Respiration and ecological niche influence bacterial membrane lipid compositions

Respiration and ecological niche influence bacterialmembrane lipid compositions

Denice C. Bay, Sean C. Booth andRaymond J. Turner*Department of Biological Sciences, University ofCalgary, Calgary, AB T2N 1N4, Canada.

Summary

Bacterial membrane compositions vary widelybetween phyla and within related species. The typesof lipids within membranes are as diverse as theselective pressures that influence bacterial lifestylessuch as their mode of respiration and habitat. Thisstudy has examined the extent that respiration andhabitat affect bacterial fatty acid (FA) and polar lipid(PL) compositions. To accomplish this, over 300 FAand PL profiles from 380 previously characterizedspecies were assembled and subjected to multivari-ate statistical analyses in order to determine lipid tohabitat/respiration associations. It was revealed thatPL profiles showed a slight advantage over FA pro-files for discriminating taxonomic relationshipsbetween species. FA profiles showed greater correla-tion with respiration and habitat than PL. This studyidentified that respiration did not consistently favouruniform FA or PL changes when lipid profiles werecompared between examined phyla. This suggeststhat although phyla may adopt similar respirationmethods, it does not result in consistent lipid attrib-utes within one respiration state. Examination of FAand PL compositions were useful to identify taxo-nomic relationships between related species and pro-vides insight into lipid variations influenced by theniche of its host.

Introduction

Lipids play an essential role in the architecture and func-tion of biological membranes and their importance forcellular processes extends well beyond their role as apermeability barrier. The lipid composition of biologicalmembranes can have a profound impact on the integrity

(Janmey and Kinnunen, 2006; Zhang and Rock, 2008),virulence (Matsuura, 2013), intracellular signalling(Brumell and Grinstein, 2003) and permeability (Ytzhaket al., 2013) of all biological membranes including micro-bial lipids. Lipid content profiling applies to many topics inmicrobiology such as microbial biodiversity andextremophilic tolerances (for examples refer to Driessenet al., 1996; Nicolaus et al., 2001), provides insights intophysicochemical membrane structural variations (asreviewed by Denich et al., 2003) as well as membraneproteins (as reviewed by Driessen et al., 1996; Dowhanet al., 2004; Lee, 2004) and can identify unique lipids thatserve as important biomarkers to aid in microbial commu-nity taxonomic profiling (reviewed by Zelles, 1999).Towards these ends, fatty acid (FA) [and to a lesser extentpolar lipid (PL)] compositions for many pure-cultured indi-vidual species have been determined over the last 40years to assist in expanding this knowledge of the diver-sity of microbial lipid profiles (Lechevalier and Moss,1977; Lechevalier, 1982; Ratledge and Wilkinson, 1988).The use of pure-cultured bacterial lipid profiles havehelped reveal the influence of factors such as respirationon the composition and abundance of particular lipidspresent in diverse communities (Quezada et al., 2007).Most information regarding the influence of respirationand habitat have focused on examining lipid profiles fromisolated microorganisms such as Bacillus subtilis(Beranova et al., 2010) and within selected species from aphylum such as Gammaproteobacteria (Ivanova et al.,2000). By comparison, relatively little information is knownin this regard for intra-phylum and inter-phyla lipid varia-tion. The insights gained from the study of particular lipidprofiles and their biosynthesis in individual species, espe-cially Escherichia coli grown under defined conditions areoften generally applied to all bacteria (as reviewed byCronan, 2003; Zhang and Rock, 2008) despite knowndifferences in microbial lipid compositions (as discussedin Dowhan, 1997).

Bacterial lipid compositions are highly diverse in PL andFA content, and they can vary considerably betweenspecies (reviewed by Lechevalier and Moss, 1977;Lechevalier, 1982; Ratledge and Wilkinson, 1988;Dowhan, 1997). Bacterial PL describes the polar headgroup moiety that is associated to one or more FA by a

Received 11 July, 2014; revised 22 September, 2014; accepted 22September, 2014. *For correspondence. E-mail [email protected];Tel. (+1) 403 220 4308; Fax (+1) 403 289 9311.

bs_bs_banner

Environmental Microbiology (2014) doi:10.1111/1462-2920.12637

© 2014 Society for Applied Microbiology and John Wiley & Sons Ltd

mailto:[email protected]

glycerol [(phospholipid/glycophospholipid) or aminoalcohol [aminolipid, ceramide/sphingolipid and glycolipid(GL)] backbone. Bacterial PL content is highly diverseand is often characterized by its head group moietyand charge, for example phosphatidylglycerol (anionic),phosphatidylethanolamine (PE) (zwitterionic) andceramide/sphingosine (cationic) (Lechevalier and Moss,1977; Fahy et al., 2005). Bacterial PL head group chargesare known to influence membrane fluidity (Goldfine, 1984)and protein interactions within and outside of the lipidbilayer (Janmey and Kinnunen, 2006; Bogdanov et al.,2008a,b).

FA features are generally denoted by the total numberof carbon atoms (C) present from (and including) thecarboxyl group and the number of double bonds (D)present in unsaturated versions, providing a lipid numberdenoted as C : D respectively. Saturated and unsaturatedacyl chain (C:1 ≥) content in bacterial membranes canvary considerably as changes in the presence/absence ofFA, length of FA and relative proportion of FA can sub-stantially differ between Gram-negative and Gram-positive species (Fahy et al., 2005). Bacterial FA alsocontain chemically modified acyl chains that result inbranched [iso-, antesio- (Kaneda, 1991) and methylated(Jackson et al., 2007)], cyclopropane (Grogan andCronan, 1997) and hydroxylated (Raetz et al., 2007;Rezanka and Sigler, 2009) FA forms.

To explore FA and PL composition variation amongtaxonomically diverse bacteria, FA and PL compositionsfrom 380 bacterial species representing 17 different phylawere collected from previously published sources toassemble two lipid profile data sets. Both data sets werestatistically assessed to determine how effective FA andPL content profiles were at identifying bacterial phyla andto determine how host respiration and habitat accountedfor variability in bacterial classifications. As lipid composi-tion, respiration states and environmental niches wereexpected to be different within major phyla, a correlationbetween lipid profile and respiration/environmental nichewas predicted to be observed by multivariate statisticaltechniques. The outcome of this study revealed that bothlipid properties were useful in discriminating taxonomicrelationships. Agglomerative hierarchical clustering andprinciple component analysis (PCA) of PL profiles showedphylum-based associations, whereas respiration influ-ences were more apparent in the FA profile, which alsoreflected the host’s ecological niche. Additionally, closercomparison of lipids from different phyla that sharedsimilar respiration modes also identified respiration-specific FA and PL differences. Altogether, this study vali-dates the use of both PL and FA data sets for otherbioinformatics analyses and provides an importantcomparison between PL and FA membrane features inbacteria.

