Phosphatidylserine Synthase-1 and -2 Are Localized to Mitochondria-associated Membranes*

We report the subcellular localization of enzymes involved in phosphatidylserine biosynthesis in mammalian cells. Several lines of evidence suggest that phosphatidylserine synthase-1 (PSS1) is highly enriched in mitochondria-associated membranes (MAM) and is largely excluded from the bulk of the endoplasmic reticulum (ER). Taking advantage of the substrate specificity of PSS1, we showed that (i) MAM contain choline exchange activity, whereas this activity is very low in the bulk of the ER, (ii) serine exchange activity is inhibited by choline to a much greater extent in MAM than in ER, and (iii) MAM use phosphatidylcholine and phosphatidylethanolamine as substrates for phosphatidylserine biosynthesis, whereas the ER utilizes only phosphatidylethanolamine. According to immunoblotting of proteins from both CHO-K1 cells and murine liver, PSS1 is localized to MAM, and in hepatoma cells stably expressing PSS1 this protein is highly enriched in MAM. Since the ER contains serine and ethanolamine exchange activities, we had predicted that PSS2 would account for the serine exchange activity in the ER. Unexpectedly, using immunoblotting experiments, we found that (i) PSS2 of CHO-K1 cells is present only in MAM and (ii) PSS2 is restricted to MAM of McArdle cells expressing recombinant PSS2. These data leave open the question of which enzyme imparts PSS activity to the ER and suggest that a third isoform of PSS might be located in the ER.

Many enzymes involved in the final stages of lipid biosynthesis are integral membrane proteins of the endoplasmic reticulum (ER). 1 Examples include cholinephosphotransferase (1,2), ethanolaminephosphotransferase (1,2), phosphatidylethanolamine (PtdEtn) N-methyltransferase (1,2), diacylglycerol acyltransferase (2), glycerol-3-phosphate acyltransferase, acyl-CoA:cholesterol acyltransferase (2), 3-hydroxy-3methylglutaryl-CoA reductase (2), and phosphatidylserine (PtdSer) synthase (1,2). Upon closer examination, however, some of these enzyme activities are found to be present in additional cellular locations such as peroxisomes, mitochondria, or mitochondria-associated membranes (MAM) (3). MAM are ER-like membranes that co-isolate with mitochondria but can be separated from the latter by centrifugation of crude mitochondria on a Percoll gradient (1). MAM have many, but not all, properties of the ER (1,2). In particular, MAM are enriched, compared with the bulk of ER, in several lipid biosynthetic enzyme activities, such as acyl-CoA:cholesterol acyltransferase (2), diacylglycerol acyltransferase (2), some enzymes involved in biosynthesis of glycosylphosphatidylinositol anchors of proteins (4), and phosphatidylserine synthase (PSS). The specific activity of PSS is typically 2-4-fold higher in MAM than in ER (1,2). A unique protein marker for MAM is PtdEtn N-methyltransferase-2 (5), an isoform of the enzyme that converts PtdEtn to phosphatidylcholine (PtdCho) in liver (6). Although ER and MAM contain similar specific activities for PtdEtn methylation, an antibody generated against PtdEtn N-methyltransferase-2 recognizes a protein in MAM that is not detectable in the ER (5). These data indicate that two isoforms of PtdEtn N-methyltransferase exist; one is specifically located in MAM and excluded from the ER, whereas the other is presumably responsible for the phospholipid methylation activity in the ER. The function of MAM has not been firmly established, but an attractive hypothesis is that MAM are involved in the import of PtdSer into mitochondria via a membrane collision-based mechanism (1,7). Indeed, essentially all mitochondrial PtdEtn is produced from decarboxylation of imported PtdSer (7). Moreover, MAM appear to be associated with contact sites between mitochondrial inner and outer membranes (8,9).
Two PtdSer synthases with distinct substrate specificities have been identified in mammalian cells. Each catalyzes a base exchange reaction between serine and preexisting phospholipids. The existence of two isoforms of PSS was confirmed by isolation and sequencing of two distinct PSS cDNAs encoding PSS1 and PSS2 that share limited (32%) sequence identity (10). PSS1 and PSS2 both catalyze serine exchange activity, whereas only PSS1, and not PSS2, catalyzes choline exchange (11). In CHO-K1 cells expressing either only PSS1 or only PSS2, PSS1 uses almost exclusively PtdCho as the donor of the phosphatidyl group, whereas PSS2 uses only PtdEtn (12). The reason why PSS1 can use both PtdEtn and PtdCho in in vitro enzymatic assays but only PtdCho in intact cells is not clear.
