Changes in the Distribution of Membrane Lipids during Growth of Thermotoga maritima at Different Temperatures: Indications for the Potential Mechanism of Biosynthesis of Ether-Bound Diabolic Acid (Membrane-Spanning) Lipids

ABSTRACT Membrane-spanning lipids are present in a wide variety of archaea, but they are rarely in bacteria. Nevertheless, the (hyper)thermophilic members of the order Thermotogales harbor tetraester, tetraether, and mixed ether/ester membrane-spanning lipids mostly composed of core lipids derived from diabolic acids, C30, C32, and C34 dicarboxylic acids with two adjacent mid-chain methyl substituents. Lipid analysis of Thermotoga maritima across growth phases revealed a decrease of the relative abundance of fatty acids together with an increase of diabolic acids with independence of growth temperature. We also identified isomers of C30 and C32 diabolic acids, i.e., dicarboxylic acids with only one methyl group at C-15. Their distribution suggests they are products of the condensation reaction but are preferably produced when the length of the acyl chains is not optimal. Compared with growth at the optimal temperature of 80°C, an increase of glycerol ether-derived lipids was observed at 55°C. Our analysis only detected diabolic acid-containing intact polar lipids with phosphoglycerol (PG) head groups. Considering these findings, we hypothesize a biosynthetic pathway for the synthesis of membrane-spanning lipids based on PG polar lipid formation, suggesting that the protein catalyzing this process is a membrane protein. We also identified, by genomic and protein domain analyses, a gene coding for a putative plasmalogen synthase homologue in T. maritima that is also present in other bacteria producing sn-1-alkyl ether lipids but not plasmalogens, suggesting it is involved in the conversion of the ester-to-ether bond in the diabolic acids bound in membrane-spanning lipids. IMPORTANCE Membrane-spanning lipids are unique compounds found in most archaeal membranes, but they are also present in specific bacterial groups like the Thermotogales. The synthesis and physiological role of membrane-spanning lipids in bacteria represent an evolutionary and biochemical open question that points to the differentiation of the membrane lipid composition. Understanding the formation of membrane-spanning lipids is crucial to solving this question and identifying the enzymatic and biochemical mechanism performing this procedure. In the present work, we found changes at the core lipid level, and we propose that the growth phase drives the biosynthesis of these lipids rather than temperature. Our results identified physiological conditions influencing the membrane-spanning lipid biosynthetic process, which can further clarify the pathway leading to the biosynthesis of these compounds.

M embrane lipids play a fundamental role in forming the physical barrier surrounding the cells, the cytoplasmic membrane. In Bacteria, membrane lipids are commonly composed of glycerol, derived from glycerol-3-phosphate (G3P), attached via ester bonds to fatty acid molecules, forming a bilayer structure. On the other hand, the opposite stereoisomer of glycerol, derived from glycerol-1-phosphate and linked to isoprenoid alkyl chains via ether bonds, is characteristic for the membrane lipids of Archaea (1,2). Most of the archaeal lipid membranes are organized as monolayers of lipids spanning the membrane (i.e., membrane-spanning lipids), formed of isoprenoidal tetraether lipids called glycerol dialkyl glycerol tetraethers (GDGTs), thought to be more rigid and impermeable than the bilayer membranes from bacteria and eukaryotes (3). However, some bacterial groups also possess membranes organized as monolayers that contain lipids resembling the archaeal GDGT lipids but with the bacterial G3P stereochemistry and comprised of branched alkyl chains rather than isoprenoidal chains.
Previous reports on bacteria from the order of Thermotogales have described the core lipid composition composed of straight-chain fatty acids (FAs) (predominantly C 14 and C 16 ) as well as the C 30 and C 32 DAs (5,6), which are arranged as tetraesters, mixed ether-ester compounds, or tetraethers (7), and detected diacylglycerol lipids bearing unusual polar head groups (glycolipid 1 and glycolipid 2; Fig. 1 shows the structure) (12). Although these studies have provided an excellent overview of the membrane lipid composition itself, changes in the relative distribution of the membrane lipids due to membrane lipid adaptation to physiological conditions remain to be explored. Furthermore, it remains unclear which intact polar lipids (IPLs; i.e., including the polar head group) contain DAs in members of the Thermotogales.
