Electrospray Ionization Mass Spectrometry and Exogenous Heavy Isotope-labeled Lipid Species Provide Detailed Information on Aminophospholipid Acyl Chain Remodeling*

Mammalian cells maintain the phospholipid compositions of their different membranes remarkably constant. Beside de novo synthesis, degradation, and intracellular trafficking, acyl chain remodeling plays an important role in phospholipid homeostasis. However, many key details of this process remain unresolved, largely because of limitations of existing methodologies. Here we describe a novel approach that allows one to study metabolism of individual phospholipid species in unprecedented detail. Forty different phosphatidylethanolamine (PE) or -serine (PS) species with a deuterium-labeled head group were synthesized and introduced to BHK21 or HeLa cells using cyclodextrin-mediated transfer. Their metabolism was then monitored in detail by electrospray ionization mass spectrometry. Atypical PE and PS species were rapidly remodeled at both sn1 and sn2 position, yielding a molecular species profile similar to that the endogenous PE and PS. In contrast, remodeling of exogenous species identical or similar to major endogenous ones was more limited and much slower. Major differences in remodeling pathways and kinetics were observed between species within a class, as well as between corresponding PE and PS species. These data along with those obtained with pharmacological inhibitors strongly suggest that multiple lipid class-specific A-type phospholipases and acyl transferases are involved in aminophospholipid remodeling. In conclusion, the approach described here provides highly detailed information on remodeling of exogenously added (amino)glycerophospholipids and should thus be very helpful when elucidating the proteins and processes maintaining molecular species homeostasis.

i.e. exchange of fatty acid residues (2)(3)(4)(5)) is yet another process, which is considered to play an important role in phospholipid homeostasis of mammalian (6), yeast (7), and bacterial (8) membranes. Although the precise role of acyl chain remodeling in membrane phospholipid homeostasis remains unresolved, this process is known to be involved in several important phenomena including (i) arachidonic acid-dependent signaling (9), (ii) generation of cardiolipin species required for proper functioning of enzymes in the inner mitochondrial membrane (10,11), (iii) induced exo-and phagocytosis (12,13), (iv) synthesis of disaturated phospholipid species necessary for functional alveolar surfactant (14), (v) temperature adaptation of poikilothermic animals (15), and (vi) replacement of oxidized fatty acids of phospholipids (16 -18). However, despite its crucial role in several important biological processes, many aspects of phospholipid remodeling remain unresolved. For example, to what extent does acyl chain remodeling contribute to the steady state phospholipid compositions of cells or their subcellular organelles? What are the molecular characteristics rendering a phospholipid molecule susceptible for remodeling, i.e. how do the phospholipases involved recognize their targets? Are there specific phospholipases for each phospholipid class, or even for different species within a class? Which acyltransferases/transacylases are involved? Where does the hydrolysis and reacylation take place? How are remodeling and degradation (turnover) coordinated?
A key problem in resolving these questions has been the lack of suitable methods. Traditionally, phospholipid remodeling has been studied using radiolabeled phospholipid precursors, such as glycerol, choline, or ethanolamine and reversed-phase chromatographic separation of the molecular species (5,19). More recently, mass spectrometry has been proven to be a convenient tool for such studies (7, 13, 20 -22). However, even when labeled precursors are employed, it is virtually impossible to determine the remodeling pathways and kinetics of individual species because numerous molecular species are labeled already during the pulse (7,20,(23)(24)(25). Here, we describe a new approach that provides detailed information on remodeling of aminophospholipids and presumably other glycerophospholipids as well. We synthesized phosphatidylethanolamine (PE) 3 and -serine (PS) molecular species with a deuterium-labeled head group, introduced them one at a time into BHK21 or HeLa cells using cyclodextrin-enhanced transfer from donor vesicles, and then monitored their conversion to other species using electrospray ionization mass spectrometry (ESI-MS) and head group-specific scanning modes (24,26). Because initially only a single labeled species is present in the cell, remodeling pathways and kinetics could be resolved in unprecedented detail. As it is likely that the exogenous species are remodeled similarly to endogenous ones, at least qualitatively, this new approach should be very useful for the identification the specific phospholipases and acyl transferases involved in remodeling of (amino)phospholipids.

