The CDP-ethanolamine Pathway and Phosphatidylserine Decarboxylation Generate Different Phosphatidylethanolamine Molecular Species*

In mammalian cells, phosphatidylethanolamine (PtdEtn) is mainly synthesized via the CDP-ethanolamine (Kennedy) pathway and by decarboxylation of phosphatidylserine (PtdSer). However, the extent to which these two pathways contribute to overall PtdEtn synthesis both quantitatively and qualitatively is still not clear. To assess their contributions, PtdEtn species synthesized by the two routes were labeled with pathway-specific stable isotope precursors, d3-serine and d4-ethanolamine, and analyzed by high performance liquid chromatography-mass spectrometry. The major conclusions from this study are that (i) in both McA-RH7777 and Chinese hamster ovary K1 cells, the CDP-ethanolamine pathway was favored over PtdSer decarboxylation, and (ii) both pathways for PtdEtn synthesis are able to produce all diacyl-PtdEtn species, but most of these species were preferentially made by one pathway. For example, the CDP-ethanolamine pathway preferentially synthesized phospholipids with mono- or di-unsaturated fatty acids on the sn-2 position (e.g. (16:0-18:2)PtdEtn and (18:1-18:2)PtdEtn), whereas PtdSer decarboxylation generated species with mainly polyunsaturated fatty acids on the sn-2 position (e.g. (18:0-20:4)PtdEtn and (18:0-20:5)PtdEtn in McArdle and (18: 0-20:4)PtdEtn and (18:0-22:6)PtdEtn in Chinese hamster ovary K1 cells). (iii) The main PtdEtn species newly synthesized from the Kennedy pathway in the microsomal fraction appeared to equilibrate rapidly between the endoplasmic reticulum and mitochondria. (iv) Newly synthesized PtdEtn species preferably formed in the mitochondria, which is at least in part due to the substrate specificity of the phosphatidylserine decarboxylase, seemed to be retained in this organelle. Our data suggest a potentially essential role of the PtdSer decarboxylation pathway in mitochondrial functioning.

Phosphatidylethanolamine (PtdEtn) 2 is the second most abundant phospholipid subclass in mammalian cells, compris-ing 15-25% of total phospholipids (1). Three pathways are present for PtdEtn biosynthesis. The majority of PtdEtn is synthesized via the CDP-ethanolamine pathway and PtdSer decarboxylation, whereas the third route, calcium stimulated base-exchange, is of little significance (2). In the CDP-ethanolamine (CDP-Etn) pathway, ethanolamine is converted to PtdEtn by the sequential actions of ethanolamine kinase, CTP: phosphoethanolamine cytidylyltransferase and finally choline/ ethanolaminephosphotransferase. Choline/ethanolaminephosphotransferase has a dual specificity, as it can use both CDPcholine (CDP-Cho) and CDP-Etn as substrates for the biosynthesis of phosphatidylcholine (PtdCho) and PtdEtn, respectively (3). In addition to choline/ethanolaminephosphotransferase, a CDP-choline-specific cholinephosphotransferase is available for PtdCho biosynthesis (4).
The PtdSer decarboxylation pathway for PtdEtn biosynthesis was first described by Borkenhagen et al. (5). In this pathway PtdSer synthesized from PtdCho or PtdEtn by phosphatidylserine synthase-1 and -2, respectively (2,6,7), is decarboxylated by the enzyme phosphatidylserine decarboxylase (PSD) to generate PtdEtn. To date, only one mammalian PSD has been cloned (8), and the enzyme was shown to be located on the external aspect of the inner mitochondrial membrane (9,10). Because PtdSer synthesis occurs in the ER and especially in ER-related membranes termed mitochondria-associated membranes (MAM) (2,11), PtdSer decarboxylation requires transport of PtdSer from its site of synthesis to the inner mitochondrial membrane, where PSD is located (12)(13)(14).
The relative importance of the CDP-Etn and PtdSer decarboxylation pathways to overall PtdEtn biosynthesis appears to vary depending on cell type and the availability of the substrates ethanolamine and serine, respectively. From studies in Chinese hamster ovary (15,16) and baby hamster kidney (17) cells that were cultured in medium with fetal bovine serum being the sole source of ethanolamine, it was concluded that PtdSer decarboxylation was the major pathway for PtdEtn synthesis. However, observations in hamster heart (18) and in rat liver, hepatocytes, heart, and kidney (19,20) illustrated that the vast majority of PtdEtn is synthesized via the CDP-Etn pathway. A possible explanation for these opposite results is the availability of exogenous ethanolamine. All studies mentioned above employed incorporation of radioactive serine and ethanola-mine into PtdEtn. Because of the presence of endogenous pools of unlabeled ethanolamine and serine, the quantitative contributions of both pathways to overall PtdEtn synthesis are still unclear.