From an evolutionary perspective, the formation of membrane-spanning lipids in bacteria is an open question and concerns the evolutionary divergence of the lipid composition, or lipid divide, of all cellular forms of life, as these membrane lipids have mixed archaeal and bacterial traits that could enlighten how this division occurred (2,13,14). To tackle this question, the enzymatic steps leading to the formation of the bacterial membrane-spanning lipids need to be elucidated. The molecular mechanism that leads to the biosynthesis of membrane-spanning lipids in bacteria involves the condensation of two FA molecules (15,16). Although the enzymatic activity catalyzing this coupling mechanism has been detected using extracts of intact or disrupted cells of anaerobic bacteria (16,17) or other (iso) diabolic acid producers (16), the enzyme(s) involved remains unknown. Likewise, the bacterial enzymes responsible for the transformation of the ester bond between the glycerol and DA into an ether bond have not been elucidated. To constrain the missing pieces of the puzzle leading to the formation of bacterial membranespanning lipids, physiological studies involving factors that could affect the membrane lipid composition are required.
Here, we identify specific physiological conditions resulting in differences in lipid membrane composition of a membrane-spanning lipid producer of the bacterial order of Thermotogales, T. maritima. The hyperthermophilic bacterium T. maritima can grow from 55°C up to 90°C with an optimum at 80°C (18). We determined the composition of core and intact polar membrane lipid of T. maritima at optimal (80°C) and suboptimal (55°C) temperature across growth phases, which resulted in differences in cell morphology at the different temperatures independent of similarities in lipid composition. Based on these findings and on genomic analyses, we identified a potential homologue of the gene involved in the formation of ether bonds in the diabolic acids of T. maritima, and we propose a hypothetical pathway for the biosynthesis of these membrane-spanning lipids.

RESULTS
Cultivation experiments with T. maritima at optimal (80°C) and suboptimal (55°C) growth temperatures were performed in batch cultures, and lipid and microscopic analysis was performed at different (early exponential, exponential, and stationary) growth phases.
Isomers of the C 30 and C 32 DAs and the C 30 and C 32 DA glycerol monoethers possessing a similar mass spectrum eluted approximately 0.4 to 0.6 min later from the corresponding DA or DA glycerol monoethers. These isomers have been recognized before in cultures of the Thermotogales (7). Careful inspection of the mass spectrum of the isomer eluting after the C 30 DA suggested that it represents a C 30 dicarboxylic acid dimethyl ester with only one methyl group at C-15 ( Fig. 2A). To confirm this, the diacids were converted to diols by treatment with LiAlH 4 and then to hydrocarbons by treatment with HI and subsequent hydrogenation. Indeed, the mass spectrum of the hydrocarbon eluting after 13,14-dimethyloctacosane was consistent with 15-methylnonacosane, characterized by a relatively strong m/z 224 fragment ion in its mass spectrum (Fig. 2C). Its measured Kovats retention index (RI; 2,933) is in good agreement with the calculated RI value (2,930) (19,20). Hence, the C 30 diacid isomer was assigned as a 15-methylnonacosanedioic acid. Similarly, the C 32 diacid isomer was assigned as a 15-methylhentriacontanedioic acid. The mass spectrum of its dimethyl ester derivative (Fig. 2B) revealed characteristic fragment ions at m/z 238 and 266. The mass spectrum of the corresponding hydrocarbon showed relatively strong fragment ions at m/z 224 and 252 (Fig. 2D), and the measured RI (3,130) is in good agreement with   the calculated RI value (3,130) for 15-methylhentriacontane. For the C 32 diacids, monomethyl diacid occurred only in relatively low abundance at optimal growth temperature; the ratio of the monomethyl and dimethyl diacids was ca. 0.1 (Fig. 3B, Table 1). Remarkably, however, this ratio for the C 30 dimethyl and monomethyl was much higher, at ca. 0.7, during all analyzed growth phases. A similar observation was made for the corresponding glycerol derivatives with a ratio of 0.1 for the C 32 and 0.6 for the C 30 diacids (Fig. 3B). Similar ratios were observed when cells were grown at suboptimal growth temperatures (Fig. 3B, Table 1). Distribution of lipid components across growth phases. The growth of T. maritima was monitored at the optimal growth temperature (80°C) with a doubling time of approximately 19 h (Fig. 4). The distribution of the FA and glycerol monoethers changed across growth phases (Fig. 3A). The dominant FA was C 16 , accounting for 43% of the total lipids in the early exponential phase, decreasing to 33% during the exponential phase and 22% by the stationary phase ( Table 1). The C 14 FA increased from 2.2 to 4.4%, and the C 18 content decreased from 13 to 1.3% between the early exponential and the stationary phase. Conversely, the major C 32 diacids increased from 28% in the early exponential phase to 42% in the exponential phase and 46% in the stationary phase, and the C 30 diacids increased from 1.3 to 3.2% between the early exponential and stationary phases. The corresponding C 32 and C 30 monomethyl diacids also increased from 2.3 to 4.8% and from 0.8 to 2.2%, respectively, between the early exponential and stationary phases.