We now report the subcellular localization of the two PSS isoforms, PSS1 and PSS2. Previous immunoblotting studies of CHO cells by Saito et al. (13) detected PSS1 in microsomes as well as in MAM. However, since microsomes contain MAM, as well as ER and other organelle membranes, these studies did not exclude the possibility that the microsomal PSS1 was contributed by MAM alone and was absent from the ER per se. On the basis of our previous findings that both MAM and the ER contain PSS activity, we hypothesized that the two PSS iso-forms might be present in spatially distinct locations in the cell, with one isoform being located in MAM and the other in the ER. We demonstrate that PSS1 protein is found exclusively in MAM and is not detectable in the ER. In addition, and contrary to our predictions, we found that PSS2 is also restricted to MAM. These findings leave open the question of which PtdSer biosynthetic enzyme is responsible for serine exchange activity in the ER and suggest that this activity might be due to a third, presently unknown, PtdSer synthase isoform.  (14).

Materials
Antibodies-Anti-PSS1 and anti-PSS2 antibodies were a generous gift from Dr. M. Nishijima (National Institute of Health, Japan). These antibodies were raised in rabbits against synthetic peptides corresponding to amino acids 1-17 of PSS1 from CHO-K1 cells and amino acids 458 -474 of PSS2 from CHO-K1 cells. The rabbit anti-rat proteindisulfide isomerase polyclonal antibody was a generous gift from Dr. M. Michalak (University of Alberta) (15). Anti-myc antibody was obtained from clone 9E10 hybridoma tissue culture supernatant and used directly for immunoblotting experiments. The rabbit anti-rat PtdEtn N-methyltransferase-2 polyclonal antibody, generated against a C-terminal dodecapeptide of PtdEtn N-methyltransferase-2, was provided by Dr. D. E. Vance (University of Alberta) (5). The rabbit anti-rat calnexin polyclonal antibody was from Stressgen Biotechnologies Corp. (Victoria, Canada).
Preparation of myc Epitope-tagged PSS1 and PSS2 cDNAs-The myc epitope (EQKLISEEDL) was appended to the 5Ј-end of murine PSS1 cDNA (14) and either the 5Ј-or 3Ј-end of murine PSS2 cDNA (16) by the polymerase chain reaction. A myc-PSS1 mutant was generated by deleting the last two codons (encoding two lysine residues) of murine PSS1 cDNA. The cDNA fragments were sequenced by the DNA core facility at the University of Alberta to confirm the modifications. Each of the cDNAs encoding myc-PSS1, myc-PSS1 with two C-terminal lysine residues deleted (myc-PSS1:⌬cKK), and myc-PSS2 with the myc tag appended to either the N terminus (N-myc-PSS2) or the C terminus (C-myc-PSS2) was inserted into the eukaryotic expression vector pcDNA 3.1 (Invitrogen). McArdle rat hepatoma cell lines were transfected with 10 g of the cDNAs using the calcium phosphate precipitation method (17). Stable transfectants were selected by culturing the cells in medium containing 600 g/ml G418. Individual colonies were isolated. When cell lines had been established, the concentration of G418 was reduced to 200 g/ml. Control cells for experiments with McArdle cells were transfected with the expression vector lacking a cDNA insert.
