Phosphatidic Acid and N-Acylphosphatidylethanolamine Form Membrane Domains in Escherichia coli Mutant Lacking Cardiolipin and Phosphatidylglycerol*

The pgsA null Escherichia coli strain, UE54, lacks the major anionic phospholipids phosphatidylglycerol and cardiolipin. Despite these alterations the strain exhibits relatively normal cell division. Analysis of the UE54 phospholipids using negativeion electrospray ionization mass spectrometry resulted in identification of a new anionic phospholipid, N-acylphosphatidylethanolamine. Staining with the fluorescent dye 10-N-nonyl acridine orange revealed anionic phospholipid membrane domains at the septal and polar regions. Making UE54 null in minCDE resulted in budding off of minicells from polar domains. Analysis of lipid composition by mass spectrometry revealed that minicells relative to parent cells were significantly enriched in phosphatidic acid and N-acylphosphatidylethanolamine. Thus despite the absence of cardiolipin, which forms membrane domains at the cell pole and division sites in wild-type cells, the mutant cells still maintain polar/septal localization of anionic phospholipids. These three anionic phospholipids share common physical properties that favor polar/septal domain formation. The findings support the proposed role for anionic phospholipids in organizing amphitropic cell division proteins at specific sites on the membrane surface.

A unique lipid composition and lipid-protein interactions appear to exist at the transient membrane domain that defines the division site in bacterial cells (1). Using the cardiolipin (CL) 4 -specific fluorescent dye 10-N-nonyl acridine orange (NAO), we previously found CL-enriched membrane domains located at cell poles and near potential division sites in Escherichia coli (2). Subsequently others reported similar CL domains in Bacillus subtilis (3) and Pseudomonas putida (4). In addition, cell pole and division site enrichment in CL in E. coli was confirmed by lipid analysis of minicells spontaneously budded off from the cell poles of a ⌬minCDE mutant (5). We suggested that formation of CL domains at cell pole/division sites plays an important role in selection and recognition of the division site by amphitropic cell cycle and cell division proteins, such as DnaA (initiation of DNA replication at oriC), MinD (a part of MinCDE system preventing positioning of the divisome at cell poles in E. coli), and FtsA (bacterial actin, which is a linker protein for cytoskeletal protein FtsZ (bacterial tubulin), responsible for targeting the Z-ring to the mid-cell membrane domain). They interact directly with membrane phospholipids through specific amphipathic motifs enriched in basic amino acids, which confers the preference for anionic lipids (for references see Ref. 1). In E. coli the ATP-bound form of MinD recruits an inhibitor of Z-ring formation, MinC, to the membrane, whereas the topological regulator, MinE, induces hydrolysis of ATP bound to MinD resulting in release of MinD, and consequently MinC, from the membrane into the cytoplasm. As a result, all three proteins oscillate between the cell poles maintaining the maximum concentration of the inhibitor MinC at the cell poles and its minimum concentration at the cell center. Pole-to-pole oscillation of Min proteins occurs by dynamic redistribution of the proteins within a helical oligomeric structure that winds around the cell (for recent review and references see Ref. 6).
Our previous study of a mutant lacking phosphatidylethanolamine (PE) and containing highly elevated levels of phosphatidylglycerol (PG) and CL demonstrated a strong inhibition of cell division and aggregation of MinD and FtsZ/FtsA proteins at domains enriched in CL (7,8). To further investigate the role of lipids in the process of cell division, we chose an E. coli mutant with an opposite extreme in phospholipid composition to PElacking mutants, namely a ⌬pgsA mutant (pgsA encodes phosphatidylglycerol phosphate synthase, which catalyzes the committed step to PG and CL synthesis (9)). This mutant is devoid of PG and CL (contribute ϳ20 mole % of phospholipids in wild type) and contains higher levels of PE (ϳ90 mole % versus 80 mole % in wild type) (10,11). Interestingly, the ⌬pgsA null mutant accumulates elevated amounts of the phospholipid precursors, phosphatidic acid (PA) (ϳ4 mole %) and CDP-diacylglycerol (ϳ3 mole %), which are also anionic lipids that were proposed to fulfill the structural and functional roles of PG and CL (10,11). These results suggest that a minimum of 5-10% anionic lipid is required to support viability. Another minor anionic phospholipid, N-acylphosphatidylethanolamine was suggested to be present in wild-type E. coli (12), which, if proven true, might be also elevated in this mutant. Finally, if anionic lipids are essential for cell division, then we would expect these normally minor lipids to segregate into similar anionic lipid domains, as does CL.
