A phosphoethanolamine-modified glycosyl diradylglycerol in the polar lipids of Clostridium tetani.

The polar lipids of the anaerobic bacterium Clostridium tetani, the causative agent of tetanus, have been examined by two-dimensional thin layer chromatography, ESI mass spectrometry, and NMR spectroscopy. Plasmalogen and di- and tetra-acylated species of phosphatidylethanolamine, phosphatidylglycerol, cardiolipin, and N-acetylglucosaminyl diradylglycerol were the major lipids present in most strains examined except for strain ATCC 10779, the parent of strain E88, the first C. tetani strain to have its genome sequenced. This strain contained the same di- and tetra-acylated species but did not contain plasmalogens. All strains contained a novel derivative of N-acetylglucosaminyl diradylglycerol in which a phosphoethanolamine unit is attached to the 6’-position of the sugar, as judged by selective 31P-decoupled, 1H-detected NMR difference spectroscopy. The N-acetylglucosamine (GlcNAc) residue is presumably linked to the 3-positon of the diradylglycerol moiety, and it has the β-anomeric configuration. Very little plasmalogen component was detected by mass spectrometry in the precursors phosphatidic acid and phosphatidylserine, consistent with the idea that plasmalogens are formed from diacylated phospholipids at a late stage of phospholipid assembly in anaerobic clostridia.

ninhydrin in ethanol, respectively, followed by heating at 120°C for 10 min. Glycolipids were detected with ␣ -naphthol.

Lipid analysis
For quantifi cation of the polar lipids of C. tetani strains ATCC 454 and ATCC 10779, the cells were labeled during growth in 10 ml reinforced clostridial medium with 10 Ci [1- 14 C] acetate (60 mCi/mmol Ϫ 1 , Perkin Elmer Life Sciences, Waltham, MA) at 37°C for 24 h. Cells were harvested by centrifugation and washed, and the lipids extracted as described above. As needed, cellular lipids were also extracted according to Benning and Somerville ( 6 ) to be added as carrier for TLC. 2D-TLC was performed using 14 Clabeled material (7,500 cpm) with added carrier, as described above. Radiolabeled lipids were visualized and quantifi ed with a PhosphoImager (Typhoon 9410, Amersham Biosciences, Arlington Heights, IL), equipped with ImageQuant software. Percent diacyl PtdEtn was calculated using the formula: % diacylPtdEtn = % (of total cpm) in diacylPtdEtn after acid hydrolysis / ⌺ % diacylPtdEtn + %LysoPtdEtn + % aldehydes.
A similar formula was used for the other lipids. The % plasmalogen = 100 Ϫ % diacyl lipid.

High-resolution ESI/MS
High-resolution ESI mass spectra were acquired on a QSTAR XL quadrupole time-of-fl ight tandem mass spectrometer (Applied Biosystems, Foster City, CA) equipped with an electrospray source. For ESI/MS analysis, the lipid extracts from 500-ml cultures were each redissolved in 200 l of chloroform/methanol (2:1, v/v). Typically, 10 l of this solution was diluted into 200 l of chloroform/methanol (1:1, v/v) containing piperidine (1%, v/v), and then immediately infused into the ESI source at 5-10 l/min. The negative and positive electrospray voltages were set at Ϫ 4200 V and +5500 V, respectively. Other MS settings were as follows: CUR = 20 psi (pressure), GS1 = 20 psi, DP = Ϫ 55 V, and FP = Ϫ 265 V. For MS/MS, collision-induced dissociation was performed with collision energy ranging from 40 V to 70 V (laboratory frame of energy) with nitrogen as the collision gas. Data acquisition and analysis were performed using the Analyst QS software.

LC/MS
LC/MS of lipids was performed using a Shimadzu LC system, which comprised a solvent degasser, two LC-10A pumps, and a SCL-10A system controller, coupled to a QSTAR XL quadrupole time-of-fl ight tandem mass spectrometer (as above). LC was operated at a fl ow rate of 200 l/min with a linear gradient as follows: 100% of mobile phase A was held isocratically for 2 min, and then linearly increased to 100% mobile phase B over 14 min, and held at 100% B for 4 min. Mobile phase A consisted of methanol/acetonitrile/aqueous 1 mM ammonium acetate (60/20/20, v/v/v). Mobile phase B consisted of 100% ethanol containing 1 mM ammonium acetate. A Zorbax SB-C8 reversed-phase column (5 m, 2.1 × 50 mm) was obtained from Agilent (Palo Alto, CA). The postcolumn splitter diverted ‫ف‬ 10% of the LC fl ow to the ESI source of the mass spectrometer.

