hydrophobic effect: thermodynamics of cationic & anionic surfactant
TRANSCRIPT
Vilnius universityFaculty of natural sciences
IVth year student of Biophysics BSc
Povilas Norvaišas
Bachelor thesis
Hydrophobic effect: Thermodynamics of cationic &
anionic surfactant interaction and laws of additivity in
the structure-based drug design
Reviewed by: Supervisors:
Dr. Vytautas Smirnovas Dr. Daumantas Matulis
Dr. Visvaldas Kairys
Vilnius, 2012
Vilniaus universitetasGamtos mokslų fakultetas
Biofizikos studijų programos IV kurso studentas
Povilas Norvaišas
Bakalauro darbas
Hidrofobinis efektas: katijoninių ir anijoninių
detergentų sąveikos termodinamika bei adityvumo
dėsniai struktūra paremtame vaistų kūrime
Recenzavo: Darbo vadovai:
Dr. Vytautas Smirnovas Dr. Daumantas Matulis
Dr. Visvaldas Kairys
Vilnius, 2012
Bakalauro darbas
Hidrofobinis efektas: katijoninių ir anijoninių
detergentų sąveikos termodinamika bei
adityvumo dėsniai struktūra paremtame vaistų
kūrime
Vilniaus Universitetas Biotechnologijos institutas
Biotermodinamikos ir vaistų tyrimo skyrius
Įvertinimas:
Darbo vadovai:
Dr. Daumantas Matulis
Dr. Visvadas Kairys
Studentas:
Povilas Norvaišas
Vilnius, 2012
Vilnius University
Faculty of Natural Sciences
Department of Biochemistry and Biophysics
IVth year student of Biophysics BScPovilas Norvaišas
Bachelor thesis
Hydrophobic effect: Thermodynamics of cationic &
anionic surfactant interaction and laws of additivity
in the structure-based drug design
Summary
Hydrophobic interactions are ubiquitous in the biological systems. They are involved in
protein folding, membrane formation, protein to ligand binding and many other processes. Even
though these interactions are crucial, it is yet impossible to comprehend them in a direct way – we
need simplifications. This thesis describes two different, but interrelated approaches of employing
such simplifications. In the first study, thermodynamics of hydrophobic effect were investigated
by employing a simple model system of oppositely charged surfactants. The results gathered
with isothermal titration calorimetry indicate that the hydrophobic aggregation in such systems
is driven primarily by the huge negative enthalpy and not positive entropy change – contrary
to the common theory. In the second study different computational modelling approaches and
simplified solvent effect representations were compared for their accuracy in predicting binding
energy of inhibitors to the human carbonic anhydrase II. The results suggested that solvent
representation needed not to depend on the binding event and that the simple docking method
might also be the best one.
Vilnius, 2012
Vilniaus universitetas
Gamtos mokslų fakultetas
Biochemijos ir biofizikos katedra
Biofizikos studijų programos IV kurso studentasPovilas Norvaišas
Bakalauro darbas
Hidrofobinis efektas: katijoninių ir anijoninių
detergentų sąveikos termodinamika bei adityvumo
dėsniai struktūra paremtame vaistų kūrime
Santrauka
Hidrofobinė sąveika biologinėse sistemose yra esminė. Jos įnašas svarbus daugybėje
reakcijų: baltymų susivyniojime, membranų susidaryme, baltymų ir ligandų jungimesi ir t.t.
Nepaisant šios sąveikos svarbos, mes dar nesugabame tirti tokių sudėtingų biologinių procesų
tiosiogiai – mums būtini supaprastinti modeliai. Šiame bakalauro darbe pristatomi du skirtingi,
tačiau susiję tyrimai, kuriuose buvo panaudoti tokie modeliai. Pirmajame, izoterminio titrav-
imo kalorimetrijos metodu buvo tirta hidrofobinio efekto termodinamika, kaip modelinę sistemą
naudojant priešingą krūvį turinčius detergentus. Rezultatai parodė, kad hidrofobinė agregacija
tokiose sistemose yra nulemta didelio neigiamo entalpijos, o ne entropijos, įnašo – priešingai
negu teigia klasikinė hidrofobinio efekto teorija. Antrajame tyrime buvo lyginami įvairūs kompi-
uterinio modeliavimo metodai ir tirpiklio reprezentacijos juose, siekiant nustatyti optimaliausią
metodą vertinant slopiklių sąveiką su žmogaus karbonanhidraze II. Analizė atskleidė, kad pa-
prasčiausi modeliavimo metodai gali būti ir geriausi, o tirpiklio įtaka gali būti reprezentuota ne
vien jungimosi reakcijos sąvybėmis.
Vilnius, 2012
Povilas Norvaišas CONTENTS
Contents
1 Introduction 6
1.1 Thermodynamics of cationic & anionic surfactant interaction . . . . . . . . . . . . 6
1.2 Laws of additivity in the structure-based drug design . . . . . . . . . . . . . . . . . 6
2 Literature overview 8
2.1 The Hydrophobic effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.1 Historic overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.2 ”Classical” hydrophobic effect . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.3 ”Non-classical” hydrophobic effect? . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.4 The fingerprint of the hydrophobic effect . . . . . . . . . . . . . . . . . . . . 11
2.2 Surfactants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.1 Mixtures of cationic and anionic surfactants . . . . . . . . . . . . . . . . . . 13
2.2.2 Interaction of cationic and anionic surfactants . . . . . . . . . . . . . . . . . 14
2.3 Isothermal Titration Calorimetry (ITC) . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.1 Adiabatic and heat flux ITC . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.2 Planning of the ITC experiment . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4 Additivity – the 4th law of thermodynamics . . . . . . . . . . . . . . . . . . . . . . 18
2.5 Carbonic anhydrases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.5.1 Structural properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.5.2 Drug design for CAs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.5.3 Inhibitors of CAs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.6 Computational structure-based drug design . . . . . . . . . . . . . . . . . . . . . . 23
2.7 Docking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.7.1 Posing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.7.2 Scoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.7.3 Poisson-Boltzman surface area method . . . . . . . . . . . . . . . . . . . . . 25
3 Thermodynamics of Cationic and Anionic Surfactant Interaction 27
3.1 Materials & Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.1.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.1.2 Isothermal Titration Calorimetry . . . . . . . . . . . . . . . . . . . . . . . . 28
3.1.3 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.1.4 Statistical analysis of the data . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 The mathematical model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.3.1 Dependence on the order of titration . . . . . . . . . . . . . . . . . . . . . . 33
1
3.3.2 Dependence on the surfactant concentration . . . . . . . . . . . . . . . . . . 34
3.3.3 Dependence on the aliphatic chain length . . . . . . . . . . . . . . . . . . . 36
3.3.4 Dependence on the ionic force of the solvent . . . . . . . . . . . . . . . . . . 38
3.3.5 Dependence on the experimental temperature . . . . . . . . . . . . . . . . . 39
3.3.6 Structure of the aggregate . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4 Laws of additivity in the structure-based drug design 44
4.1 Materials & methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.1.1 Protein and its structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.1.2 Inhibitors and their structures . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.1.3 Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.1.4 Database . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.1.5 Docking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.1.6 PBSA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.2.1 Docking with Vdock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2.2 PBSA potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2.3 Expressions of total modelled binding energy . . . . . . . . . . . . . . . . . 48
4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5 Conclusion 53
6 Acknowledgements 54
References 63
7 Supplementary material 64
2
Nomenclature
[cell] Concentration of a surfactant in the cell of the calorimeter
[M ]0 Concentration of the compounds in the reaction cell with respect to the dilution effect.
Used in ITC data analysis
[X]0 Concentration of the ligand from the syringe in the reaction cell. Used in ITC data analysis
∆∆Gpre Change in the Gibbs free energy upon addition of methylene group
∆Cp Heat capacity change at constant pressure
∆Cliquidp Heat capacity for a liquid aggregate.
∆Csolidp Heat capacity for a solid aggregate.
∆G Gibbs free energy change for reaction
∆Gagg Entropy change for aggregation multiplied by absolute temperature
∆Gagg Gibbs free energy change for aggregation
∆Gexp Gibbs free energy change determined in TSA experiment.
∆Gmod Gibbs free energy change determined by docking
∆H Enthalpy change for reaction
δH Partial enthalpy change for one injection.
∆Hagg Enthalpy change for aggregation
∆Hexp Enthalpy change of a reaction, determined by direct integration of the dosing curve.
∆Hfit Enthalpy change of a reaction, determined by a fitting the dosing curve
∆Hfus Enthalpy of fusion.
∆Hpre Enthalpy change fro precipitation
∆Hstr Enthalpy of structural rearrangements prior to fusion.
∆SASA Solvent accessible surface area change upon binding
∆T∆Sobs Observed entropy change upon addition of one methylene group
∆∆G Change in the Gibbs free energy upon addition of one methylene group
∆∆Gobs Observed Gibbs free energy change upon addition of one methylene group
∆∆Hexp Enthalpy change upon addition of one methylene group
3
Povilas Norvaišas CONTENTS
∆aggHalk Empirical enthalpy change for linear alkanes.
∆aggHalm Modelled enthalpy change for the surfactants system.
c Coefficient, used to evaluate ITC’s capability of determining Kb at the given experimental
concentration
cmc Critical micelle concentration
cmcmix Critical micelle concentration for a particular mixture of surfactants
cmt Critical micelle temperature. Analogoud to TK
Kb Binding constant
Kfitb Binding constant determined by fitting the dosing curve
Koff Disassociation constant, term used in the SPR studies.
Kon Association constant, term used in the SPR studies. Analogous to Kb.
m Total aliphatic chain length for the system of surfactants
N Stoichiometry of the reaction
n Aliphatic chain length for a single surfactant
Nfit Stoichiometry determined by fitting the dosing curve.
SAvdw Van der Waals surface area.
SASAHyd Hydrophobic solvent accessible surface area
T∆S Entropy change for reaction multiplied by temperature
TK Krafft temperature. Analogous to cmt.
V Colintr Coulomb electrostatic potential for the protein-ligand interaction
V Colint Coulomb electrostatic potential for ligand
V Dih Dihedral potential for ligand
V P B Poisson-Boltzmann electrostatic potential
V vdwattr Van der Waals attraction potential for protein-ligand interaction
V vdwint Van der Waals potential for ligand
V vdwrep Van der Waals repulsion potential for protein-ligand interaction
CPC Cetylpyridinium chloride C21H33CN ·H2O.
DPCl Decylpyridinium chloride C15H26C1N ·H2O
4
Povilas Norvaišas CONTENTS
FFT Fast Fourier Transformation
MCS Micro Calorimetry System
SDS Sodium dodecyl sulfate, C12H25OSO3Na)
SPR Surface plasmon resonance
TSA Thermal shift assay (ThermoFluor)
5
1 Introduction
Biological systems are believed to have emerged in the aqueous medium and even now,
billions of years past, both most primitive and sophisticated forms of life are fully dependent on
it. A significant role is played here by the hydrophobic effect. The biochemical machinery has
developed in such a way that it employs hydrophobicity for its own needs in the interactions
between molecules or formation of biological compartments. Even though these processes are as
ubiquitous as life is, we can not yet comprehend them by direct approach – we need simplifications.
This is is the main theme of my bachelor thesis – simplifications for complex phenomenon regarding
hydrophobic effect and the contribution of solvent in general. It is further divided into two research
projects which tackle the same main idea from completely different perspectives.
1.1 Thermodynamics of cationic & anionic surfactant interaction
The main goal behind this research project was to investigate thermodynamics of hy-
drophobic effect for highly hydrophobic compounds by employing a simplistic system of cationic
and anionic surfactants. This could only be accomplished by investigating a number of proper-
ties regarding the thermodynamics of cationic and anionic surfactant complex formation with the
technique of isothermal titration calorimetry:
• dependence on the surfactant concentration
• dependence on the aliphatic chain length of the surfactants
• dependence on the ionic strength of the solvent
• dependence on the temperature
• structure of the oppositely charged surfactant complex.
1.2 Laws of additivity in the structure-based drug design
Overall different approach was taken in this study - testing of different computational
methods and representations of the possible solvent effect to find the best one to evaluate ligand
to protein binding energies in silico for human carbonic anhydrase II and inhibitors synthesised
in the Department of Biothermodynamics and Drug Design of Vilnius University Institute of
Biotechnology. In order to accomplish this goal several computational modelling methods had
to be taken:
• docking of inhibitors with VDock
• estimation of solvent effect according to Poisson-Boltzmann surface area (PBSA) method
6
Povilas Norvaišas Introduction
• development and evaluation of total modelled energy expressions with different representa-
tions of possible solvent effect.
7
2 Literature overview
2.1 The Hydrophobic effect
If water is considered to be the reason behind life on earth, so must be the hydrophobic
effect (Chaplin 2012). Formation of membranes, cellular compartments and the folding of polypep-
tides into native proteins is believed to be driven largely by hydrophobic interactions (Tanford 1978;
1997). Due to them the whole set of amphiphilic compounds as lipids or polypeptides are capable
of establishing sophisticated, self-assembled structures at the molecular scale and the only thing
required is aqueous media. Therefore, the hydrophobic effect can be regarded as an ”organiser” of
the biological world at the molecular scale. Besides all of this, it is hard to find any interactions
between biological compounds which would not be influenced by hydrophobicity. It might be a
ligand binding to a protein, protein folding, assembly of the protein complex, signal transduction
and many more, plenty of which have enormous biological significance. Therefore, the fundamental
processes behind hydrophobic effect are profoundly interesting and research on them might as well
help us better understand many biological/biochemical phenomena at the molecular scale.
2.1.1 Historic overview
The structural reason behind solubility and insolubility of various compounds was first
identified in 1891 by german scientist Isidor Traube (Tanford 1997). During his studies of thin
layers of amphiphilic compounds, he discovered that such molecules orient their hydrophobic parts
out of the aqueous solution. Further research done by Irvin Langmuir extended the ideas of Traube
and resulted into the series of publications, because of which Langmuir is considered the ”father”
of the science of surfactants (Tanford 1997). At the beginning of XXth century both chemistry and
physics underwent a revolution which expanded the toolkit of scientists and enabled observation of
phenomena in atomistic scale. In 1919 Moore together with Winmill hypothesised hydrogen bonds
(Moore and Winmill 1912). It was not long before 1920 when it was understood that there should
be an extensive network of such bonding in water (Latimer and Rodebush 1920) that might be
responsible for the extraordinary features of this solvent (Chaplin 2012; Bernal and Fowler 1933).
In 1945 Frank and Evans suggested that it was not an attractive force behind the aggregation of hy-
drophobic compounds in aqueous medium, but structure of the solvent itself (Mizutani 2011). Soon
it became clear that such theory would have wide implications when applied to biological systems.
The research led by Charles Tanford and Walter Kauzmann revealed significance of hydrophobic
effect in the formation of micelles, membranes and folding of the proteins (Tanford 1978; 1997;
1980; Kauzmann 1959). The main thermodynamic quantities of the dissolution reactions were
determined and the basis for the research on hydrophobic effect in biological systems established
8
Povilas Norvaišas Literature overview
during the period between 50s and 70s. These ideas were later widely applied in many fields of life
sciences – drug design, computational modelling of protein-ligand interaction and protein folding,
etc.
2.1.2 ”Classical” hydrophobic effect
The main contribution of Frank & Evans to the research on hydrophobic effect was that
they hypothesised an extensive network of hydrogen bonds in the liquid water which is responsible
for the strong cohesion of solvent molecules and repulsion of non-hydrated molecules (Frank and
Evans 1945). It is now known that water molecules form 3.6 hydrogen bonds on average with
one another. This property is unique to water, as there are no compounds of similar molecular
mass, which would form such tight bonding in between themselves (Chaplin 2012). As a result of
such network of bonds, any substance incapable of incorporating and forming sufficient number of
hydrogen bonds with the solvent will be forced to aggregate with similar non-polar compounds –
the hydrophobic effect will happen (Tanford 1978). According to the classical theory this process
is primarily driven by the positive entropy change due to a gain of motional freedom of water
molecules upon loss of contact with the hydrophobic compound. The bulk water molecules are
capable of freely rotating due to brownian motion, because in every direction there will be another
water molecule capable of forming favorable hydrogen bond. However, in the vicinity of hydropho-
bic compound, water molecules will be constrained forming hydrogen bonds only in the direction of
bulk water and therefore their motional freedom will be restricted and entropy of motion decreased.
The only way of avoiding this unfavorable interaction for water molecules is to decrease contact
with non-polar compound. It is accomplished by aggregating hydrophobic molecules together and
thus reducing their overall interface with the solvent. The aggregation is accomplished by brownian
motion of the solute and solvent – once hydrophobic molecules are aggregated, it is unfavorable
for them to be dissolved again. This theory was formulated by Frank & Evans and named the
”Iceberg theory” after the hypothetical layer of constrained ice-like water molecules surrounding
the hydrophobic molecule (Frank and Evans 1945). However, there was and still is a discussion
going whether the surface area of the hydrophobic solute is the best measure of hydrophobic effect.
The theory of protein folding due to the need of concealing hydrophobic residues from the aque-
ous medium formulated by Kauzmann was based on the idea of reduction of hydrophobic surface
(Kauzmann 1959) as well as the research done by Tanford (Tanford 1980; 1979; 1978). But there
are several alternatives like the molar volume of the hydrophobic compound (Hildebrand 1979) or
the number of hydrogen bonds in the first layer of hydration (Madan and Sharp 1997).
9
Povilas Norvaišas Literature overview
2.1.3 ”Non-classical” hydrophobic effect?