In this report we identified N-acyl-PE in E. coli and, along with PA, its enrichment in polar/septal membrane domains of the ⌬pgsA mutant UE54 (11) lacking PG and CL. Thus E. coli has a mechanism for preferential segregation of anionic phospholipids to the polar/septal regions where several amphitropic proteins, which show preference for interaction with anionic phospholipids in vitro, are functionally located.
Growth of Cells-All cells were grown at 30°C in LB medium. For mass spectrometry (MS) analysis strains were grown in LB medium to an A 600 of 1.0, harvested by centrifugation at 500 ϫ g for 15 min at 4°C, washed once with phosphate-buffered saline (PBS) (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 ) and then resuspended in the same buffer. Pellets were frozen at Ϫ80°C until lipids were extracted.
Isolation of Minicells-Minicells were isolated according to a previous study (5) with modifications. Briefly, cells were grown to an optical density of A 450 of 0.2 in LB medium with 0.1% glucose, harvested, and resuspended in 100 ml of BSG buffer (1.5 M NaCl, 20 mM KH 2 PO 4 , 50 mM Na 2 HPO 4 (pH 7.7), 0.1% (w/v) gelatin). Large cells were isolated by centrifugation at 500 ϫ g for 10 min, and the pellet was resuspended in 10 ml of 50 mM KH 2 PO 4 /K 2 HPO 4 (pH 7.2), 5 mM MgSO 4 (buffer A) and stored at Ϫ80°C for lipid analysis. The supernatant was centrifuged at 20,000 ϫ g for 20 min. The resulting pellet was resuspended in 5 ml of BSG, loaded onto a sucrose gradient generated from layers of 3-ml 20%, 12-ml 10%, and 18-ml 3% (w/v) sucrose solutions in BSG, and then centrifuged at 2500 ϫ g for 10 min in a swinging bucket rotor. The minicell fraction in the 3% sucrose layer was collected. The sucrose gradient centrifugation step was repeated up to six times. The purity was checked by 4Ј,6Ј-diamidino-2-phenylindole (DAPI, Molecular Probes, Inc) staining and fluorescence microscopy (see below) to verify an enrichment of minicells over large cells. The final minicell preparation was washed in buffer A, pelleted by centrifuge at 2500 ϫ g for 10 min, resuspended in 1-2 ml of the same buffer, and frozen at Ϫ80°C.
Liposome Preparation-CL from beef heart and 1-palmitoyl 2-oleoylphosphatidic acid were purchased from Avanti Polar Lipids (Alabaster, AL). Liposomes were prepared by water bath sonication for 1-2 h at 0°C in a buffer containing 25 mM Tris-HCl, pH 7.5, and 50 mM KCl at a lipid concentration of 2-4 mg/ml, and were stored at Ϫ80°C.
TLC of Lipid Extracts-The cell pellet was resuspended in 0.1 ml of 0.5 M NaCl in 0.1 N HCl, and lipids were extracted by chloroform/methanol and examined by TLC separation according to a previous study (11).
Microscopy Study-To observe FtsZ-GFP localization or GFP-MinD movement in UE541 or UE542, overnight cultures were grown in LB medium with 10 M chloramphenicol or kanamycin to an A 600 of 1.6 and diluted 1:50 in the same growth medium followed by growth to A 600 of 0.6 at 30°C in the presence of 10 M isopropyl ␤-D-1-thiogalactopyranoside. This condition was chosen because it resulted in levels of FtsZ-GFP or GFP-MinD high enough to be detected by the camera but low enough to have practically no impact on MinD oscillation and cell division in wild-type cells.
Both types of cells were immobilized on microscope slides in 1% agarose as described before (7). Fluorescence images and differential interference contrast (DIC) were observed with a 100ϫ oil immersion objective on an Olympus BX60 microscope fitted with a GFP filter cube, and captured with a lightsensitive Photometrics CoolSnap FX cooled charge-coupled device camera driven by QED image capturing software and saved as Adobe Photoshop TIF files. Cell lengths were measured with Object Image software (Norbert Vischer). Timelapse images of GFP-MinD were taken every 6 -8 s, each with 2-to 4-s exposures using 2 ϫ 2 binning.
For microscopic examination of minicells and nucleoids of living cells, samples were stained with DAPI at a final concentration of 1 g/ml and viewed with a standard DAPI filter. The membrane of UE54 cells was stained with lipophilic FM4-64 (Molecular Probes, Inc.) at a final concentration of 4 g/ml as described before (15). For staining of anionic lipids in UE54 cells, NAO was added directly to the culture (final concentration 400 nM) in the growth medium and incubated for 1 h at room temperature. Liposomes (1 M of lipids) were stained with 2 M NAO in buffer containing 25 mM Tris-HCl (pH 7.5) and 50 mM KCl, washed by centrifugation, and resuspended in the same buffer. The stained cells or liposomes were immobilized on a cover glass with 1% agarose and viewed for green and red fluorescence images (2) as described above.