Results

Mean abundance of FA and PL highlight the diversity ofbacterial lipid profiles and identify phylum-specificlipid signatures

Prior to the analysis of bacterial lipid compositions, asummary of each lipid parameter was determined toobserve the mean abundance of FA and PL componentsidentified within each phylum (Table 1 and Fig. S1).During initial lipid analyses at the phylum level, it becameevident that Firmicutes required further subdivision (byclass and for Lactobacilli by order) to adequately identifyand account for the variation observed within this phylum.This was the only exception to phylum-level analyses, asclass-level subdivisions of other phyla did not show anyconsistent trends (data not shown). Table 1 summarizesthe number of bacterial species within each phylum andthe number of species with FA and/or PL informationavailable, and Fig. S1 provides the mean proportion ofeach lipid by phylum. Observation of mean FA abundancevalues for each of the 18 phyla surveyed in this studyrevealed that C16:0, C16:1, C18:1 and i15:0 FA werepresent in bacterial membranes at the highest proportionand were found within the most diverse phyla, highlightingtheir overall importance in bacterial membranes (Fig. S1).All remaining saturated and unsaturated FA were distrib-uted throughout all phyla, whereas the occurrencehydroxylation (OH)- and methylation (Me)-modified FAtypes were more restricted, validating their usefulness asphylum-specific chemotaxonomic markers (Fig. S1). Theabundances of saturated and unsaturated FA (C10–C15,C17, C19–C26) were typically low (ranging from 0.1% to8.0% total FA abundance), but one trend emerged: assaturated and unsaturated acyl chain length increased,species and phylum diversity diminished in accordancewith niche diversity. This trend also extended to some ofthe longer poly-unsaturated FA (C16:2, C18:2-C18:4,C20:2-C20:4) that were also present in fewer phyla bycomparison with their respective mono-unsaturated orsaturated FA (Fig. S1).

Examination of inter-phylum lipid profiles also appearsto contrast some of the reported FA trends, distinguishingGram-positive bacteria from Gram-negatives. Odd-numbered saturated and unsaturated FA (C15:0, C15:1,C17:0 and C17:1) that reportedly predominate in Gram-positive species (Kaneda, 1991; Parsons and Rock,2013) were widely distributed in both Gram-positive andGram-negative phyla (Fig. S1). Odd-numbered branched(iso- and antesio) FA are also known to be Gram-positivechemotaxonomic markers (Kaneda, 1991), but in thisstudy, iso- branched odd-numbered FA (i11:0, i13:0,i15:0, i17:0 and i17:1) were present in many Gram-negatives, in particular, non-proteobacterial Acidobacte-ria, Bacteroidetes and Spirochaetes (Fig. S1). Shorter

2 D. C. Bay, S. C. Booth and R. J. Turner

© 2014 Society for Applied Microbiology and John Wiley & Sons Ltd, Environmental Microbiology

antesio-branched FA (a11:0, a12:0 and a14:0) were foundprimarily in Gram-positives (Fig. S1), in agreement withpreviously reported trends (Kaneda, 1991; Parsons andRock, 2013). Longer antesio-branched FA (a15:0 anda17:0) were identified in a variety of Gram-negativesalbeit at much lower proportions than in Gram-positives(Fig. S1). Cyclopropane (cyclic) FA distributions in thisstudy also contrasted trends, stating that they were pri-marily enriched in Gram-negatives, as evident for 19:CyBacilli and Lactobacilli and to a lesser extent for 17:Cy(Fig. S1). Shorter hydroxylated FA (C10:0-OH to C15:0-OH) and unexpectedly iso-branched FA (i13:0-OH, i15;0-OH, i17:0-OH) were concentrated within Gram-negatives, but C16:0-OH and i13:0-OH (and to a lesserextent C17:0-OH and C18:0-OH) were also present insimilar proportions in many Gram-positive phyla (Fig. S1).

Examination of mean PL abundance values acrossdiverse phyla demonstrated that PE, phosphatidylglycerol(PG) and diphosphatidylglycerol (DPG) PL types werepredominant across all bacteria surveyed, and generally,one or more of these phospholipids were present inalmost all species (Fig. S1). GL/glycophospholipid (GPL)

were also predominant in bacterial membranes and weretypically found in all phyla with the exception of mostproteobacteria. Mean PE abundance was greatest in allproteobacterial phyla surveyed, but were also present inGram-positive phyla such as Actinomycetes and Bacilli.PG was identified within all phyla surveyed, making it oneof the most useful lipids for inter-phylum comparisonsbased on the relatively large variation in its abundanceamong different phyla (Fig. S1). DPG (cardiolipin) wasfound in most phyla almost as frequently as PG, butgenerally at a far lower abundance. All remaining PL typeswere often enriched within a particular phylum, especiallyin the Gram-negatives, indicating that phosphatidylserine(PS), phosphatidylcholine (PC) and their glycosylatedversions [phosphatidylinositolmannoside (PIM), phos-phatidyl (di- and mono-) mannosyl-ethanolamine(PME), glycophospholipid (GPL)] can serve as usefulchemotaxonomic markers (Fig. S1).

Based on this analysis, only a few FA (C16:0, C16:1,C18:1 and i15:0) and PL (PE, PG, DPG and GL) wereuniversally identified in all bacterial membranes, but atdifferent relative abundance values, making them useful

Table 1. A summary of all 380 bacterial species included in this study.

Phyluma/bTotal numberof species

Number of species withknown lipid compositions Species respiration

FA PL Ac FACe ANd

Actinobacteria (At) 102 100 94 75 16 9Bacilli (Ba)a 37 37 31 21 16 –Lactobacilli (La)b 18 18 17 – 18 –Clostridia (Cl)a 17 17 12 – – 17Deinococci (De) 5 5 4 5 – –Acidobacteria (Ac) 5 5 4 5 – –Bacteroidetes (Bc) 34 34 28 18 7 9Spirochaetes (Sp) 11 11 9 – 3 8Chlorobia (Ch) 3 3 2 – – 3Chloroflexi (Cf) 6 6 3 – 4 5Cyanobacteria (Cy) 19 19 6 1 7 11Planctomycetes (Pl) 4 4 3 3 1 –Aquificae (Aq) 3 3 3 3 – –Alphaproteobacteria (α) 27 27 26 22 4 1Betaproteobacteria (β) 21 21 16 19 2 –Gammaproteobacteria (γ) 43 43 43 13 30 –Deltaproteobacteria (δ) 14 14 7 5 – 9Epsilonproteobacteria (ε) 9 9 5 1 7 1Chlamydiae (Cd) 2 2 2 – 2 –Total surveyed 380 378 315 191 171 72

a. Represents a class designation within Firmicutes.b. Represents an order within Bacilli.c. Includes terms: obligate aerobe, microaerophilic and aerobic.d. Includes terms: obligate anaerobe and anaerobe.e. Includes terms: facultative aerobe and facultative anaerobe or for species described as both aerobic and anaerobic. A discussion of thefacultative respiration designation criteria is provided in supplementary materials.References for all lipid compositions are provided in Table S1.Lipid abbreviations: tetradecanoic acid (C14:0), tetradecenoic acid (C14:1), pentadecanoic acid (C15:0), iso- pentadecanoic acid (i15:0), antesio-pentadecanoic acid (a15:0), hexadecanoic acid (C16:0), hexadecenoic acid (16:1), iso-hexadecanoic acid (i16:0), heptadecanoic acid (C17:0),heptadecenoic acid (C17:1), iso-heptadecanoic acid (i17:0), antesio-heptadecanoic acid (a17:0), octadecanoic acid (C18:0), octadecenoic acid(18:1), iso-octadecanoic acid (i18:0), cyclo-heptadecanoic acid (17:cy), cyclo-nonadecanoic acid (19:cy), phosphatidylethanolamine (PE),phosphatidylglycerol (PG), diphosphatidylglycerol/ cardiolipin (DPG), phosphatidylcholine (PC), phosphatidylserine (PS), phosphatidylinositol (PI),phosphatidylmannosylethanolamine (PME), aminophospholipid (AL), ornithine phospholipid (OL), glycolipid (GL), and glycophospholipid (GPL).

Bacterial lipids are influenced by host niche 3


for further inter-species and inter-phyla comparisons. Therelative distributions and abundances of most FA and PLindicated the existence of phyla-specific lipid profiles thatcan serve as chemotaxonomic markers.