EXPERIMENTAL PROCEDURES
Lipids and Other Chemicals-Media and reagents for cell culture were obtained from Invitrogen, cholesterol and other unlabeled lipids from Avanti Polar Lipids (Alabaster, AL), D 3 -serine and D 4 -ethanolamine from CIL (Andover, MA), [ 13 C 18 ]oleic acid, methyl-␤-cyclodextrin, hydroxylamine, MAFP, NH 4 Cl, Triacsin C, and phospholipase D (Streptomyces sp.) from Sigma. S-BEL and R-BEL were from Cayman Europe (Tallinn, Estonia) and the solvents (HPLC-grade) from Merck. The PE and PS species with a deuterium-labeled head group were synthesized from corresponding phosphatidylcholine species and D 4 -ethanolamine or D 3 -serine, respectively, using phospholipase D-mediated transphosphatidylation as described previously for unlabeled lipids (27) except that the reaction volume was reduced 5-fold. The products were purified by normal phase HPLC (28), and their purity was confirmed by mass spectrometry. PE and PS labeled both to the fatty acid and head group were prepared by acylating sn1-oleyl-or sn1-palmitoyl-lysoPC by [ 13 C 18 ]oleate (29) followed by conversion of the sn1-oleoyl-and sn1-palmitoyl- [2-13 C 18 ]oleyl-PC to corresponding PE or PS species as described above. The concentrations were determined by phosphate analysis (30).
Cell Culture-Baby hamster kidney (BHK21) cells were grown as previously (31). HeLa-cells were cultured in DMEM containing 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin. Mycoplasma contamination was excluded using an ELISA method.
Introduction of Labeled PS and PE Species to Cells-Cholesterol, 1-palmitoyl-2-oleyl-phosphatidylcholine (POPC), a labeled PE or PS species, and di-22:0 phosphatidylcholine (5:4: 1:0.1, mol/mol) were mixed in chloroform, and the solvent was evaporated under a nitrogen stream followed by high vacuum for 1 h. 1 ml of PBS was added, and the sample was probesonicated for 3 ϫ 2 min with 30-s intervals. The sample was then centrifuged for 5 min at 3000 ϫ g to pellet any undispersed lipid and probe particles, and the small unilamellar vesicles (SUVs) in supernatant were used as the donors in cell labeling.
Cells grown to ϳ80% confluency on 6-cm dishes were washed twice with DMEM, SUV's (1 mol lipid), and methyl-␤-cyclodextrin (m-␤-CD) (8 mM) in 2 ml of DMEM were added, and the cells were incubated for 1 h at 37°C. After washing three times with DMEM, the cells were chased in DMEM with or without 10% fetal calf serum up to 24 h, washed with phosphate-buffered saline, scraped into 0.25 M sucrose, and moved to silane-treated screw cap tubes. The lipids were then extracted according to Folch et al. (32), except that the solvent contained 0.1 M HCl. For quantification of the lipid species by mass spectrometry, a mixture of internal standards was added at the one-phase stage of extraction (33). After evaporation of the solvents, the lipids were redissolved in chloroform/methanol (1:2, v/v) and stored at Ϫ20°C.