Extraction of Lipids From Cells for MS-Lipids were extracted from UE54 and MG1655 whole cells, UEM543 large cells separated from minicells, and purified minicells with a chloroform/methanol/aqueous Bligh and Dyer extraction system (16). Whole cell extracts were prepared as follows. Cell pellets resulting from the growth of MG1655 or UE54 culture in 50 ml of LB as described above were resuspended in 4 ml of PBS. Chloroform (5 ml) and methanol (10 ml) were added to generate a single-phase neutral extract (1:2:0.8, chloroform:methanol:PBS) (16). Cell debris was removed from the extract by centrifugation. The supernatant was transferred to a fresh tube and converted to a two-phase neutral extract (2:2:1.8, chloroform:methanol:PBS) by the addition of 5 ml of chloroform and 5 ml of PBS. Phases were resolved by centrifugation, and the upper phase was discarded. The lower phase was washed with 15 ml of pre-equilibrated neutral lower phase and centrifuged a second time to resolve the phases. The lower phase was dried under nitrogen, and the dried lipid films were stored at Ϫ20°C until analysis.
Lipid extracts were prepared from large cells and minicells as follows. Fifty microliters of the large cells or 150 l of minicells were used for each extraction. The aqueous volume was increased to 0.4 ml by the addition of PBS. Chloroform (0.5 ml) and methanol (1.0 ml) were added to generate a single-phase neutral extract (1:2:0.8 chloroform:methanol:PBS). Protein separation from the samples, conversion of the single-phase extract into two-phase system, and obtaining of the dry lipid films from organic phase were performed as described above and previously (17).
For quantification of lipid species using MS synthetic lipid standards were co-extracted with the cell pellets. For quantification of N-acyl-PE, 391 pmol of synthetic 1,2-dioleoylsn-glycero-3-phosphoethanolamine-N-nonadecanoyl (55:2 N-acyl-PE) standard (Avanti Polar Lipids, Alabaster, AL) was added to the resuspended cell pellets prior to addition of chloroform and methanol. For quantification of the ratio of PA to PE, 1.9 nmol of synthetic 12:0, 13:0 PA, and 1.82 nmol of 12:0, 13:0 PE (Avanti Polar Lipids, Alabaster, Al) were added to the initial extraction mixture. For quantification of the ratio of N-acyl-PE to PE, 0.98 nmol of synthetic 55:2 N-acyl-PE was used in place of the PA standard.
DEAE Fractionation of Total Lipid Extracts-Total lipid extracts derived from UE54 and MG1655 whole cells were fractionated on DE-52 cellulose (Whatman) as described previously (18).
Mass Spectrometry-For analysis of total lipid composition dried lipid films were dissolved in 100 l of chloroform/methanol (2:1) and infused into a quadrupole time-of-flight tandem mass spectrometer (QStar XL, Applied Biosystems) at 6 l/min. The mass spectra were obtained by scanning from 40 -2000 atomic mass units in the negative-ion, multichannel acquisition mode with the ESI source operating at the following settings: nebulizer gas, 20.0 p.s.i.; curtain gas, 20.0 p.s.i.; ion spray voltage, Ϫ4200 V; declustering potential, Ϫ55 V; focusing potential, Ϫ265 V; and declustering potential 2, Ϫ15 V. Spectra were acquired for 1 min at 1 scan per second and analyzed using Analyst QS 1.1 software (Applied Biosystems).
Collision-induced decomposition MS (MS/MS) was performed using collision energy of Ϫ40.0 V and N 2 as the collision gas. Spectra were acquired for up to 5 min at 1 scan per second.
Quantification of lipids was accomplished using reversedphase liquid chromatography MS (LC-MS). For the quantification of N-acyl-PE in whole cells, the dried lipid film was resuspended in 300 l of CHCl 3 , 200 l was transferred to a Target DP vial containing a Target Polyspring Inset (National Scientific), dried under nitrogen, and finally, resuspended in 100 l of mobile phase A containing 10% CHCl 3 . For quantification of the ratio of PA to PE the dried lipid film was resuspended in 0.8 ml of CHCl 3 , 200 l was transferred to a Target DP vial containing a Target Polyspring Inset dried under nitrogen, and then resuspended in 100 l of mobile phase solvent A containing 10% CHCl 3 .