Even though positive entropy change was and still is considered the main driving force
behind hydrophobic effect, several research groups have presented contradicting results. As early
as 1969 Jencks separated two classes of hydrophobic effect: the ”classical” one, driven by the
positive entropy change (∆S > 0) and the ”non-classical” – driven by negative enthalpy change
(∆H < 0). Some experiments done in 60s and 70s gave results in favour of the notion of ”non-
classical” hydrophobic effect (Mizutani 2011; Fernández-Vidal et al. 2010). However, in many cases
the reason behind the negative enthalpy change was not clear and therefore it was attributed to
different processes than hydrophobic effect, as it was done in the research which can be called
a direct predecessor to our study (Papenmeier and Campagnoli 1969). Now it is clear that the
whole notion of separating hydrophobic effect into two different processes was misleading. Most
of the initial research concerning thermodynamics of hydrophobic effect done by Tanford (Nozaki
and Tanford 1971; Tanford 1979) concentrated on small hydrophobic compounds like glycine pep-
tides and separate amino acids. The results were obtained in so-called ”fractioning” experiments,
when solubility of hydrophobic compound was measured in the ethanol-water mixtures at various
concentrations of ethanol. The solubility in pure water, which could not be determined directly,
was calculated by making an extrapolation from the collected data. Indeed, dissolution reactions
of these compounds were driven by positive entropy change, however the limitations of methodol-
ogy prevented from testing much larger and more hydrophobic compounds, not mentioning that
solubility and in result the free energy of dissolution were obtained by extrapolation and not by
a direct measurement. At that time if was determined that the thermodynamics of dissolution
depend linearly on the number of aliphatic atoms (CH2 group) of the compounds (Mizutani 2011),
but once again it was tested only for small molecules. Only much later research done by Matulis
has shown that the situation for long chained (12-14 CH2 groups) largely hydrophobic aliphatic
amines is quite different (Matulis and Bloomfield 2001a;b). The dissolution and aggregation of
such compounds is driven by large negative enthalpy change (∆H << 0) and therefore should
be regarded as an example of ”non-classical” hydrophobic effect. However, further studies and
dissection of the thermodynamic parameters for the dissolution of alkanes, alcohols and amines
(Matulis 2001) stated that the Gibbs free energy of precipitation retains the same proportional-
ity per one methylene group: ∆∆G = −3.58 kJ/mol. Therefore, it is the same process, which
exhibits different contributions from enthalpy and entropy for compounds with different levels of
hydrophobicity. Besides, thermodynamics of aggregation depend significantly on the temperature
and concentration of the compounds. If the hydrophobic molecules aggregate into a liquid phase,
their dissolution reaction is primarily driven by the positive entropy change, as the gain in motional
freedom of water molecules is high compared to the restrain of liquid aggregate. However, when the
aggregate forms a solid state, molecules experience much more constrain and their loss of motional
freedom outweighs the positive entropy gain of water – process is dominated by enthalpy driven
10
Povilas Norvaišas Literature overview
interactions in a tightly packed aggregate (Matulis 2001; Norvaisas et al. 2012). Even though
the contribution of enthalpy and entropy might differ significantly, the overall value of ∆∆G per
one CH2 group is the same. This phenomenon is called ”Enthalpy-entropy compensation” and is
observed in various systems (Cornish-Bowden 2002; Chen et al. 1998; Tanford 1997). However,
there is a discussion happening for some time, whether it is a true fact of a physical world or just
an artefact of experimental limitations. Main thermodynamic parameters – enthalpy, entropy and
Gibbs free energy are related to another by equation 2. The problem is, that only enthalpy and
Gibbs free energy (through the binding constant Kb (Eq.1) or any other measure of difference in
chemical potentials) can be measured experimentally.
∆G = −RTln(Kb) (1)
No method for direct evaluation of binding entropy has been found yet. Therefore, ∆S is always
obtained by using the equation 2 and the ”compensation” is unavoidable, as one parameter is
dependant on the values of the other two (Cornish-Bowden 2002). Even though the relationship
stated in Eq.2 might be perfectly right, the values of ∆S have an error of both enthalpy and
entropy and should not be treated as certain.
∆G = ∆H − T∆S (2)
2.1.4 The fingerprint of the hydrophobic effect
In respect to the dither concerning the nature of hydrophobic effect and its true fin-
gerprint, one another property was considered – the change in the heat capacity of hydrophobic
reactions ∆Cp. It was discovered that this change is highly negative for the reactions of dissolution
and conversely – positive for the hydration (Matulis 2001; Matulis and Bloomfield 2001b; Baldwin
1986; Sturtevant 1977). The main reason behind it is the same ”iceberg” of highly structured
water molecules surrounding hydrophobic compound in the solvent. Upon hydration it increases
the heat capacity of the solution of hydrophobic compounds because additional energy will be
spent in breaking the constrained hydrogen bonding as if the ice would melt. And vice versa for
dissolution – decrease in the number of structured water molecules leads to the decrease of heat
capacity. This change is now considered the main fingerprint of hydrophobic effect as other pa-
rameters might change significantly and sometimes in rather unpredictable fashion (Shimizu and
Chan 2000).
2.2 Surfactants
Alkyl sulfates, alkane sulfonates and alkyl amines (Tab.1) are so called detergents – de-
terge agents. Together with most lipids and molecules of a comparable structure they belong to
11
Povilas Norvaišas Literature overview
Surfactants General formula
Alkyl sulfates CnH2n+2OSO–3
Alkane sulfonates CnH2n+2SO–3
Alkyl amines CnH2n+2NH+3
Table 1: General formulas of surfactants used in this study.
a much bigger class of compounds called surfactants – surface active agents. The main structural
elements of surfactants are the polar head-group which is hydrophilic and non-polar tail with hy-
drophobic properties. Such bivalent structure is responsible for the properties of these amphiphilic
compounds. Hydrophilic head makes detergents soluble in polar solvents, whereas hydrophobic
tail is responsible for their tendency to form a plentitude of complex colloidal phases, prononced
lyotrophic & thermotrophic properties, and solubility in non-polar solvents which depends on the
ratio of the size of hydrophobic and hydrophilic parts (Nagarajan 1991). Surfactants have been
given such name because of their non-linear influence on the physical properties of the solution.
Upon increase in the concentration of surfactants, surface tension and the conductivity of a so-
lution changes in an abrupt steps (Aniansson et al. 1976; Somasundaran et al. 1964). In order
to explain such anomalous behaviour, micelle theory was created (Hoerr et al. 1943), which was
largely based on the ideas of Irvin Langmuir and then recently established theory of the hydropho-
bic effect (Tanford 1997). According to the ”classical” theory once certain concentration of the
surfactant is reached it becomes more favourable to form micelles thus concealing hydrophobic
tails than orienting these tails out of the aqueous medium. In such a process, surfactants lose their
motional freedom, but it is recovered by an increase in the entropy of water upon reduction of
hydrophobic interface. The balance between two opposing forces is reached when the concentration
of the surfactant equals cmc – critical micelle concentration. Analogous parameter is the cmt –
critical micelle or Krafft (TK) temperature. If the temperature is higher than the cmt – micelles
will not form, and the surfactants will remain as fully dissolved monomers. The formation of mi-
celles is the first step towards various phase changes that the surfactants can undergo. However,
micellar phase itself can not be considered a ”true phase” on its own. For a phase transition like
melting of ice or evaporation change in the Gibbs free energy equals zero (∆G = 0). Even though
micelles have properties of a separate phase, it is not really correct to consider them so, because
micellisation is usually accompanied by non-zero change in the Gibbs free energy. Therefore, they
should be treated as a pseudophase (Stellner et al. 1988). In solution, formation of the micelles
is followed by disassemble and the equilibrium is expressed through the half-lifes of two states
(Kume et al. 2007). Alternatively to the micellisation and other colloidal states, where surfactans
are still at least partly dissolved in the solution, they may precipitate out of the solution upon loss
of the charge in the polar group. This can be caused by change in pH (Matulis and Bloomfield
2001a), neutralisation with acid/base (Matulis and Bloomfield 2001b) or any other substance of
12
Povilas Norvaišas Literature overview
opposite charge. Purely aliphatic compounds with similar hydrophobic tail length like alkanes are
almost completely insoluble (McAuliffe 1969). Therefore, such precipitation (aggregation) is highly
favourable and is accompanied by huge negative change in the Gibbs free energy (Matulis 2001;
Matulis and Bloomfield 2001b).
2.2.1 Mixtures of cationic and anionic surfactants
In the beginning of the XXth century, it was already much known about the physi-
cal properties and thermodynamics of homogeneous surfactant solutions (Saito 1982; Tartar and
Wright 1939; Hoerr and Ralston 1942a; Hoerr et al. 1943; Hoerr and Ralston 1943; 1942b; Ralston
et al. 1941; 1942). Because of their resemblance to lipids, surfactants can be used as a simple
model system to investigate general properties of membrane formation and other biochemical phe-
nomena. However, for the most part research on the surfactants was initiated not as an approach
to model biologically relevant systems with simpler compounds, but because of industry’s inter-
est. As it was mentioned, upon aggregation with oppositively charged compounds, surfactants
precipitate out of aqueous medium and become inactive. This property is especially important in
making multicomponent cleaners, where both cationic and anionic surfactants are mixed (Holland
and Rubingh 1992). Therefore, it became necessary to investigate systems of mixed surfactants
and at least determine their cmcmix values. Detailed research on the system of sodium dodecyl
sulfate (SDS, C12H25OSO3Na) and decylpyridinium chloride (DPCl, C15H26C1N ·H2O) revealed
that such systems form several colloidal phases in the solution (Fig.1) and may precipitate in a
wide range of concentrations (Stellner et al. 1988). Here is the description of the main phases of
the system:
• The concentration of both surfactants is sufficiently low and they remain as fully dissolved
monomers (Fig.1: [SDS] ≈ [DPCL] → 0).
• Concentration ratio is highly imbalanced – homogeneous micelles of only one surfactant form
(Fig.1: [SDS]/[DPCL] » 1 or [SDS]/[DPCL] « 1, if cmc is reached).
• Concentration ratio is more balanced and close to one – mixed micelles form (Fig.1: [SDS]/[DPCL]
> 1 or [SDS]/[DPCL] < 1, if cmcmix is reached)
• Concentration ratio is close or equal to one – surfactants precipitate (Fig.1: [SDS]/[DPCL]
→ 1, when [SDS] and [DPCL] >> 0 (solid aggregate in Tab.2 and Matulis (2001); Norvaisas
et al. (2012))
• With certain ratio of concentrations and proper temperature – coacervate may form which
resembles liquid hydrophobic droplets in an aqueous medium.
13
Povilas Norvaišas Literature overview
Figure 1: Phase diagram for the mixture of SDS and DPCL. Taken from Stellner et al. (1988)
2.2.2 Interaction of cationic and anionic surfactants
In the 1969 for the first time it was tried to speculate how such compounds interact and
what structural components are responsible for the corresponding thermodynamics (Papenmeier
and Campagnoli 1969). Group led by Campagnoli investigated reactions in various concentrations
of SDS and cetylpyridinium chloride (CPC, C21H33CN ·H2O). The main goal of this research
was to determine thermodynamic parameters of the micelle formation and possible precipitation
of the surfactants. The results suggested that the interaction is highly favoured by enthalpy –
∆H = −75.3 kJ/mol . At that time it was established that the hydrophobic reactions should
be primarily driven by positive entropy change (∆S > 0). Sadly, the idea of enthalpy driven
hydrophobic effect hasn’t been formulated back then and therefore, it was not speculated what
process or structural units responsible for such thermodynamics (Papenmeier and Campagnoli
1969). The study also determined Gibbs free energy change for the system. However, due to
technological limitations, the values of ∆G relied heavily on the model used to describe kinetics of
the reaction. ∆G equaled−62.76 kJ/mol if hypothetical micelles were assumed to be a pseudophase
and −37.6 kJ/mol if the micellar phase was neglected. In equimolar mixtures of the detergents,
micelles were really ”hypothetical”, because the researchers were instead observing the formation
of precipitate. This effect was especially pronounced at higher concentrations of both detergents,
as the precipitate forms gradually with increased concentration. Due to this effect, enthalpy of
the reaction is dependent on the concentration of the detergents, until no micelles are formed at
all and equilibrium fully shifts towards the formation of precipitate (see Fig.1 here and Table II
in Papenmeier and Campagnoli (1969)). What Campagnioli and colleagues also didn’t knew, was
14
Povilas Norvaišas Literature overview
that the critical micelle concentration for the mixture of cationic and anionic surfactants (cmcmix)
is always significantly lower than the one of pure surfactants (Stellner et al. 1988). Therefore, in
the experiments they observed micelle formation at the lower concentration than initially expected
and precipitate formation where the formation of micelles was assumed. All in all, even with some
invalid assumptions, research presented some interesting results. However, no speculations have
been made on the nature of such thermodynamics of binding.
Further research on the interaction of cationic and anionic surfactants have revealed some
other interesting properties (Amante 1991). Study focused on the interaction between DPCl and
alkyl sulfates of various aliphatic chain length (n = 8 − 12). With the increasing aliphatic chain
length, boundary between monomer and precipitate shifts towards lower concentrations, primarily,
because of smaller cmc values. If only one of the detergents is changed, then the whole phase
diagram shifts in the direction of decreased cmc (Fig.5 in Amante (1991)). Also, it was observed
that the change in the Gibbs free energy upon addition of one methylene group to the aliphatic
chain (∆∆G) is larger for the formation of the precipitate (∆∆Gpre = −4.14 kJ/mol ) than that
of micelles (∆∆Gpre = −2.96 kJ/mol ). Therefore, surfactants in the mixture are more likelly
to be found in precipitate than in micelles. Values of ∆Hpre were determined with then widely
used but prone to error method of Van’t Hoff analysis (Tellinghuisen 2006; Liu and Sturtevant
1997b; Chaires 1997; Liu and Sturtevant 1997a), when enthalpy values are estimated from the ∆G
dependence on the temperature. It was also suggested that with sufficiently small n, aggregation
might not happen at all, however it was not extensively tested. Only much later it was shown
that the short chained surfactants indeed can not aggregate or require concentration above the
experimental capabilities (Matulis and Bloomfield 2001b).
Influence of electrolytes on the hydrophobic effect has been acknowledged for some time
(Ben-Naim and Yaacobi 1974) and it was suggested that the salts act according to the Hoffmeister
series (Fernandes et al. 2010). However it is not clear whether salt changes structure of the water
(Nucci and Vanderkooi 2008) or directly interfere with the interaction between molecules (Zhang
and Cremer 2006). In the case of mixed surfactant solutions it has been observed that increase
in the salt concentration interferes with the hydrophobic effect and increases cmc values for the
surfactants. Therefore, precipitation would happen at higher concentration (Kume et al. 2007;
Amante 1991).
Even though research was done on many popular surfactant systems, it mostly concen-
trated on determining characteristics of the specific compounds and not the fundamental processes
behind. There is no unified theory explaining the thermodynamics of protein denaturation with
surfactants (Otzen 2011) or membrane partitioning – processes, that both exhibit ”non-classical”
hydrophobic effect. Some questions have been partly answered (Matulis and Bloomfield 2001a;b;
Matulis 2001), but the research needs to be extended to even more hydrophobic, partly polar
15
Povilas Norvaišas Literature overview
compounds which resemble biologically relevant molecules.
2.3 Isothermal Titration Calorimetry (ITC)
The name of this methodology itself explains main principle behind - measurement of a
heat flow created by reaction, while titrating at constant temperature. First commercially avail-
able isothermal titration calorimeters appeared in 1988 (Olofsson and Loh 2009; Wiseman et al.
1989). Quickly, it became one of the most widely used laboratory methodologies in various sorts
of research. Calorimetry has been in various forms for quite some time, like the differential scan-
ning calorimetry (DSC) in the research done by Papenmeier (Papenmeier and Campagnoli 1969).
However, only during the last decade of XXth century it became very popular and it’s not surpris-
ing why. ITC doesn’t require any additional preparation of the reagents, experiments are short
and easy to perform (Jelesarov and Bosshard 1999). It enables a direct measurement of reaction
enthalpy and binding constant Kb (by fitting the data according to the mathematical model of
reaction kinetics), thus instantly providing 2 out of 3 parameters of the reaction thermodynamics.
Newest ITC calorimeters are extremely sensitive and can measure binding energies of the com-
pounds in nanomolar concentration. The main competitors of the ITC have both advantages and
drawback of their own. Surface plasmon resonance (SPR) technique enables more detailed insight
into the binding constant, as it can measure both the constant of association Kon and dissociation
Koff in the same experiment. The main disadvantage of this method is the requirement to immo-
bilise one of the reagents onto the experimental surface. Thermal shift assay (TSA) in its regard
has much greater throughput, because of the use of the 96 well plates, which enables measurement
of the affinity for various ligands or same ligands with different pH, buffers or proteins during the
same experiment. However, it cannot measure the ∆H in a direct way, as ITC does(Matulis 2008).
ITC is widely employed in the research on protein-ligand interactions in the structure-
based drug design and measurements of various other kinds of interactions like protein-protein;
protein and metal, nucleic acid, polysacharids, polymers; nucleic acids and small molecule, deter-
mination of enzyme kinetics and many other (Velázquez Campoy and Freire 2005; Heerklotz and
Seelig 2000; Otzen 2011; Falconer et al. 2010).
2.3.1 Adiabatic and heat flux ITC
There are two types of ITC calorimeters currently available: heat flux and adiabatic ones.
The reaction cell of heat flux calorimeters is surrounded by a heat-sink. Between the heat sink
and the cell there is a layer of thermopile, which registers heat flux in or out of the cell. The main
advantage of such calorimeters is wide scale of sensitivity – from 10 µW to 10 mW. However, one
experiment with the heat flux calorimeter can last up to several hours, because thermal equilibrium
16
Povilas Norvaišas Literature overview
is reached slowly after the injection, even with high thermal conductivity. These calorimeters are
capable of measuring intensive heat flow from huge samples (Olofsson and Loh 2009).
Adiabatic calorimeters are made for a completely different purpose. They are fast and
sensitive enough to work with nanomolar amounts of substances. These calorimeters are in fact very
precise thermometers, measuring the temperature difference between reaction and experimental
cell. Cooling system surrounding the cells acts as a heat sink and the whole apparatus is covered
with adiabatic shielding. During the reaction, heat released or absorbed by the experimental cell
is measured by comparing the power needed to keep both experimental and reference cells at the
constant temperature. Because of such system, adiabatic calorimeters can measure heat flow of less
than 1 µcal/s. The most popular calorimeters are made by USA established company MicroCal®,
they have a reaction vessel of 1 mL (MCS ITC®, VP-ITC®), but the more sensitive ones have a
cell of 250 µL (ITC200®). Adiabatic calorimeter has been also used during this study, therefore the
planing of the experiment will concern this type of ITC.