An intriguing question is why two pathways for PtdEtn biosynthesis exist, whereas only one pathway is available for the de novo synthesis of PtdCho in most mammalian cells. First, a second biosynthetic pathway could serve as a backup pathway under conditions where one of the two pathways is not able to function properly. Second, the two pathways could serve mainly to "locally" supply certain organelles with PtdEtn for maintaining specific molecular species profiles within these organelles. Finally, it is possible that the two pathways yield different molecular species profiles as was shown for the synthesis of PtdCho (CDP-Cho pathway versus PtdEtn methylation) in rat hepatocytes (21). The large molecular diversity of PtdEtn and other phospholipid subclasses is dictated by the combination of different lengths, number of unsaturations, and types of linkages of the hydrocarbon chains. An ester linkage at the sn-1 position defines a diacyl molecular subspecies, whereas an ether linkage at this position defines a plasmanyl subspecies, and a plasmalogen subspecies is defined by a vinyl ether bond at the sn-1 position, as in all three subspecies the hydrocarbon chain at the sn-2 position is known to be linked to the glycerol backbone via an ester bond (22).
The development and refinement of mass spectrometry in combination with the availability of deuterated pathwayspecific precursors has opened the possibility of specifically displaying the PtdEtn species synthesized via the CDP-Etn or PtdSer decarboxylation pathways. We report here that McA-RH7777 cells, when cultured at equimolar concentrations of ethanolamine and serine, prefer the CDP-Etn pathway over PtdSer decarboxylation in a ratio of ϳ2:1, with the decarboxylation route having a preference for the synthesis of long chain, polyunsaturated species.
Incorporation of Deuterium-labeled Precursors into PtdEtn, PtdCho, and PtdSer-Cells were grown in "full" Dulbecco's modified Eagle's medium to 60 -80% confluency in 175-cm 2 culture flasks. For deuterium-label studies, cells were washed twice with phosphate-buffered saline and incubated in serineand choline-free Dulbecco's modified Eagle's medium (supplemented with serum as mentioned earlier) for 6, 24, or 72 h in the presence of 400 M d 9 -choline (d 9 -Cho) and either 400 M d 4 -ethanolamine (d 4 -Etn) or d 3 -serine (d 3 -Ser). To maintain similar substrate concentrations during incubation, 400 M unlabeled serine was supplemented to the culture medium of the d 4 -ethanolamine incubations and vice versa. Incubations were stopped by washing the cells three times with ice-cold phosphate-buffered saline, and cells were scraped into methanol. Total lipids were extracted, and phospholipid classes were isolated by normal-phase high performance liquid chromatography (HPLC) and analyzed as described in one of the sections below.
Subcellular Fractionation of McA-RH7777 Cells-Mitochondrial and microsomal fractions were isolated from labeled McA-RH7777 cells essentially as described for rat liver by Shiao et al. (12). Briefly, cells were scraped into phosphate-buffered saline, pelleted, and homogenized in ice-cold isolation medium (250 mM mannitol, 5 mM HEPES (pH 7.4), 0.5 mM EGTA, and 0.1% bovine serum albumin). After removal of nuclei and cell debris, the supernatant was centrifuged at 10,000 ϫ g for 10 min to pellet crude mitochondria. The resulting supernatant was centrifuged at 100,000 ϫ g for 1 h to pellet microsomes. The mitochondrial pellet was further purified from MAM by hand homogenization in isolation medium and layering of the homogenate on top of Percoll medium (225 mM mannitol, 25 mM HEPES (pH 7.4), 1 mM EGTA, and 0.1% bovine serum albumin and 30% (v/v) Percoll) followed by centrifugation for 30 min at 95,000 ϫ g. Purity of mitochondria and characterization of microsomes was assessed by assaying for succinate dehydrogenase (mitochondrial marker enzyme) and aryl esterase (ER marker) and by Western blotting for cytochrome c (mitochondrial marker) and calnexin (ER marker).
