Identification of phosphatidylserylglutamate: a novel minor lipid in Escherichia coli.

Advances in mass spectrometry have facilitated the identification of novel lipid structures. In this work, we fractionated the lipids of Escherichia coli B and analyzed the fractions using negative-ion electrospray ionization mass spectrometry to reveal unknown lipid structures. Analysis of a fraction eluting with high salt from DEAE cellulose revealed a series of ions not corresponding to any of the known lipids of E. coli. The ions, with m/z 861.5, 875.5, 887.5, 889.5, and 915.5, were analyzed using collision-induced dissociation mass spectrometry (MS/MS) and yielded related fragmentation patterns consistent with a novel diacylated glycerophospholipid. Product ions arising by neutral loss of 216 mass units were observed with all of the unknowns. A corresponding negative product ion was also observed at m/z 215.0. Additional ions at m/z 197.0, 171.0, 146.0, and 128.0 were used to propose the novel structure phosphatidylserylglutamate (PSE). The hypothesized structure was confirmed by comparison with the MS/MS spectrum of a synthetic standard. Normal phase liquid chromatography-mass spectrometry analysis further showed that the endogenous PSE and synthetic PSE eluted with the same retention times. PSE was also observed in the equivalent anion exchange fractions of total lipids extracted from the wild-type E. coli K-12 strain MG1655.

dylglycerol phosphate ( 3 ). For each of these lipids, there can exist several molecular species arising from the different lengths, unsaturation, and/or cyclopropane analogs of the acyl chains ( 4 ). The complexity of the E. coli lipidome, however, is even greater than can be explained by acyl chain heterogeneity. Numerous additional minor lipids are present in wild-type cells, as judged by isotopic labeling experiments and two-dimensional TLC ( 3 ). Many of these species cannot be identifi ed by their migration with standards of known lipids or biosynthetic precursors.
Electrospray ionization-mass spectrometry (ESI-MS) is well suited to the analysis of intact lipids ( 5 ). Fragmentation during ionization is minimized and the sensitivity is high ( 6,7 ). In addition, collision-induced dissociation mass spectrometry (MS/MS) allows for the structural analysis of the lipid ion of interest and, when combined with the high mass accuracy of time-of-fl ight mass spectrometers, can be used to propose a molecular formula for a particular ion.
Current applications of mass spectrometry to lipid analysis have focused mainly on the quantifi cation of known lipid species, often coupling liquid chromatography directly to the mass spectrometer (8)(9)(10). These analyses compare levels of known lipid species present under various growth conditions or disease states. However, important changes in levels of unknown or minor lipids are diffi cult to analyze without knowledge of their structures and the availability of appropriate standards. Furthermore, some important, minor E. coli lipids, such as CDP-diacylg- ergy of Ϫ 50.0 V (laboratory frame of reference) and N 2 as the collision gas. Data acquisition, analysis, and elemental composition calculations were performed using the Analyst QS 1.1 software. Exact masses of lipid species and product ions were obtained using CS Chem Draw Pro, version 8.0. The PSE synthetic standard was diluted to 0.1 µg/ml using CHCl 3 :CH 3 OH (2:1, v/v) and analyzed by direct infusion mass spectrometry and MS/MS, as described above.

Normal phase liquid chromatography-mass spectrometry
Normal phase liquid chromatography-mass spectrometry (LC-MS) was adapted from Becart, Chevalier, and Biesse ( 16 ). An Agilent 1200 system with quaternary pumps was used with three mobile phase solvents and coupled to the QSTAR XL quadrupole time-of-fl ight mass spectrometer. The LC was operated with an Ascentis silica column (5 µm, 25 cm × 2.1 mm) at a fl ow rate of 300 µl/min with a three solvent gradient program.
is increased from 0% to 100% over the next 15 min. Mobile phase C is held at 100% C for 2 min. The column was regenerated by decreasing mobile phase C from 100% to 0% as mobile phase B is increased from 0% to100% over 5 min and then held at 100% mobile phase B for 2 min. Finally, mobile phase B was decreased from 100% to 0% as mobile phase A was increased from 0% to 100% over 5 min and then held at 100% mobile phase A for 18 min.
In a typical run, the dried lipid fi lm from a pooled fraction (see Table 1 ) was redissolved in 2.0 ml of CHCl 3 :CH 3 OH (2:1, v/v), and 0.5 ml was transferred to a separate tube, dried under nitrogen gas, and then redissolved in 0.3 ml of CHCl 3 :CH 3 OH: H 2 O (73:23:3, v/v/v). Twelve microliters was injected onto the column at 300 µl/min with mobile phase A. The column fl ow was split 1 to 10 prior to introduction into the mass spectrometer. Mass spectra were obtained scanning from 200 to 2000 Da in negative-ion mode with the electrospray ionization source operating at the following settings: nebulizer gas, 21 kPa, curtain gas, 27 kPa, ion-spray voltage, Ϫ 4,500 V, declustering potential, Ϫ 55 V, focusing potential, Ϫ 265 V, declustering potential 2, Ϫ 15 V, ionization temperature, ambient room temperature. In some LC-MS runs, collision-induced dissociation was performed using information-dependent acquisition mode (IDA), a collision energy of Ϫ 52 V (laboratory frame of reference) and N 2 as a collision gas. During an IDA analysis, Collision-induced decomposition (MS/MS) spectra were obtained for ions between m/z 300 and 2000 with an ion intensity of at least 10 intensity units. A given ion subjected to MS/MS during IDA was excluded for 45 s of the LC run. For normal phase LC-MS analysis of the PSE synthetic standard, 12 µl of a 0.01 ng/µl solution was analyzed as described above.

