OPEN

ORIGINAL ARTICLE

Evidence for quorum sensing and differentialmetabolite production by a marine bacterium inresponse to DMSP

Winifred M Johnson1, Melissa C Kido Soule2 and Elizabeth B Kujawinski21MIT-WHOI Joint Program in Oceanography/Applied Ocean Science and Engineering, Department of MarineChemistry and Geochemistry, Woods Hole Oceanographic Institution, Woods Hole, MA, USA and2Department of Marine Chemistry and Geochemistry, Woods Hole Oceanographic Institution, Woods Hole,MA, USA

Microbes, the foundation of the marine foodweb, do not function in isolation, but rather rely onmolecular level interactions among species to thrive. Although certain types of interactions betweenautotrophic and heterotrophic microorganisms have been well documented, the role of specificorganic molecules in regulating inter-species relationships and supporting growth are only beginningto be understood. Here, we examine one such interaction by characterizing the metabolic response ofa heterotrophic marine bacterium, Ruegeria pomeroyi DSS-3, to growth on dimethylsulfoniopropio-nate (DMSP), an abundant organosulfur metabolite produced by phytoplankton. When cultivated onDMSP, R. pomeroyi synthesized a quorum-sensing molecule, N-(3-oxotetradecanoyl)-L-homoserinelactone, at significantly higher levels than during growth on propionate. Concomitant with theproduction of a quorum-sensing molecule, we observed differential production of intra- andextracellular metabolites including glutamine, vitamin B2 and biosynthetic intermediates of cyclicamino acids. Our metabolomics data indicate that R. pomeroyi changes regulation of itsbiochemical pathways in a manner that is adaptive for a cooperative lifestyle in the presence ofDMSP, in anticipation of phytoplankton-derived nutrients and higher microbial density. Thisbehavior is likely to occur on sinking marine particles, indicating that this response may impact thefate of organic matter.The ISME Journal (2016) 10, 2304–2316; doi:10.1038/ismej.2016.6; published online 19 February 2016

Introduction

The carbon cycle is essential to the regulation ofconditions on Earth from the climate to the pHbuffering capacity of the ocean. Atmospheric carbonstores (750 Pg) are dwarfed by the 31 000 Pg ofdissolved carbon present in the ocean, of which660 Pg comprise dissolved organic matter. Thiscomplex pool of organic molecules contains sub-strates that serve as the primary catabolic energysource for heterotrophic bacteria, as well as com-pounds that are transferred among microbes topromote growth or cooperative community function.Because microbes have varying requirements fororganic substrates, organic matter compositioninfluences the microbial community and vice versa

(Baña et al., 2014; Sharma et al., 2014). Somemicrobes are metabolically specialized, whereasother microbes are generalists that can adapt to avariety of conditions (Lauro et al., 2009). Recentstudies have documented complex relationshipsbetween primary producers and heterotrophicbacteria (for example, Amin et al., 2015; Durhamet al., 2015). Experiments with the diatomThalassiosira pseudonana, showed increased intra-cellular production of amino acids in response toco-culturing with the heterotroph Dinoroseobactershibae (Paul et al., 2012). In another study,the Roseobacter Phaeobacter gallaeciensis wasfound to produce phytotoxins in response top-coumaric acid, an algal cell wall breakdownproduct (Seyedsayamdost et al., 2011). More gen-erally, vitamins such as cobalamin, biotin andthiamin are produced by some heterotrophic bacteriaand assimilated by eukaryotic organisms that areauxotrophic for these essential biochemicals (Croftet al., 2006). Likewise, organosulfur metabolites arerequired by some marine bacteria, such as theubiquitous SAR11, which cannot reduce inorganicsulfate (Tripp et al., 2008). Finally, some molecules,

Correspondence: WM Johnson, MIT/WHOI Joint Program inOceanography/Applied Ocean Sciences and Engineering,Department of Marine Chemistry and Geochemistry, Woods HoleOceanographic Institution, 360 Woods Hole Road, MS#4, WoodsHole, MA 02543, USA.E-mail: wjohnson@whoi.eduReceived 14 October 2015; revised 17 December 2015; accepted24 December 2015; published online 19 February 2016

The ISME Journal (2016) 10, 2304–2316© 2016 International Society for Microbial Ecology All rights reserved 1751-7362/16

www.nature.com/ismej

brought to you by COREView metadata, citation and similar papers at core.ac.uk

provided by Woods Hole Open Access Server

including algal break-down products and acylhomoserine lactones, act as infochemicals to signalevents such as a nutrient pulse or populationchanges (Sandy and Butler, 2009; Seyedsayamdostet al., 2011). Although these interactions betweenmicrobes and molecules occur on very small spatialscales, they culminate in the production and remi-neralization of 50–100 Pg of carbon annually (Martinet al., 1987).

DMSP is a widely recognized intermediate in thesetypes of interactions. It is produced predominantlyby dinoflagellates, coccolithophores, diatoms andhigher plants like the marsh grass Spartina (Stefels,2000). Intracellularly, it is present at high concentra-tions ranging from 0.1 to 1 M (Stefels, 2000) and bulkparticulate concentrations can reach 120 nM (Kieneand Slezak, 2006), making it a quantitatively sig-nificant organic substrate in the ocean (Simó et al.,2009). Not surprisingly, it has been demonstrated tohave a vital role in the marine food web as a sourceof reduced carbon and sulfur (González et al., 2003;Simó et al., 2009) and may act as an antioxidant insome organisms (Sunda et al., 2002).

In the open ocean, dissolved DMSP concentrationsare generally found in the sub-nanomolar range butcan reach low nanomolar levels around a phyto-plankton bloom and are likely elevated in theimmediate wake of sinking aggregates of dyingphytoplankton cells and/or fecal pellets (Yoch,2002; Kiene and Slezak, 2006; Levine et al., 2012).Consequently, DMSP is associated with the release oflabile organic matter (Biddanda and Benner, 1997;Grossart, 2010; Stocker, 2012). Thus, heterotrophicorganisms seeking hotspots of organic nutrientscould reasonably use DMSP as a proxy for other,more complex, molecules. Indeed, previous workhas shown that increased colonization occurs onDMSP-spiked agar particles relative to non-amendedagar particles (Kiørboe et al., 2002). Along theselines, DMSP acts as a chemo-attractant for organismsranging from heterotrophic bacteria to herbivorousdinoflagellates (Seymour et al., 2010). In one study,the presence of DMSP triggered the bacterial produc-tion of a plant growth hormone to promote growth ofa phytoplankton species and thus augment organicmatter production (Seyedsayamdost et al., 2011).

Bacteria with relatively large genomes can takeadvantage of transient nutrient pulses and thus couldbe well-suited to respond to DMSP-derived signals.To examine the metabolic response of a hetero-trophic bacterium to DMSP, we focus here on themetabolically flexible bacterium, Ruegeria pomeroyiDSS-3, a coastal isolate of the Roseobacter cladeof marine α-proteobacteria (Moran et al., 2004).R. pomeroyi possesses genes for a variety of metabolicstrategies including lithoheterotrophy using carbonmonoxide or sulfide, flagellar motility and transpor-ters for algal osmolytes and carboxylic acids(Moran et al., 2004). These attributes, which havebeen confirmed in culture, enable R. pomeroyi toexploit rapid chemical changes in its environment.