The glycerol ether compounds consisted of the C 16 glycerol monoether, which decreased from 1.1 to 0.3% across growth phases, and the C 30 and C 32 DA monoethers. The relative abundance of these components increased from 0.5 to 1.7% and 6% to 11%, respectively, while the corresponding monomethyl diacid C 30 and C 32 monoethers also increased from 0.3 to 1.2% and from 1.0 to 1.2% at the same growth interval ( Table 1).
The intact polar lipid (IPL) composition of T. maritima grown under optimal conditions (80°C) was analyzed from freeze-dried cells harvested at three different growth phases, resulting in the detection and identification of a range of IPLs (Table 2; Fig. 1 shows structures). A number of the IPLs were non-membrane-spanning, including phosphatidylglycerols (PGs) with a diacylglycerol (DAG) core (C 16 /C 16 ) and two PGs with a dietherglycerol (DEG) core (C 16 /C 14 and C 16 /C 16 ). Their fractional abundance decreased from 2.4 to 0.1% for the DAG core and increased from 1.9% to 8% for the DEG core across the and C 32 MM-GE (15-methyl-31-glyceryloxyhentriacontanoic acid). Bars represent the mean fractional abundance. The error indicated shows the standard deviation from three biological replicates. ANOVA significant statistical differences (P , 0.0001) between the early exponential and the stationary phases on the C 16 and C 32 lipids across growth phases at each temperature condition are shown. (B) The monomethyl/dimethyl diacid ratio and their glycerol ether derivatives as a function of number of carbon atoms of the diacid and growth temperature and phase. E. Exp, early exponential. three growth phases (Table 2). Additionally, PGs with a mixed acyl-etherglycerol (AEG) core (C 16 /C 14 and C 16 /C 16 ) were detected; their sum decreased from 6.9 to 0.9%. A range of lipids with dihexose head groups was detected ( Table 2). These are presumably diglucosyls (diGlu) based on previous detailed analysis of the glycolipids of T. maritima using nuclear magnetic resonance (NMR) spectroscopy (12). Four diGlus with DAG cores (C 16 /C 14 , C 16 /C 16 , C 16 /C 18 , and C 18 /C 18 ) decreased from 14.8 to 10.2% across the three growth phases, while the abundance of a single diGlu with an AEG core (C 16 /C 16 ) showed little change (i.e., from 1.0 to 2.8%). A minor IPL with a monohexose head group was assigned as a monoglucosyl (Glu) with a DAG core (C 16 /C 16 ) and increased slightly in abundance from 0.3 to 0.5%. Additionally, four IPLs with an unusual polar head group, a decanoyl-diGlu described previously in T. maritima by Manca et al. (12) (Fig. 1), were detected (see Table S1 in the supplemental material for accurate mass of fragments). Two decanoyl-diGlus had C 16 /C 14 and C 16 /C 16 DAG cores and their sum increased from 7.3 to 8.7% over the growth phases, and two had C 16 /C 14 and C 16 /C 16 AEG cores and their sum increased modestly from 5.9 to 8.3% (Table 2). Membranespanning IPLs were also detected but only with PG head groups. Fifteen different diglycerols, with a C 30 /C 30 , C 30 /C 32 , and C 32 /C 32 core, ranging in their ether/ester composition between tetraether and tetraester, were identified ( Table 2). All the PGs with a tetraether C 30 /C 30 , C 30 /C 32 , and C 32 /C 32 core, triether/monoester, monoether/triester C 30 /C 30 , C 30 /C 32 , and diether/diester C 30 /C 30 , C 30 /C 32 cores increased in their fractional abundance from 15.6 to 25.3%. Those with monoether/triester, tetraester, and diether/diester containing the C 32 /C 32 core decreased in their fractional abundance from 43.2 to 32.6% (Table 2) over the three growth phases. C 30 and C 32 DAs attached to a glycerol moiety also were detected, and their fractional abundance increased from 0.5 to 2.5% with time ( Table 2). Potentially these compounds are formed from the membrane-spanning IPLs by loss of the head group.
