A comprehensive insight into the lipid composition of Myxococcus xanthus by UPLC-ESI-MS.

Analysis of whole cell lipid extracts of bacteria by means of ultra-performance (UP)LC-MS allows a comprehensive determination of the lipid molecular species present in the respective organism. The data allow conclusions on its metabolic potential as well as the creation of lipid profiles, which visualize the organism’s response to changes in internal and external conditions. Herein, we describe: i) a fast reversed phase UPLC-ESI-MS method suitable for detection and determination of individual lipids from whole cell lipid extracts of all polarities ranging from monoacylglycerophosphoethanolamines to TGs; ii) the first overview of a wide range of lipid molecular species in vegetative Myxococcus xanthus DK1622 cells; iii) changes in their relative composition in selected mutants impaired in the biosynthesis of α-hydroxylated FAs, sphingolipids, and ether lipids; and iv) the first report of ceramide phosphoinositols in M. xanthus, a lipid species previously found only in eukaryotes.

were of LC-MS grade. HPLC grade ammonia (Sigma) was used as a 10% aqueous solution and formic and acetic acid (both Carl Roth) were of analytical grade. Glyceryl tristearate ‫ف‬ 99% {TG (18:

Extraction procedures
Cells were harvested by centrifugation at 15,000 g for 15 min at room temperature and washed once with the same volume of distilled H 2 O.
A modifi ed method of Bligh and Dyer ( 22 ) was applied to extract all lipids. The cell pellets were thoroughly resuspended in 2 ml of distilled H 2 O and 7.5 ml of a mixture of MeOH/DCM 2:1 (v/v) was added. After incubation for 20 min under agitation, phase separation was induced by the addition of 2.5 ml of DCM and H 2 O, respectively. The lower phase was removed after centrifugation at 750 g for 20 min and the cells and upper phase was re-extracted twice with 2 ml DCM. The combined lower phases were fi ltered through solvent-resistant 0.22 M polytetrafl uoroethylene membrane fi lters and dried under a stream of nitrogen at 40°C in weighted vials for the exact determination of the mass of the extracts. For analysis, the lipid extracts were dissolved in a monoacylglycerophosphoethanolamines and ether lipids, as well as glycerolipids, needed to be monitored ( 12 ). The availability of a single analytical method, able to detect all lipid species, will facilitate rapid analogous investigations.
Results from TLC resolved 32 PO 4 3 Ϫ metabolic labeling studies available for Stigmatella aurantiaca ( 13 ), Myxococcus fulvus ( 14 ), and M. xanthus ( 15 ) provide some information about the nature and molar ratio of phosphate-containing lipids present in those strains. However, there is as yet no comprehensive information on the complete lipidome of myxobacteria. This is a prerequisite to visualize and evaluate the changes in lipid composition of mutant strains during vegetative growth or fruiting body formation, as well as providing a basis for further investigations on lipid biomarkers or signals for fruiting body formation.
We sought to develop a LC-MS method that: i ) is able to resolve as many lipid species as possible, covering lysophospholipids, glycerophospholipids, and glycerolipids; ii ) gives MS and MS 2 spectra for most of the molecular species in a single run for structure elucidation and identifi cation; iii ) involves the standard setup of ultra-performance (UP)LC-MS instrumentation used for routine analysis; iv ) solely uses readily available software; and v ) needs only a very short measuring time.