In particular, it can catabolize DMSP via twodifferent pathways and incorporate atoms derivedfrom DMSP into its biomass (González et al., 2003;Reisch et al., 2013). There are also indications thatDMSP may provide a signal to prime R. pomeroyi’smetabolism to utilize algal lysates and exudatesthrough increased transcription of transporter genesas well as a potential quorum-sensing response(Bürgmann et al., 2007). Thus, R. pomeroyi is likelyto have the metabolic capability to respond to DMSPas an indicator of a shift in nutrient availability.

We examined R. pomeroyi’s relationship withDMSP by comparing its metabolic response to DMSPas a carbon source relative to propionate, a three-carbon carboxylic acid that lacks the dimethylsulfidefunctional group. If DMSP serves only as an organicsubstrate for R. pomeroyi, then the primary meta-bolic difference in our experiment should be aresponse to the reduced sulfur in DMSP (Reischet al., 2013) because propionate catabolism sharesthe last two steps of one of the DMSP catabolicpathways. Alternatively, if DMSP indeed functionsas a signal to R. pomeroyi of changing environmentalconditions, we would anticipate a suite of metabolicchanges that facilitate adaptation to a DMSP-containing microzone.

The microzones in the ocean where DMSP isprevalent have several characteristics in common.Specifically, these microzones have higher nutrientavailability compared with the surrounding watercolumn, particularly organic substrates, which canprovide nitrogen and phosphorus, and they host adense microbial community whose members com-pete for, and exchange, valuable substrates andnutrients (Biddanda and Benner, 1997; Grossart,2010). In response to these characteristics, R. pomeroyiand similar metabolically flexible bacteria couldchange their metabolism to take advantage of the shiftin chemical regime. Here we examine three hypothesesusing intracellular and extracellular metabolic profiles.First, higher cell densities facilitate cell–cell commu-nication (or quorum sensing) and resultant coordinatedbehaviors, and thus we expect to resolve and identifychemical signaling molecules. Second, the mostenergetically favorable means of obtaining nutrients,particularly nitrogen, would be to acquire themdirectly from the labile organic matter ubiquitous inthe microzones (Kirchman et al., 1989, 2003; Vallinoet al., 1996). Bürgmann et al. (2007) demonstrated thatR. pomeroyi increases transcription of genes associatedwith transporters for nitrogen-containing organic sub-strates when grown on DMSP. We anticipate that achange in nutrient acquisition strategy would alsoresult in differences in the concentrations of regulatorymetabolites. Last, we suggest that a mutually beneficialexchange of high-value metabolites with otherbacteria or phytoplankton would occur (Aminet al., 2015; Durham et al., 2015), leading to anadaptive and cooperative lifestyle similar to thatobserved during bacterial colonization of organicparticles (Grossart et al., 2003).

Differential metabolites in response to DMSPWM Johnson et al

2305

The ISME Journal

Our study examines the impact of DMSP relativeto propionate on metabolite production and organicmatter excretion by R. pomeroyi. We used comple-mentary untargeted and targeted metabolomicsmethods to analyze the suite of molecules producedby R. pomeroyi, both within the cells and in theexternal medium. These metabolites ranged fromthose associated with primary metabolism, whichshed light on fundamental cellular processes, tothose associated with secondary metabolism, whichhelp us understand the cell’s response to shiftingenvironmental conditions.

Materials and methods

The complete methods are given in the SupplementaryInformation. Briefly, R. pomeroyi DSS-3 [DSM 15171]was grown in the dark at 23 °C in 0.2 μm-filtered(Omnipore (PTFE), EMD Millipore, Billerica, MA,USA) seawater amended with 3mM carbon (1mM

sodium propionate or 0.6mM DMSP), 4mM ammo-nium chloride, 30 nM monosodium phosphate, 100 nM

ferrous chloride-ethylenediaminetetraacetic acid,100 nM zinc chloride, 100 nM manganese(II) chloride,1 nM cobalt(II) chloride and 1ml l−1 medium of f/2vitamin solution. Duplicate cultures were killed at 0,32, 38(propionate)/43(DMSP), 48, 60 and 72 h withcell counts in the range of 106–107 cellsml−1

(Supplementary Figure S1a). The third time pointwas offset between the propionate and DMSP treat-ments so that the same cell abundance was sampled ineach treatment. Cells were captured on a 0.2 μm filter(Omnipore (PTFE), EMD Millipore) using a combustedglass vacuum filtration apparatus (vacuum nevergreater than 10mmHg). Filters were frozen in cryo-vials at −80 °C immediately after filtration. Intracellu-lar metabolites were extracted from filters using cold40:40:20 acetonitrile:methanol:water+0.1 M formic acidsolution, adapted from previously published protocols(Rabinowitz and Kimball, 2007). Half of the extract wasfurther extracted using Agilent Bond Elut PPL (styrene-divinylbenzene polymer) solid-phase extraction resinfor untargeted analysis. The extracellular metaboliteswere extracted from acidified filtrate (pH 2–3) usingPPL cartridges (Dittmar et al., 2008; after modificationby Longnecker, 2015).

All the samples were separated with the sameliquid chromatography method on a C18 reversed-phase column with a water and acetonitrile gradient(Kido Soule et al., 2015). The targeted method used aThermo Scientific triple quadrupole mass spectro-meter with a heated electrospray ionization source.Untargeted analyses were conducted using a hybridlinear ion trap—7T Fourier transform ion cyclotronresonance mass spectrometer (FT-ICR-MS; LTQ FTUltra, Thermo Scientific) with an electrosprayionization source.

The targeted data were processed using Xcalibur(Thermo Scientific software) and a 5 to 7-pointstandard curve for relative quantification. The

untargeted mass analysis files were converted fromThermo RAW files to mzML files using MSConvert(Chambers et al., 2012). The resulting data wereprocessed using XCMS (Smith et al., 2006;Tautenhahn et al., 2008; Benton et al., 2010) andCAMERA (Kuhl et al., 2012). Peak picking andalignment were carried out separately for theintracellular and extracellular samples generating alist of features (defined as a unique combination ofmass-to-charge value (m/z) and retention time) foreach sample set (Longnecker et al., 2015). Featurelists were refined through quality control require-ments and are reported as mass spectral peak areanormalized to cell abundance (intra- and extracel-lular; peak area per cell) or volume (extracellular;peak area per liter). Features in the untargeted datawere putatively identified by comparison of m/z tothe METLIN database (Smith et al., 2005), and ifavailable, by comparison of the associated fragmen-tation spectrum with experimentally generated frag-mentation patterns in the METLIN database or to insilico-derived fragmentation patterns from MetFrag(Wolf et al., 2010). When possible, commerciallyavailable analogs for putatively identified metabo-lites were analyzed using identical protocols.

Differences in the concentration of individualmetabolites between the substrate treatments werecompared with analysis of variance. In the intracel-lular targeted data, P-values o0.0013 were consid-ered significant due to the Bonferroni correction formultiple comparisons.