PL compositions improve bacterial classificationscompared with FA, but are less influenced by respirationand niche

To further examine lipid variability within diverse bacterialphyla, agglomerative hierarchical clustering analysis wasperformed on both FA and PL data sets to determine howeffective lipid compositions were in grouping relatedspecies together and to identify how host respirationand/or ecological niche may influence these associations.Clustering analysis of FA and PL composition (Figs 1 and2) revealed that species separated into two main clades,where each clade was enriched with either Gram-negative or Gram-positive species. Closer examination ofFA clustering (Fig. 1) showed that the cladistic separationoccurred among the three most prevalent FA types(C18:1, C16:1 and C16:0) and the remaining lipids formeda secondary clade influenced by iso- (i15:0 and i16:0) andantesio- (a15:0, a17:0) branched FA, corresponding totrends observed in mean FA abundance (Fig. S1). Theleast influential FA was the remaining acyl and unsatu-rated lipids, likely due to their relatively low abundanceand asymmetric distribution (Fig. 1). It is important to notethat many of the least abundant chemotaxonomically sig-nificant saturated and unsaturated FA (> 2% of total FA)can vary significantly when physicochemical parameterssuch as salinity, pH and temperature are altered, as hasbeen observed in studies of E. coli and other unrelatedbacteria (Mrozik et al., 2004) (Fig. S1).

PL clustering was largely determined based on thepresence or absence of PE, whereas the remaining PLcomponents formed two subclusters. The first was influ-enced by the distribution of PG, DPG and PI and theother by lower abundance aminophospholipid (APL), GLPC and PS. PL cluster analysis showed greater enrich-ment of species according to phylum/class in compari-son with FA clustering analysis. The PE cluster wasenriched with proteobacteria, whereas the DPG, PG, PIcluster was enriched with Actinomycetes and Firmicutes.The final cluster of APL, GL PC and PS contained theremaining Gram-negative non-proteobacterial species(Fig. 2). While it was clear that agglomerative hierarchi-cal clustering of PL better distinguished 16S rRNA taxo-nomical associations compared with FA profiles, theapplication of different distance did not significantly altercladistic outcomes (data not shown), suggesting that theclassification of species based on FA or PL compositionswas not absolute and is likely influenced by additionalfactors.

To explore if respiration and ecological niche mayaccount for the variation observed between unrelatedspecies in FA and PL clustering analyses, the respiration[aerobic (A); anaerobic (AN); facultative (FAC)] and envi-ronmental habitats of each species were examined. Theaddition of individual species respiration information ontopre-existing cluster analyses revealed that clustering pat-terns of many branches that contained unrelated speciesshared a common mode of respiration (Figs 1 and 2).Regions of respiration enrichment were particularly appar-ent in FA cluster profiles as compared with PL clusters (FAclusters occurred 22% more frequently than PL), suggest-ing that FA profiles reflected respiration influences to agreater extent than PL. This finding supports observationsmade for E. coli studies comparing FA and PL changesunder varying temperatures (Morein et al., 1996), whereFA content changes significantly varied to a greater extentthan those observed for PL. When species habitats wereincluded, as either a terrestrial and/or aquatic environ-ment, clusters of enrichment were observed within manyof the FA branches, but to a lesser extent in PL clusters.

Classifying bacterial species using commonly distrib-uted lipids by their relative overall abundance shows thatPL profile compositions improve phylum/class-level asso-ciations between species in comparison with FA. Con-versely, FA profiles were influenced by respiration andhabitat to a greater extent than the use of PL profiles.

Principle component analyses of FA and PL profilesprovide different advantages to identify phylum-levelassociations as compared with habitats

PCA was applied to both lipid data sets in an attempt todetermine how effective FA and PL profiles were in distin-guishing phylum-level associations at the species level.This technique was applied to reduce the dimensionalityof a data set containing many interrelated variables (inthis case lipids) by calculating a series of ‘principal com-ponents’, which are uncorrelated, and subsequentlyordering them so the first few explain the majority of allvariations present in the original data set. By examiningthe scores (species samples) and loadings (lipid vari-ables) calculated by this statistical technique, it was pos-sible to determine how species FA and PL profiles couldbe used to identify phylum associations. It also deter-mined if the segregation of lipid profiles according tospecies respiration influenced associations between itsphylum and lipids.

A total of 17 FA and 9 PL were selected as lipid vari-ables for PCA on the basis of their total distribution within> 20% of the 380 species and their presence in a least twoor more phyla (Fig. 3). The proportion of total explainedvariance within the first three principle components for theanalysis of the entire FA and PL data sets was 41.2% and



Fig. 1. Agglomerative hierarchically clustered heatmap analysis of 377 bacterial species FA profiles and its associated respiration andhabitat(s). Dendrograms shown on the top- and left-hand sides of the agglomerative hierarchically clustered heatmap show similaritiesbetween each FA (top dendrogram columns) and between FA compositions of bacterial species (left dendrogram and rows) as determinedusing a Euclidean distance and average linkage methods. Heatmap colouring (dark red to yellow) in the centre of this panel indicates therelative scaled proportion of FA identified (refer to the legend in the top left corner) for each bacterial species surveyed. Coloured rowsindicate the main taxonomic designation of each bacterial species as shown in the legend at bottom left corner of the panel. Secondaryheatmaps on the right-hand side of species labels indicate the respiration state, aerobic (red), facultative (yellow) and anaerobic (green), andenvironment/ habitat index (refer to the legend in this panel) for each species. Refer to Table 1 for phylum and FA abbreviation definitions.Details for each species FA composition is provided in Table S1.



Fig. 2. Agglomerative hierarchical cluster heatmap analysis of 316 bacterial species PL profiles and their associated respiration and habitat(s).Dendrograms shown on the top- and left-hand sides of the panel show similarities between each PL (top dendrogram columns) and betweenPL compositions of bacterial species (left dendrogram and rows) as determined using a Euclidean distance and average linkage methods.Heatmap colouring (dark red to yellow) in the centre of the panel indicates the relative scaled proportion of PL identified (refer to the legend inthe top left corner) for every bacterial species surveyed. Coloured rows indicate the main taxonomic designation of each bacterial species asshown in the legend at bottom left corner of the panel. Legends for Secondary heatmaps on the right-hand side of the panel are identical tothose described in Fig. 1. Phylum and PL abbreviation definitions are provided in Table 1. Details regarding individual species PL compositionsare provided in Table S1.



Fig. 3. PCA biplots of major FA and PL profiles from diverse bacterial species separated by respiration state(s). A–H show the first two PCAprinciple components (PC1 and PC2) as biplots and their explained variance by percentage for each component (listed in parentheses). Allbiplots show species scores and loadings (grey arrows for each lipid parameter) for the total FA (A) or PL (E) data sets and the remainingpanels only show lipids from aerobic (B, F), facultative (FAC; C, G) and anaerobic (AN; D, H) species. FA and PL scores for each specieswere numbered (Table S1), coloured (refer to legend in panel) and labelled according to its abbreviated two letter phylum tag number (Table 1and Table S1).A–D. PCA biplots of 17 FA types (refer to grey arrow loading labels).E–H. PCA biplots of 9 PL types (refer to grey arrow loading labels).