Mass Spectrometry and Data Analysis-After addition of aqueous NH 3 (4%), the sample was infused (6 l/min) into a Micromass Quattro Micro triple-quadrupole mass spectrometer operated as previously (33). The D 4 -labeled and unlabeled PE species were selectively detected using constant neutral-loss (NL) scanning of 145 and 141, respectively, in the positive ion mode (34). The D 3 -labeled and unlabeled PS species were detected using NL of 90 and 87, respectively, in the negative ion mode. The peaks were identified and quantified using the recently developed LIMSA software (35). Quantification was carried out using three internal standards for both PE and PS (36). For identification of the acyl chain sn-positions of the endogenous species, as well as the remodeling products derived from the exogenous species, PE and PS classes were isolated by normal phase HPLC (28), and the individual species were then subjected to collisionally activated dissociation (CAD) and product ion analysis. Albeit some previous and ESI-MS studies have indicated that it is possible to identify the sn-positions of the acyl chain of phospholipids by collisional fragmentation and product ion scanning (37)(38)(39), controversial data exist (39,40). Therefore, we carried out detailed studies with several synthetic standards to establish the acyl chain and sn-position-dependent fragmentation patterns of PE and PS. We then isolated the PE and PS classes from BHK21 cells using normal phase HPLC (28) and determined the sn position of the individual species based on the data obtained with standards as detailed in Supplemental Note S1.

Cyclodextrin-mediated Introduction of Aminophospholipid
Species to Cells-We have previously demonstrated that certain cyclodextrins can greatly enhance translocation of hydrophobic fluorescent phospholipids from lipid vesicles to cells (34,43). To test if this method is applicable for natural lipids as well, BHK21 cells were incubated in the presence of methyl-␤-cyclodextrin (m-␤-CD) and POPC/cholesterol-vesicles containing 10 mol% of a deuterium-labeled PE or PS. Indeed, most PE and PS species tested were efficiently incorporated into cells and, in many cases, rapidly metabolized (see below).
The amount of labeled PE or PS taken up by the cells varied from 0.3 to 4.5 nmol/dish, depending on the molecular species. This corresponds to ϳ0.5-5 mol% of total cellular phospholipid. The degree of incorporation diminished nearly exponentially with lipid hydrophobicity (data not shown). This is characteristic for CD-mediated monomeric lipid transfer (44), and therefore the labeled PE and PS species were most probably incorporated into cells (probably the exoplasmic leaflet of the plasma membrane) as lipid monomers rather than as components of adhering donor vesicles cf (45). Supporting this conclusion, only negligible amounts of the donor vesicle marker 22:0/22:0-PC, which is too hydrophobic to be efficiently transferred by cyclodextrins, was detected in cells (not shown).
M-␤-CD is known to extract cholesterol from cells (46), which could compromise some cellular functions. However, we found that if 50 mol% of cholesterol was included in donor vesicles, virtually no change in cellular cholesterol content was observed during the incubation (not shown), as expected from earlier studies (47). A modest elevation of POPC, the major component of the donor vesicles was observed in the cells after labeling, but its level was normalized during the 24-h chase. No other significant changes in the endogenous lipid composition were observed. Neither did the labeling procedure compromise cell viability as judged from Trypan Blue exclusion and lactate dehydrogenase assays (not shown) as well as by the metabolic data discussed below. Importantly, the less hydrophobic species, e.g. 14:1/14:1-PE and -PS, could also be introduced to cells without m-␤-CD albeit in significantly lesser amounts. Their remodeling was essentially identical to that observed in the presence of m-␤-CD (not shown), thus indicating that the metabolism of exogenous PE or PS is not significantly influenced by the presence of m-␤-CD or by the amount of lipid introduced into cells.
Remodeling of Exogenous Phosphatidylethanolamine Species-To study how the structure and the sn position of the acyl chains affects the metabolism of PE, we synthesized a variety of D 4 -labeled PE species, introduced them into BHK21 cells and then determined their remodeling pathways and kinetics using ESI-MS as detailed under "Experimental Procedures." Experiments carried out with 20 different exogenous PE species showed that remodeling of atypical (i.e. endogenously absent or minor) PE species was rapid and extensive, whereas those similar or identical to major endogenous ones were remodeled to a lesser degree (Figs. 1-3). Fig. 1A displays spectra for cells labeled with 14:0/14:0-PE and chased for 0, 3, 7, or 24 h. The metabolism of this atypical PE species was extremely rapid since a variety of labeled species, many of them more abundant than the precursor itself, were detected already after the 1-h pulse. During the chase, the precursor and most of the early remodeling products disappeared completely, and the profile of labeled species became similar to that of the endogenous ones by 24 h. Based on the relative abundance of the precursor and its remodeling products versus the chase time ( Fig. 2A) as well as information on the sn-position of the acyl chains (see Supplemental Note S1), the remodeling pathways of 14:0/14:0-PE could be deduced (Supplemental Fig. S1A).