Reversed-phase LC-MS was performed as described previously (19) by using a Shimadzu LC system (comprising a solvent degasser, two LC-10A pumps and a SCL-10A system controller) coupled to a QSTAR XL quadrupole time-of-flight tandem mass spectrometer (Applied Biosystems/MDS Sciex). A Zorbax SB-C8 reversed-phase column (Agilent, Palo Alto, CA) was used for all LC-MS analyses. For each LC-MS analysis 30 l of sample (prepared as described above) was injected on to the reversed-phase column. LC was operated at a flow rate of 200 l/min with a linear gradient as follows: 100% mobile phase A (methanol:acetonitrile:aqueous 1 mM ammonium acetate, 60:20:20) was held for 2 min and then linearly increased to 100% mobile phase B (100% ethanol containing 1 mM ammonium acetate) over 14 min and held at 100% mobile phase B for 4 min. The column was re-equilibrated to 100% mobile phase A for 2 min prior to the next injection. The post-column split diverted ϳ10% of the LC flow to the ESI source of the mass spectrometer. All MS data were analyzed using Analyst QS software (Applied Biosystems/MDS Sciex). Individual lipid species were quantified by taking the ratio of the peak area of the extracted ion current from the LC-MS chromatogram to the peak area of the extracted ion current for the appropriate synthetic internal standard (19).

Lack of PG and CL Decreases the Length of the UE54 Mutant
Cell-TLC analysis of total lipid extracts from the ⌬pgsA strain (UE54) according to a previous study (11) showed no differences in its phospholipid composition compared with other ⌬pgsA strains reported before (11). As with the other ⌬pgsA strains, UE54 is viable only if it carries two additional gene mutations. An lpp mutation is necessary to prevent accumulation of the nascent major outer membrane lipoprotein, which requires PG for its modification (10,20,21). The Rcs phosphorelay system is constitutively activated in ⌬pgsA mutants resulting in the cell lyses at 37-42°C, which is prevented by a ⌬rcsF mutation (22). We compared the growth rates of UE54 lacking PG and CL and its two parental pgsA ϩ strains, MG1655 and UE53 (lpp-2 ⌬rcsF), with wild-type phospholipid composition. To exclude additional factors contributing to cell growth and division at high temperatures all strains were grown at 30°C. In agreement with previous findings for the ⌬pgsA mutant (10), the doubling times of UE54 were ϳ90 min versus 60 min for the two parental strains (data not shown). During the exponential phase of growth a similar average cell length was found for MG1655 and UE53 (3.66 Ϯ 0.1 m), whereas the UE54 cells were ϳ20% shorter (2.93 Ϯ 0.06 m) (Fig. 1A). Cell width was the same (1.25 Ϯ 0.05 m) for all three strains.
FtsZ-GFP Localization and GFP-MinD Oscillation-FtsZ-GFP (green fluorescent protein), produced from a single copy gene fusion, was introduced into UE54 to generate UEM542. Fig. 1B shows the normal mid-cell localization of FtsZ-GFP in the majority of UEM542 cells. FtsZ rings were observed in ϳ77% of cells (206 cells analyzed). We also examined the dynamic localization of GFP-MinD in the UEM541 cells expressing a single-copy gene fusion of GFP-MinD along with MinE (see "Experimental Procedures"). In these cells, GFP-MinD oscillated from pole to pole with cycle times between 60 and 80 s (Fig. 1C) similar to that of UE53 with wild-type phospholipid composition (data not shown).
Staining of Anionic Lipid Membrane Domains with NAO-To study possible formation of anionic lipid membrane domains in ⌬pgsA cells, we used a visualization technique developed previously for staining CL domains in bacteria with the fluorescent dye NAO (2) and successfully used by other groups (3,4,23). Association of NAO with CL results in the appearance of a red emission maximum in the dye fluorescence spectrum, and CL membrane domains emitting both green and red fluorescence were previously demonstrated in E. coli and Bacillus subtilis with wild-type phospholipid composition (2,3,23,24). In contrast to NAO bound to CL, NAO bound to other anionic phospholipids does not exhibit specific red emission. Only green fluorescence of the membrane in ⌬cls mutants of B. subtilis (3) and E. coli (2,23) or the PG-containing membrane domain of Streptococcus pyogenes (25) was observed. In vitro experiments with PA containing liposomes demonstrated this phospholipid also binds NAO, but only green fluorescence of the liposome was observed under the microscope ( Fig. 2A).
Staining of UE54 mutant cells with NAO revealed green fluorescence membrane domains in the regions of the cell poles and at the cell center (Fig. 2, B and C). Some cells in the population showed only cell pole domain fluorescence while other cells exhibited both cell pole and mid-cell or only mid-cell domain positioning of NAO, which may be a result of cells at different stages of the cell cycle. Staining of cells with a nonspecific lipophilic fluorescent dye, FM4-64, demonstrated a uniform distribution along the membrane (Fig. 2D). Comparison of green fluorescence domain positioning in UE54 and UE53 cells (Fig. 2E) suggests that in contrast to UE53 the polar staining in UE54 cells containing a mid-cell domain appears less frequently than staining of the central domain. This might be an indication of preferential targeting of anionic phospholipids to division site in UE54 cells. However, we cannot completely exclude that the domains might not be equally visible due to different orientations of the UE54 cells in the field.