Purifi cation of the unknown lipid
C. tetani ATCC 10779 was grown in anaerobic jars in twelve 500-ml cultures and harvested as described above. The total yield of cells was 12.6 g wet weight. Total lipids were extracted with chloroform/methanol/water as described above, yielding 61.0 mg of dry lipid. The lipids, dissolved in chloroform/methanol (7:3, v/v), were fractionated on DEAE-cellulose. The DEAE-cellulose column was packed in chloroform/methanol/2.4 M aqueous ammonium acetate (2:3:1, v/v/vol) to give a bed volume of We initially examined the polar lipids of C. tetani ATCC 10779, the parent strain of E88 (G. Gottschalk and H. Brüggemann, unpublished observations), and another strain of C. tetani from the collection of the Department of Microbiology, University of Pennsylvania School of Medicine. We found by two-dimensional thin layer chromatography that they both contained PtdEtn, PtdGro, cardiolipin, and several other polar lipids. However, the two strains differed in that the University of Pennsylvania strain contained plasmalogens, whereas the ATCC 10779 strain did not. We therefore examined four additional strains from the American Type Culture Collection and found that all contained plasmalogens. The lipids were further characterized by ESI mass spectrometry and NMR spectroscopy. All strains of C. tetani contained a novel phosphoethanolaminederivatized glycosyldiacylglycerol, which we identifi ed as phosphoethanolamine-6 ′ -D-GlcNAc-( ␤ ,1 ′ -3)-diradylglycerol. Although many glycosyl diradylglycerols have been identifi ed previously in bacterial and plant systems, we are not aware of any GlcNAc-diradylglycerols modifi ed with phosphoethanolamine. Although each of the major polar lipids contained a large proportion of plasmalogen, very little plasmalogen was seen in the precursors phosphatidic acid and phosphatidylserine, consistent with the hypothesis that plasmalogens are derived from diacylated glycerophospholipids in anaerobic bacteria.

Bacterial cultures
C. tetani strains ATCC 454, 9441, 10709, 10779, and 19406 were obtained from the American Type Culture Collection (Manassas, VA). For lipid isolation all strains of C. tetani were grown on reinforced clostridial medium without agar, which contained the following per liter: yeast extract, 10.5 g; peptone, 12.5 g; glucose, 5 g; soluble starch, 1 g; NaCl, 5 g; sodium acetate, 3 g; and cysteine HCl·H 2 O, 0.6 g. The pH was adjusted to ‫ف‬ 7.1 with NaOH. The glucose solution was autoclaved separately.

Lipid extraction
Cells were harvested by centrifugation at 2900 g for 10 min and washed twice in 20 mM MOPS buffer, pH 7.2. The wet cell pellets were extracted with chloroform/methanol/water by the method of Bligh and Dyer ( 4 ), as modifi ed ( 5 ). The lipid extracts were dried under a stream of nitrogen while being warmed in a heating block. They were weighed, dissolved in chloroform, and stored at Ϫ 20°C.

TLC
Two-dimensional thin-layer chromatography (2D-TLC) was performed on silica gel 60 and 10 × 10 cm thin-layer plates. The solvents were chloroform/methanol/concentrated ammonia/ water, 65:30:2.5:2.5 (v/v/v/v) in the fi rst dimension and chloroform/methanol/acetic acid/water, 80:18:12:5 (v/v/v/v) in the second dimension. For acid hydrolysis of lipids on TLC plates after separation in the fi rst dimension, the area of the plates containing the lipids was suspended over boiling HCl for 40 s, after which the plates were dried under a stream of nitrogen and reactivated under vacuum. The plates were then chromatographed in the second dimension. Phospholipids or amine-containing lipids were detected using 0.3% (w/v) molybdenum blue or 0.3% stained with both phosphomolybdate and ␣ -naphthol and was, therefore, identifi ed as a phosphorus-containing glycolipid of unknown structure ( Fig. 1A ).
Because plasmalogens are commonly found in anaerobic bacteria, including members of the genus Clostridium, we carried out exposure to HCl fumes between the fi rst and second dimensions of the 2D-TLC ( 9 ). Plasmalogens are acid-labile, and treatment with HCl fumes hydrolyzes the alk-1-enyl bond, releasing a long-chain aldehyde and a lysolipid. Treatment of the lipids from C. tetani ATCC 10779 with acid fumes resulted in no change in the mobility of the polar lipids (data not shown); however, treatment of the similar polar lipids from a strain of C. tetani from the collection of the Department of Microbiology, University of Pennsylvania School of Medicine revealed extensive hydrolysis of all the major lipids, indicating that all contained both all acyl and alk-1-enyl acyl species.
We obtained four additional ATCC strains of C. tetani : ATCC 19406, ATCC 10709, ATCC 454, and ATCC 9441. The 2D-TLC analysis of the lipids extracted from these cells with and without acid hydrolysis indicated that all of these strains contained plasmalogens. Thus, strain ATCC 10779 appeared to be an exception. A representative 2D-TLC with acid hydrolysis of the lipids between dimensions from strain ATCC 454 is shown in Fig. 1B , demonstrating the extensive formation of lyso-phospholipids compared with Fig. 1A .

