (which are

widely known for inducing starch and lipid formation), the degradation rates were

expressed as above to avoid excessive formation of starch and lipid molecules and maintain

the pool of active biomass. However, and as will be shown in the subsequent chapter, the

degradation rates used here were further improved to avoid unfeasible scenarios.



123

Time-dependent kinetic expressions.

The dynamics of the carbon-based model variables (X, S, L, x*, and A) were expressed as:

𝑑𝑋

𝑑𝑡= 𝜇 ∙ 𝑋

𝑑𝑆

𝑑𝑡= 𝑅1 − 𝑅2

𝑑𝐿

𝑑𝑡= 𝑅3 − 𝑅4

𝑑𝑥∗

𝑑𝑡= 𝜇 ∙ 𝑋 + 𝑅2 + 𝑅4 − (𝑅1 + 𝑅3)

The dynamics of X, S, L, and x* were obtained by the mass conservation balance (where

X = S + L + x*), and as per Figure 1 in the main text.

The dynamics of substrate (acetate) can be expressed by employing the substrate (acetate)

to biomass yield coefficient, YX/A, a parameter typically used to describe substrate uptake

in microbial kinetics (Shuler and Kargi, 1992), as in:

𝑑𝐴

𝑑𝑡= −

1

𝑌𝑋/𝐴∙ 𝜇 ∙ 𝑋 = −

1

𝑌𝑋/𝐴∙

𝑑𝑋

𝑑𝑡

However, the following consideration was made: whilst the specific growth rate expression

employed here dictates microalgal biomass (X) mixotrophic growth to be regulated by both

the carbon substrate (A) and the light intensity (I), the consumption of acetic acid depends

solely on the heterotrophic rate (i.e. 𝜇𝐻(𝐴)). Therefore, an additional fractional term was

incorporated into the acetate dynamics that accounts for heterotrophically-consumed

carbon (𝜇𝐻(𝐴)) as the culture grows mixotrophically:

𝑑𝐴

𝑑𝑡= −

1

𝑌𝑋/𝐴∙

𝜇𝐻(𝐴)

𝜇𝐻(𝐴) + 𝜇𝐼(𝐼)∙

𝑑𝑋

𝑑𝑡

* Note: the expression above could potentially be simplified if the weighing functions (as

employed in the specific growth rate) are included in the fractional term, so that:



124

𝑑𝐴

𝑑𝑡= −

1

𝑌𝑋 𝐴⁄∙

𝑤𝐻 ∙ 𝜇𝐻(𝐴)

𝑤𝐻 ∙ 𝜇𝐻(𝐴) + 𝑤𝐼 ∙ 𝜇𝐼(𝐼)∙

𝑑𝑋

𝑑𝑡

𝑑𝐴

𝑑𝑡= −

1

𝑌𝑋/𝐴∙

𝑤𝐻 ∙ 𝜇𝐻(𝐴)

𝑤𝐻 ∙ 𝜇𝐻(𝐴) + 𝑤𝐼 ∙ 𝜇𝐼(𝐼)∙ [��𝑀,𝑚𝑎𝑥(𝐴, 𝐼) ∙ (1 −

𝑞𝑁,0

𝑞𝑁)] ∙ 𝑋

𝑑𝐴

𝑑𝑡= −

1

𝑌𝑋/𝐴∙ 𝑤𝐻 ∙ 𝜇𝐻(𝐴) ∙ [𝜇𝑚𝑎𝑥 ∙ (1 −

𝑞𝑁,0

𝑞𝑁)] ∙ 𝑋

The simplified expression, however, was not applicable (data fitting was deemed

inadequate) to the microalgal system employed in this work.

C.2. Fitting data & parameter estimation.

As shown above, 8 different experimental datasets were obtained from this study. Each

dataset contained 8 state variables, measured along 7 – 8 points in time (Tables A.1 – A.8).

The total number of data points available was 488, from which 304 were used during the

optimisation-based fitting methodology, and 104 were used for validation purposes. Given

that the model shown within Contribution 1 employs 29 kinetic parameters, the data to

parameter ratio is: 304/29 = 10.48.

As already described within the main text, the fitting methodology yielded a set of kinetic

parameter that allowed to the model to adequately simulate nitrogen-limited mixotrophic

growth dynamics. However, it should be mentioned that despite the good level of

agreement observed between model outputs and experimental data, fitting methodologies

that involve a large number of kinetic parameters should be approached with judicious care

since the estimated parameters may not have an optimal degree of accuracy: parameters

might lack a physiological interpretation, comparisons with literature data may be non-

existent, and/or experimental data used for appropriate fitting may be scarce. In this regard,

the set of kinetic parameters identified in this work (which are again presented in Table

C.1), can be considered to be optimal for the microalgal cultivation system presented here,

which validates the model structure. However, numerical parameter values may need to be

re-identified: i) if additional phenomena is taken into account, ii) a diiferent algal strain is

employed, or iii) the cultivation is carried out via a different operating mode (e.g. fed-



125

batch, continuous). In fact, and as will be shown in subsequent chapters, some kinetic

parameter values were re-estimated to improve the model’s performance after

incorporating phosphorous-limited growth (see Chapter 4), or a pulse-assisted fed-batch

strategy (see Chapter 5).

Table C.1. List of kinetic parameter values* employed in the developed model.

Symbol Value Units Symbol Value Units

µmax 0.106 h-1 r1 0.049 gC gC-1

qN,0 0.876 gN gC-1 r2 0.003 gN gC-1h-1

Ks,A 1.79 gC L-1 r3 0.162 gC gC-1

ki,A 0.109 gC L-1 r4 0.005 gN gC-1h-1

Ks,I 1.4 µmol m-2s-1 Ks,S 0 gN L-1

ki,I 186.5 µmol m-2s-1 ki,S 0.214 gN L-1

YX/A 0.059 gC gC-1 nS 4.14 -

Ϭ 1 L gC-1 m-1 k1 0.110 -

ρN,max 40.45 gN gC-1h-1 ФS 0.675 L gC-1

K* 0.313 gN L-1 Ks,L 0.023 gN L-1

n 18.18 - ki,L 0.086 gN L-1

ФN 137.5 L gC-1 nL 1.81 -

Ks,N 0.163 gN L-1 k2 0.213 -

ki,N 0.113 gN L-1 ФL 0 L gC-1

Ks,A:N 1.004 gC L-1 KH 4.65 L gC-1 h-1

ki,A:N 1.098 gC L-1

* Values are rounded up to no more than 4 significant figures, and at least 3

decimals for parameters < 1.

C.3. Normalised sensitivity analysis.

The sensitivity analysis shown in Appendix B was carried out by calculating numerical

sensitivities using central finite differences (Eq. B.1). To assess further the effect of the

estimated model on the model state variables, a sensitivity analysis was performed by

calculating the normalised sensitivity, as follows (Cachon and Diviés, 1994):



126

𝑆𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 =

𝜕𝑍𝑖

𝜕𝑃𝑖∙

𝑃𝑖,𝑜

𝑍𝑖,𝑜=

𝜕𝑍𝑖/𝑍𝑖,𝑜

𝜕𝑃𝑖/𝑃𝑖,𝑜 Eq. C1

Where 𝜕𝑍𝑖 𝜕𝑃𝑖⁄ denotes the response change in a model state variable with respect to a

corresponding change in a model parameter, and 𝑍𝑖,𝑜 is the response of the model state

variable when the parameter is set to 𝑃𝑖,𝑜. This sensitivity reflects the effect of a parameter

in the model and indicates whether a change in the parameter leads to overpredicting or

underpredicting variables: if sensitivity > 1, a change in the parameter increases the

response of the model variable; if sensitivity < 1, a change in the parameter decreases the

response of the model variable. Meanwhile, the greater the sensitivity, the greater the effect

of the parameter.

For calculations, parameters were increased by 1 % with respect to their estimated value,

whilst keeping all other parameters constant (local sensitivity). To observe the full effect

of each parameter on the all variables, the sensitivity was computed over a 200 h cultivation

period, rather than at a single time. To solve the model and calculate sensitivities, the initial

conditions were set equivalent to those of standard [TAP] medium. The results are shown

in Figure C.1, where each plot shows the computed sensitivities for the model state

variables with respect to each kinetic parameter. To facilitate reading, parameters were

colour-labelled based on their association (as per the model equations) to biomass,

nitrogen, or starch and lipid dynamics.

As observed in Figure C.1, the maximum specific growth rate, 𝜇𝑚𝑎𝑥, and the minimum

nitrogen quota, 𝑞𝑁,0, are among the most significant parameters in the model given that a

small change can affect all state variables to a great extent, up to (-5 : 5), in the case of

𝜇𝑚𝑎𝑥, and (-2 : +4) in the case of 𝑞𝑁,0. These two parameters form the basis of the Droop-

type growth kinetics employed in this model, so their significance is expected as they

portray the growth of microalgal biomass (X) as the available external nitrogen (N), i.e. the

limiting nutrient, is consumed and stored in the form of an intracellular nitrogen quota

(𝑞𝑁).



127

Figure C.1. Normalised sensitivity of the model state variables with respect to a 1 %

increase in each model parameter, over a 200 h cultivation period. Parameter

colours denote: green – associated to biomass growth, purple – associated to N

uptake, black – associated to starch and lipid formation, red – associated to pH.



128







129





The effect of the other parameters associated to growth (i.e. 𝐾𝑆,𝐴, 𝑘𝑖,𝐴, 𝐾𝑆,𝐼, 𝑘𝑖,𝐼, 𝜎, and

𝑌𝑋/𝐴) on the model variables is mild, as indicated by their sensitivity values, although it

can be observed that the parameters 𝐾𝑆,𝐴, 𝑘𝑖,𝐴, and 𝑌𝑋/𝐴 only affect acetic acid dynamics,

as their computed sensitivity for the other variables is sufficiently low to be considered



130

insignificant. Meanwhile, the effect of the parameters 𝑘𝑖,𝐼 and 𝜎, was deemed insignificant

for all variables, as indicated by their low sensitivity values. Since these parameters are

used to portray the phototrophic growth and the light attenuation (as per the Beer-Lambert

law), their low sensitivity may indicate that under the mixotrophic conditions explored in

this work, the heterotrophic rate (rather than the phototrophic rate) is the main driver

behind the growth of biomass. These parameters were therefore set to 𝑘𝑖,𝐼 = 1.4 (Mairet

et al., 2011), and 𝜎 = 1 (nominal value).

The effect of the parameters associated to nitrogen uptake dynamics (i.e. 𝜌𝑁,𝑚𝑎𝑥, 𝐾∗, 𝑛,

𝜙𝑁 , 𝐾𝑆,𝑁 , 𝑘𝑖,𝑁, 𝐾𝑆,𝐴:𝑁 , and 𝑘𝑖,𝐴:𝑁) is observed to be significant for all model variables,

including nitrogen itself. However, the model variables more heavily influenced by

nitrogen uptake dynamics are acetic acid (A) and starch (S) as a small change to these

parameters yielded the lowest (-5 in the case of 𝜌𝑁,𝑚𝑎𝑥) or the highest (+6 in the case of

𝜙𝑁) sensitivity value.

As per the model formulation and experimental evidence, nitrogen availability plays an

important role in regulating the extent of biomass growth, which is a direct consequence

of carbon (i.e. acetic acid) assimilation; accumulation of starch (a carbon-based

compound), on the other hand, is also more significantly affected by nutrient-limited

conditions which intensify as external nitrogen is consumed. The latter can thus explain

the greater sensitivity of acetic acid and starch dynamics to changes in the parameters

responsible for the uptake of nitrogen.

In particular, the significance of the two parameters 𝜌𝑁,𝑚𝑎𝑥 and 𝜙𝑁 , which display the

greatest sensitivities, can be linked to their association with the maximum nitrogen uptake

rate and its portrayal of the “luxury uptake”, as described in the main text. Whilst the

sensitivity analysis was carried out locally (i.e. changing one parameter at a time), it is

noteworthy to mention that these two parameters are heavily related to each other as the

sensitivity exhibited by one parameter changes dramatically as the other is changed. As

observed in Figure C.2, for instance, the magnitude of the sensitivity exhibited by 𝜙𝑁

increases (and shifts) as 𝜌𝑁,𝑚𝑎𝑥 is increased by 0 %, 2%, and 4%.



131


increase in the N uptake coefficient, 𝝓𝑵, over a 200 h cultivation period, and a 0 %,

2 %, or 4 $ increase in the maximum N uptake rate, 𝝆𝑵,𝒎𝒂𝒙.

With regards to the parameters associated to starch (i.e. 𝑟1, 𝐾𝑠,𝑆, 𝑘𝑖,𝑆, 𝜙𝑆, 𝑘1, 𝑟2, and 𝑛𝑆)

and lipid (i.e. 𝑟3 , 𝐾𝑠,𝐿 , 𝑘𝑖,𝐿 , 𝜙𝐿 , 𝑘2 , 𝑟4 , and 𝑛𝐿 ) dynamics, their computed sensitivities

indicate that the effect of these parameter is only significant to starch (S) or lipid (L)

accumulation, respectively, and also to active biomass (𝑥∗) formation but to a lesser extent.

The effect of these parameters on the remaining model variables is negligible (i.e.

sensitivity=0), which was expected given that X, N, qN, and A are expressed as functions

of each other and are completely independent of S, L, and x*. As per the analysis, two

parameters were deemed to have a negligible effect on the model since their computed

sensitivity was considerably low: the saturation constant for starch formation, 𝐾𝑠,𝑆, and the

regulation coefficient for lipid formation, 𝜙𝐿. Indeed, and as explained in the main text,

the model’s predictive performance remained unchanged by setting both parameters to 0.

The remaining starch- and lipid-associated parameters have a moderate to high effect on

the formation of starch and lipid molecules, indicating they are crucial to maintain the

model’s performance. Particularly, the effect of the inhibition constant, 𝑘𝑖,𝑆 , for starch

formation is observed to be dramatically high (sensitivity as high as +15), which can be

attributed to starch synthesis being highly sensitive to the internal nitrogen concentration,

as reflected by the model which was developed to account for a higher formation of storage

molecules as the internal nitrogen concentration decreases.



132

In a similar manner to the co-relation found between 𝜌𝑁,𝑚𝑎𝑥 and 𝜙𝑁, the synthesis (and

degradation) rates for starch and lipids are heavily correlated between each other as the

change of one reaction rate influences the other. In Figure C.3, for example, the computed

sensitivities for the reaction rate 𝑟1 (starch synthesis) show that the significant effect of this

parameter gradually shifts from starch to lipids as the reaction rate 𝑟3 (lipid synthesis) is

increased. These reaction rates are therefore highly important as they dictate whether starch

or lipid accumulation will be favoured during biomass growth.


increase in the reaction rate (starch), 𝒓𝟏, over a 200 h cultivation period, and a 0 %,

2 %, or 4 % increase in the reaction rate (lipids), 𝒓𝟑.

It is also worth mentioning that the sensitivity computed for most of the starch- and lipid-

associated parameters (with the exception of 𝐾𝑠,𝑆 and 𝐾𝑠,𝐿) increases or decreases without

reaching a steady value, indicating that a small parameter change worsens the model’s

predictive power over time. This can be explained by the model similarly predicting

unsteady starch and lipid concentration profiles. Upon further analysis (see Figure C.6 in

appendix C.4), the dynamics of starch and lipid formation predicted by the developed

model do not reach in certain cases the steady-state and exhibit instead unfeasible (e.g.

negative) concentrations.



133

Finally, the effect of the pH coefficient, 𝐾𝐻 , is only significant to pH itself as the

sensitivities are zero for all the other model variables. As above, this is in line with the

model formulation since none of the variables is expressed in terms of the medium pH.

Appendix C.4. Model analysis.

As shown in the main text, the model presented in this work was shown to be capable of

predicting the nitrogen-limited mixotrophic dynamics of microalgal growth. Its predictive

capacity was additionally showcased (and validated) by carrying out an optimisation study

targeting maximised starch and lipid concentrations. The usefulness of the validated

model, along with the optimal set of kinetic parameters, is exploited here to extract relevant

information of the system dynamics.

The specific growth rate, 𝜇 , which portrays the relationship between growth-limiting

substrates and biomass growth, was extracted from the model over various initial

concentrations of nitrogen, acetic acid, and incident light intensities, and computed at

different times (t = 24, 48, and 72 h) during the cultivation period. Results are shown in

Figure C.4 below.

Figure C.4. Specific growth rate measured at different times, as predicted by the

model, subject to different initial concentrations of nitrogen (No), acetic acid (Ao),

and incident lights (Io).



134

The effects of different initial nitrogen and acetic acid concentrations on the specific

growth rate (Figure C.4) follow inhibition-type kinetics. The inhibitory effect of nitrogen,

however, is more pronounced than that exhibited by acetic acid. The light-dependent

growth rate similarly exhibits inhibition-type kinetics, which is generally the case for

microalgal systems. Nevertheless, since the model developed here was not experimentally

validated against data obtained at different light intensities, the light-dependent specific

growth rates shown in Figure C.4 should only be interpreted qualitatively. In all cases, the

specific growth rate decreases with increasing time, which corresponds to nutrients being

exhausted, or light being attenuated, as the time increases.

The dynamics of the model for biomass growth were also evaluated for three different

conditions: low nitrogen (Low N, where 𝑁0 = 0.335 gN L−1), standard medium (TAP,

where 𝑁0 = 0.382 gN L−1 and 𝐴0 = 0.42 gC L−1), and high acetic acid (High A, where

𝐴0 = 1.26 gC L−1). Results are shown in Figure C.5. As expected, biomass concentration

increases at High A concentrations and decreases at Low N concentrations.

Figure C.5. Predicted dynamics of biomass and associated model equations, for

three different initial conditions: Low N, TAP and High A.

Although the specific growth rate, 𝜇, follows similar dynamics for all three conditions, the

developed expression is capable of reacting to various changes in nutrient concentrations.

This is achieved by incorporating: i) a N-limited growth rate dependent on the nitrogen

quota, and a mixotrophic growth expression dependent on acetic acid and light. As

observed in Figure C.5, the specific growth rate is mainly controlled by the N-limited



135

growth rate: both rates follow the same trend and approach zero (as nitrogen quota

decreases). The positive mixotrophic rate, an additive expression, indicates that acetic acid

has not been exhausted (i.e. A > 0) and/or that light has not been completely attenuated

(i.e. I > 0).

The model developed here exhibited a high predictive capacity for starch and lipid

dynamics. However, when these variables were further evaluated for the three different

conditions employed before (i.e. TAP, Low N, and High A), it was observed that starch

and lipid concentrations may not reach a steady state even after the time of simulation is

increased to 𝑡 = 600 ℎ (Figure C.6). Instead, their concentrations might decrease

continuously to eventually attain negative concentrations (starch in TAP, or lipids in Low

N), or else, increase continuously (starch in Low N, or lipids in TAP), These accumulation

scenarios, despite being outside of the cultivation times explored in this work, are

considered to be unfeasible.

Figure C.6. Predicted dynamics of starch, lipids, and active biomass, for three

different initial conditions: Low N, TAP and High A.



136

As observed in Figure C.6, such unfeasible accumulation scenarios are correlated with

their corresponding accumulation rates (accounting for synthesis and degradation) which

reach a steady state different to zero, preventing storage molecule concentrations from

reaching the desired steady state: negative accumulation rates lead to negative

concentrations, and positive rates lead to a steady increase in concentration. The starch and

lipid accumulation rates thus need to be enhanced to avoid such issues.

References

Bekirogullari, M., Fragkopoulos, I.S., Pittman, J.K., Theodoropoulos, C., 2017. Production

of lipid-based fuels and chemicals from microalgae: An integrated experimental and

model-based optimization study. Algal Res. 23, 78–87.

Bougaran, G., Bernard, O., Sciandra, A., 2010. Modeling continuous cultures of

microalgae colimited by nitrogen and phosphorus. J. Theor. Biol. 265, 443–54.

Cachon, R., Diviés, C., 1994. Generalized model of the effect of pH on lactate fermentation

and citrate bioconversion in Lactococcus lactis ssp. Lactis biovar. diacetylactis. Appl.

Microbiol. Biotechnol. 41, 694–699.

Contois, D.E., 1959. Kinetics of Bacterial Growth: Relationship between Population

Density and Specific Growth Rate of Continuous Cultures. J. Gen. Microbiol. 21, 40–

50.

Droop, M.R., 1968. Vitamin B12 and Marine Ecology. IV. The Kinetics of Uptake, Growth

and Inhibition in Monochrysis Lutheri. J. Mar. Biol. Assoc. United Kingdom 48, 689–

733.

Mairet, F., Bernard, O., Masci, P., Lacour, T., Sciandra, A., 2011. Modelling neutral lipid



Shuler, M.L., Kargi, F., 1992. Bioprocess Engineering: Basic Concepts. Prentice Hall.

137

Chapter 4

Optimisation of Microalgal Starch and Lipid Formation

via Nitrogen and Phosphorus Co-limitation

4.1. Introduction.

The preceding Chapter showcased how the optimisation of nitrogen-limited conditions can

significantly increase starch and lipid formation in mixotrophically-grown algae species,

provided that their specific organic carbon requirements are also optimally supplied. The

studies reviewed in Chapter 2 indicate that starch and lipid accumulation is also

significantly induced as a result of phosphorus limitation, which then widens the possibility

of implementing more robust biofuel-oriented cultivation systems if limitation by both

nitrogen and phosphorus is taken into account during optimisation protocols.

The adequate evaluation of different degrees of nitrogen and phosphorus co-limitation

within a mixotrophic environment can enable the identification of nutrient-enhanced, and

potentially low-cost, cultivation media that favours the sought-after balance between

biomass and storage molecule accumulation. Nevertheless, the simultaneous effects of

these two important macronutrients on mixotrophic algal growth, and most importantly

their positive or negative effects on starch and lipid metabolism have not been intensively

explored.

Although the optimisation of nitrogen and phosphorus co-limitation for microalgal

biofuels production requires a more detailed and complex experimental analysis than that

required for single-nutrient limitation, the work presented previously evidenced the value

of predictive models as fast and reliable optimisation tools. However, and as discussed in

the Literature Review (see Chapter 2), models capable of accounting for the combined

effects of nitrogen and phosphorus on starch and lipid formation by mixotrophic algae

species are scarce or limited in their application.

Chapter 4 – Optimisation of Microalgal Starch and Lipid Formation


138

Therefore, in the paper that follows, the validated multi-parametric kinetic model

developed in Figueroa-Torres et al. (2017) (Contribution 1, Chapter 3) was updated to

additionally portray the dynamics of microalgal growth co-limited by nitrogen and

phosphorus, subject to mixotrophic conditions. To do so, a number of laboratory-scale

cultivation experiments* were again carried out to evaluate the effects of different

nitrogen, phosphorus, and acetic acid concentration regimes on the growing dynamics of

C. reinhardtii. The experimental observations from these new datasets were then employed

to develop the additional kinetic expressions that describe the growth-limiting effects of

phosphorus. Experimental datasets obtained from nitrogen and acetate dependent

experiments were additionally accounted for to preserve the predictiveness of the model

for these two factors.

Furthermore, a deeper analysis of the model’s predictive features (see Supplementary

Information 1, Chapter 3), indicated that as the cultivation period was prolonged the starch

and lipid concentration profiles could attain negative values or increase infinitely. Whilst

such prolonged periods (up to 600 h) greatly deviate from the batch cultivation times

explored in this study (up to 200 h), the observed concentrations scenarios are deemed to

be unfeasible and undermine the model’s predictive capacity. In fact, longer cultivation

times become relevant for strategies involving fed-batch (later addressed in Chapter 5) or

continuous operation where nutrients are not simply supplied once.

As per the analysis of the model dynamics carried out in the preceding Chapter, the

occurrence of such unfeasible starch (S) and lipid (L) concentrations was deemed to be a

consequence of their corresponding accumulation rates (synthesis - degradation) reaching

non-zero steady state values (i.e. dS⁄dt≠0 and dL⁄dt≠0): negative accumulation rates favour

negative concentrations, and positive accumulation rates favour ever-increasing

concentrations. To avoid such unfeasible scenarios, the model developed in Chapter 3

was further refined by incorporating in each accumulation rate a Contois-type saturating

function dependent on either starch or lipids. As will be shown within the following paper,

the saturating function was included within the starch and lipid degradation rates, R2 and

R4, respectively, so that:

𝑅2 =𝑟2

𝑞𝑁∙ 𝑋 is changed to: 𝑅2 = 𝑟2 ∙

𝑋

𝑞𝑁∙

𝑆 𝑋⁄

𝑆 𝑋⁄ +𝑘𝑠𝑎𝑡,𝑆



139

𝑅4 =𝑟4

𝑞𝑁∙ 𝑋 is changed to: 𝑅4 = 𝑟4 ∙

𝑋

𝑞𝑁∙

𝐿 𝑋⁄

𝐿 𝑋⁄ +𝑘𝑠𝑎𝑡,𝐿

The use of the saturating functions shown above, containing two saturating parameters,

allow the starch and lipid accumulation rates to adequately reach a steady-state value of

zero (see the Supplementary Information 2 of this Chapter). More importantly, they

avoid any negative concentration since degradation rates would stop if starch or lipids

reach a value of zero.

The simulation and optimisation value of the resulting kinetic model presented in the paper

that follows is particularly highlighted by: i) the construction of a set of ternary diagrams

displaying the predicted formation of biomass, starch, and lipids when the concentrations

of nitrogen, phosphorus, and acetic acid are set as the degrees of freedom, and ii) the

identification and validation of an improved set of optimal cultivation strategies for

maximal starch and lipid production.

* Note: As already mentioned in the preceding paper, the selected microalgae strain

employed in this research grows mixotrophically in Tris-Acetate-Phosphate (TAP)

medium, where the major carbon, nitrogen, and phosphorous sources originate from acetic

acid, ammonium chloride, and potassium phosphates salts. The detailed preparation of

TAP medium and the concentration of all components are presented in Appendix A.



140



141


Figueroa-Torres GM, Pittman JK, Theodoropoulos C. (2018). Optimisation of Microalgal

Starch and Lipid Formation via Nitrogen and Phosphorus Co-limitation. Submitted to:

Biotechnology and Bioengineering.

Authors’ Contributions:





revised the manuscript.



142

Optimisation of Microalgal Starch and Lipid

Formation via Nitrogen and Phosphorus Co-limitation

Gonzalo M. Figueroa-Torres a, Jon K. Pittman b, Constantinos Theodoropoulos a,*



b School of Earth and Environmental Sciences, The University of Manchester, Manchester, M13

9PL

*Corresponding author:






143

ABSTRACT

Microalgal biomass is regarded as a promising and sustainable feedstock for carbohydrate

and lipid-based fuels due to its ability to accumulate starch and lipid molecules. However,

in order for this promising and renewable biomass to compete with current biofuel

feedstock technologies, optimal algal cultivation systems targeting maximal starch and

lipid formation need to be established. Nitrogen and phosphorus limitation have been the

most validated cultivation strategies for increased starch and lipid formation, but these

strategies must be robust enough to: i) prevent a reduction in overall biomass growth, and

iii) adequately portray the intracellular distribution of the carbon pool, simultaneously

directed to starch and lipid synthesis. In this work, nutrient enhanced cultivation strategies

for maximised starch and lipid formation were successfully established by means of a

multi-parametric predictive kinetic model accounting for mixotrophic algal growth

dynamics co-limited by nitrogen and phosphorus. The model’s predictive capacity was

experimentally validated against datasets obtained from laboratory-scale cultures of

Chlamydomonas reinhardtti CCAP 11/32C subject to various initial nutrient regimes. The

identified model-based optimal cultivation strategies were validated experimentally and

yielded significant increases in starch (+270 %) and lipid (+74 %) production against a

non-optimised strategy.

Key words: modelling, biofuels, starch, lipids, Chlamydomonas



144

1. Introduction.

Due to their biological and renewable nature, biofuels are considered as promising and

sustainable substitutes for fossil-based fuels. The commercialisation of biofuels,

however, has been severely restricted by current feedstock technologies which largely

rely upon the use of traditional food-based (i.e. corn, sugarcane, molasses) and

lignocellulosic biomass (i.e. agricultural and forest residues) (Gouveia, 2011; Köpke and

Dürre, 2011; Nigam and Singh, 2011; Scaife et al., 2015). The on-going search for

sustainable and renewable alternatives for traditional food-based or lignocellulosic

feedstocks has led to the recognition of microalgae as a promising long-term biomass

capable of meeting global biofuel demands (Gouveia, 2011; Scaife et al., 2015; Suganya

et al., 2016).

Microalgae’s potential is highlighted by its ability to intracellularly accumulate high

concentrations of carbohydrates and lipids, which are sugar-based (e.g. bioethanol,

biobutanol) and oil-based (biodiesel) biofuel precursors (Chen et al., 2013; Choix et al.,

2012). Plenty of studies have targeted microalgal oil for its potential for biodiesel

production (Cakmak et al., 2012; Griffiths and Harrison, 2009; Rodolfi et al., 2009;

Sakarika and Kornaros, 2017), but microalgae-based carbohydrates (mainly in the form

of starch) are also attractive because of their lignin-free composition, which facilitates the

process by which carbohydrates are hydrolysed to soluble sugars, i.e. saccharification

(Asada et al., 2012; Brányiková et al., 2010; Chen et al., 2013; Markou et al., 2012).

However, commercialisation of algae-based fuels is unlikely to become a reality unless

large-scale algal cultivation with high biomass productivity becomes a cost-effective

technology.

Nutrient stress (e.g. nitrogen or phosphorous limitation) is demonstrated as a simple,

cost-effective strategy for enhanced starch and lipid formation (Bajhaiya et al., 2016;

Ball et al., 1990; Dragone et al., 2011; Markou et al., 2012; Yao et al., 2013).

Counteracting responses to stress affect the microalgae’s fixation mechanism governing

carbon assimilation and its intracellular distribution (Cade-Menun and Paytan, 2010),

leading to compositional changes favouring starch and/or lipid accumulation. However,

two major nutrient-limited outcomes should be carefully addressed: i) the nutrient-

stressed carbon distribution between starch and lipid is not equally proportional, but



145

greatly dependent on the limiting nutrient and the extent of limitation (Bajhaiya et al.,

2016; Dragone et al., 2011; Markou et al., 2012); ii) nutrient limitation often drastically

reduces algal growth, and consequently limits starch and lipid productivities (Markou et

al., 2012). Mixotrophically grown strains (i.e. those that assimilate organic carbon

sources in addition to inorganic carbon dioxide) have been shown to attain higher growth

rates than typical phototrophic strains (Bekirogullari et al., 2017; Chapman et al., 2015;

Johnson and Alric, 2013), but optimal nutritional requirements balancing the trade-off

between algal growth and starch and lipid formation should be identified.

Implementing starch/lipid-enhancing strategies, which relies on the optimisation of

media composition, can be facilitated by using predictive models reflecting the nutrient-

dependent dynamics of carbon assimilation towards cell growth and its partitioning

between the starch and lipid pools. Plenty of models thus far can predict nutrient-limited

growth (Lee et al., 2015), and more recent models have begun to address starch and lipid

dynamics (Bekirogullari et al., 2017; Klok et al., 2013; Kumar et al., 2016; Mairet et al.,

2011; Packer et al., 2011; Sinha et al., 2017). However, further work is still required to

fully exploit such modelling frameworks and identify optimal nutritional requirements

for fuel-oriented algae cultivation. We previously developed a predictive kinetic model

for nitrogen-limited, mixotrophic algal growth accounting for starch and lipid formation

(Figueroa-Torres et al., 2017). Here, we enhance the model’s predictive capacity by: i)

incorporating phosphorus limitation, thus making the model responsive to nitrogen,

phosphorus, and organic carbon concentrations, and ii) by refining the dynamics of starch

and lipid formation to avoid unfeasible accumulation scenarios. The model was

additionally exploited to establish nutrient-enhanced cultivation strategies maximising

starch and lipid formation.



Experiments were carried out with the wild-type strain Chlamydomonas reinhardtii

CCAP 11/32C. The strain was grown mixotrophically in Tris-Acetate-Phosphate (TAP)

medium (Harris, 1989): 2.42 g of tris-base, 25 mL of TAP salts (15 g L-1 NH4Cl, 4 g L-1

MgSO4.7H2O, 2 g L-1 CaCl2.2H2O), 0.387 mL of phosphate buffer 2.7 M (288 g L-1

K2HPO4, 144 g L-1 KH2PO4), 1 mL of trace components (Hutner et al., 1950), and 1 mL



146

of acetic acid, brought to 1 L with deionised water. For nutrient-dependent experiments

an algal inoculum was propagated in 150 mL of TAP medium until the late stationary

phase (5-7 days), reaching a cell dry weight of 0.001 g mL-1 (5.47x106 cells mL-1). The

inoculum was placed in an orbital shaker at 150 rpm, 25 °C, and illuminated from above

(125 μmol m-2 s-1) in a light/dark photoperiod of 16/8 h.

2.2. Nutrient-dependent cultures.

Mixotrophic growth dynamics co-limited by nitrogen and phosphorus were evaluated by

growing algal cultures under different initial nitrogen (N0), phosphorus (P0), and acetic

acid (A0) concentrations (Table 1) with respect to standard [TAP] medium. Cultures

were grown in duplicate in 500 mL of sterile medium, inoculated with 1 mL of active

algal inoculum, and kept at the environmental conditions described above. Cultures were

fully harvested (sacrificed) during cultivation (days 2, 3, 4, 6, 7, and 8) to analyse

biomass and metabolites. Data was statistically analysed by one-way ANOVA in Origin

Pro 2017 (b9.4.1.354).

During media preparation, the initial nitrogen concentration was altered by modifying the

concentration of ammonium chloride (NH4Cl) in the TAP salts solution. Initial

phosphorus concentration was altered by modifying accordingly the volume of phosphate

buffer (maintaining a 2:1 ratio for K2HPO4:KH2PO4). In phosphorus-limited media,

potassium chloride (KCl) was uniformly added to compensate for the loss of potassium

ions. Initial acetic acid concentration was altered by modifying the volume of acetic acid.

The concentration of all other TAP components remained unchanged, and the initial

medium pH was adjusted to 7 with HCl 3M or KOH 3M, as appropriate.

2.3. Analytical methods.

2.3.1. Cell growth.

The dry cell weight (DCW) was quantified by centrifuging algal cultures for 3.5 min at

3,000 g in an Eppendorf centrifuge 5424. The residual cell pellets were placed in pre-

weighed tubes and allowed to dry for 24 h at 70 °C, after which the DCW was

determined gravimetrically. Dried pellets were kept in sealed containers and analysed for

their lipid content.



147

2.3.2. Starch and lipid contents.

For analysis of microalgal starch, 2 mL aliquot samples of algal cultures were pelleted by

centrifugation at 13,000 g for 3 min. Chlorophyll was removed by washing pelleted cells

in 500 μL of 80% ethanol for 5 min at 85 °C. Washed cells were re-centrifuged at 13,000

g for 3 min, and cellular starch was then solubilised as described in Bajhaiya et al.