These results are consistent with the targeting of anionic lipids to the cell poles and division sites of UE54 cells. To reveal the nature of anionic lipids concentrated at cell poles/septal sites, we analyzed the phospholipid composition of minicells (as described below), which are produced due to abnormal cell division occurring at the cell poles.

MS of Lipid Extracts from UE54 and MG1655 Cells-Initially
we compared lipid composition of UE54 to MG1655. Analyses were performed by negative-ion electrospray ionization timeof-flight MS (ESI-MS). As expected the lipid composition of the two strains differed drastically. MG1655 showed predominant ions corresponding to PG (Fig. 3A). The ions at m/z 691. 44 Although PE represents roughly 80% of the total phospholipids in wild-type cells (9), it ionizes much less efficiently than PG in the negative-ion mode, and, therefore its true abundance is not reflected by its ionization (17). To confirm the presence of PE, CL, and PA in MG1655, the total lipid extract was fractionated on DEAE-cellulose (18), which separates the phospholipids according to charge. ESI-MS of the appropriate fractions revealed ions corresponding to PE, CL, and PA as expected (data not shown).
In contrast, the mass spectrum of the lipid extract derived from UE54 cells contained predominately ions corresponding to PE (Fig. 3B) consistent with the inability of these cells to make PG and CL (11). The ions at m/z 688. 47 ions of PA with 32:1 and 34:1 acyl chains, respectively. Mass spectrometry of the UE54 lipid extract following fractionation on DEAE-cellulose also reveals additional species of PA (see below) and the presence of ions corresponding to CDP-diacylglycerol (data not shown), both precursors to PG and CL. All lipids were identified by exact mass and MS/MS, verifying by more rigorous means the previously reported phospholipid composition of this strain (11).
Inspection of the UE54 spectrum in the range m/z 900 -1000 revealed the presence of several singly charged ions that did not correspond to any of the known lipids of E. coli (data not shown). Although these ions, m/z of 924.7, 926.7, 952.7, 954.7, 978.7, and 980.7, are abundant in UE54, in the MG1655 extract the ions were obscured by abundant ions corresponding to acyl-PG (26). Analysis of the lipid extracts using reversed-phase LC-MS more clearly revealed the presence of these ions in UE54 (Fig. 4A) as well as in MG1655 (Fig. 4B)  Identification of N-acyl-PE-We set forth to determine the identity of the unknown ions that appear to be enriched in the UE54 lipid extract. One atomic mass unit separates the monoisotopic peak from the ion corresponding to the molecule with one 13 C indicating that these are singly charged ions. The even nominal monoisotopic mass of each unknown ion is suggestive of the presence of an odd number of nitrogen atoms in the molecule. The difference in mass among the unknown ions corresponds to the differences expected due to acyl chain heterogeneity. To determine the identity of the lipid responsible for these ions, MS/MS analysis was performed as shown in   (Fig. 5B) correspond to loss of an acyl chain from this product ion as a fatty acid (RCO 2 H), 16:0 and 18:1 acyl chain remaining esterified, respectively. The remaining product ions at m/z 378.23 and 404.26 (Fig. 5B) could not be explained as simple acyl chain losses from the product ion at m/z 716.51. These data and comparison of the exact mass with putative product ions suggest that the ion at m/z 980.8 is derived from N-acyl-PE having 52 total carbons in the three acyl chains, two of which contain one unsaturation. Fig. 5C shows two possible molecular species consistent with the product ion spectra in Fig.   5A. The product ions at m/z 378.24 and 404.26 can now be explained as N-acyl-PE in which the acyl chain is 16:0 or 18:1, respectively (Fig. 5, B and C). Analysis of the other unknown ions by MS/MS yielded similar fragmentation patterns consistent with N-acyl-PE species with different acyl chain compositions (data not shown).