2.3.2 Planning of the ITC experiment
One of the most important parameters which must be decided or at least guessed before
the actual experiment is the concentration of the reagents used (Doyle 1997). If the concentration
is too high, excess heat from the cell won’t be registered, if too low – heat flux will not be sufficient
to accurately determine reaction enthalpy. Same applies to the binding constant Kb. High con-
centration will increase the steepness of the binding curve, so that the titration is finished in one
injection. Low concentration, however, will cause the gradual ending of the reaction, which might
also be too flat for fitting model (Doyle 1997; Olofsson and Loh 2009; Jelesarov and Bosshard
1999). As it can be seen, planning requires finding a golden ratio between accuracy and techno-
logical limitations. Such golden ratio is called ”c” coefficient, which determines relation between
concentration of the substance in the calorimeter cell and the binding constant Kb:
c = [cell]×Kb. (3)
The range of ”good” c values is not consistent among different researchers. Most popular range
is 1 − 1000 (Wiseman et al. 1989), but the other authors suggest that the limitations are more
stringent and c ∈ 10 − 100 (Jelesarov and Bosshard 1999). However, it was proven that with
c < 0.1 it is still possible to determine binding constant Kb with sufficient level of confidence and
it becomes less dependent on the error in stoichiometry (Tellinghuisen 2008).
One of the technological experts of the ITC methodology, Joel Tellinghuisen have deter-
mined rules for the most effective ITC experiment (Tellinghuisen 2006):
1. titration should be carried out in 10 injections
17
Povilas Norvaišas Literature overview
2. concentration of the substances in the syringe and the cell (c) selected in such a way, that
after the last injection ratio of the titrant [X]0 and titrand [M ]0 concentrations in the cell is
determined by the relationship in Eq.4:
[X]0[M ]0
≡ Rm = 6.4c0.2 + 13
c, with Rm > 1.1 (4)
3. titration must be started with highest possible concentration, but c < 1000
When all of the mentioned rules are satisfied, the error in ∆H and Kb, which depends on
the technological limitations of the method is considered to be 1% (Tellinghuisen 2006). Even then
there is a possibility that the calorimeter has a systematic error. To check it, every calorimeter has
built in heat pulse generator which enables calibration of the instrument. However, even then there
might be some inconsistencies, which can only be ruled out by testing calorimeter with standard
reactions of high and accurately determined enthalpy. The examples of such reactions could be
the NaCl dilution into water (Tellinghuisen 2007a), neutralisation reaction HNO3 + TrisBase or
interaction of two salts NaI + AgNO3 (Baranauskiene et al. 2009).
2.4 Additivity – the 4th law of thermodynamics
The additivity of contribution made by chemical groups in terms of their physical, chem-
ical properties and thermodynamics is one of the most widely used principles in biochemistry and
other physical sciences. It is indeed so popular, that might be called the unofficial, 4th law of
thermodynamics (Dill 1997). Almost everything has been tried to be summed up: contribution to
the overall Gibbs free energy, entropy and enthalpy of the reaction, solubility, hydrophobic surface
area and the energy of chemical bonds. However, in many cases the assumptions are too stringent
to be true. Parameters of chemical groups should be independent of one another and each group
should experience more or less similar environment (Dill 1997). Besides, usually the additivity
models have a significant level of redundancy – for example Gibbs free energy can be calculated
in terms of the number of polar and non polar groups, types and energetics of the interactions
formed, primary and secondary structural components, etc. The only systems, which seem to obey
this 4th law are polymers, like alkanes, which are made of similar chemical groups (Dill 1997).
Good linear dependence of the parameters on the structure of such compounds is most likely to
be a consequence of similar bonding and environment that each atom experiences. The properties
of heterooligomeric structures with only a small number of differing monomers can also be calcu-
lated with a degree of certainty (Plyasunov et al. 2004; 2006; Plyasunov 2000a; Plyasunova et al.
2005; Plyasunov et al. 2000; Plyasunov 2000b; Plyasunov and Shock 2001; Matulis and Bloomfield
2001a;b; Matulis 2001). A good example of additivity concerning the hydrophobic effect are the
thermodynamics of aggregation for various classes of compounds (Tab.2).
18
Povilas Norvaišas Literature overview
∆Gagg, kJ/mol ∆Hagg, kJ/mol T∆Sagg, kJ/molA
ggre
gatio
n
Into
liqui
d Alkanes (m = 5− 17) −3.58m− 0.69 −1.25m + 9.25 2.33m + 9.94
Alcohols (m = 1− 12) −3.58m + 14.33 −1.25m + 12.70 2.33m− 1.63
Alkylamines (m = 3− 11) −3.58m + 19.30 −1.25m + 27.70 2.33m + 8.40
Into
solid Alkanes (m = 18− 20) −3.58m− 0.69 −5.20m + 18.41 −1.62m + 3.76
Alcohols (m = 13− 20) −3.58m + 14.33 −5.67m + 25.08 −2.06m− 5.09
Alkylamines (m = 12− 20) −3.58m + 19.30 −5.25m + 45.71 −1.67m + 11.00
Table 2: Dependency of the thermodynamic parameters of aggregation on aliphatic chain length
n for alkanes, alcohols and alkylamines at 25℃. Taken from Matulis (2001).
The attempts to calculate analogous dependencies for a much more complex biological
systems like protein-ligand complexes are complicated. In heterogeneous environment contribution
of various chemical groups is governed not only by additivity, but also by cooperativity – groups
influence one another. Polar groups could have a significant contribution to the change in the
entropy of the system and hydrophobic interactions might constitute a large part of the enthalpy
potential (Baum et al. 2010). Therefore, it becomes hard to dissect contributions of individual
groups, especially when it is only possible to observe the overall result of the pleiad of interac-
tions. Three main thermodynamic parameters (∆G, ∆H and T∆S) represent the final result of
a reaction. With such a level of abstractiveness a significant part of the information concerning
the system is lost and can only be recovered by the supplementary methods (X-ray crystallogra-
phy, FRET microscopy, etc.). The extent of ”non-additivity” becomes apparent in analyzing large
databases of protein-ligand interaction thermodynamics (Ladbury 2010; SCORPIO 2012). When
the thermodynamics are resolved according to structural units of the ligands, no correlation can
be found between, for example, hydrophobic surface area of the ligand and the entropy change of
the reaction (according the material of lecture by John E Ladbury). The conclusion regarding the
4th law of thermodynamics is clear – it is viable only in very simple and homogeneous systems.
These are the ones that can really help in understanding some basic processes behind more complex
phenomenon.
2.5 Carbonic anhydrases
Carbonic anhydrases (CAs) are ancient enzymes present in virtually every tissues, cell
types, subcellular organelles, and in organisms ranging from unicellular cyanobacteria to mammals
(Imtaiyaz Hassan et al. 2012). The reason behind such ubiquity is their role in the process of
respiration. CAs catalyse reaction of carbon dioxide (CO2) conversion to bicarbonate (HCO–3) and
19
Povilas Norvaišas Literature overview
vice versa
CO2 + H2O←−→ HCO−3 + H+. (5)
Bicarbonate is insoluble in lipid membranes and needs to be transported, while carbon dioxide is
more soluble, therefore it diffuses freely in and out of the cell (Imtaiyaz Hassan et al. 2012).
There are five evolutionarily unrelated gene families of CAs: the α-, β -, γ -, δ -, and
ζ -CAs. The α, β and δ-CAs contain a Zn2+ ion at the active site, the γ-CAs are probably
Fe2+ enzymes (but they are active also with bound Zn2+ or Fe2+ ions), while the metal ion is
usually replaced by Cd2+ in the ζ -CAs (Alterio et al. 2012). These enzymes are found in all
the kingdoms of life and are involved in respiration, photosynthesis in eukaryotes and cyanate
degradation in prokaryotes. In plant cells, they are related with the photosynthetic fixation of
CO2 in the presence of chloroplasts (Imtaiyaz Hassan et al. 2012). Behind such wide spectrum of
functions and occurrence, however, the main mechanism of activity is the same. Three histidine
residues (His94, His96 and His119) which coordinate the metal ion are conserved thorough all
families. In the case of a zinc-bound enzymes, catalitic reaction can be divided into two steps
(Alterio et al. 2012). In a direction of hydration, Zn2+ bound hydroxide makes a nucleophilic
attack on CO2, with consequent formation of bicarbonate HCO–3 and release of the proton H+.
The second step is rate limiting and involves regeneration of the catalytically active hydroxide.
During this step a protein is transferred from a water molecule bound to the Zn2+ to the bulk
solvent. It is accomplished by a network of structured water molecules which reside in the binding
pocket (Imtaiyaz Hassan et al. 2012; Alterio et al. 2012).
All human carbonic anhydrases belong to the α family. Up to now there are 15 iso-
forms identified, which differ significantly by molecular features, oligomeric arrangement, cellular
localization, distribution in organs and tissues, expression levels, kinetic properties and response
to diffierent classes of inhibitors. Twelve isoforms (CAs I−IV, VA −VB, VI −VII, IX, and XII−
XIV) show a variable degree of enzymatic activity, whereas three isoforms (VIII, X, and XI), the
so-called CA-related proteins (CARP’s) are devoid of catalitic activity (Alterio et al. 2012). It is
predicted that CAIV and CAVI has been the oldest mammalian isozymes, while CAI, CAII and
CAIII deviated in recent times (Imtaiyaz Hassan et al. 2012). Isoforms of CAs are involved in di-
verse physiological functions including pH regulation, ion transport, bone resorption and secretion
of gastric, cerebrospinal fluid and pancreatic juices (Alterio et al. 2012). In result of such wide range
of functions CAs differ for subcellular localizaton: CA I, II, III, VII and XIII exist in cytosol, CA
IV, IX, XII, and XIV are membrane associated, CA VA and VB reside in mitochondria, whereas
CA VI is secreted in saliva and milk (Alterio et al. 2012). CAs can be regarded as the central
enzymes for both transport and metabolic processes at the cellular level (Imtaiyaz Hassan et al.
2012). For example in the metabolically active tissue of a muscle, CA facilitates the CO2 transport
and in other cases it works as a key protective enzyme in elevated CO2 concentration (Imtaiyaz
Hassan et al. 2012; Alterio et al. 2012). Membrane associated CA is important in acidifying the
20
Povilas Norvaišas Literature overview
outer boundary layer through the catalyzed hydration of excreted CO2.
2.5.1 Structural properties
X-ray crystal structures are already available for the majority of the twelve catalytically
active members of the human CA family. All isoforms are highly homologous by their sequence
similarity and even with differing sequences preserve more or less similar structure (Imtaiyaz Hassan
et al. 2012). The active site of the enzyme is located in a large, conical cavity, approximately 12
Å wide and 13 Å deep, which spans from the protein surface to the center of the molecule. The
catalytic zinc ion is located at the bottom of this cavity, exhibiting a tetrahedral coordination
with three conserved His residues and a water molecule/hydroxide ion as ligands. On the basis
of hydrophobic and hydrophilic nature, the active site cavity is characteristically divided into two
halves. Hydrophobic part contains Ala121 and 135, Val 207, Phe91, Leu131, 138, 146, 109, and
Pro201, 202 while the hydrophilic part of the cavity consists of His64, 67, 200, Asn69, Gln92,
Thr199, Tyr7 and Val62 (Imtaiyaz Hassan et al. 2012). Such bipartite environment is required
for the carbon dioxide to bicarbonate conversion reaction catalysed by CAs. Hydrophobic part
is responsible for CO2 capture and orientation in the right direction, whereas hydrophilic side
arranges water molecules for the proton transfer from zinc-bound water molecule to the bulk
solvent. During the first step of the mentioned hydration, tetrahedrally coordinated hydroxide is
involved in a network of hydrogen bonds which helps to enhance its nucleophilicity. In particular,
it is hydrogen bonded with the hydroxyl moiety of a conserved Thr residue (Thr199) and with
two water molecules, located on two opposite sides: the first one, also called the “deep water”, is
located in a hydrophobic cavity (Alterio et al. 2012). The molecular dynamics studies, however,
suggest that up to six water molecules might be involved in the whole process (Tripp et al. 2001).
2.5.2 Drug design for CAs
Besides the the variety of roles, CAs assume at a normal tissue, they have also been
reported to be involved in various disease mechanisms. Therefore, extensive drug design studies
have been taken to find the potential inhibitors and stimulators of CAs (Alterio et al. 2012). In
the cases where memory and learning is impaired, like Alzheimer’s and other neurodegenerative
diseases – activators are sought for (Alterio et al. 2012). However, majority of potential CA
targeting drugs act as the inhibitors, like diuretics, antiglaucoma agents, antiepileptics, and in the
management of altitude sickness. The more novel generation compounds are undergoing clinical
investigation as antiobesity, and antitumor drugs/diagnostic tools and even as anti-malaria drugs
(Alterio et al. 2012). In the case of tumours, CAs have quite a special role as enzymes at least partly
responsible for the survival of cancerous cells in an hypoxic environment and the acidification of
the extracellular space (Pastorekova et al. 2008). There is now a solid evidence that CAs actively
21
Povilas Norvaišas Literature overview
contribute to adaptive responses of tumour cells to physiological stresses (Pastorekova et al. 2008).
Indeed, cancerous cells are remodelling expression of CA genes to gain new phenotypic properties
to become capable of rebuilding their metabolism and increase migration and invasion. This
adaptation to physiological stresses in microenvironment is an essential aspect of tumor progression
that significantly contributes to metastasis and resistance to anticancer therapy (Pastorekova et al.
2008).
Even though there is a handful of crystallised hCA structures, most of the reported
complexes with inhibitors regards just isozyme II, the most thoroughly characterized CA isoform
(Imtaiyaz Hassan et al. 2012; Alterio et al. 2012). Cytosolic CAII shows the highest distribution,
it covers almost every tissue and organ, including osteoclasts in bone, choroid plexus epithelia,
retinal muller cells, hepatocytes, kidney, oligodendrocytes in brain, salivary glands, erythrocytes
and platelets (Imtaiyaz Hassan et al. 2012). The major role of CAII is in the contribution of H+
production and acid–base homeostasis, pH balance, metabolic acidosis. It also triggers the CO2
exchange in the erythrocytes and lungs (Imtaiyaz Hassan et al. 2012). As a side-effect of such
properties, hCAII also participates in the mechanisms of glaucoma, edema, epilepsy and altitude
sickness (Alterio et al. 2012).
2.5.3 Inhibitors of CAs
The inhibitors of CAs can be classified into two major categories: those that are bound
to the active site but do not interact directly with the metal ion, and those that bind to the enzyme
active site by anchoring themselves to the catalytic zinc ion. The first class includes phenols (some
of which are natural compounds), polyamines, coumarins, and antiepileptic drug lacosamide. The
second class is more established and includes the ureates/hydroxamates, the mercaptophenols, the
metal-complexing anions (cyanide, azide, hydrogen sulfide and trithiocarbonate) and the sulfon-
amides and their bioisosteres, such as sulfamates and sulfamides. The majority of the clinically
used inhibitors of CAs are sulfonamides – R−SO2−NH2. The inhibitory properties of these com-
pounds are based on the coordination of the deprotonated sulfonamide nitrogen to the catalytic
Zn2+ ion, with consequent substitution of the zinc-bound water molecule. Thus, amine of the
sulfonamide group occupies the place of the hydroxide ion and prohibits nucleophilic attack of the
CO2.
Even though there is a handful of CA related drugs which already are in the clinical
trials (Alterio et al. 2012), none of the currently clinically used CA inhibitors shows selectivity for
a specific isozyme. This is especially important, because, for e.g., CA IX and XII in tumors should
be inhibited by compounds, which do not affect the activity of CA I, II, VA, and VB (Alterio et al.
2012). As CAs are ubiquitous and responsible for many vital functions, collective inhibition might
lead to the undesired side effects and even more extensive damage than that done by the initial
22
Povilas Norvaišas Literature overview
disease (Alterio et al. 2012).
2.6 Computational structure-based drug design
Structure-based drug design and computational drug design are interconnected method-
ologies in the process of new drug discovery or optimization (Kitchen et al. 2004). The basic idea
behind structure-based drug design is, that molecular structures of the target and the drug give
sufficient information to determine and predict the energetics of their interaction. This approach
is largely influenced by the biochemical laws of additivity and therefore it assumes, that the overall
energetics of the target-drug complex formation can be approximated by the sum of contribution
of the individual chemical groups (Dill 1997). Computational methods in their regard, provide
tools to investigate this affinity in silico by the use of the structural data. Nowadays, these meth-
ods have become a crucial component of many drug discovery programmes from hit identification
to lead optimization and beyond. When only the structure of a target and its active or binding
site is available, high-throughput docking is primarily used as a hit-identification tool. However,
similar calculations are often also used later on during lead optimization, when modifications to
known active structures can quickly be tested in computer models before the actual synthesis of
the compound. This, at least partly, is an effect of the improvements in techniques for structure
determination, such as high-throughput X-ray crystallography. The increase in the number of crys-
tallised protein structures allows the identification of new target and testing of prospective drugs
in silico. There are now a number of drugs, like inhibitors for HIV protease, whose development
was heavily influenced and, in some cases, based on the screening strategies or other computational
techniques (Kitchen et al. 2004).
2.7 Docking
One of the key methodologies in the computational structure-based drug design — dock-
ing of small molecules to protein binding sites was pioneered during the early 1980s (Kitchen et al.