(2016). Total starch was quantified as per a Total Starch enzymatic assay kit (Megazyme

International) where released free D-glucose is measured colourimetrically against a D-

glucose standard curve. The lipid content of cells (previously pulverised) was determined

by solvent extraction (using hexane at 155 °C) in a SOXTEC Unit 1043 following a

three-stage extraction protocol as described in Bekirogullari et al. (2017). Extracted

lipids were quantified gravimetrically.

2.3.3. Metabolites concentrations.

Acetic acid was quantified by High Pressure Liquid Chromatography (HPLC) in a HPX-

87H column (8μm, 300x7.7 mm, Bio-Rad), coupled to a UV detector set at 210 nm.

Sulfuric acid (H2SO4) 5 μM was used as the mobile phase at a flow rate of 0.6 mL min-1

and a temperature of 50 °C. Total nitrogen was measured in a Total Organic

Carbon/Total Nitrogen unit (TOC-VCSH/TNM-1 Shimadzu) as per manufacturer’s

instructions. For calibration standards, ammonium chloride (NH4Cl) was used as the

nitrogen source. Phosphorus was measured by Inductively Coupled Plasma – Optical

Emission Spectroscopy (ICP-OES) in a Varian Vista MPX set at 213 nm. All samples

and calibration standards were filtered through 0.45 μm nitrocellulose membranes

(Millipore Ltd.) and diluted accordingly in Type 1 grade water. The nitrogen and

phosphorus cellular quotas were estimated as follows:

𝑞𝑁 =𝑁𝑜−𝑁

𝑋; 𝑞𝑃 =

𝑃𝑜−𝑃

𝑋 (1)

where N0 (gN L-1) and P0 (gPO4 L-1) are the initial nitrogen and phosphorus medium

concentrations, respectively, and N, P, and X are the residual concentrations of nitrogen,

phosphorus, and biomass, respectively (Bougaran et al., 2010).

2.3.4. Active biomass and carbon equivalent concentration.

The fraction of active biomass (i.e. starch and lipid free biomass) was determined by

subtracting starch and lipid concentration from the total biomass (DCW). Acetic acid,



148

starch, lipids, and biomas are reported on a carbon basis by means of conversion factors

(gC g-1): 0.40 acetate, 0.44 starch, 0.77 lipids, and 0.504 biomass. C. reinhardtii cells

were assumed to have the elemental composition reported by Eriksen et al. (2007).

3. Mathematical modelling.

We previously developed a kinetic model capable of predicting the cultivation dynamics

of algal growth, starch, and lipids as a function of the initial nitrogen and acetate

concentrations (Figueroa-Torres et al., 2017). Here, we improved the model’s predictive

capabilities by: i) taking into account the effects of phosphorus concentration on the algal

cultivation dynamics, ii) incorporating the average, rather than local, light intensity

received by the algal culture, and iii) improving the starch and lipid formation rates. The

model state variables include: total biomass (X, gC L-1), starch (S, gC L-1), lipids (L, gC

L-1), active biomass (x*, gC L-1), nitrogen (N, gN L-1), nitrogen quota (qN, gN gC-1),

phosphorus (P, gPO4 L-1), phosphorus quota (qP, gPO4 gC-1), and acetic acid (A, gC L-1).

Total biomass is assumed to be comprised by a functional compartment made up of

active biomass and a storage compartment made up of starch and lipids, as shown in

Figure 1.

3.1. Specific growth rate.

The specific growth rate, 𝜇, is described by a quadruple-factor function incorporating the

combined effects of nitrogen, phosphorus, acetic acid, and light:

𝜇 = ��𝑀,𝑚𝑎𝑥(𝐴, 𝐼) ∙ 𝑚𝑖𝑛[𝜇𝑁(𝑞𝑁), 𝜇𝑃(𝑞𝑃)] (2)

The nitrogen-limited, 𝜇𝑁, and phosphorus-limited, 𝜇𝑃, growth rates are subject to a

minimum law (Cherif and Loreau, 2010), and are each expressed as Droop functions

(Droop, 1968) of the nitrogen quota, 𝑞𝑁, and the phosphorus quota, 𝑞𝑃, respectively:

𝜇𝑁(𝑞𝑁) = 1 −𝑞𝑁,0

𝑞𝑁; 𝜇𝑃(𝑞𝑃) = 1 −

𝑞𝑃,0

𝑞𝑃 (3)

Here, 𝑞𝑁,0 and 𝑞𝑃,0 are the minimum nitrogen and phosphorus quotas required to sustain

growth, respectively. The maximum mixotrophic specific growth rate, ��𝑀,𝑚𝑎𝑥(𝐴, 𝐼), is

regulated by the acetate-driven heterotrophic growth rate, 𝜇𝐻, and the light-driven

phototrophic growth rate, 𝜇𝐼, both described by Andrews functions (Andrews, 1968) to

portray substrate-inhibition and photoinhibition, respectively:



149

��𝑀,𝑚𝑎𝑥(𝐴, 𝐼) = 𝜇𝑚𝑎𝑥 ∙ [𝑤𝐻 ∙ 𝜇𝐻(𝐴) + 𝑤𝐼 ∙ 𝜇𝐼(𝐼)] (4)

𝜇𝐻(𝐴) =𝐴

𝐴 + 𝐾𝑆,𝐴 + 𝐴2 𝐾𝑖,𝐴⁄; 𝜇𝐼(𝐼) =

𝐼

𝐼 + 𝐾𝑆,𝐼 + 𝐼2 𝐾𝑖,𝐼⁄ (5)

Here, 𝐾𝑆,𝐴 and 𝐾𝑖,𝐴 are the acetate-associated half-saturation and inhibition constants,

respectively, and 𝐾𝑆,𝐼 and 𝐾𝑖,𝐿 are the light-associated half-saturation and inhibition

constants, respectively; 𝑤𝐻 and 𝑤𝐼 are weighing functions controlling the magnitude of

the heterotrophic and phototrophic growth rates, respectively (Figueroa-Torres et al.,

2017). The light, I, received by the culture of a given depth, z, is often described by the

Beer-Lambert law, which assumes an exponential decrease in light with increasing

biomass growth:

𝐼(𝑧) = 𝐼0 ∙ 𝑒−𝜎∙𝑋∙𝑧 (6)

where 𝐼0 is the incident light intensity and 𝜎 is a light attenuation coefficient. However, a

more accurate representation of the light received by the culture throughout the vessel is

obtained by computing an average light intensity between the surface (𝑧 = 0), and the

total depth (𝑧 = 𝐿) of the vessel:

𝐼 =𝐼𝑜

𝐿∫ 𝑒−𝜎∙𝑋∙𝑧 ∙ 𝑑𝑧 =

𝐼0

𝜆∙ (1 − 𝑒−𝜆)

𝐿

0

(7)

where 𝜆 = 𝜎 ∙ 𝑋 ∙ 𝐿 is the optical depth. It should be noted that the optical depth is often

further improved by considering that light attenuation depends not only on biomass

growth, but also on the concentration of chlorophyll and other pigments (Bernard, 2011).

3.2. Nitrogen and phosphorus uptake rates.

The nitrogen uptake rate, 𝜌𝑁, incorporates inhibition-type kinetics, as per the Andrews

model (Andrews, 1968), dependent on the nitrogen, 𝑁, and acetate, 𝐴, medium

concentrations. The inhibition terms were incorporated given that high concentrations of

nitrogen and acetic acid were observed to be inhibitory for nitrogen uptake (which

regulates biomass growth):

𝜌𝑁 = ��𝑁,𝑚𝑎𝑥(𝑁0, 𝑋) ∙𝑁

𝑁 + 𝑘𝑠,𝑁 + 𝑁2 𝑘𝑖,𝑁⁄∙

𝐴

𝐴 + 𝑘𝑠,𝐴:𝑁 + 𝐴2 𝑘𝑖,𝐴:𝑁⁄∙ 𝑓(𝑞𝑃) (8)

Here, 𝑘𝑠,𝑁 and 𝑘𝑖,𝑁 are nitrogen-associated half-saturation and inhibition constants,

respectively, and 𝑘𝑠,𝐴:𝑁 and 𝑘𝑖,𝐴:𝑁 are acetate-associated half-saturation and inhibition



150

constants, respectively. In Eq. (8), ��𝑁,𝑚𝑎𝑥(𝑁𝑜, 𝑋) is the maximum nitrogen uptake rate,

which accounts for the luxury uptake of nitrogen of microalgal cells (i.e. a phenomenon

where the uptake of nutrient is fast immediately after inoculation). Given that the extent

of luxury uptake was thought to be dependent on the nutrient concentration of the “fresh”

medium and the cell density (Droop, 1983), the maximum nitrogen uptake rate is

regulated by the initial nitrogen medium concentration, 𝑁0, and the biomass

concentration, 𝑋, as:


𝑛


𝑛 ∙ 𝑒−𝜙𝑁∙𝑋 (9)

where 𝜙𝑁 is an uptake regulation coefficient, 𝑛 is a shape-controlling parameter, and 𝐾∗

is a saturation constant. In Eq. (9), the effect of the initial nitrogen concentration is

described as per saturation-type kinetics, whereas the effect of biomass is expressed by

an exponential term indicating that the uptake of nitrogen decreases exponentially with

increasing biomass concentration.

The above formulation follows the structure proposed in our previous work. However,

experimental data (Figure 2.b and Figure 3.b) suggested that the uptake of nitrogen was

greatly affected by phosphorus limitation: the consumption of nitrogen decreased in

those cultures grown in low phosphorous concentrations (P-limitation). The negative

effect of phosphorus limitation on the cellular mechanisms controlling nitrogen uptake

has been previously reported, and is explained by a shortage of nutrient transport energy

supplied by phosphorus-containing molecules such as ATP (Bougaran et al., 2010). In

order to replicate this scenario, the nitrogen uptake rate (Eq. 8) was thus additionally

regulated by a Droop function of the phosphorus quota, 𝑓(𝑞𝑃), so that nitrogen uptake

decreases under phosphorus-limited conditions (i.e. low P quotas):

𝑓(𝑞𝑃) = (1 −𝐾𝑃

𝑞𝑃) (10)

Here, 𝐾𝑃 denotes the minimum P quota below which nitrogen uptake stops: (i.e. if 𝑞𝑃 <

𝐾𝑞𝑃, 𝜌𝑁 = 0).

The uptake of phosphorus (unlike nitrogen), was not affected by acetic acid, and was thus

solely expressed in terms of the residual phosphate concentration, P, by means of

inhibition-type kinetics, as in:



151

𝜌𝑃 = 𝜌𝑃,max ∙𝑃

𝑃 + 𝑘𝑠,𝑃 + 𝑃2 𝑘𝑖,𝑃⁄∙ 𝑓(𝑞𝑁) (11)

Here, 𝜌𝑃,max is the maximum phosphorous uptake rate, and 𝑘𝑠,𝑃 and 𝑘𝑖,𝑃 are the

phosphorus-associated half-saturation and inhibition constants, respectively. In Eq. 11,

𝑓(𝑞𝑁), is a regulating function dependent on the N quota which accounts for the negative

effects of nitrogen stress on phosphorus uptake, described as:

𝑓(𝑞𝑁) = [1 + (𝜌𝑃,𝑚𝑎𝑥

𝑞𝑁)

2

]

−1

(12)

This function regulates phosphorus uptake as follows: the uptake of phosphorus

decreases as the nitrogen quota decreases (i.e. nitrogen-limited conditions). It should be

noted that the regulating function shown in Eq. (12) is an inhibitory function which

differs from that in Eq. (10) (which requires an additional parameter: KP) given that the

negative effects of N-limitation were observed to be less pronounced (Figure .d, Figure

.d) than those of P-stress on nitrogen uptake.

3.3. Formation of starch and lipids.

The dynamics of starch and lipids are regulated by their synthetic rates, 𝑅1 and 𝑅3, and

their degradation rates, 𝑅2 and 𝑅4, respectively, as in Figure 1. The synthetic rates (Eq.

(13) and Eq. (14)) are dependent on: i) the internal nitrogen concentration, i.e. 𝑁𝑖 = 𝑞𝑁 ∙

𝑋, and ii) the bioavailable carbon concentration, i.e. 𝐴𝑖 = 𝐴 − 𝐴0.

𝑅1 = 𝑟1 ∙

𝑁𝑖𝑛𝑠

𝑁𝑖𝑛𝑠 + 𝑘𝑠,𝑆

𝑛𝑆 + (𝑁𝑖2 𝑘𝑖,𝑆⁄ )

𝑛𝑠∙

𝑘1

𝑘1 + 𝑁 𝑁𝑜⁄∙ [1 +

1

𝜇∙ 𝑒𝜙𝑆∗𝐴𝑖]

∙ 𝜇 ∙ 𝑥∗

(13)

𝑅3 = 𝑟3 ∙

𝑁𝑖𝑛𝐿

𝑁𝑖𝑛𝐿 + 𝑘𝑠,𝐿


𝑛𝐿∙

𝑘2

𝑘2 + 𝑁 𝑁0⁄∙ [1 +

1

𝜇∙ 𝑒𝜙𝐿∗𝐴𝑖]

∙ 𝜇 ∙ 𝑥∗

(14)

Here, 𝑟1 and 𝑟3 are the rate constants for starch and lipid synthesis, respectively; 𝑘𝑠,𝑆 and

𝑘𝑠,𝐿 are saturation constants; 𝑘𝑖,𝑆 and 𝑘𝑖,𝐿 are inhibition constants; 𝑛𝑆 and 𝑛𝐿 are shape-

controlling parameters; 𝜙𝑆 and 𝜙𝐿 are regulation coefficients; and 𝑘1 and 𝑘2 are

constants regulating starch and lipid formation with respect to nitrogen consumption.

Meanwhile, starch and lipid degradation rates are described by:



152

𝑅2 = 𝑟2 ∙𝑋

𝑞𝑁∙

𝑆 𝑋⁄

𝑆 𝑋⁄ + 𝑘𝑠𝑎𝑡,𝑆 (15)

𝑅4 = 𝑟4 ∙𝑋

𝑞𝑁∙

𝐿 𝑋⁄

𝐿 𝑋⁄ + 𝑘𝑠𝑎𝑡,𝐿 (16)

Here, 𝑟2 and 𝑟4 are the rate constants for starch and lipid degradation, respectively; and

𝑘𝑠𝑎𝑡,𝑆 and 𝑘𝑠𝑎𝑡,𝐿 are saturation constants that control the extent of degradation and avoid

unfeasible accumulation scenarios suffered by our previous model (see Supplementary

Information). The saturation-type functions incorporated in Eq. (15) and Eq. (16) above

follow the formulation proposed by Contois (Contois, 1959).

3.4. Time-dependent equations.

The accumulation rates of the carbon-based cell components (i.e. biomass, starch, lipids,

and active biomass) are described by the following set of ordinary differential equations:

𝑑𝑋

𝑑𝑡= 𝜇 ∙ 𝑋 (17)

𝑑𝑆

𝑑𝑡= 𝑅1 − 𝑅2 (18)

𝑑𝐿

𝑑𝑡= 𝑅3 − 𝑅4 (19)

𝑑𝑥∗

𝑑𝑡=

𝑑𝑋

𝑑𝑡− (

𝑑𝑆

𝑑𝑡+

𝑑𝐿

𝑑𝑡) (20)

The extracellular and intracellular (i.e. cell quotas) nutrient dynamics are described by:

𝑑𝑁

𝑑𝑡= −𝜌𝑁 ∙ 𝑋 (21)

𝑑𝑞𝑁

𝑑𝑡= 𝜌𝑁 − 𝜇 ∙ 𝑞𝑁 (22)

𝑑𝑃

𝑑𝑡= −𝜌𝑃 ∙ 𝑋 (23)

𝑑𝑞𝑃

𝑑𝑡= 𝜌𝑃 − 𝜇 ∙ 𝑞𝑃 (24)

𝑑𝐴

𝑑𝑡= −

1

𝑌𝑋/𝐴∙

𝜇𝐻

𝜇𝐻 + 𝜇𝐼∙

𝑑𝑋

𝑑𝑡 (25)



153

3.5. Estimation of model parameters.

The model presented here (Eq. (17) – Eq. (25)) is comprised by 37 kinetic parameters, all

appropriately defined in Table 2. The values of 12 kinetic parameters (associated to

growth and nitrogen uptake dynamics) were set equivalent to those previously identified

in Figueroa-Torres et al. (2017). The remaining 25 kinetic parameters were estimated by

minimising the squared relative error between experimental and predicted data:

min 𝐺(𝑃) = ∑ ∑ ∑ (𝑍ℎ,𝑖,𝑘

𝑃𝑟𝑒𝑑(𝑃) − 𝑍ℎ,𝑖,𝑘𝐸𝑥𝑝

𝑍ℎ,𝑖,𝑘𝐸𝑥𝑝 )

2𝑛𝑘

𝑘=1

𝑛𝑖

𝑖=1

𝑛ℎ

ℎ=1

(26)

Here, 𝐺 is the objective function, P is the parameter set, and Z is the set of predicted or

experimental data. Predicted data was generated by solving the model using initial values

equivalent to those of nutrient-dependent experiments. nh, ni, and nk denote the number

of data points in time, number of fitting experimental datasets, and number of state

variables, respectively. Parameters were restricted by lower (lb) and upper (ub) bounds as

per data obtained from literature or experimental analysis. It should be noted that from

the 25 estimated kinetic parameters, 19 had already been previously identified in

Figueroa-Torres et al. (2017), but their original values were refined here to account for

the effects of phosphorus (Table 2). The minimisation problem was solved via a

stochastic optimisation routine (simulated annealing) subject to multiple re-starts to

approximate the solution around a global minimum. The stochastic solution was then

used as initial guess in a deterministic routine (sequential quadratic programming) to

generate the final parameter set (Vlysidis et al., 2011). Both routines were coded in-

house in Matlab R2015a.

4. Results and discussion.

4.1. Biomass, starch and lipid responses to different media composition.

In order to evaluate the effects of initial nutrient availability on algal growth and starch

and lipid accumulation, cultures were grown under different nitrogen, phosphorus, and

acetic acid concentration regimes until they attained stationary phase (8 days). The

culture grown in standard [TAP] medium was selected as the reference case against

which all nutrient-dependent cultures were statistically compared. Results are

summarised in Table 3. [TAP] medium composition yielded a biomass concentration of

0.318 gC L-1, consisting of 5.6 % starch and 14.1 % lipid. When compared to [TAP], all



154

nutrient-limited conditions caused reductions in biomass concentration of -11 % [Low

N], -16% [Low P], and -18 % [Low A]. However, the only significant reduction in

biomass concentration associated to nutrient limitation was observed in the culture grown

under [Low N:Low P] conditions (p=0.048, one-way ANOVA), where biomass

concentration dropped -22 % with respect to [TAP]. Increasing acetic acid concentration,

on the other hand, significantly increased biomass concentrations up to 23 % in [High A]

(p=0.043) and 30% in [High A+] (p=0.009). Acetate-associated induced growth in C.

reinhardtii has been previously described as a consequence of enhanced mixotrophic

growing conditions (Bekirogullari et al., 2017; Chapman et al., 2015). High acetic acid

concentration subject to low phosphorus [High A : Low P] similarly supported higher

biomass concentrations with respect to [TAP], unlike those subject to low nitrogen [High

A : Low N] where biomass decreased significantly (p=0.026), which may indicate

nitrogen plays a more important role in sustaining algal growth.

Despite the adverse effects of nitrogen stress on algal growth, this strategy increased

starch and lipid contents up to 16.8% and 21.2 %, respectively, observed in the [Low N]

culture (Table 3). The increased accumulation of storage molecules translated into a

significant increase of starch and lipid concentrations. On the other hand, only one of the

phosphorus limited scenarios, [Low P], significantly induced starch accumulation up to

11.3 % which correlated with a significant increase (p<0.006) of the starch medium

concentration. Lipid formation, however, was not significantly different under any of the

two phosphorus limited scenarios. The culture grown in [Low N : Low P] conditions

exhibited a similar behaviour, where only starch concentrations increased significantly

(p<0.001) with respect to [TAP]. Accumulation of starch rather than lipid molecules

during phosphorus limitation could be explained simply by starch synthesis being the

preferred product of carbon assimilation in C. reinhardtii (Fan et al., 2012), or by the

phosphate-associated inhibition of the enzyme ADP-glucose pyrophosphorylase which

regulates starch synthetic pathways (Gomez-Casati et al., 2003; Heldt et al., 1977). It is

worth noting that greater degrees of phosphorus and nitrogen limitation than those

employed in this study have previously been shown to induce higher lipid contents in C.

reinhardtii (Bajhaiya et al., 2016). However, these scenarios were not explored since

such extreme starvation substantially reduced biomass concentration (data not shown)

below levels required for adequate lipid quantification via SOXTEC extraction.



155

Although C. reinhardtii cultures subject to high acetic acid concentrations attained higher

starch and lipid concentrations with respect to [TAP], such increases were mainly

associated to the higher biomass supported by the acetate boost, with the exception of the

[High A: Low N-] culture which accumulated significantly more starch and lipid than

[TAP] due to the combined effect of the acetate boost with nitrogen stress. Increased

lipid concentrations during acetate-enhanced cultivation was similarly reported by

Bekirogullari et al. (2017) for the wild-type C. reinhardtii, and considerably higher

accumulation of TAG has been similarly observed in the starch-less (sta6) strain when

subject to an acetate boost and nitrogen limitation (Goodenough et al., 2014; Goodson et

al., 2011).

Extreme high nutrient concentrations ([HIGH N], [HIGH P], and [HIGH A]) inhibited

biomass growth and yielded no increases in starch and lipid concentration (Table 3),

indicating such strategies are inappropriate for large-scale algal cultivation. Although

nitrogen and phosphorus limitation have been the most extensively proven algal

cultivation strategies for starch and lipid accumulation, studies have mainly evaluated

such strategies under either complete starvation or single nutrient limitation (Ball et al.,

1990; Markou et al., 2012; Philipps et al., 2012; Xin et al., 2010). Few works have

explored storage molecules accumulation under multiple stresses such as nutrient co-

limitation (Bajhaiya et al., 2016; Dragone et al., 2011), characterised by a trade-off

between biomass growth and starch and lipid contents. The model proposed in this work

was thus employed to explore the full effect of nutrient composition on algal dynamics

and identify starch and lipid enhancing strategies.

4.2. Modelling mixotrophic algal growth co-limited by nitrogen and phosphorus.

The multi-parametric model proposed here was evaluated in terms of its capacity to

predict algal growth dynamics subject to different nitrogen, phosphorus, and acetic acid

concentrations. The values of the estimated parameters, as per the fitting methodology

shown in Section 3.5, are shown in Table 2. The predicted algal growth dynamics of the

cultivation scenarios employed for parameter fitting ([TAP], [Low P], and [High A:Low

P]) were observed to be in good agreement with the corresponding experimental datasets

(Figure 2). The model’s predicted dynamics were further validated against different

cultivation scenarios ([Low N:Low P], [Med N], and [High A+]), indicating the model



156

adequately portrayed growth, nutrient uptake, and starch and lipid formation in C.

reinhardtii (Figure 3).

The level of agreement between experimental and predicted data for each model variable

is additionally displayed in the parity plots shown in Figure 4. The computed mean

correlation coefficient (r2) for both fitting and validating datasets averaged r2=0.95,

highlighting the model’s high predictive capacity. The model was thus exploited to

compute the formation of biomass, starch, and lipids at the 8th day (t=192 h) of

cultivation, subject to various initial nitrogen (0.25 – 0.75 gN L-1), phosphorus (0 – 0.14

gPO4 L-1), and acetic acid (0 – 3.5 gC L-1) concentrations. The results are presented as

three individual ternary diagrams (Figure 5), each showing predicted biomass, starch,

and lipids (model outputs) in response to initial nutrient concentrations (model inputs).

The ternary diagrams show the corresponding changes in starch and lipid formation when

subject to nitrogen and phosphorus co-limitation, and allow to identify the required level

of nutrient limitation to maximise starch and lipid formation during mixotrophic growth.

4.3. Model-based cultivation strategies for enhanced starch and lipid formation.

The ternary diagrams were employed to identify the optimal nutritional requirements (i.e.

nitrogen, phosphorus, and acetic acid) maximising starch and lipid concentrations,

identified as: i) “starch-enhancing” medium: [No=0.330 gN L-1, P0=0.052 gPO4 L-1,

Ao=0.96 gC L-1], yielding 0.33 gC L-1 biomass with 21 % starch and 22 % lipids, and ii)

“lipid-enhancing” medium [No=0.365 gN L-1, P0=0.041 gPO4 L-1, Ao=1.00 gC L-1],

yielding 0.38 gC L-1 biomass with 15 % starch and 21 % lipids. The predicted outcome of

the optimised scenarios was additionally verified by growing two lab-scale cultures of C.

reinhardtii subject to the above optimal medium compositions. As observed in Figure 6,

both of the model-based optimal cultivation scenarios agreed well with the corresponding

experimental data. Compared to [TAP] medium, starch-enhancing conditions yielded

increases of 270 % and 56 % in starch and lipid concentrations, respectively, whereas

lipid-enhancing conditions yielded increases of 203 % and 74 % in starch and lipid

concentrations, respectively. In line with these optimal scenarios, co-limitation by

nitrogen and phosphorus can significantly induce starch and lipid formation, but provided

that reduced growth rates are overcome via the supply of sufficient acetic acid. Although

from an economic perspective the organic carbon requirements (e.g. acetate) restrict



157

mixotrophic cultivation, such an issue could be avoided by adequately integrating

wastewater effluents rich in organic matter with microalgal growth (Adeniyi et al., 2018;

Zhan et al., 2017). The validated optimal nutrient compositions identified here, however,

offer a promising and sustainable outlook for the scaling-up of biofuel-oriented algal

cultivation systems where the supply of nitrogen and phosphorus can be managed

efficiently whilst simultaneoulsy reducing the environmental impacts of nitrogen

fertilisers or the overuse of inorganic phosphorus, a non-renewable resource (Usher et al.,

2014).

5. Conclusions.

A multi-parametric kinetic model for mixotrophic algal growth developed to predict

starch and lipid formation via nitrogen and phosphorus co-limitation. The model’s

predictiveness was validated against different nutrient-dependent cultivation scenarios,

and model-based starch-enhancing and lipid-enhancing cultivation strategies were

subsequently established by identifying optimal nutrient compositions. The optimised

strategies were validated experimentally and yielded increases of 270 % starch and 74 %

lipids, compared to non-optimised cultivation conditions. The cultivation strategies

maximising starch and lipids highlight the benefits of exploiting modelling frameworks

as optimisation tools for the development of fuel-oriented algal cultivation.

Acknowledgements.

Gonzalo M. Figueroa-Torres kindly acknowledges the Mexican National Council of

Science and Technology (CONACyT) for its financial support.



158

References.

Adeniyi OM, Azimov U, Burluka A. 2018. Algae biofuel: Current status and future

applications. Renew. Sustain. Energy Rev. 90:316–335.

Andrews JF. 1968. A mathematical model for the continuous culture of microorganisms

utilizing inhibitory substrates. Biotechnol. Bioeng. 10:707–723.

Asada C, Doi K, Sasaki C, Nakamura Y. 2012. Efficient Extraction of Starch from

Microalgae Using Ultrasonic Homogenizer and Its Conversion into Ethanol by

Simultaneous Saccharification and Fermentation. Earth Environ. Sci. 3:175–179.

Bajhaiya AK, Dean AP, Driver T, Trivedi DK, Rattray NJW, Allwood JW, Goodacre R,

Pittman JK. 2016. High-throughput metabolic screening of microalgae genetic

variation in response to nutrient limitation. Metabolomics 12:9.

Ball SG, Dirick L, Decq A, Martiat J-C, Matagne R. 1990. Physiology of starch storage

in the monocellular alga Chlamydomonas reinhardtii. Plant Sci. 66:1–9.

Bekirogullari M, Fragkopoulos IS, Pittman JK, Theodoropoulos C. 2017. Production of

lipid-based fuels and chemicals from microalgae: An integrated experimental and

model-based optimization study. Algal Res. 23:78–87.

Bernard O. 2011. Hurdles and challenges for modelling and control of microalgae for

CO2 mitigation and biofuel production. J. Process Control 21:1378–1389.

Bougaran G, Bernard O, Sciandra A. 2010. Modeling continuous cultures of microalgae

colimited by nitrogen and phosphorus. J. Theor. Biol. 265:443–54.

Brányiková I, Maršálková B, Doucha J, Brányik T, Bišová K, Zachleder V, Vítová M.

2010. Microalgae—novel highly efficient starch producers. Biotechnol. Bioeng.

108:766–776.

Cade-Menun BJ, Paytan A. 2010. Nutrient temperature and light stress alter phosphorus

and carbon forms in culture-grown algae. Mar. Chem. 121:27–36.

Cakmak T, Angun P, Demiray YE, Ozkan AD, Elibol Z, Tekinay T. 2012. Differential

effects of nitrogen and sulfur deprivation on growth and biodiesel feedstock

production of Chlamydomonas reinhardtii. Biotechnol. Bioeng. 109:1947–1957.

Chapman SP, Paget CM, Johnson GN, Schwartz J-M. 2015. Flux balance analysis

reveals acetate metabolism modulates cyclic electron flow and alternative glycolytic

pathways in Chlamydomonas reinhardtii. Front. Plant Sci. 6:474.

Chen C-Y, Zhao X-Q, Yen H-W, Ho S-H, Cheng C-L, Lee D-J, Bai F-W, Chang J-S.

2013. Microalgae-based carbohydrates for biofuel production. Biochem. Eng. J.

78:1–10.

Cherif M, Loreau M. 2010. Towards a more biologically realistic use of Droop’s

equations to model growth under multiple nutrient limitation. Oikos 119:897–907.

Choix FJ, De-Bashan LE, Bashan Y. 2012. Enhanced accumulation of starch and total



159

carbohydrates in alginate-immobilized Chlorella spp. induced by Azospirillum

brasilense: I. Autotrophic conditions. Enzyme Microb. Technol. 51:294–9.

Contois DE. 1959. Kinetics of Bacterial Growth: Relationship between Population

Density and Specific Growth Rate of Continuous Cultures. J. Gen. Microbiol.

21:40–50.

Dragone G, Fernandes BD, Abreu AP, Vicente A a., Teixeira J a. 2011. Nutrient


Energy 88:3331–3335.

Droop MR. 1968. Vitamin B12 and Marine Ecology. IV. The Kinetics of Uptake,


Kingdom 48:689–733.

Droop MR. 1983. 25 Years of Algal Growth Kinetics A Personal View. Bot. Mar. 26:99.

Eriksen N, Riisgård F, Gunther W, Lønsmann Iversen J. 2007. On-line estimation of O2

production, CO2 uptake, and growth kinetics of microalgal cultures in a gas-tight

photobioreactor. J. Appl. Phycol. 19:161–174.

Fan J, Yan C, Andre C, Shanklin J, Schwender J, Xu C. 2012. Oil accumulation is


reinhardtii. Plant Cell Physiol. 53:1380–1390.

Figueroa-Torres GM, Pittman JK, Theodoropoulos C. 2017. Kinetic modelling of starch

and lipid formation during mixotrophic, nutrient-limited microalgal growth.

Bioresour. Technol. 241:868–878.

Gomez-Casati DF, Cortassa S, Aon MA, Iglesias AA. 2003. Ultrasensitive behavior in

the synthesis of storage polysaccharides in cyanobacteria. Planta 216:969–975.

Goodenough U, Blaby I, Casero D, Gallaher SD, Goodson C, Johnson S, Lee J-H,

Merchant SS, Pellegrini M, Roth R, Rusch J, Singh M, Umen JG, Weiss TL, Wulan

T. 2014. The Path to Triacylglyceride Obesity in the sta6 Strain of Chlamydomonas

reinhardtii. Eukaryot. Cell 13:591–613.

Goodson C, Roth R, Wang ZT, Goodenough U. 2011. Structural Correlates of

Cytoplasmic and Chloroplast Lipid Body Synthesis in Chlamydomonas reinhardtii

and Stimulation of Lipid Body Production with Acetate Boost. Eukaryot. Cell

10:1592–1606.

Gouveia L. 2011. Microalgae as a Feedstock for Biofuels. In: . Microalgae as a Feed.

Biofuels SE - 1. Springer Berlin Heidelberg. SpringerBriefs in Microbiology, pp. 1–

69.

Griffiths M, Harrison SL. 2009. Lipid productivity as a key characteristic for choosing

algal species for biodiesel production. J. Appl. Phycol. 21:493–507.

Harris EH. 1989. The Chlamydomonas sourcebook : a comprehensive guide to biology

and laboratory use. Academic Press 780 p.



160

Heldt HW, Chon CJ, Maronde D, Herold A, Stankovic ZS, Walker DA, Kraminer A,

Kirk MR, Heber U. 1977. Role of Orthophosphate and Other Factors in the

Regulation of Starch Formation in Leaves and Isolated Chloroplasts. Plant Physiol.

59:1146 LP-1155.

Hutner SH, Provasoli L, Schatz A, Haskins CP. 1950. Some Approaches to the Study of

the Role of Metals in the Metabolism of Microorganisms. Proc. Am. Philos. Soc.

94:152–170.

Johnson X, Alric J. 2013. Central Carbon Metabolism and Electron Transport in

Chlamydomonas reinhardtii: Metabolic Constraints for Carbon Partitioning between

Oil and Starch. Eukaryot. Cell 12:776–793.

Klok AJ, Verbaanderd JA, Lamers PP, Martens DE, Rinzema A, Wijffels RH. 2013. A

model for customising biomass composition in continuous microalgae production.


Köpke M, Dürre P. 2011. Handbook of biofuels production. Handb. Biofuels Prod.

Elsevier 221-257 p.

Kumar A, Guria C, Chitres G, Chakraborty A, Pathak AK. 2016. Modelling of



bubble column. Bioresour. Technol. 218:1021–1036.

Lee E, Jalalizadeh M, Zhang Q. 2015. Growth kinetic models for microalgae cultivation:

A review. Algal Res. 12:497–512.

Mairet F, Bernard O, Masci P, Lacour T, Sciandra A. 2011. Modelling neutral lipid



Markou G, Angelidaki I, Georgakakis D. 2012. Microalgal carbohydrates: an overview

of the factors influencing carbohydrates production, and of main bioconversion

technologies for production of biofuels. Appl. Microbiol. Biotechnol. 96:631–645.

Nigam PS, Singh A. 2011. Production of liquid biofuels from renewable resources. Prog.

Energy Combust. Sci. 37:52–68.

Packer A, Li Y, Andersen T, Hu Q, Kuang Y, Sommerfeld M. 2011. Growth and neutral

lipid synthesis in green microalgae: a mathematical model. Bioresour. Technol.

102:111–7.

Philipps G, Happe T, Hemschemeier A. 2012. Nitrogen deprivation results in

photosynthetic hydrogen production in Chlamydomonas reinhardtii. Planta

235:729–745.

Rodolfi L, Chini Zittelli G, Bassi N, Padovani G, Biondi N, Bonini G, Tredici MR. 2009.

Microalgae for oil: strain selection, induction of lipid synthesis and outdoor mass

cultivation in a low-cost photobioreactor. Biotechnol. Bioeng. 102:100–12.