Analysis of Phosphatidylethanolamine Molecular Species-Total lipids were extracted from cells and mitochondrial and microsomal fractions according to the method of Bligh and Dyer (23). The obtained total lipid extract was dissolved in hexane/isopropanol/acetone (82:17:1, v/v/v). Lipid classes were separated on a normal-phase HPLC column as described (24), and the PtdEtn fraction was collected manually from the column effluent using a flow splitter, dried under nitrogen, and stored at Ϫ20°C till analysis. The fraction was dissolved in chloroform/methanol (1:1, v/v), and PtdEtn species were separated on two Synergi 4 m MAX-RP 18A columns (250 ϫ 3 mm) (Phenomenex, CA) in series as described (25) with a slightly modified mobile phase of acetonitrile/methanol (2:3, v/v). Identification of molecular species was performed by online tandem mass spectrometry in the negative-ion mode on an API 4000 Q Trap mass spectrometer fitted with an Atmospheric Pressure Chemical Ionization source (Sciex, Ontario, Canada). Analysis of PtdEtn molecular species compositions and deuterium labeling was performed by on-line single quadrupole mass spectrometry in the negative-ion mode on an API 3000 triple stage quadrupole mass spectrometer fitted with an Atmospheric Pressure Chemical Ionization source (Sciex). Nitrogen was used as nebulizer gas and curtain gas. PtdEtn molecular species compositions were determined by extracting the (labeled) molecular ions and isotope peaks of the various species from the negative Q1 chromatogram representing total PtdEtn, which was verified by detection with a Varex MKIII evaporative light scattering detector (Alltech, Deerfield, IL) operated at 100°C at a gas flow of 1.8 liters/min. Determination of the position of the ester linkage of fatty acids to glycerophosphoethanolamine was performed according to Brouwers et al. (25). Evaporative light scattering detector data were analyzed using EZChrom software (Scientific Software, San Ramon, Canada), and mass spectrometric data were analyzed using Analyst1.4 software (Sciex).
Analysis of Phosphatidylcholine Molecular Species-The PtdCho fraction of cells and mitochondrial and microsomal fractions was isolated from total lipid extracts by normalphase HPLC as described in the previous section. The fraction was dissolved in chloroform/methanol (1:1, v/v), and PtdCho molecular species were separated on two LiChrospher 100 RP18-e columns (5 m, 250 ϫ 4.6 mm; Merck) in series as previously (24), with a slightly modified mobile phase of acetonitrile/methanol/triethylamine (25:24:1, v/v/v). Identification of molecular species was performed by on-line tandem mass spectrometry in the positive-ion mode on an API 4000 Q Trap mass spectrometer fitted with an electrospray ionization source (Sciex). Deuterium labeling of the various molecular species was determined by on-line mass spectrometry in the positive-ion mode on an API 2000 Q Trap mass spectrometer operated in enhanced (trapping) mode fitted with an electrospray ionization source (Sciex). PtdCho molecular species compositions were determined by evaporative light scattering detector detection as described above. Determination of the position of the ester linkage of fatty acids to glycerophosphocholine was performed according to Brouwers et al. (25). Evaporative light scattering detector and mass spectrometric data were analyzed using software as described above.
Analysis of Phosphatidylserine Molecular Species-The Ptd-Ser fraction of cells and mitochondrial and microsomal fractions was isolated from total lipid extracts by normal-phase HPLC as described above. The fraction was dissolved in chloroform/methanol (1:1, v/v), and PtdSer molecular species were separated on a Synergi 4 m MAX-RP 18A column (250 ϫ 3 mm) (Phenomenex, CA) with a mobile phase of acetonitrile/methanol/H 2 O (15:22.5:12.5, v/v/v) containing 1 M serine and 2.5 mM ammonium acetate. Analysis and identification of molecular species was performed by on-line (tandem) mass spectrometry in the negative enhanced mass spectrometry ion mode on an API 4000 Q Trap mass spectrometer fitted with an electrospray ionization source. Ptd-Ser species compositions were determined in PtdSer fractions obtained from unlabeled cells by extracting the molecular ions and isotope peaks of various species from the negative neutral loss 87-chromatogram representing total PtdSer.
Interpretation of Mass Spectra-The incorporation of d 4 -Etn into PtdEtn and d 9 -Cho into PtdCho was calculated for all molecular species by determining the intensities of the unlabeled and labeled molecular ion peak in the negative or positive Q1 mass spectrum and expressing the intensity of the labeled molecular ion as the percentage of the sum of the unlabeled and labeled molecular ion. d 3 -Ser incorporation into PtdEtn and PtdSer was calculated similarly from (enhanced) negative Q1 spectra. Because d 3 -Ser labeling also, unexpectedly, yielded significant d 2 -Ser labeling (see "Discussion"), this d 2 -Ser incorporation had to be taken into account for determination of the total d 3 -Ser incorporation into the various PtdEtn and PtdSer species. All PtdCho, PtdEtn, and PtdSer molecular ions display isotope peaks in Q1 mass spectra ( Fig. 1, C and D), originating from the natural presence of 13 C in these large biomolecules. To determine total d 3 -Ser incorporation into PtdEtn and PtdSer molecular species, the natural contribution of 13 C to the intensities of the isotope peaks of each molecular ion was subtracted from the actual, measured intensities of the isotope peaks in the Q1 mass spectra obtained, the remaining "isotope peak" intensities (exemplified for PtdEtn in Fig. 1D) representing total d 3 -Ser incorporation into the various PtdEtn and PtdSer species.