Growth of E. coli
E. coli K-12 strain MG1655 was cultured at 37°C in Luria broth (LB) consisting of 10 g of NaCl, 5 g of yeast extract, and 10 g of tryptone per liter ( 17 ). The cells were grown overnight in LB medium at 37°C and then diluted into 2 liters of LB medium to an A 600 of 0.01. The culture was grown at 37°C, shaking at 225 rpm until the A 600 was about 3.0. Cells were harvested by centrifugation for 45 min at 2,600 g and washed with 1 liter of PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , and 1.8 mM KH 2 PO 4 ). lycerol or the peptidoglycan precursor lipid II, are not detectable unless the total lipids are fi rst fractionated to remove the major species ( 11,12 ).
In this work, we prefractionated 0.5 g of E. coli B lipids using anion exchange chromatography. A fraction that eluted just after cardiolipin with high salt from a DEAE cellulose column was analyzed by negative-ion ESI-MS. Here, we report the identifi cation of the novel glycerophospholipid phosphatidylserylglutamate (PSE), in which the seryl-glutamate dipeptide is linked to the phosphate moiety via the serine hydroxyl group.

Ion exchange chromatography
The total lipid of E. coli B (0.5 g) was dissolved in 100 ml of CHCl 3 /CH 3 OH/H 2 O (2:3:1, v/v/v) and loaded onto a 15 ml DEAE-cellulose column (Whatman DE52) preequilibrated with ammonium acetate as the counter-ion in the same solvent ( 13 ). The column was washed with 500 ml of CHCl 3 /CH 3 OH/H 2 O (2:3:1, v/v/v) and then with 75 ml each of CHCl 3 /CH 3 OH/ NH 4 Ac (2:3:1, v/v/v) containing sequentially 15, 30, 60, 90, 120, 240, and 480 mM NH 4 Ac in the aqueous component. Fractions (15 ml) were collected, and a portion of each was spotted onto an HPTLC plate. The plate was developed in CHCl 3 /CH 3 OH/H 2 O/ CH 3 COOH (25:15:4:2, v/v/v/v) and charred with 10% sulfuric acid in ethanol. This TLC analysis was used to direct the pooling of fractions as shown in Table 1 . The pooled fractions were converted to a two-phase Bligh-Dyer system ( 14 ) through the addition of water and chloroform (fi nal ratios, CHCl 3 :CH 3 OH:aqueous, 2:2:1.8, v/v/v). Upper and lower phases were resolved by centrifugation for 15 min at 2,600 × g in Tefl on-lined centrifuge bottles. The upper phase was discarded and the lower phase dried by rotary evaporation. For subsequent analysis, the dried, pooled fractions were redissolved CHCl 3 :CH 3 OH (2:1, v/v) as shown in Table 1 to adjust for the pooled volume. Twenty microliters were spotted on a silica-gel 60 HPTLC plate (EMD Chemicals, NJ) and developed as described above. Following TLC analysis, the pooled fractions were dried under nitrogen and stored at -20°C.