Based on the results shown in Figs Notably, the turnover of the total pool of labeled PE was quite similar for all exogenous species studied (not shown), indicat-  FEBRUARY 8, 2008 • VOLUME 283 • NUMBER 6

Aminophospholipid Acyl Chain Remodeling
ing that selective precursor degradation, i.e. conversion to water soluble products, does not significantly contribute to the observed differences in the remodeling kinetics. Remodeling of several PE species was also studied in HeLa cells with analogous results (Supplemental Fig. S2).
Remodeling of Exogenous Phosphatidylserine Species-We next studied remodeling of exogenous PS species in BHK21 cells to see if the lipid head group affects remodeling pathways and kinetics. Analogously to the corresponding PE species, 14:0/14:0-PS was first converted to 14:0/18:1, which was then further remodeled to 18:0/18:1 and 18:1/18:1 (Fig. 4A), the two major endogenous PS species in these cells. Consequently, the profile of labeled species became similar to that of the endoge-nous ones (Fig. 4) by 24 h of chase. Notably, remodeling of 14:0/14: 0-PS was significantly slower than that of corresponding PE species (see Fig. 3
Because the comparison of the remodeling kinetics of corresponding PS and PE species is complicated by that different PS species are decarboxylated at different rates (31,34), we also carried out control experiments in the presence of 1 mM hydroxylamine, which blocks the PS decarboxylase without compromising cell viability (48). The results from these studies (not shown) were consistent with the conclusions drawn above regarding PS remodeling.
In summary, because both the remodeling pathways and kinetics of most PS species differed significantly from those of corresponding PEs, the remodeling PLAs and acylating enzymes acting on these two aminophospholipids in BHK21 cells are likely to be different.
Effect of PLA Inhibitors on Remodeling of PE-The data presented above demonstrate that both PLA 1 and PLA 2 activities are involved in remodeling of exogenous PE. Whereas little is known of the PLA 1 s potentially involved in remodeling of phospholipids, several types of PLA 2 s have been proposed to partic-  FEBRUARY 8, 2008 • VOLUME 283 • NUMBER 6 JOURNAL OF BIOLOGICAL CHEMISTRY 3681 ipate in this process. Particularly the Group VIA Ca 2ϩ -independent phospholipase iPLA 2 ␤ has been frequently implicated (49), albeit contradictory data have been obtained with some cell types (50). Also certain Group IV cytosolic phospholipases (cPLAs) may participate in phospholipid remodeling (51,52).

Aminophospholipid Acyl Chain Remodeling
To complement the pathway data indicating the involvement of multiple PLAs in aminophospholipid remodeling, we studied the effects of methyl arachidonyl fluorophosphonate (MAFP) and R-and S-enantiomers of bromoenol lactone (BEL) on remodeling of select PE species in BHK21 cells. MAFP effectively inhibits all cPLA 2 s as well as iPLA 2 ␤ (53), while S-BEL and R-BEL should inhibit iPLA 2 ␤ and iPLA 2 ␥, respectively, but not cPLA 2 s (54). Fig. 5A displays a spectrum for cells labeled with 14:1/14: 1-PE and then chased for 3 h in the absence of inhibitors. As can be seen, the precursor has virtually disappeared and significant In the case of PS, 1 mM hydroxylamine was added to block decarboxylation. The spectra have been normalized to the highest peak. Major intermediates and end products are indicated in the uppermost panels (A and F; CTRL), whereas only the remodeling products that are different from these are indicated in the other panels. The region corresponding to the remodeling products in which both of the original acyl chains have been exchanged is indicated by sn1 ϩ sn2. The experiments were repeated three times with similar results, and representative data are shown. Consistent results were also obtained with the other chase times tested (see Fig. 6).