Total organic carbon (TOC; unfiltered) anddissolved organic carbon (DOC; 0.2-μm filtered)samples were acidified and then analyzed using aShimadzu TOC-VCSH Total Organic Carbon Analyzer,according to standard practices. Methanethiol anddimethylsulfide were measured using gas chromato-graphy (adapted from Levine et al. (2012)).

Results

R. pomeroyi growth parametersR. pomeroyi cells exhibited similar growth rates(0.1 h− 1) on both substrates (DMSP or propionate),and remineralization of the organic carbon substratewas confirmed by the diminishing concentrations ofDOC and TOC in the media (Supplementary FigureS1b). Methanethiol and dimethylsulfide (DMS),which are products of two different DMSP degrada-tion pathways, were produced during growth onDMSP (Supplementary Figure S1c; Reisch et al.,2011, 2013). In this study, we did not quantify thetwo gases as we only sought confirmation that bothpathways were active.

Quantifying the metabolome (targeted metabolites)In targeted metabolomics, the goal is to obtainrelative concentrations of a pre-defined list oforganic compounds. Most of the intracellular and

Differential metabolites in response to DMSPWM Johnson et al

2306

The ISME Journal

extracellular metabolites quantified by our methodshowed no significant difference between thesubstrate treatments. For example, intracellularmethionine concentrations were similar in bothtreatments (Figure 1d). The measured compoundsin the targeted method include vitamins, aminoacids, nucleotides and precursors to these molecules,implying that primary cellular processes remainedsimilar regardless of substrate (see SupplementaryFigure S2 for the list of intracellular metabolitesquantified). However, within the targeted dataencompassing both intra and extracellular samples,significant changes were observed for one of theamino acids (glutamine), several vitamins and inter-mediates in cellular nitrogen and sulfur cycling.Intracellular glutamine concentrations by time pointranged from 3.5 to 157 times lower during growth onDMSP relative to growth on propionate (Figure 1a).This varied by time in the growth curve and theconcentrations were particularly low in the DMSPtreatment. Conversely, the B vitamins thiaminmonophosphate (phosphorylated vitamin B1) andriboflavin (vitamin B2) were both elevated within thecell at some time points in the DMSP treatment(Figures 1b and c, respectively). Extracellularly, themaximum concentration of riboflavin was almosteight times higher in the DMSP treatment (Figure 2a).

The intracellular concentrations of 5′-deoxy-5′-methylthioadenosine (MTA), a sulfur-containing

metabolite, were not significantly different betweenthe two treatments (Figure 3b). However, in theextracellular fraction, MTA concentrations wereapproximately three times higher in the DMSPtreatment than in the propionate treatment(Figure 3c), indicating increased production andrelease of MTA into the medium during DMSP-based growth. Given its sulfur content, the elevatedMTA concentration could be derived from excesssulfur generated during DMSP catabolism. We testedthis hypothesis with an experiment that monitoredMTA concentrations in three treatments: 100%propionate, 90% propionate/10% DMSP and 100%DMSP, where all the treatments contained the sameamount of dissolved organic carbon (3mM). Whennormalized to cell abundance, the extracellular MTAconcentrations in the full DMSP and the 10%DMSP treatments were seven and six times higher,respectively, than in the propionate treatment(Supplementary Figure S3a). Therefore, despite the90% reduction in sulfur, there was not a correspond-ing decrease in cell-normalized MTA concentration.

Profiling the metabolome (untargeted metabolites)The goal of the untargeted method is to profile asmuch of the metabolome as possible. Unlike thetargeted method it is only semi-quantitative, andrelative abundances cannot be converted into molar

0 12 24 36 48 60 720

0.01

0.02

0.03

0.04

0.05

Time (hrs)

atto

mol

es/c

ell

0 12 24 36 48 60 720

0.01

0.02

0.03

0.04

0.05

Time (hrs)

atto

mol

es/c

ell

0 12 24 36 48 60 720

0.2

0.4

0.6

0.8

1

Time (hrs)

atto

mol

es/c

ell

0 12 24 36 48 60 720

0.05

0.1

0.15

0.2

Time (hrs)

atto

mol

es/c

ell

p = 0.0004

p = 0.0009

p = 0.000005glutamine thiamin monophosphate

riboflavin methionine

Figure 1 Intracellular metabolite profiles indicate differential concentrations of glutamine and two B vitamins while methionineconcentrations are the same. All metabolites presented here were measured in the targeted method: (a) glutamine, (b) thiaminmonophosphate, (c) riboflavin, (d) methionine. Each line reflects the average of two biological replicates, shown in open and closedsymbols: red triangles—DMSP treatment; blue squares—propionate treatment. The P-value from an analysis of variance test is reported formetabolites with a significant difference in concentration between the substrate treatments across all time points.

Differential metabolites in response to DMSPWM Johnson et al

2307

The ISME Journal

0 12 24 36 48 60 720

1

2

3

4

5x 10

7

Time (hrs)

Pea

k A

rea/

L

0 12 24 36 48 60 720

2

4

6

8

10

12

Time (hrs)

nM

riboflavin

3-(methylthio)propionate

0 12 24 36 48 60 720

2

4

6

8

10

12

14x 10

6

Time (hrs)

Pea

k A

rea/

L

α-ribazole

0 12 24 36 48 60 720

0.5

1

1.5

2

2.5x 10

7

Time (hrs)

Pea

k A

rea/

L

0 12 24 36 48 60 720

1

2

3

4

5

6

7x 10

7

Time (hrs)

Pea

k A

rea/

L

0 12 24 36 48 60 720

2

4

6

8

x 106

Time (hrs)

Pea

k A

rea/

L

3-dehydroshikimate shikimate

chorismate

0 12 24 36 48 60 720

0.5

1

1.5

2

2.5

3x 10

6

Time (hrs)

Pea

k A

rea/

L

phenylpyruvate

Figure 2 Energetically high-value metabolites are found in the medium at higher concentrations in the DMSP treatment, along with ametabolite associated with DMSP degradation. Each subplot shows the extracellular concentration profile of a metabolite over time. Ineach subplot, the line reflects the average of two biological replicates, shown in open and closed symbols: red triangles—DMSP treatment;blue squares—propionate treatment. (a) Riboflavin, measured in the targeted method and reported in nM. The remaining metabolites weremeasured in the untargeted method and the concentration is reported as peak area per liter. (b) α-Ribazole, (c) 3-dehydroshikimate, (d)shikimate, (e) chorismate*, (f) phenylpyruvate*, (g) 3-(methylthio)propionate* (an asterisk (*) denotes putative metabolite identificationswith an identification ranking of 3 (see Supplementary Table S3).).

Differential metabolites in response to DMSPWM Johnson et al

2308

The ISME Journal

quantities. However, metabolites measured in boththe targeted and untargeted methods showhighly similar instrument responses between thetwo methods. This is shown here for MTA(Supplementary Figure S3b) and previous work hasdemonstrated this for other metabolites (Fiore et al.,2015). This confirms that relative abundances offeatures (that is, molecules defined by a uniquecombination of m/z value and retention time) withinthe untargeted data set can be compared andstatistically evaluated.