2004). The term ”docking” is used for the computational modelling techniques, which ”dock” the
ligand into the rigid or semi-rigid protein and ”score” their potential complementarity to the bind-
ing site. The first step – prediction of ligand conformation and orientation is called posing. This in
itself is challenging, as even relatively simple organic molecules can contain many conformational
degrees of freedom. Sampling these degrees of freedom must be performed with sufficient accuracy
to identify the conformation that best matches the receptor structure, and must be fast enough to
permit the evaluation of thousands of compounds in relatively short time. These algorithms are
complemented by the scoring functions that are designed to predict the biological activity through
the evaluation of interactions between compounds and potential targets. In general, there are two
23
Povilas Norvaišas Literature overview
aims of docking studies: accurate structural modelling and correct prediction of activity (Kitchen
et al. 2004; Taylor et al. 2002).
Relatively simple scoring functions continue to be heavily used, at least during the initial
stages of drug discovery and docking simulations (Kitchen et al. 2004). From one side, they let the
very fast approximation of the binding energy, which is required for screening the huge number of
compounds. From the other side, they also include a number of assumptions which significantly
decrease their predictive value. A number of factors is not included, like inherent flexibility,
induced fit or other conformational changes that occur on binding, and the participation of water
molecules in protein–ligand interactions. Docking, by its nature can not evaluate the entropic term
of the interaction and bias the enthalpic one (Kitchen et al. 2004). These imperfections of scoring
functions continue to be a major limiting factor.
Because of all the mentioned weaknesses of the docking algorithms, re-selected conformers
are often further evaluated using more complex scoring schemes with more detailed treatment of
electrostatic and van der Waals interactions, and inclusion of at least some solvation or entropic
effects (Taylor et al. 2002).
2.7.1 Posing
Finding the right posing for the selected ligands is not an easy task. Most biological
molecules have a number of rotatable bonds and if we consider the rotational freedom in the ligand,
we deal with the problem called ”conformational explosion”. The number of possible conformations
Nconformations =N∏
i=1
ninc∏j=1
360θi,j
. (6)
increses exponentialy with the number of rotatable bonds N (ninc – number of incremental rota-
tional angles, θi,j – incremental rotational angle). Docking algorithms must somehow sample the
immense number of conformations and find the valid ones. Possible treatments of the problem can
be divided into three main categories (Kitchen et al. 2004):
• Systematic methods (incremental construction, conformational search, databases) – which
try to explore all the degrees of freedom and ultimately face the combinatorial explosion.
Ligands can be incrementally ”grown” into the binding site or be separated into rigid and
flexible parts.
• Stochastic methods (Monte Carlo, genetic algorithms, tabu search) – which operate by mak-
ing random changes to a conformation of a ligand or a number of ligands. It is done by
the ”random-walk” of a Markov chain or ”mutations” to the previous pose in the genetic
algorithm.
24
Povilas Norvaišas Literature overview
• Simulation methods (molecular dynamics, energy minimization) – which employ the molec-
ular dynamics simulation of the ligand.
2.7.2 Scoring
The generation of the right ligand conformation is not enough by itself, because it must
be identified and selected out of all other conformations. This is accomplished by the scoring
functions. Generally, there are three unique types of scoring functions available and one as a
complement.
• Force-field based scoring – use of molecular mechanics forcefields to quantify the binding
energy by the sum of two energies: protein-ligand interaction and internal energy (usually
only for the ligand). Interaction between the protein and the ligand is most often described
in terms of electrostatic Coulomb and steric Van de Waals potentials:
V col(r) =NA∑i=1
NB∑j=1
qiqj
4πϵrij(7)
where A and B are two atoms and rij – distance between them, and
V vdw(r) =NA∑i=1
NB∑j=1
4ϵ
[(σij
rij
)12
−(
σij
rij
)6]
(8)
where ϵ – well depth of the potential, σij – collision diameter for the respective atoms A
and B, and rij – distance between them. This method neglects the effect of the solvent or
approximates it in terms of the distance dependent dielectric (Kitchen et al. 2004).
• Empirical scoring – fit to reproduce experimental data, such as binding energies and/or con-
formations, as a sum of several parameterized functions. The design of empirical scoring
functions is based on the idea that binding energies can be approximated by a sum of indi-
vidual uncorrelated terms. Takes solvation into account, but depends heavily on the quality
of the empirical data (Kitchen et al. 2004).
• Knowledge based scoring – designed to reproduce experimental structures rather than binding
energies. May include solvent-accessibility corrections, but also depend on the number of high
quality protein-ligand complex structures (Kitchen et al. 2004).
• Consensus – combines information from different scores to balance errors in single scores and
improve the probability of identifying ”true” ligands (Kitchen et al. 2004).
2.7.3 Poisson-Boltzman surface area method
Poisson-Boltzman surface area method, or PBSA is one of the most popular methods
to calculate electrostatic potential in a salty solvent without explicitly modelling water molecules
25
Povilas Norvaišas Literature overview
(implicit solvation). Because of such capability it is also called the electrostatic continuum solvation
model (Kitchen et al. 2004). Due to a high computational cost it is mostly used as supplement to
the traditional methods (Kitchen et al. 2004) and applied after the initial pose detection by docking
or the molecular dynamics simulations (Massova 2000; Kollman et al. 2000; Fogolari et al. 2002).
PBSA acts as a scoring function and does not change the given pose of the ligand. It operates
by performing two separate tasks: calculation of the electrostatic potential by numerically solving
Poisson-Boltzman differential equation (Eq.9) and estimation of the change in the solvent accessible
surface area upon ligand binding to the protein (Fogolari et al. 2003).
∇⃗[ϵ(r⃗)∇⃗Ψ(r⃗)
]= −ρf (r⃗)−
∑i
c∞i ziqλ(r⃗) exp
[−ziqΨ(r⃗)
kBT
](9)
26
3 Thermodynamics of Cationic and Anionic Surfactant In-
teraction
3.1 Materials & Methods
3.1.1 Materials
Linear surfactants with differently charged polar head-groups and various aliphatic chain
length were used in this study to dissect the thermodynamic parameters of their aggregation.
Alkylsulfates and alkylsulfonates were chosen as negatively charged surfactants and alkylamines –
as the positively charged ones. The main reason behind such choice was the simple, non-branching
structure of these compounds, which is necessary in order to apply the laws of additivity (Dill 1997).
Moreover, all of the selected compounds have appropriate physicochemical properties which are
relevant to this study. First of all, given surfactants did not form micelles at any experimental
concentration (concentration of the surfactant in the cell and syringe was always lower than the
cmc) (Lide et al. 2009). At the lowest experimental temperature of 25℃ all surfactants were in a
solid crystalline form and only at greater temperatures they underwent a phase change (Lide et al.
2009). However, there also were several parameters, which were specific to a particular mixture of
oppositely charged surfactants and therefore could not be evaluated in advance. These were cmcmix
and Krafft temperature TK (Amante 1991; Stellner et al. 1988). It was not known whether all
systems of surfactants will form aggregates at the given conditions, but this was easily determined
from the thermodynamics of the reactions. In the case of no aggregation happening, ∆Hexp was
very small and represented only the ionic pairing of the headgroups (concentration of surfactants
was smaller than cmcmix and experimental temperature greater than TK). Theoretically, it should
be possible to observe aggregation reactions in all systems of surfactants with greater concentrations
and lower temperatures, but this was not extensively tested in this study.
The surfactants were bought from several sources. Acros Organics (New Jersey, USA:
1-800-ACROS-01; Geel, Belgium: +32 14 57 52 11) provided undecylamine (98%), dodecylamine
hidrochloride (99%) and tridecylamine (98%). Sigma Chemical Co. (P.O. Box 14508, St. Louis,
MO 63178 USA 314-771-5750): 1-decanesulfonic acid, sodium salt (98%), dodecylsulfate, sodium
salt (99%). Aldrich Chem. Co. (P.O. Box 355, Milw. WI 53201, 414-273-3850): 1-dodecanesulfonic
acid, sodium salt (99%). Pfaltz Bauer Inc. (375 Fairfield Ave., Stamford, Conn. 06902): 1-
nonanesulfonic acid, sodium salt (98%). Nonylamine (≥ 97%) was obtained from Fluka, Sigma-
Aldrich (CH-9471 Buchs, 081/75525 11). At room temperature surfactants were in a form of white
crystalline powder and only short-chained compounds as nonylamine appeared as a near-molten
wax. All surfactants were kept in the original packaging and used without further purification.
27
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
Aqueous solutions were prepared using distilled milli-Q grade water, which was also boiled to
remove dissolved CO2. Solutions of alkylamines were kept in tightly sealed containers to minimize
chances of reaction with atmospheric CO2, which might lead to a decreased concentration of an
active surfactant.
3.1.2 Isothermal Titration Calorimetry
All experiments were performed with Microcal (Northampton, MA) Micro Calorime-
try System (MCS) calorimeter in temperature range 25 −− 65℃. Circulating cooling bath was
attached to calorimeter in order to perform experiments bellow or near room temperature (Olof-
sson and Loh 2009). The temperature of the refrigerant was constantly kept at 17℃. The ITC
unit was calibrated using its built-in electronic heat pulse generator and validated by reaction
NaI+AgNO3 −−→ AgI ↓ . The enthalpy change determined in five experiments ∆H = (−109.4±1.8)
kJ/mol was very close to that found in the literature: ∆H = −110.9 kJ/mol (Baranauskiene et al.
2009), thus confirming the accuracy of the device. Prior to every experiment the cell was washed
multiple times with water and once in every month treated with 20% ”Contrad 70” cleaning agent,
followed by rinsing with 0.1 M CH3COOH and 0.1 M NaOH at the temperature of 65℃. Similarly,
injection syringes were cleaned by pumping 100 mL of water through their volume and then leaving
them to passively dry out. The cell was always prerinsed with a portion of the same surfactant
solution before the actual filling to minimize dilution of reagents due to small water droplet which
might have not dried out. Reactant solutions were cooled at least 5℃ below experimental tem-
perature just before filling to assure quick equilibration of the calorimeter. The cell (1,3438 mL)
and the syringe were loaded with solutions of oppositely charged surfactants – 0.33 mM and 5
mM respectively. Titration was performed in 40 injections of 6,25 µL at intervals of 180 s, with a
250 µL injection syringe. During the period of equilibration syringe was set to spin in short bursts
at the maximum speed of 1200 rpm to buoy any microscopic air bubbles attached to the cell walls.
Afterwards, the syringe rotation was set to a constant speed of 400 rpm. At least 180 s of data was
collected prior to the first injection to check the stability of the baseline. The ”Refference offset”
parameter of calorimeter was changed according to every system of surfactants to keep baseline
level well above the 0 µJ/s as in the figure 2a. This setting assured that the calorimeter registered
all the heat flow in or out of the cell and it was not dissipated by the cooling system.
There were cases when the baseline did not return to the preinjection level after the
injection (as in Fig.2a). This effect became more pronounced for the systems of greater aliphatic
chain length. Yet the exact cause is not known, it can be speculated that after the injection
aggregate underwent slow structural rearrangements because of a change in the ratio of charges.
This can be understood by inspecting figure 1 in this text. As the titration happens, we are diluting
one surfactant and increasing concentration of the other one in the cell. Because of this, system
28
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
goes through different phases, which are determined by the charge and concentration ratio of the
surfactants (Stellner et al. 1988). These processes stopped when stoichiometry reached unity and
surfactants formed insoluble aggregate. It was tried to avoid this effect by increasing the intervals
between injections to the maximum of 380 s (Fig.3), however it did not affect the integral enthalpies
of interaction (∆Hexp) and therefore, periods of 180 s were used to reduce the overall length of the
experiments.
3.1.3 Data Analysis
Pow
er (μ
J/s)
50
60
70
80
90
100
110
Time (min)0 20 40 60 80 100 120
(a)
δH (k
J/mol
)
−80
−60
−40
−20
0
Molar ratio0 1 2 3
Parameters:ΔHexp = -64.98 kJ/mol ΔHfit = -70.69 kJ/mol Kb
fit = 1.47x105 N = 1.03
(b)
Figure 2: ITC data of dodecylammonium binding to decane sulfonate at 25℃. (a) Raw data,
straight line represents the baseline. (b) Dosing curve, line represent the fit according to the one
site binding model.
Raw binding curves obtained from the ITC experiments were analysed with Origin 5.0
software package, which has Microcal plug-in. In addition to usual functionality of the Origin, this
plug-in enabled fitting of the data according to the models of ”one-on-one” or ”two-site” binding,
thus providing the possibility to extract valuable information about the reaction kinetics. Most
of the experiments could be treated as an ”one-on-one” binding reactions despite the aggregation
that followed and negligible baseline drift due to structural rearrangement (as in Fig.2). Such data
didn’t require any sophisticated approach as the analysis was straightforward:
1. Conversion of units from calories (cal) to joules (J)
2. Automated fitting of the baseline and it’s straightening with FFT (Fast Fourier Transforma-
tion) filter, which removes ”high-frequency” fluctuations in the baseline without changing its
overall shape OR approximation of the baseline as a straight line if the power supplied to the
cell does not differ significantly before and after experiment (baseline drift ≤1 − 2 µJ/s).
3. Integration of the raw data curve (Power, µJ/s over time, s) and automatic conversion to the
dosing curve (δH, kJ/mol over Molar ratio)
29
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
4. Subtraction of the dilution energy, which is determined by the last injections of constant
partial enthalpy change (δH)
5. Integration of the dosing curve, thus acquiring ∆Hexp
6. Fitting curve with ”one-on-one” binding model, thus acquiring binding constant (Kfitb , M−1),
enthalpy change of the reaction (∆Hfit, kJ/mol), stoichiometry (N) and entropy change (∆S,
J/(mol K)), which was calculated using the equation 2
In some cases baseline drift was greater than 1 − 2 µJ/s and therefore straight baseline
was preferred over the automatically determined one in order to keep consistency in the data
analysis. Such a difference in heat-flux supplied to the calorimeter cell before and after experiment
was primarily caused by periodical temperature change in the room during the course of the day.
These fluctuations were described in the manual of MCS and are natural for such a temperature
sensitive instrument as calorimeter is.
The dilution enthalpy of a higher concentration surfactant solution being injected into one
of lower concentration was determined to be comparably small (overall ≈ 2 kJ/mol) and constant
during the course of titration. Therefore it was decided to simplify the process of data analysis
and subtract the dilution enthalpy acquired in the last injections of each titration experiment if
two conditions were satisfied: dilutions’ partial enthalpy (δH) was smaller than 2 kJ/mol and it
remained constant after the end of the aggregation.
Not all experiments could be analysed with the mentioned straight-forward approach,
especially ones for a longer overall aliphatic chain length m . These reactions exhibited greater
influence of aggregation process after the ionic pairing. Thus, the second part of the reaction
became more apparently expressed and the usual ”one-in-one” reaction scheme could not be applied
(Fig.3). There is no appropriate model devised to evaluate kinetics of such a two part – ionic pairing
and aggregation/dissolution reaction. However, it was possible to acquire rough estimates with a
”two binding site” or ”alosteric/cooperative” binding model. This model was devised for a case,
when protein has two active sites which alosterically regulate one another. During the course of
reaction, binding of a ligand to one of the active sites increases its affinity to the other. In our
case ionic pairing represented kinetics of the first binding site and aggregation – of the second
binding site. Thermodynamic parameters Kfitb and ∆Hfit, determined via such binding model
were considered less reliable. In many cases it was not possible to obtain reasonable values of
∆Hfit due to sharp increase in δH near the point of stoichiometric unity (Fig.3b). The accuracy
of ∆Hexp, however, didn’t suffer from these problems as it was obtained by direct integration of
the dosing curve. Therefore ∆Hexp and not ∆Hfit was used as a measure of reaction enthalpy
change in this study.
30
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
Pow
er (μ
J/s)
10
20
30
40
50
60
70
Time (min)0 50 100 150 200
(a)
δH (k
J/mol
)
−200
−150
−100
−50
0
Molar ratio0 1 2 3
Parameters:ΔHexp = -95.07 kJ/molΔHfit = - - kJ/molKb
fit = 5.5x106
N = 1.01
(b)
Figure 3: ITC data of dodecylammonium binding to dodecane sulfate at 25℃. (a) Raw data,
straight line represent the baseline. (b) Dosing curve, line represent the fit according to the II site
binding model.
3.1.4 Statistical analysis of the data
Most reactions have been carried out more than once, therefore it was possible to perform
statistical analysis of the data. First of all it was determined whether data sets were normally
distributed by employing Shapiro-Wilk W test from Statistica 8.0 software package. If such an
assumption could not be ruled out – mean value and confidence intervals of 95% were calculated.
Values of Kfitb have not been analysed directly, but instead were transformed to more informative
form of ∆G (q.1). Such transformation was viable, because Kfitb would have been nevertheless
transformed with the ln Kb to increase the normality of distribution. Confidence intervals of entropy
change T∆S were fully dependent on the ones of ∆G and ∆Hexp, therefore equation 10 has been
used to calculate them. The results for every use of error bars in the figures can be found in the
section of supplementary material (Sec.7).
Err.T ∆S =√
Err.2∆G + Err.2∆H (10)
3.2 The mathematical model
In order to explain the thermodynamics of cationic and anionic surfactant interaction, Dr.
Daumantas Matulis has devised a mathematical model which describes main processes happening
upon formation of the precipitate (Norvaisas et al. 2012). As it was already mentioned, the process
can be divided into two arbitrary stages – ionic pairing and hydrophobic aggregation. The first
part can be expressed via reaction 11:
R1NH+3 + R2SO−
3 ←−→ R1NH+3 · · ·R2SO−
3 , (11)
31
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
where R1 and R2 denote aliphatic chains of both surfactants. The parameter describing overall
chain length of the conjugate – m = R1 +R2 has been used extensively in this study. In the second
stage (reaction 12), following a loss of charge of the surfactants, aggregate of indefinite size ν is
formed which precipitates out of aqueous solution:
ν(R1NH+3 · · ·R2SO−
3 )←−→ (R1NH+3 · · ·R2SO−
3 )ν ↓ . (12)
These two processes are then approximated as independent reactions – ion pair formation in am-
monium sulfate aqueous solution and precipitation of m length alkane. By using such approach it
is then possible to calculate Gibbs free energy of the alkylammonium interaction with alkane sul-
fonate by summing up individual contributions of ionic pairing (∆AGion) and alkane precipitation
(∆AGalk):
∆AG = ∆AGion + ∆AGalk. (13)
The Gibbs free energy of ammonium sulfate aggregation at 1 M reference state is equal to
∆AGion = −RT ln(Aion) = RT ln(Sion), (14)
where Sion is the solubility Aion = 1/Sion – aggregation parameter of ammonium sulfate. For
concentrations different than the 1 M reference state, concentration parameter C was introduced:
∆aggGion = −RT ln(AionC). (15)
The Gibbs free energy of alkane aggregation can be obtained from the following equation (Matulis
2001; Matulis and Bloomfield 2001a):
∆aggGalk = −RT ln(AalkC) = −RT (m ln(∆w) + ln(w0) + ln(C)), (16)
where ∆w is an increase in aggregation upon addition of one methylene (CH2) group and is equal
to 4.241 (Matulis 2001). The parameter w0 is an empirical coefficient and is equal to 1.32 for
n-alkanes (Matulis 2001).