Quantifi cation of C. tetani polar lipids
The polar lipid composition of C. tetani strains ATCC 10779 ( Fig. 1A ) and ATCC 454 ( Fig. 1B ) could be quantifi ed isotopically, given that the cells had been grown for 24 h with [1-14 C] acetate ( Table 1 ). Aside from the presence of plasmalogens in strain ATCC 454 and their absence in strain ATCC 10779, there are other differences in the lipid class compositions. Strain ATCC 454 contains almost twice as much PtdEtn as ATCC 10779, but it has very little GlcNAcdiacylglycerol, compared with 21.5% of total polar lipid in the latter strain. However, the other plasmalogen-rich strains ATCC 9441, ATCC 10709, and ATCC 19406 contain larger amounts of GlcNAc-diacylglycerol than ATCC 454, as visualized after 2D-TLC (data not shown). Strains ATCC 10779 and ATCC 454 have comparable amounts of the unknown phosphorus-containing glycolipid.

ESI/MS of total C. tetani polar lipids
To confi rm the identities of the major lipids and obtain additional information about the structure of the unknown compound, we subjected the total polar lipids of all C. tetani strains to ESI/MS. The identities of all major lipids of strains ATCC 10779 and ATCC 454 assigned by TLC were confi rmed by mass spectroscopy ( Table 2 and supplementary Figs. I-III ). In the other strains (data not shown), similar results were obtained for the all-acyl lipids and the corresponding plasmalogens, which were observed at 16 amu lower than those for the all-acyl lipids ( Table 1 ). For cardiolipin from ATCC 454, in addition to the species that contain a single alkenyl chain, there is evidence for cardiolipin containing two alkenyl chains. TLC after HCl treatment shows a lipid 13 ml. The column was then washed with chloroform/methanol, 7:3 (by volume). After application of the lipids, the column was washed with the same solvent mixture: 50 ml (fraction 1). The column was then eluted with chloroform/methanol 1:1, 50 ml, and the same solvent, 60 ml (designated fractions 2 and 3). The column was then eluted with chloroform/methanol 2:1, containing 0.25% ammonium acetate, 100 ml (fraction 4). All fractions were evaporated in a rotary evaporator and transferred to weighed tubes. A portion of each fraction (50 g) was applied to a TLC plate and chromatographed in chloroform/methanol/ acetic acid/water, 200:40:30:12.5 (v/v/v/v) (solvent I), sprayed with 10% sulfuric acid in ethanol, and charred on a hot plate. Almost all of the unknown lipid was found in fraction 2.
A silicic acid column, 1.0 ×18 cm, was packed in Solvent I, and the lipid from fraction 2 was applied in the same solvent. The column was eluted with 60 ml of the same solvent, and 1-ml fractions were collected. Lipids from every other fraction were chromatographed on a TLC plate, as described for the DEAE-cellulose fractions, and the unknown lipid was eluted in fractions 15-25, as judged by charring. Fractions 15-25 were pooled, dried, and taken up in CDCl 3 :CD 3 OD:D 2 O for NMR spectroscopy.