Measuring the Intracellular Deuterated to "Cold" Serine and Ethanolamine Ratio-Cells were labeled with 400 M d 3 -Ser or d 4 -Etn for various times up to 6 h and washed twice with icecold phosphate-buffered saline before extracting the water-soluble components (23). The water-methanol phase was collected and evaporated to dryness, and the primary amines Ser and Etn were subsequently derivatized with fluorescamine exactly as described (26). The deuterated to cold ratio was determined using mass spectrometry.
Determination of the Substrate Specificity of Phosphatidylserine Decarboxylase-The substrate (d 3 -labeled egg-PtdSer and various PtdSer molecular species) required to determine the substrate specificity of PSD by using a mass spectrometry approach was synthesized from their respective PtdCho species. Briefly, 2.5 mg of PtdCho was dissolved in 1 ml of chloroform, after which 25 mg of silica (kieselgel 60 for column chromatography) was added. The mixture was stirred for 30 min and carefully dried under a gentle stream of nitrogen. Subsequently, 250 l of 100 mM acetate buffer (pH 5.6) containing 100 mM CaCl 2 , 50 mg/ml d 3 -serine, and 10 units of phospholipase D (Streptomycin species) was added to the silica, and the suspension was incubated for 36 -48 h at 30°C while shaking continuously. The reaction was stopped by adding 470 l of H 2 O and 80 l of 6 M HCl, and phospholipids were extracted (23,27). The amount of d 3 -PtdEtn formed was quantified using the phosphorus assay (28).
PSD activity was measured in mitochondria, prepared as described in the subcellular fractionation section, as the formation of d 3 -PtdEtn from d 3 -PtdSer (29). The reaction mixture (final volume 0.4 ml) consisted of 100 mM KH 2 PO 4 (pH 6.8), 10 mM EDTA, 0.5 mg of Triton X-100/ml of assay mixture, 50 M PtdSer (d 3  After removing the Triton X-100 using small silica (kieselgel 60) columns, the amount of d 3 -PtdSer (substrate) and d 3 -PtdEtn (product) was quantified as described above.

Pathway-specific Monitoring of Phosphatidylethanolamine Biosynthesis in
McA-RH7777 Cells-To get insight into the qualitative and quantitative contributions of the CDP-Etn pathway and PtdSer decarboxylation to overall PtdEtn synthesis in mammalian cells, McA-RH7777 (McArdle) cells were incubated in the presence of deuterated, pathway-specific precursors. After various times, total lipids were extracted and subfractionated into phospholipid subclasses, and the PtdEtn fraction was analyzed by HPLC-mass spectrometry (MS). A typical chromatogram of the HPLC separation of McArdle PtdEtn molecular species is shown in Fig. 1A, with the peak identification given in Table 1. The major advantage of on-line HPLC separation before MS is that isobaric molecular species, i.e. species having the same mass but different radyl groups in their diacylglycerol backbones, are largely separated, thus allowing individual analysis.
Labeling of cells with d 4 -Etn or d 3 -Ser in combination with mass spectrometric phospholipid analysis allowed us to distinguish between PtdEtn species synthesized via the CDP-Etn pathway and species formed by Ptd-Ser decarboxylation. d 4 -Etn-labeled PtdEtn species de novo synthesized via the CDP-Etn pathway could be easily discriminated from unlabeled PtdEtn species because of a 4-Da mass difference (Fig. 1C) between the molecular ions in the (negative) Q1 spectrum (Fig. 1B). In d 3 -Ser, three carbon-bound protons of serine are replaced by deuterium atoms. Once the d 3 -Ser label is incorporated into the head group of PtdSer species, the 3-Da mass difference between labeled PtdSer species and their non-labeled counterparts is retained upon decarboxylation to PtdEtn ( Fig. 1D; the unexpected appearance of d 2 -PtdSer will be explained under "Discussion" and is corrected for, see "Experimental Procedures"). In addition to PtdSer decarboxylation, (labeled) serine can be incorporated into PtdEtn via two other routes; (i) it can enter the CDP-Etn pathway as (labeled) phosphoethanolamine, generated as an intermediate of sphingomyelin metabolism (30), and (ii) it can be incorporated into the diacylglycerol moiety (31). However, experiments with ␤-chloro-L-alanine, a potent inhibitor of sphingomyelin synthesis (32), revealed that phosphoethanolamine liberated from sphingomyelin breakdown only marginally (Ͻ2%) contributed to serine labeling of PtdEtn in our experimental system (data not shown). Furthermore, only a slight amount of deuterium label was detected in the diacylglycerol moiety of PtdEtn and PtdCho in McArdle cells labeled with d 3 -Ser for 24 h, which was clear from the fact that less than 4% of the PtdCho molecules were detected with a mass up to 3 mass units heavier than the parental molecular species. Therefore, it was concluded that in McArdle cells deuterated Ser-labeled PtdEtn species were derived from PtdSer decarboxylation (Ͼ95%).