Hundreds of intracellular features wereresolved with the untargeted metabolomics method(Supplementary Table S1). Most of these (43%,negative ion mode; 96%, positive ion mode)were detected in both substrate treatments buttheir relative abundances illustrate more strikingdifferences between the treatments (SupplementaryTable S2). At 60 h, a majority (86%, negative ionmode; 64%, positive ion mode) of the intracellularfeatures were five times higher in the propionatetreatment. In contrast, a minority (9%, negative ionmode; 17%, positive ion mode) of the intracellularfeatures were at least five times higher in the DMSPtreatment relative to the propionate treatment(Supplementary Table S2).

Hundreds of extracellular features were similarlyobserved in these samples (Supplementary Table S1).

A non-metric multidimensional scaling analysis ofthe extracellular metabolic profiles shows thatsamples cluster more by cell density and/or growthphase than by substrate treatment (SupplementaryFigure S4). Carbon substrate is of secondary impor-tance but has a discernible impact, particularly atlater time points. A comparison of peak area per cellof the individual extracellular features at 60 h showsthat, depending on substrate and ionization mode,between 19 and 36% of these features were at leastfive times higher in one of the substrate treatments(Supplementary Table S2).

Identification of features in the untargeted data setThe structural diversity of metabolites and the gapsin the characterization of metabolic pathways pre-sent major challenges to the identification of featuresin the untargeted data set. Our approach was toidentify as many features as possible within theranking system proposed by Sumner et al. (2007) anddescribed in Longnecker et al. (2015). Metabolitesare identified with increasing confidence first by anm/zmatch, then a fragmentation match to a database,and, for the highest confidence level, comparison toan authentic standard analyzed with our method toconfirm m/z value, retention time and fragmentationpattern. In some cases, putative identifications of

0 12 24 36 48 60 720

2

4

6

8

10

12x 105

Time (hrs)

Pea

k A

rea/

L

0 12 24 36 48 60 720

1

2

3

4

Time (hrs)

nM

0 12 24 36 48 60 720

0.5

1

1.5

2

2.5

3

3.5 x 10−3

Time (hrs)

atto

mol

es/c

ell

Intracellular MTA

Extracellular MTA

N-(3-oxotetradecanoyl)-L-homoserine lactone

0 12 24 36 48 60 72108

107

108

Time (hrs)

Cel

ls/m

L

DMSPPropionate

Cell count

Figure 3 MTA linked to AHL production. (a) Growth curve for R. pomeroyi. Red solid line is the DMSP treatment and the blue dashedline is the propionate treatment. Error bars represent one standard deviation of replicate bottles. In the remaining subplots, the line reflectsthe average of two biological replicates, shown in open and closed symbols: red triangles—DMSP treatment; blue squares—propionatetreatment. (b) The intracellular concentrations of MTA over time, normalized to the number of cells. (c) The extracellular MTAconcentrations over time. (d) Extracellular concentrations of the quorum-sensing compound N-(3-oxotetradecanoyl)-L-homoserine lactone.The structure of N-(3-oxotetradecanoyl)-L-homoserine lactone is shown in the inset of d.

Differential metabolites in response to DMSPWM Johnson et al

2309

The ISME Journal

metabolites with only an m/z match can be con-sidered more significant if intermediates in the samemetabolic pathway are identified at higher rankings(Suhre and Schmitt-Kopplin, 2008). These stringentrequirements result in a relatively small subset ofmetabolites that we can reliably identify. However,we can expect to be able to identify a larger set ofmetabolites in future studies as the field of metabo-lomics matures; these additional identifications willadd greater nuance to our understanding of meta-bolic responses and their ecological implications.

We focused our most rigorous identificationefforts, including purchasing standards, on featuresthat were differentially produced in the twotreatments. MTA was quantified by our targetedmethod and found to be elevated in the extracellularfraction of the DMSP growth treatment; it wassimilarly detected in the untargeted method(Supplementary Figure S3b). MTA can be producedby two different pathways, as a by-product ofspermidine/spermine synthesis or acyl homoserinelactone (AHL) biosynthesis (Parveen and Cornell,2011; Supplementary Figure S5). Structurally, AHLsconsist of a lactone head-group linked by an amide to

a carbon chain of varying lengths and degrees ofsaturation (Figure 3d, inset). AHLs are a class ofquorum-sensing chemicals and R. pomeroyi isknown to produce three different AHL compounds(Moran et al., 2004). We searched the untargeted dataset for predicted m/z values for all AHLs up toa 16-carbon chain length. We identified N-(3-oxote-tradecanoyl)-L-homoserine lactone in our data set,based on exact mass and fragmentation patternmatches in the METLIN database (Smith et al.,2005). Its size was consistent with an AHL observedby Moran et al. (2004). Comparison with anauthentic standard revealed a positive identificationfor this compound (Figure 4). The identified AHL iselevated in the extracellular medium by a factorof 8–15 in the DMSP treatment at the last four timepoints of the experiment (Figure 3d), where MTAconcentrations are also high.

We also putatively identified precursors to thecyclic amino acids (tryptophan, phenylalanineand tyrosine) including chorismate, shikimate, phe-nylpyruvate and their associated intermediates(Supplementary Table S3; Supplementary Figure S6).These features occur at higher relative concentrations

0

20

40

60

80

100 23.7

22 24 26 28 30 320

20

40

60

80

100 23.4

0

20

40

60

80

100 225.3

102.3

270.3

50 100 150 200 250 300 350 4000

20

40

60

80

100 225.3

102.1

270.3

Rel

ativ

e A

bund

ance

Rel

ativ

e A

bund

ance

m/zTime (min)

O

O

HN

O O

O

O

NH

O O225.2

102.0

[M+H]+ = 326.2326

1 µg/mL Standardin Sample Matrix

DMSP TreatmentTime Point 5

1 µg/mL Standardin Sample Matrix

DMSP TreatmentTime Point 5

Figure 4 AHL identification. N-(3-Oxotetradecanoyl)-L-homoserine lactone commercial standard was spiked into original extracellularextract from the experiment approximately 6 months after the experiment. (a) Extracted ion chromatograms displaying the mass rangewithin 1 Dalton of the mass of N-(3-oxotetradecanoyl)-L-homoserine lactone. (b) The same data in the original experimental sample.The difference in retention time was confirmed to be owing to changes in the column over the 6-month lag between analyses.N-(3-Oxotetradecanoyl)-L-homoserine lactone fragments into a lactone head group (102.0m/z) and a tail group fragment (225.2m/z).(c) Fragmentation of the commercial standard. (d) Fragmentation of the feature identified in the experiment. (e) The structure ofN-(3-oxotetradecanoyl)-L-homoserine lactone and its primary fragments.

Differential metabolites in response to DMSPWM Johnson et al

2310

The ISME Journal

in the DMSP treatment (Figure 2c–f). Other putativeidentifications include the first metabolite in theDMSP demethylation pathway, 3-(methylthio)propionate, which was exclusively observed inthe DMSP treatment in both extracellular andintracellular fractions (Figure 2g; SupplementaryTable S3 and S4). In addition, in the untargeteddata, α-ribazole, a component of vitamin B12,was putatively identified and found to accumulatein the media over time (Figure 2b; SupplementaryTable S3). While the current annotations ofR. pomeroyi’s genome do not predict directsynthesis of α-ribazole, its production may resultfrom an un-annotated biosynthetic pathway ornon-enzymatic degradation of vitamin B12.