amounts of primary as well as secondary remodeling products (m/z region indicated by "sn1 ϩ sn2") were present. When MAFP (25 M) was present, only a single remodeling product, i.e. 18:1/14:1-PE could be observed (Fig. 5B). In addition R-BEL markedly inhibited the remodeling, (Fig. 5C), whereas S-BEL had a rather weak effect (Fig. 5D). We next studied the effect of the inhibitors on remodeling of 14:0/18:0-PE and 16:0/20:4-PE. Rather than showing multiple spectra for each, the fraction of the precursor of the total labeled lipids versus time in the absence and presence of inhibitors is plotted in Fig. 6. In the case of 14:0/18:0-PE, neither R-nor S-BEL had any significant effect, whereas MAFP inhibited modestly (Fig. 6B). Interestingly, none of the inhibitors had any effect on remodeling of 16:0/20:4-PE (Fig. 6C).
Collectively, these data indicate that at least three different of PLA 2 s are involved in remodeling of exogenous PE species: one which is inhibited by MAFP ϾR-BEL ϾS-BEL; another, which is inhibited by MAFP only, and a third one, which is not inhibited by any of the inhibitors.
Effect of PLA Inhibitors on Remodeling of PS- Fig. 5F shows a spectrum for cells labeled with 14:1/14:1-PS and then chased for 3 h in the absence of inhibitors. Because this PS species is rapidly decarboxylated (see Figs. 3F and 6D), we studied the effect of the inhibitors in the presence of hydroxylamine, which blocks PS decarboxylase. In the absence of PLA inhibitors the precursor has nearly disappeared, and significant amounts of primary as well as secondary remodeling products are present after the 3-h chase. MAFP markedly inhibited the formation of secondary remodeling products (Fig. 5G), as observed for the corresponding PE species. However, in marked contrast to corresponding PE, both R-BEL and S-BEL strongly inhibited the remodeling of 14:1/14: 1-PS with virtually identical efficiency (Figs. 5, H and I, and 6D). Unlike 14:1/14:1-PS, MAFP had no effect on remodeling of 14:0/18:0-PS, whereas it was nearly eliminated by S-BEL. R-BEL had only a minor effect (Fig. 6E). None of the PLA 2 inhibitors affected remodeling of 16:0/20:4-PS (Fig. 6F) as was observed for the corresponding PE. Collectively, these data suggest that remodeling of exogenous PS involves at least three different of PLA 2 s: one which is strongly inhibited with MAFP and by R-and S-BEL; another which is strongly inhibited by S-BEL, but not with R-BEL and MAFP, and a third one which is not inhibited by any of the inhibitors. In conclusion, the inhibitor data strongly support the pathway data indicating that several different PLAs, capable of distinguishing both the lipid head group as well as the acyl chains, participate in remodeling of aminophospholipids in BHK21 cells.

Involvement of Acyl-CoA-dependent and -independent Acyltransferases in Remodeling of Exogenous PE and PS Species-To
investigate if multiple types of acylating enzymes are involved in remodeling of exogenous PE and PS species, we studied the effect of Triacsin C, a potent inhibitor of long chain acyl-CoA synthase (55). As shown in Fig. 6A, Triacsin C had no notable effect in the kinetics of remodeling of 14:1/14:1-PE, but markedly increased species with 18:0 and 16:0 at the sn1 position (Fig. 5E). Such species were nearly absent in the control cells (Fig.  5A). Similar results were obtained for 14:0/14:0-and 16:0/20:4-PE, i.e. Triacsin C had no significant effect on the remodeling rate (Fig.  6, B and C), but again a marked increase of remodeling products with 18:0 in the sn1 position was observed (not shown). Whereas the reason for this phenomenon is not clear, these results show that acyl-CoA-dependent acyl transferases/transacylases are not likely to be required for remodeling of PE. In contrast to PE, remodeling of all PS species studied was markedly inhibited by Triacsin C (Fig. 6, D-F). 14:1/14:1 PS was remodeled to 18:0/14:1, 18:1/14:1, and 16:0/14:1 indicating that remodeling of the sn2, but not the sn1 position of PS was inhibited (Fig. 5J). Parallel results were obtained for several other PS species studied (not shown). Accordingly, acyl-CoA-dependent acyltransferases or transacylases are implicated in sn2 remodeling of PS, whereas sn1 remodeling appears to be acyl-CoA-independent.