Qualitative and Quantitative Contributions of the CDP-ethanolamine Pathway and PtdSer Decarboxylation to Overall Phosphatidylethanolamine Synthesis-We first determined a suitable labeling time to study the contribution of the various pathways to phospholipid biosynthesis by labeling McArdle cells with 400 M d 9 -Cho and 400 M d 4 -Etn for 6, 24, and 72 h. Because McArdle cells have only one route for de novo PtdCho synthesis, namely the CDP-Cho pathway, it was expected that all PtdCho species would be labeled to the same extent when remodeling was completed. Fig. 2A shows that already after 6 h PtdEtn versus 5% in PtdCho) and plasmalogen (ϳ10% of total PtdEtn) species than PtdCho (Ͻ5%). In line with the data presented in Fig. 2B, the various PtdEtn species were labeled with d 3 -Ser to a different level (Figs. 3 and 4). PtdEtn species that were only modestly labeled with d 3 -Ser after 24 h (e.g. (16:0-18:1)PtdEtn, d 3 -Ser labeled for 14%) displayed a high d 4 -Etn incorporation (66%) and vice versa (e.g. (18:0-20:5)PtdEtn was d 4 -Etn-labeled for 21% and labeled with d 3 -Ser for 36%). To obtain the quantities (Fig. 3) by which various phospholipid species are synthesized, label incorporation data were coupled to the molecular species compositions (Tables 1-3).

TABLE 1 Phosphatidylethanolamine molecular species composition in McA-RH7777 cells, mitochondria and microsomes
The PtdEtn fraction was collected from total lipid extracts of McA-RH7777 cells, mitochondria and microsomes, and the molecular species compositions were determined as described under "Experimental Procedures." Shown in this table is the PtdEtn molecular species composition (first row) of the total cell homogenate (third row), the mitochondrial fraction (fourth row), and the microsomal fraction (fifth row). The second row indicates the chromatogram peak in which the various molecular species elute in the HPLC-MS chromatogram displayed in Fig. 1A. Data are expressed as the mean Ϯ S.D. of three experiments, performed in triplicate.

PtdEtn species
Peak (Fig.1A Fig. 2A) that because all PtdCho species were labeled to a similar extent, the picture illustrating the amount by which the various Ptd-Cho species were synthesized de novo after 24 h of incubation (Fig. 3) reflected the PtdCho molecular species composition as presented in Table 2 To determine whether the observed preference of the PtdSer dearboxylation route to synthesize PtdEtn species with polyunsaturated fatty acids at the sn-2 position is a more common phenomenon and not restricted to McArdle cells, a non-rat liver-derived cell line was studied. CHO cells were labeled for 24 h with d 4 -Etn or d 3 -Ser, and the amounts by which the various PtdEtn species were de novo-synthesized via the CDP-Etn pathway and PtdSer decarboxylation are shown in Fig. 5. In CHO cells the CDP-Etn pathway is favored over PtdSer decarboxylation in a ratio of 3:1. The higher contribution of the Kennedy pathway to PtdEtn synthesis in the CHO cells in comparison with McArdle cells is most likely due to the elevated amount of PtdEtn plasmalogens in CHO cells (CHO (ϳ40%) versus McArdle (ϳ10%)), as they are predominantly synthesized by the CDP-Etn pathway (Fig. 5). The only PtdEtn species synthesized for more than 50% by the  Table 3. No significant difference in PtdSer species distribution between the whole cell, mitochondria, and microsomes was observed. The PtdSer "parent molecules" of (18:0-20:

TABLE 2 Phosphatidylcholine molecular species composition in McA-RH7777 cells, mitochondria, and microsomes
The PtdCho fraction was collected from total lipid extracts of McA-RH7777 cells, mitochondria, and microsomes, and the molecular species compositions were determined as described under "Experimental Procedures." Shown in this table is the PtdCho molecular species composition (first row) of the total cell homogenate (second row), the mitochondrial fraction (third row), and the microsomal fraction (fourth row). Data are expressed as the mean Ϯ S.D. of three experiments, performed in triplicate.