Subcellular Fractionation of Cultured Cells-Microsomes, MAM, and mitochondria were isolated from cultured cells by slight modifications of a method previously described (7). Cells were scraped into phosphate-buffered saline, pelleted by centrifugation at 500 ϫ g for 5 min, and resuspended in 8 ml of homogenization buffer (0.25 M sucrose and 10 mM Hepes (pH 7.4)). The cells were gently disrupted by 15 up-and-down strokes in a Potter-Elvehjem motor-driven homogenizer. The homogenate was centrifuged twice at 600 ϫ g for 5 min to remove cellular debris and nuclei, and the supernatant was centrifuged at 10,300 ϫ g for 10 min to pellet crude mitochondria. The resultant supernatant was centrifuged at 100,000 ϫ g for 1 h in a Beckman Ti 70.1 rotor at 4°C to pellet microsomes, which were resuspended in homogenization buffer. The mitochondrial pellet was resuspended in 300 l of isolation medium (250 mM mannitol, 5 mM Hepes (pH 7.4), and 0.5 mM EGTA) and layered on top of 8 ml of Percoll medium (225 mM mannitol, 25 mM Hepes (pH 7.4), 1 mM EGTA, and 30% Percoll (v/v)) in a 10-ml polycarbonate ultracentrifuge tube and then centrifuged for 30 min at 95,000 ϫ g. A dense band containing purified mitochondria was recovered from approximately 3 ⁄4 down the tube. The mitochondrial band was removed, diluted with isolation medium, and washed twice by centrifugation at 6300 ϫ g for 10 min to remove the Percoll, after which the mitochondria were resuspended in isolation medium. MAM were removed from the Percoll gradient as the diffuse white band located above the mitochondria. Isolation medium was added, and the suspension was centrifuged at 6300 ϫ g for 10 min. The supernatant containing MAM was centrifuged at 100,000 ϫ g for 1 h in a Beckman Ti 70.1 rotor, and the resulting MAM pellet was resuspended in homogenization buffer.
Subcellular Fractionation of Murine Liver Homogenates-MAM and mitochondria were isolated from murine liver as described previously for rat liver (1). ER fractions enriched in rough ER (ER1) and smooth ER (ER2) were prepared by the method of Croze and Morre (18) as modified by Vance and Vance (19). ER1 was isolated from the final discontinuous sucrose gradient at the interface between sucrose solutions of 1.5 and 2.0 M. ER2 was isolated from the same gradient at the interface between sucrose solutions of 1.3 and 1.5 M.
Measurement of PtdSer Synthase Activity-PtdSer synthase activity was measured in subcellular fractions of mouse liver as described previously (19). Cultured cells were scraped from 100-mm dishes and disrupted by sonication (2 ϫ 10 s) with a probe sonicator in 10 mM HEPES buffer (pH 7.5) containing 0.25 M sucrose. Lysates were centrifuged for 2 min at 600 ϫ g to pellet cellular debris, and PSS activity was measured in the supernatant in the presence of 10 mM calcium using  (19). In some experiments, as indicated, unlabeled choline or ethanolamine was added to the reaction mixture to final concentrations of 0.5, 5, or 50 mM at pH 7.4.
Measurement of PSS Activity Using PtdCho or PtdEtn Dispersions-100 l of 50 mM PtdCho or PtdEtn dissolved in chloroform were evaporated to dryness under a stream of N 2 . The phospholipid residue was resuspended in 0.04% Triton X-100 to a final concentration of 7 mM, incubated on ice for 30 min with frequent vortexing, and then sonicated until the solution cleared. PtdCho or PtdEtn (60 l of solution) was added to each PSS assay to a final concentration of 2 mM.
Trypsin Treatment of Membranes-An aliquot of ER membranes or MAM was incubated with trypsin (ratio of membrane protein to trypsin, 2:1 (w/w)) at 37°C for 15 min. The reaction was terminated by the addition of 20 l of trypsin inhibitor (40 mg/ml). Treated membranes were diluted in isolation medium, and 25 g of protein from each fraction was used for the PSS assay.
Other Methods-Protein concentrations were determined by the BCA method (Pierce) using bovine serum albumin as a standard.