Discussion

The primary goal of this study was to examinethe difference between the two growth substrates(that is, DMSP and propionate) beyond the obviousdifferences in direct degradation of the carbonsubstrate. We hypothesized that these differenceswould be observed either in sulfur-containingmetabolites (Thiel et al., 2010) or in metabolitesrelated to a re-tooling of metabolism for amore cooperative lifestyle (Geng et al., 2008;Seyedsayamdost et al., 2011). While a few sulfur-containing metabolites, such as MTA, had ahigher abundance in the DMSP treatment, we didnot find any indication of far-reaching effects onsulfur-containing metabolites under DMSP-growth.For instance, an essential metabolite such asmethionine, a sulfur-containing amino acid, hadequivalent concentrations in the two treatments(Figure 1d) suggesting its intracellular concentra-tion is highly regulated and insensitive to theinflux of sulfur. Thus, rather than an impacton only S-containing molecules, we observeda broad shift in metabolite patterns, suggesting apervasive impact of DMSP on the R. pomeroyimetabolome.

DMSP degradationThe fate of DMSP during degradation by R. pomeroyihas been extensively studied (Reisch et al., 2013) andthe concentrations of metabolites within the DMSPdegradation pathways are consistent with previousgenomics research. On the basis of the metabolitesmeasured, both the DMSP demethylation andthe cleavage pathways were active in the DMSPtreatment (Figure 2g; Supplementary Table S4;Supplementary Figure S1c). After the production ofthese molecules, the atoms derived from DMSP areincorporated into intermediates within central car-bon metabolism and are impossible to discernrelative to propionate metabolism (without isotopelabeling).

MTA as a proxy for possible quorum sensing inR. pomeroyiMTA, which accumulates in the DMSP treatment, isan intermediate in the methionine salvage pathway,which allows the cell to recycle sulfur atoms formethionine synthesis. R. pomeroyi lacks the genesthat complete the classic methionine salvage path-way, suggesting that excess MTA produced byR. pomeroyi would be excreted rather than recycled.MTA is a by-product of AHL synthesis and ofspermidine/spermine synthesis (SupplementaryFigure S5; Parveen and Cornell, 2011). R. pomeroyi’sgenome suggests that it can synthesize spermidineand spermine, but that this occurs through analternative pathway that does not produce MTA.Therefore, of the characterized pathways that gen-erate MTA as a by-product, the R. pomeroyi genomeis only known to contain genes required for AHLsynthesis (Supplementary Figure S5). Furthermore,Bürgmann et al. (2007) have demonstrated increasedtranscription of a gene annotated as a ‘transcriptionalregulator, LuxR family’ (luxR is the AHL receptorgene) in response to DMSP. Although luxR andluxI (the AHL synthase gene) did not showincreased transcription in the Bürgmann et al.(2007) study, the regulation of these systems iscomplex, as shown in species closely related toR. pomeroyi (Zan et al., 2014), and might not bevisible in the transcriptome for a variety of reasons.On the basis of this genomic information, MTA wasmost likely found at higher concentrations in theDMSP treatment owing to increased AHL synthesis.This hypothesis was supported in the untargeteddata by the detection of one AHL, N-(3-oxotetrade-canoyl)-L-homoserine lactone. The coherence ofMTA and AHL dynamics suggests that the elevatedMTA concentrations are linked to increased synth-esis of N-(3-oxotetradecanoyl)-L-homoserine lactoneand thus the presence of a possible quorum-sensingresponse by R. pomeroyi to elevated DMSP concen-trations. A linear decrease in MTA concentrationwas not observed between a culture grown on 100%DMSP and one grown on 10% DMSP/90% propio-nate, further implying that the elevated accumula-tion of MTA in the DMSP treatment is not simplyowing to the increased reduced sulfur pool, butinstead is the result of a regulatory responsetriggered by DMSP.

Phenotypes regulated by a quorum-sensing systemgenerally depend on a high density of cells to beeffective. Quorum sensing can regulate activitiessuch as antibiotic production, exoenzyme synthesisor biofilm formation (Miller and Bassler, 2001).R. pomeroyi can produce exoenzymes to break downlarge biopolymers (Christie-Oleza et al., 2015) andhas genes for flagellar motility, which can beregulated by a quorum-sensing system (Moranet al., 2004). The microzones with likely elevatedDMSP concentrations generally have high microbialpopulation densities. With a population-dependentcell–cell communication pathway, such as AHL-based

Differential metabolites in response to DMSPWM Johnson et al

2311

The ISME Journal

quorum sensing, R. pomeroyi could coordinate ashift in its metabolism to exploit the increase inorganic nutrient availability. This hypothesis isconsistent with previous characterization of aquorum-sensing phenotype of Ruegeria sp. KLH11,a marine sponge symbiont (Zan et al., 2012), andwith recent findings of measurable quantities of anumber of AHLs in association with sinking marineparticles (Hmelo et al., 2011). Further work will beneeded to understand the specificity of R. pomeroyi’sresponse to DMSP including experiments thatevaluate R. pomeroyi metabolic (and transcrip-tomic/proteomic) profiles under exposure to theAHL identified here, as well as experiments withluxR/luxI knock-out mutants to isolate the affectedmetabolic pathways.

Glutamine as a proxy for altered nitrogen assimilationstrategyIt is particularly striking that intracellular glutamineconcentrations were relatively reduced in the DMSPtreatment because it is a metabolite that is highlysensitive to nitrogen availability and is integral to acell’s nitrogen cycle (Brauer et al., 2006). Glutaminecan serve as a nitrogen storage compound because itis produced when ammonium is assimilated into thecell. Once present, glutamine is an essential pre-cursor for the synthesis of other amino acids anddonates an amino group during the synthesis ofpurines, pyrimidines, amino sugars and NAD+

(Reitzer, 2003). High glutamine concentrations indi-cate that sufficient ammonium concentrations arepresent. In contrast, low glutamine concentrationssignal a decline in ammonium concentrations andresult in the upregulation of alternative nitrogenassimilation pathways. This has been observed notonly in E. coli, but also in organisms such asSaccharomyces cerevisiae (Brauer et al., 2006). Inour experiments, excess ammonium was present inthe growth media in both treatments, and thusinorganic nitrogen availability can be discounted asa controlling factor of intracellular glutamineconcentrations.