48.5%, respectively, indicating that only a portion of vari-ance in lipid profiles could be accounted for by PCA. InitialPCA of the entire FA and PL data sets included most of thechemotaxonomically significant lipid variables [short-branched acyl chains, hydroxylated or methylated FA(Quezada et al., 2007) and GL (SDQG and AGL), PG, PC(Brandsma et al., 2012)], but reduced the total proportionof explained lipid variance within the first three principlecomponents to 10–25% (Fig. S2). Taxonomic associa-tions between related species in scores plots were alsonoticeably reduced compared with PCA restricted tomajor FA and PL (Fig. 3). The reduction in explainablevariance and phylum-level associations involving uniquechemotaxonomically significant lipids is likely due to theirrelatively low abundance and selective distribution.Although many unique lipids are useful to identify speciespresent within mixed microbial communities of a specifichabitat and/or respiration (as reviewed by Zelles, 1999),their usefulness within this broadly sampled and environ-mentally diverse data set was hindered by this extremediversity. Closer examination of PCA scores plots of lipids(coloured according to phylum) demonstrated phylum-enriched clusters, indicating that PCA was useful in dis-tinguishing taxonomic relationships for many species(Fig. 3A and E). In particular, a clear division betweenGram-positive and Gram-negative Proteobacteria wereobserved in the PCA scores plots of both FA and PL.Similar to agglomerative hierarchical clustering analyses,phylum-level associations between species were slightlyimproved for PCA of PL in comparison with FA; however,FA showed much closer groupings of related species(Fig. 3A and E). PCA loading plots of complete FA and PLprofiles also identified the contribution of the main lipidsresponsible for the variation in lipid profiles of the sur-veyed bacteria, as well as the phyla that correspondedwith each lipid type.

Using the same 17 FA and 9 PL variables, separatePCA of three subsets of lipid profiles containing only A,FAC and AN species, respectively, was performed(Fig. 3B–D and F–H). This resulted in small increases (FA5–12%; PL 6–13%) in the total explained variance withinthe first three principle components for all respirationtypes by comparison with the first three principle compo-nents from total lipid analyses. PCA scores plots alsoshowed improvements in the phylum associations of PLand FA separated by respiration type, which was apparentfor phylum clusters observed for non-proteobaterialGram-negatives such as Bacteroidetes and Spirochaetes(Fig. 3B–D and F–H). To determine if the ecological nicheaccounted for some or all of the unrelated taxonomicassociations observed in FA and PL profiles, the mostcommon habitats of each organism were also examinedin the PCA scores and loadings (Fig. S3) and in separateanalyses (Fig. 4). Colour coding the same PCA scores

plots as Fig. 3 according to habitat rather than phylumdemonstrated that FA profiles improved niche-basedsubgroupings (separation of terrestrial versus aquaticspecies) far more than PL. PCA separated by respirationshowed the clearest subgroups according to habitat in theFA FAC respiration plot and to a lesser extent, the PL plot.When species were separated according to soil, aquatic,plant and human/animal microflora, PCA showed modestimprovements in explained variance in the first three prin-ciple components (FA 5–14%; PL 7–16%) similar to thesegregation of lipids according to species respiration(Figs 3 and 4). PCA of FA and PL separated according tohabitat-enhanced phylum clustering in scores plots(Fig. 4A–C and D–G). Human/animal microflora lipidprofile separation using PCA demonstrated the poorestphylum clustering, particularly for FA, confirming that PLprofiles were useful taxonomic markers for bacterialspecies discrimination (Fig. 4D and H). The improvementin phylum associations for habitat-restricted FA and PLprofile PCA indicate that niche influences lipid-basedtaxonomic species associations in addition to the respira-tion state. The ability of PCA to improve the clustering oftaxonomically related species when these environmentalparameters were used to divide the data set supports theirstrong influence on both FA and PL profiles in mostspecies.

PL and FA changes among A, FAC and AN speciesidentified respiration-specific lipid changes within phyla

The next aim of this study was to identify specific lipidchanges that occur within phyla/ classes that containspecies with diverse respiratory states. Lipids favouredunder a particular respiration condition within eachphylum can be determined by measuring the mean lipid %difference between pairwise comparisons of species thatcan adopt A, FAC and AN lifestyles. Overall, 12 of the 18phyla included in this study contained species with morethan one respiration method, whereas the remaining 6phyla/classes were composed of species with a singlemode of respiration (Table 1).

Examination of FA and PL differences between respira-tory modes for each phylum identified a set of uniquelipids favoured under different respiration conditions(Figs S4 and S5). Comparison of FA and PL valuesbetween A and FAC/AN conditions revealed the greatestdifferences, but no single respiration state was consist-ently favoured by any of the major FA or PL types. Com-parisons between FA and PL from FAC and AN speciesalso showed unique patterns of lipid gains and losses, butat much lower difference values than A comparisons.Similar to A/AN mean lipid % differences, the patterns ofFA or PL gains and losses were not uniformly biasedtowards one respiration type for all phyla when FAC to AN



Fig. 4. PCA biplots of major bacterial FA and PL profiles separated according to different ecological niches. A total of 17 FA (A–D) and 9 PL(E–H) were examined by PCA in each panel. A–D show loadings plots of 17 major FA (grey arrows) and scores plots (coloured and labelledby phylum tag) for 244 soil (A) species, 110 aquatic species (B), 37 plant (C) associated species and 135 human/animal microflora/ pathogen(D) species. E–H show loadings plots of 9 main PL (grey arrows) and scores plots (coloured and labelled by phylum tag) for 215 soil (E)species, 87 aquatic species (F), 36 plant (G) associated species and 117 human/animal microflora (H) species. The x- and y-axes of allpanels show the first two principle components (PC1 and PC2) and its explained variance as a percentage (in parentheses).



conditions were examined (Fig. S4C and F). This phe-nomenon is specifically apparent when following the mostabundant and widely distributed FA (C16:0, C16:1, C18:1and i15:0) and PL (PE, PG, DPG and GL/GLP). Forexample, C16:0 was present at greater levels in the mem-branes of FAC phyla with the exception of three A phyla:Spirochaetes, Epsilonproteobacteria and Alphaproteo-bacteria (Fig. S4A). Conversely, C16:0 values betweenFAC and AN conditions identified that Spirochaetesand Epsilonproteobacteria favoured this FA under FACconditions, whereas Alphaproteobacteria favoured ANconditions (Fig. S4B). Similar to the findings of FA, PLdifferences between respiration modes were alsoobserved for PE, PG, DPG and GL (Fig. S4D).

To complement the analysis of lipid differencesbetween bacterial respiration states, a comparison of FAand PL abundance profiles including respiration andhabitat were examined for each phylum using circularrelationship plots (Fig. 5 and Fig. S6). Circular relation-ships for phyla that had a single respiration state,extremophilic aerobes Acidobacteria, Aquifex andDeinococcus; anaerobes Chlorobia and Clostridia; andFAC Firmicutes class Lactobacilli were also determined(Fig. S6). Comparisons of mean lipid profiles by phylumrespiration using circular relationship plots highlightedwhich lipids were present between different respirationstates (Fig. S4), with the added advantage of includinghabitats for all respiration conditions (Fig. 5). The habitatsof A species typically demonstrated the greatest environ-mental variation, with the exception of Chloroflexi,Cyanobacteria and Epsilonproteobacteria. A lipid profilestypically demonstrated increased FA and PL diversity withincreased habitat diversity. When habitat diversity waslimited among aerobes within a phylum (or aerobes werecompletely absent), FAC lipid profiles reflected the trendobserved in A phyla. AN generally had the most restrictedhabitats, which would be expected for oxygen-deprivedenvironments, and often demonstrated lower FA and/orPL diversity and abundance in comparison with A andFAC profiles. A notable exception was observed in the ANFA and PL profiles of Spirochaetes and FA profiles of

Deltaproteobacteria, which showed the presence of addi-tional lipids and variations in lipid abundance (Fig. 5).