Mass spectrometry
The dried lipid fi lm remaining in the tube following solvent removal was redissolved in 2.0 ml of CHCl 3 :CH 3 OH (2:1, v/v). This solution was directly infused into the Turbo electrospray ionization source of a QSTAR XL quadrupole time-of-fl ight tandem mass spectrometer (ABI/MDS-Sciex, Toronto, Canada) at 6 µl/min ( 15 ). The instrument was calibrated using polypropylene glycol (PPG) (Applied Biosystems, Foster City, CA). The resolution of the instrument under normal operating conditions was 10,000 to 15,000. The mass accuracy of the instrument was between 5 and 20 ppm, and as such measured masses are given to three decimal places. MS/MS was performed with a collision en-NH 4 Ac (2:3:1, v/v/v) elution, contained residual traces of PE and some PG. Fractions 3-5 contained the majority of the PG. PA and some CL eluted at higher NH 4 Ac concentrations in fractions 6 and 7. Most of the CL eluted in fractions 8 and 9. Fraction 10 contained the least lipid ( Fig. 1 ). Small amounts of CL are present as well as at least two more slowly migrating, unknown species. Initial mass spectrometry analysis was conducted on this fraction because it contained the least amount of a known, major glycerophospholipid and would allow for more effi cient ionization of a novel minor lipid species.  13 C-isotope ions, did not correspond to any of the known lipids of E. coli ( Fig. 2, inset). However, this pattern of ions is consistent with a series of related E. coli glycerophospholipids; the difference of 28 mass units corresponds to two methylenes (CH 2 ); the difference of 26 mass units corresponds to two methylenes and an unsaturation; and the difference of 14 amu refl ects the presence of cyclopropane fatty acids in E. coli ( 24 ).

Elucidation of the structure of the unknown ions by ESI-MS/MS
To determine the structures of the compounds responsible for these ions, MS/MS was performed on each of the ions, revealing similar fragmentations patterns.

Extraction of E. coli MG1655 total lipids
The fi nal cell pellet was resuspended in 50 ml of PBS and transferred to a Tefl on-lined centrifuge bottle. The cellular lipids were extracted using the method of Bligh and Dyer ( 14 ). Briefl y, 62.5 ml of chloroform and 125 ml of methanol were added to the cell suspension to generate a single phase extraction mixture of CHCl 3 :CH 3 OH:PBS (1:2:0.8, v/v/v) . After incubation at room temperature for 20 min, the mixture was centrifuged at 2,600 g for 15 min. The supernatant was transferred to a clean bottle and converted to a two-phase Bligh-Dyer extraction mixture (CHCl 3 :CH 3 OH:PBS, 2:2:1.8, v/v/v) by the addition of 62.5 ml of chloroform and 62.5 ml of PBS. The extraction mixture was centrifuged as above to resolve the phases. The lower phase was washed with 237 ml of preequilibrated neutral upper phase, and the fi nal lower phase dried using rotary evaporation. The MG1655 total lipid extract was redissolved in 150 ml of CHCl 3 :CH 3 OH:H 2 O (2:3:1, v/v/v) and fractionated on a 15 ml DEAE cellulose column as described above. The fractions that eluted with CHCl 3 : CH 3 OH:240 mM NH 4 Ac (2:3:1, v/v/v) and CHCl 3 :CH 3 OH:480 mM NH 4 Ac (2:3:1, v/v/v) were pooled and converted to a two-phase Bligh-Dyer extraction mixture. The lower phase was dried down and the lipids redissolved in 100 µl of CHCl 3 :CH 3 OH:H 2 O (73:23:3, v/v/v). Seventy microliters of the sample was analyzed by normal phase LC-MS as described above.