Sakarika M, Kornaros M. 2017. Kinetics of growth and lipids accumulation in Chlorella



161

vulgaris during batch heterotrophic cultivation: Effect of different nutrient limitation

strategies. Bioresour. Technol. 243:356–365.

Scaife MA, Merkx-Jacques A, Woodhall DL, Armenta RE. 2015. Algal biofuels in

Canada: Status and potential. Renew. Sustain. Energy Rev. 44:620–642.

Sinha SK, Kumar M, Guria C, Kumar A, Banerjee C. 2017. Biokinetic model-based

multi-objective optimization of Dunaliella tertiolecta cultivation using elitist non-

dominated sorting genetic algorithm with inheritance. Bioresour. Technol. 242:206–

217.

Suganya T, Varman M, Masjuki HH, Renganathan S. 2016. Macroalgae and microalgae

as a potential source for commercial applications along with biofuels production: A

biorefinery approach. Renew. Sustain. Energy Rev. 55:909–941.

Usher PK, Ross AB, Camargo-Valero MA, Tomlin AS, Gale WF. 2014. An overview of

the potential environmental impacts of large-scale microalgae cultivation. Biofuels

5:331–349.

Vlysidis A, Binns M, Webb C, Theodoropoulos C. 2011. Glycerol utilisation for the

production of chemicals: Conversion to succinic acid, a combined experimental and

computational study. Biochem. Eng. J. 58–59:1–11.

Xin L, Hong-ying H, Ke G, Ying-xue S. 2010. Effects of different nitrogen and

phosphorus concentrations on the growth, nutrient uptake, and lipid accumulation of

a freshwater microalga Scenedesmus sp. Bioresour. Technol. 101:5494–5500.

Yao C-H, Ai J-N, Cao X-P, Xue S. 2013. Characterization of cell growth and starch

production in the marine green microalga Tetraselmis subcordiformis under

extracellular phosphorus-deprived and sequentially phosphorus-replete conditions.

Appl. Microbiol. Biotechnol. 97:6099–6110.

Zhan J, Rong J, Wang Q. 2017. Mixotrophic cultivation, a preferable microalgae

cultivation mode for biomass/bioenergy production, and bioremediation, advances

and prospect. Int. J. Hydrogen Energy 42:8505–8517.



162

TABLES

Table 1. Initial nutrient concentrations employed during nutrient-dependent

cultures.

Treatment Label Nitrogen a Phosphorus Acetic acid

gN L-1 (gN L-1) gPO4 L-1 gC L-1

TAP b 0.382 (0.098) 0.096 0.42

[Low N : Low P] 0.335 (0.042) 0.0096 0.42

[Low N] 0.335 (0.042) 0.0960 0.42

[Med N] 0.356 (0.070) 0.0960 0.42

[Low P] 0.382 (0.098) 0.0096 0.42

[Med P] 0.382 (0.098) 0.0480 0.42

[Low A] 0.382 (0.098) 0.0960 0.21

[High A] 0.382 (0.098) 0.0960 0.75

[High A +] 0.382 (0.098) 0.0960 1.26

[High A : Low N-] 0.315 (0.032) 0.0960 1.26

[High A : Low P] 0.382 (0.098) 0.0096 1.26

[HIGH N] 0.742 (0.450) 0.0960 0.42

[HIGH P] 0.382 (0.098) 0.3860 0.42

[HIGH A] 0.382 (0.098) 0.0960 2.52 a First column refers to total nitrogen concentration; second column (in

parenthesis) refers only to the nitrogen concentration from NH4Cl. b Initial nutrient concentrations in standard TAP medium.



163

Table 2. List of kinetic parameters employed in the proposed model for the

mixotrophic growth of C. reinhardtii co-limited by nitrogen and phosphorus.

Type Symbol Parameter description Value Units Reference A

sso

ciat

ed

to g

row

th

µmax Maximum specific growth rate 0.106 h-1 Figueroa-Torres et al.

(2017)

qN,0 Minimum nitrogen quota 0.877 gN gC-1 Figueroa-Torres et al.

(2017)

qP,0 Minimum phosphorus quota 0.016 gPO4 gC-1 This work

Ks,A Acetate saturation constant 1.789 gC L-1 Figueroa-Torres et al.

(2017)

ki,A Acetate inhibition constant 0.110 gC L-1 Figueroa-Torres et al.

(2017)

Ks,I Light saturation constant 1.4 µmol m-2s-1 Mairet et al. (2011)

ki,I Light inhibition constant 186.5 µmol m-2s-1 Figueroa-Torres et al.

(2017)

YX/A Acetate yield coefficient 0.059 gC gC-1 Figueroa-Torres et al.

(2017)

Ϭ Light attenuation coefficient 1 L gC-1 m-1 Figueroa-Torres et al.

(2017)

Ass

oci

ated

to

N &

P -

up

tak

e

ρN,max Maximum N uptake rate 44.01 gN gC-1h-1 This work *

K* Saturation constant, No 0.300 gN L-1 This work *

n Shape-controlling parameter 14.54 - This work *

ФN N Uptake regulation coefficient 143.9 L gC-1 This work *

Ks,N Uptake saturation constant, N 0.163 gN L-1 Figueroa-Torres et al.

(2017)

ki,N Uptake inhibition constant, N 0.113 gN L-1 Figueroa-Torres et al.

(2017)

Ks,A:N Uptake saturation constant, A:N 1.004 gC L-1 Figueroa-Torres et al.

(2017)

ki,A:N Uptake inhibition constant, A:N 1.098 gC L-1 Figueroa-Torres et al.

(2017)

KP P quota supporting N uptake 0.057 gPO4 gC-1 This work

ρP,max Maximum P uptake rate 21.10 gPO4 gC-1h-1 This work

Ks,P Uptake saturation constant, P 2.299 gPO4 L-1 This work

ki,P Uptake inhibition constant, P 0.004 gPO4 L-1 This work

Ass

oci

ated

to

Sta

rch

& L

ipid

fo

rmat

ion

r1 Starch formation rate (R1) 0.058 gC gC-1 This work *

Ks,S Saturation constant (R1) 0.000 gN L-1 This work *

ki,S Inhibition constant (R1) 0.205 gN L-1 This work *

nS Shape parameter (R1) 4.17 - This work *

k1 Regulation constant (R1) 0.108 - This work *

ФS Regulation coefficient (R1) 0.775 L gC-1 This work *

r2 Starch degradation rate (R2) 0.005 gC gC-1 This work *

ksat,S Starch saturation constant (R2) 0.018 - This work

r3 Lipid formation rate (R3) 0.191 gN gC-1h-1 This work *

Ks,L Saturation constant (R3) 0.012 gN L-1 This work *

ki,L Inhibition constant (R3) 0.091 gN L-1 This work *

nL Shape parameter (R3) 2.01 - This work *

k2 Regulation constant (R3) 0.153 - This work *

ФL Regulation coefficient (R3) 0.000 L gC-1 This work *

r4 Lipid degradation rate (R4) 0.007 gN gC-1h-1 This work *

ksat,L Lipid saturation constant (R4) 0.079 - This work * Parameter values were re-identified from those established in Figueroa-Torres et al., 2017.



164

%

[TA

P]

0.3

18

±0.0

10

-5.6

%0.0

179

±0.0

003

-14.1

%0.0

448

±0.0

053

-

[Lo

w N

: L

ow

P]

0.2

47

±0.0

24

*16.8

%***

0.0

414

±0.0

035

***

17.7

%0.0

436

±0.0

022

[Lo

w N

]0.2

81

±0.0

01

16.8

%***

0.0

473

±0.0

008

***

21.2

%***

0.0

596

±0.0

009

*

[Med N

]0.3

05

±0.0

03

10.1

%*

0.0

309

±0.0

004

***

18.6

%*

0.0

566

±0.0

003

[Lo

w P

]0.2

67

±0.0

05

11.3

%**

0.0

302

±0.0

001

***

15.6

%0.0

415

±0.0

010

[Med P

]0.2

94

±0.0

11

7.1

%0.0

208

±0.0

009

14.3

%0.0

419

±0.0

016

[Lo

w A

]0.2

59

±0.0

12

4.9

%0.0

128

±0.0

008

14.8

%0.0

383

±0.0

010

[Hig

h A

]0.3

90

±0.0

50

*5.6

%0.0

220

±0.0

000

17.1

%0.0

666

±0.0

008

***

[Hig

h A

+]

0.4

14

±0.0

14

**

9.2

%0.0

380

±0.0

008

***

18.3

%0.0

758

±0.0

006

***

[Hig

h A

: L

ow

N-]

0.2

34

±0.0

13

*22.9

%***

0.0

536

±0.0

021

***

20.5

%**

0.0

479

±0.0

002

[Hig

h A

: L

ow

P]

0.3

72

±0.0

05

8.2

%0.0

304

±0.0

013

***

16.7

%0.0

620

±0.0

065

**

[HIG

H N

]0.1

68

±0.0

02

***

8.4

%0.0

141

±0.0

000

14.4

%0.0

242

±0.0

024

***

[HIG

H P

]0.2

94

±0.0

35

5.3

%0.0

155

±0.0

016

14.7

%0.0

431

±0.0

075

[HIG

H A

]0.2

94

±0.0

17

6.1

%0.0

178

±0.0

012

14.3

%0.0

421

±0.0

008

Tre

atm

ent

La

bel

gC

L-1

gC

L-1

gC

L-1

Bio

ma

ss (

CD

W)

%

Sta

rch

Lip

ids

Tab

le 3

. B

iom

ass

, st

arc

h,

an

d l

ipid

con

cen

trati

on

s q

uan

tifi

ed a

t th

e 8

th d

ay

(t

= 1

92

h)

of

gro

wth

in

cu

ltu

res

of

C.

rein

hard

tii

CC

AP

11/3

2c s

ub

ject

to v

ari

ou

s in

itia

l n

utr

ien

t re

gim

es. S

tars

(*)

den

ote

sig

nif

ican

t d

iffe

ren

ces

(p <

0.0

5*,

0.0

1**, 0.0

01*

**)

wit

h r

esp

ect

to [

TA

P],

as

per

on

e-w

ay A

NO

VA

. D

ata

are

th

e m

ean

of

two b

iolo

gic

al

rep

lica

tes.



165

FIGURES

Figure 1. Schematic representation of the cellular compartments and flows used in the

kinetic model for mixotrophic algal growth co-limited by nitrogen and phosphorus. X,

total biomass; μ, specific growth rate; 𝝆𝑵, nitrogen uptake rate; 𝝆𝑷, nitrogen uptake rate;

𝑹𝟏, starch synthetic rate; 𝑹𝟑, lipid synthetic rate; 𝑹𝟐, starch degradation rate; 𝑹𝟒, lipid

degradation rate.

Acetate (Carbon)

A

Nitrogen

N

Starch

S

Lipids

L

Active

biomass

X*

Total Biomass: X

X = x* + S + L

Irradiance

I

N quota

qN

Phosphorus

PP quota

qP

𝑅1

𝑅2

𝑅4

𝑅3

𝜇

𝜌𝑁

𝜌𝑃

Mixotrophic

growth



166

Figure 2. Comparison between the predicted time-profile (lines) of the cultivation

variables and the experimental datasets (points) used for parameter fitting: [TAP]:

N0=0.382 gN L-1, P0=0.096 gPO4 L-1, A0=0.42 gC L-1, [Low P]: 0.382 gN L-1, 0.0096

gPO4 L-1, 0.42 gC L-1, and [High A:Low P]: 0.382 gN L-1, 0.0096 gPO4 L

-1, 1.26 gC L-1.

Data and standard deviation are the mean of two experimental replicates.



167

Figure 3. Comparison between the predicted time-profile (lines) of the cultivation

variables and the experimental datasets (points) used for model validation: [Low N:Low

P]: N0=0.335 gN L-1, P0=0.0096 gPO4 L-1, A0=0.42 gC L-1, [Med N]: 0.354 gN L-1, 0.096

gPO4 L-1, 0.42 gC L-1, and [High A+]: 0.382 gN L-1, 0.096 gPO4 L

-1, 1.26 gC L-1. Data

and standard deviation are the mean of two experimental replicates.



168

Figure 4. Parity plots comparing predicted and experimental data for both fitting and

validating datsets. r2 is the computed mean correlation coefficient for: a) biomass, b)

nitrogen, c) nitrogen quota, d) phosphorus, e) phosphorus quota, f) acetic acid, g) starch,

h) lipids, and i) active biomass. Data are the mean of two experimental replicates.



169

Figure 5. Ternary diagrams for: a) biomass, b) starch, and c) lipid formation (as

predicted by the model) in C. reinhardtii CCAP 11/32c (at t=192 h) subject to different

initial nitrogen, phosphorus, and acetic acid concentration sets.

0.00 0.88 1.75 2.63 3.500.00

0.04

0.07

0.11

0.140.25

0.38

0.50

0.63

0.75P

hosp

hate, gP

O4 L

-1Nit

roge

n, g

N L

-1

Acetate, g C L-1

0.00

0.0460

0.0920

0.138

0.184

0.230

0.276

0.322

0.368

0.414

Biomass, gC L-1

0.25 0.63 1.00 1.38 1.750.00

0.04

0.07

0.11

0.140.25

0.38

0.50

0.63

0.75

Phosp

hate, gP

O4 L

-1Nit

roge

n, g

N L

-1

Acetate, g C L-1

0.0

0.0078

0.016

0.023

0.031

0.039

0.047

0.054

0.062

0.070

Starch, gC L-1

0.25 0.63 1.00 1.38 1.750.00

0.04

0.07

0.11

0.140.25

0.38

0.50

0.63

0.75

Phosp

hate, gP

O4 L

-1Nit

roge

n, g

N L

-1

Acetate, g C L-1

0.0

0.0093

0.019

0.028

0.037

0.047

0.056

0.065

0.075

0.084

Lipids, gC L-1

a)

b)

c)



170

Figure 6. Comparison between the predicted and experimental data for C. reinhardtii

cultures grown under: non-optimised medium [No=0.382 gN L-1, P0=0.096 gPO4 L-1,

A0=0.42 gC L-1], starch-enhancing medium [N0=0.331 gN L-1, P0=0.051 gPO4 L-1,

A0=0.96 gC L-1], and lipid-enhancing medium [N0=0.363 gN L-1, P0=0.039 gPO4 L-1,

A0=1.00 gC L-1]. Data and standard deviation are the mean of 2 experimental replicates.



171



presented next.



172


Associated to:

Optimisation of Microalgal Starch and Lipid Formation via Nitrogen and

Phosphorus Co-limitation

Gonzalo M. Figueroa-Torres a, Jon K. Pittman b and Constantinos Theodoropoulos a,*




9PL







173

1. Statistical analysis.

Table S.1 includes the p-values obtained by the one-way ANOVA analysis (tukey test) of

the experimental data compared against the control culture (i.e. [TAP]). Analysis was

carried out in Origin Pro 2017 (b9.4.1.354).

Table S.1. p-values obtained by one-way ANOVA. Highlighted cells denote

significant differences between each treatment pair (p<0.05).

p values

Treatment pair X S (%) S (gC/L) L (%) L (gC/L)

[TAP] - - - - -

[TAP] [Low N : Low P] 0.0483 4.24E-06 0.00E+00 0.1524 1.0000

[TAP] [Low N] 0.7886 4.73E-06 0.00E+00 0.0007 0.0179

[TAP] [Med N] 1.0000 0.0398 8.47E-06 0.0429 0.0883

[TAP] [Low P] 0.3837 0.0063 1.70E-05 0.9720 0.9970

[TAP] [Med P] 0.9850 0.9764 0.6502 1.0000 0.9992

[TAP] [Low A] 0.2189 1.0000 0.0684 1.0000 0.7367

[TAP] [High A] 0.0438 1.0000 0.2327 0.2928 0.0005

[TAP] [High A+] 0.0099 0.1675 2.97E-08 0.0628 8.99E-06

[TAP] [High A : Low N-] 0.0269 0.00E+00 0.00E+00 0.0021 0.9989

[TAP] [High A : Low P] 0.3421 0.5626 1.37E-05 0.5414 0.0052

[TAP] [HIGH N] 0.0001 0.4311 0.3261 1.0000 0.0009

[TAP] [HIGH P] 0.9842 1.0000 0.8638 1.0000 1.0000

[TAP] [HIGH A] 0.9838 1.0000 1.0000 1.0000 0.9996

2. Sensitivity analysis (by central differences).

A sensitivity analysis was carried out for all model parameters by computing the numerical

sensitivities, as per central differences, for each variable with respect to each kinetic

parameter subject to a ±10 % change:

𝑁𝑢𝑚𝑒𝑟𝑖𝑐𝑎𝑙 𝑠𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 =𝑉𝑎𝑟(𝑡, 𝑃 + ∆𝑃) − 𝑉𝑎𝑟(𝑡, 𝑃 − ∆𝑃)

2 ∙ ∆𝑃 (S.1)

Sensitivities were computed at 5 different cultivation times (t=24, 48, 120, 168, and 192

h) to account for changes throughout the cultivation time. A sensitivity threshold of 0.02

was used (Figueroa-Torres et al. 2017), so that parameters with sensitivities higher than

0.02 were deemed sensitive. Table S.2 exemplifies the above by showing the results

obtained from the sensitivity analysis for the maximum specific growth rate, µmax, and the



174

starch saturation constant (R1), Ks,S. It was observed that µmax was highly sensitive since

the computed average numerical sensitivity was higher than 0.02 for all 9 state variables.

Meanwhile, Ks,S, was deemed to be not sensitive since for all 9 state variables, the

computed average sensitivity was below the 0.02 threshold.

Table S.2. Parameter sensitivities for the maximum specific growth rate, µmax, and

the starch saturation constant (R1), Ks,S. Cells highlighted are > 0.02.

Time (hours) Parameter Variable 24 48 120 168 192 Average

µmax X 0.051 0.431 0.712 2.894 3.207 1.459

N 0.338 0.824 2.915 2.915 2.915 1.982

qN 177.775 400.668 14.714 1.084 0.287 118.906

P 0.017 0.043 0.720 0.613 0.561 0.391

qP 20.833 68.337 3.787 0.594 1.038 18.918

A 0.135 1.129 0.242 0.422 0.443 0.474

S 0.002 0.025 0.336 0.423 0.500 0.257

L 0.006 0.038 0.133 0.354 0.480 0.202

x* 0.042 0.369 0.243 2.963 3.227 1.369

Ks,S X 0.000 0.000 0.000 0.000 0.000 0.000

N 0.000 0.000 0.000 0.000 0.000 0.000

qN 0.000 0.004 0.000 0.000 0.000 0.000

P 0.000 0.000 0.000 0.000 0.000 0.000

qP 0.000 0.000 0.000 0.000 0.000 0.000

A 0.000 0.000 0.000 0.000 0.000 0.000

S 0.000 0.000 0.000 0.000 0.000 0.000

L 0.000 0.000 0.000 0.000 0.000 0.000

x* 0.000 0.000 0.000 0.000 0.000 0.000

The results of the sensitivity analysis for all parameters is summarised in Table S.3, which

presents the computed average (across all 5 time points) numerical sensitivity for all state

variables with respect to each kinetic parameter. As per Table S.3: 6 kinetic parameters

were deemed highly sensitive (µmax, ki,P, ki,N, K*, KP, and Ks,N) since their numerical

sensitivities were higher than 0.02 for all state variables; 28 parameters were deemed

sensitive for at least one state variable; and 3 were deemed not sensitive for any state

variable (Ϭ, Ks,S, and ФL). The light attenuation coefficient, Ϭ, was set to nominal value of

1. Meanwhile, the starch saturation constant, Ks,S, and the lipid regulation coefficient, ФL,

were set to 0 since it was observed that this did not affect the model’s predictive capacity.



175

Table S.3. Average numerical sensitivities of the 37 model parameter. Cells

highlighted are > 0.02.

Parameter X N qN P qP A S L x* Mean

µmax 1.459 1.982 118.906 0.391 18.918 0.474 0.257 0.202 1.369 15.995**

ki,P 0.677 1.005 38.328 1.954 95.936 0.128 0.097 0.090 0.687 15.434**

ki,N 0.914 1.305 40.699 0.217 6.130 0.179 0.130 0.121 0.925 5.624**

K* 0.314 0.435 12.201 0.073 1.837 0.068 0.049 0.037 0.327 1.705**

KP 0.163 0.251 9.315 0.049 1.930 0.035 0.022 0.022 0.164 1.328**

r2 0.001 0.001 0.007 0.000 0.002 0.001 4.809 0.861 3.946 1.070*

r4 0.000 0.000 0.000 0.000 0.000 0.000 0.505 3.932 3.427 0.874*

Ks,N 0.218 0.286 6.101 0.039 0.843 0.040 0.031 0.029 0.220 0.868**

Ks,A:N 0.120 0.167 4.688 0.026 0.687 0.023 0.017 0.017 0.120 0.651*

qP,0 0.046 0.055 4.093 0.001 0.108 0.021 0.009 0.003 0.053 0.488*

YX/A 0.088 0.105 1.427 0.014 0.175 0.696 0.039 0.017 0.110 0.297*

qN,0 0.181 0.017 1.696 0.006 0.357 0.034 0.008 0.005 0.188 0.277*

ki,L 0.000 0.000 0.021 0.000 0.003 0.000 0.105 0.703 0.598 0.159*

r1 0.001 0.000 0.031 0.000 0.003 0.000 0.458 0.093 0.364 0.106*

ki,A:N 0.016 0.023 0.662 0.004 0.098 0.003 0.002 0.002 0.016 0.092*

ki,S 0.000 0.000 0.012 0.000 0.001 0.000 0.397 0.079 0.317 0.090*

ksat,S 0.000 0.000 0.001 0.000 0.000 0.000 0.281 0.050 0.231 0.063*

ki,A 0.016 0.019 0.287 0.002 0.036 0.058 0.004 0.003 0.018 0.049*

r3 0.000 0.000 0.010 0.000 0.002 0.000 0.028 0.188 0.159 0.043*

k1 0.000 0.000 0.015 0.000 0.002 0.000 0.181 0.037 0.144 0.042*

Ks,P 0.001 0.002 0.069 0.009 0.244 0.000 0.000 0.000 0.001 0.036*

k2 0.000 0.000 0.011 0.000 0.002 0.000 0.023 0.154 0.131 0.036*

ksat,L 0.000 0.000 0.000 0.000 0.000 0.000 0.016 0.123 0.107 0.027*

Ks,L 0.001 0.005 0.170 0.000 0.000 0.002 0.013 0.033 0.021 0.027*

ρN,max 0.004 0.006 0.166 0.001 0.024 0.001 0.001 0.001 0.004 0.023*

Ks,A 0.000 0.001 0.056 0.000 0.007 0.072 0.003 0.000 0.003 0.016*

n 0.002 0.003 0.103 0.001 0.016 0.001 0.000 0.001 0.002 0.014*

Ks,I 0.001 0.001 0.078 0.000 0.012 0.000 0.000 0.000 0.001 0.010*

ФS 0.000 0.000 0.000 0.000 0.000 0.000 0.038 0.009 0.029 0.008*

nL 0.000 0.000 0.001 0.000 0.000 0.000 0.005 0.035 0.030 0.008*

nL 0.000 0.000 0.001 0.000 0.000 0.000 0.005 0.035 0.030 0.008*

ki,I 0.001 0.000 0.047 0.000 0.007 0.000 0.000 0.000 0.001 0.006*

nS 0.000 0.000 0.000 0.000 0.000 0.000 0.022 0.006 0.016 0.005*

ФN 0.001 0.002 0.031 0.000 0.004 0.000 0.000 0.000 0.001 0.004*

ρP,max 0.000 0.001 0.025 0.000 0.000 0.001 0.000 0.000 0.000 0.003*

ФL 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.013 0.012 0.003

Ϭ 0.000 0.000 0.002 0.000 0.001 0.000 0.000 0.000 0.000 0.000

Ks,S 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.000

** Parameters deemed highly sensitive

* Parameters deemed sensitive

3. Starch and lipid dynamics.

As explained in the main text (section 3.3), the starch and lipid degradation rates, Eq . (15)

and Eq. (16), incorporate two saturating functions (using ksat,s, and ksat,L) which avoided

unfeasible starch and lipid accumulation scenarios suffered by the unsaturated structure

proposed in our previous work (Figueroa-Torres et al., 2017).



176

As observed in Figure S.1.(a), the starch and lipid dynamics of the culture grown in [TAP],

as predicted by the “unsaturated” model (Figueroa-Torres et al., 2017), become unbounded

and attain either negative (starch) or ever-increasing (lipids) concentrations. The original

model formulation was therefore improved by the use of saturating functions which, as

shown in Figure S.1.(b), bound starch and lipid dynamics within steady concentration

profiles.

a)

b)

Figure S.1. Predicted and experimental dynamics of starch, lipids and active

biomass: a) without saturating functions, and b) with saturating functions (this

work).

The dynamics predicted by the improved saturated model were further evaluated using

cultivation conditions subject to [Low N] and [High A] concentrations. As observed in

Figure S.2, the starch and lipid accumulation rates predicted by the improved model

formulation clearly reach a value of zero (i.e. accumulation stops), which in turn allows

starch, lipids, and active biomass to reach the steady state. As observed in Figure S.3, such

dynamics are not predicted by our previous model.



177

Figure S.2. Cultivation dynamics subject to three cultivation conditions, as

predicted by the improved saturated model (this work).

Figure S.3. Cultivation dynamics subject to three cultivation conditions, as

predicted by the unsaturated model presented in Figueroa-Torres et al., 2017.



178

4. Normalised sensitivity analysis.

To assess further the effect of the kinetic parameters on the model variables, a normalised

sensitivity analysis was performed by calculating:

𝑆𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 =

𝜕𝑍𝑖

𝜕𝑃𝑖∙

𝑃𝑖,𝑜

𝑍𝑖,𝑜=

𝜕𝑍𝑖/𝑍𝑖,𝑜

𝜕𝑃𝑖/𝑃𝑖,𝑜 (S.2)

Where 𝜕𝑍𝑖 𝜕𝑃𝑖⁄ denotes the response change in a model state variable with respect to a

corresponding change in a model parameter, 𝑍𝑖,𝑜 is the response of the model state variable

when the parameter is set to 𝑃𝑖,𝑜. The sensitivity reflects the outcome of a change in the

parameter value: for sensitivities > 1, a change in the parameter increases the response of

the model variable; for sensitivities < 1, a change in the parameter decreases the response

of the model variable; the greater the sensitivity, the greater the effect of the parameter.

For computations: i) parameters were increased by 1 %, whilst keeping all other parameters

constant , and ii) the sensitivity was computed over a 200 h period. Model outputs were

generated using initial conditions equivalent to those used in [TAP], with the exception of

the minimum P quota, 𝑞𝑃,0, given that the effect of this parameter becomes significant as

the initial P concentrations decreases (Figure S.4). This is a consequence of the minimum

law adopted within the specific growth rate (Eq. 2 in the main text). The results of the

sensitivity analysis are presented in Figure S.5.

Figure S.4. Normalised sensitivity of the model variables with respect to a 1 %

increase in 𝒒𝑷,𝟎, over a 200 h period, subject to different initial P concentrations.



179

Figure S.5. Normalised sensitivity of the model state variables with respect to a 1 %



uptake, orange – associated to P uptake, and black – associated to starch and lipid

formation. * The sensitivities for 𝒒𝑷,𝟎 were obtained by setting P0=0.0096 gPO4 L-1.



180





formation.



181





formation.



182

Figure S.5. (cont.) Normalised sensitivity of the model state variables with respect to

a 1 % increase in each model parameter, over a 200 h cultivation period. Parameter



formation.

The computed sensitivities of the parameters are similar to those obtained when the

analysis was carried out using the previous (unsaturated) model formulation, and therefore

similar conclusions are obtained (see Supplementary Information, Chapter 3). Indeed, the

parameters deemed not sensitive in this improved model remain the same as those

previously identified before: 𝜎 , 𝑘𝑆,𝐼 , 𝐾𝑠,𝑆 , and 𝜙𝐿 . The values of these parameters, as

mentioned above, were set to: 𝜎 = 1, 𝑘𝑆,𝐼 = 1.4, 𝐾𝑠,𝑆 = 0, and 𝜙𝐿 = 0.

The improved model, however, includes phosphorous as an additional state variable. The

sensitivity analysis shown in Figure S.5 thus allows to observe that the effects of the model

parameters associated to biomass growth and N uptake are as significant to phosphorous

as they are for nitrogen. However, whilst nitrogen-associated parameters have an effect

on all model variables, the three phosphorus-associated parameters (i.e. 𝜌𝑃,𝑚𝑎𝑥, 𝐾𝑆,𝑃 and

𝑘𝑖,𝑃) only affect phosphorous since their computed sensitivities are considerably low for

the remaining variables.

As observed with the previous model, the effect of the starch- and lipid-associated

parameters (including the two new parameters 𝑘𝑠𝑎𝑡,𝑆 and 𝑘𝑠𝑎𝑡,𝐿 ) is, as expected, only



183

significant for starch and lipids. However, it should be mentioned that the computed

sensitivities of these parameters noticeably tend to a steady state, which is due to the

corrected model formulation which prevents starch and lipids from attaining unsteady

concentration profiles.



184

185

Chapter 5

An Experimental and Model-based Evaluation of Fed-

Batch Microalgal Cultivation for Biofuels Production

5.1. Introduction.

As evidenced by those works discussed previously in the Literature Review (see Chapter

2), and also by the research findings presented thus far (see Chapter 3 and Chapter 4),

one of the main challenges of microalgal cultivation for the purposes of biofuel production

is the well-known trade-off between algal biomass growth and starch and lipid

accumulation. As established before, those cultivation conditions that typically increase

starch and lipid formation (e.g. nitrogen and/or phosphorus limitation) can substantially

reduce biomass densities, making them unfit for the production of microalgal biofuels at a

commercial scale.

Fed-batch operations, which rely on appropriate nutrient feeding strategies, are already

widely and typically employed in various biological industrial processes to increase both

cell life and production yields. Therefore, fed-batch microalgal cultivation represents a

suitable strategy to sustain nutrient-limited microalgal growth whilst simultaneously

favouring starch or lipid accumulation. The optimal implementation of fed-batch

strategies, however, requires optimal nutrient feeding strategies to be identified.

Model-based optimisation can reduce and simplify optimisation tasks (see Chapter 3 and

Chapter 4), but most of the existing modelling approaches found in literature (see

Chapter 2) have been developed to describe batch dynamics. Indeed, the model developed

thus far throughout this thesis was fundamentally built based on experimental observations

obtained from multiple batch cultivation scenarios. Therefore, to generate useful data that

can be used to evaluate and further adapt the model’s predictive capacity for fed-batch

Chapter 5 – An Experimental and Model-based Evaluation of Fed-batch

Microalgal Cultivation for Biofuels Production

186

dynamics, an experimental analysis of such a cultivation strategy was carried out and will

be presented next.

The fed-batch mode employed here involved the addition of acetic acid (i.e. the

mixotrophic carbon source) pulses throughout the cultivation period with the aim of

prolonging microalgal growth as other nutrients became exhausted (i.e. nutrient stress).

Such a feeding strategy was experimentally evaluated in laboratory-scale cultures of the

strain employed in this thesis: C. reinhardtii. To favour the presence of nutrient-limited

conditions, the nitrogen and phosphorus sources were supplemented only at the beginning

of the cultivation as per their corresponding standard concentrations (see Appendix A),

and were allowed to naturally deplete as cells continued to grow.

Preliminary experiments indicated that the acidic nature of the acetic acid pulses (pH ~2.5

after preparation) caused cultivation medium pH to drop drastically after the pulse was

injected. This sudden reduction in pH was unfavourable since it caused cell mass to decline

rapidly in all cases, defeating the purpose of the fed-batch strategy. In order to avoid cell

death due to acid pH values, the acetic acid pulse was neutralised with potassium

hydroxide. However, as will be shown throughout the text, the increased presence of

potassium hydroxide seemed to be inhibitory for cell growth and restricted the use of pulses

with higher acetic acid concentrations.

The following paper will show the favourable outcome of the fed-batch strategy employing

up to three pulses of acetic acid throughout the cultivation period. Compared to the base

case (batch culture), the fed-batch strategy yielded significant increases in biomass, starch,

and lipid concentrations, highlighting its potential implementation. In addition, and in

order to provide a means to simulate the dynamics of the pulse-assisted cultivation, the

model developed previously in Chapter 1 was evaluated based on its capacity to predict

the experimental observations. As will be shown within the text, the model’s original

structure was not capable of accounting for a number of biological processes exhibited by

the cells when they were supplemented with the pulses.

To overcome the above, an adaptation of the model’s structure and a re-parametrization

procedure was carried out, allowing the model to simulate fairly the outcome of a single-

pulse of acetic acid. However, the model failed to predict the dynamics of cultures that



187

were subject to two consecutive pulses. The major drawback of the model, which will be

better addressed within the main text of the manuscript, was derived from its inability to

predict the clear increase in biomass concentration attained by the cultures grown in fed-

batch mode.

It is necessary to acknowledge that whilst the experimental results obtained from this work

are novel and validate the advantage of implementing a pulse-assisted fed-batch strategy

for increased biomass growth, the predictive capacity of the proposed model formulation

is limited. Further work and refining is envisaged before the following paper is deemed fit

for final publication, but to maintain consistency with the style selected for this thesis all

the major experimental and computational findings are presented in manuscript format.



188



189


Figueroa-Torres GM, Pittman JK, Theodoropoulos C. (2018). An experimental and

Model-based Evaluation of Fed-Batch Microalgal Cultivation for Biofuels Production. To

be submitted to: Algal Research.






revised the manuscript.



190

An Experimental and Model-based Analysis of Fed-

Batch Microalgal Cultivation for Biofuels Production

Gonzalo M. Figueroa-Torresa, Jon K. Pittmanb, Constantinos Theodoropoulosa,*




9PL

*Corresponding author:






191

ABSTRACT

Third-generation biofuels, produced from microalgal carbohydrates and lipids, have

attracted attention since their environmental impacts are much lower than those of first-

and second-generation biofuels. Producing third-generation biofuels requires the mass-

scale cultivation of microalgal biomass suitable for the final biofuel downstream processes.

However, conventional batch cultivation of phototrophic algal strains is unfit for biofuels

production purposes due to the low cell densities attained. Fed-batch cultivation of

mixotrophic strains is suggested as a more reliable strategy for sustained microalgal growth

via the implementation of optimal nutrient feeding strategies. This work presents a fed-

batch cultivation strategy consisting of intermittent acetic acid (carbon substrate) pulses

resulting in significantly increased biomass densities, and consequently carbohydrate

(starch) and lipid formation. The strategy was evaluated in bench-scale mixotrophic

cultures of Chlamydomonas reinhardtii CCAP 11/32C, and yielded an increase in

concentration of 94 % biomass, 217 % starch, and 167 % lipids with respect to the batch

case. In addition, and based on experimental observations, a kinetic model previously

developed for microalgal growth in batch mode was adapted to simulate pulse-assisted

cultivation and its predicted dynamics were evaluated.

Keywords: microalgae, fed-batch, modelling, biofuels, nutrient limitation.