Cultivation of T. maritima at suboptimal temperature (55°C) resulted in a doubling time of approximately 32 h (Fig. 4). As observed at optimal growth temperature, the relative distribution of FA and DA core lipids changed significantly across growth phases (Fig. 3A). The fractional abundance of the C 16 FA accounted for 46% at the early exponential phase and decreased to 21% in the stationary phase ( Table 1). The C 14 FA increased gradually from 2% to 4% while the C 18 FA decreased gradually from 15% to 2% across growth phases. The relative abundance of the C 30 diacids increased from 0.5 to 2.5%, while the C 32 diacids increased from 13% to 23% and increased up to 39% during the batch cultivation. The fractional abundance of the corresponding C 32 and C 30 monomethyl dicarboxylic acids also increased across growth phases from 0.8 to 3.7% and from 0.3 to 1.6%, respectively. The fractional abundance of the C 16 glycerol monoether decreased from 3.4 to 0.4%, whereas those of the C 30 and C 32 diacid glycerol monoethers increased from 1% to 3% and from 15% to 19%, respectively, from the early exponential to stationary phase of growth. The fractional abundance of  the corresponding monomethyl C 30 and C 32 monoethers also increased from 0.6 to 2.0% and from 1.2 to 2.1% at the same growth interval (Table 1, Fig. 3A). The same IPL species as those observed under the optimal growth conditions (80°C) were detected in T. maritima at suboptimal growth conditions (55°C). The fractional abundance of non-membrane-spanning PG-DAGs decreased, from 8% to 0%, as did the PG-AEGs, from 9 to 0.3%, while the fractional abundance of the PG-DEGs increased across the three growth phases to the same extent as the observed at 80°C, from 0.3% to 9% ( Table 2). The fractional abundance of the four diGlu-DAGs was much higher at the early exponential phase but decreased substantially, i.e., from 32% to 11%. The single diGlu-AEG increased, from 1.8 to 3.2%, as did the Glu-DAG, from 0.1 to 0.9%. The fractional abundance of the membrane-spanning PGs with tetraether, triether/monoester, triester/monoether, and diether/diester C 30 /C 30 , C 30 /C 32 , and C 32 /C 32 cores increased substantially, from 18% to 49%, while that of the membrane-spanning PGs with a tetraester C 32 /C 32 core decreased from 19% to 8% (Table 2) over the three growth phases. Finally, the C 30 and C 32 diacids attached to a glycerol moiety were not detected at the first growth stage, but fractional abundance was ca. 3% at the end of the growth curve, a scale increase similar to that seen at 80°C.
Cell morphology and membrane integrity. To determine changes in the morphology and the membrane integrity of T. maritima cells associated with changes in the membrane lipid composition, we used fluorescence microscopy and image analysis on cells collected at the exponential and stationary phase. A specific membrane dye (FM4-64) allowed the visualization of cell lipid boundaries and morphologies (Fig. 5), and the stained DNA (49,6-diamidino-2-phenylindole dihydrochloride [DAPI]) was utilized to quantitatively measure various FIG 5 Fluorescence microscopy of Thermotoga maritima at optimal (80°C) and suboptimal (55°C) temperature growth conditions. T. maritima cells from exponential and stationary phases at optimal temperature and exponential-and stationary-phase cells grown at suboptimal temperature are shown. Cellular profiles were imaged after the selection of the appropriate filter for each specific stain. Membrane stain was FM4-64 (red). DNA stains were membrane permeable, DAPI (blue) membrane-impermeable, and SYTOX (green). Scale bar, 5 mm. The arrows indicate coccoid cells. morphological cell parameters (area, length, and circularity). To analyze the cell membrane permeability properties, fluorescence intensity values were calculated from imaged cells simultaneously stained with membrane-permeable (DAPI) and membrane-impermeable (SYTOX green) DNA dyes.