Many lipidomic methods that involve the direct measurement of lipid species from lipid extracts are based on direct infusion into ESI sources or ionization by MALDI ( 16,17 ), in which the detection of less abundant molecular species is typically impeded by ion suppression. This effect can be circumvented by separation of the analytes using chromatography ( 18 ). The development of a comprehen sive LC method for lipids is challenging due to their highly diverse physical properties, ranging from partly water (H 2 O) soluble (lysophospholipids) over amphiphilic (glycerophospholipids) to highly nonpolar (glycerolipids), nonionic [glycerolipids, ceramides (Cers)], and strongly pH dependent zwitter, cationic, and anionic molecules (glycerophospholipids). All chromatographic parameters (different eluents, nature, and concentration of additives, as well as temperature and pH) have to be carefully controlled over the whole chromatographic process. The use of eluent additives is limited, as they have to be compatible with MS coupling.
To date, many LC methods have been introduced that involve normal phase chromatography, hydrophilic/ lipophilic interaction chromatography, and reversed phase (RP) chromatography, in most cases to address particular analytical challenges (19)(20)(21). RP chro mato graphy, which can be regarded as "the" gold standard in routine small molecule analytics, is ideal for resolving molecules with even small differences in polarity, and it is also our method of choice, not least because it is part of our system's standard instrumentation.   +140 V (positive mode); nitrogen fl ow, 11.83 l/min; drying gas temperature, 300°C; and nebulizing gas pressure, 44 psi. Both MS and MS/MS data were recorded in the same run using the "ultra" scan mode with a trap drive of 64.3 (negative mode) and 53.3 (positive mode), respectively, a scan speed of m/z 25,000 per second and a mass resolution of m/z 0.6 over a range of m/z 100-2,000, with one spectrum having an average of seven scans in the MS mode and three scans in the MS 2 mode. For autoMS/MS fragmentation, an absolute threshold of 25,000 cts was chosen, and already fragmented precursor ions were excluded for further fragmentation after two spectra for 0.5 min. The isolation width was m/z 1.5. In a fi rst step, a LibraryEditor library was established to enable automatic identifi cation of lipid species from the lipid extracts. MS 2 spectra that were acquired during the runs were automatically extracted from the data set using the "Compound Ϫ autoMS(n)" algorithm with the following settings: intensity threshold positive, 1,000,000; intensity threshold negative, 10,000; maximum number of compounds, 350; retention time window (min), 0.25; spectrum type, line spectra only; background subtraction, none. Lipid species were identifi ed from the MS 2 spectra and added to the library together with their retention times. This library was applied to all HPLC-MS runs using the "Identify Mass Spectra" function of data analysis with the following settings: desired score, 900; minimum score, 600; minimum parameter matching score, 250; maximum number of spectra, 1; accepted retention time deviation, 0.1 min; maximum retention time deviation, 0.2 min.