Ion exchange chromatography of E. coli lipids
E. coli B lipids were separated on a DEAE cellulose column based on their charge. The fractions were pooled as shown in Table 1 . Care was taken to minimize the overlap among the major lipid species, PE, PG, phosphatidic acid (PA), and CL and further analyzed by TLC ( Fig. 1 ). As expected, PE did not adhere to the column and was found predominantly in the run through ( Fig. 1 , fraction 1). Fraction 2, the fi rst part of the CHCl 3 :CH 3 OH:15 mM Ϫ , had the same elemental composition as the [M-H] Ϫ ion of the amino acid glutamate. The product ion at m/z 128 is consistent with water loss from one of the carboxyl groups of the glutamate ion. From this data, we hypothesized that the novel head group consisted of a serine residue with a glutamate moiety attached via an amide linkage to the carboxyl of serine. The proposed structure of the 889.5 ion is shown in the inset of Fig. 3 , a new glycerophospholipid, which we designate as PSE. Figure 3B shows the proposed structures of the product ions derived from the head group of PSE. These product ions are all consistent with glutamate and serine linked via an amide bond between the carboxyl group of the serine and the amine group of the glutamate. The alternative linkage, between the amine of the serine and either carboxyl of the glutamate, would not yield the glutamate product ion at m/z 146.0459 ( Fig. 3B ).
The other major peaks in the product ion spectra are attributable to losses of acyl chains and/or the head group from the precursor ion.  ( Fig. 3B ). The ion at m/z 197.057 corresponds to water loss from the 215.068 ion. The ion at m/z 171.078 is interpreted as loss of CO 2 from the ion at m/z 215.068. The ion at m/z 128.036 corresponds to water loss from the ion at m/z 146.047.
We used an elemental composition calculator to obtain a list of possible molecular formulas for the product ion at m/z 146.047. With a mass error tolerance of 30 ppm, 13 possible molecular formulas containing the elements C, H, O, N, S, and P (maximum 20 of each) were predicted ( Table  2 ). Several of the formulas were not consistent with naturally occurring molecules. However, it was observed that the  ) and has a predicted [M-H] Ϫ of m/z 915.524 amu, isobaric with an endogenous PSE esterifi ed with two 18:1 fatty acids, presumed to be cis -vaccenate residues (18:1 ⌬ 11 ) based on the fatty acid composition of E. coli ( 32,35 ). The presence of additional fragment ions, such as m/z 267.215 and m/z 295.249, in the endogenous PSE spectrum strongly suggest the presence of such isobaric PSE species in fraction 10. As stated above, we cannot defi nitively assign these product ions as cyclopropane fatty acids using this mass spectrometry analysis.
To confi rm the proposed structure of the unknown lipids, normal phase LC-MS was performed to demonstrate that the endogenous PSE species have the same retention time as the synthetic PSE. Fraction 10 and the synthetic PSE were independently analyzed by normal phase LC-MS. The extracted ion current of the synthetic PSE ion at m/z 915.5 and its mass spectrum are shown in Fig. 6 A and B. It elutes between minutes 21 and 22. The endogenous PSE ion at m/z 889.5 from fraction 10 also eluted between 21 and 22 min ( Fig. 6 C). The mass spectrum of the fraction 10 lipids eluting between minutes 21 and 22 of is shown in Fig. 6 D. Several ions corresponding to PSE with 32:1, 33cp, 34:2, 34:1, 35:1cp, and 36:2 acyl chains (the number of carbons in the acyl chains:number of unsaturations or cyclopropane units in the acyl chains; see Table 3 ) are detected ( Fig. 6 D). In addition, when synthetic PSE was added to a portion of fraction 10 and analyzed by LC-MS, the standard and the endogenous PSE coeluted, as seen by an increase in the ion intensity of the ion at m/z 915.5 ( Fig.  6 E, F).

Detection of PSE in MG1655 wild-type E. coli K-12
PSE was fi rst identifi ed in lipids of E. coli B (ATCC 11303). To demonstrate that PSE is not unique to E. coli B strains, we looked for PSE in wild-type E. coli K-12 MG1655. Following anion exchange fractionation, PSE was again detected in fraction 10 using normal phase LC-MS as the fi nal step in the identifi cation. As with the synthetic PSE ( Fig. 6A ) and the endogenous PSE from E. coli B ( Fig. 6 C), ions corresponding to PSE were detected in MG1655 fraction 10 between minutes 21 and 22 ( Fig. 7 A , peak marked with an asterisk). The mass spectra of the lipids eluting between minutes 21 and 22 are shown in Fig. 7 B. The ions at m/z 861.531, 863.534, 873.542, 875.540, 887.546, and neutral loss of 216 mass units, the ion at m/z 215, and the other head group derived ions were all observed (data not shown). The differences in the product ion masses are attributable to the differences in the acyl chain composition among the different ions. Table 3 summarizes the variety of PSE molecular species detected and their acyl chain compositions, as inferred from the exact mass and MS/MS analysis. MS/MS can suggest the location of the acyl chain on the glycerol backbone but does not determine it unequivocally ( 27,31 ). Accordingly, we report the total number of carbons and unsaturations in the acyl chains and the fatty acid product ions observed in the MS/MS. However, the acyl chain distribution in PSE is consistent with the overall fatty acid composition of the major glycerophospholipids of E. coli ( 8,(32)(33)(34).
Lastly, the 13 C isotope distribution of the PSE ion with m/z 889.5 matches the distribution predicted for a molecule with the molecular formula of PSE, C 45 H 82 N 2 O 13 P Ϫ exactly ( Fig. 4 ). This strongly suggests that the molecular formula provided above is the correct one.