Through a cascade of signals, glutamine alsoregulates alternative forms of nitrogen acquisition.In E. coli, for example, low glutamine concentrationsare correlated with increased transport of organicnitrogen substrates such as amino acids, peptidesand polyamines into the cell to offset the decline inextracellular ammonium (Zimmer et al., 2000;Chubukov et al., 2014). While the regulation ofnitrogen assimilation has not been studied inR. pomeroyi, R. pomeroyi’s genome annotationsindicate the presence of genes for both pathways forammonium assimilation (glutamate dehydrogenaseand glutamine synthetase) as well as for the nitrogenregulatory protein P-II and a uridylyltransferase, allof which would be required to regulate nitrogenassimilation in a manner similar to E. coli. Bürgmannet al. (2007) previously observed that R. pomeroyi’s

transcriptional response to DMSP involved upregu-lation of genes specifically associated with aminoacid and polyamine transport. This accumulatedevidence suggests that R. pomeroyi altered itsnitrogen assimilation strategy in response to thepresence of DMSP. Specifically, we hypothesizethat R. pomeroyi is regulating intracellular gluta-mine concentrations to promote organic nitrogenuptake and utilization. Scavenging nitrogen fromorganic sources rather than utilizing ammoniumcould benefit R. pomeroyi by decreasing the cellularenergy expenditure associated with ammoniumassimilation and/or full biosynthesis of nitrogenousmetabolites. To test this further, experiments areneeded that compare the uptake of organic nitrogensources and the differential transcription of nitro-gen assimilation genes, in both the presence andabsence of DMSP.

Production of extracellular metabolites with potentialfor microbial cross-feedingReduced glutamine concentrations can also affectthe diverse biosynthetic pathways that rely on thismetabolite. For example, during nitrogen starvationin E. coli, decreased glutamine concentrations resultin lower tryptophan concentrations and accumula-tion of phenylpyruvate, a precursor of phenylala-nine, owing to decreased de novo synthesis ofphenylalanine from phenylpyruvate (Brauer et al.,2006). Tryptophan shares a common precursor,chorismate, with the other cyclic metabolites phe-nylalanine, tyrosine, 2,3-dihydroxybenzoic acid and4-aminobenzoic acid (Supplementary Figure S6). Inour experiment, intracellular and extracellular tryp-tophan concentrations remain the same in bothtreatments, while its precursors, 3-dehydroshiki-mate, shikimate, chorismate and phenylpyruvate,all accumulate in the extracellular fraction of theDMSP treatment. This suggests that the precursorsfor tryptophan synthesis were not being utilized byR. pomeroyi either owing to decreased synthesis ofthe amino acid or to increased production of theprecursors.

The cellular release of these metabolically high-value biosynthetic intermediates supports the thirdcomponent of our hypothesis, namely that switchingto a cooperative lifestyle triggered by DMSP wouldrequire the release of some metabolites that could beused in cross-feeding. The increased extracellularconcentrations of precursors like chorismate andphenylpyruvate could provide labile intermediatesto surrounding microbes with incomplete aminoacid biosynthetic pathways; these organisms mightin turn produce valuable metabolites that R. pomer-oyi could use. According to the annotatedR. pomeroyi genome, R. pomeroyi can synthesizeall amino acids except asparagine and possiblyhistidine. The analysis of 3062 available bacterialgenomes in the Integrated Microbial Genomesdatabase reveals that biosynthetic capability in

Differential metabolites in response to DMSPWM Johnson et al

2312

The ISME Journal

bacteria ranges from 0 to 20 amino acids, averagingonly 7.9 complete amino acid biosynthetic path-ways. Bacteria are most commonly auxotrophic fortyrosine, phenylalanine, lysine or histidine. In 124eukaryote genomes, complete biosynthetic pathwaysare present, on average, for only 4.1 out of 20 aminoacids with auxotrophy most commonly found forhistidine, lysine, serine or leucine and no speciesable to synthesize more than 14 of the 20 essentialamino acids (Mee and Wang, 2012). The enhancedexcretion of these biosynthetically high-value cyclicintermediates under DMSP-derived growth may belinked in part to changes in nitrogen assimilationstrategy used by R. pomeroyi. However, thesemetabolites would also serve as valuable currencyfor microbial cross-feeding as evidenced by theubiquity of incomplete de novo amino acid synthesispathways.

Other metabolites that could be used in microbialcross-feeding are riboflavin (vitamin B2) and α-ribazole, a component of vitamin B12. Both of thesemolecules accumulated in the extracellular matrix ofthe DMSP treatment (riboflavin to 8X and α-ribazoleto 3X at 72 h; Figure 2a and b). It is reasonable tospeculate that the availability of a precursor such asα-ribazole, as well as the vitamin, riboflavin, wouldbe valuable within a dense microbial consortium.For example, auxotrophy for α-ribazole has beendemonstrated in the bacterium, Listeria innocua,which has a specific transporter to acquire thiscompound from its environment (Gray andEscalante-Semerena, 2010). Although α-ribazole,as well as the cyclic amino acid precursors

mentioned earlier, do not directly provideamino acids and vitamins, they would alloworganisms to synthesize essential metabolites atlower metabolic cost and could lead to precursorauxotrophy (the ‘black queen’ hypothesis (Morriset al., 2012)). Further analysis of marine genomesmay improve our ability to link metaboliteproduction to the demands of different types ofauxotrophic organisms but clearly exchangeof metabolites, as proposed here, would bemutually beneficial to microbes in a high-densityconsortium.

Conclusions

Differences in the phylogenetic diversity offree-living versus surface-associated microbialpopulations have been well-documented (DeLonget al., 1993). However, many bacteria mayswitch between these two lifestyles, and the factorsgoverning this switch are only beginning to beunderstood (Grossart, 2010). Our work expands onthese studies by showing that specific metabolites,such as algal-derived DMSP, can trigger metabolicshifts that could be expressed under high populationdensities (Figure 5). If groups of bacteria areusing some phytoplankton-derived metabolites assignals (Seyedsayamdost et al., 2011), rates ofparticle degradation, and more broadly of carbonremineralization, may be intimately tied to particlecomposition and source. These results emphasizethe need to understand the molecular-level

Figure 5 Visual interpretation of R. pomeroyi’s response to DMSP. The left-hand panel portrays the diffuse oligotrophic ocean populatedby free-living organisms. In contrast, the right-hand panel shows dying plankton cells coalescing to form a colloid or particle composed of,and exuding, organic matter. The phytoplankton cells leak DMSP, which triggers production of a signaling molecule in R. pomeroyi(orange cells). Once the AHL is present at a sufficient concentration, an array of metabolic shifts occur, including reduced intracellularglutamine which may cause increased uptake of dissolved organic nitrogen (DON), increased output of high-value metabolites such asalpha-ribazole, riboflavin, and shikimate, and release of sulfur metabolites such as methanethiol (MeSH; Supplementary Figure S1c),dimethylsulfide (DMS; Supplementary Figure S1c), and MTA (image credit: Jack Cook, WHOI).

Differential metabolites in response to DMSPWM Johnson et al

2313

The ISME Journal

composition of organic matter in the ocean because asingle molecule, such as DMSP, can alter themetabolism of organisms that encounter it, withlikely impact on their biogeochemical function.

Conflict of Interest

The authors declare no conflict of interest.