PtdCho species
Whole cell Mitochondria Microsomes  (50%) (Fig. 3), which is lower than the amount of newly synthesized PtdCho (see Fig. 2A) due to the lower specific activity of deuterated Ser. The similar extent of labeling of all PtdSer species implicates that the relative amounts of the various PtdSer species, synthesized in 24 h, reflected the PtdSer molecular species composition shown in Table 3.

Contribution of CDP-Etn Pathway and PtdSer Decarboxylation to Newly Synthesized PtdEtn in
McA-RH7777 Mitochondria and Microsomes-PtdEtn synthesized via the CDP-Etn pathway originates from the ER (1,10,33,34), whereas PtdSer decarboxylation occurs in the mitochondria (1, 2, 11). Shiao et al. (12) concluded from radioactive labeling studies that in Chinese hamster ovary cells the majority of mitochondrial PtdEtn was derived from PtdSer decarboxylation and that essentially no PtdEtn was imported from the ER. To get a picture of "local" PtdEtn and PtdCho synthesis in our cell system, a similar experiment as described in the previous section was performed using McArdle cells labeled with the three pathway-specific deuterated precursors (d 4 -Etn, d 9 -Cho, and d 3 -Ser) for 24 h.
No significant difference was detected in the PtdCho species profile between the whole cell and the mitochondrial and microsomal fractions (Table 2), nor were any significant differences found in subcellular d 9 -Cho incorporation. As already described for the whole cell, PtdCho species in the mitochondria and microsomes were labeled to the same degree (70%) (data not shown). PtdCho is synthesized in the Golgi via CPT and in the ER via choline/ethanolaminephos- When the cells were about 75% confluent, cells were labeled with d 3 -Ser or d 4 -Etn for 24 h. After labeling, the cells (C) were collected for lipid extraction or subcellular fractions (mitochondria (M) and microsomes (E)) were prepared before lipid extraction. Subsequently lipids were separated into the major phospholipid classes by HPLC followed by molecular species profiling of the various phospholipids classes by HPLC-MS as described under "Experimental Procedures." The black bars represent the contribution of PtdSer decarboxylation, and the gray bars the contribution of the CDP-Etn pathway to newly synthesized PtdEtn. Data are expressed as the mean Ϯ S.D. of three experiments, performed in triplicate.

TABLE 3 Composition of some phosphatidylserine molecular species in McA-RH7777 cells, mitochondria, and microsomes
The PtdSer fraction was collected from total lipid extracts of McA-RH7777 cells, mitochondria, and microsomes, and the molecular species compositions were determined as described under "Experimental Procedures." Shown in this table is the composition of six selected PtdSer molecular species (first row) that were abundant and/or interesting with respect to the research question in the total cell homogenate (second row), the mitochondrial fraction (third row), and the microsomal fraction (fourth row). Data are expressed as the mean Ϯ S.D. of three experiments, performed in triplicate. photransferase (33); therefore, these data suggest a tendency of newly synthesized PtdCho to equilibrate relative quickly between its place of synthesis, mitochondria, and presumably other cellular membranes.