The given model is already sufficient to calculate Gibbs free energy of the surfactant
precipitation. Lets consider a reaction between decane sulfonate and dodecylammonium (see Fig.2
in the section 3.1.3 Data analysis). Required parameters are: C = 0.33 mM, Sion = 4.1 M at
T = 25℃ (Dawson et al. 1986), Aion = 0.244 M−1 and m = 22. Ionic contribution according to
equation (15) is then estimated to be ∆aggGion = 23.4 kJ/mol and alkane precipitation according
(16) – ∆AGalk = −59.6 kJ/mol. Upon summation, the overall expected Gibbs free energy of the
reaction is ∆aggG = −36.2 kJ/mol. The observed Gibbs free energy of the mentioned system
was measured with ITC and equals −30.6 kJ/mol. Given at hand the assumptions made, such
a discrepancy is surprisingly small, but larger than the expected error of the calorimeter of 1
kJ/mol . After the further investigation it was discovered that the experimentally observed ∆Gfit
was always smaller by ≈ 5 kJ/mol . Therefore correction by empirical coefficient B = 4.03× 10−5
was introduced.
32
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
The final expression of modelled association constant Kmodb could then be derived. First
of all, association constants for the reactions of ionic pairing (11)
Aion = [R1NH+3 · · ·R2SO−
3 ][R1NH+
3 ][R2SO−3 ]
(17)
and aggregation (12)
Aalk = [R1NH+3 · · ·R2SO−
3 ↓ ][R1NH+
3 · · ·R2SO−3 ]
(18)
were determined. Then, the factors of concentration C and empirical correction coefficient B were
introduced, thus giving the final expression of
Kmodb = [R1NH+
3 · · ·R2SO−3 ↓ ]
[R1NH+3 ][R2SO−
3 ]BC = AionBAalkC. (19)
The modelled Gibbs free energy can then be calculated with
∆aggGmod = −RT ln(Kmodb ). (20)
The enthalpy of alkane aggregation into a solid phase at 25℃ can be estimated by an
empirical equation (Matulis and Bloomfield 2001b)
∆aggHalk = −5.2m + 18.41, (21)
where m is the number of aliphatic atoms. In the case of alkylammonium interaction with the
alkyl sulfate or alkane sulfonate, rough approximation of the enthalpy can be derived by adding
the enthalpy of ammonium sulfate crystallization (−6.62 kJ/mol, (Dean 1999)) to the equation 21,
which then becomes
∆aggHsurf = −5.2m + 11.79. (22)
3.3 Results
3.3.1 Dependence on the order of titration
First of all it was necessary to determine whether the reaction thermodynamics were
dependent only on the state of the system and not on the route to it. This was done by performing
titration reactions where the contents of the cell and syringe were swapped. If we analyse phase
diagram of the mixed surfactant system (Fig.1), it can be seen that the precipitation occurs on
the diagonal line. During titration, surfactant in the cell is slightly diluted and concentration of
the surfactant from the syringe increases steadily until and after the equimolar ratio is reached.
If we switch the contents of the syringe and the cell – same state will be approached by going
through different pseudophases in the phase diagram (Fig.1) (Amante 1991; Stellner et al. 1988).
This was tested in several systems of the surfactants and as it can be seen in the example of
dodecylammonium/decyl sulfate (Fig.4 and Tab.3) – no significant difference was observed.
33
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
δH (k
J/mol
)
−80
−60
−40
−20
0
Molar ratio0 0.5 1 1.5 2
cell - decane sulfonatesyringe - dodecylammonium
syringe - decane sulfonatecell - dodecylammonium
Figure 4: Dodecylammonium titrated to decane sulfonate and vice versa at 25℃, with [cell] = 0.33
mM, [syringe] = 5 mM.
Cell Syringe ∆Hexp Kfitb Nfit
C10H21SO–3 (0.33 mM ) C12H25NH+
3 (5 mM) −70.69 1.47× 105 1.03
C12H25NH+3 (0.33 mM ) C10H21SO–
3 (5 mM) −70.93 1.96× 105 1.05
Table 3: Dodecylammonium titrated to decane sulfonate at 25℃: dependence on the order of
titration.
3.3.2 Dependence on the surfactant concentration
The association constant of an usual reaction does not depend on the experimental
concentration. However, if the reaction is followed by an aggregation as in the case of the positively
and negatively charged surfactants, linear dependency in the apparent binding constant Kfitb can
be observed. This was determined with the system of dodecylammonium/decyl sulfonate at various
concentrations. Upon the proportional increase in the concentration of surfactant in the cell and
syringe (and thus concentration of the final aggregate), Kfitb increased (Fig.5,6) so that the ratio
of the concentrations is approximately proportional to the ratio of Kfitb (Fig.6). Therefore, 2-
fold increase in concentration leads to the 2-fold increase in the Kfitb . The enthalpy, however,
was constant for all concentrations within the experimental error, thus we could be sure that the
precipitation is full for m = 22. These results agree with the predictions of the mathematical
model (see supplementary material, tables 6,2,7).
34
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
δH (k
J/mol
)
−80
−60
−40
−20
0
Molar ratio0 0.5 1 1.5 2 2.5
[cell] (mM) 0.66 0.33 0.165
Figure 5: Dodecylammonium binding to decane sulfonate at 25℃: Dosing curves of 0.66 (open
triangles), 0.33 (solid circles), and 0.165 (open squares) decane sulfonate concentrations in the cell.
Dodecylammonium concentrations in the syringe were 10 mM, 5mM and 2.5 mM, respectively.
K bfit
(M-1
)
0
105
2×105
3×105
4×105
[cell] (mM)0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Figure 6: Dodecylammonium binding to decane sulfonate at 25℃: the observed binding constant
dependence on decane sulfonate concentration in the cell. For the statistical analysis data see
supplementary material, table 7).
35
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
δH (k
J/mol
)
−100
−80
−60
−40
−20
0
Molar ratio0 0.5 1 1.5 2
m22 23 24
Figure 7: Dodecylammonium binding to alkane sulfonate of varying aliphatic chain length. Filled
circle – decane sulfonate, open square – undecane sulfonate and filled diamond – dodecane sulfonate.
3.3.3 Dependence on the aliphatic chain length
For the systems of cationic and anionic surfactants all three thermodynamic parameters
exhibited clear dependence on the total aliphatic chain length m. This was anticipated from the
results of previous studies (Matulis and Bloomfield 2001b) and the predictions of the mathematical
model (comparison in the supplementary material, Tab.6). The experiments have been carried
out for surfactants with m = 18 − 25 at the experimental temperature of 25℃ and surfactant
concentration in the cell 0.33 mM. For each chain length m there were at least two different
surfactant systems tested. Surfactants with the total chain length of m = 18 − 21 did not fully
precipitate (see supplementary material Fig.2). However, it was hypothesized that this might
be due to the low experimental concentration, which is not sufficient for the full aggregation.
Indeed, it was determined that the enthalpy for the systems with m = 21 is dependent on the
concentration and reaches plateau with the double concentration of 0.66 mM, whereas m = 22
showed no significant difference in enthalpy for various concentrations (see supplementary material
Fig.2 and Tab.7). Therefore data of m = 21 at the 0.66 mM experimental concentration was used
in the final analysis.
In this study oxygen atom between aliphatic tail and the headgroup of alkyl sulfates
was approximated as an additional methylene group. With such approach, the energetics of the
reactions have added up precisely. However, the enthalpies of alkyl sulfate reaction with alky-
lammoniums were systematically less exothermic when compared with alkane sulfonate reactions
with alkylammoniums. These differences might be attributed to the distinct phase and solubility
36
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
(kJ/m
ol)
−120
−100
−80
−60
−40
−20
0
m, aliphatic chain length20 21 22 23 24 25 26 27
ΔGobs
TΔSobs
ΔHexp
Figure 8: Dependence of the thermodynamic parameters on the alyphatic chain length m with
linear fits. For statistical analysis data see supplementary material tables in 7,7 and 6.
behaviour of both species (Chen et al. 2004) and possibly, less ordered packaging in the aggregate.
Enthalpy: As it can be seen in figure 8, experimentally observed values of enthalpy
follow a linear trend and becomes more negative for longer aliphatic chains. The linear regression
of the data gave a result of
∆Hexp = −8.38×m + 112.33 (kJ/mol), (23)
with correlation coefficient r = 0.976 and the significance p = 0.004. The contribution of methylene
group to the ratio of the concentrations approximately ∆∆Hexp = −8.38 kJ/mol per CH2.
Gibbs free energy: Gibbs free energies were increasingly more negative upon increasing
aliphatic chain length of the surfactant in all tested series of surfactants (Fig.8). However, due
to the use of ”two binding site” fitting model and steep binding curve, values of Kb seemed to
be underestimated. Besides, according to the Kfitb values with the mathematical model (see
supplementary material table 7), with m = 24 − 25 at the concentration of 0.33 mM, limits of
the calorimeter sensitivity were reached (c ≈ 1000)(Tellinghuisen 2007b). Linear regression gave a
result of
∆Gfit = −1.87×m + 12, 267 (kJ/mol), (24)
with correlation coefficient r = 0.74 and the significance p = 0.15. The contribution of methylene
group to the association process: ∆∆Gfit = −1.87 kJ/mol per CH2.
Entropy: The entropy values were determined by using equation 2 and therefore they
were completely dependent on enthalpy and Gibbs free energy. This thermodynamic parameter
should be considered the least reliable one. It was only possible to give a qualitative conclusion
37
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
δH (k
J/mol
)
−80
−60
−40
−20
0
Molar ratio0 0.5 1 1.5 2 2.5
[NaCl] (M) 0 0.5 1.0
Figure 9: Dosing curves of the dodecylammonium binding to decane sulfonate at 25℃ with surfac-
tant concentration in the cell 0.33 mM: dependence of the partial enthalpy on the concentration
of NaCl
that the aggregation of long chained cationic and anionic surfactants was primarily driven by the
huge negative enthalpy change and opposed by unfavorable entropy (Fig.8). T∆Sobs data has been
approximated with the linear fit of
T∆Sobs = −7.28×m + 119, 33 kJ/mole, (25)
with correlation coefficient r = 0.9784 and the significance p = 0.004. The contribution of methy-
lene group to the association process: T∆∆Sobs = −7.28 kJ/mole per CH2 at 25℃.
3.3.4 Dependence on the ionic force of the solvent
The series of titration experiments were carried out with dodecylammonium/decane sul-
fonate and different concentrations of salt in both cell and syringe to dissect the role of the elec-
trostatics in the reaction. If the association is primarily driven by the ionic interaction of the
headgroups, then the saline solution with heightened ionic force would inhibit it. However, the
results suggested that the ionic interaction is not so important as it was suggested (Bai et al. 2001),
(Fig.9). The effect of salt only appeared when 0.2 − 0.5 M of NaCl was added – concentration
600− 1500 times higher than that of a surfactant in the cell (0.33 mM). In the extreme case with
the 1 M of NaCl (3000 higher than [cell]), reaction enthalpy determined according to the model
(∆Hfit) significantly decreased and stoichiometry shifted to the higher values (Fig.9). However,
change in the ∆Hexp which is determined by integrating the dosing curve was not so pronounced.
38
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
Shift in the stoichiometry means, that more surfactant from the syringe was need for the precipi-
tate to form. It seems that the salt ions (Na+, Cl–) were competitively inhibiting the association
of the headgroups of surfactants (R1NH+3 , R2SO–
3). Therefore, a higher concentration of differ-
ently charged surfactant was needed to shift the equilibrium back to the formation of complex
R1NH+3 · · ·R2SO–
3.
3.3.5 Dependence on the experimental temperature
The series of experiments with dodecylammonium/decane sulfonate were performed at
various temperatures to determine the heat capacity ∆Cp, temperature of melting Tm and enthalpy
of fusion ∆Hfus for this system. It was observed that the aggregate undergoes a phase transition
into a liquid state at the Tm ≈46−51℃ (Fig.10a). From the slope of the enthalpy dependency on
the temperature, heat capacity for solid and liquid state of the aggregate was determined to be
∆Csolidp = −1.56 kJ/(mol K) and ∆Cliquid
p = −0.23 kJ/(mol K) respectively.
It was interesting to compare physical properties of surfactant aggregate with those of a
pure alkane of similar aliphatic chain length m = 22 (Fig.10b). Surprisingly enough, these both
compounds melt at nearly the same temperature and with the same enthalpy of fusion (∆Hfus ≈
−75 kJ/mol ).
3.3.6 Structure of the aggregate
We found no studies concerning the structure of our investigated surfactant systems,
apart from chemical composition (Stellner et al. 1988). Therefore, a schematic model of surfactant
packing was developed using crystallographic data kindly provided by Dr. Musti J. Swamy, of
the most smilar conjugate available O-lauroylethanolamine-dodecylsulfate (Tarafdar et al. 2010).
O-lauroylethanolamine was modified to dodecylammonium and the resulting structure optimized
using the semiempirical quantum chemistry algorithm MOPAC (Stewart 2008) in the open source
program Gabedit QC (Allouche 2011). During the optimization, decylsulfate molecules as well as
dodecylammonium headgroups were frozen assuming no qualitative difference in hydrogen bonding.
11 shows the modeled structure of dodecylamine bound to dodecyl sulfate. The oppositely charged
ionic headgroups bind to each other stoichiometrically as determined by ITC. Aliphatic tails form
well-packed layer of parallel hydrophobic tails.
3.4 Discussion
This research investigated the thermodynamics of oppositely charged surfactant hy-
drophobic aggregation for a series of surfactant systems with varying aliphatic chain length. By
39
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
ΔH
(kJ
/mo
l)
−100
−80
−60
−40
−20
0
(b)
(a)
C22H46
ΔH
(kJ
/mo
l)
−100
−80
−60
−40
−20
0
T (oC)20 30 40 50 60 70
ΔHfus
ΔHfus+ΔHstr
C18H
38
C19H
40
C20H
42
C21H
44
C22H
46
C23H
48
C24H
50
C25H
52
C26H
54
Figure 10: (a) Enthalpy dependence on temperature for decane sulfonate reaction with dodecy-
lammonium. For statistical analysis data see supplementary material table 7 (b) Enthalpies of
phase transitions of alkanes containing various number of carbon atoms (from 18 to 26) (Lide
et al. 2009): narrow grey bars represent the enthalpy changes of solid-to-liquid phase transition
(∆Hfus), while black bars represent the sum of enthalpy changes ∆Hfus + ∆Hstr, where ∆Hstr
is related to structural rearrangements of molecules in solid phase. The grey patterned column
marks temperature interval at which both the structural and solid-to-liquid phase transitions in
C22H46 occurs. (Temperature and enthalpy data taken from (Dirand et al. 2002)).
40
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
Figure 11: Packing diagrams of dodecylamine complex with dodecyl sulfate: (a) crystalline lat-
tice unit cell, (b) and (c) – orthogonal enlarged views of the hydrogen bonding network with
the distances between ionic headgroups. The model was build according to the structure of O-
Lauroylethanolamine complex with dodecylsulfate provided by Tarafdar et al. (2010).
41
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
changing the experimental conditions in respect to the temperature, surfactant and salt concentra-
tion it was possible to dissect thermodynamic additivity rules for the chemical groups (Dill 1997;
Matulis 2001) and several interesting properties of such systems were determined for the first time.
Thermodynamics of reactions between the cationic and anionic surfactants were shown
to depend on concentration in terms of the association constant Kb. This effect is unusual and pri-
marily arises due to the aggregation reaction which follows the ion pair formation of the surfactant
headgroups. However, this was anticipated from the results of the mathematical model describing
the process.All thermodynamic parameters of the reactions showed linear dependence on the total
aliphatic chain length m of the surfactant systems. The linear regression of data suggests that
there are deviations from the expected values determined by model. However, in the case of Gibbs
free energy it could be explained by a limiting factor of calorimeters capability of determining
association constants with high values – Kb > 107 − 108, at the given experimental concentration.
And also the influence of the fitting model, which, in this case, was not completely appropriate. In
the case of ∆Hexp, mathematical model approximated the contribution of one methylene group for
the aggregation of surfactants as that of a pure alkane (see Tab.2, Eq.22 section 3.2 and Matulis
and Bloomfield (2001b)). When this assumption is taken into account, the discrepancy actually
appears to be surprisingly small.
The ionic contributions to the overall process seems to be non-critical. Only very high
concentration of salt could inhibit the association of the headgroups. This is mainly caused by
the overall unfavourable energetics of ammonium sulfate formation as both headgroups are highly
soluble. Therefore, the process was primarily driven by the association of the hydrophobic tails
and the effect of saline environment was dampened. This is rather unexpected result, because in
other studies on hydrophobic effect, much of the energetics of hydrophobic reactions have been
attributed to the ionic pair formation (Bai et al. 2001).
The association reactions between cationic and anionic surfactants were driven by huge
negative enthalpy change and the contribution of entropy was unfavourable. Analogous reactions
of polymer-polymer or polymer-surfactant association (Courtois and Berret 2010) are favored by
entropy. However, according to the crystallographic data (Tarafdar et al. 2010) (Fig.11), such high
enthalpy in the association of the oppositely charger surfactants could be attributed to the tight
packaging of aliphatic tails in the aggregate. These observations expand the conventional under-
standing of the hydrophobic effect, which states that the entropy drives hydrophobic interactions.