NMR spectroscopy
Approximately 5 mg of the purifi ed unknown lipid was dissolved in 0.6 ml of CDCl 3 /CD 3 OD/D 2 O (2:3:1, v/v/v) in a 5 mm NMR tube. Spectra were obtained at 25°C using a Varian Inova 500 spectrometer equipped with a Dell 390 computer and a 5-mm Varian probe. Proton and carbon chemical shifts are reported relative to internal tetramethylsilane (TMS) at 0.00 ppm. The 2 H signal of CD 3 OD was used as a fi eld frequency lock with the residual signal of CD 3 OD serving as the secondary reference at 49.5 ppm for carbon spectra. 31 P NMR chemical shifts were referenced to 85% H 3 PO 4 at 0.00 ppm. Use of the CDCl 3 /CD 3 OD/D 2 O solvent system introduced four solvent 1 H resonances. The signals from CH 3 OD (3.3 ppm) and CHCl 3 (7.6 ppm) did not overlap with the lipid resonances. The HOD (4.5 ppm) and CD 3 OH (4.8 ppm) signals were removed with a presaturation sequence. 1 H NMR spectra at 500 MHz were obtained with spectral width (SW) of 4.5 kHz, a 67° pulse fl ip angle (6 msec), a 5.0 s acquisition time (AT), and a 1.2 s relaxation delay (RD); they were digitized into 45k points yielding a digital resolution of 0.2 Hz/pt. 1 H-decoupled 13 C NMR spectra were recorded at 125.7 MHz with a spectral window of 31421.8 Hz digitized into 80,000 data points (digital resolution: 0.78 Hz/point or ‫ف‬ 0.006 ppm/point), a 60° pulse fl ip angle (8 µs), and a 2.0 s repeat time. 1 H-decoupled 31 P NMR spectra were recorded at 202.3 MHz with a spectral window of 12143.3 Hz digitized into 25,280 data points (digital resolution: 1 Hz/point or ‫ف‬ 0.005 ppm/point), a 60° pulse fl ip angle (8 µs), and a 1.6 s repeat time. Two-dimensional NMR experiments [correlation spectroscopy (COSY), heteronuclear multiplequantum correlation (HMQC), heteronuclear multiple-bond correlation (HMBC )] and 1 H-detected difference spectra derived from selective 31 P-decoupled NMR experiments were performed as previously described ( 7,8 ).

TLC and acid sensitivity of C. tetani polar lipids
Two-dimensional TLC of the total lipids of C. tetani ATCC 10779 revealed fi ve major polar lipids, four of which were tentatively identifi ed as PtdEtn, PtdGro, cardiolipin, and a GlcNAc-diacylglycerol (see below), based on their relative mobility and staining with phosphomolybdate, ninhydrin, or ␣ -naphthol. The fi fth compound, labeled 6, 891.58, and 893.59 were observed in the total lipids of strain ATCC 10779 ( Table 2 and  Evaluation of the covalent structure of the unknown lipid by 31 P and 1 H NMR spectroscopy 31 P NMR spectroscopy of the 5-mg sample of purifi ed unknown lipid from C. tetani ATCC 10779, which contained no plasmalogen ( Table 2 ), revealed large (76%) and small (24%) phosphorus resonances near 1.511 and more polar than lyso-cardiolipin, which may represent dilysocardiolipin. Mass spectrometry shows prominent peaks at 1207.8, 1233.9, 1235.8, 1261.9, 1287.9, and 1289.9, which correspond to the dialkenyl species of cardiolipin, 56:0, 58:1, 58:0, 60:1, 62.2, and 62.1, respectively ( Table 2 ).
Phosphatidic acid and phosphatidylserine (PtdSer) were detected in strain ATCC 454 (and other strains), but in contrast to the major lipids, no plasmalogen species were detected (supplementary Figs. V and VI), consistent with previous studies indicating that plasmalogens are formed from diacylated phospholipids at a late stage of phospholipid assembly in clostridia (10)(11)(12)(13).