In contrast to what was observed for PtdCho, the PtdEtn species distribution over the mitochondria and microsomes was very distinct (Table 1). Whereas some species were more or less equally distributed over both fractions, (16:0-18:1)PtdEtn, The preferential synthesis of (18:0-20:4)PtdEtn and (18:0-20: 5)PtdEtn via PtdSer decarboxylation, as observed in whole cells (Fig. 4, cf. Fig. 2B), was even more pronounced in mitochondria (Fig. 4); ϳ80% of the newly synthesized molecules of these PtdEtn species were derived from PtdSer decarboxylation (black bars). In addition, newly synthesized PtdEtn of several other species was found to be preferentially derived from PtdSer decarboxylation in the mitochondria, including For all other newly synthesized PtdEtn species (except the ethanolamine plasmalogens) in the mitochondria, the balance was shifted toward PtdSer decarboxylation as well but only to a modest extent; the vast majority of de novo synthesis for these species had taken place via the CDP-Etn pathway. In mitochondria the ratio CDP-Etn versus PtdSer decarboxylation-derived PtdEtn was shifted toward the decarboxylation route in comparison to the whole cell. The balance Kennedy pathway/PtdSer decarboxylation observed in the microsomal fraction ( Fig. 4)

DISCUSSION
The CDP-Etn pathway and PtdSer decarboxylation are present in all mammalian cells. Here, we described the detailed analysis of the qualitative and quantitative contributions of the various pathways responsible for PtdEtn, PtdSer, and PtdCho biosynthesis in McA-RH7777 cells by pathway-specific deuterium labeling coupled to HPLC-MS analysis. At the 24-h labeling time point, the amount of newly synthesized PtdSer is somewhat lower than that of PtdCho and PtdEtn via their respective CDP pathways (Figs. 2, A and B). This can be explained as, in contrast to Cho and Etn, McArdle cells synthesize Ser de novo. As a result, the intracellular pool-size of Ser, which is composed of d 3 -Ser, d 2 -Ser, and Ser is only for 80 Ϯ 5% deuterated (ratio d 3 -Ser to d 2 -Ser is 1:1), whereas in the case of Etn the pool is for 94 Ϯ 2 deuterated. Therefore, it can be concluded that in the Figs  shows the quantitative contribution of the CDP-Etn route to the synthesis of the various PtdEtn species independent of the d 3 -Ser labeling experiments. We expected the observation that not all PtdEtn species were labeled to the same extent at 24 and 72 h to be due to the fact that not all PtdEtn species were made via the CDP-Etn pathway in the same quantity and not due to incomplete remodeling of some species. This was supported by the observations that (i) the two plasmalogen PtdEtn species show a similar profile as PtdCho (see Figs. 2, A and B), which is in agreement with the fact that plasmalogens are exclusively synthesized via the CDP-Etn pathway and (ii) the PtdEtn species with a relative low labeling percentage from the CDP-Etn pathway were rapidly synthesized via the decarboxylation route, as was confirmed by the d 3 -Ser incorporation studies (Figs. 3 and 4). dria and subjected to remodeling before decarboxylation; (ii) all Ptd-Ser species are transported from ER/MAM to mitochondria with equal efficiency, and selectivity is introduced by substrate specificity of the PtdSer decarboxylase (PSD); (iii) PSD is indiscriminative toward imported PtdSer species, and selectivity is introduced by remodeling of PtdEtn synthesized by PSD in the mitochondria.
To assess the possibility that PtdSer species were preferentially imported into the mitochondria (i), the PtdSer species composition and d 3 -Ser incorporation into this phospholipid class was determined in cells, mitochondria, and microsomes. There was no difference in PtdSer species composition between the whole cell, mitochondria, and microsomes, and more importantly, all PtdSer species were labeled with d 3 -Ser with equal efficiency in all three fractions. Based on these results, it seems that PtdSer, once synthesized in the ER/MAM, readily equilibrates over the cell. Therefore, we consider it unlikely that the species selectivity observed in the PSD pathway originates from selectivity in PtdSer import into mitochondria or specific remodeling of PtdSer in the mitochondria.
The second possibility (ii), i.e. that the species selectivity originates from substrate specificity of the PSD, seems a likely explanation as an in vitro PSD assay, using isolated mitochondria as source for the decarboxylase, showed a clear substrate preference for PtdSer species with a 20:4 fatty acyl chain at the sn-2 position (see Fig. 6). So the substrate specificity of the decarboxylase explains at least in part the preferential synthesis of PtdEtn species with long-chain polyunsaturated fatty acids at the sn-2 position. This observation is in contrast to earlier reports, where no clear preference of PSD for polyunsaturated fatty acids was detected in in vitro studies (35,36). We have no ready explanation for this discrepancy except that different tissue, cell lines, and methodologies were used.
Alternatively, selectivity is introduced by rapid remodeling of PtdEtn generated by decarboxylation of PtdSer (iii). Although it has been shown that the de novo-synthesized phospholipids are extensively remodeled by deacylation-reacylation (37)(38)(39), studies on phospholipid remodeling have yielded conflicting results. Because the enzymes involved in remodeling are reported to reside in several subcellular compartments (40 -43), including microsomes and mitochondria, remodeling of PtdEtn "freshly" synthesized by PtdSer decarboxylation in the mitochondria is a possible explanation for the preferential syn- thesis of (18:0-20:4)PtdEtn and (18:0-20:5)PtdEtn via this route that cannot be excluded by our data.