192

1. Introduction.

Biofuels produced from microalgae, so-called third generation biofuels, are considered

sustainable energy alternatives fit to replace fossil-based fuels. The environmental impacts

(e.g. land use, water use, fertiliser use, and greenhouse gas emissions) associated to

microalgal biofuels production are estimated to be much lower than those associated to

traditional food-based biofuels or those produced from lignocellulosic substrates [1,2].

Since microalgae are aquatic photosynthetic organisms which grow on a variety of fresh

or marine water environments, the cultivation of microalgae for the purpose of biofuels

production avoids one of the most controversial disadvantages of food-based feedstocks:

the competition for food and arable land destined for human activities [3].

The cellular composition of microalgal biomass is generally rich in carbohydrates and

lipids, carbon-based storage molecules that act as raw substrates in the conversion

processes (e.g. fermentation or transesterification) for bioethanol, biobutanol, or biodiesel

[4]. Therefore, the successful commercialisation of third-generation biofuels relies on the

adequate establishment of mass-scale microalgal cultivation systems generating high

density biomass containing the biofuel precursors (carbohydrate and lipid). In this regard,

nutrient-limited cultivation strategies (i.e. those where algal cells undergo stressed growth

as a consequence of reduced nutrient availability) have been proven to induce the

accumulation of storage molecules [5,6].

Limitation by nitrogen or phosphorus, particularly, has significantly increased the

intracellular carbohydrate and lipid contents in green microalgae [7–10], including the

model species Chlamydomonas reinhardtii, which accumulates carbohydrate in the form

of starch granules [11,12]. Nutrient limitation can be exploited for biofuel-oriented

microalgal cultivation, but it must be optimally implemented to avoid a reduction of

biomass growth that could ultimately be unfit for biofuels production. The challenge of

maintaining high biomass densities whilst nutrient-limited conditions simultaneously

induce starch and/or lipid formation has been approached via the use of mixotrophic

species whose growth requirements are satisfied by both inorganic carbon dioxide and

organic carbon sources [13–15].



193

Additional alternatives which avoid the undesired reduction of biomass (or further increase

it) during nutrient limitation include the use of two-stage or fed-batch systems. Two-stage

strategies rely on transferring cells from a nutrient-replete stage which allows adequate cell

densities to be attained, to a nutrient-limited stage which allows starch and/or lipid

accumulation to be induced [16,17]. This approach can meet the desired targets if

implemented correctly, but its scalability is deemed economically uncertain given that the

process of harvesting cells and transferring them between each stage will require large

energy inputs [5,9].

Fed-batch systems, on the other hand, are already one of the preferred operating modes for

industrial bioprocesses targeting increased cell life or productivity [18,19]. Fed-batch

operation has been proven to yield high microalgal biomass densities (which facilitates

downstream processes) and to additionally favour starch or lipid formation [20–23],

making it a promising strategy for biofuel production purposes. The implementation of

fed-batch microalgal cultivation strategies, however, relies on the identification of the most

appropriate nutrient feeding regime.

Modelling and simulation tools can enable the fast identification and further optimisation

of bioprocessing strategies whilst diminishing the costs and time associated to

experimental analysis [24]. Mathematical models capable of describing microalgal growth

dynamics during batch and fed-batch operation can thus facilitate the identification of

nutrient feeding regimes suitable for increased biomass. We previously developed a model

with a high predictive capacity for biomass growth and starch and lipid formation in C.

reinhardtii under various nutrient concentration regimes, although subject to batch

operation [25].

In this work we present a fed-batch cultivation strategy whereby microalgal biomass

densities are not only sustained, but further increased, therefore favouring starch and lipid

formation. The fed-batch nutrient feeding regime employed here consisted of acetic acid

(i.e. the carbon substrate) pulses supplied intermittently during cultivation. We additionally

assess and improve the predictive capacity of our previously developed model by adapting

the model’s original structure as per fed-batch experimental observations.



194



All experiments were carried out with the wild-type strain Chlamydomonas reinhardtii

CCAP 11/32C, grown mixotrophically in Tris-Acetate-Phosphate (TAP) medium [26].

Prior to fed-batch cultivation experiments, an active algal inoculum was prepared by

growing the strain for 7 days (up to late stationary phase) in 150 mL of sterile TAP

medium. The algal inoculum was kept in a rotary shaker at a rotating speed of 150 rpm.

Temperature was maintained at 25 °C, and light was supplied at an incident intensity of

125 μmol m-2s-1 (one-side illumination, from above) in a photoperiod of 16 h light and 8 h

dark.

2.2. Fed-Batch cultivation strategy.

Fed-batch cultivation experiments were performed in duplicate in 500 mL of sterile TAP

medium under the environmental conditions described above. Growth was initiated by

inoculating all culture vessels simultaneously with 1 mL of algal inoculum. All cultures

were first allowed to grow in batch mode up to the beginning of the exponential phase,

after which they were subjected to a feeding strategy involving intermittent pulses of acetic

acid (i.e. the organic carbon source in standard TAP medium) at different concentrations

and cultivation times. The volume of the pulses was set to 10 mL, and their acetic acid

concentration ranged from 5.1 gC L-1 to 31 gC L-1, so that the corresponding increase in

the medium concentration after pulse addition ranged from + 0.1 gC L-1 to + 0.6 gC L-1

(e.g. a 10 mL pulse with 5.1 gC L-1 of acetic acid, supplied to 500 mL of a growing culture,

increased the residual medium concentration by 0.1 gC L-1, and so on). For reference,

standard TAP medium contains an initial acetic acid concentration of 0.42 gC L-1

(equivalent to 1.05 g mL-1). To improve clarity and interpretation of the results, the

different pulses, Pi, evaluated here are identified by their corresponding increase of the

acetic acid medium concentration.

Pulses were prepared by diluting the required concentration of acetic acid in standard TAP

salts solution free of nitrogen and phosphorus sources to maintain nutrient-limited

conditions. To avoid a drastic reduction of pH in the culture medium, the pH of the pulses



195

was set to 4.5 with potassium hydroxide (KOH) 3M, and final volume was brought up to

10 mL. All pulses were sterilised prior to use. Fed-batch cultivations with 1 single pulse,

2 consecutive pulses, or 3 consecutive pulses were evaluated. During sampling, cultures

were fully harvested for analysis of the biomass (cell dry weight) and residual metabolites

concentrations. All data was statistically analysed by one-way ANOVA in origin Pro 2017

(b9.4.1354).

2.3. Analytical methods.

2.3.1. Cell dry weight.

The biomass cell dry weight (CDW) was measured by centrifuging microalgal cultures for

10 min at 7,500 rpm in an Avanti J-26S XP centrifuge (Beckman Coulter). Pelleted cells

were placed in pre-weighed tubes and left to dry at 70 °C for 24 h. Pellets were cooled

down to room temperature in a desiccator and the cell dry weight was calculated

gravimetrically in a M-Pact AX221 fine balance (Sartorius). Samples of the supernatant

were kept in Falcon tubes and stored at -20 °C for further analysis. The medium pH was

measured in a bench-type HI-2211 pH meter (Hanna Instruments).

2.3.2. Residual nutrients concentration.

The concentration of acetic acid was quantified via HPLC analysis using a HPX-87H

column (300 x 7 mm) and a UV detector at a wavelength of 210 nm. The mobile phase

(H2SO4 0.005 M) was set at a flow rate of 0.6 mL min-1 and a temperature of 50 °C. The

concentration was measured from the area of the chromatographic peaks read against a

calibration curve. The concentration of total nitrogen was quantified in a Total Organic

Carbon/Total Nitrogen unit (TOC-VCSD/TNM-1 Shimadzu) following manufacturer’s

instructions and using a calibration curve prepared with ammonium chloride as nitrogen

source. The intracellular nitrogen concentration at any given time was assumed to be

equivalent to the nitrogen consumed by cells (i.e. the difference between the initial and

residual nitrogen concentration), and the nitrogen quota, 𝑞𝑁 (gN gC-1), was estimated by

dividing the intracellular nitrogen concentration by the biomass (CDW) concentration [25].



196

2.3.3. Starch and lipid quantification.

Microalgal starch was measured through a Total Starch kit (Megazyme International,

Ireland). Prior to analysis, harvested cells were pre-treated as in [27] to remove chlorophyll

pigments, break cells, and solubilise starch. Cells were then subjected to a two-stage

enzymatic hydrolysis (as per manufacturer’s instructions) to release D-glucose, after which

the concentration was measured colourimetrically (at 508 nm) against a standard

calibration curve. Starch concentration was calculated by multiplying D-glucose

concentration by 0.9 (162/180), a factor adjusting free D-glucose to Anhydrous D-glucose.

Microalgal lipids were measured by solvent extraction in a ST-243 SoxtecTM (FOSS). Prior

to lipid extraction, dried cell pellets (as obtained from CDW measurements) were manually

pulverised using mortar and pestle alternated with liquid nitrogen supply. Cells were

weighed and placed in cellulose extraction thimbles (26 x 60 mm, 603, Whatman®), and

the lipids were then extracted in a three-stage program (extraction 2 h, rinsing 40 min, and

solvent recovery 20 min) set as in [15]. Extracted lipids were allowed to cool down to room

temperature in a desiccator and the concentration was then calculated gravimetrically.

3. Modelling of Cultivation Dynamics.

3.1. Kinetic model for batch dynamics.

We developed a kinetic model to describe the dynamics of batch microalgal growth and

starch and lipid formation subject to mixotrophic growing conditions (using acetic acid as

carbon source), during nitrogen limitation. This model adopted a compartmentalised

approach, so that the cell is made up of three carbon-based pools: starch, lipids, and active

biomass (i.e. biomass free of starch and lipids) [25]. In this work, we use this model to

evaluate the dynamics of microalgal cultivation, which takes into account the following

state variables: biomass, X (gC L-1), nitrogen, N (gN L-1), nitrogen quota, qN (gN gC-1),

acetic acid, A (gN L-1), starch, S (gC L-1), lipids, L (gC L-1), active biomass, x* (gC L-1),

and pH, H. The rate of accumulation of each model state variable is described by the

following set of differential equations:

𝑑𝑋

𝑑𝑡= 𝜇 ∙ 𝑋 Eq. (1)



197

𝑑𝑁

𝑑𝑡= −𝜌𝑁 ∙ 𝑋 Eq. (2)

𝑑𝑞𝑁

𝑑𝑡= 𝜌𝑁 − 𝜇 ∙ 𝑞𝑁 Eq. (3)

𝑑𝐴

𝑑𝑡= −

1

𝑌𝑋/𝐴∙

𝜇𝐻

𝜇𝐻 + 𝜇𝐼∙

𝑑𝑋

𝑑𝑡 Eq. (4)

𝑑𝑆

𝑑𝑡= 𝑅1 − 𝑅2 Eq. (5)

𝑑𝐿

𝑑𝑡= 𝑅3 − 𝑅4 Eq. (6)

𝑑𝑥∗

𝑑𝑡=

𝑑𝑋

𝑑𝑡− (

𝑑𝑆

𝑑𝑡+

𝑑𝐿

𝑑𝑡) Eq. (7)

𝑑𝐻

𝑑𝑡= 𝐾𝐻 ∙

𝑑𝑥∗

𝑑𝑡

Eq. (8)

In Eq. 1, the specific growth rate, 𝜇 , is portrayed by an interactive formulation that

incorporates: i) nitrogen-limited growth, 𝜇𝑁(𝑞𝑁), dependent on the nitrogen quota, ii)

heterotrophic growth, 𝜇𝐻(𝐴), dependent on acetic acid concentration, and iii) phototrophic

growth, 𝜇𝐼(𝐼), dependent on the average light intensity received by the culture, as in:

𝜇 = 𝜇𝑚𝑎𝑥 ∙ [𝑤𝐻 ∙ 𝜇𝐻(𝐴) + 𝑤𝐼 ∙ 𝜇𝐼(𝐼)] ∙ 𝜇𝑁(𝑞𝑁) Eq. (9)

where μmax is the maximum specific growth rate. The nitrogen-limited growth rate employs

Droop kinetics [28] as in Eq. 10, and the heterotrophic and phototrophic growth rates

employ Andrews kinetics [29] to account for substrate inhibition and photoinhibition,

respectively, as in Eq. 10:

𝜇𝑁(𝑞𝑁) = 1 −𝑞𝑁,0

𝑞𝑁 Eq. (10)

𝜇𝐻(𝐴) =𝐴

𝐴 + 𝐾𝑆,𝐴 + 𝐴2 𝐾𝑖,𝐴⁄; 𝜇𝐼(𝐼) =

𝐼

𝐼 + 𝐾𝑆,𝐼 + 𝐼2 𝐾𝑖,𝐼⁄ Eq. (11)

Here, qN,0 is the minimum nitrogen quota required to sustain growth; KS,A and Ki,A are half-

saturation and inhibition constants associated to acetic acid, respectively; and KS,A and Ki,A

are light-associated half-saturation and inhibition constants, respectively. The weighing

functions in Eq. 9, wH and wI, regulate the extent of the heterotrophic and phototrophic

rates, respectively, and are described by Eq. 12:


𝐴 𝐾𝑠,𝐴⁄ +𝐼 𝐾𝑠,𝐼⁄; 𝑤𝐼 =


𝐴 𝐾𝑠,𝐴⁄ +𝐼 𝐾𝑠,𝐼⁄ Eq. (12)



198

The average light intensity, 𝐼, received by the microalgal culture between the surface (𝑧 =

0) and the depth (𝑧 = 𝐿) of the vessel was computed as in Eq. 13 [30]:

𝐼 =𝐼𝑜

𝐿∫ 𝑒−𝜎∙𝑋∙𝑧 ∙ 𝑑𝑧 =

𝐼0

𝜎 ∙ 𝑋 ∙ 𝐿∙ (1 − 𝑒−𝜎∙𝑋∙𝐿)

𝐿

0

Eq. (13)

where Io is the incident light (at 𝑧 = 0) and 𝜎 is the light attenuation coefficient. The

nitrogen uptake rate, 𝜌𝑁, is described by:

𝜌𝑁 = ��𝑁,𝑚𝑎𝑥(𝑁0, 𝑋) ∙𝑁

𝑁 + 𝑘𝑠,𝑁 + 𝑁2 𝑘𝑖,𝑁⁄∙

𝐴

𝐴 + 𝑘𝑠,𝐴:𝑁 + 𝐴2 𝑘𝑖,𝐴:𝑁⁄ Eq. (14)

��𝑁,𝑚𝑎𝑥(𝑁𝑜, 𝑋) = 𝜌𝑁,𝑚𝑎𝑥 ∙𝑁𝑜

𝑛


𝑛 ∙ 𝑒−𝜙𝑁∙𝑋 Eq. (15)

In Eq. 14, 𝑘𝑠,𝑁 and 𝑘𝑖,𝑁 are half-saturation and inhibition constants associated to nitrogen,

respectively; and 𝑘𝑠,𝐴:𝑁 and 𝑘𝑖,𝐴:𝑁 are half-saturation and inhibition constants associated

to acetic acid, respectively. In Eq. 15, 𝜌𝑁,𝑚𝑎𝑥, is the maximum nitrogen uptake rate, 𝐾∗ is

a half-saturation constant, n is a shape-controlling parameter, and 𝜙𝑁 is an uptake

regulation coefficient. These last 4 parameters (𝜌𝑁,𝑚𝑎𝑥, 𝐾∗, 𝑛 and 𝜙𝑁) account for the

luxury uptake of nitrogen, as observed in batch cultures of C. reinhardtii when grown

under various nitrogen-limited scenarios [25].

The accumulation of starch and lipids (Eq. 5 and Eq. 6) is regulated by their synthetic rates,

R1 and R3, and their degradation rates, R2 and R4, respectively. These rates are described

by Eq. 16 –Eq. 19:

𝑅1 = 𝑟1 ∙𝑁𝑖

𝑛𝑠

𝑁𝑖𝑛𝑠 + 𝑘𝑠,𝑆

𝑛𝑆 + (𝑁𝑖2 𝑘𝑖,𝑆⁄ )

𝑛𝑠∙

𝑘1

𝑘1 + 𝑁 𝑁𝑜⁄∙ [1 +

1

𝜇∙ 𝑒𝜙𝑆∗𝐴𝑖] ∙ 𝜇 ∙ 𝑥∗ Eq. (16)

𝑅3 = 𝑟3 ∙𝑁𝑖𝑛𝑡

𝑛𝐿

𝑁𝑖𝑛𝑡𝑛𝐿 + 𝑘𝑠,𝐿

𝑛𝐿 + (𝑁𝑖𝑛𝑡2 𝑘𝑖,𝐿⁄ )

𝑛𝐿∙

𝑘2

𝑘2 + 𝑁 𝑁𝑜⁄∙ [1 +

1

𝜇∙ 𝑒𝜙𝐿∗𝐴𝑖] ∙ 𝜇 ∙ 𝑥∗ Eq. (17)

𝑅2 = 𝑟2 ∙𝑋

𝑞𝑁∙

𝑆 𝑋⁄

𝑆 𝑋⁄ + 𝑘𝑠𝑎𝑡,𝑆 Eq. (18)

𝑅4 = 𝑟4 ∙𝑋

𝑞𝑁∙

𝐿 𝑋⁄

𝐿 𝑋⁄ + 𝑘𝑠𝑎𝑡,𝐿 Eq. (19)

In Eq. 16 and Eq. 17, 𝑁𝑖 = 𝑞𝑁 ∙ 𝑋 is the internal nitrogen concentration, and 𝐴𝑖 = 𝐴𝑜 − 𝐴

is the bioavailable carbon concentration; 𝑁0 and 𝐴0 are the initial nitrogen and acetic acid

concentrations (at 𝑡 = 0), respectively; 𝑟1 and 𝑟3 are the starch and lipid synthetic rate



199

constants, respectively; 𝑘𝑠,𝑆 and 𝑘𝑠,𝐿 are half-saturation constants; 𝑘𝑖,𝑆 and 𝑘𝑖,𝐿 are

inhibition constants; 𝑛𝑆 and 𝑛𝐿 are shape-controlling parameters; 𝜙𝑆 and 𝜙𝐿 are

regulation coefficients; and 𝑘1 and 𝑘2 are constants regulating starch and lipid synthesis

with respect to nitrogen consumption.

In Eq. 18 and Eq. 19, 𝑟2 and 𝑟4 are the starch and lipid degradation rate constants,

respectively, and 𝑘𝑠𝑎𝑡,𝑆 and 𝑘𝑠𝑎𝑡,𝐿 are half-saturation constants regulating the extent of

starch and lipid degradation as employed in [31]. The batch dynamics of pH throughout

cultivation are considered in the model, as per Eq. 8, by assuming that the observed

increase in pH throughout cultivation (a consequence of H+ removal as acetate is

consumed), is proportional to the corresponding change in active biomass, x*, by means of

the proportionality coefficient, 𝐾𝐻 [25].

3.2. Model adaptation for fed-batch dynamics.

The predictive capacity of the kinetic model described above (Eq. 1 - Eq. 7) was

experimentally validated against batch cultures of C. reinhardtii CCAP 11/32C grown in

different nitrogen and acetic acid concentration regimes. However, when the model was

evaluated for its capacity to simulate the dynamics of the fed-batch feeding strategy

employed here (i.e. intermittent pulses of acetic acid), the model was observed to be unable

to replicate the ability of cells to regain their capacity for nitrogen uptake following the

addition of a pulse (see Figure 3.b). To account for this experimental observation, the

original expression describing the maximum uptake rate was updated as follows:

��𝑁,𝑚𝑎𝑥(𝑁∗, 𝑋) = 𝜌𝑁,𝑚𝑎𝑥 ∙𝑁∗

𝑛

𝑁∗𝑛 + [

𝐾∗

1 + 𝑋∗ 𝑁∗⁄]

𝑛 ∙ 𝑒−𝜙𝑁∙(𝑋−𝑋∗) Eq. (20)

The improved nitrogen uptake rate (Eq. 20) differs from the initial formulation (Eq. 15) in

that: i) nitrogen uptake is now dependent on the corresponding increase in biomass

concentration with respect to the cell density at the time of pulse (i.e. 𝑋 − 𝑋∗) rather than

by the residual biomass concentration (i.e. 𝑋), and ii) the half-saturation constant 𝐾∗ is

regulated similarly by the cell density, 𝑋∗, and by the residual nitrogen, 𝑁∗, concentrations

at the time of pulse. This change allows the microalgae’s nitrogen uptake capacity to re-

start appropriately when a pulse of acetic acid is supplemented, as per experimental

observations.



200

The dynamics of the medium pH, (Eq. 8), which as explained before is proportional to the

change in active biomass, were also updated to reflect the pulse-assisted strategy. The

adaptation was carried out by including a fractional term that scales the proportional

increase in pH with respect to the corresponding change in active biomass at the time of

pulse addition (i.e. 𝑥∗ − 𝑥𝑜∗, where 𝑥0

∗ is the active biomass concentration at the time of

pulse):

𝑑𝐻

𝑑𝑡= 𝐾𝐻 ∙ (

𝑥∗ − 𝑥𝑜∗

𝑥∗) ∙

𝑑𝑥∗

𝑑𝑡 Eq. (21)

It is worth mentioning that the change in culture volume caused by nutrient feeding regimes

is often an important consideration in fed-batch systems and requires models to account

for any dilution effects. In this work, the addition of 10 mL pulses on the growing cultures

yielded only a 2 % increase in volume, which was considered to be negligible and thus

dilution terms were not incorporated in the model.

3.2. Parameter estimation.

The adapted model described above contains 33 kinetic parameters (Table 1), which were

previously identified in Figueroa-Torres et al. (2017) by fitting model outputs to datasets

obtained from various laboratory-scale batch experiments. In this work, however, the

values of the 4 kinetic parameters included within the nitrogen uptake rate expression (i.e.

𝜌𝑁,𝑚𝑎𝑥 , 𝐾∗ , 𝑛 and 𝜙𝑁 ), were refined in order to make sure that the model’s validated

predictive capacity for batch dynamics was not lost after the original expression (Eq. 15)

was updated to account for fed-batch dynamics (Eq. 20). The values of all the remaining

kinetic parameters were kept constant.

The values of these 4 kinetic parameters were quantified by re-fitting the outputs predicted

by the adapted model to experimental data, but allowing a ±15 % change with respect to

their previously identified values. The fitting protocol was carried out as in our previous

work: briefly, the sum of the squared relative error (Eq. 22) between the predicted data and

experimental data (of the 8 cultivation variables) was minimised via an optimisation-based

routine combining stochastic (simulated annealing, SA) and deterministic (successive

quadratic programming, SQP) algorithms, thus avoiding the chances of getting trapped in

local minima [32].



201

min 𝐺(𝑃) = ∑ ∑ ∑ (𝑍ℎ,𝑖,𝑘

𝑃𝑟𝑒𝑑(𝑃) − 𝑍ℎ,𝑖,𝑘𝐸𝑥𝑝

𝑍ℎ,𝑖,𝑘𝐸𝑥𝑝 )

2𝑛𝑘

𝑘=1

𝑛𝑖

𝑖=1

𝑛ℎ

ℎ=1

where 𝑃 = [𝜌𝑁.𝑚𝑎𝑥, 𝐾∗, 𝑛, 𝜙𝑁]

Eq. (22)

Here, G is the objective function (i.e. the squared relative error), P is a vector with the

kinetic parameters to be estimated, and Z is a vector with the predicted (subject to P) or

the experimental variables of the model. nh, ni, and nk denote the number of data points in

time, the number of datasets used for fitting, and the number of state variables,

respectively. The resulting parameter values obtained from this fitting protocol, as well as

all other parameter employed in the proposed model, are included in Table 1 along with

their definitions and corresponding units.



202

Table 1. List of kinetic parameters and values employed in the updated model to

account for batch and fed-batch cultivation [31].

Type Symbol Parameter description Value Units

Ass

oci

ated

to

bio

mas

s gro

wth

µmax Maximum specific growth rate 0.106 h-1

qN,0 Minimum nitrogen quota 0.876 gN gC-1

Ks,A Acetate saturation constant 1.789 gC L-1

ki,A Acetate inhibition constant 0.110 gC L-1

Ks,I Light saturation constant 1.4 µmol m-2s-1

ki,I Light inhibition constant 186.5 µmol m-2s-1

YX/A Acetate yield coefficient 0.059 gC gC-1

Ϭ Light attenuation coefficient 1 L gC-1 m-1

Ass

oci

ated

to

Nit

rogen

upta

ke

ρN,max Maximum N uptake rate a 34.56 gN gC-1h-1

K* Saturation constant, No a 0.311 gN L-1

n Shape-controlling parameter a 19.10 -

ФN N uptake regulation coefficient a 138.7 L gC-1

Ks,N Uptake saturation constant, N 0.163 gN L-1

ki,N Uptake inhibition constant, N 0.113 gN L-1

Ks,A:N Uptake saturation constant, A:N 1.004 gC L-1

ki,A:N Uptake inhibition constant, A:N 1.098 gC L-1

Ass

oci

ated

to

Sta

rch &

Lip

id f

orm

atio

n

r1 Starch formation rate (R1) 0.058 gC gC-1

Ks,S Saturation constant (R1) 0 gN L-1

ki,S Inhibition constant (R1) 0.205 gN L-1

nS Shape parameter (R1) 4.17 -

k1 Regulation constant (R1) 0.108 -

ФS Regulation coefficient (R1) 0.775 L gC-1

r2 Starch degradation rate (R2) 0.005 gC gC-1

ksat,S Starch saturation constant (R2) 0.018 -

r3 Lipid formation rate (R3) 0.191 gN gC-1h-1

Ks,L Saturation constant (R3) 0.012 gN L-1

ki,L Inhibition constant (R3) 0.091 gN L-1

nL Shape parameter (R3) 2.01 -

k2 Regulation constant (R3) 0.153 -

ФL Regulation coefficient (R3) 0 L gC-1

r4 Lipid degradation rate (R4) 0.007 gN gC-1h-1

ksat,L Lipid saturation constant (R4) 0.079 -

pH KH pH coefficient b 4.65 L gC-1 h-1 a Parameters values were refined from those previously obtained in [25].



203

4. Results and Discussion.

4.1. Effect of acetic acid pulses on biomass, starch and lipids.

Fed-batch cultivation experiments were carried out by supplementing one or two pulses of

acetic acid at the 4th and 9th day of cultivation, respectively. The pulses increased the

residual medium concentration of acetic acid in a range of + 0.1 to + 0.6 gC L-1. Cultures

subject to one pulse were fully harvested at the 9th day of cultivation, whereas cultures

subject to two consecutive pulses were harvested at the 13th day of cultivation and cells

were analysed for their cell dry weight, and starch and lipid content. For comparison and

statistical analysis, a control culture was grown in standard batch mode. The data obtained

from these experiments are shown in Figure 1 and Figure 2.

The culture grown in batch mode (no pulses) attained a biomass concentration of 0.305 gC

L-1 and accumulated 5.15 % and 14.94 % of its dry weight as starch and lipids, respectively.

When compared against the batch case (Figure 1.a), the cultures subject to one pulse of

acetic acid attained higher biomass concentrations up to a maximum value of 0. 459 gC L-

1 (50 % more than batch), after which biomass concentration decreased gradually with

increasing acetic acid pulse concentration to a value of 0.26 gC L-1 biomass (15 % less

than batch). Although in our previous work we observed that high concentrations of acetic

acid were inhibitory for the growth of C. reinhardtii, such concentration was at least higher

than 1.26 gC L-1 [25]. Thus, the increases in acetic acid medium concentration caused by

the addition of pulses (from P1 = 0.1 gC L-1 to P1 = 0.6 gC L-1) were not sufficiently high

to be considered inhibitory. The reduction in biomass with increasing pulse concentration

may instead be explained by: i) an inhibitory effect of potassium hydroxide (the buffering

agent employed to neutralise the pulses) which increased correspondingly with acetic acid

concentration, ii) the reduction of the natural buffering capacity of the culture solution due

to the gradual consumption of Tris-base, a biochemical component present in standard

TAP medium, or iii) the insufficient mixing of the medium after pulse addition since this

protocol was carried out off-line.



204

Figure 1. Effect of acetic acid pulses on: a) biomass (CDW) and b) biomass

composition in C. reinhardtii CCAP 11/32C. Batch culture and cultures subject to

various [P1] were harvested at day 9. Cultures subject to *[P1] and various [P2]

were harvested at day 12. Results and S.D. are the mean of two biological replicates.

Stars denote significant differences (p < 0.05*, 0.01**, 0.001***) with respect to the

batch culture.

For the cultures subject to two consecutive pulses, the first pulse was supplemented (at day

4) using the pulse that yielded the highest biomass in the single-pulse experiments, which

corresponded to P1=0.2 gC L-1 of acetic acid. The cultures were then supplemented with

[P1] at day 4 *[P1]+[P2] at day 9Batch


****

**

* **

**



205

the second pulse at the 9th day, with concentrations ranging from P2=0.1 to P2=0.3 gC L-1

of acetic acid. As observed in Figure 1.a, all three cultures subject to the two consecutive

pulses attained significantly higher concentrations than the batch culture (p<0.01, as per

one-way anova). The highest biomass concentration of 0.593 gC L-1 was attained by the

culture with a pulse of P2=0.2 gC L-1, which corresponded to a 29 % increase with respect

to that obtained with one pulse, and a 94 % increase with respect to batch conditions.

A similar inhibitory trend with increasing acetic concentration was observed in two-pulse

experiments, but as explained above, this was deemed a consequence of the pulse

preparation and injection protocols rather than of the acetic acid itself. Thus, potentially

higher biomass concentrations could be attained by implementing an improved on-line

system suitable for the intermittent (or continuous) feeding of acetic acid at higher

concentrations than those evaluated here. For example, a recent fed-batch system for the

growth of C. reinhardtii CC-2937 consisting of a semi-continuous on-line feeding of acetic

acid coupled with pH control (maintained between 6.9 and 7.1) yielded a biomass density

of 23.69 g L-1 after a period of 168 h, which was much higher than that of the culture grown

in batch which attained a biomass density of 2.33 g L-1 after a period of 123 h [21].

Meanwhile, in another study that explored the heterotrophic fed-batch growth of C.

reinhardtii (CS-51), it was observed that a 1.8-fold increase in biomass (with respect to

batch) was attained by increasing by 4-fold the concentration of acetic acid in the feed [33].

Regarding biomass composition (Figure 1.b), both starch and lipids remained constant or

increased slightly following the addition of the acetic acid pulses. Up to 18.15 % of lipids

were accumulated by the cultures subject to P1=0.2 gC L-1 of acetic acid, whereas up to

6.76 % of starch was accumulated by the culture subject to P1=0.3 gC L-1 of acetic acid.

In the cultures subject to two pulses, starch and lipid contents increased further up to 8.46

% (in P2=0.2 gC L-1) and 20.92% (in P2=0.3 gC L-1), respectively. The starch contents

attained by all the cultures subject to two pulses were significantly different to those

attained by batch conditions (p <0.05 for P2=0.1, and p < 0.01 for P2=0.2 and P2=0.3, as

per one-way ANOVA), but the lipid contents were not. The increase in starch and lipid

contents following pulse addition was attributed to the gradual consumption of the nitrogen

(and phosphorus) sources over the cultivation period, leading to nutrient starvation which

is well-known to induce storage molecule accumulation [5].



206

Figure 2. Effect of acetic acid pulses on: a) starch and b) lipid concentrations in C.

reinhardtii CCAP 11/32C. Batch culture and cultures subject to various [P1] were

harvested at day 9. Cultures subject to *[P1] and various [P2] were harvested at day

12. Results and S.D. are the mean of two biological replicates. Stars denote

significant differences (p < 0.05*, 0.01**, 0.001***) with respect to the batch

culture.



**

******

***

******

*

**

*** ***



207

As observed in Figure 2, the increase in biomass concentrations coupled with the

corresponding increase in starch and lipid contents led to significant increases in the starch

and lipid medium concentrations. In the single-pulse cultures, P1=0.2 gC L-1 yielded the

highest starch (0.03 gC L-1) and lipid (0.083 gC L-1) concentrations, which corresponded

to 92 % and 83% more starch and lipids than those obtained in batch mode, respectively.

Meanwhile, in the cultures subject to two pulses the maximal starch (0.049 gC L-1) and

lipid (0.122 gC L-1) concentrations were attained by the culture subject to P2 = 0.2 gC L-1.

These concentrations corresponded to 65 % and 46 % more starch and lipids than those

obtained by a single pulse, respectively; and 217 % and 167 % more starch and lipids than

those obtained in batch conditions, respectively. It is worth noting that the increase in lipid

concentration in the 2-pulse strategy (167 %) is much higher than that obtained under the

lipid-optimised batch scenario identified in Figueroa-Torres et al. (2017), where a 66 %

increase in lipids was attained. The latter indicates that lipid production is strongly linked

to biomass growth and can thus similarly benefit by biomass-enhancing strategies such as

fed-batch cultivation.

4.2. Modelling of fed-batch dynamics.

The experimental dynamics of the microalgal cultures subject to pulses of acetic acid was

obtained by harvesting and analysing samples at different intervals before and after the

addition of pulses. To simulate fed-batch dynamics, our previously developed model for

batch algal growth, consisting of 8 state variables (Eq. 1 – Eq. 7, and Eq. 21) and 33 kinetic

parameters (Table 1), was adapted in line with experimental observations. A set of 4

kinetic parameters associated to the maximum nitrogen uptake rate (i.e. 𝜌𝑁,𝑚𝑎𝑥, 𝐾∗, 𝑛 and

𝜙𝑁) were refined to account for the adapted model’s structure.

When using the re-identified parameter values, the adapted model’s predictive capacity for

microalgal batch growth dynamics was observed to be maintained under cultivation

scenarios subject to standard (TAP medium) conditions, low nitrogen, or high acetic acid

concentration regimes (see Supplementary Information). Therefore, the model’s adapted

structure was preserved, and fed-batch dynamics were evaluated. However, it should be

noted that since the model’s structure contains a large number of kinetic parameters, the

parameter values obtained by the optimisation-based fitting methodology may not



208

necessarily be the true values and should rather be regarded as a set of estimates that yield

a satisfactory fit to the available data.

The predicted cultivation dynamics for the culture subject to a single-pulse of acetic acid

(P1=0.2 gC L-1) supplemented at day 4 (𝑡 = 100 h) are presented in Figure 3, where it is

observed that the predicted dynamics of the cultivation variables are in fair agreement with

the data. However, the predictions for nitrogen, starch, and lipids deviated from the trends

observed experimentally. As explained previously in Section 3.2, the model was re-

structured to account for the observed ability of cells to consume more nitrogen after a

pulse was supplemented, contrary to the batch culture where nitrogen consumption stopped

completely. Nevertheless, although the updated equation for nitrogen uptake (Eq. 20)

allowed the model to replicate this observation (Figure 3.b), the model predicts a sharp

decrease in nitrogen consumption which does not match with the slower trend measured

by experimentation. Given that the modelling equations for starch (Eq. 5) and lipid (Eq. 6)

formation are strongly dependent on nitrogen (both extracellular and intracellular), the

deviation of the predicted nitrogen dynamics was deemed responsible for the

disagreements observed in the concentration profiles of the storage molecules (Figure 3.e

and Figure 3.f).