Cells visualized with the membrane dye FM4-64 showed the inner and outer membranes combined in one strong signal surrounding the edges of the cell. At the exponential phase, T. maritima culture was composed mostly of cells with a rod shape (Table 3). Two morphologies were detected at the stationary phase, being rod-shaped (Fig. 5) and coccoid cells. The latter were mostly detected during prolonged stationary phase (Fig. S1). Cells from stationary phase showed a significant decrease in the area from 1.3 to 0.9 mm 2 , from 2.3 to 1.8 mm in length, and increased their circularity values from 0.5 to 0.6 compared to the exponentialphase cells (Table 3, Fig. S2). The detected fluorescent signal for the DNA dyes showed no significant changes between the growth phases ( Table 3, Fig. S2).
Cells grown at 55°C and visualized with the membrane dye FM4-64 showed a fluorescent signal belonging to the inner and outer membrane, mainly distributed over the cell ends, forming the toga's membranous structure surrounding the cytoplasmic membrane (Fig. 5). Bacterial cell profiles at the exponential phase were dominated by elongated rods, and at the stationary phase, the main morphology detected was the rod-shaped bacteria (Fig. 5, Table 3). Cells from stationary phase showed a significant decrease in area from 3 to 2.2 mm 2 , from 4.1 to 3.5 mm in length, and with no significant changes in their circularity values compared to the exponential-phase cells (Table 3, Fig. S2). The detected fluorescent signal for both DNA dyes showed significant increased values, from 65 to 81 arbitrary units (AU) of intensity for DAPI and from 35 to 61 AU for SYTOX green within the growth phases (Table 3, Fig. S2).

DISCUSSION
Dynamic biosynthesis of membrane-spanning lipids across growth phases. In this study, we confirmed the presence of previously reported core lipids (Table 1), identified new lipids (Tables 1 and 2), and extended our analyses to the core lipid composition at optimal and suboptimal growth temperatures in T. maritima across growth phases. A previous study on the stable carbon isotopic fractionation associated with the biosynthesis of FAs in T. maritima across growth phases showed a decrease in the relative abundance of the C 16 FA concomitant with an increase of higher-molecular-weight lipids that were speculated to be DAs (21). Our core lipid analysis provides compelling evidence that the C 16 FAs are utilized as building blocks for the synthesis of C 32 DAs at both optimal and suboptimal temperature growth conditions. Statistical analysis (ANOVA) showed the relative abundance of C 16 FA and C 32 DA to be significantly different (P , 0.0001) between the early exponential and stationary phases with a marked decrease of the relative abundance of the C 16 FA and a concomitant increase of the relative abundance of the membrane-spanning C 32 (Fig. 3A). This strongly supports the direct relationship between substrate and product across growth phase, independent of growth temperature.
The core lipid analysis also revealed that the lipid composition at the stationary phase is dominated by DAs (Fig. 3A) with an overall degree of membrane-spanning lipids, increasing from 41% to 71% at optimal growth temperature and from 32% to 71% at a Data represent the mean 6 SD, n = 3 independent images with 100 cells measured on each image. AU, arbitrary units.