The Molecular Species Profiles Synthesized by the Kennedy Pathways Are Similar-The final reaction of the Kennedy pathways for PtdEtn and PtdCho synthesis is catalyzed by a phosphotransferase. Mammalian cells express a phosphotransferase with a dual specificity for CDP-Etn and CDP-Cho. Because the phosphotransferase reaction supplies the diacylglycerol part of the PtdEtn and PtdCho molecules, one would expect, when remodeling is of minor contribution, the species profiles of Ptd-Cho and PtdEtn made by the respective Kennedy pathways to be similar. As shown in Fig. 3, this seems to be largely the case, as the CDP-Etn/Cho pathways synthesize predominantly saturated and mono-and di-unsaturated phospholipid species. Thus, the large variety of PtdEtn species (versus PtdCho) (Tables 1 and 2) is mainly caused by the presence of an additional pathway, the PtdSer decarboxylation route.
Involvement of a Serine Racemase Explains the Synthesis of d 2 -PtdSer from d 3 -serine-In mammalian cells PtdSer is synthesized by a calcium-dependent base-exchange reaction in which the head group of pre-existing PtdCho and PtdEtn is replaced with L-serine, catalyzed by PtdSer synthase 1 and 2, respectively (44). The molecular species profile of PtdSer agrees with that of its substrate, microsomal PtdCho and PtdEtn, except for the enrichment in (18:0-18:1)PtdSer (Tables 1-3). Whether this is caused by a substrate preference of the PtdSer synthases or by the existence of specific PtdSer pools between endoplasmic reticulum and MAM, cannot be concluded from our data. The analysis of d 3 -Ser incorporation into PtdSer and PtdEtn showed an unexpected finding; in addition to endogenous, unlabeled phospholipid species and d 3 -Ser labeled species, significant d 2 -Ser incorporation in all PtdEtn and PtdSer species was detected (Fig. 1D). This phenomenon could not be correlated to impurity of the d 3 -Ser (L-[2,3,3-d 3 ]-serine) label used nor to a-specific, non-enzymatic exchange of deuterium atoms from d 3 -Ser with protons from water molecules during incubation (not shown). One possible explanation is that during the exchange of the head group from PtdCho or PtdEtn with serine, a deuterium atom is lost. This seems unlikely, as the hydroxyl group participating in the PtdSer synthase-catalyzed reaction is not deuterated in the precursor (L-[2,3,3-d 3 ]-serine). More likely may be the involvement of a serine racemase activity. This enzyme, which is present in various tissues like brain, heart, kidney, and liver (45), catalyzes the conversion of L-serine to D-serine and vice versa. The reaction mechanism of amino acid racemases, such as alanine and serine racemase (46,47), shows that the hydrogen attached to the C2 (␣-atom) of the amino acid gets detached from the molecule, resulting in the conversion (2,3,3-d 3 -labeled) L-serine to (3,3-d 2 -labeled) D-serine. The presence of a substantial amount of d 2 -Ser-labeled PtdSer and PtdEtn suggests that one or both PtdSer synthases may possibly use D-serine as substrate provided that a racemase is indeed the cause of the d 2 -Ser formation. However, Kuksis and Itabashi recently showed that in rat brain, a tissue with a relatively high serine racemase activity, the serine in PtdSer was exclusively in the L-configuration (48). This observation suggests that (3,3-d 2 -labeled) D-serine is most likely converted back to (3,3-d 2 -labeled) L-serine despite a 10ϫ higher k m for the D-to L-serine than for the L-to D-serine conversion before it is used by the PtdSer synthase.
Local Phosphatidylethanolamine Synthesis Suggests an Essential Role of the PtdSer Decarboxylation Route in Mitochondria-The main PtdEtn species newly synthesized from the Kennedy pathway in the ER/microsomes appeared to rapidly equilibrate between ER and mitochondria and seemed to be actively imported into mitochondria. The rapid equilibration of PtdEtn de novo synthesized by the CDP-Etn pathway but also of the newly formed PtdSer and PtdCho observed in our study is in agreement with previous reports (13). Because the Kennedy pathway was found to be responsible for the bulk of the PtdEtn synthesis, rapid equilibration over the whole cell is probably required to keep up with the cellular demand for this phospholipid. In contrast, (18:0-20:4)PtdEtn and (18:0-20:5)Ptd Etn, mainly formed by PtdSer decarboxylation in the mitochondria, appeared to be actively retained within this organelle, as these two species are notably enriched in the mitochondria but are hardly present in the microsomal fraction. Although the biochemical significance of this observation is not known yet, Steenbergen et al. (49) recently showed that embryos of mice lacking PSD activity die in the uterus and have fragmented and misshapen mitochondria, an abnormality that likely contributes to the embryonic lethality of the PSD knock-out.