It was observed in our previous study that cells inoculated in fresh medium had a large

capacity for nitrogen consumption [25]. This phenomenon in which algal cells initially

consume rapidly large amounts of nutrients (which then disseminate into new generations

of cells) has been referred to as luxury consumption and thought to be coupled to

intracellular nutrient levels [34]. For example, proteins and enzymes associated to nutrient

transport and assimilation across cell membranes may be either supressed when the

intracellular nutrients are sufficient, or activated when nutrient levels are low [35]. The

observed slower consumption of nitrogen following the addition of the pulse, which

occurred days after inoculation, may thus be explained by the cells having already attained

the specific levels of intracellular nitrogen satisfying their growth.



209

Fig

ure

3.

Co

mp

ari

son

bet

wee

n t

he

pre

dic

ted

co

nce

ntr

ati

on

tim

e-p

rofi

le (

lin

es)

an

d e

xp

erim

enta

l d

ata

(p

oin

ts)

usi

ng t

he

ad

ap

ted

fed

-batc

h m

od

el. D

ata

an

d S

.D. are

th

e m

ean

of

two b

iolo

gic

al

rep

lica

tes

ob

tain

ed f

rom

cu

ltu

res

of

C.

rein

hard

tii

gro

wn

in

batc

h o

r su

bje

ct t

o o

ne

pu

lse

of

ace

tic

aci

d (

wh

ere

P1

= 0

.2 g

C L

-1).



210

It is worth noting that in our proposed model, the parameters identified to be

responsible for regulating the magnitude of nitrogen consumption are the maximum

nitrogen uptake rate, 𝜌𝑁,𝑚𝑎𝑥 , and the nitrogen uptake coefficient, 𝜙𝑁 . Whilst these

parameters were fitted to account for both batch and fed-batch dynamics, the deviation

in the model from the experimental trend may indicate that these parameters should

have a set of values used only for batch conditions, and a different set used specifically

for fed-batch (i.e. after pulse addition). Nevertheless, this was deemed inappropriate

as having a unified model is desired for major bioprocess applications such as scale-

up, control, or optimisation.

The model was also assessed for its capacity to predict the dynamics of cultures subject

to two pulses of acetic acid, with a first pulse of P1=0.2 gC L-1 supplemented at day 4

(𝑡 = 100 h), followed by second pulse of P2=0.2 gC L-1 supplemented at day 9 (𝑡 =

216 h). However, although the model was able to predict the increase in biomass with

one pulse, it failed to simulate the biomass increase obtained after two pulses (Figure

4.a), which in turn affected the predicted dynamics of all other state variables. This

particular limitation of the model was deemed to be a direct consequence of the

predicted nitrogen quota reaching its minimum subsistence value (estimated as 𝑞𝑁,0 =

0.876 gN L-1, see Table 1), thus preventing the model from simulating further growth.

As observed in Figure 5.b, the predicted nitrogen quota reaches the minimum

subsistence value shortly after the 2nd pulse is added to the system. Whilst the model

accurately follows the behaviour predicted by the Droop formulation (with cellular

growth stopping when 𝑞𝑁 = 𝑞𝑁,0), the fact that cells continued to grow further after

this point can be a consequence of more complex intracellular metabolic processes

which the model does not account for. For example, it is observed that the low

extracellular nitrogen concentrations drop further after the second pulse (Figure 4.b).

Under this stressed environment, the acetic acid pulse may stimulate growth processes

by making cells switch their carbon metabolism to a heterotrophic fixation.

On the other hand, it is reported that nutrient-stressed cells of C. reinhardtii can

increase enzymes responsible for the degradation of nitrogen-containing (or

phosphorous-containing) molecules such as proteins and nucleic acids [12,36], which

may allow them to replenish their nutrient quotas and thus assimilate additional carbon



211

sources. A thorough analysis of these potential scenarios is thus required to enhance

the model’s predictive capacity.

Figure 4. Comparison between the predicted concentration time-profile (lines)

and experimental data (points) using the adapted fed-batch model. Data and

S.D. are the mean of two biological replicates obtained from cultures of C.

reinhardtii grown in batch or subject to one or two pulses of acetic acid (where

P1=P2= 0.2 gC L-1).

In line with the above, and to measure the potential increase in biomass by a three-

pulse strategy, an additional experiment was carried out by subjecting cultures to three



212

consecutive pulses of acetic acid, where P1=P2=P3=0.2 gC L-1, supplemented at days

4th (𝑡 = 100 h), 9th (𝑡 = 216 h), and 14th (𝑡 = 336 h), respectively.

Figure 5. Comparison between the predicted concentration time-profile (lines)

and experimental data (points) using the adapted fed-batch model. Data and

S.D. are the mean of two biological replicates obtained from cultures of C.

reinhardtii grown in batch or subject to two consecutive pulses of acetic acid

(where P1=P2=0.2 gC L-1). The y axis of the nitrogen quota is zoomed to

improve readability of the data and the minimum subsistence quota.

The biomass (as well as starch) concentration profiles obtained from this experiment

are presented in Figure 6. Both biomass and starch concentrations increased further

after the addition of a third pulse, with biomass reaching a concentration of 0.64 gC L-

1 (126 % higher than batch) and starch reaching a concentration of 0.047 gC L-1 (67 %

higher than batch). Although the increase in biomass was favourable, the difference in

biomass with respect to the second pulse was only 0.06 gC L-1, much lower than the

difference between batch and a single pulse (0.157 gC L-1), and between the first and

second pulses (0.143 gC L-1). This indicates that biomass begins to reach a saturation

point due to the exhaustion of other essential nutrients in the medium. In particular,



213

the gradual imbalance of the nitrogen:carbon (or phosphorous:carbon) ratio in the

nutrient-starved cells will prevent further pulses to be assimilated. An improved fed-

batch nutrient feeding strategy will thus have to take into account such optimal nutrient

balance by additionally supplying nitrogen or phosphorous sources to replenish the

intracellular nutrient pools.

Figure 6. Biomass and starch concentrations attained by cultures grown in

pulse-assisted fed-batch mode. Data and S.D. are the mean of two biological

replicates obtained from cultures of C. reinhardtii grown in batch or subject to

two consecutive pulses of acetic acid (where P1=P2=P3=0.2 gC L-1).



214

5. Conclusions.

A fed-batch cultivation strategy for increased biomass formation was evaluated with

C. reinhardtii subject to a nutrient feeding strategy involving intermittent pulses of

acetic acid. The biomass concentration attained by the cultures subject to a single or

two consecutive pulses was observed to increase significantly by 50 % and 94 %,

respectively, with respect to the batch case. Meanwhile, the starch and lipid

concentrations of the cultures subject to two pulses of acetic acid increased by 218 %

and 168%, respectively, with respect to the batch case. A kinetic model was

additionally adapted to fed-batch operation, but its application was observed to be

restricted to a single-pulse scenario. The major limitation of the model was its inability

to portray the observed growth of microalgal biomass despite cells reaching their

minimum nitrogen subsistence quota, which suggested the occurrence of more

complex biological processes taking place during extended periods of nutrient

starvation. However, the significant increases in biomass, starch, and lipids attained

by the simple pulse feeding strategy employed in this work highlighted its potential

use as a suitable cultivation strategy targeting biofuels production.

Acknowledgements.

GMFT kindly acknowledges the financial support of the National Mexican Council

for Science and Technology (CONACyT).

Declarations of interest: none.

References.

[1] M.J. Groom, E.M. Gray, P.A. Townsend, Biofuels and Biodiversity:

Principles for Creating Better Policies for Biofuel Production, Conserv. Biol.

22 (2008) 602–609. doi:10.1111/j.1523-1739.2007.00879.x.

[2] L.A. Ribeiro, P.P. da Silva, Surveying techno-economic indicators of

microalgae biofuel technologies, Renew. Sustain. Energy Rev. 25 (2013) 89–

96. doi:10.1016/j.rser.2013.03.032.

[3] M.A. Scaife, A. Merkx-Jacques, D.L. Woodhall, R.E. Armenta, Algal biofuels

in Canada: Status and potential, Renew. Sustain. Energy Rev. 44 (2015) 620–

642. doi:10.1016/j.rser.2014.12.024.

[4] T. Suganya, M. Varman, H.H. Masjuki, S. Renganathan, Macroalgae and



215

microalgae as a potential source for commercial applications along with

biofuels production: A biorefinery approach, Renew. Sustain. Energy Rev. 55

(2016) 909–941. doi:10.1016/j.rser.2015.11.026.

[5] G. Markou, I. Angelidaki, D. Georgakakis, Microalgal carbohydrates: an


bioconversion technologies for production of biofuels, Appl. Microbiol.

Biotechnol. 96 (2012) 631–645. doi:10.1007/s00253-012-4398-0.

[6] L. Rodolfi, G. Chini Zittelli, N. Bassi, G. Padovani, N. Biondi, G. Bonini,

M.R. Tredici, Microalgae for oil: strain selection, induction of lipid synthesis

and outdoor mass cultivation in a low-cost photobioreactor., Biotechnol.

Bioeng. 102 (2009) 100–12. doi:10.1002/bit.22033.

[7] I. Brányiková, B. Maršálková, J. Doucha, T. Brányik, K. Bišová, V.

Zachleder, M. Vítová, Microalgae—novel highly efficient starch producers,

Biotechnol. Bioeng. 108 (2010) 766–776. doi:10.1002/bit.23016.

[8] B.J. Cade-Menun, A. Paytan, Nutrient temperature and light stress alter

phosphorus and carbon forms in culture-grown algae, Mar. Chem. 121 (2010)

27–36. doi:10.1016/j.marchem.2010.03.002.

[9] G. Dragone, B.D. Fernandes, A.P. Abreu, A. a. Vicente, J. a. Teixeira,

Nutrient limitation as a strategy for increasing starch accumulation in

microalgae, Appl. Energy. 88 (2011) 3331–3335.

doi:10.1016/j.apenergy.2011.03.012.

[10] T. Lacour, A. Sciandra, A. Talec, P. Mayzaud, O. Bernard, Neutral lipid and

carbohydrate productivities as a response to nitrogen status in Isochrysis sp.

(T-ISO; Haptophyceae): Starvation versus limitation 1, J. Phycol. 48 (2012)

647–656. doi:10.1111/j.1529-8817.2012.01154.x.

[11] A.K. Bajhaiya, A.P. Dean, T. Driver, D.K. Trivedi, N.J.W. Rattray, J.W.

Allwood, R. Goodacre, J.K. Pittman, High-throughput metabolic screening of

microalgae genetic variation in response to nutrient limitation, Metabolomics.

12 (2016) 9. doi:10.1007/s11306-015-0878-4.

[12] X. Johnson, J. Alric, Central Carbon Metabolism and Electron Transport in

Chlamydomonas reinhardtii: Metabolic Constraints for Carbon Partitioning

between Oil and Starch, Eukaryot. Cell. 12 (2013) 776–793.

doi:10.1128/EC.00318-12.

[13] S.G. Ball, L. Dirick, A. Decq, J.-C. Martiat, R. Matagne, Physiology of starch

storage in the monocellular alga Chlamydomonas reinhardtii, Plant Sci. 66

(1990) 1–9. doi:10.1016/0168-9452(90)90162-H.

[14] M.A.M. Mirzaie, M. Kalbasi, S.M. Mousavi, B. Ghobadian, Investigation of

mixotrophic, heterotrophic, and autotrophic growth of Chlorella vulgaris

under agricultural waste medium, Prep. Biochem. Biotechnol. 46 (2016) 150–

156. doi:10.1080/10826068.2014.995812.



216

[15] M. Bekirogullari, I.S. Fragkopoulos, J.K. Pittman, C. Theodoropoulos,

Production of lipid-based fuels and chemicals from microalgae: An integrated

experimental and model-based optimization study, Algal Res. 23 (2017) 78–

87. doi:10.1016/j.algal.2016.12.015.

[16] S.-H. Ho, W.-M. Chen, J.-S. Chang, Scenedesmus obliquus CNW-N as a

potential candidate for CO2 mitigation and biodiesel production, Bioresour.

Technol. 101 (2010) 8725–8730. doi:10.1016/J.BIORTECH.2010.06.112.

[17] C.H. Ra, C.-H. Kang, N.K. Kim, C.-G. Lee, S.-K. Kim, Cultivation of four

microalgae for biomass and oil production using a two-stage culture strategy

with salt stress, Renew. Energy. 80 (2015) 117–122.

doi:10.1016/j.renene.2015.02.002.

[18] A. Kiparissides, M. Koutinas, C. Kontoravdi, A. Mantalaris, E.N.

Pistikopoulos, ‘Closing the loop’ in biological systems modeling — From the

in silico to the in vitro, Automatica. 47 (2011) 1147–1155.

doi:10.1016/J.AUTOMATICA.2011.01.013.

[19] M.L. Shuler, F. Kargi, Bioprocess Engineering: Basic Concepts, Prentice Hall,

1992.

[20] C. Jeffryes, J. Rosenberger, G.L. Rorrer, Fed-batch cultivation and bioprocess

modeling of Cyclotella sp. for enhanced fatty acid production by controlled

silicon limitation, Algal Res. 2 (2013) 16–27. doi:10.1016/j.algal.2012.11.002.

[21] F.J. Fields, J.T. Ostrand, S.P. Mayfield, Fed-batch mixotrophic cultivation of

Chlamydomonas reinhardtii for high-density cultures, Algal Res. 33 (2018)

109–117. doi:10.1016/J.ALGAL.2018.05.006.

[22] T. Wang, X. Tian, T. Liu, Z. Wang, W. Guan, M. Guo, J. Chu, A two-stage

fed-batch heterotrophic culture of Chlorella protothecoides that combined

nitrogen depletion with hyperosmotic stress strategy enhanced lipid yield and

productivity, Process Biochem. 60 (2017) 74–83.

doi:10.1016/J.PROCBIO.2017.05.027.

[23] F. Ji, Y. Zhou, A. Pang, L. Ning, K. Rodgers, Y. Liu, R. Dong, Fed-batch

cultivation of Desmodesmus sp. in anaerobic digestion wastewater for

improved nutrient removal and biodiesel production, Bioresour. Technol. 184

(2015) 116–122. doi:10.1016/j.biortech.2014.09.144.

[24] A. Velayudhan, Overview of integrated models for bioprocess engineering,

Curr. Opin. Chem. Eng. 6 (2014) 83–89. doi:10.1016/J.COCHE.2014.09.007.

[25] G.M. Figueroa-Torres, J.K. Pittman, C. Theodoropoulos, Kinetic modelling of

starch and lipid formation during mixotrophic, nutrient-limited microalgal

growth, Bioresour. Technol. 241 (2017) 868–878.

doi:10.1016/j.biortech.2017.05.177.

[26] E.H. Harris, The Chlamydomonas sourcebook : a comprehensive guide to

biology and laboratory use, Academic Press, 1989.



217

[27] A.K. Bajhaiya, A.P. Dean, L.A.H. Zeef, R.E. Webster, J.K. Pittman, PSR1 Is a

Global Transcriptional Regulator of Phosphorus Deficiency Responses and

Carbon Storage Metabolism in Chlamydomonas reinhardtii, Plant Physiol. 170

(2016) 1216–1234. doi:10.1104/pp.15.01907.

[28] M.R. Droop, Vitamin B12 and Marine Ecology. IV. The Kinetics of Uptake,

Growth and Inhibition in Monochrysis Lutheri, J. Mar. Biol. Assoc. United

Kingdom. 48 (1968) 689–733. doi:10.1017/S0025315400019238.

[29] J.F. Andrews, A mathematical model for the continuous culture of

microorganisms utilizing inhibitory substrates, Biotechnol. Bioeng. 10 (1968)

707–723. doi:10.1002/bit.260100602.

[30] O. Bernard, Hurdles and challenges for modelling and control of microalgae

for CO2 mitigation and biofuel production, J. Process Control. 21 (2011)

1378–1389. doi:10.1016/j.jprocont.2011.07.012.

[31] G.M. Figueroa-Torres, J.K. Pittman, C. Theodoropoulos, Optimisation of

microalgal starch and lipid formation via nitrogen and phosphorus co-

limitation, Biotechnol. Bioeng. Submitted (2018).

[32] A. Vlysidis, M. Binns, C. Webb, C. Theodoropoulos, Glycerol utilisation for

the production of chemicals: Conversion to succinic acid, a combined

experimental and computational study, Biochem. Eng. J. 58–59 (2011) 1–11.

doi:10.1016/j.bej.2011.07.004.

[33] X.-W. Zhang, F. Chen, M.R. Johns, Kinetic models for heterotrophic growth

of Chlamydomonas reinhardtii in batch and fed-batch cultures, Process

Biochem. 35 (1999) 385–389. doi:10.1016/S0032-9592(99)00082-5.

[34] M.R. Droop, 25 Years of Algal Growth Kinetics A Personal View, Bot. Mar.

26 (1983) 99. doi:10.1515/botm.1983.26.3.99.

[35] J.A. Bonachela, M. Raghib, S.A. Levin, Dynamic model of flexible

phytoplankton nutrient uptake, Proc. Natl. Acad. Sci. 108 (2011) 20633–

20638. doi:10.1073/pnas.1118012108.

[36] A. Grossman, Acclimation of Chlamydomonas reinhardtii to its Nutrient

Environment, Protist. 151 (2000) 201–224. doi:10.1078/1434-4610-00020.



218



219



presented next.



220


Associated to:

An Experimental and Model-based Analysis of Fed-Batch Microalgal

Cultivation for Biofuels Production

Gonzalo M. Figueroa-Torres a, Jon K. Pittman b and Constantinos Theodoropoulos a,*



b School of Earth and Environmental Sciences, The University of Manchester, Manchester,

M13 9PL







221

5. Statistical analysis.

Table S.1 includes the p-values obtained by the one-way ANOVA analysis (tukey test)

of the experimental data from fed-batch cultivation via a single pulse of acetic acid.

All treatments were compared amongst themselves. Analysis was carried out in Origin

Pro 2017 (b9.4.1.354).

Table S.1. p-values obtained by one-way ANOVA. Highlighted cells denote

significant differences between each treatment pair (p<0.05).

p-values

Treatment pair X (gC L-1) S (gC L-1) L (gC L-1) S (%) L (%)

Batch: Batch - - - - -

Batch: [P1]=0.1 0.34656 0.00103 0.09824 0.70134 0.62476

Batch: *[P1]=0.2 0.08267 4.15E-05 0.01832 0.36649 0.53368

Batch: [P1]=0.3 0.2854 1.37E-04 0.16697 0.24248 0.93101

Batch: [P1]=0.4 0.53015 0.08766 0.98028 1 0.98631

Batch: [P1]=0.6 0.96903 0.73838 0.94613 1 0.99648

Batch: *[P1]+[P2]=0.1 0.00405 1.40E-07 0.00123 0.04686 0.50696

Batch: *[P1]+[P2]=0.2 0.00165 0 1.09E-04 0.00537 0.07599

Batch: *[P1]+[P2]=0.3 0.00466 0 2.15E-04 0.00515 0.05563

6. Model’s predictability.

As mentioned within the main text of the manuscript, a set of 4 kinetic parameter

(from the maximum nitrogen uptake, Eq. 19) were refined to maintain the model’s

predictive capacity for batch dynamics. Figure S.1 shows the model predictions

(using the adapted model for fed-batch) against experimental data obtained from

batch experiments subject to: standard TAP medium, low nitrogen medium, and

high acetate medium.



222

Figure S.1. Comparison between the predicted time-profile (lines) and

experimental data (points) for the cultures grown in: TAP (Ao=0.42 gC L-1,

No=0.3824 gN L-1), Low N (Ao=0.42 gC L-1, No=0.356 gN L-1), and High A

(Ao=1.26 gC L-1, No=0.3824 gN L-1). Data and S.D. is the mean of two

biological replicates as obtained in Figueroa-Torres et al. (2017).

223

Chapter 6

Microalgal Biomass as a Biorefinery Platform for

Biobutanol and Biodiesel Production: A Case Study

6.1. Introduction.

Microalgae-based biorefineries are regarded as the most economically viable approach for

the co-production of fuels and other value-added chemicals by exploiting all potential

biomass conversion routes (Barsanti and Gualtieri, 2018; Suganya et al., 2016). Given the

rich chemical composition of microalgae, the biorefinery concept allows for all potential

side-products and/or waste to be re-valorised, thereby increasing energy efficiency and

process profitability by capitalising from both the “high volume low price” and the “low

volume high price” typical of bioprocessing strategies (Figure 6.1).

Figure 6.1. Market sizes and volumes for conventional microalgal products.

Adapted from Zhu (2015).

The successful implementation of microalgae-based biorefineries, however, requires the

evaluation and analysis of all possible co-production scenarios rather than adopting a

single-route approach. A visual representation of the many microalgal products and their

Chapter 6 – Microalgal Biomass as a Biorefinery Platform for Biobutanol

and Biodiesel Production: A case study

224

corresponding conversion routes is presented in Figure 6.2. It can be observed that the

adequate exploitation of microalgal biomass can yield important energy-rich products such

as syngas, bio-oil, electricity, methane, or liquid biofuels via thermochemical or

biochemical conversion processes.

Among the biochemical routes, anaerobic digestion is already a well-known technology

whereby organic matter is sequentially metabolised by bacteria into a final biogas

comprised of methane and carbon dioxide (Adeniyi et al., 2018). Microalgal biomass, rich

in organic compounds, could be thus converted into methane and recover energy estimated

to be as high as that obtained from lipid extraction. The performance of anaerobic digestion

processes, however, decreases when the organic feed contains a high protein content,

which may be the case in microalgal biomass. Co-digestion of algae and other low-protein

wastes is therefore suggested as a more viable approach (Barsanti and Gualtieri, 2018;

Suganya et al., 2016).

Figure 6.2. A schematic representation of the various microalgal conversion routes

suitable for the co-production of liquid biofuels (highlighted) and other value-added

chemicals. Adapted from Suganya et al. (2016).



225

The other biochemical conversion routes shown in Figure 6.2 correspond to the production

of liquid biofuels: biodiesel production via transesterification, and bioethanol or biobutanol

production via fermentation processes. As stated in the introduction of this thesis (see

Chapter 1), this research aimed to identify microalgal cultivation strategies suitable for

the production of liquid biofuels. Whilst microalgal biodiesel has been the subject of

intensive research, the production of microalgal biobutanol (a sugar-based biofuel with

much superior physical properties than ethanol) has been less explored. Therefore, a case

study was carried out to provide a quantifiable measure of the biorefinery potential for both

biobutanol and biodiesel production from the microalgal biomass employed in this work.

Biobutanol, a promising gasoline-replacement, is biochemically produced by the ABE

fermentation, a biological process catalysed by Clostridium bacterial species. This

fermentation has been studied for over 100 years, and its metabolic pathways are also well

identified (Moon et al., 2016). Despite its renowned potential, however, the ABE

fermentation is still not fit for commercial application due in part to the poor biobutanol

conversion yields that it typically attains (Xue et al., 2014). Such low biobutanol yields

can be addressed through the isolation of super-productive strains (via metabolic

engineering) and/or the development of optimal fermentation technologies.

Taking the above into consideration, the case study that follows focused predominantly on

biobutanol production by first establishing a set of optimal fermentation conditions through

a number of glucose-based experiments and extrapolating them to microalgae-based

fermentations. Microalgal biomass residues from fermentation experiments were then

evaluated for their lipid and biodiesel content. The conversion routes as well as

experimental protocols involved will be described in detail, and the obtained biofuel yields

are also compared against relevant studies available in the literature.



226



227

6.2. Contribution 4. A biofuels production case study.

Figueroa-Torres GM, Wan Mahmood WMA, Pittman JK, Theodoropoulos C. Microalgal

Biomass as a Platform for Biobutanol and Biodiesel Production: A Case Study. To be

submitted to: Biochemical Engineering Journal.


Gonzalo M. Figueroa-Torres performed experimental and computational tasks

associated to this work, analysed data, and wrote the case study. Specifically, the

experimental tasks included: microalgal biomass cultivation and stock collection, ABE

fermentation experiments, and microalgal lipid extraction.

Wan M. Asyraf Wan Mahmood performed experimental tasks associated to this work,

analysed data, and co-wrote Section 2.3.4 and Section 4.2 of this case study. Specifically,

the experimental tasks included: transesterification and FAME composition analysis.

Jon K. Pittman co-supervised the research.

Constantinos Theodoropoulos contemplated and supervised the research, and reviewed

the manuscript.



228

Microalgal Biomass as a Biorefinery Platform for

Biobutanol and Biodiesel Production: A case study

Gonzalo M. Figueroa-Torresa, Wan M. Asyraf Wan Mahmooda, Jon K. Pittmanb,

Constantinos Theodoropoulosa,*




9PL







229

ABSTRACT

Microalgal biofuels have been regarded as the most sustainable energy alternatives to

conventional fossil-based fuels. However, due to the high costs associated to microalgal

cultivation, the economic viability of microalgal biofuels remains uncertain. Microalgal

biorefineries have thus emerged as the most economically viable option for the co-

production of biofuels and waste re-valorisation. Microalgae can be directed towards the

production of both biodiesel (via the transesterification of microalgal lipids) and

biobutanol (via the fermentation of microalgal carbohydrates). Whilst microalgal biodiesel

production has been studied extensively, microalgal biobutanol has received less attention

due to the low product yields of the ABE fermentation, the biochemical process from

which biobutanol is obtained. Therefore, this case study evaluated the potential of a

microalgae-based biorefinery by: i) optimising the ABE fermentation via a surface

response analysis, suitable for biobutanol production from microalgal biomass (in raw and

hydrolised form); and by ii) quantifying the production of biodiesel via transesterification

of microalgal biomass. Product yields of 10.31 % and 10.07 % butanol were attained by

the raw microalgae or the microalgal hydrolysate, respectively, under optimal fermentation

conditions. Meanwhile, the fermented microalgal biomass and the hydrolysed biomass

residues yielded up to 3.29 % and 3.82 % biodiesel, respectively. Results validate

microlgal biorefineries as a viable option for the co-production of biofuels.

Keywords: biobutanol, biodiesel, biorefineries, fermentation, microalgae.



230

1. Introduction.

Microalgal biofuels, so-called third generation biofuels, have been praised due to their

potential to become sustainable replacements for fossil-based fuels, large contributors to

greenhouse gas emissions and crude oil depletion. The production of microalgal biofuels

is estimated to require less land, water, and fertiliser usage than first generation food-based

biofuels (e.g. ethanol from corn or sugarcane, biodiesel from rapeseed oil) or second

generation lignocellulosic-based biofuels (e.g. butanol from wheat straw) (Groom et al.,

2008; Qureshi et al., 2007; Suganya et al., 2016). However, whilst the production of third

generation biofuels has been proven technically feasible, their financial viability and

commercial success remains unclear (Richardson et al., 2012; Zhu, 2015).

Microalgal biofuels naturally require the mass-scale generation of microalgal biomass, but

this is currently restricted by the high costs and energy required to first cultivate and

subsequently harvest biomass, which translates into increased biofuel prices that are

uncompetitive against those of well-established fossil fuels (Hariskos and Posten, 2014).

A competitive way of maximising the profitability of third generation biofuels is thus

through the implementation of microalgal biorefineries, where the production of biofuels

as well as any other value-added chemicals is fully exploited (Barsanti and Gualtieri, 2018;

Suganya et al., 2016; Trivedi et al., 2015).

The biorefinery concept offers plenty of opportunities given that microalgae’s chemical

composition is rich in major carbon-based compounds (e.g. carbohydrates, lipids, proteins,

vitamins) which can act as the raw precursors for a large variety of human health-related

products (e.g. vitamins and antioxidants), but also biofuels production (Chen et al., 2013;

Enamala et al., 2018). However, before microalgal biorefineries become commercially

successful it is necessary to evaluate and optimally integrate all possible bioprocessing

conversion and/or waste re-valorisation scenarios (Hariskos and Posten, 2014).

Microalgal lipids can be directed towards biodiesel production via the transesterification

reaction, where tryacilglycerides are converted into fatty acid methyl esthers (i.e.

biodiesel). Meanwhile, microalgal carbohydrates can be directed towards bioethanol or

biobutanol production via fermentative processes (Trivedi et al., 2015). The potential for

biodiesel and bioethanol production from microalgae has been well documented and



231

investigated in literature (Chisti, 2007; Harun et al., 2010; Harun et al., 2014; Kim et al.,

2014; Lage and Gentili, 2018; Liu et al., 2011; Rojan P et al., 2011; Velasquez-Orta et al.,

2014; Xu et al., 2006), but microalgal biobutanol has received less attention (Efremenko

et al., 2012; Ellis et al., 2012; Wang et al., 2016).

Biobutanol is a much superior biofuel than bioethanol since it can be blended with fuels at

much higher ratios without requiring modifications to engines (Bankar et al., 2013).

However, biobutanol production is severely restricted due to the low product yields

attained by the ABE fermentation, the biochemical route from which biobutanol is

produced, along with acetone and ethanol (García et al., 2011). Thus, before biobutanol

production (whether from microalgae or any other biofuel feedstock) becomes an

economically viable alternative to the most dominant bioethanol, the low ABE

fermentation yields should be improved.

The ABE fermentation is carried out by microbial species of the genus Clostridia in two-

stages: acidogenesis and solventogenesis. During the acidogenesis stage, carbohydrate rich

biomass is first metabolised into organic acids (e.g. acetic acid and butyric acid) and causes

pH levels to drop, which in turn induces the start of the solventogenic phase where

accumulated acids are concerted into the final ABE products (Köpke and Dürre, 2011;

Kumar and Gayen, 2011). Manipulation of the fermentation media (e.g. type of carbon

substrate) including external supplementation of acetic acid and butyric acid (i.e. the major

precursors for solventogenesis) have been suggested as strategies for improving butanol

yields, but provided that an optimal composition is previously identified (Al-Shorgani et

al., 2018; Matta-El-Ammouri et al., 1987; Zhou et al., 2018).

Therefore, to evaluate the potential for biobutanol and biodiesel production within a

microalgal biorefinery framework, this case study aimed to quantify the co-production of

microalgal butanol via the ABE fermentation and the production of microalgal biodiesel

via the transesterification reaction. Analysis associated to biobutanol production involved:

i) the identification of an optimal fermentation media composition subject to the

concentrations of acetic acid, butyric acid, and nitrogen, via glucose-based

experimentation; and ii) the subsequent evaluation of microalgal biomass as a fermentation

substrate in raw or hydrolysed form. The analysis associated to biodiesel production



232

involved the extraction, quantification, and profiling of lipids from the microalgal biomass,

before and after its use as a fermentation substrate.


2.1. Preparation of microalgal biomass.

The microalgal biomass employed in this study originated from Chlamydomonas

reinhardtii CCAP 11/32C. The strain was grown mixotrophically in Tris-Acetate-

Phosphate (TAP) medium (Harris, 1989) under the environmental conditions described in

(Figueroa-Torres et al., 2017). Microalgal cultures were harvested at the 7th day by

centrifuging cells for 10 min at 7,500 rpm in an Avanti J-26S XP centrifuge (Beckman

Coulter). The pelleted cells were dried for 24 hours at 70 °C and kept in sealed containers.

The process was repeated successively until a stock of ~ 25 g of dried biomass was

collected. Prior to analyses, dried cells were pulverised manually with mortar and pestle

with intermittent supply of liquid nitrogen. Pulverised cells were washed in 70 % ethanol

to remove chlorophyll pigments and allowed to dry a second time (24 h, 70 °C). The

resulting microalgal biomass (MB) was then used as required in ABE fermentations and

transesterification. The biofuel routes evaluated in this case-study are presented in Figure

1.

2.2. ABE Fermentation.

2.2.1. Strain and maintenance.

ABE fermentation experiments were carried out with Clostridium acetobutylicum DSM

792 (purchased from the Leibniz Institute DSMZ-German Collection of Microorganisms

and Cell cultures). The strain was delivered in freeze-dried from and activated as per

manufacturer’s instructions. The active strain was then incubated anaerobically in 100 mL

of Reinforced Clostridial Medium (RCM, Sigma-Aldrich®) at 37 °C for a period of 48 h.

The culture was preserved as: i) glycerol stocks kept at -80 °C, and ii) agar plates (solid

RCM) kept at 37 °C in an anaerobic jar. An active culture of the stain was maintained via

weekly inoculation into fresh sterile RCM medium.



233

2.2.2. Batch fermentation.

All fermentation experiments were performed in batch in 125 mL serum glass bottles (54

mm x 107 mm, Wheaton) containing 100 mL of P2 medium (g L-1): 60 glucose, 0.2

MgSO47H2O, 0.01 MnSO4H2O, 0.01 FeSO47H2O, 0.25 K2HPO4, 0.25 KH2PO4, 2

NH4CL, 1 yeast extract, and 5 CaCO3 (Raganati et al., 2015). The medium employed either

glucose or microalgal biomass as carbon source. Prior to sterilisation, the medium pH was

adjusted to a starting value of 6.5, using KOH 3M or HCL 3M, as required. The nitrogen

sources (NH4Cl and yeast extract) were prepared and sterilised separately from all other

P2 medium components, and then mixed aseptically upon cooling to room temperature. To

create anaerobic conditions, sterile P2 medium was flushed with oxygen-free nitrogen gas

for 5 min, after which bottles were immediately capped with rubber stoppers and

aluminium caps. To release the pressure from accumulated fermentation gasses, all capped

bottles were connected to a water trap. Fermentation was initiated by inoculating P2

medium with 5 mL of an active Clostridial culture previously grown on RCM at 37 °C for

20 h. To measure the cell growth and metabolites, 2 mL samples of the fermentation broth

were removed aseptically at regular times and stored at -20 °C for analysis. Data was

statistically analysed by one-way ANOVA in Origin Pro 2017 (b9.4.1.354).

Glucose-based fermentations were carried out to identify key factors affecting biobutanol

production. The fermentation conditions and components that were evaluated included:

release of accumulated gases and supplementation of CaCO3, and the effects of different

butyric acid (𝐵𝐴0 = 2, 4, and 8 g L−1), acetic acid (𝐴𝐴0 = 2, 4, and 8 g L−1), and nitrogen

(𝑁0 = 4 and 6 g NH4Cl L−1), medium concentrations. In nitrogen-dependent experiments,

only the concentration of ammonium chloride (NH4Cl) was varied, whilst keeping yeast

extract concentration constant. All other P2 medium components remained unchanged.

Microalgae-based fermentations were carried out under the standardised glucose-based

fermentation conditions, but using microalgal biomass (MB), or microalgal biomass

hydrolysate as the main carbon substrate. In fermentations with MB, medium was prepared

by replacing glucose with microalgal biomass, and autoclaved at 120 °C for 20 min.

Microalgal biomass was supplemented at a concentration of 10 g L-1. In fermentations with

microalgal hydrolysate, hydrolysis was performed by mixing 10 g of MB in 150 mL of 4

% (w/v) sulfuric acid (H2SO4), followed by autoclaving at 120 °C for 20 min. The



234

hydrolysed microalgal biomass (MB-H) residues were separated by centrifugation (7,500

rpm, 10 min). The hydrolysate was neutralised with CaCO3, and used in fermentation along

with all other P2 medium components.