suboptimal temperature (Table 1). In their natural environment, bacterial cells are predominantly found in the stationary phase or limited growth rather than exponential phase (22,23), undergoing changes in morphology and cell physiology (24,25). Based on the high proportion of C 32 DAs compared to the C 16 FAs in the stationary phase, we therefore hypothesize that the membrane-spanning DAs probably play an essential role in determining the physical properties of the membrane in natural T. maritima communities and their fitness and survival. A former labeling experiment of the DA producer Butyrivibrio fibrisolvens (15) with [16-2 H 3 ]-palmitic acid (C 16 ) and [14-2 H 2 ]-palmitic acid demonstrated that the C 32 DA is biosynthesized by a condensation reaction between two C 16 FAs at the v-1 position. This must be induced by abstraction of a hydrogen at C-15, indicating that this most likely is an enzyme-mediated reaction, as the resulting DA consists of only one stereoisomer (15). A C 30 DA would likewise be biosynthesized by a similar condensation of a C 14 and a C 16 FA. In our analysis of T. maritima lipids, we identified the monomethyl C 30 and C 32 diacids. Their molecular structure indicates that instead of being formed by condensation of the two v -1 carbon atoms of the FA precursors, these compounds are likely produced by condensation of v-1 and the v position of the two FAs. The ratio of the C 30 mono-and dimethyl diacids was ca. 0.67 at both growth temperatures (Fig. 3B, Table 1), suggesting only a slight preference for the v -1/v -1 condensation. Conversely, for the C 32 diacids this ratio is ca. 0.1 (i.e., 7-to 8-fold lower; Fig. 3B, Table 1). A similar trend was observed for the monoether-bound diacids (Fig.  3B, Table 1). Based on this observation and the relatively low abundance of the C 30 compounds, we hypothesize that the enzymatic condensation leading to a membrane-spanning lipid is specifically designed for the v -1/v -1 condensation of two C 16 alkyl chains. With one shorter (C 14 ) chain instead of the two C 16 chains, the condensation is less specific and results in much more by-product. Note that despite the presence of some C 18 FAs (Table 1), we never identified DAs based on C 18 alkyl chains in T. maritima lipids.
The enzymatic activity mediating this reaction has been confirmed in membrane enrichments and cell extracts of Butyrivibrio and Sarcina (17). Although this specific carboncarbon bond formation between the membrane FAs depends on enzymatic activity, the enzyme or the biosynthetic complex has not been yet identified.
The IPL analysis revealed a range of the membrane-spanning core lipids derived from C 30 /C 30 , C 30 /C 32 , and C 32 /C 32 diglycerol compounds. Interestingly, all the C 30 /C 30 , C 30 /C 32 , and C 32 /C 32 diglycerols detected were bound to a phosphoglycerol (PG) head group at one side of the membrane-spanning molecule (Fig. 6), with no polar head group attached to the other glycerol moiety. The presence of the PG head group bound to one side of the C 30 /C 30 , C 30 /C 32 , and C 32 /C 32 diglycerol spanning lipids (Fig. 6) strongly suggests that the biosynthesis of the DA membrane-spanning lipids depends on preformed diacylglycerol lipids, the biosynthesis of a PG phospholipid intermediate (Fig. 6), and a membrane-bound enzyme. These observations suggest that the condensation reactions resulting in the formation of membrane-spanning lipids take place when the membrane lipid bilayer is already in place.
We also investigated the membrane lipid biosynthetic pathways in the genome of T. maritima (Table 4). The acylation of the bacterial G3P with specific FA is mediated by the activity of the PlsX/PlsY acyltransferase system followed by the PlsC acyltransferase, giving rise to the phosphatidic acid intermediate (PA) (Fig. 6). Subsequently, PA is activated by CDP-DAG synthase (CdsA) and transformed into glycerol-phosphate by the activity of the PgsA enzyme, converting PA to phosphatidylglycerol phosphate (PGP). Finally, the removal of the inorganic phosphate (P i ) (probably by an unidentified phosphatase) would lead to the anticipated intermediate for the membrane-spanning IPLs, i.e., the PG phospholipid (Fig. 6). It is conceivable that through these steps the PG phospholipid could be synthesized de novo, and the two FA molecules of C 16 (or C 14 ) required for the synthesis of DAs would be incorporated into the glycerol backbone (Fig. 6), leading to C 32 /C 32 with the PG head group (Fig. 1  and 6) on one side of the compound. Similarly, the FAs attached to the diglyceride backbone found on the opposite side of the membrane-spanning PG compound could be formed through the mentioned steps forming the PA with the diacylglycerol (DAG) precursor, followed by the consequent removal of the phosphate group by a phosphatase. Other metabolic pathways could lead to the formation of the DAG compounds in the cell (26), but it is possible that the IPL glycolipids found on T. maritima also serve as the precursor of DAG prior removal of the Glu or diGlu head group.