As with other multiply acylated glycerophospholipids, a given molecular mass of N-acyl-PE can represent several distinct molecular species. At least two species can explain the ion at m/z 980.83 (Fig.  5C), one with an 18:1 acyl chain attached to the terminal amine and a 16:0 acyl chain esterified to the glycerol (left structure) and one in which a 16:0 acyl chain is attached to the amine and an 18:1 acyl chain is esterified to the glycerol (right structure). In addition, the MS/MS technique we employed does not allow for definitive assignment of the acyl chains to the sn-1 or sn-2 position. The structures shown are based on the biological activity of the glycerol 3-phosphate and monoacylglycerol 3-phosphate acyl transferases (28,29). Although we can definitively identify some of the acyl chains attached to the amine of N-acyl-PE fragment ions, it is likely, especially due to low yield of the N-acyl-PE fragment ions, that other acyl chain combinations are present. Therefore, we present the molecular species of N-acyl-PE by the total number of carbons and unsat-  urations in the acyl chains. Eleven different N-acyl-PE molecular species were detected in E. coli (Table 1).
To confirm the interpretation of the MS/MS spectrum, synthetic N-acyl-PE (Avanti Polar Lipids, Alabaster, AL) was analyzed by MS/MS. Fig. 5D shows the MS/MS spectrum of the synthetic standard (55:1, m/z 1022.82). The fragmentation pattern is very similar to that shown in Fig. 5 (A and B). The ion yield of the product ions in the range of 370 -900 atomic mass units is similarly low. In addition, similar ions corresponding to N-acylphosphoethanolamine are present in the two spectra. The synthetic standard yields a single N-acylphos-phoethanolamine fragment ion with 19:0 acyl chain (m/z 420.29) consistent with the structure and synthesis. The ion at m/z 402.28 in the MS/MS spectrum of the standard corresponds to loss of water from the product ion at m/z 420.29. Additional differences in the product ion spectra are explained by the difference in acyl chain composition between the two samples. Taken together these data strongly suggest that the unknown ions revealed in the UE54 lipid extract are various molecular species of N-acyl-PE. This constitutes the definitive identification of N-acyl-PE in E. coli, a lipid well known to be present in eukaryotic cells.
Quantification of N-Acyl-PE Levels in UE54 and MG1655 Cells-There appeared to be enrichment in the amount of N-acyl-PE in UE54 as compared with MG1655. The absolute level of N-acyl-PE in the two cell lines was determined by extracting the cellular lipids in the presence of the 55:2 N-acyl-PE synthetic standard. The ions derived from the standard are not isobaric with any ions detected in either the MG1655 or UE54 lipid extracts. The linear range of the MS response to the amount of the 55:2 N-acyl-PE was from 1 to 30 ng of the standard injected into the reversed-phase LC column (data not shown). Fig. 6 shows the amounts of each N-acyl-PE species in the UE54 and MG1655 lipid extracts. When the picomoles of all N-acyl-PE molecular species are summed there is ϳ180 pmol of N-acyl-PE/mg protein in MG1655 and 790 pmol of N-acyl-PE/mg protein in UE54. Overall, there is ϳ4-fold more N-acyl-PE in UE54 than in MG1655, and N-acyl-PE represents ϳ0.4% and 0.1% of the total lipid content, respectively (9).
Lipid Composition of Minicells-Next we studied the distribution of phospholipids along the cell membrane in UE54 mutant cells. Minicells are small anucleated cells produced by min mutants of E. coli as a result of aberrant cell division at the cell poles and were previously used to study protein and lipid composition of the E. coli division site/cell poles (5,30). In a minicell-forming derivative of UE54 (UEM543) asymmetric division at the cell pole results in formation of a minicell and a short filament with at least two nucleoids, because no division at mid-cell between nucleoids occurs (31). In contrast to "minicells" we define short filaments and normal size cells in the UEM543 culture as "large" cells. Fig. 7A (left) shows a number of UE543 cells with aberrant division at the cell poles. The phospholipid composition of the purified UEM543 minicells (Fig. 7A  (right)) was compared with the phospholipid composition of UEM543 large cells using ESI-MS. Spectra of UEM543 large and minicells, as was the case of its parental strain UE54, showed peaks for PE, PA, diacylglycerol, and N-acyl-PE species (data not shown).
To characterize the phospholipid composition of the large cells and mini cells in more detail and to compare ratios of PA to PE or N-acyl-PE to PE in large cells and minicells, reversed-phase LC-MS was performed. Synthetic standards for PA, N-acyl-PE and PE were co-extracted with the samples to quantify the phospholipids. Table 2 shows the variety of PA and PE species found both in minicells and large cells. Fig. 7B demonstrates differences in the ratios of individual PA species to total PE between minicells and the large cells.
There is approximately a 2.5-fold increase in the overall PA/PE ratio in minicells compared with the large cells. Because the ionization yields for PE species (32) are inversely proportional to chain length, the PA/PE ratios in large cells and minicells does not reflect the absolute ratios in each cell type due to normalization to a short chain PE. However, because the distribution of lipid species was the same in large cells and minicells, the calculation for the relative increase of each PA species in minicells is not affected.