2.3. Analytical Methods.

2.3.1. Cell growth.

The growth of C. acetobutylicum DSM 792 was quantified by measuring the optical

density (OD) at 680 nm. Samples taken from the fermentation broth were diluted in

distilled water (2:10) and measured in a UVmini-1240 spectrophotometer (Shimadzu). The

OD was co-related to the cell dry weight by 1 OD = 0.4 g CDW (Raganati et al., 2015).

The residual medium pH was measured off-line in a bench-type HI 2211 pH meter (Hannah

Instruments).

2.3.2. Fermentation substrate and metabolites.

The residual glucose concentration was measured by High Pressure Liquid

Chromatography (HPLC) in a Dionex Ultimate 3000 instrument, using an Aminex HPX-

87H (300 x 7.8 mm) column coupled to an RI detector. Sulphuric acid (H2SO4) was

employed as mobile phase at a flow rate of 0.6 mL min-1 and a temperature of 50 °C. The

residual concentrations of acetic acid, butyric acid, acetone, butanol, and ethanol, were

measured by Gas Chromatography in an Agilent 7820A system coupled to a Flame

Ionisation Detector (GC-FID). Analysis was carried out in a Poraplot Q-HT fused silica

column (10 m x 0.32 mm) using helium as a carrier gas (160 kPa). The temperature of the

detector was set to 250 °C. The oven temperature program was set initially isothermal at

90 °C for 2 min, then ramped up to 200 °C at a rate of 10°C min-1, and finally kept

isothermal at 200 °C for 4 min. The injection volume and temperature were 2 μL (splitless

mode) and 225 °C, respectively. The concentration of all metabolites (substrate, organic

acids and solvents) was measured from the area of the chromatographic peaks (from either

HPLC or GC) using calibration standards of known composition. All samples and

calibration standards were diluted in Type-1 grade water and filtered in 0.45 m

nitrocellulose membrane filters.



235

2.3.3. Microalgal lipid contents.

The microalgae employed in this work was evaluated for its biodiesel potential by

evaluating the lipid content and composition of three different biomass conditions: i) the

washed microalgal biomass (MB), ii) the hydrolysed microalgal biomass (MB-H), and also

iii) the washed microalgal biomass (non-hydrolysed) post-fermentation (MB-F). The lipid

content of all three microalgal biomass conditions (MB, MB-H, and MB-F) was quantified

by solvent extraction in a ST 243 SOXTEC unit (FOSS), employing a three-stage

extraction protocol (extraction, rinsing, and evaporation) as in (Bekirogullari et al., 2017).

Hexane was used as extracting solvent at a temperature of 155 °C. Prior to extraction, dried

pulverised samples of the microalgal biomass were placed in 25 x 60 mm cellulose

extraction thimbles (Whatman®). The Bligh and Dyer method was used as a reference

method for the quantification of Total Lipids (Bligh and Dyer, 1959).

2.3.3. Microalgal lipid composition.

To analyse lipid composition, the crude lipids extracted from each microalgal biomass

condition were subject to a base-catalysed room temperature transesterification as

suggested by other studies (Orr et al., 2016; Wan Mahmood et al., 2017). Briefly: i)

extracted lipids were diluted in 5 mL hexane and mixed with 1 mL of freshly prepared 2M

methanolic KOH solution, ii) the mixture was shaken for 10 minutes at 50 rpm and phase

separation was then achieved by centrifugation at 4000 rpm for 15 minutes, iii) 1 mL of

distilled water was added to the mixture to dissolve any unreacted methanolic KOH and

other impurities, and the mixture was left standing for 2 hours to achieve a biphasic layer,

iv) the organic layer was evaporated and reconstituted with 1 mL of dichloromethane,

suitable for gas chromatography analysis. After transesterification, the profiling and

quantification of the fatty acid methyl esthers (FAMEs) of all microagal conditions was

carried out in a gas chromatography unit (Shimadzu) with mass spectrometry detection

(GC-MS). Analysis was performed in a BPX70 column (60 m x 0.25 mm x 0.25mm) using

helium as a carrier gas (1.5 mL min-1). The temperature of the detector was set to 250 °C.

The oven temperature program was set initially isothermal at 100 °C for 1 min, then

ramped up to 250 °C at a rate of 5°C min-1, and finally kept isothermal at 250 °C for 10

min. The injection volume and temperature were 1 μL (split mode 1:50) and 250 °C,



236

respectively. For FAME quantification, an internal standard of tripentadecanoin (in TAG)

was used, incorporated into all microalgal samples prior to crude lipid extractions.

3. Response surface analysis (RSA) for ABE solvents.

The experimental results obtained from the glucose-based fermentations were employed

to evaluate the influence of butyric acid, BA, and acetic acid, AA, initial medium

concentrations on the production of acetone, butanol, and ethanol. Response surfaces were

generated by fitting second-order polynomial equations to experimental data via multiple

regression analysis. The two-factor quadratic model equations for each of the fermentation

products were expressed as in Eq. 1:

�� = 𝛼0 + 𝛼1 ∙ 𝑋1 + 𝛼2 ∙ 𝑋2 + 𝛼3 ∙ 𝑋12 + 𝛼4 ∙ 𝑋2

2 + 𝛼5 ∙ 𝑋1 ∙ 𝑋2 Eq. (1)

Where �� is the response variable (i.e. acetone, butanol, or ethanol), 𝑋1 and 𝑋2 are the

indepedent variables (i.e. acetic acid and butyric acid), 𝛼0 is the interception coefficient,

𝛼1 and 𝛼2 are the linear regression coefficients, 𝛼3 and 𝛼4 are the quadratic regression

coefficients, and 𝛼5 is the interactive regression coefficient.

3.1. Model fitting.

The regression coefficients, �� = [𝛼0, 𝛼1, 𝛼2, 𝛼3, 𝛼4, 𝛼5], of the quadratic model (Eq. 1)

were estimated by minimising the squared error, E, between predicted, ��, and experimental

values, 𝑦, as in Eq. 2:

min 𝐸 = ∑(��𝑘(��) − 𝑦𝑘)2

𝑛

𝑘=1

Eq. (2)

Here, n represents the number of experimental runs dependent on acetic and/or butyric

acid. Data employed for fitting corresponded to the means obtained from three

experimental replicates. Eq. 2 was minimised by means of an optimisation algorithm

combining both stochastic and deterministic routines (Vlysidis et al., 2011). The

significance of the estimated regression coefficients in Eq. 1 was evaluated by one-way

ANOVA in Origin Pro 2017 (b9.4.1.354).



237

4. Results and Discussion.

The production of microalgal biobutanol and biodiesel via the ABE fermentation and

transesterification, respectively, was evaluated and quantified using microalgal biomass

from the model species C. reinhardtii, according to the routes shown in Figure 1.

4.1. Biobutanol from microalgae.

4.1.1. Glucose-based fermentation.

Glucose-based experiments were first carried out to standardise fermentation protocols and

obtain butanol production yields comparable to those reported in literature. Initial

experiments indicated that the fermentation gasses needed to be released from the system

to avoid a build-up of pressure that could potentially damage cells and prevent solvent

formation. It was also observed that unbuffered medium lead to the drop of pH to values

below 4.5 which prevented the formation of ABE solvents, a phenomenon that has been

referred to as “acid crash” (Bryant and Blaschek, 1988; Wang et al., 2011a). Experimental

data obtained from an unbuffered fermentation is included as Supplementary Information.

To avoid the so-called acid crash, P2 medium was therefore supplemented with CaCO3 at

a concentration of 5 g L-1, as in (Raganati et al., 2015). Although this buffering agent has

not been frequently used (or reported) for batch butanol fermentations, the use of CaCO3

allowed pH to remain above 4.5 which in turn triggered solventogenesis.

After identifying the above fermentation conditions (i.e. gas release and CaCO3

supplementation), subsequent experiments aimed to evaluate the effects of butyric acid

(BA0), acetic acid (AA0), and nitrogen (N0), on the ABE fermentation dynamics. Figure

2 shows the experimental datasets obtained from 4 different fermentations: i) the control

fermentation (P2 medium, where BAo=0 g L-1, AAo=0 g L-1 , and No=2 g NH4Cl L-1), ii)

a fermentation with butyric acid at BAo=4 g L-1, iii) a fermentation with acetic acid at

AAo=4 g L-1 , and iv) a fermentation with nitrogen at No=4 g NH4Cl L-1. All other P2

components remained constant. All other experimental datasets are included as

Supplementary Information.

As observed in Figure 2.a, cell growth followed the typical stages of microbial growth

kinetics (Shuler and Kargi, 1992), with a short stationary phase followed by cell death. The



238

first 20-24 h of fermentation corresponded to the initial lag phase, led by an exponential

phase lasting up to the second day (48 h). Most of the glucose was completely consumed

during this phase (Figure 2.b), which corresponded to the accumulation of organic acids

characteristic of the acidogenic phase (Figure 2.c and Figure 2.d). Fermentation gases

(released from the system) are produced during this acidogenic period and are made up of

H2 and CO2 (Köpke and Dürre, 2011). The accumulation of organic acids is additionally

associated to the initial drop of medium pH (Figure 2.h).

Both the medium pH and the accumulated organic acids play a key role in triggering the

start of solventogenesis (Wang et al., 2011b; Yang et al., 2013). It has been suggested that

for solventogenesis to initiate, the pH of a batch fermentation should fall to within a range

of 4.3 – 5.2 (Yerushalmi et al., 1986). The shift between phases is also associated to a

specific threshold of accumulated organic acids (mostly in undissociated form) which later

permeate through the cell membrane to be reassimilated into the ABE solvents (Bryant and

Blaschek, 1988). However, medium pH and organic acid titres must reach optimal levels

since high concentration of undissociated acids at low pH values have also been known to

induce “acid crash”, whereby only acids are formed at the expense of solvents (Yang et

al., 2013).

The production of ABE solvents started at the 48 h mark and continued up to 100 h, after

which concentrations remained relatively constant. Solvent accumulation is responsible

for the final increase of pH typical of this fermentation. The accumulation of butanol,

particularly, is inhibitory for clostridial growth since it disrupts cellular membranes,

affecting their permeability and in consequence nutrient uptake processes (Kumar and

Gayen, 2011). The inhibitory effects of accumulated butanol can be observed in Figure

2.a, where cell dry weight started to decline with increasing butanol concentration.

The final fermentation titres and yields of all glucose-based fermentations are summarised

in Table 1. The control culture, grown in standard (buffered) P2 medium, attained final

concentrations of 1.88 g L-1 acetone, 12.67 g L-1 butanol, and 0.92 g L-1 ethanol. For

comparison, these concentrations are: i) higher than those reported for batch fermentations

of C. acetobutylicum DSM 792: ~1.9 g L-1 acetone, ~5.9 g L-1 butanol, and ~1 g L-1 ethanol

(Survase et al., 2012), and ii) slightly lower than those obtained by C.



239

saccharoperbutylacetonicum N1-4 grown in batch mode: 3.43 g L-1 acetone, 13.64 g L-1

butanol, and 1.24 g L-1 ethanol (Ellis et al., 2012).

The highest butanol concentration and ABE titres were attained by the culture

supplemented with 4 g L-1 of butyric acid. Supplementation of butyric acid (at 4 g L-1) has

also been shown to increase butanol titres (from 13.5 g L-1 to 16.5 g L-1) and yields (from

0.27 to 0.34) in C. acetobutylicum YM1 when grown in batch without pH control (Al-

Shorgani et al., 2018). The butanol yield of C. pasteurianum DSM 525 has also been

reported to increase (from 0.31 to 0.38 g g-1) by butyric acid addition when grown in batch,

but with pH controlled at 5.3 (Regestein et al., 2015). Despite the favourable effect of

butyric acid, the higher butyric acid concentration of 8 g L-1 yielded the lowest substrate

consumption yield (64 %), and the lowest acetone (0.79 g L-1), butanol (8.72 g L-1) and

ethanol (0.41 g L-1) titres. High concentrations of butyric acid (particularly in undissociated

form) are thought to be inhibitory for clostridial species by affecting the balanced pH

gradient across the cell membrane (Bryant and Blaschek, 1988). Therefore, results suggest

that butyric acid medium concentrations should be optimised to avoid potential inhibitory

effects.

Acetic acid supplementation did not have a pronounced effect on butanol or ethanol

production, but cultures grown in 4 g L-1 or 8 g L-1 of acetic acid yielded higher acetone

titres than the control (Table 1). Increased acetic acid can improve the production of

fermentation solvents, although generally favouring acetone production. For example, the

supplementation of acetic acid (up to 1 g L-1) induced the formation of acetone (with no

visible effects on butanol) in C. acetobutylicum 77 (mutant from ATCC 824) (Matta-El-

Ammouri et al., 1987). Meanwhile, in C. saccharoperbutylacetonicum N1-4, the addition

of 4 g L-1 of acetic acid increased both acetone (from 3.42 to 6.67 g L-1) and butanol (from

8.90 to 13.2 g L-1). In addition to acetic acid acting as an additional carbon source, the

increases of both acetone and butanol concentration via acetic acid supplementation are

attributed to a corresponding increase in the activity of enzymes responsible for the

formation of these solvents, acetoacetate decarboxylase and butanol dehydrogenase (Zhou

et al., 2018).

Interestingly, results obtained in this study showed that both nitrogen treatments similarly

resulted in higher acetone titres. The culture grown in 6 g L-1 of NH4Cl actually attained



240

the highest acetone titre (2.97 g L-1) in all the glucose-based fermentations. As observed

in Figure 2.c., nitrogen treatment resulted in an elevated concentration of acetic acid

compared to other treatments, which explains the increased formation of acetone. Nitrogen

plays a major role in providing the necessary growth molecules (e.g. nucleic acids and

aminoacids) and metabolic energy from the transport of NH4+ ions, and can therefore

regulate clostridial metabolism and in turn solvent production (Roos et al., 1985).

However, the specific effects of nitrogen availability on the ABE fermentation have not

been yet evaluated in detail in the literature and require further analysis.

Despite the above differences, the one-way ANOVA analysis of the experimental data

presented in Table 1 revealed that the only significant difference against the control

corresponded to the butanol titre (p=0.013) and glucose consumption yield (p < 0.001)

obtained by the culture grown in 8 g L-1 of butyric acid. Therefore, further evaluation is

required to identify significant concentrations, and/or additional fermentation variables,

that can be artificially manipulated to optimise fermentation yields.

4.1.2. Response surface analysis (RSA).

To evaluate the responses of acetone, butanol, and ethanol to the initial concentrations of

butyric acid and acetic acid, surface response curves were generated by fitting a polynomial

equation model (Eq. 1) to the experimental data in Table 1. The estimated regression

coefficients of the corresponding polynomial equations are presented in Table 2. The

polynomial models for acetone, butanol, and ethanol formation displayed co-relation

coefficients (R2) of 0.87, 0.91, and 0.82, respectively, indicating a fair level of fit to explain

the formation of ABE solvents (i.e. the dependent variables) in terms of the initial acetic

acid and butyric acid concentrations (i.e. the independent variable). The 3D surface

responses obtained by each of the resulting model equations are presented in Figure 3.

According to the surface curves, butyric acid exhibits a stronger effect than acetic acid on

the formation of all the fermentation solvents. In particular, the surface curves show the

inhibitory effects of increased butyric acid concentration on the ABE solvents, which

agrees with experimental observations. The pronounced effects of butyric acid over acetic

acid are also evidenced by the statistical significance of the regression coefficients directly

associated to butyric acid (Table 2). As per the analysis, regression coefficients with a p-



241

value lower than 0.05 indicate a high significance for its corresponding factor and thus a

high effect on the response variable. As observed in Table 2, the regression coefficients

with p-values higher than 0.05 (not significant) are mostly associated to acetic acid,

specifically its quadratic effects (i.e. 𝛼3 ∙ 𝑋12) on all three response variables (acetone,

butanol, and ethanol), and its linear effect (i.e. 𝛼1 ∙ 𝑋) on ethanol. Butyric acids effects can

be considered insignificant only by their interaction with acetic acid (i.e. 𝛼5 ∙ 𝑋1 ∙ 𝑋2) on

butanol and ethanol.

3D surface responses are a graphical tool to evaluate the effects of multiple factors on a

response variable, and have been used to optimise biological processes (Dragone et al.,

2011; Khunchantuek and Fiala, 2017). Here, the response surface curve for butanol

(Figure 3.b) indicated that a maximum butanol concentration of 14.77 g L-1 could be

attained by supplementing fermentation medium with 0.1 g L-1 of acetic acid and 3 g L-1

of butyric acid. However, the following points should be made clear: i) although the

variables used in surface response analyses should be noise-free, the variables measured

here (i.e. acetone, butanol, and ethanol) are subject to various fermentation-related factors

difficult to control, and whilst standard deviations were computed, the analysis was carried

out using only the means of three experimental replicates ; ii) the analysis presented here

was done a posteriori, and despite the fair level of fit of the polynomial models, surface

response analyses require the evaluation of more factor levels (i.e. concentrations) and the

interactions between them.

4.1.3. Microalgae-based fermentation.

The results of the fermentations carried out with microalgal biomass (MB), or microalgal

hydrolysate (derived from hydrolysed microalgal biomass, MB-H) are presented in Figure

4. In both cases, fermentation media was supplemented with ~ 3 g L-1 of butyric acid since

this concentration was identified as the optimal for glucose-based fermentations via the

surface response analysis. As observed in Figure 4.a, cell growth took place in both

fermentations, but the fermentation carried out in microalgal hydrolysate yielded the

highest cell dry weight. No lag-phase was observed in neither of these fermentations, and

unlike in glucose-based fermentations, cells remained in a stationary phase which can be

explained due to the lack of butanol inhibition.



242

The HPLC analysis of the microalgal hydrolysate yielded a glucose concentration of 10.94

g L-1 of glucose, which was then slightly diluted to an initial substrate concentration of

8.86 g L-1 upon mixing with the remaining P2 medium components. For comparison

purposes, the fermentation with microalgal biomass (MB) was carried out at a

concentration of 10 g L-1 of biomass (dry weight). It should be mentioned that 60 g L-1 of

substrate are typically employed in ABE fermentations to obtain appropriate solvent titres

(Qureshi et al., 2006), but the limited stock of microalgal biomass collected in this study

restricted this concentration to be used. As observed in Figure 4.b, most of the glucose in

the hydrolysate was completely consumed within 24 h. Glucose was not detected in the

fermentation with non-hydrolysed microalgal biomass (MB), except during the last days

in low concentrations which could have resulted from a gradual solubilisation effect by

starch degrading enzymes of clostridial species (Jones and Woods, 1986). Both

fermentations produced organic acids, but the initial rate of acid production was higher in

the fermentation with microalgal hydrolysate, possibly due to the substrate (glucose) being

already available in soluble form.

According to analysis, fermentations produced up to 1.031 g L-1 (with microalgal biomass)

and 0.89 g L-1 (with microalgal hydrolysate) of butanol (Figure 4.f). Although the

difference in butanol titres is not pronounced, the lower concentration attained by the

hydrolysate could be the result of certain inhibitory compounds that are typically released

during acid treatments, such as furfural or phenolic compounds (Yang et al., 2015). It was

also observed that butanol formation started sooner in the fermentation with hydrolysate

possibly due to having accumulated organic acids at a faster rate. Acetone and ethanol

solvents were produced in very low quantities or were not detected during analysis. Results

thus indicate that the substrate to butanol yield of the fermentation with microalgal

hydrolysate was 10.07 % (i.e. 100 x 0.89/8.86 g g-1 of substrate). Meanwhile, the butanol

yield of the fermentation with microalgal biomass was estimated as 10.31 % (i.e. 100 x

1.03/10 g g-1 of CDW).

The butanol yields of microalgae-based fermentations were much lower than those

obtained by any of the glucose-based fermentations, which was expected given the low

substrate concentrations employed for the fermentations. However, as shown in Table 4,

butanol yields are not far from those reported by other studies where butanol production



243

by microalgae has also been evaluated. Nevertheless, it is clear that more experimentation

is required to adequately evaluate, and further optimise, the overall performance of

microalgae-based fermentations.

4.2. Biodiesel from microalgae.

Crude lipids from microalgae biomass consist of neutral, polar, and non-fatty acid lipids

and other contaminants such as ketones, chlorophyll, and proteins. However, only neutral

or non-polar lipids (e.g. FAMEs such as TAG) are saponifiable, meaning they can be

converted by transesterification to biodiesel. Therefore, due to a non-polar nature with high

selectivity towards TAGs, hexane has been the preferred solvent to favour extraction of

saponifiable lipids (Orr et al., 2016).

The hexane-extracted crude lipids from the microalgal biomass (MB), the hydrolysed

microalgal biomass (MB-H), and the fermented microalgal biomass (MB-F), were 4.25 %,

6.8 %, and 6.5 %, respectively. The percentage of saponifiable lipids extracted from all

three microalgal conditions was determined by applying Eq. 3:

Saponifiable lipids (%)= FAME (mg/g)

Extracted crude lipids (mg/g)∙ 100 Eq. (3)

The FAMEs content and composition were quantified by chromatographic analysis, and

results are shown in Figure 5. The microalgal biomass in raw conditions yielded the lowest

saponifiable lipids (MB at 21.62 %) against the hydrolysed microalgal biomass (MB-H at

56.1 %) or the fermented microalgal biomass (MB-F at 49.9 %). Given that both the

hydrolysed and the fermented samples originated from the same microalgal biomass, the

FAME profile (e.g. saturated, unsaturated, and polyunsaturated fatty acids) was essentially

the same (Figure 5.b) in all three biomass conditions However, the change in saponifiable

content can be explained by the different degrees of cell disruption obtained during their

pre-treatment steps. Disruption processes can improve the efficiencies of lipid extraction

by favouring a higher interfacial area between solvent and cells (Orr et al., 2016). Whilst

all biomass samples were pulverised manually with liquid nitrogen, MB-F was additionally

autoclaved (at 120 °C for 20 min) before being used as fermentation substrate, and MB-H,

which showed the highest saponifiable content, was both autoclaved (at 120 C for 20 min)

and hydrolysed with sulfuric acid.



244

4.2.1. Fatty acid methyl ester (FAME) profile and biodiesel yield.

The complete fatty acid methyl esther (FAME) profile, as obtained by chromatographic

analysis, is shown in Table 3. The most dominant FAMEs are palmitic acid (C16:0), oleic

acid (C18:1), linoleic acid (C18:2), and linolenic acid (C18:3). Other FAMEs were also

present such as palmitoleic acid (C16:1), hexadecadienoic acid (C16:2), stearic acid

(C18:0). These FAMEs correspond to the most relevant fatty acids for biofuel production

(Halim et al., 2011; Wan Mahmood et al., 2017). In addition, linolenic acid (C18:3) is an

essential fatty acid for human health and has also been found in Picochlorum sp. (Yang et

al., 2014). All microalgal biomass conditions showed similar FAME profiles where most

of the important fatty acids were dominantly present.

Tripentadecanoin, a triacylglyceride (TAG), was used as an internal standard for

chromatographic quantification of FAMEs. Although studies may employ FAMEs as their

internal standard (Halim et al., 2012; Pan et al., 2016), the use of a TAG allows to account

for the potential loss of analyte during both extraction and transesterification processes

since similar losses would be experienced by the standard and the microalgal lipids.

Besides, tripentadecanoin is also a good internal standard since it consists of 3 C15:0 and

does not interfere with microalgal lipids which only contain TAGs with even numbers

(Wan Mahmood et al., 2017).

The yield of biodiesel (FAMEs) produced from all three biomass conditions was calculated

using Eq. 4 (Wan Mahmood et al., 2017):

Biodiesel (FAME) yield %= FAME content (mg/g CDW)

Total lipid content (mg/g CDW)∙ 100 Eq. (4)

The Total lipid content (provided by the Bligh and Dyer method) was 158 mg g-1 of cell

dry weight (CDW) of microalgal biomass (or 15.58 %). As shown in Table 3, the highest

FAME content was derived from the MB-H condition at 38.2 mg g-1 CDW (i.e. 3.82 %

FAME per CDW), followed by MB-F condition at 32.9 mg g-1 CDW (i.e. 3.29 % per

CDW). The lowest FAME content was provided by the initial MB, with only 9.2 mg g-1

CDW (i.e. 0.92 % per CDW). This indicated that MB-H and MB-F provided the highest

biodiesel (FAME) yields, at 24.16% and 20.80%, respectively. Meanwhile, MB provided

a biodiesel yield of 5.82%. Again, the higher biodiesel yields of MB-H and MB-F can be



245

attributed to their higher level of cell disruption. A comparison between different

microalgal pre-treatment steps and final biodiesel (FAME) yields obtained in this study

and others reported in the literature is presented in Table 4.

5. Conclusions.

This study quantified the production of biobutanol (via the ABE fermentation) and

biodiesel (via transesterification) using microalgal biomass from the model species C.

reinhardtii. The microalgal biomass was subjected to different pre-treatment methods

including pulverisation, sterilisation, and acid hydrolysis at high temperature. Biobutanol

yields of 10.31 % and 10.07 % were obtained when using microalgal biomass or microalgal

hydrolysate, respectively. The biodiesel (FAME) content of the fermented microalgal

biomass and the hydrolysed biomass residues were 3.29 % and 3.82 %, respectively, much

higher than that of the raw microalgal biomass (0.92 %), which highlighted the benefits of

cell disruption via different pre-treatment methods. Results demonstrate the potential use

of microalgae as a biofuel substrate within a biorefinery framework.

References.

Al-Shorgani NKN, Kalil MS, Yusoff WMW, Hamid AA. 2018. Impact of pH and butyric

acid on butanol production during batch fermentation using a new local isolate of

Clostridium acetobutylicum YM1. Saudi J. Biol. Sci. 25:339–348.

Bankar SB, Survase S a., Ojamo H, Granström T. 2013. Biobutanol: the outlook of an

academic and industrialist. RSC Adv. 3:24734.

Barsanti L, Gualtieri P. 2018. Is exploitation of microalgae economically and

energetically sustainable? Algal Res. 31:107–115.

Bekirogullari M, Fragkopoulos IS, Pittman JK, Theodoropoulos C. 2017. Production of

lipid-based fuels and chemicals from microalgae: An integrated experimental and

model-based optimization study. Algal Res. 23:78–87.

Bligh EG, Dyer WJ. 1959. A rapid method of total lipid extraction and purification. Can.

J. Biochem. Physiol. 37:911–917.

Bryant DL, Blaschek HP. 1988. Buffering as a means for increasing growth and butanol

production byClostridium acetobutylicum. J. Ind. Microbiol. 3:49–55.



78:1–10.

Chisti Y. 2007. Biodiesel from microalgae. Biotechnol. Adv. 25:294–306.



246



Energy 88:3331–3335.

Efremenko EN, Nikolskaya a B, Lyagin I V, Senko O V, Makhlis T a, Stepanov N a,

Maslova O V, Mamedova F, Varfolomeev SD. 2012. Production of biofuels from

pretreated microalgae biomass by anaerobic fermentation with immobilized

Clostridium acetobutylicum cells. Bioresour. Technol. 114:342–8.

Ellis JT, Hengge NN, Sims RC, Miller CD. 2012. Acetone, butanol, and ethanol

production from wastewater algae. Bioresour. Technol. 111:491–5.

Enamala MK, Enamala S, Chavali M, Donepudi J, Yadavalli R, Kolapalli B, Aradhyula

TV, Velpuri J, Kuppam C. 2018. Production of biofuels from microalgae - A

review on cultivation, harvesting, lipid extraction, and numerous applications of

microalgae. Renew. Sustain. Energy Rev. 94:49–68.

Figueroa-Torres GM, Pittman JK, Theodoropoulos C. 2017. Kinetic modelling of starch

and lipid formation during mixotrophic, nutrient-limited microalgal growth.


Gao K, Orr V, Rehmann L. 2016. Butanol fermentation from microalgae-derived

carbohydrates after ionic liquid extraction. Bioresour. Technol. 206:77–85.

García V, Päkkilä J, Ojamo H, Muurinen E, Keiski RL. 2011. Challenges in biobutanol

production: How to improve the efficiency? Renew. Sustain. Energy Rev.

15:964–980.

Groom MJ, Gray EM, Townsend PA. 2008. Biofuels and Biodiversity: Principles for

Creating Better Policies for Biofuel Production. Conserv. Biol. 22:602–609.

Halim R, Danquah MK, Webley PA. 2012. Extraction of oil from microalgae for

biodiesel production: A review. Biotechnol. Adv. 30:709–732.

Halim R, Gladman B, Danquah MK, Webley PA. 2011. Oil extraction from microalgae

for biodiesel production. Bioresour. Technol. 102:178–185.

Hariskos I, Posten C. 2014. Biorefinery of microalgae - opportunities and constraints for

different production scenarios. Biotechnol. J. 9:739–52.

Harris EH. 1989. The Chlamydomonas sourcebook : a comprehensive guide to biology

and laboratory use. Academic Press 780 p.

Harun R, Danquah MK, Forde GM. 2010. Microalgal biomass as a fermentation

feedstock for bioethanol production. J. Chem. Technol. Biotechnol. 85:199–203.

Harun R, Yip JWS, Thiruvenkadam S, Ghani WAWAK, Cherrington T, Danquah MK.

2014. Algal biomass conversion to bioethanol – a step-by-step assessment.

Biotechnol. J. 9:73–86.

Jones DT, Woods DR. 1986. Acetone-butanol fermentation revisited. Microbiol. Rev.

50:484–524.



247

Khunchantuek C, Fiala K. 2017. Optimization of Key Factors Affecting Butanol

Production from Sugarcane Juice by Clostridium beijerinckii TISTR 1461.

Energy Procedia 138:157–162.

Kim KH, Choi IS, Kim HM, Wi SG, Bae H-J. 2014. Bioethanol production from the

nutrient stress-induced microalga Chlorella vulgaris by enzymatic hydrolysis and

immobilized yeast fermentation. Bioresour. Technol. 153:47–54.

Köpke M, Dürre P. 2011. Handbook of biofuels production. Handb. Biofuels Prod.

Elsevier 221-257 p.

Kumar M, Gayen K. 2011. Developments in biobutanol production: New insights. Appl.

Energy 88:1999–2012.

Lage S, Gentili FG. 2018. Quantification and characterisation of fatty acid methyl esters

in microalgae: Comparison of pretreatment and purification methods. Bioresour.

Technol. 257:121–128.

Liu J, Huang J, Sun Z, Zhong Y, Jiang Y, Chen F. 2011. Differential lipid and fatty acid

profiles of photoautotrophic and heterotrophic Chlorella zofingiensis: Assessment

of algal oils for biodiesel production. Bioresour. Technol. 102:106–110.

Matta-El-Ammouri G, Janati-Idrissi R, Junelles A-M, Petitdemange H, Gay R. 1987.

Effects of butyric and acetic acids on acetone-butanol formation by Clostridium

acetobutylicum. Biochimie 69:109–115.

Orr VCA, Plechkova N V, Seddon KR, Rehmann L. 2016. Disruption and Wet

Extraction of the Microalgae Chlorella vulgaris Using Room-Temperature Ionic

Liquids. ACS Sustain. Chem. Eng. 4:591–600.

Pan J, Muppaneni T, Sun Y, Reddy HK, Fu J, Lu X, Deng S. 2016. Microwave-assisted

extraction of lipids from microalgae using an ionic liquid solvent

[BMIM][HSO4]. Fuel 178:49–55.

Qureshi N, Li X-L, Hughes S, Saha BC, Cotta M a. 2006. Butanol production from corn

fiber xylan using Clostridium acetobutylicum. Biotechnol. Prog. 22:673–80.

Qureshi N, Saha BC, Cotta M a. 2007. Butanol production from wheat straw hydrolysate

using Clostridium beijerinckii. Bioprocess Biosyst. Eng. 30:419–27.

Raganati F, Procentese A, Olivieri G, Götz P, Salatino P, Marzocchella A. 2015. Kinetic

study of butanol production from various sugars by Clostridium acetobutylicum

using a dynamic model. Biochem. Eng. J. 99:156–166.

Regestein L, Doerr EW, Staaden A, Rehmann L. 2015. Impact of butyric acid on butanol

formation by Clostridium pasteurianum. Bioresour. Technol. 196:153–159.

Richardson JW, Johnson MD, Outlaw JL. 2012. Economic comparison of open pond

raceways to photo bio-reactors for profitable production of algae for

transportation fuels in the Southwest. Algal Res. 1:93–100.



248

Rojan P J, Anisha GS, Nampoothiri KM, Pandey A. 2011. Micro and macroalgal

biomass: a renewable source for bioethanol. Bioresour. Technol. 102:186–93.

Roos JW, McLaughlin JK, Papoutsakis ET. 1985. The effect of pH on nitrogen supply,

cell lysis, and solvent production in fermentations ofClostridium acetobutylicum.

Biotechnol. Bioeng. 27:681–694.

Shuler ML, Kargi F. 1992. Bioprocess Engineering: Basic Concepts. Prentice Hall 479 p.

Suganya T, Varman M, Masjuki HH, Renganathan S. 2016. Macroalgae and microalgae

as a potential source for commercial applications along with biofuels production:

A biorefinery approach. Renew. Sustain. Energy Rev. 55:909–941.

Survase SA, van Heiningen A, Granström T. 2012. Continuous bio-catalytic conversion

of sugar mixture to acetone--butanol--ethanol by immobilized Clostridium

acetobutylicum DSM 792. Appl. Microbiol. Biotechnol. 93:2309–2316.

Trivedi J, Aila M, Bangwal DP, Kaul S, Garg MO. 2015. Algae based biorefinery—How

to make sense? Renew. Sustain. Energy Rev. 47:295–307.

Velasquez-Orta SB, Garcia-Estrada R, Monje-Ramirez I, Harvey A, Orta Ledesma MT.

2014. Microalgae harvesting using ozoflotation: Effect on lipid and FAME

recoveries. Biomass and Bioenergy 70:356–363.

Vlysidis A, Binns M, Webb C, Theodoropoulos C. 2011. Glycerol utilisation for the

production of chemicals: Conversion to succinic acid, a combined experimental

and computational study. Biochem. Eng. J. 58–59:1–11.

Wan Mahmood WMA, Theodoropoulos C, Gonzalez-Miquel M. 2017. Enhanced

microalgal lipid extraction using bio-based solvents for sustainable biofuel

production. Green Chem. 19:5723–5733.

Wang SH, Zhang YP, Dong HJ, Mao SM, Zhu Y, Wang RJ, Luan GD, Li Y. 2011a.

Formic acid triggers the ``acid crash’’ of acetone-butanol-ethanol fermentation by

clostridium acetobutylicum. Appl Env. Microb 77.

Wang S, Zhang Y, Dong H, Mao S, Zhu Y, Wang R, Luan G, Li Y. 2011b. Formic Acid

Triggers the “Acid Crash” of Acetone-Butanol-Ethanol Fermentation by

Clostridium acetobutylicum. Appl. Environ. Microbiol. 77:1674–1680.

Wang Y, Guo W, Cheng C-L, Ho S-H, Chang J-S, Ren N. 2016. Enhancing bio-butanol

production from biomass of Chlorella vulgaris JSC-6 with sequential alkali

pretreatment and acid hydrolysis. Bioresour. Technol. 200:557–564.