An ethanolamine-P-containing glycosyl diradylglycerol in C. tetani
A series of singly charged peaks seen during ESI/MS in the negative ion mode at m/z 837.526, 863.54, 865.55, Abbreviations: GlcNAc, N-acetylglucosamine; PtdEtn, phosphatidylethanolamine; PtdGro, phosphatidylglycerol. a Mean ± SD, n = 4. b Mean ± SD, n = 3. c Difference in PtdEtn content of the two strains is statistically signifi cant, P < 0.05. d Numbers in parentheses are the percentage plasmalogen form. This could not be calculated for PtdGro and cardiolipin separately, but together they were 52 ± 6% plasmalogen. e Spot 6 in Fig. 1A . f Spot 2 in Fig. 1A and spot 6 in Fig. 1B . As this spot is greatly decreased in intensity after treatment with acid vapor with no corresponding hydrolysis products, it is unlikely to be a lipid. It is more likely a compound that becomes volatile upon acidifi cation. 1.242 ppm, respectively ( Fig. 3A ). These 31 P NMR shifts are consistent with the presence of two distinct phosphodiester linkages. The larger signal arises from the major diacyl component, while the smaller signal arises from contaminating lysophosphatidylethanolamine (lysoPE), which elutes together with the unknown compound. Reverse phase chromatography allows separation of the unknown lipid from lysoPE (not shown).
The 500-MHz 1 H NMR spectrum of the unknown lipid, dissolved in CDCl 3 /CD 3 OD/D 2 O (2:3:1, v/v) ( Fig. 3C, Fig.  4 , and Table 3 ), shows relatively well-resolved resonances  Table 2 were identifi ed by tandem mass spectrometry in conjunction with exact mass measurement. For example, the collision-induced dissociation mass spectra of PtdEtn in the sugar (3.2-5.8 ppm) and acyl chain regions (0.8-2.7 ppm). An expansion of the sugar region ( Fig. 3C ) reveals a prominent envelope signal integrating to fi ve protons near 4.1 ppm, three overlapped multiplets each integrating to two protons near 3.6, 3.5, and 3.2 ppm, respectively, and fi ve resolved single proton resonances.
The positions of the individual protons of the glycerol moiety were derived from 2D-COSY NMR ( Fig. 4 and Table 3 ). The multiplet at 5.18 ppm arises from the glycerol H-2 proton (see Fig. 2 inset for the numbering scheme). A pair of COSY cross peaks from H-2 ( Fig. 4 ) locate the H-1a signal at 4.33 ppm (dd, J 12 =2.5, J ab =12.0 Hz) and the H-1b multiplet as one of fi ve overlapped signals near 4.1 ppm. A second pair of COSY cross peaks from H-2 connect to H-3a at 3.93 ppm (dd, J 23 =5.1, J ab =10.7 Hz) and H-3b as one of two protons signals near 3.63 ppm.
The 2D-COSY analysis also revealed a strong cross peak between the resolved ␤ -CH 2 (3.19 ppm) of the phosphoethanolamine moiety ( 8 ) and the ␣ -CH 2 for the same residue, which contributes to two of the fi ve overlapped proton signals near 4.1 ppm ( Fig. 3C, Fig. 4 , and Table 3 ).
The positions of the individual protons of the putative N -acetylglucosamine residue were also derived from 2D-COSY ( Fig. 4 and Table 3 ). The doublet at 4.41 ppm ( J 12 =8.4) arises from the glucosamine anomeric H-1 ′ proton ( Fig. 2 , inset). The large J 12 coupling (8.4 Hz) indicates that H-1 of the GlcNAc unit is in the ␤ confi guration and the axial orientation, so that the glycerol group is linked equatorially ( Fig. 2 , inset). The COSY cross peak from H-1 ′ ( Fig.  4 ) locates the H-2 ′ signal at 3.62 ppm (which overlaps with the glycerol H3b). A second COSY cross peak from H-2 ′ connects to overlapped H-3 ′ (3.49 ppm) and H-4 ′ (3.50 ppm) multiplets. H-4 ′ , in turn, is coupled to H-5 ′ at 3.40 ppm (a broad multiplet). Further tracing of the COSY connectivities locate H-6a ′ and H-6b ′ as the remaining two overlapped proton signals near 4.1 ppm.
The 2D-COSY ( Fig. 4 ) also revealed strong cross peaks between the resolved ␣ -CH 2 ( ‫ف‬ 2.32 ppm) and the ␤ -CH 2 ( ‫ف‬ 1.59 ppm) resonances, as well as the terminal -CH 3 (0.86 ppm) and the bulk CH 2 units (1.25 ppm) of the acyl chains (see Fig. 2 inset for labeling). The CH multiplet at 5.32 ppm connects to a CH 2 resonance near 2.0 ppm. The 2.0 ppm CH 2 signal shows a further COSY connectivity to the bulk CH 2 signals near 1.25 ppm. These resonances arise from allylic CH 2 groups in the acyl chains and refl ect the presence of a double bond in these chains ( Table 3 ). The 2.0 ppm CH 2 multiplet also overlaps with an N -acetyl CH 3 singlet resonance at 1.99 ppm.
The location of the linkage between the phosphoethanolamine moiety and the proposed GlcNAc unit of the unknown lipid was evaluated by subtraction of two 1 H NMR spectra, obtained with on-and off-resonance selective decoupling of the 1.511 ppm phosphate signal. This strategy revealed simultaneous changes at the H-6 CH 2 signal of the putative GlcNAc residue and the ␣ -CH 2 signal of the phosphoethanolamine moiety ( Fig. 3B ), establishing the presence of a single phosphate group linking the glucosamine C-6 ′ carbon and the ␣ -CH 2 carbon of the phosphoethanolamine unit, as shown in the inset of Fig. 2 . multiplets correlate to a carbon signal at 68.31 ppm (C-3), while the H-1a and H-1b multiplets connect to a carbon resonance at 63.67 ppm (C-1). The ␣ -CH 2 and the ␤ -CH 2 1 H signals of the phosphoethanolamine moiety correlate to CH 2 carbon resonances at 62.57 and 41.17 ppm, respectively. The upfi eld shift of the ␤ -CH 2 carbon refl ects the nitrogen substitution of the methylene carbon by the NH 2 group.
The anomeric H-1 ′ proton signal of the presumed GlcNAc residue correlates to the anomeric carbon resonance at 102.6 ppm (C-1 ′ ). This carbon position further verifi es Evaluation of the carbon structure of unknown lipid by 13