Examination of phylum-level lipid profile variationsunder different respiratory states and some environmen-tal habitats identified many FA and PL differences. Theselipid content differences highlight important variations inthe presence and absence or the relative lipid abun-dance and emphasize the influence of respiration andcertain environmental niches on bacterial lipid contentdiversity.

Discussion

The findings of this study revealed that FA and PL profilesof individual species were capable of distinguishingGram-positive from Gram-negative species and validatedlipid profile trends previously reported for both bacterialdivisions. The wide breadth of taxonomic diversity of lipidprofiles surveyed in this work also helped to identify newtrends in phylum-specific lipid distributions. Prior to thisanalysis, odd-numbered saturated, unsaturated andbranched FA reportedly predominated in Gram-positives(as reviewed by Kaneda, 1991; Parsons and Rock, 2013),whereas cyclic and hydroxylated FA typified Gram-negatives (Grogan and Cronan, 1997; Raetz et al., 2007)in many cases based on the examination of a few keyspecies. The present study, which included expandedspecies diversity, demonstrated that these trends werenot strictly associated to Gram-negative or Gram-positivephyla and highlighted many FA exceptions that should beincluded to update lipid profile trend generalizations. Thisis most apparent when considering the distribution ofbranched-chain FA, which should be considered carefullywhen distinguishing Gram-positive species from Gram-negatives as these FA were found in many non-proteobacterial Gram-negatives. The presented data alsoindicated that branched-chain FA may be useful aschemotaxonomic markers when examining both respira-tion and niche of the organisms.

Another important observation from this study was thatincreasing acyl chain lengths corresponded to reductions

Fig. 5. Circular relationship ring plots comparing mean lipid % abundance values and habitat diversity among aerobic, facultative andanaerobic phyla. In each panel, a summary of mean lipid % abundance and habitat niche values are summarized for all species within asingle phylum. Each panel is labelled by its two letter phylum abbreviation (refer to Table 1). Three coloured tabs present on each ring indicatethe mean values for aerobic (A; green), facultative (FAC; orange) and anaerobic (AN; red) respiration states of species with a single phylum.Coloured lines (red, green or orange) emanating from each of the three respiration tabs to lipid and/or habitat variables (small unfilled circlesaround the ring) identified the presence of the variable within all species under that condition. Line thickness is proportional to the mean valuefor the indicated variable (habitat/lipid) for all species that were associated with the specified respiration state. The upper left-hand ring of eachphylum ring displays 8 PL values and the bottom left-hand side shows 12 summed FA categories. Each FA category represents thesummation of mean FA % abundance values according to 1 of the 12 major FA chemical features: saturated (sat), saturated-methylated(sat-Me), mono- unsaturated (mono-unsat), poly- unsaturated (poly-unsat), cyclopropylated (cyclic), iso-branched saturated (i-sat),iso-branched hydroxylated saturated (i-sat-OH), iso-branched unsaturated (i-unsat), antesio-branched saturated (a-sat) and antesio-branchedunsaturated (a-unsat). The remaining right-hand ring provides the mean habitat values based on descriptions for all species within thespecified phylum represented as a value (refer to Experimental procedures for scoring details). Due to limited PL data available, PL profileswere not shown for anaerobic Chloroflexi and facultative Planctomycetes species.



c

AN

FAC

c

AN

FAC

c

AN

FAC

c

AN

FAC

Pl

a b

g

e

d

c

AN

FAC

c

AN

FAC



in phylum diversity. Despite a noticeable decrease inphylum diversity as saturated and mono-unsaturated FAincreased in length, the limited species that were identi-fied with longer FA (C20-C26) were found in relativelydiverse phyla lacking any common niche or respiration,indicating their usefulness solely as chemotaxonomicmarkers. This may reflect physicochemical requirements,since longer FA increases membrane stability at highertemperatures, pressures and bilayer thickness (asreviewed by Mrozik et al., 2004), which may be requiredto compensate for habitat limitations for these species.

The outcome of FA and PL clustering analyses demon-strated only a partial resemblance to known 16S rRNA-based species phylogeny (Garrity and Holt, 2001)suggesting that lipid compositions are useful in distin-guishing some species within the 18 phyla surveyed, butnot others. It also highlighted that FA and PL lipid compo-sitions were not completely adequate to distinguishphylum/class-level differences among the surveyedspecies and that these compositions may be influencedby other parameters that include respiration and habitat.The results from PCA were consistent with the findingsfrom agglomerative hierarchical clustering (Figs 1 and 2),and demonstrated that FA profiles, in particular FACspecies, exhibited better niche-related clustering com-pared with PL profiles. PCA also served as a useful sta-tistical method to determine the strengths of lipid profilesto identify phylum-level relationships between lipid pro-files from highly diverse species. Although chemotaxono-mically significant lipids, such as short-branched acylchains, hydroxylated or methylated FA and GL (SDQGand AGL), PG, PC were shown to be useful for distin-guishing taxa in mixed microbial communities (Quezadaet al., 2007; Zhang et al., 2007), their low abundance andinconsistent distribution in this environmentally diverselipid data set likely caused the diminished varianceexplainable by PCA in this study.

The use of FA and PL restricted by habitat (Fig. 4) andrespiration states (Fig. 3) demonstrated improvements inphylum-based clustering and revealed that a species’environment confers some influence on its lipid composi-tion. Although habitat and respiration could not completelyaccount for unrelated taxonomic associations betweenspecies using agglomerative hierarchical clustering andPCA, it does support the use of FA and PL lipid profiles aschemotaxonomic markers. All of the microorganismsselected for this study were cultivated under laboratorystandard growth conditions and almost all species hadsequenced genomes. Many of these species (68%) werereportedly isolated from multiple habitats (Bergey’smanuals; references provided in Table S1) and the ubiq-uitous lifestyle of these species may have also obscuredchanges in lipid content. Currently, there are no experi-mental data examining changes in FA or PL content for a

specific microbial species grown/isolated from differentenvironmental habitats, making this analysis useful forfuture studies.

Examination of both PL and FA profiles together alsosuggest that PL profiles were better than FA for determin-ing taxonomic associations between species (Figs 1–4).This was finding was surprising, because FA profiling(99% of surveyed profiles) appears to dominate overavailable PL profiles (in this study 82%) for pure-culturedspecies (Table 1). Most lipid-profiling experiments focuson bacterial FA content using fatty acid methyl esters(FAME) and mass spectroscopy (MS) techniques,whereas PL profiles are reported primarily in systematicjournals (refer to references in cited Table S1). Previousstudies of Gram-positive PL content throughout the 1960–1980s (Lechevalier and Moss, 1977; Lechevalier, 1982;Ratledge and Wilkinson, 1988) demonstrated that PLcomposition is useful in distinguishing phyla- and evenclass-level differences, as observed in this study for theFirmicutes classes Bacilli (Lactobacilli) and Clostridia.With regards to Gram-negatives, PL profile analysisappears to have fallen out of favour possibly due to thehigher overall PE levels present in many Proteobacteria.Our findings highlight that differences in the amount of PEdid not prevent phylum-level associations between Gram-negatives because many non-Proteobacterial phyla com-pletely lack detectable PE (Fig. 5 and Fig. S6). Lipidprofiles of non-Proteobacterial Gram-negatives such asBacteroidetes and Spirochaetes appeared to have themost difficulty forming phylum-specific groupings byagglomerative hierarchical clustering or by PCA, indicat-ing that both FA and PL profiles were subject to the influ-ence of additional factors including respiration and nicheas shown in Fig. 5 and Fig. S4.