Data processing
For the calculation of the relative percent of lipid species, extracted ion chromatograms (EICs) were created for every single precursor ion of the identifi ed lipid species. All chromatographic traces were smoothed using a Gaussian algorithm (1 pt width, 2 cycles) to enhance peak detection. Those EICs from the identifi ed lipid species were integrated by applying the "Compounds Ϫ Chromatograms" algorithm with the following settings: signal-tonoise threshold, 5; area threshold, 10%; intensity threshold, 10,000 for the negative ion mode and 1,000,000 for the positive ion mode; skim ratio, 0; smoothing width, 1. Charts containing the identifi ed autoMSn compounds and the areas under the curve (AUCs) of the EICs were exported into Excel charts. Those charts were manually inspected for false identifi cations and missing EIC peak integrations.

GC-MS FAME analysis
Derivatization procedures and GC-MS conditions are described elsewhere ( 23 ).

HPLC-MS conditions
For every data set, 5 l of lipid extract was separated on an ACQUITY UPLC TM BEH C18 column (Waters) with a length of 50 mm, an inner diameter of 2.1 mm, and a mean particle size of 1.7 M connected to an UltiMate 3000 (Thermo) HPLC system coupled to an AmaZon X (Bruker) mass spectrometer at a constant fl ow rate of 0.6 ml/min and a temperature of 50°C using a binary gradient. Eluent A consisted of H 2 O/ACN 50/50 (v/v) and solvent system B of ACN/isopropanol 10/90 (v/v). Both eluents contained 10 mM NH 4 AcO as a buffering additive and were adjusted to pH 5.5 using formic acid in order to maintain the analytes in the same charge state during the separation process. The gradient was programmed as follows: start at 1% B, increased to 95% within 10 min in a linear manner, hold at 95% for 2 min and re-equilibrate for 2 min at 1% B.
The ESI parameters for the negative ion mode were: capillary voltage, +4.5 kV (negative mode) Ϫ 4.5 kV (positive mode); end plate offset, Ϫ 500 V; capillary exit, Ϫ 140 V (negative mode) Relative AUCs given as per thousand of the total AUC of all identifi ed lipid molecular species determined separately for positive and negative ionization modes. MM (molecular mass) is the calculated molecular weight of monoisotopic mass of respective lipid species. See supplementary Tables I, II for details.  [PE(P)s], and CLs from the HPLC-MS runs in the negative ionization mode and the DGs and TGs in the positive ion mode were separately summed and normalized to 1,000, and the peak area of each individual EIC peak was calculated as per milliliter of the total integral of all peaks (see supplementary Table I and II for the complete data set and Table 1 for a summary). MATLAB release 2013b (The MathWorks, Inc., Natick, MA) was used for the visualization of the base peak full scan UPLC-ESI-MS chromatograms and plotting the lipid profi les. The base peak full scan UPLC-ESI-MS chromatograms were created by using the functions "meshgrid" and "surf". The "stem3" function was applied to create the 3D lipid profi les.

Identifi cation of lipids by MS
The experiments of the positive ionization and the negative ionization mode were combined in order to identify a maximum possible number of lipids. The above-mentioned standards were used in order to evaluate the chromatographic and fragmentation behavior of those lipid species. This information plus the reference spectra and software tools from LIPID MAPS ( 24 ) were used as the bases for the MS-based structure elucidation of the lipid species . The fragmentation behavior of PI-Cers have been described previously ( 25,26 ). ESI-MS data for the identifi cation and structure elucidation of PGs, PEs, and CLs, as well as glycerolipids (DGs and TGs) are also available (27)(28)(29)(30)(31)(32). The presence of PGs, PEs, and N -acylsphinganines (Cers), as well as the presence of distinctive ␣ -hydroxylated FAs together with the complete FA profi les of the investigated strains have been published ( 9,11,14,15  In general, isobaric molecular species could not be resolved by HPLC, and thus yielded one MS 2 spectra (see Tables 1, 3 ; supplementary Figs. I-VII ) For the creation of the lipid profi les, only the integrals from the negative ion mode for the PEs, PGs, PI-Cers, and Cers and from the positive ion mode for the DGs and TGs were used. PE and Cer ions were also clearly detectable in positive ion mode, but not used for the creation of lipid profi les.

High resolution MS analyses
Determination of the exact mass of Cer molecular species from lipid extracts was carried out using a Dionex Ultimate 3000 RSLC coupled to a Bruker micrOTOF-Q II equipped with an ESI source set to negative ionization mode. HPLC settings were identical to those described above, though the column bed was 100 mm in length and the gradient length was doubled to 20 min accordingly. The mass spectrometer was calibrated using a sodium formate calibrant solution (10 mM). MS data were acquired within the mass range of m/z 200-2,000.

Nomenclature
For the sake of uniformity and comprehensibility, we used the nomenclature of the LIPID MAPS Lipid Classifi cation System ( 24 ) for the names and abbreviations . RESULTS A binary eluent system consisting of H 2 O and ACN in solvent A and ACN and isopropanol in solvent B (for details see the Experimental section) proved to be useful for the separation and elution of an extensive range of lipid standards on a RP-18 column, including glycerolipids, glycerophospholipids, and sphingolipids with different degrees of acylation, as well as acidic, basic, and amphoteric polar head groups ( Fig. 1 ). The use of a 10 mM ammonium acetate buffer adjusted with formic acid to pH 5.5, a pH at which all major phospholipids have an equable charge state ( 34,35 ), gave an acceptable separation of the lipid extract of M. xanthus DK1622 ( 36 ) as depicted in