Instead, here we confirm our earlier findings with alkylamines (Matulis 2001), that long aliphatic
chain association in water is an enthalpy-driven and entropy-opposed process.
ITC experiments of dodecylammonium binding to decane sulfonate carried at various
temperatures revealed physical properties of the aggregate. Negative heat capacity change ∆Cp
indicated that indeed, the observed reactions were driven by the hydrophobic effect (Baldwin 1986).
42
Povilas Norvaišas Thermodynamics of Cationic and Anionic Surfactant Interaction
Properties of the liquid aggregate at temperature higher than ≈ 50℃ were significantly different
from those of a solid one. Enthalpy contribution was measured to be only ≈ −30 kJ/mol instead
of nearly -100 kJ/mol. Therefore, it may be concluded that if aliphatic chains are unable to form
a solid phase, the enthalpic contribution will be much smaller than if the solid phase is formed.
This was the first research to extensively investigate thermodynamics of long chained
aliphatic compound aggregation. Result might not have a huge relevance from the practical point
of view, however, the selected model system of surfactants can help us better understand the
fundamental processes behind formation of biological membranes and surfactant induced protein
denaturation. The additivity values for chemical groups which were derived in this study can be
used in approximation of the thermodynamics for hydrophobic effect induced aggregation.
The results were published as: Norvaišas, P., Petrauskas, V., & Matulis, D. (2012).
Thermodynamics of Cationic and Anionic Surfactant Interaction. The Journal of Physical Chem-
istry B, 116(7), 2138-44. doi:10.1021/jp2095888
43
4 Laws of additivity in the structure-based drug design
4.1 Materials & methods
Figure 12: Schematic depiction of the modelling process.
4.1.1 Protein and its structure
The crystallographic structure of human carbonic anhydrase II was obtained from the
”RSCB PDB” protein data bank (Berman 2000). The priorities in selecting the particular data
were high resolution and non-ambiguously resolved position of dynamic amino acid residues. The
requirements were satisfied by the structure PDB ID: 3M96, resolved in 1.4 Å (Čapkauskaitė
et al. 2010). The protein was further prepared by removing ligand and water molecules, then
adding charges and explicit hydrogens. Upon addition of the hydrogens it was selected that the
δ1 nitrogens of histidine residues were protonated, except the case of His119, whose δ1 nitrogen
forms a bond with Zn2+ and therefore ϵ1 nitrogen is protonated.
4.1.2 Inhibitors and their structures
The inhibitors of hCAII were synthesised by Dr. Virginija Dudulienė (Labanauskas et al.
2009; Baranauskienė et al. 2010; Dudutiene et al. 2007) and Edita Čapkauskaitė (Čapkauskaitė
et al. 2010; Capkauskaitė et al. 2012). They were named according to the initials of the chemist
and therefore will be referred to as VD-compounds and E-compounds. All of these inhibitors
were benzensulfonamide derivatives, however, their structural properties differ significantly and
therefore analysis was performed on two separate sets of data.
Initially, data of inhibitor structures was collected and imported to the Instant JChem
database, to check whether there are no duplicates. After standardising the structures, adding
the explicit hydrogens and performing initial optimisation, they were exported in the MDL .mol
44
Povilas Norvaišas Laws of additivity in the structure-based drug design
Figure 13: An example of the structures of the inhibitors. (a) E-compound, (b) VD-compound.
formate. Each structure was then optimized with OBabel (O’Boyle et al. 2011) by using MMFF94
forcefield (Halgren 1996) and selecting conformer with the lowest energy out of 20 candidates.
4.1.3 Software
Various software packages were required to perform this study. First of all, programs
which are capable of visualizing protein and ligand structures, like Accelrys Discovery Studio 3.0
(Accelrys Software 2012), UCSF Chimera 1.5.3. (Pettersen et al. 2004) and Avogadro (Avogadro
2012). Then – command line tools to automate various transformations and adjustments made to
the number of structures. The mostly used one was OBabel (O’Boyle et al. 2011) – management and
conversion tool for all popular chemistry file formats. Where possible, processes were automated
with scripts written in Bash and Perl.
4.1.4 Database
During the process of this study it became clear that the management of such a huge
amount of data concerning structures of the inhibitors, their physical, structural, chemical proper-
ties and the results of their experimentally determined and modelled binding affinities can not be
accomplished without a proper tool. Therefore, an academic license for the ChemAxons’ JChem
cheminformatics software package (ChemAxon 2012) was acquired and the database established.
All the data collected in the study was managed and analysed with this software.
4.1.5 Docking
For the docking VDock program was used (Kairys and Gilson 2002), which implements
Mining minima algorithm (David et al. 2001). In this program ligand posing is performed by the
genetic algorithm and scoring is force-field based. Initially other tools like Autodock (Morris et al.
1998) or Autodock Vina (Trott and Olson 2010) were tested. However they lacked the proper
45
Povilas Norvaišas Laws of additivity in the structure-based drug design
Interaction energy Internal energy
• Van der Waals attraction (V vdwattr )
• Van der Waals repulsion (V vdwrep )
• Coulomb electrostatics (V Colintr)
• Van der Waals (V vdwint )
• Coulomb electrostatics (V Colint )
• Dihedral (V Dih)
Table 4: Potentials of protein-inhibitor interaction, determined by docking with Vdock.
parametrisation of the Zn2+ ion in the binding pocket of CA’s and were incapable to restrain
certain parts of the ligand – options, both of which were implemented in the VDock.
After import to the program each ligand’s translational center was placed on a sulfon-
amide amino group and freely rotating bonds were automatically detected. Dreiding force field
(Mayo and Olafson 1990) with the VeraChem’s (Gilson et al. 2003) partial atomic charges were
used for the ligands. In the case of a protein, binding box with the 5Å sides was placed in the
vicinity of the Zn2+ ion. The binding box was oriented in such a way that the Zn2+ was included at
the one corner and diagonal line of the box went into the direction out of the binding pocket. Dur-
ing the docking simulation, translational center of the ligand is always kept in the box, therefore
assuring interaction of the amino group of the ligand with the zinc and prohibiting the generation
of ligand poses where such interaction is not present. The NB buffer – extension of the binding
box where interaction potentials were also calculated, was set to incorporate a whole side of the
protein with the binding pocket. For the protein, CHARMM22 forcefield (Mackerell 2004) was
used. Most of the parameters were set to their default values. However, 20 000 trial conformations
were tested for each ligand before selecting 20 of the highest binding energy instead of the default
value of 3 000 trials per 10 conformation. This let the more extensive testing of ligands with huge
number of rotatable bonds and did not significantly increase the computation time. The effect
of the solvent was modelled with the distance-dependent dielectric approach with ϵij = 4ij . As a
result, VDock has determined various potentials concerning the interaction between protein and
the inhibitor (Tab.4). The total modelled binding energy ∆Gmod is simply expressed as a sum of
these potentials
∆Gmod = V vdwattr + V vdw
rep + V Colintr + V vdw
int + V Colint + V Dih. (26)
Data from the VDock output files was extracted and summarized by using Bash scripts.
4.1.6 PBSA
Poisson-Boltzman surface area method was employed to investigate the effect of solvent
in more detailed way. For each ligand, 5 poses with highest energy were selected and underwent
calculation of solvent accessible surface area change upon binding (∆SASA) by program SIMS
(Vorobjev and Hermans 1997). Afterwards, APBS (Baker et al. 2001) was used to calculate PB
46
Povilas Norvaišas Laws of additivity in the structure-based drug design
VD-compounds E-coumpoundsPB
SAV
Doc
k
ΔG
mod
(kJ/
mo
l)
0
−10
−20
−30
−40
−50
−60
ΔG
mod
(kJ/
mo
l)
0
−20
−40
−60
−80
−100
ΔGexp (kJ/mol)0 −10 −20 −30 −40 −50 −60
ΔGexp (kJ/mol)0 −10 −20 −30 −40 −50 −60
Figure 14: Binding energies obtained by docking with VDock and PBSA method against the
experimental results for both VD- and E-compounds. Lines mark linear regression of the data.
potential (V P B). The overall modelling binding energy could then be determined:
∆Gmod = V P B + γ∆SASA. (27)
The coefficient γ describes the non-polar contribution to the desolvation. In this study, the value
of 6 cal/mol/Å2 (0.025 kJ/mol/Å2) was used (Luo and Sharp 2002).
4.2 Results
The results were gathered and analysed at three stages during the whole modelling pro-
cess: after the initial docking with VDock, PBSA evaluation and the use of fitting models. Accord-
ing to the ranking of VDock, 5 best poses were selected their binding energies compared against the
experimental results collected exclusively in thermal shift assay (TSA) studies made at the depart-
ment (Capkauskaitė et al. 2012; Baranauskienė et al. 2010; Čapkauskaitė et al. 2010). All in all,
47
Povilas Norvaišas Laws of additivity in the structure-based drug design
152 ligand underwent complete modelling cycle and results of 132 of them were compared against
the experimental data. Unfavourable binding energies (∆Gmod > 0) determined with VDock, were
mostly caused by steric clashes of huge ligands in a constrained binding pocket. Results of these
ligands were not included in the final figures and linear regression.
4.2.1 Docking with Vdock
The initial docking with VDock already showed promising results, especially with VD-
compounds, however performed poorly with the E-coumpounds. As it can be seen in the figure
14, VDock has managed to differentiate ligands and distributed them more or less according to
their real binding energies but also made significant errors. The linear approximation of the data
suggested the overall trend to be ∆Gmod ≈ 0.83×∆Gexp +7.72 (r = 0.55, p = 0.00). These results
indicate that there is a significant amount of information about possible ligand binding affinities
in the potentials determined by VDock.
In the case of the E-compounds situation was quite different (Fig.14). VDock didn’t
managed to differentiate good and bad binders and there was no correlation between modelled and
experimental results (∆Gmod ≈ 0.13×∆Gexp − 39.96, r = 0.14, p = 0.03).
4.2.2 PBSA potential
The PBSA analysis returned results more or less similar to those of VDock. Again, VD-
compounds seemed to be better differentiated than E-compounds (Fig.14). However, results for the
VD series were more scattered than in the case of VDock ranking – ∆Gmod ≈ 0.4×∆Gexp−37.68,
r = 0.23, p = 0.00.
E-compounds, again, could not be properly ranked and results were even worse than in
the case of VDock analysis – ∆Gmod ≈ 0.02×∆Gexp − 34.70, r = 0.02, p = 0.79.
4.2.3 Expressions of total modelled binding energy
After the initial analysis of the VDock and PBSA results, three different expressions of
the total modelled binding energy ∆Gmod for different representations of the effect of the solvent
were tested. One already used by Dr. Visvaldas Kairys (Capkauskaitė et al. 2012) and two as
modifications to it. Expression of Dr. Kairys, represented effect of the solvent in terms of the
solvent accessible surface area change upon binding ∆SASA, whereas two models of mine used
hydrophobic solvent accessible surface area of the ligand (SASAHyd) and the Van der Waals
surface area of the ligand (SAvdw). They were named according to the differing approach to the
solvent – ∆SASA-model, SASAHyd-model and the SAvdw-model accordingly. All three expression
48
Povilas Norvaišas Laws of additivity in the structure-based drug design
VD-compounds E-compounds
SA
SA
vd
wS
AH
yd
∆S
AS
A
ΔG
mod
(kJ/
mo
l)
0
−20
−40
−60
−80
ΔG
mod
(kJ/
mo
l)
0
−10
−20
−30
−40
−50
−60
ΔG
mod
(kJ/
mo
l)
0
−10
−20
−30
−40
−50
−60
ΔGexp (kJ/mol)0 −10 −20 −30 −40 −50 −60
ΔGexp (kJ/mol)0 −10 −20 −30 −40 −50 −60
Figure 15: Modelled binding energies obtained by fitting models to the experimental data. Lines
mark linear regression of the data.
49
Povilas Norvaišas Laws of additivity in the structure-based drug design
incorporated coefficients α, β, γ and δ for different potentials involved in the binding process (more
information in Tab.5). They represented the weights for different potentials used in the expressions.
Values of these coefficients were determined by equating the expression to the experimental data
∆Gexp and finding the best fit with the least squares method.
Contribution Coefficient Representation
Electrostatic α V Colintr + V P B
Van der Waals β V vdwattr + V vdw
rep
Solvent effect γ ∆SASA, SASAhyd or SAvdw
Linear error δ –
Table 5: Potentials and their coefficients used in the expressions of total modelled binding energy
∆Gmod.
The ∆SASA-model:
α(V Colintr + V P B) + β(V vdw
attr + V vdwrep ) + γ∆SASA + δ = ∆Gexp (28)
Results obtained by using the ∆SASA-model were somewhere intermediate between those purely of
VDock and PBSA. Once again VD-compounds were less scattered and exhibited greater correlation
with the experimental results (Fig.15): ∆Gmod ≈ 0.24×∆Gexp − 33.95, r = 0.50, p = 0.00.
Fitted binding energies of E-compounds once again did not show any significant correla-
tion with the experimental data (Fig.15): ∆Gmod ≈ 0.11×∆Gexp − 36.97, r = 0.22, p = 0.00.
The SASAHyd-model:
α(V Colintr + V P B) + β(V vdw
attr + V vdwrep ) + γSASAhyd + δ = ∆Gexp (29)
The approach of SASAHyd-model was to include property, which depends only on the ligand and
not on the binding event like ∆SASA does. Quite surprisingly, results for VD-compounds were
just as good as with the ∆SASA-model (Fig.15): ∆Gmod ≈ 0.50 × ∆Gexp − 34.22, r = 0.51,
p = 0.00.
For the E-compounds, however, it was the worst case, with even negative correlation
coefficient (Fig.15): ∆Gmod ≈ −0.02×∆Gexp − 60.76, r = −0.01, p = 0.92.
The SAvdw-model:
α(V Colintr + V P B) + β(V vdw
attr + V vdwrep ) + γSAvdw + δ = ∆Gexp (30)
Much like the SASAHyd-model, this approach included property dependent on the ligand alone –
its Van der Waals surface area (SAvdw). However, neither VD-compounds (∆Gmod ≈ −43.15 +
0.09×∆Gexp, r = 0.36, p = 0.00.) nor E-compounds (∆Gmod ≈ +0.01×∆Gexp− 40.73, r = 0.05,
p = 0.41.) were ranked correctly with this model.
50
Povilas Norvaišas Laws of additivity in the structure-based drug design
4.3 Discussion
The approach of modelling taken in this study can be regarded as somewhat a mixture
of screening (Grüneberg et al. 2002) and the ”precise” docking techniques (Capkauskaitė et al.
2012). As in the screening studies, multiple ligands were ranked without paying attention to their
correlation to the crystallographic data. But, 152 ligands used in this study is nothing near to the
thousands or hundreds of thousands compounds used in the screening approaches. Similarly to the
usual ”precise” docking studies, more sophisticated methods of solvent approximation and data
analysis were employed. But, none of the ligand poses were analysed in very detail or compared
to the crystallographic data. Besides, the ligands weren’t grouped according to the details of their
structures (Capkauskaitė et al. 2012) and only according their main scaffold.
The initial binding energies obtained with VDock, exhibited good correlation with the
experimental data for VD-compounds. However, the method was not capable of distinguishing
good binders in the set of E-compounds. Overall, if the E-compounds are neglected, VDock gave
the best experimental and modelled energies correlation with the least effort.
The systematically lower correlation for the modelled energies of E-coumpounds rep-
resents quite different structural properties of these inhibitors. VD-compounds are comparably
bigger and because of that, Van der Waals interactions dominates their binding mode. Whereas,
E-compounds are smaller and more influenced by the the electrostatic potential. However, the ef-
fect might also been caused by high structural variability of the E-compounds. This concludes, that
the method used in this study cannot be employed to rank structurally very distinct compounds
when they are the same set.
Comparison of the four methods used to evaluate the effect of solvation provided interest-
ing results. Firstly, it approved the ∆SASA-model used by Dr. Kairys (Capkauskaitė et al. 2012),
which overly gave quite satisfactory result for the VD-compounds (Fig.15). Secondly, SASAHyd
method, which approximated the solvation effect by using hydrophobic solvent accessible surface
area of each ligand, gave comparable results with those of ∆SASA model. This indicates that
SASAHyd might carry relevant information concerning the possible binding energy of each ligand
without having any concerning the particular binding event. Van der Waals surface area of each
ligand (SAvdw), in its regard did not carry such information at all and PBSA method performed
surprisingly poorly.
The results obtained with such an approach also appear intermediate between the two
mentioned major techniques. With the respect of the screening studies, the observed tendencies
have a comparable level of correlation between the experimental and modelling results (Grüneberg
et al. 2002), especially in the case of VDock and ∆SASA-model output for VD-compounds. How-
ever, these tendencies are only very approximate when compared to the usual, ”precise” docking
51
Povilas Norvaišas Laws of additivity in the structure-based drug design
(Capkauskaitė et al. 2012), primarily because of the use of large number of structurally distinct
compounds.
There is also a huge practical aspect of this research project. For the first time such
number of ligands developed in the Department of Biothermodynamics and Drug Design were
analysed in one study. The workbench has been created, primarily in terms of the database and
the scripts written. The database enables more general observation of various tendencies within the
data and better management of it. Hopefully, it will be further developed for all the proteins and
ligands investigated in the department. The scripts will allow the similar study to be performed
more easily in an automated way.