AcknowledgementsThis research is funded in part by the Gordon and BettyMoore Foundation through Grant GBMF3304 as well asby the National Science Foundation (Grants OCE-0928424and OCE-1154320). The instruments in the WHOI FT-MSFacility were purchased with support from the GBMF andNSF. Support for WMJ came from a National DefenseScience and Engineering Fellowship. In addition, wethank the following people for their assistance: KristaLongnecker and Cara Fiore for help with data analysis;Cara Fiore and Isa Howard-Åkerfeldt for technicalassistance; Jamey Fulton and Naomi Levine for assistancewith the GC-PFPD; Bethanie Edwards for instrumentinstruction; and Ben Van Mooy, Mary Ann Moran andPhil Gschwend for feedback on this work. The raw massspectrometry files and processed peak files have beenarchived at MetaboLights (MTBLS157, Release date:30 January 2015) run by the European MolecularBiology Laboratory-European Bioinformatics Institute(EMBL-EBI).

ReferencesAmin SA, Hmelo LR, van Tol HM, Durham BP, Carlson LT,

Heal KR et al. (2015). Interaction and signallingbetween a cosmopolitan phytoplankton and associatedbacteria. Nature 522: 98–101.

Baña Z, Ayo B, Marrasé C, Gasol JM, Iriberri J. (2014).Changes in bacterial metabolism as a responseto dissolved organic matter modification duringprotozoan grazing in coastal Cantabrian andMediterranean waters. Environ Microbiol 16:498–511.

Benton HP, Want EJ, Ebbels TMD. (2010). Correction ofmass calibration gaps in liquid chromatography-massspectrometry metabolomics data. Bioinformatics 26:2488–2489.

Biddanda B, Benner R. (1997). Carbon, nitrogen, andcarbohydrate fluxes during the production of particu-late and dissolved organic matter by marine phyto-plankton. Limnol Oceanogr 42: 506–518.

Brauer MJ, Yuan J, Bennett BD, Lu W, Kimball E,Botstein D et al. (2006). Conservation of the meta-bolomic response to starvation across two divergentmicrobes. Proc Natl Acad Sci USA 103:19302–19307.

Bürgmann H, Howard EC, Ye W, Sun F, Sun S, Napierala Set al. (2007). Transcriptional response of Silicibacterpomeroyi DSS-3 to dimethylsulfoniopropionate (DMSP).Environ Microbiol 9: 2742–2755.

Chambers MC, Maclean B, Burke R, Amodei D, Ruderman DL,Neumann S et al. (2012). A cross-platform toolkit

for mass spectrometry and proteomics. Nat Biotechnol30: 918–920.

Christie-Oleza JA, Scanlan DJ, Armengaud J. (2015).‘You produce while I clean up,’ a strategy revealedby exoproteomics during Synechococcus-Roseobacterinteractions. Proteomics 15: 3454–3462.

Chubukov V, Gerosa L, Kochanowski K, Sauer U. (2014).Coordination of microbial metabolism. Nat RevMicrobiol 12: 327–340.

Croft MT, Warren MJ, Smith AG. (2006). Algae need theirvitamins. Eukaryot Cell 5: 1175–1183.

DeLong EF, Franks DG, Alldredge AL. (1993). Phylo-genetic diversity of aggregate-attached vs. free-livingmarine bacterial assemblages. Limnol Oceanogr 38:924–934.

Dittmar T, Koch B, Hertkorn N, Kattner G. (2008). A simpleand efficient method for the solid-phase extraction ofdissolved organic matter (SPE-DOM) from seawater.Limnol Oceanogr Methods 6: 230–235.

Durham BP, Sharma S, Luo H, Smith CB, Amin SA,Bender SJ et al. (2015). Cryptic carbon and sulfurcycling between surface ocean plankton. Proc NatlAcad Sci USA 112: 453–457.

Fiore CL, Longnecker K, Kido Soule MC, Kujawinski EB.(2015). Release of ecologically relevant metabolites bythe cyanobacterium, Synechococcus elongatus CCMP1631. Environ Microbiol 17: 3949–3963.

Geng H, Bruhn JB, Nielsen KF, Gram L, Belas R. (2008).Genetic dissection of tropodithietic acid biosynthesisby marine roseobacters. Appl Environ Microbiol 74:1535–1545.

González JM, Covert JS, Whitman WB, Henriksen JR,Mayer F, Scharf B et al. (2003). Silicibacter pomeroyisp. nov. and Roseovarius nubinhibens sp. nov.,dimethylsulfoniopropionate-demethylating bacteriafrom marine environments. Int J Syst Evol Microbiol53: 1261–1269.

Gray MJ, Escalante-Semerena JC. (2010). A new pathwayfor the synthesis of α-ribazole-phosphate in Listeriainnocua. Mol Microbiol 77: 1429–1438.

Grossart H, Kiørboe T, Tang K, Ploug H. (2003). Bacterialcolonization of particles: growth and interactions.Appl Environ Microbiol 69: 3500–3509.

Grossart H-P. (2010). Ecological consequences of bacter-ioplankton lifestyles: changes in concepts are needed.Environ Microbiol Rep 2: 706–714.

Hmelo LR, Mincer TJ, Van Mooy BAS. (2011).Possible influence of bacterial quorum sensing on thehydrolysis of sinking particulate organic carbon inmarine environments. Environ Microbiol Rep 3:682–688.

Kido Soule MC, Longnecker K, Johnson WM,Kujawinski EB. (2015). Environmental metabolomics:analytical strategies. Mar Chem 177: 374–387.

Kiene RP, Slezak D. (2006). Low dissolved DMSP con-centrations in seawater revealed by small volumegravity filtration and dialysis sampling. Limnol Ocea-nogr Methods 4: 80–95.

Kirchman DL, Hoffman KA, Weaver R, Hutchins DA.(2003). Regulation of growth and energetics of a marinebacterium by nitrogen source and iron availability.MarEcol Prog Ser 250: 291–296.

Kirchman DL, Keil RG, Wheeler PA. (1989). The effectof amino acids on ammonium utilization and regen-eration by heterotrophic bacteria in the subarcticPacific. Deep Sea Res Part A Oceanogr Res Pap 36:1763–1776.

Differential metabolites in response to DMSPWM Johnson et al

2314

The ISME Journal

Kiørboe T, Grossart H-P, Ploug H, Tang K. (2002).Mechanisms and rates of bacterial colonizationof sinking aggregates. Appl Environ Microbiol 68:3996–4006.

Kuhl C, Tautenhahn R, Bottcher C, Larson TR, Neumann S.(2012). CAMERA: an integrated strategy for compoundspectra extraction and annotation of liquid chromato-graphy/mass spectrometry data sets. Anal Chem 84:283–289.

Lauro FM, McDougald D, Thomas T, Williams TJ, Egan S,Rice S et al. (2009). The genomic basis of trophicstrategy in marine bacteria. Proc Natl Acad Sci USA106: 15527–15533.

Levine NM, Varaljay VA, Toole DA, Dacey JWH, Doney SC,Moran MA. (2012). Environmental, biochemical andgenetic drivers of DMSP degradation and DMSproduction in the Sargasso Sea. Environ Microbiol 14:1210–1223.

Longnecker K, Futrelle J, Coburn E, Kido Soule MC,Kujawinski EB. (2015). Environmental metabolomics:databases and tools for data analysis. Mar Chem 177:366–373.

Longnecker K. (2015). Dissolved organic matter innewly formed sea ice and surface seawater. GeochimCosmochim Acta 171: 39–49.