C and HMQC spectroscopy
To confi rm the assignments derived from the 1 H NMR analysis, 13 C data for the unknown lipid (5 mg) were obtained both directly through 1D 13 C NMR spectroscopy, and indirectly through 1 H-detected 2D-HMQC and 2D-HMBC NMR experiments ( Table 3 ). The partial 2D-HMQC 1 H-13 C correlation map ( Fig. 5A ) reveals the fi ve direct 1 H-13 C single-bond correlations in the glycerol unit. The glycerol H-2 proton signal correlates to the glycerol CH carbon resonance at 70.87 ppm (C-2). The H-3a and H-3b proton P decoupling of the 1.511 ppm phosphorus signal showed that the 6-CH 2 group of the putative N-acetylglucosamine is linked via a phosphodiester to the ␣ -methylene unit of the phosphoethanolamine group. C: The partial one-dimensional 500 MHz 1H NMR spectrum of the donor lipid shows six GlcNAc and various glycerol and phosphoethanolamine proton signals (see Table 3). The resonances at 4.6 ppm and 3.3 ppm are due to residual water and methanol solvent signals. The small resolved multiplets at 3.99 and 3.86 ppm arise from impurities, presumably from co-eluting lysophosphatidylethanolamine, the presence of which is well detected in the 31 P nmr spectrum of Panel A. Note that an impurity peak also overlaps H-5 near 3.66 ppm. The proposed structure of the novel lipid is shown in the inset to Fig. 2 along with the numbering scheme.  ( Fig. 1A, spot 6 ) is shown in the inset, along with the numbering scheme used for the NMR analysis.
substitution. The results of the HMQC correlations, along with data for the acyl chains (not discussed), are summarized in Table 3 .

Evaluation of the carbon structure of unknown lipid by HMBC spectroscopy
HMBC analysis ( Fig. 5B ) reveals distinct multibond correlations from the glycerol H-3a and H-3b at 3.93 and 3.63 ppm to the glucosamine C-1 ′ (102.6 ppm) and the reciprocal multibond correlation from the glucosamine H-1 ′ (4.21 ppm) to the glycerol C-3 (68.31 ppm), thus establishing the linkage of the glycerol to C-1 of the sugar.
The remaining HMBC multibond correlations ( Fig. 5B ) in general verify the glycerol, phosphoethanolamine, and glucosamine assignments derived from the COSY and HMQC. For example, the glycerol H-3a and H-3b at 3.93 and 3.63 ppm show distinct multibond correlations to the glycerol C-2 (70.87 ppm) and C-1 (63.68 ppm). Similarly, the resolved ␤ -CH 2 proton signal of the phosphoethanolamine moiety shows a strong multibond correlation to ␣ -CH 2 carbon resonance at 62.57 ppm ( Table 3 ).