The influences of A, FAC and AN respiration and to alesser extent niche were more apparent in FA profilescompared with PL (Figs 1 and 3). This finding supportsstudies of E. coli (Morein et al., 1996) and Bacillus cereus(Haque and Russell, 2004), which noted lipid contentdifferences under growth conditions such as temperaturewere far more significant in FA. Comparing respiratorydifferences between FA and PL profiles within phyla alsoidentified different lipid profile patterns under a particularrespiration condition (Fig. 5 and Fig. S4). Studies report-ing individual Gram-negative species lipid profile changesunder different respiration states prior to this study werelimited and available only for Rhodopseudomonasacidophila (alphaproteobacterium) (Russell et al., 2002)and Vibrio sp. ABE-1 (alphaproteobacteria) (Morita et al.,1992) species. Rhodopseudomonas acidophila PL andFA profiles between A and AN growth conditions demon-strated that A PG and DPG contents were higher than inAN conditions because of the concomitant increase in PC(Russell et al., 2002). Major FA C16:1 and C18:1 in



R. acidophila also increased under AN growth, whereasC16:0 decreased and the presence of the uniquehydroxylated FA C16:0-OH also increased under ANconditions (Russell et al., 2002). The findings forR. acidophila and V. sp. ABE-1 agreed with the findingsshown for mean PL and/or FA % differences inalphaproteobacterial and gammaproteobacterial betweenA and FAC/AN conditions (Figs S4 and S5).

With regards to Gram-positive organisms, the resultsfrom studies of FA profiles (C16:0, C18:0, i15:0 and a15:0)from cold-adapted strains of Bacillus subtilis culturedunder A and AN conditions (Beranova et al., 2010) werealso supported by FA findings for Bacilli under A and ANrespiration conditions (Fig. S4). Furthermore, this studyidentified that respiration and habitat influenced lipid-specific differences within and between phyla (Fig. 5 andFig. S4). For example many Spirochaetes included in thissurvey are obligate pathogens that reportedly lack all FAbiosynthesis genes, gaining the majority of FA from itshosts despite the presence of complex branched FA andunique PL that could not be synthesized or provided by itshost (Das et al., 2000). The analysis of diverse bacteriallipid profiles according to respiration and habitat per-formed in this study can help support the existence of ayet unidentified lipid biosynthetic pathway(s) and identifytheir presence within uncharacterized phyla.

In conclusion, the bioinformatic analysis of 380 bacte-rial lipid profiles has revealed insights into taxonomic clas-sifications involving FA and PL and revealed lipid patternsfavoured under various respiration conditions. It alsorevealed that habitat has some influence on FA and PLprofiles; however, the diversity and inherent bias of theseselected species likely preclude the use of these data setsfor in-depth habitat analysis. These lipid data sets will beuseful in other bioinformatics studies: a limited version ofthis data set was recently used to examine how the topo-logical insertion bias of membrane protein transporterswas influenced by bacterial PL head group charges andFA thickness (acyl length) (Bay and Turner, 2012). Thesuccess of this study has demonstrated that bacterial lipidprofiles determined from isolated lab-cultivated bacteriacan provide useful insights into many physiological facetsof bacterial life and hypothesis testing well beyond itsintentioned use as chemotaxonomic markers.

Experimental procedures

Assembly of membrane composition matrices usedin this study

Lipid compositions examined in this study were assembledusing published FA content (by mol%) and PL content (byrelative % abundance) values from characterized plasmamembranes of 380 bacterial species (Table S1). Selection

preference for lipid composition profiles favoured bacterialspecies with sequenced genomes (87% of total database) soeach data set could be integrated with genomic data in futurestudies, whereas the remaining profiles were selected torepresent extremophilic species. Due to the limited or incom-plete characterization of most bacterial plasma membranes,the lipid parameters examined herein did not reflect theasymmetrical lipid distributions present in bacteria (Goldfine,1984). Both FA and PL compositions for individual specieswere determined under optimal laboratory growth conditions.

The relative abundance of FA compositions were describedusing five different acyl chain parameters, which involvedsaturated acyl chain lengths (C10:0 to C26:0), unsaturatedacyl chains (one to four unsaturated acyl chains), iso-branched acyl chains (ranging from i6 to i19), antesio-branched acyl chains (a12:0 to a19:0), and cyclopropane acylchains (17:Cy to 20:Cy). Additionally, Me or OH modificationsof all five parameters listed were also included in this study.FA examined in this study were collected from previouslypublished experiments providing FA compositions from iso-lated plasma membranes by percentage (%) weight or byaverage % mass and references are provided in Table S1. FAcompositions were provided as either underivatized FAs or asFAME from mass spectroscopy (MS), gas chromatography orhigh-performance liquid chromatography (HPLC) techniques.Experimental FA contents were summated and representedas a percentage of total abundance in order to comparecontents between species. Values were normalized (0–1)using the function (ei-Emin/Emax-Emin); where ei is the variable,Emax is the maximum value of all variables and Emin isminimum value of all variables examined.

PL content determined in this study was available for 315species (Table S1) and was determined from published thinlayer chromatography (TLC) and HPLC experiments. PLcontent surveyed within species included cardiolipin/DPG,PG, PS, PI, PIM, PE, LPE, PME, PC, APL, lysylphosphati-dylglycerol, GL, GPL, ceramide (C), sphingophospholipid,aminoglycophospholipid and sulfoquinovosyl diacylglycerol.The amount of PL per species collected for this study wasconverted into a percentage total based on total amounts ordirectly from total % proportions cited in published experi-ments (Table S1). In some cases, PL values were determineddirectly from TLC chromatograms or HPLC spectra. In theseinstances, TLC spot image quantification was performedusing KODAK 1D gel imaging software (Pizzonia, 2001), wherethe measured intensities of particular lipid types were dividedby the total number quantified values.

FA and PL values determined for each bacterial speciesmembrane were sorted into groups according to their bacte-rial phylum, class and species according to taxonomic des-ignations set by Bergey’s Manual of Systematic Bacteriology(Garrity, 2001) and the NCBI taxonomy browser (Bensonet al., 2009; Sayers et al., 2009). Bacterial respiration andecological habitat(s) or isolation niche(s), were obtained fromBergey’s manual or from references cited in Table S1. Bac-terial respiration was categorized into three headings, A, FACand AN, was and scored as 0 (absent) or 1 (present). Fordetails regarding respiration designations descriptions referto Table 1. Habitat/environmental niche descriptions weredetermined as search terms and scored as described forrespiration when more than one term could be scored. A total



of 28 habitat descriptions were selected to describe the eco-logical niche(s) of each species included in this analysis,which fell into three main categories: aquatic, terrestrial andanimal/plant.

Statistical analysis of lipid data sets

Statistical analysis of lipid profile data sets were initially per-formed using MICROSOFT EXCEL 2007 edition software tocalculate mean lipid abundance values for FA and PL compo-nents in phyla and classes using the formula (Σ x/ n), where xrepresents the value of single type of FA or PL and n repre-sents the total number of species within a phylum with therespective lipid. To identify FA or PL exhibiting major differ-ences between respiration modes, mean lipid % differenceswere calculated using the formula [(Σ x/n)Resp1 − (Σ x/n) Resp2],where Resp1 and Resp2 formulas are identical to the mean lipidabundance values for the same FAor PL, except that values ofResp1 and Resp2 represent all species within a phylum withdifferent respiration modes (A, FAC and AN). By this calcula-tion, lipids favouring a particular respiration mode will have ahigh mean % difference value in one of the two axes repre-senting each respiration state. Lipids with similar mean %difference values between both respiration states will remainconstant and produce the lowest mean % difference values. Atwo way analysis of variation calculation was used to assessstatistically significant differences between respiration lipid %differences (P ≥ 0.05) and only significant values for phyla areshown here. To avoid multiple testing problems that may occurwithin phylum of limited species number, P-values wereassessed using a Bonferroni correction (Shaffer, 1995).