Membrane Domains in E. coli Lacking Cardiolipin
Nearly the same variety of N-acyl-PE species were found in the UE543 minicell and large cell extracts as in the UE54 cells extract (Fig. 7C versus Fig. 6). Approximately a 5-fold increase in the overall N-acyl-PE/PE ratio was found for minicells relative to the large cells using lipid standards as above.
These experiments demonstrate significant enrichment of the anionic phospholipids PA and N-acyl-PE at the cell pole and division site membrane domains in the UE54 mutants lacking the major wild-type anionic phospholipids, PG and CL. The increase in these ratios in the minicell preparations is likely significantly underestimated, because there is only enrichment of minicells in the final samples rather than a homogenous population of minicells.

DISCUSSION
In this study ESI-MS analysis of the E. coli mutant UE54, which completely lacks the anionic phospholipids CL and PG, demonstrated that cells up-regulate the levels of other anionic phospholipids to compensate for the loss of PG and CL. Consistent with previous studies using TLC (10,11), we observed elevated levels of PA and CDP-diacylglycerol in the mutant by MS. Direct quantification of CDP-diacylglycerol was not pursued because ion-exchange chromatography is required to effectively observe this lipid using MS and was not feasible because of the low yield of minicells. However, analysis of this strain also revealed yet another method by which cells can increase the levels of anionic phospholipids for survival through elevated levels of N-acyl-PE by making an anionic phospholipid directly from the abundant PE. The lack of PG, and therefore PG, in UE54 allowed the definitive identification of this lipid in E. coli. N-Acyl-PE was enriched at least 4-fold in the anionic phospholipid-deficient mutant relative to the parental strain.
Despite these major alterations in lipid composition UE54 cells divide in a near normal manner. This is in contrast to the situation of profound inhibition of the cell division in the PElacking filamentous ⌬pssA mutant, in which membranes contain only anionic phospholipids primarily made up of PG, CL, and PA (8). UE54 cells, in contrast, were even shorter than the parental cells. Interestingly, a similar decrease in the cell length was found in the case of a cls mutant of Pseudomonas putida. The authors concluded that changing the proportion of anionic phospholipids led to cell division before the cell had reached the average size of wild-type cells (4). The lack of CL might result in alteration in the energy coupling process in the mutant, because many energy-transducing enzymes require CL or PG for optimal function (33). This might lead to slower growth and a decrease in the cell length. High percentage of FtsZ ring positioning in the slow growing short UE54 cells might also be a symptom of the loss of coordination between growth rate and cell cycle specific events (34).
MinD oscillation in UE54 did not differ from its parental strain UE53 with wild-type phospholipid composition. However, based on our previous finding on the preference of MinD for anionic phospholipids (7), we expected that the large decrease in the level of anionic phospholipids should somehow affect the MinD dynamics and cell division. Preference for anionic phospholipids implies that their level in the membrane should regulate special temporal behavior of MinD in the cell. Indeed, the high content of PG and CL in the ⌬pssA mutant increased the dwell time for MinD on the membrane surface thus disrupting the normal cycling pattern of MinD (7). The paradox of the UE54 mutant deficient in the major anionic phospholipids lies in its normal MinD dynamics despite the apparent affinity of MinD for anionic phospholipids. On the other hand, MinD preference for anionic phospholipids might play a role at the step of the nucleation of the MinD polymers at the cell pole (35) where anionic phospholipid domains appear to exist. The assumption for a nucleation site for MinD polymerization was formulated previously in our hypothesis (36) and was suggested in the "Min proteins polymerization-depolymerization model" by (37). The multistranded MinDE polymerization model (38) also supports this assumption. In the case of UE54 cells the above models would require enrichment and localization at the cell poles of the minor residual anionic phospholipids found in this mutant, which would explain the paradox of this mutant and further support the involvement of anionic lipid domains in the cell division process.