In summary, because Triacsin C modulated sn1 remodeling of PE but not of PS and, conversely, sn2 remodeling of PS but not of PE, it appears that the enzymes reacylating the sn1 and sn2 positions of PE are probably different from each other, as well from those reacylating the corresponding positions of PS.

Accumulation of Remodeling Products Corresponding to the Major Endogenous PE and PS Species Is Due to Their Slow
Hydrolysis Rather Than Acyl Chain Recycling-In principle, the accumulation of 18:1/18:1-PE and 18:0/18:1-PS as the main end products of remodeling of exogenous PE and PS species, respectively, could be due to 1) their slow hydrolysis by remodeling PLAs or 2) rapid hydrolysis followed by rapid reacylation with identical fatty acids. To choose between these options, we synthesized 16:0/ 13 C-18:1-D 4 PE, i.e. a species in which both the head group and the sn2 acyl chain is labeled, and then studied its metabolism in BHK21 cells. As shown in Fig. 7A, this doubly labeled PE was metabolized slowly to either 18:1/ 13 C-18:1-D 4 PE or 16:0/18:1-D 4 PE via exchange of the sn1 or sn2 chain, respectively. At later chase times, some 18:1/18:1-D 4 PE appeared indicating that both original fatty acids had been replaced (Fig. 7A). The 18:1/ 13 C-18:1, 16:0/18:1 and 18:1/18:1 species increased at comparable rates, indicating that both acyl chains turn over with similar kinetics. Notably, this turnover was much slower that that of atypical PE species (see Figs. 1-3), because the precursor still represented Ͼ40% of the total D 4 -labeled PE after 24h chase (Fig. 7A). Whereas we cannot exclude the possibility that the unlabeled sn1-palmitate would be recy-cled, i.e. hydrolyzed and reacylated with an identical fatty acid, this seems improbable because only a very minor fraction of exogenous 18:1/ 13 C Ϫ18:1-D 4 PE was converted to 16:0/18:1 (see Fig. 7B); i.e. there is no strong tendency to add a 16:0 chain to the sn1 position of PE.
In conclusion, these results strongly support the notion that remodeling PLA 1 s and PLA 2 s display a low activity toward major endogenous PE and PS species, such as 16:0/18:1 and 18:1/18:1, meaning that the kinetics remodeling are determined by the substrate specificity of those PLAs. On the other hand, the specificities of acyl transferases/ transacylases are crucial as they, together with the availability of different fatty acids, strongly influence the spectrum of remodeling products eventually formed.
Remodeling of Endogenously Synthesized PE Species-When BHK21 cells were briefly labeled with D 4 -ethanolamine, which incorporates to PE via the CDP-ethanolamine pathway, the profile of newly synthesized D 4 -PE species was initially quite different from that of pre-existing (unlabeled) PE (Supplemental Fig. S3). However, when the cells were chased in the presence of unlabeled ethanolamine, the labeled PE species were rapidly remodeled to yield a molecular species profile similar to that of unlabeled PE. After a 24-h chase, labeled PE consisted mainly of 18:1/18:1 and few other species. These data resemble those obtained for many exogenous species, thus indicating that the latter are remodeling similarly to the endogenous ones. However, it is not possible to compare the remodeling of exogenous and endogenous species in detail because, as mentioned under the Introduction, the remodeling pathways or kinetics of individual molecular species synthesized endogenously cannot be resolved due to simultaneous labeling of multiple species.