Xu H, Miao X, Wu Q. 2006. High quality biodiesel production from a microalga

Chlorella protothecoides by heterotrophic growth in fermenters. J. Biotechnol.

126:499–507.

Yang F, Xiang W, Sun X, Wu H, Li T, Long L. 2014. A Novel Lipid Extraction Method

from Wet Microalga Picochlorum sp. at Room Temperature. Mar. Drugs.



249

Yang M, Kuittinen S, Zhang J, Vepsäläinen J, Keinänen M, Pappinen A. 2015. Co-

fermentation of hemicellulose and starch from barley straw and grain for efficient

pentoses utilization in acetone–butanol–ethanol production. Bioresour. Technol.

179:128–135.

Yang X, Tu M, Xie R, Adhikari S, Tong Z. 2013. A comparison of three pH control

methods for revealing effects of undissociated butyric acid on specific butanol

production rate in batch fermentation of Clostridium acetobutylicum. AMB

Express 3:3.

Yerushalmi L, Volesky B, Votruba J. 1986. Modelling of culture kinetics and physiology

for c. acetobutylicum. Can. J. Chem. Eng. 64:607–616.

Zhou Q, Liu Y, Yuan W. 2018. Kinetic modeling of lactic acid and acetic acid effects on

butanol fermentation by Clostridium saccharoperbutylacetonicum. Fuel 226:181–

189.

Zhu L. 2015. Biorefinery as a promising approach to promote microalgae industry: An

innovative framework. Renew. Sustain. Energy Rev. 41:1376–1384.



250

Lab

el

Ace

ton

e (

A)

Bu

tan

ol

(B)

Eth

an

ol

(E)

AB

ES

ub

str

ate

(S

)

co

nsu

mp

tio

n

Bu

tan

ol

YS

/B

AB

E

YS

,AB

E

Co

ntr

ol

(P2

) a

1.8

8 ±

0.0

312.6

7 ±

0.3

00.9

2 ±

0.0

515.4

7 ±

0.4

7100%

25%

31%

BA

0 =

2 g

L-1

1.5

9 ±

0.4

413.2

1 ±

0.7

00.7

4 ±

0.2

415.5

4 ±

1.7

189%

23%

28%

BA

0 =

4 g

L-1

2.3

6 ±

0.1

114.6

2 ±

0.2

81.0

8 ±

0.0

718.0

6 ±

0.5

8100%

30%

37%

BA

0 =

8 g

L-1

0.7

9 ±

0.0

88.7

2 ±

0.4

30.4

1 ±

0.0

19.9

3 ±

0.6

664%

30%

33%

AA

0 =

2 g

L-1

1.7

8 ±

0.2

211.0

4 ±

0.7

10.8

5 ±

0.0

713.6

7 ±

1.2

399%

24%

28%

AA

0 =

4 g

L-1

2.2

5 ±

0.0

211.4

8 ±

0.0

30.9

1 ±

0.0

214.6

3 ±

0.1

0100%

24%

30%

AA

0 =

8 g

L-1

2.6

3 ±

0.0

412.0

1 ±

1.2

91.1

1 ±

0.1

315.7

5 ±

1.7

9100%

26%

34%

BA

0,

AA

0 =

4 g

L-1

2.1

1 ±

0.0

812.7

6 ±

0.3

90.9

0 ±

0.4

15.7

7 ±

0.6

393%

27%

34%

N0 =

4 g

L-1

2.4

8 ±

0.5

311.7

4 ±

0.7

51.2

4 ±

0.1

115.4

6 ±

1.8

7100%

22%

29%

N0 =

6 g

L-1

2.9

7 ±

0.0

612.4

5 ±

0.4

31.2

7 ±

0.4

316.7

0 ±

0.5

999%

24%

32%

Tit

res,

g L

-1Y

ield

s,

%

Tab

le 1

. F

erm

enta

tion

tit

res

an

d y

ield

s fr

om

glu

cose

-base

d f

erm

enta

tion

s su

bje

ct t

o d

iffe

ren

t b

uty

ric

aci

d, ace

tic

aci

d,

an

d n

itro

gen

con

cen

trati

on

reg

imes

. S

tars

(*

) d

enote

sig

nif

ican

t d

iffe

ren

ces

(p <

0.0

5*, 0.0

1**)

wit

h r

esp

ect

to C

on

trol,

as

per

on

e-w

ay A

NO

VA

. R

esu

lts

an

d S

.D.

are

th

e m

ean

of

thre

e b

iolo

gic

al

rep

lica

tes.

a C

on

trol,

P2

med

ium

: B

A0=

0 g

L-1

, A

A0=

0 g

L-1

, a

nd =

2 g

NH

4C

lL-1

.



251

αi

Fact

ors

&

Inte

ract

ion

Valu

eM

ean

S.E

. a

pV

alu

eM

ean

S.E

. a

pV

alu

eM

ean

S.E

. a

p

α0

Inte

rcept

1.2

12

0.0

70

-11.5

48

0.6

28

-0.7

13

0.0

19

-

α1

X1

, A

ceti

c aci

d0.1

52

0.5

69

0.0

00***

-0.7

46

12.6

04

0.0

08***

0.0

00

0.0

19

1.0

00

α2

X2

, B

uty

ric

aci

d0.5

67

3.0

57

0.0

00***

2.2

18

46.3

00

0.0

00***

0.1

98

0.3

82

0.0

00***

α3

X1

2-0

.001

0.0

72

1.0

00

0.0

74

8.4

14

0.0

61

0.0

04

0.0

41

0.6

46

α4

X2

2-0

.090

2.5

11

0.0

00***

-0.3

72

42.1

02

0.0

00***

-0.0

34

0.3

67

0.0

00***

α5

X1

•X2

-0.0

37

0.1

96

0.0

00***

-0.0

08

0.6

34

1.0

00

-0.0

10

0.0

28

0.9

80

Ace

tone (

R2

= 0

.87)

Buta

nol (

R2

= 0

.91)

Eth

anol (

R2

= 0

.82)

Tab

le 2

. R

egre

ssio

n c

oeff

icie

nts

of

the

poly

nom

ial

mod

els

(Eq

. 1)

for

acet

on

e, b

uta

nol,

an

d e

than

ol.

Sta

rs d

enote

the

stati

stic

al

sign

ific

an

ce o

f th

e co

effi

cien

ts (

p <

0.0

5*, 0.0

1**, 0.0

01***

) as

per

on

e-w

ay A

NO

VA

.

a M

ean s

quare

d e

rror

for

all

exp

erim

enta

l ru

ns,

if

ai=

0 .



252

Table 3. Fatty acid methyl esther (FAME) profiles and biodiesel yield in: microalgal

biomass (MB), hydrolysed microalgal biomass (MB-H), and fermented microalgal

biomass (MB-F).

Distribution (% of FAME)

FAME MB MB-H MB-F

C12:0 0 0.1 0.1

C14:0 0.8 0.7 1

C16:0 24 22.1 19.5

C16:1 1.9 3 3.2

C16:2 1.2 1.6 2

C16:3 0.9 1.7 1.4

C18:0 5.2 9.8 4.1

C18:1 23.5 26.2 28.2

C18:2 15.4 16 17

C18:3 21.6 18.1 13.1

C20:4 0.5 0 7.7

C20:5 4.9 0 2

C22:0 0.2 0.4 0.3

C24:0 0 0.3 0.3

Saturated FA a 30.2 33.4 25.3

Monounsaturated FA 25.4 29.2 31.4

Polyunsaturated FA 44.5 37.4 43.2

FAME (mg/g DW) 9.2 38.2 32.9

Biodiesel (FAME) yield (%) b 5.82 24.16 20.8

a FA, Fatty Acids

b Biodiesel (FAME) yield, as computed by Eq.4



253

Bio

mass

AB

E f

erm

en

tati

on

Bu

tan

ol

Lip

id e

xtr

acti

on

FA

ME

Pre

-tre

atm

en

tsu

bstr

ate

g L

-1 (

%)

so

lve

nt

(% p

er

DW

)

C.

rein

ha

rdti

iP

ulv

eri

sed (

P)

--

Hexane (

155 °

C, 2h)

0.9

2

C.

rein

ha

rdti

iP

+ S

teri

lised (

S)

Alg

ae (

1 %

w/v

)1.0

31 (

10.3

1)

Hexane (

155 °

C, 2h)

3.2

9

P +

S +

Alg

al hydro

lysa

te

Hydro

lyse

d (

acid

)(S

o =

8.6

g L

-1 g

lucose

)

Chlo

rofo

rm:M

eth

anol (2

:1)

(4 °

C, overn

ight)

Ozo

flota

tion

Chlo

rofo

rm:M

eth

anol (2

:1)

Lyophili

sed

(4 °

C, overn

ight)

Chlo

rofo

rm:M

eth

anol (2

:1)

(room

tem

pera

ture

, 2 m

in)

Chlo

rofo

rm:M

eth

anol (2

:1)

(room

tem

pera

ture

, 2 m

in)

Chlo

rofo

rm:M

eth

anol (2

:1)

(room

tem

pera

ture

, 2 m

in)

Chlo

rofo

rm:M

eth

anol (2

:1)

(room

tem

pera

ture

, 2 m

in)

Mix

ed c

ulture

Hydro

lyse

d (

acid

)T

reate

d a

lgae (

10 %

w/v

)2.2

6 (

10.1

0)

--

Mix

ed c

ulture

N/A

Un-t

reate

d a

lgae (

10 %

)0.5

2 (

17.3

0)

--

Ch

lore

lla

vu

lga

ris

Fre

eze

-dri

ed

Lip

id e

xtr

acte

d* a

lgae (

8%

w/v

)6.6

3 (

23)

13.8

0

Lip

id e

xtr

acte

d* a

lgal hydro

lysa

te

(So =

28.8

8 g

L-1

glu

cose

)

Ch

lore

lla

vu

lga

ris

Fre

eze

-dri

ed

Lip

id e

xtr

acte

d* a

lgae (

8%

w/v

)4.9

9 (

15)

13.8

0

Lip

id e

xtr

acte

d* a

lgal hydro

lysa

te

(So =

36.4

5 g

L-1

glu

cose

)C

hlo

rell

a v

ulg

ari

sF

reeze

-dri

ed

5.3

4 (

15)

13.8

0

Hexane:2

-pro

panol (3

:2)

(room

tem

pera

ture

, 16 h

)

Ionic

liq

uid

[C2m

im][

EtO

H]

(room

tem

pera

ture

, 2 h

)

Gao e

t al. (

2016)

Ch

lore

lla

vu

lga

ris

Fre

eze

-dri

ed

0.4

4 (

4)

13.8

0

Scen

ed

esm

us

dim

orp

hu

sF

reeze

-dri

ed

--

~ 5

Elli

s et

al. (

2012)

~ 1

0

Scen

ed

esm

us

dim

orp

hu

sO

ven-d

ried

--

~ 1

0

Scen

ed

esm

us

dim

orp

hu

sB

oile

d (

isopro

panol)

--

~ 6

.5

Lage a

nd G

entili,

(2018)

Scen

ed

esm

us

dim

orp

hu

sF

roze

n-

-

Mix

ed c

ulture

Lyophili

sed

--

1.2

2V

ela

squez-

Ort

a

et

al. (

2014)

Mix

ed c

ulture

--

3.2

3

Mic

roalg

ae

Re

fere

nce

Th

is w

ork

C.

rein

ha

rdti

i0.8

9 (

10.0

7)

Hexane (

155 °

C, 2h)

3.8

2

Tab

le 4

. C

om

pari

son

bet

wee

n d

iffe

ren

t m

icro

alg

al

pre

-tre

atm

ents

an

d b

iob

uta

nol

an

d b

iod

iese

l



254

Mic

roa

lgal

Bio

ma

ss

(MB

)

Hy

dro

lyse

d

Mic

roa

lgal

Bio

ma

ss

(MB

-H)

Fer

men

ted

Mic

roa

lgal

Bio

ma

ss

(MB

-F)

AB

E

Fer

men

tati

on

Hyd

roly

sis

4%

(w

/v)

H2S

O4

Tem

per

atu

re

120

C, 20 m

in

Tem

per

atu

re

120

C, 20 m

in

Hyd

roly

sate

Bio

bu

tan

ol

Bio

die

sel

Tra

nse

ster

ific

ati

on

AB

E

Fer

men

tati

on

Tra

nse

ster

ific

ati

on

Tra

nse

ster

ific

ati

on

Extr

act

ion

Hex

ane

155

C, 2 h

Pu

lver

isati

on

& f

reez

e d

ryin

g

Extr

act

ion

Hex

ane

155

C, 2 h

Extr

act

ion

Hex

ane

155

C, 2 h

Fig

ure

1. S

chem

ati

c d

iagra

m o

f th

e b

iofu

el p

rod

uct

ion

rou

tes

evalu

ate

d i

n t

his

case

-stu

dy a

lon

g w

ith

th

e

corr

esp

on

din

g m

icro

alg

al

bio

mass

pre

-tre

atm

en

t st

eps.



255

Fig

ure

2.

Con

cen

tra

tion

pro

file

of

dif

fere

nt

glu

cose

-base

d f

erm

enta

tion

s: C

on

trol

(P2:

BA

0=

0 g

L-1

, A

A0=

0 g

L-1

,

N=

2 g

NH

4C

l L

-1);

𝑩𝑨

𝟎=

𝟒 𝐠

𝐋−

𝟏 (

wit

h A

A0=

0 g

L-1

an

d N

=2 g

NH

4C

l L

-1);

𝑨𝑨

𝟎=

𝟒 𝐠

𝐋−

𝟏 (

wit

h B

A0=

0 g

L-1

, a

nd

N=

2 g

NH

4C

l L

-1),

an

d N

=4 g

NH

4C

l L

-1 (

wit

h B

A0=

0 g

L-1

an

d A

A0=

0 g

L-1

). R

esu

lts

an

d S

.D. are

th

e m

ean

of

thre

e

bio

logic

al

rep

lica

tes.



256

Acetone g L-1

Butanol g L-1

Ethanol g L-1

a)

Ace

ton

eb

) B

uta

no

lc)

Eth

an

ol

Fig

ure

3.

Res

pon

se s

urf

ace

plo

ts f

or

ace

ton

e, b

uta

nol,

an

d e

than

ol,

sh

ow

ing t

he

effe

cts

of

init

ial

ace

tic

aci

d a

nd

bu

tyri

c aci

d

med

ium

con

cen

trati

on

s. B

lack

poin

ts a

re e

xp

erim

enta

l d

ata

(m

ean

) ob

tain

ed u

nd

er t

he

corr

esp

on

din

g m

ediu

m

con

cen

trati

on

s.



257

Fig

ure

4.

Con

cen

tra

tion

tim

e-p

rofi

le o

f fe

rmen

tati

on

s u

sin

g m

icro

alg

al

bio

mass

(M

B),

an

d m

icro

alg

al

hyd

roly

sate

(d

eriv

ed

from

hyd

roly

sed

mic

roalg

al

bio

mass

, i.

e. M

B-H

). R

esu

lts

an

d S

.D. are

th

e m

ean

of

two b

iolo

gic

al

rep

lica

tes



258

Figure 5. a) Saponifiable (FAME ) lipid content, and b) FAME composition of:

microalgal biomass (MB), hydrolysed microalgal biomass (MB-H), and fermented

microalgal biomass (MB-F). [SFA, saturated fatty acids; MUFA, monounsaturated

fatty acids; PUFA, polyunsaturated fatty acids].



259



presented next.



260


Associated to:

Microalgal Biomass as a Biorefinery Platform for Biobutanol and Biodiesel

Production: A case study

Gonzalo M. Figueroa-Torresa, Wan M. Asyraf Wan Mahmooda, Jon K. Pittmanb,

Constantinos Theodoropoulosa,*




9PL







261

1. Glucose-based ABE Fermentation:

As explained within the case study, glucose-based experiments aimed to standardise the

fermentation protocols so as to obtain butanol production yields comparable to other

studies. Preliminary experiments carried out in P2 medium using closed fermentation

vessels showed clear microbial growth and the accumulation of organic acids characteristic

of the acidogenic phase. However, ABE fermentation solvents were either produced in low

quantities (below 0.5 g L-1) or not produced at all. During the course of these experiments

it was evident that the microbial cells accumulated gases which were abruptly released

upon sampling. Since the build-up of gas pressure was assumed to affect the start of the

solventogenic phase, a venting system was implemented. The venting system consisted of

connecting all fermentation vessels (via sterile plastic pipes and needles) to a water trap,

allowing continuous pressure release but avoiding the inlet of oxygen gas. The

implementation of this system yielded higher butanol titres (Figure a) than when gases

were not released. However, butanol production and substrate consumption were still

deemed not appropriate when compared to those reported by other studies, where

substrates are completely consumed and butanol titres can be higher than 10 g L-1 (Cheng

et al., 2015; Wang et al., 2014).

At this point, it also became clear that the pH of these fermentations would generally drop

to values below 4.5 (Figure a.h). Although the drop in pH (characteristic of the acidogenic

phase) is deemed necessary for solventogenesis to occur, the reduction of pH can also lead

fermentations to experience “acid crash”, a phenomenon where solvent formation and

substrate uptake stop due to an over accumulation of organic acids (Bryant and Blaschek,

1988; Wang et al., 2011). To avoid the so-called acid crash, P2 medium was supplemented

with CaCO3 at a concentration of 5 g L-1, as in Raganati et al. (2015). Although this

buffering agent has not been frequently used (or reported) for batch butanol fermentations,

the use of CaCO3 allowed pH to remain above 4.5 and attain higher butanol yields, as

shown in the fermentations presented in the main text of this manuscript.

The complete concentration profiles obtained from all glucose-based fermentations subject

to different concentrations of butyric acid, acetic acid, and nitrogen, are presented in

Figure b, Figure c, and Figure d, respectively.



262

Fig

ure

a.

AB

E f

erm

enta

tio

n d

ata

ob

tain

ed w

ith

60

g L

-1 o

f glu

cose

in

P2 m

ediu

m, u

nb

uff

ered

, u

sin

g a

ven

tin

g s

yst

em. A

ceto

ne

was

not

det

ecte

d. D

ata

an

d S

.D. are

th

e m

ean

of

two b

iolo

gic

al

rep

lica

tes.



263

Fig

ure

b. A

BE

fer

men

tati

on

data

ob

tain

ed w

ith

60 g

L-1

of

glu

cose

in

P2 m

ediu

m s

ub

ject

to d

iffe

ren

t in

itia

l co

nce

ntr

ati

on

s of

bu

tyri

c aci

d. D

ata

an

d S

.D. are

th

e m

ean

of

thre

e b

iolo

gic

al

rep

lica

tes.



264

Fig

ure

c. A

BE

fer

men

tati

on

data

ob

tain

ed w

ith

60 g

L-1

of

glu

cose

in

P2 m

ediu

m s

ub

ject

to d

iffe

ren

t in

itia

l co

nce

ntr

ati

on

s of

ace

tic

aci

d. D

ata

an

d S

.D. are

th

e m

ean

of

thre

e b

iolo

gic

al

rep

lica

tes.



265

Fig

ure

d. A

BE

fer

men

tati

on

data

ob

tain

ed w

ith

60 g

L-1

of

glu

cose

in

P2 m

ediu

m s

ub

ject

to d

iffe

ren

t in

itia

l co

nce

ntr

ati

on

s of

nit

rogen

(as

NH

4C

l). D

ata

an

d S

.D. a

re t

he

mea

n o

f th

ree

bio

logic

al

rep

lica

tes.



266

References

Bryant DL, Blaschek HP. 1988. Buffering as a means for increasing growth and butanol

production byClostridium acetobutylicum. J. Ind. Microbiol. 3:49–55.

Cheng H-H, Whang L-M, Chan K-C, Chung M-C, Wu S-H, Liu C-P, Tien S-Y, Chen S-

Y, Chang J-S, Lee W-J. 2015. Biological Butanol Production from microalgae-

based biodiesel residues by Clostridium acetobutylicum. Bioresour. Technol.

Raganati F, Procentese A, Olivieri G, Götz P, Salatino P, Marzocchella A. 2015. Kinetic

study of butanol production from various sugars by Clostridium acetobutylicum

using a dynamic model. Biochem. Eng. J. 99:156–166.

Wang S, Zhang Y, Dong H, Mao S, Zhu Y, Wang R, Luan G, Li Y. 2011. Formic Acid

Triggers the “Acid Crash” of Acetone-Butanol-Ethanol Fermentation by

Clostridium acetobutylicum. Appl. Environ. Microbiol. 77:1674–1680.

Wang Y, Guo W-Q, Lo Y-C, Chang J-S, Ren N-Q. 2014. Characterization and kinetics

of bio-butanol production with Clostridium acetobutylicum ATCC824 using mixed

sugar medium simulating microalgae-based carbohydrates. Biochem. Eng. J.

91:220–230.

267

Chapter 7

Conclusions and Recommendations

7.1. Conclusions.

Efforts are being made around the globe to help us transition into a more sustainable and

efficient bio-based economy where biological resources and processes are exploited to

their full potential. In particular, increased efforts are being directed towards the

commercialisation of liquid biofuels, one of the most promising alternative transport

energies which can lead the fight against climate change by lessening our overreliance on

fossil fuels. Biofuels production technologies, however, have long been overshadowed by

the uncertainty of their economic and environmental sustainability. Commercially

available biofuels, for example, are deemed unfit for widespread use given that the food-

based feedstocks from which they are produced compete for human food and arable land.

The need to develop advanced feedstock-to-biofuel conversion technologies favoured the

positioning of microalgae as a viable non-food candidate for biofuels production.

Microalgae’s potential for biofuels production is emphasized by their ability to accumulate

starch and lipids, i.e. biofuel precursors. As discussed in Chapter 1, if microalgal biofuels

are to become competitive fossil fuels alternatives, it is necessary to implement microalgal

cultivation systems yielding high density biomass rich in starch and/or lipid molecules.

Tailor-made cultivation strategies such as nutrient limitation (in which cellular stress is

artificially inflicted by reducing nutrient availability) have been widely demonstrated for

significantly inducing starch and lipid accumulation, although often with a trade-off in

biomass growth.

There is a growing number of research works aiming to carefully balance the negative

trade-off between the growth of biomass and the formation of starch and lipids. Such an

optimisation task, however, and as is often the case with bioprocesses, has typically relied

Chapter 7 – Conclusions and Recommendations

268

on the iterative manipulation of key growth-limiting factors known to regulate

microalgae’s carbon metabolism (e.g. nutrients, light, temperature) until a desired target is

achieved.

Mathematical models representative of microalgae’s complex growing dynamics can

speed-up and increase the efficiency of optimisation tasks by allowing the simulation and

preliminary evaluation of potential cultivation strategies. Therefore, with the goal of

identifying cultivation strategies optimised for biofuels production, this Ph.D. thesis

developed, in conjunction with experimental analysis, a macroscopic model capable of

predicting the dynamics of microalgal biomass growth as well as starch and lipid formation

in response to the cultivation environment.

Although a model should ideally account for the effects of multiple growth-limiting factors

rather than a single one, the increased mathematical complexity of such a refined model

may undermine its applicability, which is why “…modelling does not make sense without

defining, before making the model, what its use is and what problem is intended to help to

solve” (Bailey, 1998).

The review of literature presented in Chapter 2 highlighted, on one hand, nitrogen and

phosphorus limitation as two of the most studied starch and lipid enhancing cultivation

strategies, and on the other, how the associated trade-off in biomass can be avoided by

using mixotrophic strains which assimilate organic carbon substrates and exhibit high

biomass productivities. Based on these considerations, the model presented in this thesis

was thus developed with the goal of accounting for the effects of nitrogen, phosphorus,

and carbon (model inputs) on the formation of biomass, starch, and lipids (model outputs).

To construct the model’s input-output relationships and ultimately assess its predictive

capacity, laboratory-scale experiments were carried out with the green species

Chlamydomonas reinhardtii (strain CCAP 11/32C), grown mixotrophically in acetic acid.

The central carbon metabolism of C. reinhardtii has been extensively studied and the

nutrient-limited responses of starch and lipid accumulation have also been demonstrated,

making it a model organism fit for the development of advanced analytical and

computational tools that could be extrapolated to other microalgal species.


269

C. reinhardtii was initially cultivated mixotrophically under various initial concentrations

of nitrogen and acetic acid (the organic carbon substrate) to evaluate their effects on

biomass, starch, and lipids. The outcome of these experiments validated the characteristic

increase in starch and lipid accumulation and the associated drop in biomass growth as the

concentration of nitrogen was reduced (i.e. nitrogen limitation). It was also observed that

increasing the concentration of acetic acid produced higher biomass concentrations but

only up to a particular level since further addition of this carbon substrate became

inhibitory for biomass. The effect of acetic acid on the starch and lipid contents was much

less significant than that induced by nitrogen limitation, so that increases in starch and lipid

medium concentrations were deemed to be a consequence of the higher biomass

concentrations attained in the acetate-boosted cultures.

Based on the experimental observations and on the integration of existing modelling

approaches exhibiting desired predictive features, a multi-parametric kinetic model

portraying nitrogen-limited mixotrophic dynamics was developed (Contribution 1,

Chapter 3). The model was built by considering total microalgal biomass (cell dry weight)

to be equivalent to the sum of three individual carbon-based pools: starch, lipids, and active

biomass (i.e. biomass free of starch and lipids). The estimation of the model parameters

was carried out by means of an optimisation-based fitting methodology using a defined

number of experimental datasets obtained from the nitrogen and acetic acid experiments,

and the model’s predictive value was then validated against different experimental

datasets.

Aiming to identify microalgal cultivation scenarios optimised for biofuels production, the

validated model was employed to identify the optimal initial concentration of nitrogen, 𝑁0,

and acetic acid, 𝐴0, maximising starch and lipid concentrations. When compared to the

non-optimised case (N0 = 0.382 g N L-1, A0 = 0.42 gC L-1), the starch-enhanced (N0 = 0.336

g N L-1, A0 = 1.06 gC L-1) and the lipid-enhanced (N0 = 0.378 g N L-1, A0 = 1.15 gC L-1)

scenarios yielded a 261 % increase in starch concentration and a 66% increase in lipid

concentration, respectively.

Having demonstrated the high predictive capacity of the model, the next step was to

enhance its potential as an optimisation tool by further accounting for the effects of

phosphorus on the cultivation dynamics. Therefore, an additional set of laboratory-scale


270

experiments was carried out to evaluate the combined effects of acetic acid, nitrogen, and

phosphorus medium concentrations in C. reinhardtii. The different outcomes in biomass

growth and increased storage molecules accumulation observed in these experiments

demonstrated the advantage of manipulating multiple (rather than single) nutrients to attain

desired starch and lipid targets, but also showed the increased level of complexity that

would be required to experimentally identify optimal nutrient compositions. However, and

as explained before, experimental observations were employed to expand the model’s

predictive capacity by further establishing the phosphorus-dependent input-output kinetic

relationships.

The resulting model (Contribution 2, Chapter 4) was used to: i) generate a set of ternary

diagrams displaying the predicted concentrations of biomass, starch, and lipids, as a

function of a wide range of initial nitrogen, 𝑁0 , phosphorus, 𝑃0 , and acetic acid, 𝐴0

concentration regimes, and ii) identify optimal concentration sets for maximal starch and

lipid formation. When compared to the non-optimised case (N0 = 0.382 g N L-1, P0 = 0.096

gPO4 L-1, A0 = 0.42 gC L-1), the starch-enhanced (N0 = 0.330 g N L-1, P0 = 0.052 gPO4 L

-1,

A0 = 0.96 gC L-1) and the lipid-enhanced (N0 = 0.365 g N L-1, P0 = 0.041 gPO4 L-1, A0 =

1.00 gC L-1) scenarios were shown to yield increases of 270 % starch and 74 % lipids,

respectively. These increases were slightly higher than those attained by solely

manipulating nitrogen and acetic acid, and although they exhibit high acetic acid

requirements, they highlighted the benefits of optimally applying nutrient co-limitation

given that the combined reduction of both nitrogen and phosphorus sources can lower

cultivation costs.

Regarding the acetic acid concentrations of the optimised cases obtained in this study, they

were in all cases higher than 50 % of those employed in the (non-optimised) base case.

Although the high acetic acid requirements can be unfavourable from an economic

perspective, they were deemed to be necessary to maintain adequate mixotrophic growing

conditions and avoid drastic reductions in biomass when subject to nutrient limitation. As

discussed previously, microalgal cultivation systems should attain high-density algal

biomass suitable to be employed in biofuel conversion processes. In this regard, fed-batch

cultivation strategies, where growth-limiting nutrients are supplied at different feeding


271

rates over the cultivation period, are similarly demonstrated to yield high biomass densities

(Literature Review, Chapter 2).

Therefore, further experimental work was carried out to: i) evaluate the performance of a

fed-batch cultivation strategy and quantify the potential increases in biomass, starch, and

lipids, with respect to a standard batch cultivation, and ii) evaluate and improve the

capacity of the developed model to predict fed-batch cultivation dynamics (Contribution

3, Chapter 5). Taking into account that the optimised cases for batch cultures relied on

high acetic acid supply, the fed-batch nutrient feeding strategy employed in this work

consisted on the intermittent addition of acetic acid pulses with various concentrations. It

was observed that, when cultures were subjected to a single pulse, the increase in the acetic

acid medium concentration yielded up to a 50 % increase in biomass with respect to

standard batch conditions. When cultures were supplemented with two consecutive pulses

biomass concentration was 94 % higher than batch cultures, which in turn led to a

significant increase of the starch (218 %) and lipid (168 %) concentrations. The increased

concentration yields attained by the pulse-assisted fed-batch system demonstrated its

viability as a biofuel-oriented cultivation strategy. The high biomass yields of the fed-batch

strategy, particularly, were favourable for lipid production since they attained much higher

concentrations than those of the lipid-enhanced batch scenarios.

The model developed previously was then evaluated based on its capacity to simulate the

pulse-assisted fed-batch dynamics. However, the model failed to replicate the clear

increase in biomass following the addition of a pulse, a drawback that was associated to

its inability to account for the regained uptake nitrogen capacity of cells after a pulse of

acetic acid, as observed from experimentation. The model was therefore adapted to

simulate both the batch and fed-batch dynamics by re-structuring the equation responsible

for nitrogen uptake, and by subsequently re-estimating a set of kinetic parameters that were

deemed to affect the predictability of the re-structured model. Although the resulting model

was capable of predicting fairly the outcome of a single-pulse scenario, it was unable to

replicate the observed increased growth when cultures were supplemented with two pulses.

This was attributed to the model not accounting, on one hand, for the inhibitory effects of

the pulse, and on the other, for more complex metabolic processes that may activate

survival mechanisms that allow growth during prolonged periods of nutrient starvation.


272

Indeed, the supplementation of a third pulse of acetic acid increased biomass concentration

by 126 % with respect to batch. However, cells began to reach a saturation point due to the

exhaustion of nutrients (e.g. nitrogen, phosphorus). An improved feeding strategy will thus

need to account for nutrient replenishment.

The increases in biomass, starch, and lipids, as obtained by the combined experimental and

model-based optimisation studies presented in this thesis, provide a favourable outlook for

the establishment of efficient microalgal cultivation systems targeting biofuels production.

However, whilst optimal cultivation strategies can strengthen the dominance of microalgae

as a superior feedstock for advanced biofuels, the success and economic viability of

microalgae-to-biofuel conversion routes is not guaranteed. It is suggested that the most, if

not the only, economically viable approach to commercialise microalgal biofuels is by

implementing biorefineries where microalgal biomass is fully exploited via the co-

production of fuels and chemicals.

Therefore, to assess the potential for biofuels production within a microalgal biorefinery

framework, a case study (Contribution 4, Chapter 6) was carried out to quantify the

production of biobutanol (via the ABE fermentation of microalgal starch) and biodiesel

(via the transesterification of microalgal lipids). By evaluating different microalgal pre-

treatments and conversion routes, it was found that up to 10.31% biobutanol and 3.82 %

biodiesel could be obtained from the same microalgal biomass employed in this thesis (C.

reinhardtii). Besides demonstrating the practical feasibility and economic potential of

microalgae as a biorefinery platform, results from this case-study open the door for future

in-house research endeavours.

In summary, this Ph.D. thesis provides a systematic optimisation framework making use

of experimental and modelling tools for the accurate identification of optimal biofuel-

oriented microalgal cultivation strategies. If extrapolated appropriately, the tools employed

in this thesis can be systematically applied to optimise other bioprocesses of industrial

significance.


273

7.2. Recommendations.

A number of limitations and recommendations for future work are presented below:

The optimal cultivation scenarios established in Chapter 3 and Chapter 4 of this

thesis yielded significant increases in starch and lipid formation by reducing

nitrogen and phosphorus sources, but at the expense of supplying more than double

the amount of acetic acid required by the base case. The organic acid requirements

of mixotrophic species is often considered a drawback given that it can lead to

increased cultivation costs. It has been suggested that microalgal cultivation should

be coupled to water treatment systems rich in organic matter to reduce costs

(Literature review, Chapter 2), but the variability of the composition of effluent

waters can complicate the control of the cultivation variables. Therefore, an

economic analysis focusing on nutrient costs is instead recommended to

additionally identify, through model-based optimisation, biofuel-oriented

cultivation strategies that can minimise the costs of nutrient supply.

The model developed in this thesis accounted for mixotrophic growing conditions

by considering the contributions of both the heterotrophic growth rate (dependent

on the acetic acid concentration) and the phototrophic growth rate (dependent on

the incident light intensity). However, although the effect of different acetic acid

concentrations was experimentally evaluated to validate model predictions, the

effect of different incident light intensities was not. The expression used to portray

the phototrophic growth rate can account for self-shading and photoinhibition

effects (as observed in the literature), but the experimental evaluation of different

light intensities and/or photoperiods is recommended to refine the kinetic

parameters associated to light-limited growth and thus improve the model’s

predictive capacity. Another recommendation is to improve the model by

considering that self-shading effects are not only a function of increasing cell

density but also of Chlorophyll pigment concentrations.

As discussed in Chapter 5, the evaluation of the specific effects of potassium

hydroxide (the pulse neutralising agent) on the dynamics of microalgal growth


274

subject to the pulse-assisted fed-batch operation were not analysed in detail. This

buffering agent, however, was thought to be one of the major factors behind the

observed inhibition of biomass as the concentration of acetic acid in the pulse

increased. Therefore, a more thorough evaluation of the growth-limiting effects of

this component, or the potential use of a more appropriate buffering agent, are

recommended to further improve the scalability of a fed-batch cultivation and to

additionally refine the modelling considerations. In particular, this evaluation can

lead to the incorporation of the growth-limiting effects of culture pH in the

modelling equations. In addition, an evaluation of the microalge’s nitrogen:carbon

balance is recommended to identify a feeding strategy that can simultaneously

sustain growth and also replenish exhausting nutrients.

The case study presented in Chapter 6 evaluated the production of biobutanol from

microalgal carbohydrates and the production of biodiesel from microalgal lipids.