The anionic IPL PG is known to be one of the principal membrane components of bacteria and is required for its proper function and stability (27). Thus, its formation and distribution across the surface are essential for bacterial survival and cellular organization. The  data provided by our study suggest that the hypothetical biosynthetic pathway to produce complex DA-containing lipids involves de novo formation of PG and might be selective for the anionic head group of PG. Since the terminal carbon ends of the FAs are known to be embedded in the hydrophobic core of the membrane, we hypothesize that the enzymatic process implicated in this coupling reaction is catalyzed by a membrane protein that can easily bind or have access to the lipid substrate. Moreover, our finding on the monomethyl diacids and the increased ratio for the C 32 diacids compared to the C 30 diacids (Fig. 3B), together with the low abundance of the C 30 compounds, suggests that the enzymatic activity leading to the membrane-spanning lipids is selective in terms of substrate requirements (i.e., chain length and head group).
Glycerol ether lipid formation. As cultures reached the stationary phase, a slight increase in the amount of hydrolysis-derived ether bonds was detected at optimal growth conditions (from 4% to 7%) ( Table 1). In contrast, growth experiments at suboptimal temperature showed that the percentage of ether bonds was slightly higher at ca. 10% and remained constant across growth phases. This may represent an adaptation to low temperatures, although typically, the presence of ether bonds in membrane lipids is interpreted as an adaptation to high temperatures, since ether bonds are thought to be more stable than ester bonds (28). In the case of T. maritima, however, these could be the result of adaptation to stressful growth conditions as its optimal temperature of growth is 80°C, while no growth below 55°C is detected (18).
The IPLs detected across growth phases under both temperature conditions showed that the fractional abundance of the sum of IPLs with DAG cores decreased from the early exponential to the stationary phase. In contrast, the IPLs with DEG core lipids showed an apparent increase in their fractional abundance. The mixed cores containing AEG, DAG/DEG, triether, and triester combinations remained constant at 80°C, while at 55°C they increased. Taking these data together, we can speculate that the concomitant decrease on DAG (ester) cores with the increase on DEG (ether) cores is the result of a growth phase-dependent synthesis of ether lipids from ester compounds and that ester lipids can be subsequently converted into ether lipids. Although it is not possible to precisely attribute the changes between ester and ether in the mixed cores to confirm the growth phase dependence for ether lipid formation, our results indicate that the synthesis of ether lipids are not a requirement to make membrane-spanning DAs, as we were able to detect mixed DAG/AEG, DAG/DAG, and DAG/DEG cores in the membrane-spanning PGs across growth phases.
It appears that a considerable part of membrane lipids of T. maritima contains ether bonds in their structure. The strong similarity in the distribution of ether-and ester-derived lipids (e.g., Fig. 3B) suggests that a biochemical mechanism exists that can convert ester bonds into ether bonds. Genes encoding enzymes involved in the biosynthesis of FA-derived ether lipids in bacteria have been described in Myxobacteria (29) and potential homologs identified in members of SD4 Acidobacteria (11). This set of enzymes is indeed capable of converting an sn-1-bound alkyl ester into an ether-bound alkyl moiety. However, homology protein BLAST searches did not result in the detection of close homologs in the genomes of T. maritima and other Thermotogales species capable of ether lipid production (7).
Recently, a biosynthetic pathway for the biosynthesis of plasmalogens (i.e., sn-1 vinyl ethers) has been identified in the strict anaerobe Clostridium perfringens, consisting of a two-gene operon, plsA and plsR (30). The proposed mechanism converting the diacyl precursor to a plasmalogen involves two sequential electron reductions, followed by protonation and dehydration (see Fig. S3 in the supplemental material). We searched in T. maritima for potential homologs of plsA-plsR from C. perfringens and identified the Tmari 0479 gene as a close homolog but with a different protein domain architecture (Fig. S3). We also identified homologs of a modified plsA gene in the genomes of Desulfatibacillum alkenivorans, Thermodesulfobacterium geofontis, and "Candidatus Kuenenia stuttgartensis" (Fig. S3). These four bacterial species have in common with T. maritima that they produce ether lipids but not plasmalogens, suggesting that the modified PlsA encoded by the Tmari 0479 gene is capable of converting an sn-1 ester bond to an sn-1 ether bond, potentially with a plasmalogen as an intermediate. In support of this hypothesis, no homologs to this gene could be detected in members of the Thermotogales (e.g., Thermosipho spp.) that are able to produce tetraester compounds but not mixed ethers/esters lipids (7). Based on this, we hypothesize that the membrane-spanning and FA-derived glycerol ether lipids in Thermotoga are synthesized by enzymatic activities that are not related to those of Myxobacteria and Acidobacteria but rather by a novel mechanism involving a modified plsA homolog (Fig. 6). Future work is required to confirm if the enzymatic activity of the modified plsA gene product leads to conversion of the ester moiety into a saturated ether in the membrane-spanning lipids of Thermotoga. Furthermore, our analysis of the IPLs of T. maritima ( Table 2) also reveals that ether bonds do occur at the sn-2 position. It remains enigmatic how these are formed.