52
5 Conclusion
In this thesis two different approaches in investigating the biologically relevant systems
in respect to the solvation and hydrophobic effect were described. The first research project con-
cerned the quantification of an already approved simple linear model of chemical groups additivity
in dissecting the thermodynamics of oppositely charged surfactants interaction. Whereas, in the
second study it was tried to identify the best method and simplistic representation of the solvent
in the protein-ligand interaction modelling. Both ”wet” laboratory experimentation methods like
isothermal titration calorimetry and computational methods like docking were employed respec-
tively. The results can be concluded as follows:
Thermodynamics of cationic and anionic surfactant reaction
• The association constant Kb depends on the concentration of surfactants when the aggrega-
tion reaction is taking place
• Ionic interaction of the surfactant heads seems to be not crucial and generally unfavourable
for the process of aggregation
• Thermodynamic parameters depend linearly on the overall aliphatic chain length of the
surfactant conjugate
• The aggregation of oppositely charged surfactants is driven by the huge negative enthalpy
contribution, whereas entropy change is unfavourable
• If the liquid aggregate is formed, the enthalpy contribution is significantly smaller .
Laws of additivity in the structure-based drug design
• VDock program was the most accurate – it provided the best results with the least effort
• Expression of modelled binding energy, which incorporated properties of the ligand alone
(SASAHyd) performed just as good as one that used properties regarding the binding event
(∆SASA)
• It was not possible to rank multitude of significantly different inhibitors with the same method
• The workbench for computational and structure-activity relation studies has been established
in the department and, hopefully, it will be developed further.
53
6 Acknowledgements
I would like to thank both of my supervisors Dr. Daumantas Matulis and Dr. Visvaldas
Kairys for their friendly and sincere attitude towards me and other students. It was a pleasure
both learning and working with them. One another colleague which was not mentioned as my
supervisor, but definitely should have been so is Dr. Vytautas Petrauskas. From him I have
gained invaluable experience, which helped me in various projects that I have been involved in. I
would also thank Lithuanian Science Council for the financial support given during my work in
both projects – Student Summer Research Practice and Student Scientific Research scholarships.
Special thanks also go to the Dr. Vytautas Smirnovas, whose review and insights helped me to
improve this thesis.
54
Povilas Norvaišas REFERENCES
References
Accelrys Software, I. (2012). Discovery Studio Modeling Environment, Release 3.1.
Allouche, A.-R. (2011). Gabedit–a graphical user interface for computational chemistry softwares.
Journal of computational chemistry, 32(1):174–82.
Alterio, V., Fiore, A. D., Ambrosio, K. D., Supuran, C. T., and Simone, G. D. (2012). Multiple
Binding Modes of Inhibitors to Carbonic Anhydrases: How to Design Specific Drugs Targeting
15 Different Isoforms? Chemical Reviews, -(-):–.
Amante, J. (1991). Precipitation of mixtures of anionic and cationic surfactants II. Effect of surfac-
tant structure, temperature, and pH. Journal of Colloid and Interface Science, 144(1):243–253.
Aniansson, E. A. G., Wall, S. N., Almgren, M., Hoffmann, H., Kielmann, I., Ulbricht, W., Zana,
R., Lang, J., and Tondre, C. (1976). Theory of the kinetics of micellar equilibria and quantita-
tive interpretation of chemical relaxation studies of micellar solutions of ionic surfactants. The
Journal of Physical Chemistry, 80(9):905–922.
Avogadro (2012). Avogadro: an open-source molecular builder and visualization tool. Version 1.0.3.
http://avogadro.openmolecules.net/, 2012 02 02.
Bai, G., Wang, Y., Wang, J., Han, B., and Yan, H. (2001). Microcalorimetric studies of
the interaction between DDAB and SDS and the phase behavior of the mixture. Langmuir,
17(12):3522–3525.
Baker, N. a., Sept, D., Joseph, S., Holst, M. J., and McCammon, J. a. (2001). Electrostatics
of nanosystems: application to microtubules and the ribosome. Proceedings of the National
Academy of Sciences of the United States of America, 98(18):10037–41.
Baldwin, R. L. (1986). Temperature dependence of the hydrophobic interaction in protein folding.
Proceedings of the National Academy of Sciences of the United States of America, 83(21):8069–72.
Baranauskiene, L., Petrikaite, V., Matuliene, J., and Matulis, D. (2009). Titration calorimetry
standards and the precision of isothermal titration calorimetry data. International journal of
molecular sciences, 10(6):2752–62.
Baranauskienė, L., Hilvo, M., Matulienė, J., Golovenko, D., Manakova, E., Dudutienė, V.,
Michailovienė, V., Torresan, J., Jachno, J., Parkkila, S., Maresca, A., Supuran, C. T., Gražulis,
S., and Matulis, D. (2010). Inhibition and binding studies of carbonic anhydrase isozymes I, II
and IX with benzimidazo[1,2-c][1,2,3]thiadiazole-7-sulphonamides. Journal of enzyme inhibition
and medicinal chemistry, 25(6):863–70.
Baum, B., Muley, L., Smolinski, M., Heine, A., Hangauer, D., and Klebe, G. (2010). Non-additivity
55
Povilas Norvaišas REFERENCES
of functional group contributions in protein-ligand binding: a comprehensive study by crystal-
lography and isothermal titration calorimetry. Journal of molecular biology, 397(4):1042–54.
Ben-Naim, a. and Yaacobi, M. (1974). Effects of Solutes on the Strength of Hydrophobic Interaction
and Its Temperature Dependence. The Journal of Physical Chemistry, 78(2):170–175.
Berman, H. M. (2000). The Protein Data Bank. Nucleic Acids Research, 28(1):235–242.
Bernal, J. D. and Fowler, R. H. (1933). A Theory of Water and Ionic Solution, with Particular
Reference to Hydrogen and Hydroxyl Ions. The Journal of Chemical Physics, 1(8):515.
Capkauskaitė, E., Zubrienė, A., Baranauskienė, L., Tamulaitienė, G., Manakova, E.,
Kairys, V., Gražulis, S., Tumkevičius, S., and Matulis, D. (2012). Design of [(2-
pyrimidinylthio)acetyl]benzenesulfonamides as inhibitors of human carbonic anhydrases. Eu-
ropean journal of medicinal chemistry, 51.
Chaires, J. B. (1997). Possible origin of differences between van’t Hoff and calorimetric enthalpy
estimates. Biophysical chemistry, 64(1-3):15–23.
Chaplin, M. (2012). Water structure and science. http://www.lsbu.ac.uk/water/index.html, 2012
05 12.
ChemAxon (2012). Instant JChem: structure database management, search and prediction.
http://www.chemaxon.com, Version: 5.
Chen, L., Xiao, J.-X., and Ma, J. (2004). Striking differences between alkyl sulfate and alkyl
sulfonate when mixed with cationic surfactants. Colloid & Polymer Science, 282(5):524–529.
Chen, L.-J., Lin, S.-Y., and Huang, C.-C. (1998). Effect of Hydrophobic Chain Length of Surfac-
tants on Enthalpy−Entropy Compensation of Micellization. The Journal of Physical Chemistry
B, 102(22):4350–4356.
Cornish-Bowden, A. (2002). Enthalpy-entropy compensation: a phantom phenomenon. Journal
of biosciences, 27(2):121–6.
Courtois, J. and Berret, J. (2010). Probing Oppositely Charged Surfactant and Copolymer Inter-
actions by Isothermal Titration Microcalorimetry. Langmuir, (12):315–318.
David, L., Luo, R., and Gilson, M. K. (2001). Ligand-receptor docking with the Mining Minima
optimizer. Journal of computer-aided molecular design, 15(2):157–71.
Dawson, R. M. C., Elliott, D. C., Elliott, W. H., and Jones, K. M. (1986). Data for biochemical
research. Clarendon Press, Oxford.
Dean, J. A. (1999). Lange’s Handbook of Chemistry. McGraw-Hill, Inc., New York, 15th edition.
56
Povilas Norvaišas REFERENCES
Dill, K. a. (1997). Additivity principles in biochemistry. The Journal of biological chemistry,
272(2):701–4.
Dirand, M., Bouroukba, M., Briard, A., Chevallier, V., Petitjean, D., and Corriou, J. (2002).
Temperatures and enthalpies of (solid+ solid) and (solid+ liquid) transitions of n-alkanes. The
Journal of Chemical Thermodynamics, 34(8):1255–1277.
Doyle, M. (1997). Characterization of binding interactions by isothermal titration calorimetry.
Current opinion in biotechnology, 8(1):31–35.
Dudutiene, V., Baranauskiene, L., and Matulis, D. (2007). Benzimidazo[1,2-c][1,2,3]thiadiazole-
7-sulfonamides as inhibitors of carbonic anhydrase. Bioorganic & medicinal chemistry letters,
17(12):3335–8.
Falconer, R. J., Penkova, A., Jelesarov, I., and Collins, B. M. (2010). Survey of the year 2008:
applications of isothermal titration calorimetry. Journal of molecular recognition : JMR,
23(5):395–413.
Fernandes, R. M., Marques, E. F., Silva, B. F., and Wang, Y. (2010). Micellization behavior of a
catanionic surfactant with high solubility mismatch: Composition, temperature, and salt effects.
Journal of Molecular Liquids, 157(2-3):113–118.
Fernández-Vidal, M., White, S. H., and Ladokhin, A. S. (2010). Membrane Partitioning: ”Classi-
cal” and ”Nonclassical” Hydrophobic Effects. The Journal of membrane biology.
Fogolari, F., Brigo, a., and Molinari, H. (2002). The Poisson-Boltzmann equation for biomolec-
ular electrostatics: a tool for structural biology. Journal of molecular recognition : JMR,
15(6):377–92.
Fogolari, F., Brigo, A., and Molinari, H. (2003). Protocol for MM/PBSA molecular dynamics
simulations of proteins. Biophysical journal, 85(1):159–66.
Frank, H. S. and Evans, M. W. (1945). Free Volume and Entropy in Condensed Systems III.
Entropy in Binary Liquid Mixtures; Partial Molal Entropy in Dilute Solutions; Structure and
Thermodynamics in Aqueous Electrolytes. The Journal of Chemical Physics, 13(11):507.
Gilson, M. K., Gilson, H. S. R., and Potter, M. J. (2003). Fast assignment of accurate partial
atomic charges: an electronegativity equalization method that accounts for alternate resonance
forms. Journal of chemical information and computer sciences, 43(6):1982–97.
Grüneberg, S., Stubbs, M. T., and Klebe, G. (2002). Successful virtual screening for novel inhibitors
of human carbonic anhydrase: strategy and experimental confirmation. Journal of medicinal
chemistry, 45(17):3588–602.
57
Povilas Norvaišas REFERENCES
Halgren, T. A. (1996). Merck molecular force field. I. Basis, form, scope, parameterization, and
performance of MMFF94. Journal of Computational Chemistry, 17(5-6):490–519.
Heerklotz, H. and Seelig, J. (2000). Titration calorimetry of surfactant-membrane partitioning and
membrane solubilization. Biochimica et biophysica acta, 1508(1-2):69–85.
Hildebrand, J. H. (1979). Is there a ”Hydrophobic Effect”? Proceedings of the National Academy
of Sciences, 76(1):194–194.
Hoerr, C., McCorkle, M., and Ralston, A. (1943). Studies on high molecular weight aliphatic
amines and their salts. X. Ionization constants of primary and symmetrical secondary amines in
aqueous solution. Journal of the American Chemical Society, 65(3):328–329.
Hoerr, C. and Ralston, A. (1942a). Studies on High Molecular Weight Aliphatic Amines and
Their Salts. IX. The Behavior of Various Salts of Dodecylamine in Water, Ethanol and Benzene.
Journal of the American Chemical Society, 64(12):2824–2829.
Hoerr, C. and Ralston, A. (1943). Studies on High Molecular Weight Aliphatic Amines and Their
Salts. XI. Transference Numbers of Some Primary Amine Hydrochlorides in Aqueous Solution
and Their Significance in the Interpretation of the Micelle Theory•. Journal of the American
Chemical Society, 65(5):976–983.
Hoerr, C. W. and Ralston, a. W. (1942b). Studies on High Molecular Weight Aliphatic Amines and
Their Salts. IX. The Behavior of Various Salts of Dodecylamine in Water, Ethanol and Benzene.
Journal of the American Chemical Society, 64(12):2824–2829.
Holland, P. and Rubingh, D. (1992). Mixed surfactant systems. In THE ACS Symposium series,
volume 1. ACS Publications.
Imtaiyaz Hassan, M., Shajee, B., Waheed, A., Ahmad, F., and Sly, W. S. (2012). Structure,
function and applications of carbonic anhydrase isozymes. Bioorganic & Medicinal Chemistry,
pages 21–24.
Jelesarov, I. and Bosshard, H. R. (1999). Isothermal titration calorimetry and differential scanning
calorimetry as complementary tools to investigate the energetics of biomolecular recognition.
Journal of molecular recognition : JMR, 12(1):3–18.
Kairys, V. and Gilson, M. K. (2002). Enhanced docking with the mining minima optimizer:
acceleration and side-chain flexibility. Journal of computational chemistry, 23(16):1656–70.
Kauzmann, W. (1959). Some factors in the interpretation of protein denaturation. Adv. Protein
Chem, 14.
Kitchen, D. B., Decornez, H., Furr, J. R., and Bajorath, J. (2004). Docking and scoring in
58
Povilas Norvaišas REFERENCES
virtual screening for drug discovery: methods and applications. Nature reviews. Drug discovery,
3(11):935–49.
Kollman, P., Massova, I., and Reyes, C. (2000). Calculating structures and free energies of com-
plex molecules: combining molecular mechanics and continuum models. Accounts of Chemical
Research, 33(12):889–97.
Kume, G., Gallotti, M., and Nunes, G. (2007). Review on Anionic/Cationic Surfactant Mixtures.
Journal of Surfactants and Detergents, 11(1):1–11.
Labanauskas, L., Dudutiene, V., and Matulis, D. (2009). Synthesis of a new heterocyclic system:
3-phenylbenzimidazo - selenadiazole. Chemistry of heterocyclic compounds, 45(9):1153–1154.
Ladbury, J. E. (2010). Calorimetry as a tool for understanding biomolecular interactions and an
aid to drug design. Biochemical Society transactions, 38(4):888–93.
Latimer, W. and Rodebush, W. (1920). Polarity and ionization from the standpoint of the lewis
theory of valence. Journal of the American Chemical, 42(7):1419–1433.
Lide, D. R., Haynes, W. M. M., Baysinger, G., Berger, L. I., Roth, D. L., Zwillinger, D., Frenkel,
M., and Goldberg, R. N. (2009). CRC Handbook of Chemistry and Physics, 2009−2010, 90th
edition, volume 131.
Liu, Y. and Sturtevant, J. M. (1997a). Significant discrepancies between van’t Hoff and calorimetric
enthalpies. II. Biophysical chemistry, 64(1-3):121–6.
Liu, Y. and Sturtevant, J. M. (1997b). Significant discrepancies between van’t Hoff and calorimetric
enthalpies. III. Biophysical chemistry, 64(1-3):121–6.
Luo, H. and Sharp, K. (2002). On the calculation of absolute macromolecular binding free en-
ergies. Proceedings of the National Academy of Sciences of the United States of America,
99(16):10399–404.
Mackerell, A. D. (2004). Empirical force fields for biological macromolecules: overview and issues.
Journal of computational chemistry, 25(13):1584–604.
Madan, B. and Sharp, K. (1997). Molecular origin of hydration heat capacity changes of
hydrophobic solutes: perturbation of water structure around alkanes. J. Phys. Chem. B,
101(51):11237–11242.
Massova, I. (2000). Combined molecular mechanical and continuum solvent approach (MM-
PBSA/GBSA) to predict ligand binding. Perspectives in drug discovery and design, (18):113–135.
Matulis, D. (2001). Thermodynamics of the hydrophobic effect. III. Condensation and aggregation
of alkanes, alcohols, and alkylamines. Biophysical chemistry, 93(1):67–82.
Matulis, D. (2008). Baltymų fizikinė chemija. Technologija, Kaunas.
59
Povilas Norvaišas REFERENCES
Matulis, D. and Bloomfield, V. a. (2001a). Thermodynamics of the hydrophobic effect. I. Cou-
pling of aggregation and pK(a) shifts in solutions of aliphatic amines. Biophysical chemistry,
93(1):37–51.
Matulis, D. and Bloomfield, V. a. (2001b). Thermodynamics of the hydrophobic effect. II. Calori-
metric measurement of enthalpy, entropy, and heat capacity of aggregation of alkylamines and
long aliphatic chains. Biophysical chemistry, 93(1):53–65.
Mayo, S. and Olafson, B. (1990). DREIDING: a generic force field for molecular simulations.
Journal of Physical, 101(Suite 540):8897–8909.
McAuliffe, C. (1969). Solubility in Water of Normal C9 and C10, Alkane Hydrocarbons. Science,
163(3866):478.
Mizutani, T. (2011). Thermodynamics of supramolecular structure formation in water. In Mizutani,
T., editor, Aplication of Thermodynamics in Biological and materials Science, chapter 34, pages
111–128. InTech.
Moore, T. and Winmill, T. (1912). The state of amines in aqueous solution. J. Chem. Soc., Trans.,
101:1635.
Morris, G. M., Goodsell, D. S., Halliday, R. S., Huey, R., Hart, W. E., Belew, R. K., and Olson,
A. J. (1998). Automated docking using a Lamarckian genetic algorithm and an empirical binding
free energy function. Journal of Computational Chemistry, 19(14):1639–1662.
Nagarajan, R. (1991). Theory of surfactant self-assembly: a predictive molecular thermodynamic
approach. Langmuir, (3):2934–2969.
Norvaisas, P., Petrauskas, V., and Matulis, D. (2012). Thermodynamics of Cationic and Anionic
Surfactant Interaction. The Journal of Physical Chemistry B, 116(7):2138–44.
Nozaki, Y. and Tanford, C. (1971). The solubility of amino acids and two glycine peptides in
aqueous ethanol and dioxane solutions. Journal of Biological Chemistry, 246(7):2211–2217.
Nucci, N. V. and Vanderkooi, J. M. (2008). Effects of Salts of the Hofmeister Series on the Hydrogen
Bond Network of Water. Journal of molecular liquids, 143(2-3):160–170.
O’Boyle, N. M., Banck, M., James, C. A., Morley, C., Vandermeersch, T., and Hutchison, G. R.
(2011). Open Babel: An open chemical toolbox. Journal of cheminformatics, 3(1):33.