The conversion pathways for these two biofuels involved a two-step sequence

starting with the ABE fermentation, followed by transesterification of fermented

or hydrolysed biomass residues. However, the opposite sequence,

transesterification followed by the ABE fermentation of lipid-extracted biomass

residues was not evaluated. It is recommended that this sequence is also analysed

to assess if the solvent extraction step (used to extract lipids) can lead to improved

(or inhibited) fermentation yields and thus establish the best biorefinery co-

production route based on biofuel conversion yields.

275

References

Adeniyi OM, Azimov U, Burluka A. 2018. Algae biofuel: Current status and future

applications. Renew. Sustain. Energy Rev. 90:316–335.

Adesanya VO, Davey MP, Scott SA, Smith AG. 2014. Kinetic modelling of growth

and storage molecule production in microalgae under mixotrophic and

autotrophic conditions. Bioresour. Technol. 157:293–304.

Andrews JF. 1968. A mathematical model for the continuous culture of

microorganisms utilizing inhibitory substrates. Biotechnol. Bioeng. 10:707–723.

Azadi P, Malina R, Barrett SRH, Kraft M. 2017. The evolution of the biofuel science.

Renew. Sustain. Energy Rev. 76:1479–1484.

Bailey JE. 1998. Mathematical Modeling and Analysis in Biochemical Engineering:

Past Accomplishments and Future Opportunities. Biotechnol. Prog. 14:8–20.

Bajhaiya AK, Dean AP, Driver T, Trivedi DK, Rattray NJW, Allwood JW, Goodacre

R, Pittman JK. 2016. High-throughput metabolic screening of microalgae genetic

variation in response to nutrient limitation. Metabolomics 12:9.

Ball SG, Deschamps P. 2009. The Chlamydomonas Sourcebook: Starch Metabolism.

Chlamydomonas Sourceb. Elsevier 1-40 p.

Ball SG, Dirick L, Decq A, Martiat J-C, Matagne R. 1990. Physiology of starch storage

in the monocellular alga Chlamydomonas reinhardtii. Plant Sci. 66:1–9.

Bankar SB, Survase S a., Ojamo H, Granström T. 2013. Biobutanol: the outlook of an

academic and industrialist. RSC Adv. 3:24734.

Baroukh C, Muñoz-Tamayo R, Steyer J-P, Bernard O. 2014. DRUM: A New

Framework for Metabolic Modeling under Non-Balanced Growth. Application to

the Carbon Metabolism of Unicellular Microalgae. PLoS One 9:e104499.

Barros AI, Gonçalves AL, Simões M, Pires JCM. 2015. Harvesting techniques applied

to microalgae: A review. Renew. Sustain. Energy Rev. 41:1489–1500.

Barsanti L, Gualtieri P. 2018. Is exploitation of microalgae economically and

energetically sustainable? Algal Res. 31:107–115.

Béchet Q, Shilton A, Guieysse B. 2013. Modeling the effects of light and temperature

on algae growth: State of the art and critical assessment for productivity

prediction during outdoor cultivation. Biotechnol. Adv. 31:1648–1663.

Behrens PW, Bingham SE, Hoeksema SD, Cohoon DL, Cox JC. 1989. Studies on the

incorporation of CO2 into starch by Chlorella vulgaris. J. Appl. Phycol. 1:123–

130.

References

276

Bekirogullari M, Fragkopoulos IS, Pittman JK, Theodoropoulos C. 2017. Production

of lipid-based fuels and chemicals from microalgae: An integrated experimental

and model-based optimization study. Algal Res. 23:78–87.

Bekirogullari M, Pittman JK, Theodoropoulos C. 2018. Multi-factor kinetic modelling

of microalgal biomass cultivation for optimised lipid production. Bioresour.

Technol.

Bernard O. 2011. Hurdles and challenges for modelling and control of microalgae for

CO2 mitigation and biofuel production. J. Process Control 21:1378–1389.

Blair MF, Kokabian B, Gude VG. 2014. Light and growth medium effect on Chlorella

vulgaris biomass production. J. Environ. Chem. Eng. 2:665–674.

Bougaran G, Bernard O, Sciandra A. 2010. Modeling continuous cultures of

microalgae colimited by nitrogen and phosphorus. J. Theor. Biol. 265:443–54.

Brányiková I, Maršálková B, Doucha J, Brányik T, Bišová K, Zachleder V, Vítová M.

2010. Microalgae—novel highly efficient starch producers. Biotechnol. Bioeng.

108:766–776.

Brennan L, Owende P. 2010. Biofuels from microalgae—A review of technologies for

production, processing, and extractions of biofuels and co-products. Renew.

Sustain. Energy Rev. 14:557–577.

Brennan L, Owende P. 2013. Biofuels from Microalgae: Towards Meeting Advanced

Fuel Standards. In: Lee, JW, editor. Adv. Biofuels Bioprod. SE - 24. Springer New

York, pp. 553–599.

Cade-Menun BJ, Paytan A. 2010. Nutrient temperature and light stress alter

phosphorus and carbon forms in culture-grown algae. Mar. Chem. 121:27–36.

Caperon J, Meyer J. 1972. Nitrogen-limited growth of marine phytoplankton—II.

Uptake kinetics and their role in nutrient limited growth of phytoplankton. Deep

Sea Res. Oceanogr. Abstr. 19:619–632.

Capros P, De Vita A, Tasios N, Siskos P, Kannavou M, Petropoulos A,

Evangelopoulou S, Zampara M, Papadopoulos D, Nakos C, Paroussos L,

Fragiadakis K, Tsani S, Karkatsoulis P, Fragkos P, Kouvaritakis N, Höglund-

Isaksson N, Winiwarter W, Purohit P, Gomez-Sanabria A, Frank S, Forsell N,

Gusti M, Havlík P, Obersteiner M, Witzke HP, Kesting M. 2016. EU Reference

Scenario 2016: Energy, Transport and GHG emissions - Trends to 2050. Eur.

Comm. Luxembourg: Publications Office of the European Union.

Castro YA, Ellis JT, Miller CD, Sims RC. 2015. Optimization of wastewater

microalgae saccharification using dilute acid hydrolysis for acetone, butanol, and

ethanol fermentation. Appl. Energy 140:14–19.

Chapman SP, Paget CM, Johnson GN, Schwartz J-M. 2015. Flux balance analysis

reveals acetate metabolism modulates cyclic electron flow and alternative

glycolytic pathways in Chlamydomonas reinhardtii. Front. Plant Sci. 6:474.

References

277



78:1–10.

Chen F, Johns MR. 1994. Substrate inhibition of Chlamydomonas reinhardtii by

acetate in heterotrophic culture. Process Biochem. 29:245–252.

Cheng H-H, Whang L-M, Chan K-C, Chung M-C, Wu S-H, Liu C-P, Tien S-Y, Chen

S-Y, Chang J-S, Lee W-J. 2015. Biological Butanol Production from microalgae-

based biodiesel residues by Clostridium acetobutylicum. Bioresour. Technol.

Cherif M, Loreau M. 2010. Towards a more biologically realistic use of Droop’s

equations to model growth under multiple nutrient limitation. Oikos 119:897–

907.

Chisti Y. 2007. Biodiesel from microalgae. Biotechnol. Adv. 25:294–306.

Chojnacka K, Noworyta A. 2004. Evaluation of Spirulina sp. growth in

photoautotrophic, heterotrophic and mixotrophic cultures. Enzyme Microb.

Technol. 34:461–465.

Chundawat SPS, Beckham GT, Himmel ME, Dale BE. 2011. Deconstruction of

Lignocellulosic Biomass to Fuels and Chemicals. Annu. Rev. Chem. Biomol. Eng.

2:121–145.

Colling Klein B, Bonomi A, Maciel Filho R. 2018. Integration of microalgae

production with industrial biofuel facilities: A critical review. Renew. Sustain.

Energy Rev. 82:1376–1392.

Contois DE. 1959. Kinetics of Bacterial Growth: Relationship between Population

Density and Specific Growth Rate of Continuous Cultures. J. Gen. Microbiol.

21:40–50.

Dabes JN, Finn RK, Wilke CR. 1973. Equations of substrate-limited growth: the case

for blackman kinetics. Biotechnol. Bioeng. 15:1159–1177.

Dean AP, Estrada B, Nicholson JM, Sigee DC. 2008. Molecular response of Anabaena

flos-aquae to differing concentrations of phosphorus: A combined Fourier

transform infrared and X-ray microanalytical study. Phycol. Res. 56:193–201.

Delrue F, Setier P-A, Sahut C, Cournac L, Roubaud A, Peltier G, Froment A-K. 2012.

An economic, sustainability, and energetic model of biodiesel production from

microalgae. Bioresour. Technol. 111:191–200.



Energy 88:3331–3335.

Droop MR. 1968. Vitamin B12 and Marine Ecology. IV. The Kinetics of Uptake,


Kingdom 48:689–733.

References

278

Droop MR. 1983. 25 Years of Algal Growth Kinetics A Personal View. Bot. Mar.

26:99.

Edgar TF, Himmelblau DM, Lasdon LS. 2001. Optimization of chemical processes

Second edi. New York: McGraw-Hill.

Enamala MK, Enamala S, Chavali M, Donepudi J, Yadavalli R, Kolapalli B,

Aradhyula TV, Velpuri J, Kuppam C. 2018. Production of biofuels from

microalgae - A review on cultivation, harvesting, lipid extraction, and numerous

applications of microalgae. Renew. Sustain. Energy Rev. 94:49–68.

Eriksen N, Riisgård F, Gunther W, Lønsmann Iversen J. 2007. On-line estimation of

O2 production, CO2 uptake, and growth kinetics of microalgal cultures in a gas-

tight photobioreactor. J. Appl. Phycol. 19:161–174.

European Commission. 2009a. EU. DIRECTIVE 2009/28/EC OF THE EUROPEAN

PARLIAMENT AND OF THE COUNCIL of 23 April 2009 on the promotion of

the use of energy from renewable sources and amending and subsequently

repealing Directived 2001/77/EC and 2003/30/EC. European Commission.

European Commission. 2009b. DIRECTIVE 2009/30/EC OF THE EUROPEAN

PARLIAMENT AND OF THE COUNCIL of 23 April 2009 amending Directive

98/70/EC as regards the specification of petrol, diesel and gas-oil and introducing

a mechanism to monitor and reduce greenhouse gas emissions and amend.

European Comission.

European Commission. 2017a. Report from the Commission to the European

Parliament and the Council in accordance with Article 9 of Directive 98/70/EC

relating to the quality of petrol and diesel fuels. Brussels.

European Commission. 2017b. Impact of higher levels of bio components in transport

fuels in the context of the Directive 98/70/EC of the European Parliament and of

the Council of 13 October 1998, relating to the quality of petrol and diesel fuels

and amending Council Directive 93/12 Report Oct. Luxembourg: Publications

Office of the European Union.

European Environment Agency. 2017. Renewable Energy in Europe - 2017 Update:

Recent growth and knock-on effects. Luxembourg: Publications Office of the

European Union.

Eurostat. 2016. Energy from renewable sources (Shares 2016 results). Energy Data.

http://ec.europa.eu/eurostat/web/energy/data/shares.

Eurostat. 2017. Energy balance sheets (2015 DATA). Luxembourg: Publications

Office of the European Union 115 p.

Fan J, Yan C, Andre C, Shanklin J, Schwender J, Xu C. 2012. Oil accumulation is


reinhardtii. Plant Cell Physiol. 53:1380–1390.

Fernandes B, Dragone G, Abreu A, Geada P, Teixeira J, Vicente A. 2012. Starch

References

279

determination in Chlorella vulgaris—a comparison between acid and enzymatic

methods. J. Appl. Phycol. 24:1203–1208.

Fields FJ, Ostrand JT, Mayfield SP. 2018. Fed-batch mixotrophic cultivation of

Chlamydomonas reinhardtii for high-density cultures. Algal Res. 33:109–117.

Friedman O, Dubinsky Z, Arad S (Malis). 1991. Effect of light intensity on growth

and polysaccharide production in red and blue-green rhodophyta unicells.


Gao K, Orr V, Rehmann L. 2016. Butanol fermentation from microalgae-derived

carbohydrates after ionic liquid extraction. Bioresour. Technol. 206:77–85.

GlobalPetrolPrices. 2018. USA Diesel prices, US gallon.

https://www.globalpetrolprices.com/USA/diesel_prices/.

Gouveia L. 2011. Microalgae as a Feedstock for Biofuels. In: . Microalgae as a Feed.

Biofuels SE - 1. Springer Berlin Heidelberg. SpringerBriefs in Microbiology, pp.

1–69.

Green EM. 2011. Fermentative production of butanol--the industrial perspective. Curr.

Opin. Biotechnol. 22:337–43.

Groom MJ, Gray EM, Townsend PA. 2008. Biofuels and Biodiversity: Principles for

Creating Better Policies for Biofuel Production. Conserv. Biol. 22:602–609.

Grover JP. 1991. Dynamics of Competition among Microalgae in Variable

Environments: Experimental Tests of Alternative Models. Oikos 62:231–243.

Guiry MD. 2012. How many species of algae are there? J. Phycol. 48:1057–1063.

Günerken E, D’Hondt E, Eppink MHM, Garcia-Gonzalez L, Elst K, Wijffels RH.

2015. Cell disruption for microalgae biorefineries. Biotechnol. Adv. 33:243–260.

Harvey BG, Meylemans HA. 2011. The role of butanol in the development of

sustainable fuel technologies. J. Chem. Technol. Biotechnol. 86:2–9.

Ho S-H, Chen W-M, Chang J-S. 2010. Scenedesmus obliquus CNW-N as a potential

candidate for CO2 mitigation and biodiesel production. Bioresour. Technol.

101:8725–8730.

Ho S-H, Huang S-W, Chen C-Y, Hasunuma T. 2013. Bioethanol production using

carbohydrate-rich microalgae biomass as feedstock. Bioresour. Technol.

135:191–198.

Izumo A, Fujiwara S, Oyama Y, Satoh A, Fujita N, Nakamura Y, Tsuzuki M. 2007.

Physicochemical properties of starch in Chlorella change depending on the CO2

concentration during growth: Comparison of structure and properties of pyrenoid

and stroma starch. Plant Sci. 172:1138–1147.

Jeffryes C, Rosenberger J, Rorrer GL. 2013. Fed-batch cultivation and bioprocess

modeling of Cyclotella sp. for enhanced fatty acid production by controlled

References

280

silicon limitation. Algal Res. 2:16–27.

Ji F, Zhou Y, Pang A, Ning L, Rodgers K, Liu Y, Dong R. 2015. Fed-batch cultivation

of Desmodesmus sp. in anaerobic digestion wastewater for improved nutrient

removal and biodiesel production. Bioresour. Technol. 184:116–122.

Jiang Y, Liu J, Jiang W, Yang Y, Yang S. 2015. Current status and prospects of

industrial bio-production of n-butanol in China. Biotechnol. Adv. 33:1493–1501.

Jiménez-González C, Woodley JM. 2010. Bioprocesses: Modeling needs for process

evaluation and sustainability assessment. Comput. Chem. Eng. 34:1009–1017.

Johnson X, Alric J. 2013. Central Carbon Metabolism and Electron Transport in

Chlamydomonas reinhardtii: Metabolic Constraints for Carbon Partitioning

between Oil and Starch. Eukaryot. Cell 12:776–793.

Jones DT, Woods DR. 1986. Acetone-butanol fermentation revisited. Microbiol. Rev.

50:484–524.

Juneja A, Ceballos R, Murthy G. 2013. Effects of Environmental Factors and Nutrient

Availability on the Biochemical Composition of Algae for Biofuels Production:

A Review. Energies 6:4607–4638.

Al Ketife AMD, Judd S, Znad H. 2016. A mathematical model for carbon fixation and

nutrient removal by an algal photobioreactor. Chem. Eng. Sci. 153:354–362.

Kim J, Yoo G, Lee H, Lim J, Kim K, Kim CW, Park MS, Yang J-W. 2013. Methods

of downstream processing for the production of biodiesel from microalgae.

Biotechnol. Adv. 31:862–876.

Kim KH, Choi IS, Kim HM, Wi SG, Bae H-J. 2014. Bioethanol production from the

nutrient stress-induced microalga Chlorella vulgaris by enzymatic hydrolysis and

immobilized yeast fermentation. Bioresour. Technol. 153:47–54.

Kiparissides A, Koutinas M, Kontoravdi C, Mantalaris A, Pistikopoulos EN. 2011.

‘Closing the loop’ in biological systems modeling — From the in silico to the in

vitro. Automatica 47:1147–1155.

Klausmeier CA, Litchman E, Levin SA. 2004. Phytoplankton growth and

stoichiometry under multiple nutrient limitation. Limnol. Oceanogr. 49:1463–

1470.

Kumar A, Guria C, Chitres G, Chakraborty A, Pathak AK. 2016. Modelling of



bubble column. Bioresour. Technol. 218:1021–1036.

Kumar M, Gayen K. 2011. Developments in biobutanol production: New insights.

Appl. Energy 88:1999–2012.

Küüt A, Ilves R, Küüt K, Raide V, Ritslaid K, Olt J. 2017. Influence of European

Union Directives on the Use of Liquid Biofuel in the Transport Sector. Procedia

References

281

Eng. 187:30–39.

Kwon HK, Oh SJ, Yang H-S. 2013. Growth and uptake kinetics of nitrate and

phosphate by benthic microalgae for phytoremediation of eutrophic coastal

sediments. Bioresour. Technol. 129:387–95.

Lam MK, Lee KT. 2012. Microalgae biofuels: A critical review of issues, problems

and the way forward. Biotechnol. Adv. 30:673–90.

Lee E, Jalalizadeh M, Zhang Q. 2015a. Growth kinetic models for microalgae

cultivation: A review. Algal Res. 12:497–512.

Lee Y-C, Lee K, Oh Y-K. 2015b. Recent nanoparticle engineering advances in

microalgal cultivation and harvesting processes of biodiesel production: A

review. Bioresour. Technol. 184:63–72.

Leong W-H, Lim J-W, Lam M-K, Uemura Y, Ho Y-C. 2018. Third generation

biofuels: A nutritional perspective in enhancing microbial lipid production.

Renew. Sustain. Energy Rev. 91:950–961.

Litchman E, Klausmeier CA, Miller JR, Schofield OM, Falkowski PG. 2006. Multi-

nutrient, multi-group model of present and future oceanic phytoplankton

communities. Biogeosciences 3:585–606.

Liu J, Huang J, Sun Z, Zhong Y, Jiang Y, Chen F. 2011. Differential lipid and fatty

acid profiles of photoautotrophic and heterotrophic Chlorella zofingiensis:

Assessment of algal oils for biodiesel production. Bioresour. Technol. 102:106–

110.

Lowrey J, Brooks MS, McGinn PJ. 2015. Heterotrophic and mixotrophic cultivation

of microalgae for biodiesel production in agricultural wastewaters and associated

challenges---a critical review. J. Appl. Phycol. 27:1485–1498.

MacIntyre HL, Cullen JJ. 2005. Algal Culturring Techniques. Ed. Robert A. Andersen.

Burlington, MA: Elsevier Academic Press 287-326 p.

Mairet F, Bernard O, Masci P, Lacour T, Sciandra A. 2011. Modelling neutral lipid



Mankad T, Bungay HR. 1988. Model for microbial growth with more than one limiting

nutrient. J. Biotechnol. 7:161–166.

Markou G, Angelidaki I, Georgakakis D. 2012a. Microalgal carbohydrates: an


bioconversion technologies for production of biofuels. Appl. Microbiol.

Biotechnol. 96:631–645.

Markou G, Chatzipavlidis I, Georgakakis D. 2012b. Carbohydrates Production and

Bio-flocculation Characteristics in Cultures of Arthrospira (Spirulina) platensis:

Improvements Through Phosphorus Limitation Process. BioEnergy Res. 5:915–

925.

References

282

Martínez Sancho M, Jiménez Castillo JM, El Yousfi F. 1997. Influence of phosphorus

concentration on the growth kinetics and stoichiometry of the microalga

Scenedesmus obliquus. Process Biochem. 32:657–664.

McNichol J, MacDougall KM, Melanson JE, McGinn PJ. 2012. Suitability of Soxhlet

Extraction to Quantify Microalgal Fatty Acids as Determined by Comparison

with In Situ Transesterification. Lipids 47:195–207.

Mirzaie MAM, Kalbasi M, Mousavi SM, Ghobadian B. 2016. Investigation of

mixotrophic, heterotrophic, and autotrophic growth of Chlorella vulgaris under

agricultural waste medium. Prep. Biochem. Biotechnol. 46:150–156.

Molina-Grima E, García-Camacho F, Sánchez-Pérez JA, Fernández-Sevilla JM,

Acién-Fernández FG, Contreras-Gómez A. 1994. A mathematical model of

microalgal growth in light-limited chemostat culture. J. Chem. Technol.


Monod J. 1949. The Growth of Bacterial Cultures. Annu. Rev. Microbiol. 3:371–394.

Moon HG, Jang Y-S, Cho C, Lee J, Binkley R, Lee SY. 2016. One hundred years of

clostridial butanol fermentation. Ed. Michael Sauer. FEMS Microbiol. Lett. 363.

Morales-Sánchez D, Kyndt J, Ogden K, Martinez A. 2016. Toward an understanding

of lipid and starch accumulation in microalgae: A proteomic study of Neochloris

oleoabundans cultivated under N-limited heterotrophic conditions. Algal Res.

20:22–34.

Moseley J, Grossman AR. 2009. Phosphate Metabolism and Responses to Phosphorus

Deficiency. Chlamydomonas Sourceb.:189–215.

Nan C, Dong S. 2004. Comparative studies on phosphorus uptake and growth kinetics

of the microalga Tetraselmis subcordiformis and the macroalga Ulva pertusa. J.

Ocean Univ. China 3:56–59.

Nigam PS, Singh A. 2011. Production of liquid biofuels from renewable resources.

Prog. Energy Combust. Sci. 37:52–68.

Nzayisenga JC, Eriksson K, Sellstedt A. 2018. Mixotrophic and heterotrophic

production of lipids and carbohydrates by a locally isolated microalga using

wastewater as a growth medium. Bioresour. Technol. 257:260–265.

Oh Y-K, Hwang K-R, Kim C, Kim JR, Lee J-S. 2018. Recent developments and key

barriers to advanced biofuels: A short review. Bioresour. Technol. 257:320–333.

Olivier B, Isabelle Q. 2010. Dynamic Models of Biochemical Processes: Properties of

Models. In: . Bioprocess Control. Wiley-Blackwell, pp. 17–45.

Packer A, Li Y, Andersen T, Hu Q, Kuang Y, Sommerfeld M. 2011. Growth and

neutral lipid synthesis in green microalgae: a mathematical model. Bioresour.

Technol. 102:111–7.

de Prada C, Hose D, Gutierrez G, Pitarch JL. 2018. Developing grey-box dynamic

References

283

process models. IFAC-PapersOnLine 51:523–528.

Pragya N, Pandey KK, Sahoo PK. 2013. A review on harvesting, oil extraction and

biofuels production technologies from microalgae. Renew. Sustain. Energy Rev.

24:159–171.

Ra CH, Kang C-H, Kim NK, Lee C-G, Kim S-K. 2015. Cultivation of four microalgae

for biomass and oil production using a two-stage culture strategy with salt stress.

Renew. Energy 80:117–122.

RAEng. 2017. Sustainability of liquid biofuels. London: Royal Academy of

Engineering.

Rashid N, Ur Rehman MS, Sadiq M, Mahmood T, Han J-I. 2014. Current status, issues

and developments in microalgae derived biodiesel production. Renew. Sustain.

Energy Rev. 40:760–778.

Raslavičius L, Semenov VG, Chernova NI, Keršys A, Kopeyka AK. 2014. Producing

transportation fuels from algae: In search of synergy. Renew. Sustain. Energy

Rev. 40:133–142.

Rasul I, Azeem F, Siddique MH, Muzammil S, Rasul A, Munawar A, Afzal M, Ali

MA, Nadeem H. 2017. Algae Biotechnology: A Green Light for Engineered

Algae. Algae Based Polym. Blends, Compos.:301–334.

Ribeiro LA, Silva PP da. 2013. Surveying techno-economic indicators of microalgae

biofuel technologies. Renew. Sustain. Energy Rev. 25:89–96.

Riekhof WR, Benning C. 2009. Glycerolipid Biosynthesis. Chlamydomonas

Sourceb.:41–68.

Rodolfi L, Chini Zittelli G, Bassi N, Padovani G, Biondi N, Bonini G, Tredici MR.

2009. Microalgae for oil: strain selection, induction of lipid synthesis and outdoor

mass cultivation in a low-cost photobioreactor. Biotechnol. Bioeng. 102:100–12.

Salama E-S, Hwang J-H, El-Dalatony MM, Kurade MB, Kabra AN, Abou-Shanab

RAI, Kim K-H, Yang I-S, Govindwar SP, Kim S, Jeon B-H. 2018. Enhancement

of microalgal growth and biocomponent-based transformations for improved

biofuel recovery: A review. Bioresour. Technol. 258:365–375.

Scaife MA, Merkx-Jacques A, Woodhall DL, Armenta RE. 2015. Algal biofuels in

Canada: Status and potential. Renew. Sustain. Energy Rev. 44:620–642.

Schwartz J-M, Soons Z. 2019. Metabolic Models. Encycl. Bioinforma. Comput.

Biol.:438–444.

Sforza E, Enzo M, Bertucco A. 2014. Design of microalgal biomass production in a

continuous photobioreactor: An integrated experimental and modeling approach.

Chem. Eng. Res. Des. 92:1153–1162.

Shuba ES, Kifle D. 2018. Microalgae to biofuels: ‘Promising’ alternative and

renewable energy, review. Renew. Sustain. Energy Rev. 81:743–755.

References

284

Shuler ML, Kargi F. 1992. Bioprocess Engineering: Basic Concepts. Prentice Hall 479

p.

Singh J, Gu S. 2010. Commercialization potential of microalgae for biofuels

production. Renew. Sustain. Energy Rev. 14:2596–2610.

Singh SP, Singh P. 2015a. Effect of temperature and light on the growth of algae

species: A review. Renew. Sustain. Energy Rev. 50:431–444.

Singh SP, Singh P. 2015b. Effect of temperature and light on the growth of algae

species: A review. Renew. Sustain. Energy Rev. 50:431–444.

Sinha SK, Kumar M, Guria C, Kumar A, Banerjee C. 2017. Biokinetic model-based

multi-objective optimization of Dunaliella tertiolecta cultivation using elitist non-

dominated sorting genetic algorithm with inheritance. Bioresour. Technol.

242:206–217.

Slade R, Bauen A. 2013. Micro-algae cultivation for biofuels: Cost, energy balance,

environmental impacts and future prospects. Biomass and Bioenergy 53:29–38.

Spalding MH. 2009. The CO2-Concentrating Mechanism and Carbon Assimilation.

Chlamydomonas Sourceb.:257–301.

Spijkerman E, de Castro F, Gaedke U. 2011. Independent Colimitation for Carbon

Dioxide and Inorganic Phosphorus. PLoS One 6:1–12.

Su C-H, Chien L-J, Gomes J, Lin Y-S, Yu Y-K, Liou J-S, Syu R-J. 2011. Factors

affecting lipid accumulation by Nannochloropsis oculata in a two-stage

cultivation process. J. Appl. Phycol. 23:903–908.

Su Y, Song K, Zhang P, Su Y, Cheng J, Chen X. 2017. Progress of microalgae

biofuel’s commercialization. Renew. Sustain. Energy Rev. 74:402–411.

Suganya T, Varman M, Masjuki HH, Renganathan S. 2016. Macroalgae and

microalgae as a potential source for commercial applications along with biofuels

production: A biorefinery approach. Renew. Sustain. Energy Rev. 55:909–941.

Surendhiran D, Vijay M, Sivaprakash B, Sirajunnisa A. 2014. Kinetic modeling of

microalgal growth and lipid synthesis for biodiesel production. 3 Biotech:1–7.

Surisetty K, Hoz Siegler HD la, McCaffrey WC, Ben-Zvi A. 2010. Robust modeling

of a microalgal heterotrophic fed-batch bioreactor. Chem. Eng. Sci. 65:5402–

5410.

Turon V, Baroukh C, Trably E, Latrille E, Fouilland E, Steyer J-P. 2014. Use of

fermentative metabolites for heterotrophic microalgae growth: Yields and

kinetics. Bioresour. Technol. 175C:342–349.

Velasquez-Orta SB, Garcia-Estrada R, Monje-Ramirez I, Harvey A, Orta Ledesma

MT. 2014. Microalgae harvesting using ozoflotation: Effect on lipid and FAME

recoveries. Biomass and Bioenergy 70:356–363.

References

285

Velayudhan A. 2014. Overview of integrated models for bioprocess engineering. Curr.

Opin. Chem. Eng. 6:83–89.

Venkata Mohan S, Rohit MV, Chiranjeevi P, Chandra R, Navaneeth B. 2015.

Heterotrophic microalgae cultivation to synergize biodiesel production with

waste remediation: Progress and perspectives. Bioresour. Technol. 184:169–178.

Vitova M, Bisova K, Kawano S, Zachleder V. 2015. Accumulation of energy reserves

in algae: From cell cycles to biotechnological applications. Biotechnol. Adv.

33:1204–1218.

Volesky B, Votruba J. 1992. Modeling and Optmization of Fermentation Processes.

Elsevier Science Publishers B.V.

Wang T, Tian X, Liu T, Wang Z, Guan W, Guo M, Chu J. 2017. A two-stage fed-batch

heterotrophic culture of Chlorella protothecoides that combined nitrogen

depletion with hyperosmotic stress strategy enhanced lipid yield and productivity.

Process Biochem. 60:74–83.

Wang X, Ruan Z, Sheridan P, Boileau D, Liu Y, Liao W. 2015. Two-stage

photoautotrophic cultivation to improve carbohydrate production in

Chlamydomonas reinhardtii. Biomass and Bioenergy 74:280–287.

Wang Y, Chu J, Zhuang Y, Wang Y, Xia J, Zhang S. 2009. Industrial bioprocess

control and optimization in the context of systems biotechnology. Biotechnol.

Adv. 27:989–995.

Wang Y, Guo W, Cheng C-L, Ho S-H, Chang J-S, Ren N. 2016. Enhancing bio-

butanol production from biomass of Chlorella vulgaris JSC-6 with sequential

alkali pretreatment and acid hydrolysis. Bioresour. Technol. 200:557–564.

Xin L, Hong-ying H, Ke G, Ying-xue S. 2010. Effects of different nitrogen and

phosphorus concentrations on the growth, nutrient uptake, and lipid accumulation

of a freshwater microalga Scenedesmus sp. Bioresour. Technol. 101:5494–5500.

Xu H, Miao X, Wu Q. 2006. High quality biodiesel production from a microalga

Chlorella protothecoides by heterotrophic growth in fermenters. J. Biotechnol.

126:499–507.

Xue C, Zhao J-B, Chen L-J, Bai F-W, Yang S-T, Sun J-X. 2014. Integrated butanol

recovery for an advanced biofuel: current state and prospects. Appl. Microbiol.


Yao C-H, Ai J-N, Cao X-P, Xue S. 2013. Characterization of cell growth and starch

production in the marine green microalga Tetraselmis subcordiformis under

extracellular phosphorus-deprived and sequentially phosphorus-replete

conditions. Appl. Microbiol. Biotechnol. 97:6099–6110.

Yao C, Ai J, Cao X, Xue S, Zhang W. 2012. Enhancing starch production of a marine

green microalga Tetraselmis subcordiformis through nutrient limitation.


References

286

Yoo SJ, Kim JH, Lee JM. 2014. Dynamic modelling of mixotrophic microalgal

photobioreactor systems with time-varying yield coefficient for the lipid

consumption. Bioresour. Technol. 162:228–35.

Yoon H, Klinzing G, Blanch HW. 1977. Competition for mixed substrates by

microbial populations. Biotechnol. Bioeng. 19:1193–1210.

Zhan J, Rong J, Wang Q. 2017. Mixotrophic cultivation, a preferable microalgae

cultivation mode for biomass/bioenergy production, and bioremediation,

advances and prospect. Int. J. Hydrogen Energy 42:8505–8517.

Zhang X-W, Chen F, Johns MR. 1999. Kinetic models for heterotrophic growth of

Chlamydomonas reinhardtii in batch and fed-batch cultures. Process Biochem.

35:385–389.

Zheng H, Gao Z, Yin F, Ji X, Huang H. 2012. Effect of CO2 supply conditions on lipid

production of Chlorella vulgaris from enzymatic hydrolysates of lipid-extracted

microalgal biomass residues. Bioresour. Technol. 126:24–30.

Zhu L. 2013. The combined production of ethanol and biogas from microalgal

residuals to sustain microalgal biodiesel: A theoretical evaluation. Biofuels,

Bioprod. Biorefining 8:7–15.

Zhu L. 2015. Biorefinery as a promising approach to promote microalgae industry: An

innovative framework. Renew. Sustain. Energy Rev. 41:1376–1384.

Zhu S, Wang Y, Huang W, Xu J, Wang Z, Xu J, Yuan Z. 2014. Enhanced

Accumulation of Carbohydrate and Starch in Chlorella zofingiensis Induced by

Nitrogen Starvation. Appl. Biochem. Biotechnol. 174:2435–2445.

287

APPENDIX A

A.1 Preparation of TAP medium.

Preparation of 1 L of standard TAP medium is carried out by mixing stock solutions (SS)

as per the quantities specified in Table 7-1. Final medium is filled up to 1 L of deionized

water and adjusted to pH 7. Adjustment of pH is done with either hydrochloric acid (HCl)

3 M, or potassium hydroxide (KOH) 3 M. All the culture media was autoclaved before

inoculation.

Table 7-1 Components of stock solutions used for 1 L of TAP medium.

Stock

solution

(SS)

Component Concentration

in SS

Qty. of SS

in 1 L of

TAP

Final concentration of

limiting nutrients

Tris base H2NC(CH2OH)3 -

2.42 g 2.42 g/L (0.28 gN/L)

TAP-Salts

NH4Cl 15 g/L 25 mL 0.375 g/L (0.098 gN/L)

MgSO4·7H2O 4 g/L 0.1 g/L

CaCl2·2H2O 2 g/L 0.05 g/L

Phosphate

solution

K2HPO4 288 g/L 0.375 mL 0.108 g/L (0.058gPO4/L)

KH2PO4 144 g/L 0.054 g/L (0.037gPO4/L)

Trace

elements

Na2EDTA·2H2O 50 g/L 1 mL 0.05 g/L (0.002 gN/L)

ZnSO4.7H2O 22 g/L 0.022 g/L

H3BO3 11.4 g/L 0.0114 g/L

MnCl2.4H2O 5.06 g/L 0.0050 g/L

FeSO4.7H2O 4.99 g/L 0.0049 g/L

CoCl2.6H2O 1.61 g/L 0.0016 g/L

CuSO4.5H2O 1.57 g/L 0.0015 g/L

(NH4)6Mo7O24.4H2O 1.1 g/L 0.0011 g/L (7.5E-5gN/L)

Acetic Acid CH3COOH - 1 mL 1.05 g/L (0.42 gC/L)

optimisation of biofuels production from microalgal biomass

Documents