Lipid profi le of M. xanthus
Applying the described method to normalized lipid extracts resulted in two sets of LC-MS data for each sample: one recorded in the positive ionization mode and the other one in the negative ionization mode. Figure 2 shows Ϫ ( 45,46 ). Molecular composition was confi rmed by high-resolution MS (see Table 3 ). the resulting base peak chromatograms of the LC-MS analysis of lipid extracts from vegetatively grown DK1622 cells in negative ( Fig. 2A ) and positive ( Fig. 2B ) ionization modes, respectively. The most intense signals are discernable in the region of R t = ‫ف‬ 6.0-8.5 min and have m/z values of approximately 550-850 in negative ion mode, with another large group of signals occurring at R t = ‫ف‬ 9.0-10.5 min and having m/z values of 700-880 in positive ionization mode. Additionally, the most intense glycerophospholipid molecular species are discernable as dimeric ions, and there is an unknown compound class with a strong abundance at R t = ‫ف‬ 9.8 min with about m / z 1,008, which could not be identifi ed.
Evaluation of autoMS 2 data resulted in the identifi cation of a large number of PE and PG species, as well as the previously identifi ed PE(P)s ( 4 ) (supplementary Figs. I-III), hydroxylated and nonhydroxylated N -acylsphinganines (Cers) (supplementary Fig. IV), and interestingly, for the fi rst time in a prokaryote, PI-Cers [ Fig. 3 ; for a summary see Table 1 and supplementary Table I and II (the full data set obtained from data analysis)]. Consistent with the relative low abundance and limited diversity of sphingoid bases, as determined by GC-MS analysis of lipid methanolyzed cell extracts ( Table 2 ), they are detectable as less abundant compounds. Finally, CLs are detectable as late eluting compounds with rather low abundances (supplementary Fig. V).
Fragmentation spectra of the late eluting molecules in positive ionization mode are consistent with ammonium adducts of DGs and TGs ( 37 ) (supplementary Figs. VI, VII), as well as protonated Cers. PEs are the only detectable glycerophospholipid species and are distinguished by the [M-141+H] + MS 2 fragment resulting from the neutral loss of the polar headgroup. They were not further evaluated, as these fragment spectra do not contain any additional structural information.
None of the evaluated MS 2 spectra displayed evidence for other glycerophospholipids, glycerolipids, or precursors of identifi ed lipids such as lysophospholipids, diacylglycerophosphoserines, or PIs.
Concerning the elution pattern of these lipid species, the following regularities can be observed (see Fig. 2 ; Table 1 ;  supplementary Tables I, II): i ) As expected, retention time increases with increasing length of the acyl chains attached to the glycerophospholipids, as hydrophobic interaction with the RP-18 stationary phase becomes more intense. ii ) Elution order for glycerophospholipids is PGs → PEs → PE(P)s → CLs, as long as the number of carbon atoms and the degree of saturation in the acyl side chains are identical . iii ) PI-Cers elute earlier than Cers, whereas ␣ -hydroxylation in the acyl side chain further decreases the retention time. iv ) For DGs and TGs, analogously to the glycerophospholipids, retention time increases with acyl chain length and degree of saturation. v ) Lipids from the same class with an identical overall carbon and double bond number generally coelute. However, data from MS 2 fragmentation of glycerophospholipids in the negative ion mode does allow the assignment of fatty acyl residues to sn -1 or sn -2 positions of the glycerol backbone of those lipid species (supplementary Figs. I-III). In total, 3 Cers, 4 PI-Cers, 52 PEs, 2 PE(P)s, 20 PGs, 32 glycerophosphoglycerophosphoglycerols, and 7 DGs were identifi able. Additionally, 46 TG molecular species were identifi ed, but the exact assignment of all three fatty acyl moieties to the three hydroxyl functions of the glycerol backbone was not possible under these conditions, as only MS 3 fragments of the [M-NH 3 -R n COOH] + ions allow the assignment of the fatty acyl residues to a particular hydroxy group of the glycerol backbone. Nevertheless, the fragment spectra of TGs indicated a high diversity of those molecules comprising almost all saturated and unsaturated FAs present in M. xanthus . Beyond that, all TG ion species consisted of several isobaric TG species which coeluted, just as observed for the PGs and PEs (supplementary Fig. VIIC). We did not test the use of adduct-forming additives such as Li + and Na + because this would only further increase the number of ionic species and would not solve the problem of overlapping tandem mass spectra; the same applies to the CLs. Only MS 3 experiments allow the exact determination of the acyl residues present in the  V).