DISCUSSION
The lipids of C. tetani are characteristic of the genus Clostridium in having PtdEtn, PtdGro, and cardiolipin in both the acyl forms and plasmalogens as major components. Absent are the glycerol acetals of plasmalogens which are characteristic of the solvent-producing species, such as C. acetobutylicum , C. beijerinckii , and the closely related C. butyricum ( 1 ). Plasmalogens are present in all strains of C. tetani examined with the exception of ATCC 10779. This strain is sometimes referred to as strain Massachusetts, which has been used in vaccine production, and is the parent strain of E88, a nonsporulating variant that has had its genome sequenced ( 3 ).
Figs. I-III ). The latter determination assumes that the diacyl and plasmalogen species are ionized with equal efficiency. Based on acid hydrolysis of the 14 C-labeled lipids, PtdGro and cardiolipin combined were found to consist of 52% plasmalogen. The novel ethanolamine phosphate-GlcNAc diradylglycerol is 48% plasmalogen based on 14 C-labeling. The determination of the plasmalogen content based on 14 C-acetate labeling assumes that the acyl and alk-1-enyl chains will be equally labeled, which is likely to be the case after growth into stationary phase. Minor lipids in C. tetani ATCC 454, such as phosphatidic acid and phosphatidylserine, which serve as precursors of the major diacylglycerolipids, were not enriched in the corresponding plasmalogen species (supplementary Figs. V and VI). This fi nding is consistent with previous studies based on isotopic labeling, which indicated that the glycerol-phosphate backbones of the diacyl lipids may serve as precursors for the plasmalogens ( 10-13 ). All of the C. tetani strains we examined contain plasmalogens, with the exception of ATCC 10779, which was used for toxoid production and was presumably passaged many times in the past. It will be interesting to study the ability of these strains to cause tetanus. Eventually, when the genes for plasmalogen biosynthesis have been discovered, a study of the pathogenicity of knockout strains will be instructive.
The biosynthesis and function of our novel phosphoethanolamine-GlcNAc diradylglycerol will require further investigation. In principle, the phosphoethanolamine moiety might be derived from CDP-ethanolamine, or alternatively, by transfer of the phosphoethanolamine unit from phosphatidylethanolamine, as occurs during lipid A modifi cation in Gram-negative bacteria ( 22 ). Identifi cation of the relevant enzyme and the corresponding structural gene, in conjunction with studies of the phenotypes of specifi c deletion mutants, should facilitate the elucidation of the function of this novel lipid.
The authors thank C. Michael Reynolds for help in the isolation of the novel phosphoethanolamine-containing glycosyl diradylglycerol.
acyl, but not the alk-1-enyl, chains in strains E4222 and 34946A. Analysis of the molecular species of the polar lipids by MS ( Table 2 and supplementary Figs. I-III ) is consistent with these earlier analyses, which were done by gas chromatography.
Evidence for the ethanolamine phosphate-GlcNAc diradylglycerol came initially from negative ion ESI/MS anal ysis of the total lipids of C. tetani ATCC 10779, which revealed prominent peaks at m/z 837.52, 863.54, and 891.57 ( Table 2 and supplementary Fig. I ). The lipid giving rise to m/z 837.52 produced fragments near m/z 627, 325, and 140, the last being a characteristic fragment of phospholipids with an ethanolamine-phosphate head group ( Fig. 2 ). The structure was confi rmed by a combination of 1 H, 13 C and 31 P NMR spectroscopy ( Figs. [3][4][5] . All the major lipids in C. tetani ATCC 454 have significant proportions of plasmalogens as determined by 2D-TLC with acid hydrolysis between the fi rst and second chromatographies and by MS ( Table 2 and supplementary  Figs. I-III ). PtdEtn had 32 ± 5% plasmalogen as determined by acid hydrolysis of 14 C-acetate-labeled lipids and an average of 43% plasmalogen as determined by the intensities of the respective peaks upon MS (supplementary H- 13 C correlations of the glycerol, phosphoethanolamine, and GlcNAc moiety are shown. The single carbon-decoupled cross peak at 3.30/49.9 ppm is due to residual methanol solvent. B: This expansion shows multibond correlations from the glycerol H3 protons to the GlcNAc C-1 carbon, and from the GlcNAc H-1 to the glycerol C-3, thus confi rming the proposed linkage of the glycerol to the C-1 carbon of the sugar (Fig. 2, inset). The two cross peaks at 49.9 ppm (from carbon coupling) are due to residual methanol solvent.