Circular relationship plots were generated for each phylumaccording to its respiration state(s) to provide mean lipidprofile values and habitat distributions. For this analysis,mean PL % abundance values within each phylum wereselected based on their overall abundance and mean FA %abundance values were summed according to its majorchemical features representing total lipid changes. Twelvehabitat designations were selected from 28 available descrip-tions based on its presence in at least 30% of the bacterialdata set. Habitat designations were scored by the presence(1) or absence (0) of the species in an overall habitat descrip-tion. Among the 12 habitat designations, 4 main habitats (soil,aquatic, plant, human/ animal microflora) were summarizedusing multiple description terms: ‘polluted’ included the term(metal-) contaminated, ‘animal’ included terms insect, inver-tebrate (fish) and mammal, ‘human’ included terms opportun-istic, pathogen, and microflora, ‘rhizosphere’ included theterm nodule and ‘plants’ included terms vegetation, decaying,cellulose degradation, legumes, leaf and tree.

All remaining statistical analyses were performed using R

STATISTICS version 2.14.1 (http://www.r-project.org). Lipiddata sets were subdivided according to phylum, class andgenus for matrix assembly using R STATISTICS ‘data.matrix(dataset)’ function (http://stat.ethz.ch/R-manual/R-patched/library/base/html/data.matrix.html).

Agglomerative hierarchical clustering analysisof lipid matrices

Agglomerative hierarchical clustering analysis of FA and PLlipid compositions per species were performed by first deter-

mining the number of clusters present within the matrix usingthe K-means partitioning method using the ‘kmeans’ cluster-ing function (http://stat.ethz.ch/R-manual/R-patched/library/stats/html/kmeans.html). The R STATISTICS library (pvclust)package (http://www.sigmath.es.osaka-u.ac.jp/shimo-lab/prog/pvclust/) and the function ‘pvclust(data, method.hclust = “Ward”, method.dist = “euclidean”)’ was used todetermine total number of predicted clusters to generate hier-archical groupings using Euclidean distance and Ward-linkage method to determine clustering similarities within thematrix. Associated heatmaps were generated for thesegroupings using the function ‘heatmap.2(data.matrix)’. TheP-value significance of cluster dendrogram nodes (shown onthe x- and y-axes) were determined after bootstrapping(either 100 or 1000 replicates) and P-values (α = 0.90) pro-vided in red and green, respectively, indicate approximatelyunbiased and bootstrap probability values (Figs 1 and 2).Similar clustering divisions of the major nodes were alsoobtained using a maximum likelihood-based model for dis-tance relationships in hierarchical clustering analysis per-formed for this study (data not shown).

Principle component analysis of lipid matrices

PCA was performed on FA and PL data sets using R STATIS-TICS function ‘prcomp(data)’ (http://stat.ethz.ch/R-manual/R-patched/library/stats/html/prcomp.html), which included amean centring function. The scores and loadings for eachPCA analysis shown in PCA biplots were prepared using‘ggbiplot(data.pca)’ as part of the ‘ggplot2’ library package.The significance of the PCA loadings was determined bybootstrapping the generated eigenvectors as described by(Peres-Neto et al., 2003) to estimate type I error rateP-values (ranging from 0.001 to 0.09) after resampling rowvariables 1000 times. Three species, Kinetococcusradiotolerans, Planococcus sp. and Citreicella aestuarii, wereoriginally omitted from initial PCA analyses due to their pres-ence as strong outliers in both scores and residuals plots withrespect to other phylum members. Closer examination ofeach species lipid profile identified that key FA and/or PLvalues were absent from the profile, so lipid profiles fromclosely related species (Table S1) measured under the sameexperimental conditions were used to replace missing datafor all three lipid profiles and were included into the final PCA.

Acknowledgements

We would like to thank Tara Winstone and Jana Vamosi forhelpful manuscript discussions. Funding for this work wasprovided by Discovery Grant Natural Sciences and Engineer-ing Research Council (NSERC) to RJT.

References

Bay, D.C., and Turner, R.J. (2012) Membrane compositioninfluences the topology bias of bacterial integral membraneproteins. Biochim Biophys Acta 1828: 260–270.

Benson, D.A., Karsch-Mizrachi, I., Lipman, D.J., Ostell, J.,and Sayers, E.W. (2009) GenBank. Nucleic Acids Res 37:D26–D31.



http://www.r-project.org

http://stat.ethz.ch/R-manual/R-patched/library/base/html/data.matrix.html

http://stat.ethz.ch/R-manual/R-patched/library/base/html/data.matrix.html

http://stat.ethz.ch/R-manual/R-patched/library/stats/html/kmeans.html

http://stat.ethz.ch/R-manual/R-patched/library/stats/html/kmeans.html

http://www.sigmath.es.osaka-u.ac.jp/shimo-lab/prog/pvclust/

http://www.sigmath.es.osaka-u.ac.jp/shimo-lab/prog/pvclust/

http://stat.ethz.ch/R-manual/R-patched/library/stats/html/prcomp.html

http://stat.ethz.ch/R-manual/R-patched/library/stats/html/prcomp.html

Beranova, J., Mansilla, M.C., de Mendoza, D., Elhottova, D.,and Konopasek, I. (2010) Differences in cold adaptation ofBacillus subtilis under anaerobic and aerobic conditions.J Bacteriol 192: 4164–4171.

Bogdanov, M., Mileykovskaya, E., and Dowhan, W. (2008a)Lipids in the assembly of membrane proteins and organi-zation of protein supercomplexes: implications for lipid-linked disorders. Subcell Biochem 49: 197–239.

Bogdanov, M., Xie, J., Heacock, P., and Dowhan, W. (2008b)To flip or not to flip: lipid-protein charge interactions are adeterminant of final membrane protein topology. J Cell Biol182: 925–935.

Brandsma, J., Hopmans, E.C., Philippart, C.J.M., Veldhuis,M.J.W., Schouten, S., and Sinninghe Damsté, J.S. (2012)Low temporal variation in the intact polar lipid compositionof North Sea coastal marine water reveals limitedchemotaxonomic value. Biogeosciences 9: 1073–1084.

Brumell, J.H., and Grinstein, S. (2003) Role of lipid-mediatedsignal transduction in bacterial internalization. CellMicrobiol 5: 287–297.

Cronan, J.E. (2003) Bacterial membrane lipids: where do westand? Annu Rev Microbiol 57: 203–224.

Das, R., Hegyi, H., and Gerstein, M. (2000) Genome analy-ses of spirochetes: a study of the protein structures, func-tions and metabolic pathways in Treponema pallidum andBorrelia burgdorferi. J Mol Microbiol Biotechnol 2: 387–392.

Denich, T.J., Beaudette, L.A., Lee, H., and Trevors, J.T.(2003) Effect of selected environmental and physico-chemical factors on bacterial cytoplasmic membranes.J Microbiol Methods 52: 149–182.

Dowhan, W. (1997) Molecular basis for membrane phospho-lipid diversity: why are there so many lipids? Annu RevBiochem 66: 199–232.

Dowhan, W., Mileykovskaya, E., and Bogdanov, M. (2004)Diversity and versatility of lipid-protein interactionsrevealed by molecular genetic approaches. BiochimBiophys Acta 1666: 19–39.

Driessen, A.J., van de Vossenberg, J.L.C.M., and Konings,W.N. (1996) Membrane composition and ion-permeabilityin extremophiles. FEMS Microbiol Rev 18: 139–148.