Cell profile modification across growth phases and changes at the lipid level. T. maritima is characterized by an unusual cell envelope, the toga, distending from the poles of the cells by the continued growth of the outer envelope as cells enter the stationary phase under physiological conditions (31). The toga was not visible in the cells grown at optimal temperature (80°C) (Fig. 5). In contrast, cells incubated at 55°C showed a fluorescent membrane signal attributed to the inner and the outer membrane (Fig. 5). Under both examined conditions (optimal and suboptimal temperatures of growth), the cell area and length were significantly reduced by 0.4 and 0.8 mm 2 and 0.5 and 0.6 mm, respectively, in stationary phase, as previously observed (18). At optimal temperature of growth, the cells become coccoid in shape in stationary phase and increased their circularity values, while at suboptimal temperature the rod-shaped morphology was dominant. Lastly, the membrane permeability was evaluated by assessing the fluorescence measurements of permeable (DAPI) and nonpermeable (SYTOX green) DNA dyes. This analysis revealed that most of the exponential-and stationary-phase cells growing either at optimal or suboptimal temperatures showed fluorescent signals for both DNA dyes. At suboptimal temperature, the fluorescence intensity for both dyes increased significantly in the stationary phase (Fig. 5, Table 3, Fig. S2). This increase at suboptimal conditions suggests these cells have increased membrane permeability relative to those grown under optimal growth conditions. Considering that at both temperatures and stationary phase of growth higher content of membranespanning lipids was detected while differences in morphology and membrane permeability were observed, it is possible that other membrane components (e.g., proteins) are involved in the adaptation to different growth temperatures.
Conclusions. Cultivation experiments combined with membrane lipid and microscopy analysis identified several environmental and physiological factors affecting the biosynthesis of membrane-spanning lipids in Thermotoga maritima. Our core and intact polar lipid analysis have demonstrated that the synthesis of membrane-spanning lipids across growth phases is dependent on the available fatty acid substrate, and that membrane-spanning lipids increased in stationary phase of growth with independence of temperature. Only intact polar lipids of the membrane-spanning lipid diabolic acid with the polar head group phosphoglycerol were detected. Overall, we have identified the physiological conditions leading to an increase of membrane-spanning lipids in T. maritima, information that is key for further investigations of the nature of their biosynthesis. In addition, we detected a putative coding gene involved in the formation of ether bonds in the membrane-spanning lipids of T. maritima. Taking all the evidence together, we propose a hypothetical biosynthetic pathway for the formation of membrane-spanning lipids of T. maritima that needs to be confirmed by future studies.

MATERIALS AND METHODS
Strains, media, and growth conditions. Thermotoga maritima (MSB8; DSM 3109) (18) was obtained from the DSMZ culture collection and was cultivated under anaerobic conditions in 120-ml batch cultures in 250-ml serum bottles. The basal medium (BM) (32) used was composed of (per liter) 1 g NH 4 Cl, 0.3 g KH 2 PO 4 , 0.3 g K 2 HPO 4 , 0.1 g KCl, 25 g NaCl, 6.5 g MgCl 2 Á6H 2 O, 0.13 g CaCl 2 Á2H 2 O, 1 g yeast extract, 10 ml Balch's trace mineral solution (33), 0.5 mg resazurin, and pH adjusted to 7 with KOH. The medium was anaerobically dispensed in 250-ml serum bottles, and a gas phase of N 2 -CO 2 (80:20, vol/vol) was applied. After sterilization, individual bottles were supplemented with 21 mM NaHCO 3 , 2.1 mM Na 2 SÁ9H 2 O, 18.7 mM glucose, 2.5 mM L-cysteine hydrochloride monohydrate. Cultures were first grown at 55 or 80°C for three subcultivation growth cycles to ensure acclimatization. The growth phase experiment was inoculated during the fourth subcultivation at the mentioned temperature. Growth was monitored by measuring the optical density