Martin J, Knauer G, Karl D, Broenkow W. (1987).VERTEX: carbon cycling in the northeast Pacific.Deep Sea Res Part A Oceanogr Res Pap 34:267–285.

Mee MT, Wang HH. (2012). Engineering ecosystems andsynthetic ecologies. Mol Biosyst 8: 2470–2483.

Miller M, Bassler B. (2001). Quorum sensing in bacteria.Annu Rev Microbiol 55: 165–199.

Moran MA, Buchan A, González JM, Heidelberg JF,Whitman WB, Kiene RP et al. (2004). Genomesequence of Silicibacter pomeroyi reveals adaptationsto the marine environment. Nature 432: 910–913.

Morris JJ, Lenski RE, Zinser ER. (2012). The blackqueen hypothesis: evolution of dependencies throughadaptive gene loss. mBio 3: 1–7.

Parveen N, Cornell KA. (2011). Methylthioadenosine/S-adenosylhomocysteine nucleosidase, a critical enzymefor bacterial metabolism. Mol Microbiol 79: 7–20.

Paul C, Mausz MA, Pohnert G. (2012). A co-culturing/metabolomics approach to investigate chemicallymediated interactions of planktonic organismsreveals influence of bacteria on diatom metabolism.Metabolomics 9: 349–359.

Rabinowitz JD, Kimball E. (2007). Acidic acetonitrile forcellular metabolome extraction from Escherichia coli.Anal Chem 79: 6167–6173.

Reisch CR, Crabb WM, Gifford SM, Teng Q,Stoudemayer MJ, Moran MA et al. (2013). Metabolismof dimethylsulphoniopropionate by Ruegeria pomeroyiDSS-3. Mol Microbiol 89: 774–791.

Reisch CR, Moran MA, Whitman WB. (2011). Bacterialcatabolism of dimethylsulfoniopropionate (DMSP).Front Microbiol 2: 172–184.

Reitzer L. (2003). Nitrogen assimilation and globalregulation in Escherichia coli. Annu Rev Microbiol57: 155–176.

Sandy M, Butler A. (2009). Microbial iron acquisition:marine and terrestrial siderophores. Chem Rev 109:4580–4595.

Seyedsayamdost MR, Case RJ, Kolter R, Clardy J. (2011).The Jekyll-and-Hyde chemistry of Phaeobactergallaeciensis. Nat Chem 3: 331–335.

Seymour J, Simó R, Ahmed T, Stocker R. (2010).Chemoattraction to dimethylsulfoniopropionate through-out the marine microbial food web. Science 329:342–345.

Sharma AK, Becker JW, Ottesen EA, Bryant JA, Duhamel S,Karl DM et al. (2014). Distinct dissolved organicmatter sources induce rapid transcriptional responsesin coexisting populations of Prochlorococcus,Pelagibacter and the OM60 clade. Environ Microbiol16: 2815–2830.

Simó R, Vila-Costa M, Alonso-Sáez L, Cardelús C,Guadayol Ò, Vázquez-Domínguez E et al. (2009).Annual DMSP contribution to S and C fluxesthrough phytoplankton and bacterioplankton ina NW Mediterranean coastal site. Aquat Microb Ecol57: 43–55.

Smith CA, Maille GO, Want EJ, Qin C, Trauger SA,Brandon TR et al. (2005). A metabolite mass spectraldatabase. Ther Drug Monit 27: 747–751.

Smith CA, Want EJ, O’Maille G, Abagyan R, Siuzdak G.(2006). XCMS: processing mass spectrometry datafor metabolite profiling using nonlinear peakalignment, matching, and identification. Anal Chem78: 779–787.

Stefels J. (2000). Physiological aspects of the productionand conversion of DMSP in marine algae andhigher plants. J Sea Res 43: 183–197.

Stocker R. (2012). Marine microbes see a sea of gradients.Science 338: 628–633.

Suhre K, Schmitt-Kopplin P. (2008). MassTRIX:mass translator into pathways. Nucleic Acids Res 36:481–484.

Sumner LW, Amberg A, Barrett D, Beale MH,Beger R, Daykin CA et al. (2007). Proposedminimum reporting standards for chemical analysis:Chemical Analysis Working Group (CAWG) Meta-bolomics Standards Initiative (MSI). Metabolomics 3:211–221.

Sunda W, Kieber DJ, Kiene RP, Huntsman S. (2002). Anantioxidant function for DMSP and DMS inmarine algae. Nature 418: 317–320.

Tautenhahn R, Böttcher C, Neumann S. (2008). Highlysensitive feature detection for high resolution LC/MS.BMC Bioinformatics 9: 504–520.

Thiel V, Brinkhoff T, Dickschat JS, Wickel S, Grunenberg J,Wagner-Döbler I et al. (2010). Identification andbiosynthesis of tropone derivatives and sulfur volatilesproduced by bacteria of the marine Roseobacter clade.Org Biomol Chem 8: 234–246.

Tripp HJ, Kitner JB, Schwalbach MS, Dacey JWH,Wilhelm LJ, Giovannoni SJ. (2008). SAR11 marinebacteria require exogenous reduced sulphur forgrowth. Nature 452: 741–744.

Vallino JJ, Hopkinson CS, Hobbie JE. (1996). Modelingbacterial utilization of dissolved organic matter:optimization replaces Monod growth kinetics. LimnolOceanogr 41: 1591–1609.

Wolf S, Schmidt S, Müller-Hannemann M, Neumann S.(2010). In silico fragmentation for computer assistedidentification of metabolite mass spectra. BMC Bioin-formatics 11: 148–160.

Yoch DC. (2002). Dimethylsulfoniopropionate: its sources,role in the marine food web, and biological degrada-tion to dimethylsulfide. Appl Environ Microbiol 68:5804–5815.

Zan J, Cicirelli EM, Mohamed NM, Sibhatu H, Kroll S,Choi O et al. (2012). A complex LuxR-LuxI type

Differential metabolites in response to DMSPWM Johnson et al

2315

The ISME Journal

quorum sensing network in a roseobacterial marinesponge symbiont activates flagellar motility and inhi-bits biofilm formation. Mol Microbiol 85: 916–933.

Zan J, Liu Y, Fuqua C, Hill RT. (2014). Acyl-homoserinelactone quorum sensing in the Roseobacter clade.Int J Mol Sci 15: 654–669.

Zimmer DP, Soupene E, Lee HL, Wendisch VF,Khodursky AB, Peter BJ et al. (2000). Nitrogenregulatory protein C-controlled genes of Escherichiacoli: scavenging as a defense against nitrogen limita-tion. Proc Natl Acad Sci USA 97: 14674–14679.

This work is licensed under a CreativeCommons Attribution 4.0 International

License. The images or other third party material inthis article are included in the article’s CreativeCommons license, unless indicated otherwise in thecredit line; if the material is not included under theCreative Commons license, users will need to obtainpermission from the license holder to reproduce thematerial. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Supplementary Information accompanies this paper on The ISME Journal website (http://www.nature.com/ismej)

Differential metabolites in response to DMSPWM Johnson et al

2316

The ISME Journal