UPLC-MS analysis allows the assignment of a particular polar headgroup to Cers and detects loss of lipid species in mutant strains
We applied the UPLC-MS method and data evaluation procedure to lipid extracts of vegetatively grown cells from various strains with mutations in different genes involved in lipid biosynthesis. Strain ⌬ 0191 lacks a FA hydroxylase, which is responsible for the formation of 2-OH FAs ( Table 2 ). Strain Ϫ MXAN_3748 bears a mutation in a serine-palmitoyl-CoA acyltransferase homolog and is impaired in sphingoid base formation ( 9 ) ( Table 2 ). Figure 4A, B shows EICs of various Cer species from wild-type, ⌬ 0191, and Ϫ MXAN_3748 extracts. Strain ⌬ 0191 shows the selective loss of hydroxylated Cers ( Fig. 4A ). Strain Ϫ MXAN_3748 no longer forms either Cers or PI-Cers, as determined by both low resolution MS and confi rmative high resolution MS experiments ( Fig. 4B , Table 3 ).  Analogously, in strain Ϫ MXAN_1528 bearing a mutation in elbD , a multidomain enzyme involved in ether lipid biosynthesis ( 12 ), the relative abundance of vinyl ether species is strongly reduced ( Table 2 , Fig. 4C ).

Lipid profi les of various mutants defi cient in lipid biosynthesis
Integration of EIC peak areas of the monoisotopic masses of the respective lipid molecular species allows the construction of lipid profi les. If only the relative proportion of all lipid species are compared among the various investigated strains, no major differences are observable, apart from the above mentioned loss of certain lipid species and the expected relative increase of the remaining molecular species (supplementary Fig. VIII).
In order to gain an in-depth look into the lipid profi le of M. xanthus , we constructed a scatter plot comprising mass, retention time, and relative abundances of all analyzed lipid molecular species given as the sum of all peak areas per thousand ( Fig. 5 ; Table 1 ; supplementary Tables I, II). The observed pattern refl ects previous results from GC-MS analysis of FAMEs and neutral lipids with a high proportion of lipid molecular species containing 15:0 and 16:1. The occurring FAs are somewhat equally distributed among the various lipid classes with an almost "Gaussian"-like distribution among the TGs and PEs ( Fig. 5 ). Like Curtis et al. ( 11 ), we identifi ed numerous PE and also PG molecular species with unsaturated FAs attached to the sn -1 position of the glycerol backbone ( Table 1 ), indicating a less specifi c glycerol-3-phosphate 1-O-acyltransferase activity compared with Escherichia coli , where those lipids are less prevalent ( 32 ). Vinyl ether-containing phospholipids were exclusively detected within the PE molecules, whereas O -alkylglycerols could not be detected under these conditions, which is consistent with previous results ( 4 ). The relatively low proportion of PGs is consistent with results from Orndorff and Dworkin ( 15 ), who specifi ed the amount of PGs as 9% of total 32 PO 4 3 Ϫ -labeled M. xanthus membranes . Analogously, CLs are minor compounds, but exhibit a high molecular diversity, as each molecule is potentially made up of four different FAs ( Table 1 ). Neutral lipids, especially TGs, gave strong signals as ammonium adducts, although they are less prevalent in vegetative cells ( 10 ). Their molecular diversity seems to be even higher than that of the polar lipids.
In an additional step, we created scatter plots that compare the relative changes in the abundance of the individual lipid species between the mutant lipid profi les and the wild-type profi le. This refl ects the changes in the abundance of individual molecular species in the analyzed mutant strains compared with the wild-type, displayed in a 3D way. Now, shifts and trends in the mutant profi les are able to be visually appraised ( Fig. 6A-C , supplementary Table I).
The lipid profi le of strain ⌬ 0191 versus wild-type shows the expected relative increase of nonhydroxylated Cers in addition to the aforementioned loss of hydroxylated Cer molecular species and a general increase in the abundance of PG molecular species ( Fig. 6A , Table 1 ). There is a conspicuous increase in PE(15:0/17:0) and a simultaneous strong decrease in the abundance of PE(15:0/15:0) and PE(15:1/15:0). These fi ndings match well with the FAME pattern of strain ⌬ 0191, which shows a higher proportion of i17:0 FA compared with the wild-type ( Table 2 ). The changes in abundance within the TG molecular species seem to be rather random, but could be attributed to a higher ratio of iso17:0 to iso15:0 FAs . Figure 6B visualizes the changes in the lipid profi le of Ϫ MXAN_3748 compared with the wild-type. A general increase in higher molecular weight lipid species within the PE, PG, and TG molecular species, as well as a greater proportion of PE(P), is observable, meaning that there is a general increase in longer and/or less saturated FAs that could compensate for the loss of the lipophilic Cer species.