Fahy, E., Subramaniam, S., Brown, H.A., Glass, C.K., Merrill,A.H., Jr, Murphy, R.C., et al. (2005) A comprehensive clas-sification system for lipids. J Lipid Res 46: 839–861.

Garrity, G.M. (2001) The Archaea and the Deeply Branchingand Phototrophic Bacteria, Vol. 1. New York, USA: Springer.

Garrity, G.M., and Holt, J.G. (2001) The Road Map to theManual. Bergey’s Manual of Systematic Bacteriology, 2ndedn. New York, NY, USA: Springer.

Goldfine, H. (1984) Bacterial membranes and lipid packingtheory. J Lipid Res 25: 1501–1507.

Grogan, D.W., and Cronan, J.E., Jr (1997) Cyclopropane ringformation in membrane lipids of bacteria. Microbiol Mol BiolRev 61: 429–441.

Haque, M.A., and Russell, N.J. (2004) Strains of Bacilluscereus vary in the phenotypic adaptation of their mem-brane lipid composition in response to low water activity,reduced temperature and growth in rice starch. Microbiol-ogy 150: 1397–1404.

Ivanova, E.P., Zhukova, N.V., Svetashev, V.I., Gorshkova,N.M., Kurilenko, V.V., Frolova, G.M., and Mikhailov, V.V.

(2000) Evaluation of phospholipid and fatty acid composi-tions as chemotaxonomic markers of Alteromonas-likeproteobacteria. Curr Microbiol 41: 341–345.

Jackson, M., Stadthagen, G., and Gicquel, B. (2007) Long-chain multiple methyl-branched fatty acid-containing lipidsof Mycobacterium tuberculosis: biosynthesis, transport,regulation and biological activities. Tuberculosis (Edinb)87: 78–86.

Janmey, P.A., and Kinnunen, P.K. (2006) Biophysical proper-ties of lipids and dynamic membranes. Trends Cell Biol 16:538–546.

Kaneda, T. (1991) Iso- and anteiso-fatty acids in bacteria:biosynthesis, function, and taxonomic significance.Microbiol Rev 55: 288–302.

Lechevalier, H.A. (1982) Lipids, 2nd edn. Boca Raton, FL,USA: CRC Press.

Lechevalier, H.A., and Moss, C.W. (1977) Lipids in bacterialtaxonomy – a taxonomist’s view. Crit Rev Microbiol 5:109–210.

Lee, A.G. (2004) How lipids affect the activities of integralmembrane proteins. Biochim Biophys Acta 1666: 62–87.

Matsuura, M. (2013) Structural modifications of bacteriallipopolysaccharide that facilitate Gram-negative bacteriaevasion of host innate immunity. Front Immunol 4: 109.

Morein, S., Andersson, A., Rilfors, L., and Lindblom, G.(1996) Wild-type Escherichia coli cells regulate themembrane lipid composition in a ‘window’ between geland non-lamellar structures. J Biol Chem 271: 6801–6809.

Morita, N., Gotoh, M., Okajima, N., Okuyama, H., Hayashi,H., Higashi, S., and Murata, N. (1992) Both the anaerobicpathway and aerobic desaturation are involved in the syn-thesis of unsaturated fatty acids in Vibrio sp. strain ABE-1.FEBS Lett 297: 9–12.

Mrozik, A., Piotrowska-Seget, Z., and Łabuzek, S. (2004)Cytoplasmatic bacterial membrane responses to environ-mental perturbations. Pol J Environ Stud 13: 487–494.

Nicolaus, B., Manca, M.C., Lama, L., Esposito, E., andGambacorta, A. (2001) Lipid modulation by environmentalstresses in two models of extremophiles isolated from Ant-arctica. Polar Biol 24: 1–8.

Parsons, J.B., and Rock, C.O. (2013) Bacterial lipids:metabolism and membrane homeostasis. Prog Lipid Res52: 249–276.

Peres-Neto, P.R., Jackson, D.A., and Somers, K.M. (2003)Giving meaningful interpretation to ordination axes:assessing loading significance in principal componentanalysis. Ecology 84: 2347–2363.

Pizzonia, J. (2001) Electrophoresis gel image processing andanalysis using the KODAK 1D software. Biotechniques 30:1316–1320.

Quezada, M., Buitron, G., Moreno-Andrade, I., Moreno, G.,and Lopez-Marin, L.M. (2007) The use of fatty acid methylesters as biomarkers to determine aerobic, facultativelyaerobic and anaerobic communities in wastewater treat-ment systems. FEMS Microbiol Lett 266: 75–82.

Raetz, C.R., Reynolds, C.M., Trent, M.S., and Bishop, R.E.(2007) Lipid A modification systems in gram-negative bac-teria. Annu Rev Biochem 76: 295–329.



Ratledge, C., and Wilkinson, S.G. (1988) Microbial Lipids, 1stedn. San Diego, CA, USA: Academic Press.

Rezanka, T., and Sigler, K. (2009) Odd-numbered very-long-chain fatty acids from the microbial, animal and plant king-doms. Prog Lipid Res 48: 206–238.

Russell, N.J., Coleman, J.K., Howard, T.D., Johnston, E., andCogdell, R.J. (2002) Rhodopseudomonas acidophila strain10050 contains photosynthetic LH2 antenna complexesthat are not enriched with phosphatidylglycerol, and thephospholipids have a fatty acyl composition that is unusualfor purple non-sulfur bacteria. Biochim Biophys Acta 1556:247–253.

Sayers, E.W., Barrett, T., Benson, D.A., Bryant, S.H.,Canese, K., Chetvernin, V., et al. (2009) Databaseresources of the National Center for Biotechnology Infor-mation. Nucleic Acids Res 37: D5–D15.

Shaffer, J.P. (1995) Multiple hypothesis testing. Ann RevPsych 46: 561–584.

Ytzhak, S., Weitman, H., and Ehrenberg, B. (2013) The effectof lipid composition on the permeability of fluorescentmarkers from photosensitized membranes. PhotochemPhotobiol 89: 619–624.

Zelles, L. (1999) Fatty acid patterns of phospholipids andlipopolysaccharides in the characterisation of microbialcommunities in soil: a review. Biol Fertil Soils 29: 111–129.

Zhang, Q.C., Wang, G.H., and Yao, H.Y. (2007) Phospholipidfatty acid patterns of microbial communities in paddy soil

under different fertilizer treatments. J Environ Sci (China)19: 55–59.

Zhang, Y.M., and Rock, C.O. (2008) Membrane lipid homeo-stasis in bacteria. Nat Rev Microbiol 6: 222–233.

Supporting information

Additional Supporting Information may be found in the onlineversion of this article at the publisher’s web-site:

Fig. S1. The mean distribution and abundance of FA and PLlipids for each bacterial phylum surveyed in this study.Fig. S2. PCA scores plots of bacterial FA and PL profiles thatinclude chemotaxonomically significant lipids.Fig. S3. PCA scores plots of major FA and PL profiles undervarious respiration states coloured according to specieshabitat.Fig. S4. Mean lipid % differences between respiration statesof all bacterial phyla surveyed in this study.Fig. S5. A summary of mean lipid % differences betweenrespiration states of all surveyed bacterial phyla.Fig. S6. Circular relationship plots summarizing mean lipidvalues and habitat diversity for phyla with a single respirationstate.Table S1. Scaled FA and PL profiles of each species includ-ing species respiration and habitat(s) that were examined inthis study.



Respiration and ecological niche influence bacterial membrane lipid compositions

Documents