Olofsson, G. and Loh, W. (2009). On the use of titration calorimetry to study the association of
surfactants in aqueous solutions. J. Braz. Chem. Soc, 20(4):577–593.
Otzen, D. (2011). Protein-surfactant interactions: A tale of many states. Biochimica et Biophysica
Acta (BBA)-Proteins & Proteomics, 1814(5):562–591.
60
Povilas Norvaišas REFERENCES
Papenmeier, G. J. and Campagnoli, J. M. (1969). Microcalorimetry. Thermodynamics of the
reaction of anionic detergent with a cationic detergent. Journal of the American Chemical
Society, 91(24):6579–6584.
Pastorekova, S., Zatovicova, M., and Pastorek, J. (2008). Cancer-associated carbonic anhydrases
and their inhibition. Current pharmaceutical design, 14(7):685–98.
Pettersen, E. F., Goddard, T. D., Huang, C. C., Couch, G. S., Greenblatt, D. M., Meng, E. C.,
and Ferrin, T. E. (2004). UCSF Chimera–a visualization system for exploratory research and
analysis. Journal of computational chemistry, 25(13):1605–12.
Plyasunov, a. (2000a). Infinite dilution partial molar properties of aqueous solutions of nonelec-
trolytes. I. Equations for partial molar volumes at infinite dilution and standard thermodynamic
functions of hydration of volatile nonelectrolytes over wide ranges of conditions. Geochimica et
Cosmochimica Acta, 64(3):495–512.
Plyasunov, a. (2000b). Thermodynamic functions of hydration of hydrocarbons at 298.15 K and
0.1 MPa. Geochimica et Cosmochimica Acta, 64(3):439–468.
Plyasunov, A. A., Plyasunova, N. N., and Shock, E. (2006). Group Contribution Values for
the Thermodynamic Functions of Hydration at 298.15 K, 0.1 MPa. 3. Aliphatic Monoethers,
Diethers, and Polyethers. Journal of Chemical &, 51(1):276–290.
Plyasunov, A. V., O’Connell, J. P., Wood, R. H., and Shock, E. L. (2000). Infinite dilution
partial molar properties of aqueous solutions of nonelectrolytes. II. equations for the standard
thermodynamic functions of hydration of volatile nonelectrolytes over wide ranges of conditions
including subcritical temperatures. Geochimica et Cosmochimica Acta, 64(16):2779–2795.
Plyasunov, A. V., Plyasunova, N. V., and Shock, E. L. (2004). Group Contribution Values for the
Thermodynamic Functions of Hydration of Aliphatic Esters at 298.15 K, 0.1 MPa. Journal of
Chemical & Engineering Data, 49(5):1152–1167.
Plyasunov, A. V. and Shock, E. L. (2001). Group Contribution Values of the Infinite Dilution
Thermodynamic Functions of Hydration for Aliphatic Noncyclic Hydrocarbons, Alcohols, and
Ketones at 298.15 K and 0.1 MPa †. Journal of Chemical & Engineering Data, 46(5):1016–1019.
Plyasunova, N. V., Plyasunov, A. V., and Shock, E. L. (2005). Group Contribution Values for
the Thermodynamic Functions of Hydration at 298.15 K, 0.1 MPa. 2. Aliphatic Thiols, Alkyl
Sulfides, and Polysulfides. Journal of Chemical & Engineering Data, 50(1):246–253.
Ralston, A., Hoerr, C., and Hoffman, E. (1942). Studies on High Molecular Weight Aliphatic
Amines and their Salts. VII. The Systems Octylamine , Dodecylamine and Octadecylamine
Water. Journal of the American Chemical Society, 64(7):1516–1523.
Ralston, A., Hoffman, E., Hoerr, C., and Selby, W. (1941). Studies on high molecular weight
61
Povilas Norvaišas REFERENCES
aliphatic amines and their salts. I. Behavior of the hydrochlorides of dodecylamine and octade-
cylamine in water. Journal of the American Chemical Society, 63(6):1598–1601.
Saito, M. (1982). Dissolution and micellization of sodium n-alkylsulfonates in water. Journal of
Colloid and Interface Science, 88(2):578–583.
SCORPIO (2012). SCORPIO - online repository of protein-ligand complexes
which have been structurally resolved and thermodynamically characterised.
http://scorpio.biophysics.ismb.lon.ac.uk/, 2012 05 10.
Shimizu, S. and Chan, H. S. (2000). Temperature dependence of hydrophobic interactions: A mean
force perspective, effects of water density, and nonadditivity of thermodynamic signatures. The
Journal of Chemical Physics, 113(11):4683.
Somasundaran, P., Healy, T. W., and Fuerstenau, D. W. (1964). Surfactant Adsorption at the
Solid–Liquid Interface–Dependence of Mechanism on Chain Length. Journal of Physical Chem-
istry, 68(12):3562–3566.
Stellner, K., Amante, J., Scamehorn, J., and Harwell, J. (1988). Precipitation phenomena in
mixtures of anionic and cationic surfactants in aqueous solutions. Journal of Colloid and Interface
Science, 123(1):186–200.
Stewart, J. J. P. (2008). MOPAC2009. Stewart Computational Chemistry, Colorado S.
Sturtevant, J. (1977). Heat capacity and entropy changes in processes involving proteins. Proceed-
ings of the National Academy of Sciences, 74(6):2236.
Tanford, C. (1978). The hydrophobic effect and the organization of living matter. Science,
200(4345):1012–8.
Tanford, C. (1979). Interfacial free energy and the hydrophobic effect. Proceedings of the National
Academy of Sciences of the United States of America, 76(9):4175–6.
Tanford, C. (1980). The Hydrophobic Effect: Formation of Micelles and Biological Membranes.
John Wiley & Sons, 2 edition.
Tanford, C. (1997). How protein chemists learned about the hydrophobic factor. Protein science
: a publication of the Protein Society, 6(6):1358–66.
Tarafdar, P. K., Reddy, S. T., and Swamy, M. J. (2010). A base-triggerable catanionic mixed
lipid system: isothermal titration calorimetric and single-crystal X-ray diffraction studies. The
journal of physical chemistry. B, 114(43):13710–7.
Tartar, H. V. and Wright, K. A. (1939). Studies of Sulfonates. III. Solubilities, Micelle Formation
and Hydrates of the Sodium Salts of the Higher Alkyl Sulfonates. Journal of the American
Chemical Society, 61(3):539–544.
62
Taylor, R. D., Jewsbury, P. J., and Essex, J. W. (2002). A review of protein-small molecule docking
methods. Journal of computer-aided molecular design, 16(3):151–66.
Tellinghuisen, J. (2006). Van’t Hoff analysis of K degrees (T): how good...or bad? Biophysical
chemistry, 120(2):114–20.
Tellinghuisen, J. (2007a). Calibration in isothermal titration calorimetry: heat and cell volume
from heat of dilution of NaCl(aq). Analytical biochemistry, 360(1):47–55.
Tellinghuisen, J. (2007b). Optimizing experimental parameters in isothermal titration calorimetry:
variable volume procedures. J. Phys. Chem. B, 111(39):11531–11537.
Tellinghuisen, J. (2008). Isothermal titration calorimetry at very low c. Analytical biochemistry,
373(2):395–7.
Tripp, B. C., Smith, K., and Ferry, J. G. (2001). Carbonic anhydrase: new insights for an ancient
enzyme. The Journal of biological chemistry, 276(52):48615–8.
Trott, O. and Olson, A. (2010). Software news and update AutoDock Vina: improving the speed
and accuracy of docking with a new scoring function, efficient optimization, and multithreading.
J. Comput. Chem, 31:455–461.
Čapkauskaitė, E., Baranauskienė, L., Golovenko, D., Manakova, E., Gražulis, S., Tumkevičius, S.,
and Matulis, D. (2010). Indapamide-like benzenesulfonamides as inhibitors of carbonic anhy-
drases I, II, VII, and XIII. Bioorganic & medicinal chemistry, 18(21):7357–64.
Velázquez Campoy, A. and Freire, E. (2005). ITC in the post-genomic era...? Priceless. Biophysical
chemistry, 115(2-3):115–24.
Vorobjev, Y. N. and Hermans, J. (1997). SIMS: computation of a smooth invariant molecular
surface. Biophysical journal, 73(2):722–32.
Wiseman, T., Williston, S., Brandts, J. F., and Lin, L. N. (1989). Rapid measurement of bind-
ing constants and heats of binding using a new titration calorimeter. Analytical biochemistry,
179(1):131–7.
Zhang, Y. and Cremer, P. (2006). Interactions between macromolecules and ions: the Hofmeister
series. Current opinion in chemical biology, 10(6):658–663.
63
7 Supplementary material
Kb
[cell], mM W p
0.66 0.97 0.88
0.33 0.91 0.52
0.165 0.78 0.09
Table 2: Dodecylammonium binding to the decane sulfonate. The normality of Kb for different
concentrations of the surfactant in the cell (decane sulfonate), according to Shapiro-Wilk (W) test.
ΔHex
p (kJ
/mol
)
−120
−100
−80
−60
−40
−20
0
m, aliphatic chain length17 18 19 20 21 22 23 24 25 26
ΔHexp(0.33mM)ΔHexp(0.66mM) ΔH=-8.4m+112
Figure 2: ∆Hexp dependence on the aliphatic chain length m at 25℃ with linear regression data.
64
Povilas Norvaišas Supplementary material
[cel
l],
mM
∆H
ex
p,
kJ/m
ol
Erro
r,
±kJ
/mol
∆H
fit
,
kJ/m
ol
Erro
r,
±kJ
/mol
Nf
itEr
ror,
±K
fit
,
M−
1
Erro
r,95
%
conf
.±
M−
1
Aa
lkC
Aio
nB
,
M−
1
∆a
ggG
,kJ/
mol
∆a
ggH
,kJ/
mol
CH
2
T∆
ag
gS
obs,
J/m
olK
fitca
lcco
nc
alk c
on
cfit
calc
alk
0.66
-68.
902.
08-7
0.85
3.74
1.01
0.03
3.49
E+05
3.98
E+04
5.45
E+05
-31.
64-3
2.75
-61.
33-7
0.85
-102
.61
-95.
9912
-39.
20
0.59
4-6
9.51
--7
2.53
-0.
99-
3.04
E+05
-4.
91E+
05-3
1.30
-32.
48-6
1.07
-72.
53-1
02.6
1-9
5.99
13-4
1.23
0.52
8-7
1.00
--7
1.65
-1.
02-
3.25
E+05
-4.
36E+
05-3
1.46
-32.
19-6
0.78
-71.
65-1
02.6
1-9
5.99
13-4
0.19
0.46
2-6
8.83
--7
0.89
-1.
01-
3.11
E+05
-3.
82E+
05-3
1.35
-31.
86-6
0.45
-70.
89-1
02.6
1-9
5.99
12-3
9.54
0.39
6-6
6.68
--6
8.22
-1.
03-
2.72
E+05
-3.
27E+
05-3
1.02
-31.
48-6
0.06
-68.
22-1
02.6
1-9
5.99
12-3
7.20
0.36
3-6
6.65
--6
9.33
-1.
00-
2.17
E+05
-3.
00E+
05-3
0.46
-31.
26-5
9.85
-69.
33-1
02.6
1-9
5.99
12-3
8.87
0.33
-65.
254.
19-7
0.48
5.16
0.98
0.05
1.86
E+05
4.31
E+04
2.73
E+05
-30.
08-3
1.03
-59.
61-7
0.48
-102
.61
-95.
9912
-40.
40
0.29
7-5
4.40
--6
8.26
-0.
96-
1.72
E+05
-2.
45E+
05-2
9.89
-30.
77-5
9.35
-68.
26-1
02.6
1-9
5.99
12-3
8.36
0.26
4-6
6.91
--7
2.83
-0.
98-
1.38
E+05
-2.
18E+
05-2
9.34
-30.
47-5
9.06
-72.
83-1
02.6
1-9
5.99
13-4
3.48
0.23
1-6
3.45
--7
2.28
-0.
96-
1.34
E+05
-1.
91E+
05-2
9.27
-30.
14-5
8.73
-72.
28-1
02.6
1-9
5.99
13-4
3.00
0.19
8-5
9.08
--6
8.50
-0.
98-
1.30
E+05
-1.
64E+
05-2
9.18
-29.
76-5
8.34
-68.
50-1
02.6
1-9
5.99
12-3
9.31
0.16
5-5
5.52
19.6
4-7
2.69
15.9
10.
990.
031.
05E+
052.
63E+
041.
36E+
05-2
8.66
-29.
31-5
7.89
-72.
69-1
02.6
1-9
5.99
13-4
4.03
Tabl
e3:
Dod
ecyl
amm
oniu
mtit
ratio
nto
deca
nesu
lfona
teat
25◦ C
:de
pend
ence
ofth
erm
odyn
amic
para
met
ers
onsu
rfact
ant
conc
entr
atio
n
65
Povilas Norvaišas Supplementary material
m∆
He
xp,
kJ/m
ol
Erro
r,
±kJ
/mol
∆H
fit
,
kJ/m
ol
Erro
r,
±kJ
/mol
Nf
itEr
ror,
±K
fit
,
M−
1
Aa
lkC
Aio
nB
,
M−
1
∆a
ggG
,kJ/
mol
∆a
ggH
,kJ/
mol
CH
2
T∆
ag
gS
obs,
kJ/m
ol
Erro
r,
±kJ
/mol
fitEr
r.ca
lc0.
33m
Mal
k 0.3
3mM
fitca
lcal
k
18-2
.92
–-3
.54
–0.
88–
2.35
E+05
8.43
E+02
-30.
65–
-16.
70-4
5.28
-2.9
2-8
1.81
-75.
192
27.2
6–
19-5
.67
––
––
––
3.58
E+03
––
-20.
28-4
8.87
-5.6
7-8
7.01
-80.
392
––
20-1
.91
3.77
-18.
79–
0.57
–5.
85E+
041.
52E+
04-2
7.21
–-2
3.86
-52.
45-1
.91
-92.
21-8
5.59
224
.88
–
21-5
4.10
4.49
-83.
2648
.50
0.97
0.09
2.76
E+04
6.43
E+04
-25.
353.
25-2
7.45
-56.
03-5
4.10
-97.
41-9
0.79
11-2
8.27
5.54
22-6
7.45
3.20
-71.
273.
221.
050.
131.
81E+
052.
73E+
05-3
0.01
0.12
-31.
03-5
9.61
-67.
45-1
02.6
1-9
5.99
13-3
6.81
3.20
23-8
1.92
10.1
6-8
4.89
14.8
00.
840.
082.
61E+
051.
16E+
06-3
0.92
0.61
-34.
61-6
3.19
-81.
92-1
07.8
1-1
01.1
916
-50.
1410
.17
24-9
1.14
7.31
-151
.79
140.
041.
230.
121.
29E+
064.
90E+
06-3
4.88
1.35
-38.
19-6
6.77
-91.
14-1
13.0
1-1
06.3
918
-55.
317.
43
25-9
5.81
7.29
-418
.58
656.
501.
060.
204.
47E+
052.
08E+
07-3
2.25
1.91
-41.
77-7
0.36
-95.
81-1
18.2
1-1
11.5
919
-62.
507.
54
Tabl
e4:
Vario
ussu
rfact
ant
syst
ems
at25
℃,c
once
ntra
tion
ofth
esu
rfact
ant
inth
ece
ll-0
.33
mM
:dep
ende
nce
ofth
erm
odyn
amic
para
met
ers
onal
ipha
tic
chai
nle
ngth
m.
m∆
He
xp,
kJ/m
ol
Erro
r,
±kJ
/mol
∆H
fit
,
kJ/m
ol
Erro
r,
±kJ
/mol
Nf
itEr
ror,
±K
fit
,
M−
1
Aa
lkC
Aio
nB
,
M−
1
∆a
ggG
,kJ/
mol
∆a
ggH
,kJ/
mol
CH
2
T∆
ag
gS
obs,
kJ/m
ol
Erro
r,
±kJ
/mol
fitEr
r.ca
lcco
nc
alk c
on
cfit
calc
alk
20-1
1.43
91.4
0-3
7.78
–1.
02–
6.80
E+04
3.03
E+04
-27.
58–
-25.
58-5
4.17
-11.
43-9
2.21
-85.
593
15.8
8264
–
21-6
5.76
2.14
-64.
7115
.09
1.32
–1.
65E+
051.
29E+
05-2
9.79
4.43
-29.
16-5
7.75
-65.
76-9
7.41
-90.
7913
-35.
365
4.91
5427
22-6
8.90
2.08
-70.
853.
741.
01–
3.48
E+05
5.45
E+05
-31.
630.
28-3
2.75
-61.
33-6
8.90
-102
.61
-95.
9914
-36.
6448
2.09
518
Tabl
e5:
Vario
ussu
rfact
ant
syst
ems
at25
℃,c
once
ntra
tion
ofth
esu
rfact
ant
inth
ece
ll-0
.66
mM
:dep
ende
nce
ofth
erm
odyn
amic
para
met
ers
onal
ipha
tic
chai
nle
ngth
m.
66
Povilas Norvaišas Supplementary material
m
∆Hexp ∆Gfit
0.33 mM 0.66 mM 0.33 mM 0.66 mM
W p W p W p W p
18 – – – –
19 – – – –
20 0.96 0.81 – – – – – –
21 0.85 0.15 0.98 0.95 0.77 0.04 0.77 0.04
22 0.88 0.27 0.78 0.08 0.93 0.56 0.93 0.56
23 0.78 0.05 0.85 0.22
24 0.82 0.06 0.90 0.33
25 0.93 0.55 0.83 0.11
Table 6: The normality of thermodynamic parameters for different lengths of aliphatic chain m,
according to Shapiro-Wilk (W) test
t, ℃ No. of exp. Avg. ∆Hexp, kJ/mol Error, ± kJ/mol
25 4 -65.25 4.19
37 5 -83.02 2.88
52 2 -26.82 1.75
69 2 -32.66 2.24
Table 7: Analysis of enthalpies of dodecylammonium binding to decane sulfonate at various tem-
peratures.
67