DISCUSSION
Such a comprehensive data set about molecular lipid species in myxobacteria has not been published before. Curtis et al. ( 38 ) reported the analysis of lipid molecular species from the PE fraction of M. xanthus by HPLC-ESI-MS, resulting in the identifi cation of 26 PE molecular species; though for four of these, no assignment of the fatty acyl residues to the sn-1 and sn-2 position of the glycerol backbone was achieved. Our UPLC-MS analysis confi rmed that ␣ -hydroxylation of FAs and sphingolipid formation are indeed metabolically linked, as the lipid extract of the ␣ -hydroxylase deletion mutant ⌬ 0191 lacks the hydroxylated sphingolipids and the spt -disruption mutant Ϫ MXAN_3748 is devoid of any detectable sphingolipids and does not contain other lipids with an ␣ -hydroxylated FA. When the results concerning the abundance of neutral lipids are compared with those obtained by high temperature GC-MS analysis ( 10 ), it seems that the latter method results in overestimation of the presence of MGs and DGs. This is very likely due to the derivatization procedure with N-Methyl-N-(trimethylsilyl) trifl uoroacetamide , which converts the free hydroxy groups of the MGs and DGs into easily ionizable trimethylsilyl derivates. In contrast, identical amounts of TGs yield more intense ESI-MS signals than underivatized MGs or DGs ( Fig. 1 ). In the end, the creation of calibration curves using molecular standards for every lipid class will be necessary to determine their individual response factors. This will be the only way to establish the exact ratio between different lipid classes. Although not further evaluated, it is interesting that the TG molecular species seem to contain fatty acyl species that are not detectable in FAME-GC-MS analysis, as the [M-NH 3 -R n COOH] + fragment at m/z 521.6 in the mass spectrum shown in supplementary We did not detect any glycerophosphoserines or lysophospholipids during our analysis, although these molecules are putative precursors of the highly abundant PEs or their degradation products, respectively. These lipids were detected in 32 PO 4 3 Ϫ -labeled TLC-resolved membrane fractions of M. xanthus , but not in its whole cell lipid extract ( 15 ), leading us to conclude that they are products of ongoing enzymatic reactions during the lengthy membrane fractionation procedure. We cannot exclude the presence of traces of these compounds with abundances lower than the limit of detection. The use of a column with a higher resolution and different MS experiments and instrumentation may lead to the identifi cation of more lipid species. However, trace component analysis was not the intention of this study.
At the same time, the low abundance of CLs, as determined by our ESI-MS analyses, is consistent with the low abundance of CLs found in the 32 PO 4 3 Ϫ -labeled cells ( 15 ). The presence of PI-Cers is an interesting fi nding of this study, as this is a class of lipid known to be present in fungi such as Saccharomyces cerevisiae ( 39 ) and Aspergillus niger  ( 40 ). They have also been found in protozoan pathogens like Trichomonas brucei ( 41 ), Trypanosoma brucei ( 41 ), Leishmania major ( 42 ), and plants ( 43 ). However, to our knowledge, the occurrence of bacterial PI-Cers has never been previously described. It seems that sphingolipids are a common component of myxobacterial membranes, as Cers were found in Cystobacter fuscus ( 44 ), Cer phosphoethanolamines in Myxococcus stipitatus ( 45 ), and Cer phosphoethanolamines, ␤ -D-glucosylsphingenines, and even phosphonosphingolipids in Sorangium cellulosum So ce56 ( 46 ), where they serve together with ornithine-containing lipids and glycerol ether lipids as substitutes for missing lipopolysaccharides in the outer membrane ( 46 ).