Recycling of Labeled Head Group Does Not Bias the Remodeling Data-The fact that very different profiles were obtained for different exogenous PE and PS species makes it unlikely that recycling of labeled head group would significantly contribute to the profiles of labeled species and would thus bias interpretation of the data. Would the labeled species derive mainly from head group recycling rather than acyl chain remodeling, identical profiles should be obtained for all exogenous species (due to use of the same endogenous DG precursor pool), which clearly was not the case. Nevertheless, to fully exclude this, we studied metabolism of 14:1/14:1-D 4 PE in the presence of unlabeled ethanolamine (1.6 mM), which should effectively dilute out the labeled ethanolamine eventually released from the exogenous species and thus prevent its reincorporation to PE. As shown in Supplemental Fig. S4, the profile of labeled PE species was essentially identical to the control lacking ethanolamine, thus confirming that recycling of labeled ethanolamine does not bias the interpretation of the remodeling data. Such studies were not carried out with PS, because serine is an intrinsic component of the cell culture medium used.

DISCUSSION
The present study demonstrates that far more detailed data on aminophospholipid remodeling pathways and kinetics can be obtained by ESI-MS/MS using exogenous phospholipids as compared with the use of labeled ethanolamine or serine as the precursors. This is because of the fact that only a single, rather than multiple, labeled species is initially present. Another key finding of this study is that among the 40 different PE and PS species studied in BHK21 cells the atypical (endogenously absent) ones were rapidly remodeled to endogenously abundant ones, while the exogenous species similar or identical to the endogenous were remodeled only to a limited degree. Analogous results were obtained for all exogenous PE and PS species tested with HeLa cells, whose endogenous PE and PS profiles differ significantly from those of BHK21 cells.
The accumulation of the 18:1/18:1-PE and 18:0/18:1-PS species as the main remodeling end products in BHK21 cells appears to be mainly due to: 1) their slow hydrolysis by remodeling PLAs and 2) preferential acylation of the lysoPE with oleate and lysoPS with oleate or stearate (Figs. 1, 4, and 7). Thus PLAs, by catalyzing the committed step, set the pace of remodeling of exogenous (and most probably endogenous) PE and PS species. However, the specificity of the acylating enzymes and the composition of the FA donor pools play a crucial role in determining the steady state profile of molecular species in the cells.
Most exogenous PE and PS species were remodeled both at sn1 and sn2 positions in agreement with earlier studies (5,56) and therefore both PLA 1 and PLA 2 activities must be involved. A novel finding is that both the kinetics and the sequence of the sn1 versus sn2 remodeling are very sensitive to the phospholipid acyl chain and head group structure. The PLA 1 s acting on PE seemingly prefer saturated and unsaturated chains, except for 18:1, which is not effectively cleaved (Figs. 1-3). In contrast, the PLA 1 s acting on PS seem to have a low affinity for saturated chains, particularly 18:0, as indicated by slow remodeling of the sn1-saturated PS species (Figs. 3 and 4). On the other hand, the PLA 2 s acting on PE seem to avidly cleave saturated chains, but poorly 18:1, while the PLA 2 s acting on PS seem to cleave preferentially short monounsaturated and polyunsaturated chains (Figs. 3 and 4).
The fact that structurally very similar aminophospholipid species, including positional isomers, were often remodeled with very different kinetics (see Fig. 3) strongly suggests that the susceptibility for remodeling is largely determined by the substrate specificities of the PLAs, rather than by e.g. differences in intracellular trafficking or partitioning between membrane domains.
While little is known of the PLA 1 s potentially involved in remodeling of phospholipids, several types of PLA 2 s, particu-larly members of the iPLA 2 and cPLA 2 families, have been implicated in this process (49,51,52), albeit most studies have concentrated on PC. The involvement of both iPLAs and cPLAs in aminophospholipid remodeling is supported by the present data obtained with the PLA 2 inhibitors MAFP, R-BEL, and S-BEL. However, these data imply that also other activities are involved. The observed differences in the effects of these inhibitors on remodeling of six different PE and PS species strongly support the involvement of multiple selective PLA 2 s, capable of distinguishing both the phospholipid acyl chains and head group, in aminophospholipid remodeling. However, more specific tools, such as RNA interference or other protein knockdown methods are needed to establish the identity of the specific proteins involved.