Analysis of PSE synthetic standard
To confi rm our interpretation of the MS/MS spectrum of the proposed PSE, we analyzed a synthetic sample of PSE (Avanti Polar Lipids). The synthetic PSE is esterifi ed , which is overlaid on the observed mass spectrum of the PSE species with an exact mass near m/z 889.5.

DISCUSSION
The advent of ESI-MS has confi rmed the complexity of the lipidome of E. coli , which was already suspected based on early radiochemical studies ( 3 ). The major ions observed in the direct negative ion ESI-MS analysis of an E. coli total lipid extract, PG, PE, PA, and CL, often obscure the additional minor lipid ions. Prefractionation of the lipids prior to analysis by mass spectrometry uncovers ions arising from some of these minor lipids. For example, in this work, we were able to identify a series of ions corresponding to molecules of phosphatidylglycerol phosphate, the immediate precursor of PG ( 21,36,37 ) only after prefractionation of the total lipids using anion exchange chromatography ( Fig. 2 ). Likewise, we have now identifi ed a new family of molecules, the PSEs, as minor lipids from two strains of E. coli .
The structure of PSE was confi rmed by accurate mass measurements, MS/MS, and comparison to a synthetic standard. In addition to the detection of this novel lipid in E. coli B lipid extracts, fi ve different PSE species were observed in fractionated lipid extracts of MG1655.
The biosynthetic pathway responsible for the formation of PSE remains to be determined. Several possibilities can be envisioned. Phosphatidylserine synthase may be able to use a Ser-Glu dipeptide, perhaps the by-product of protein degradation, instead of serine. While it has been shown that the phosphatidylserine synthase of E. coli is very specifi c for serine, it does slowly transfer the CDP-DAG phosphatidyl moiety to glycerol, glycerol 3-P, and phos-phatidylglycerol ( 38,39 ). Accordingly, purifi ed phosphatidylserine synthase needs to be tested to determine if PSE can be formed in vitro using CDP-DAG and serylglutamate as substrates ( 39 ). If so, PSE should be missing in the available pss deletion mutants ( 40 ).
Alternatively, Shibuya, Yamagoe, and Miyazaki ( 41 ) have described the formation of novel E. coli glycerophospholipids in which a phosphatidyl group is transferred to the sugar alcohol, mannitol, to form phosphatidylmannitol. Cardiolipin synthase is implicated as the phosphatidyl donor in the formation of this and other sugar alcohol analogs. A similar activity could be at work in the formation of PSE. The free hydroxyl on the serine of a Ser-Glu dipeptide could be used in an alcoholysis reaction catalyzed by cardiolipin synthase ( 41 ). If that is the case, cardiolipin synthase deletion mutants should lack PSE.
PSE also may be formed by the addition of a glutamate directly to PS. However, PS levels in wild-type E. coli are very low ( 42 ). Only when the phosphatidylserine decarboxylase is inhibited by mutation do phosphatidylserine levels become readily detectable ( 42 ). In that scenario, PSE might accumulate in strains harboring psd mutations. In addition to the favorable equilibrium for the decarboxylation of PS to PE ( 43 ), PS levels may be low because of other enzymatic reactions, such as formation of PSE.
A fi nal hypothetical possibility is that PSE may arise from the processing of a protein, modifi ed with a phosphatidyl moiety on a Ser-Glu dipeptide within its primary sequence. Such a lipid modifi cation is not without precedent. The lipidation of the yeast autophagy factor Apg8 involves the linkage of a C-terminal glycine to the free amine of phosphatidylethanolamine ( 44 ). To our knowledge, protein lipidation of this sort has not been observed in E. coli . However, the known lipoproteins of E. coli are conjugated with a diacylglycerol moiety linked via a thioether to an N-terminal cysteine residue ( 45 ). A third acyl chain is attached to the N-terminal amine of the same cysteine ( 45 ). The identifi cation and structure of PSE suggests that a new type of protein lipidation might exist in E. coli . The phosphatidyl group may be transferred to a seryl-glutamate dipeptide within a protein from CDP-diacylglycerol.
Interestingly, no ions corresponding to phosphatidylserine linked to any of the other common amino acids were detected. Phosphatidylserylaspartate, in particular, which should be enriched in the same anion exchange fraction (10) as PSE, was not present (data not shown). This fi nding suggests that PSE is not an artifact of the extraction and therefore may have a defi ned role in the cell. In future investigations, the biosynthesis and degradation of PSE will be examined to determine its role in E. coli .