In S. cerevisiae , PI-Cers are synthesized by the inositol phosphorylceramide synthase catalytic subunit (AUR1), an enzyme that transfers a phosphoinositol headgroup to ␣ -hydroxyceramides from a diacylglycerophosphoinositol moiety ( 47 ). Protein MXAN_0451 (annotated as a PAP2 family protein) is 29% homologous to AUR1 (E-value = 2 Ϫ 20 ), a further indication of the capability of M. xanthus to synthesize this lipid species. Interestingly, no diacylglycerophosphoinositols were detected in the lipid extracts, though it has been previously noted that a myo -inositol-1-phosphate synthase homolog (MXAN_0452) as well as a CDP-alcohol phosphatidyltransferase homolog (MXAN_0450) are encoded next to MXAN_0451 ( 48 ). This may be due to a fast conversion of diacylglycerophosphoinositols to PI-Cers as the occurrence of diacylglycerophosphoinositols are indeed described for M. fulvus and S. aurantiaca ( 13 ).
In general, glycosylated Cers are important signaling molecules in multicellular organisms ( 49 ), but no specifi c functions for PI-Cers are known in lower eukaryotes, except that inhibition of AUR1 is lethal to S. cerevisiae ( 50 ) and that PI-Cer is at least indirectly associated with the pathogenicity of Cryptococcus neoformans ( 51 ). A possible signaling function of PI-Cers in myxobacterial fruiting body formation or sporulation remains to be investigated. Although there have been no systematic investigations on putative signaling functions of Cers in bacteria, it was noted that sphingolipids (as determined by the sphingobases) accumulate in developing M. xanthus cells, but are dispensable for fruiting body formation ( 9 ).
In summary, the changes in the lipid profi les between wild-type and mutant strains can be used to unveil genes and their protein products that are involved in lipid biosynthesis. An overview about gene candidates with homology to genes involved in lipid biosynthesis from other organisms exists in the literature ( 48 ). Even when there is functional redundancy between paralogous genes, it is possible to deduce protein functions from shifts in the lipid profi les of the respective mutant as shown in Fig. 6A for the ⌬ 0191 mutant, in which the biosynthesis of a polar headgroup is not affected.
We are aware that fi nal statements concerning changes in lipid profi les are not possible without statistically analyzable data, although some observable changes, like the previously established decrease or loss of selected molecular species and the easily explained general increase in degree of saturation and chain length of FAs within the lipids suggest that our data analysis procedure is generally a valid approach. Moreover, the visualization of the data using scatter plots resolved by the lipid class strongly facilitates data inspection and interpretation. This becomes even more important with increasing numbers of analytes. We envision that refi nement of this method using appropriate internal standards and statistics will yield lipid profi les that can be linked to transition profiles, diffusibility of membranes, activities of selected membrane proteins and membrane proteomes, and allow us to unravel the multifactorial links between those properties and the lipid composition of membranes.