Triacsin C, an inhibitor of acyl-CoA synthase, did not inhibit remodeling of any PE species studied and thus reacylation of both sn positions of PE probably occurs by transacylation. Triacsin C also did not inhibit reacylation of the sn1 position of PS, thus again implicating CoA-independent transacylases. However, it markedly inhibited sn2 reacylation of PS, which is thus probably mediated by acyl-CoA-dependent acyltransferases/ transacylases. CoA-dependent and -independent acylation of lyso-PS has been observed in vitro (57,58). Because the sn1 position of PS was preferentially reacylated with 18:0 as found previously in vitro (59), whereas that of PE was mainly reacylated with 18:1 or 16:1, it is likely that different transacylases mediate the sn1 acylation of PS versus PE. This is consistent with previous in vitro data (6,60). In summary, at least three different enzyme activities appear to catalyze reacylation of aminophospholipids in BHK21 cells.
Beside the specificity of PLAs and acylating enzymes the pools of free/esterified fatty obviously also play an important role in determining the molecular species eventually formed, as shown by the remarkable changes in the phospholipid species profiles obtained by feeding different fatty acids to cells (61). Phospholipases and acyl transferases/transacylases putatively involved in phospholipid remodeling have been localized to the endoplasmic reticulum, the Golgi, mitochondria, nucleus, and peroxisomes (9,10,18,51,52,(62)(63)(64)(65). It is an intriguing possibility that organelles harbor different sets of remodeling enzymes, which could contribute to their specific phospholipid molecular species profiles. Evidence for such organelle-specific remodeling has been obtained for nuclear phosphatidylcholine in mammalian cells (66) and for yeast phospholipids in general (67). Thus the observed differences in remodeling of aminophospholipid molecular species might, at least in part, result from their different intracellular trafficking and distribution (68). Conversely, the remodeling process might contribute to interorganelle trafficking of phospholipids, since the rather hydrophilic lysolipid formed upon a PLA reaction, unless immediately reacylated, should be able to rapidly diffuse to another organelle wherein it might be reacylated to reform an intact phospholipid molecule.
Are exogenous aminophospholipids remodeled similarly to those synthesized endogenously? It seems very likely that their remodeling is qualitatively similar to that of the endogenous ones, because studies with labeled ethanolamine showed that the newly formed species were, like the exogenous ones, exten-sively remodeled eventually yielding a species pattern identical to the unlabeled ones (Supplemental Fig. S3). However, the kinetics of remodeling are likely to differ somewhat, because it takes time before the exogenous species fully mixes with the endogenous ones, despite the fact that intracellular trafficking of lipids is thought to be a relatively rapid process (69).
Albeit not studied here, it is very likely the exogenous species can be used to study remodeling of other glycerolipids as well. This, however, may require that the rate of transbilayer movement of the exogenous lipid to be studied is significant so that it will be readily accessible for the remodeling enzymes. While PS and PE are known to be rapidly translocated to the inner leaflet of the plasma membrane of all mammalian cells studied, flipping of e.g. PC seems to depend on the cell type (70).
Many crucial questions regarding glycerophospholipid acyl chain remodeling remain to be unanswered. First of all, it will be essential to identify the particular phospholipases and acyl transferases/transacylases involved in aminophospholipid remodeling. RNA interference (RNAi) or cells derived from knock-out animals should be very useful here. However, major efforts are required due to a large number of potential remodeling enzymes, off-target effects, and compensatory phenomena (71). Second, once the enzymes involved in remodeling have been identified, studies with model membranes with systematically varied compositions are needed to define the factors responsible for the high specificity of the remodeling process. Third, the subcellular site(s) of phospholipid remodeling needs to be established, and this probably requires in vitro studies with purified organelles. Fourth, remodeling of other glycerophospholipids such as PC, PI, and PA needs to be studied in various cell types to complete the picture. Yet, it is important to establish the relationship between the phospholipases involved in remodeling and those maintaining phospholipid class homeostasis. The approach described here should be very useful when studying many of these issues.