RESULTS AND DISCUSSION
PtdSer Synthase Exchange Activities in ER and MAM-We have previously shown that serine exchange activity is present in the ER and is relatively enriched in MAM (1, 2). As a first approach to determining the subcellular locations of PSS1 and PSS2, we compared the substrate specificities for base exchange in subcellular fractions by assessing the ability of unlabeled choline or ethanolamine to inhibit serine exchange activity. Since PSS1 exchanges all three bases (serine, choline, and ethanolamine) in in vitro assays, unlabeled choline and ethanolamine would be expected to compete with radiolabeled serine in a serine exchange assay (11). In contrast, since PSS2 catalyzes the exchange of serine and ethanolamine, but not choline, one would expect that ethanolamine, but not choline, would compete with radiolabeled serine in the in vitro serine exchange assay. The addition of 5 mM ethanolamine to the serine exchange assay mixture inhibited the serine exchange activity of the ER and MAM by 92 and 85%, respectively (Fig.  1). Although 0.5 and 5 mM choline inhibited the serine exchange activity in MAM by ϳ40 and 70%, respectively, the serine exchange activity in the ER was not inhibited by 0.5 mM choline and inhibited only slightly, if at all, by 5 mM choline (Fig. 1). Additional data presented below demonstrate that MAM also contain PSS2; therefore, one would not expect the serine exchange activity of MAM to be completely inhibited by choline. These observations suggest that choline exchange activity (i.e. PSS1) is primarily concentrated in the MAM, since the serine exchange activity was much more potently inhibited by choline in the MAM than in the ER. When the choline concentration was increased to 50 mM, inhibition of the serine exchange activity in the MAM increased to 83%, whereas the activity in the ER was inhibited by ϳ40%. We found that 50 mM choline modestly inhibits the serine exchange activity in M.9.1.1 cells (which lack PSS1), by ϳ20%. Kuge et al. (11) reported that in PSA3 cells (in which PSS1 is defective and choline exchange is Ͻ1% of that in parental CHO cells) 5 mM choline inhibited serine exchange activity by ϳ20%. The partial inhibition by 50 mM choline of the serine exchange activity in the ER probably reflects a combination of some inhibition of PSS2 serine exchange activity by the very high concentration of choline and the presence of small amounts of PSS1 in the ER.
Sensitivity of PSS1 and PSS2 Activity to Trypsinization of ER and MAM-In order to establish further the biochemical properties, subcellular locations, and topology of PSS1 and PSS2, ER and MAM were incubated with trypsin. Our laboratory has previously demonstrated that trypsin proteolysis of ER membranes results in the loss of only ϳ60% of serine exchange activity, whereas the same treatment inactivates another ER membrane protein, cholinephosphotransferase, by almost 100% (19). Choline, ethanolamine, and serine exchange activities were measured in ER and MAM that had been preincubated with or without trypsin. The data in Fig. 2 demonstrate that MAM contain choline exchange activity, whereas this activity is essentially absent from the ER, supporting the idea that PSS1 is localized to MAM, whereas the ER is almost devoid of PSS1. After proteolysis, the serine and choline exchange activities of MAM and ER were decreased by Ͼ80% (Fig. 2). In MAM, ϳ85% of the ethanolamine exchange activity was inactivated by trypsin treatment. In contrast, the ethanolamine exchange activity of the ER was reduced by only 35%. As a positive control to confirm that the trypsin was proteolytically active, trypsin treatment under the same conditions reduced the PtdEtn N-methyltransferase activity by 96% in the ER and by 84% in MAM. The relative resistance of the ethanolamine exchange activity in the ER to proteolysis suggests that the PSS(s) responsible for ethanolamine exchange in the ER and MAM might be either structurally distinct or have different topological arrangements in the membrane.
Phospholipid Substrate Preference of PSS1 and PSS2 in ER and MAM-In addition to using the free bases for the base exchange reaction, PSS also uses phospholipids as substrates. PSS1 uses both PtdCho and PtdEtn, whereas PSS2 uses only PtdEtn (20). Serine exchange activity was measured in the ER and MAM with exogenously added PtdEtn and PtdCho in the presence of the detergent Triton X-100. Very little serine exchange activity was detected in membranes assayed in the absence of exogenously added phospholipid (Fig. 3) because Triton X-100 inhibits the reaction (19). PtdEtn stimulated the serine exchange activity of both ER and MAM, but the enzymespecific activity was 39% less in the ER than in MAM (Fig. 3). In addition, PtdCho stimulated the serine exchange activity in MAM (by ϳ18-fold) to a greater extent than in the ER (by ϳ6-fold).
A combination of the results presented in Figs. 1-3 suggests that the bulk of PSS1, which is responsible for choline exchange activity, resides in MAM, whereas the ER contains little PSS1. In addition, ethanolamine exchange activity, which is contributed by both PSS1 and PSS2 in vitro, is present in both ER and MAM. These experiments do not, however, permit us to determine whether the ethanolamine exchange activity in the ER and MAM is contributed by PSS1 or PSS2.