We established by using two independent approaches, NAO staining of the cells and ESI-MS spectrometry of lipid extracts from minicells, that, despite the lack of CL and PG, anionic lipid-enriched membrane domains still formed at the cell poles/septa areas of the UE54 mutant. The visualization technique, which was developed for staining CL in bacteria and uses fluorescent dye NAO (2)(3)(4)23), was applied in this work to E. coli mutant cells lacking both CL and PG. NAO has a higher affinity for CL than for other anionic lipids, but only the association with CL results in the red emission maximum in the spectrum due tobond stacking of the dye bound to CL molecules packed in microdomains within membranes (39 -41). Interaction with the other anionic phospholipids is charac- terized by lower affinity 5 and results only in green fluorescence (25). Due to the absence of PG, which appears to create a relatively high background of NAO green fluorescence in the membrane, staining of UE54 containing low levels of anionic phospholipids allowed visualization of membrane domains formed by residual anionic lipids, PA and N-acyl-PE. The labeling of the membrane using the lipophilic stain FM4-64 indicated uniform distribution of total lipid throughout the membrane. The second approach, ESI-MS analysis of phospholipid composition of minicells and large cells produced by UE543, demonstrated that the level of both PA and N-acyl-PE was increased at the cell pole/division site membrane domains of the mutant and provides a molecular basis for the viability of ⌬pgsA mutants. This result coupled with the similar localization of CL determined by NAO labeling (2) and minicell analysis (5) strongly indicates that bacteria have a mechanism for enriching the polar/septal regions with anionic lipids of similar physical properties (see below). The fact that several amphitropic proteins, which bind with higher specificity to anionic phospholipids in vitro, co-localize to these same regions supports a functional role for anionic lipid domains.
Comparison of the properties of PA and CL shows that PA can replace CL in its role in the formation of the cell pole/septa membrane domains. An equilibrium mechanism based on lipid micro-phase separation was proposed to explain this pattern of CL localization in rod-shaped bacteria (42). According to the model clusters of lipids, such as CL or PA, that induce high negative curvature strain in a bilayer can localize to the cell poles and septa. It is well documented that PA plays a key role in recruiting proteins involved in membrane tubulation, fission, and fusion processes in mammalian cells (for review see Ref. 43). It was also suggested that PA might play a direct role in membrane bending and destabilization, due to its special biophysical properties (44). Study of the phase behavior of PA demonstrated significant similarity between PA and CL. Both phospholipids demonstrate Ca ϩ2 -dependent H II non-bilayer phase formation driven by lipid charge neutralization and headgroup dehydration, which decrease the effective size of the lipid headgroup (Ref. 44 and references within). The two pK a values of the phosphate headgroup of PA are ϳ3.5 and 8.5 (44) and are similar to the two pK a values of CL (45) but not to the other anionic phospholipids. This means that, in contrast to other non-bilayer phospholipids such as PE, the shape of PA and CL molecules can be modulated by Ca ϩ2 (and other divalent cations) and pH, which might be an important factor in the transformation of the mid-cell membrane domain into the cell pole domain (1). Interestingly, analysis of the phospholipid composition of another Gram-negative bacterium, Neisseria gonorrhoeae revealed PE, PG, PA (ϳ10 mol% of the latter), and only traces of CL (46). This example supports the in vivo mutual interchangeability of CL and PA.
Hemly et al. have previously suggested the presence of N-acyl-PE in E. coli. They identified the N-acyl-PE using TLC (12). Here we definitively identify numerous molecular species of N-acyl-PE in E. coli using ESI-MS and MS/MS analysis in comparison with a synthetic N-acyl-PE. We have also determined that N-acyl-PE represents ϳ0.1% of the total phospholipid of wild-type cells. It is surprising that Hemly et al. would have been able to detect, without pre-fractionation or radiochemical labeling, a lipid of this low abundance using TLC. The species they reported to be N-acyl-PE may indeed be the more abundant PG, which would migrate similarly in the solvent systems used in their analysis.
N-Acylation of PE to form N-acyl-PE results in the conversion of the zwitterionic PE into a negatively charged phospholipid. N-Acyl-PE species are found in plants, microbes and mammals as minor phospholipids (47,48). In mammals, N-acyl-PE is formed by transacylation of the sn-1 acyl chain of a phospholipid onto the amine of PE. N-Acyl-PE serves as the precursor to a class of bioactive lipids called N-acylethanolamines, which are produced in a variety of tissues and implicated in the modulation of pain sensation, feeding, and sleep cycles (49, 50, 51 and references within). Both in animal tissues and plants N-acyl-PE accumulates under stress conditions and plays a dual role, as a direct precursor for N-acylethanolamine signaling pathways and as a stabilizing component of the membrane (49,52). Pathways of biosynthesis and function of N-acyl-PE in E. coli cells remain to be studied. However, its role in anionic lipid domain formation in the UE54 mutant membrane was discovered in our experiments. Notably, N-acyl-PE shares with CL and PA specific physical properties, such as H ϩ -and Ca ϩ2 (or Mg ϩ2 )-facilitated transition to the H II phase (53,54). This means that N-acyl-PE can fulfill requirements for phospholipids forming micro domains at cell poles and division sites in the rod-shaped bacterial cell (42). Irrespective of the mechanisms and signals involved in N-acyl-PE formation, lipids with common biophysical properties as well as negative charge such as PA, CL, and N-acyl-PE segregate to the poles/division sites of cells where they form domains that appear to play a functional role as nucleation sites for amphitropic proteins.