Localization of PSS1 and PSS2 by Immunofluorescence Microscopy-The subcellular localization of PSS1 and PSS2 was further investigated using immunofluorescence confocal microscopy. A myc epitope tag was appended to the 5Ј-end of the cDNA encoding murine PSS1 and to the 3Ј-end of the cDNA encoding PSS2 (C-myc-PSS2), and the constructs were stably transfected into McArdle rat hepatoma cells. The localization of these proteins was compared with that of calnexin, a resident ER membrane chaperone protein, using double immunofluorescence confocal microscopy with antibodies directed against the myc epitope and calnexin. Staining for calnexin showed a punctate pattern of fluorescence throughout the cytoplasm, consistent with that expected for an ER protein (Fig. 4, B and  E). No immunoreactivity was observed in cells from which the primary antibody had been omitted (data not shown). myc-PSS1 (Fig. 4A) and C-myc-PSS2 (Fig. 4D) showed similar staining patterns that were somewhat more diffuse than that of calnexin (Fig. 4, B and E). Overall, however, PSS1 and PSS2 extensively co-localized with each other and with the ER marker, calnexin (Figs. 4, C and F).
Immunoloblotting Analysis of Localization of PSS1-We next determined by immunoblotting experiments if PSS1 protein was localized to MAM or was distributed throughout the ER. Proteins of MAM isolated from CHO-K1 and M.9.1.1 cells (mutant CHO cells defective in PSS1 activity) (20) were immunoblotted with anti-PSS1 antibody. Fig. 5A shows that PSS1 was present in MAM of wild-type CHO-K1 cells but was not detectable in MAM of M.9.1.1 cells (Fig. 5A). In addition, small amounts of PSS1 were detected in microsomes isolated from CHO-K1 cells but not in microsomes from M.9.1.1 cells (not shown). These results were as expected, since PSS1 activity is greatly reduced in M.9.1.1 cells (14,20) and these cells also lack detectable PSS1 mRNA (16). These observations are consistent with the report of Saito et al. (13), who showed that PSS1 was present in MAM and, to a lesser extent, in microsomes of CHO-K1 cells.
Microsomes are ER-enriched membranes that in addition to ER contain other organelle membranes including MAM. Consequently, the possibility existed that the presence of PSS1 in microsomes was due to MAM rather than the bulk of the ER. We therefore isolated MAM, mitochondria, and two membrane fractions highly enriched in ER from murine liver (18): ER1, which is enriched in rough ER, and ER2, which is enriched in smooth ER. Immunoblotting with anti-PSS1 antibody showed that PSS1 is localized to MAM (Fig. 5B); neither ER1 nor ER2 contained any detectable PSS1 protein (Fig. 5B). Included in Fig. 5B are immunoblots of the same membrane proteins probed with an antibody directed against rat liver PtdEtn N-methyltransferase-2, a specific marker protein for MAM (5). Similar to PSS1, and in agreement with previous observations (5), PtdEtn methyltransferase-2 was present in MAM but was absent from ER1 and ER2 (Fig. 5B). To confirm that the ER1 and ER2 fractions were indeed derived from the ER, the membrane proteins were also immunoblotted with an antibody directed against protein-disulfide isomerase, a resident ER protein. Protein-disulfide isomerase immunoreactivity was observed in ER1, ER2, and MAM and was absent from mitochondria (Fig. 5B), confirming that ER1 and ER2 are derived from the ER. These immunoblotting data demonstrate that in murine liver PSS1 resides in MAM but is undetectable in the ER and suggest that the PSS1 detected in microsomes of CHO-K1 cells was due to MAM being a constituent of the microsomal preparation.
One possible explanation for the apparent absence of PSS1 from the ER is that proteolytic cleavage of the N terminus of PSS1 occurs, generating a truncated protein that would not be recognized upon immunoblotting with the antibody used (the data in Fig. 5 were generated using an antibody directed against a peptide consisting of amino acids 1-17 of PSS1 from CHO-K1 cells). Therefore, if the ER had contained an isoform of PSS1 that lacked this N-terminal epitope, no immunoreactive protein would have been detected in the ER. However, this scenario is unlikely because Saito et al. (13) previously exam- Fluorescein isothiocyanate-conjugated anti-mouse IgG and Texas Redconjugated anti-rabbit IgG, respectively, were used to detect myc-PSS1 (A and C) and myc-PSS2 (D and F) and calnexin (B, C, E, and F). Co-localization of calnexin with PSS1 or PSS2 is indicated by yellow (C and F).
ined the subcellular location of PSS1 in CHO cells using two different antibodies, one directed against a peptide close to the N terminus (amino acids 4 -18) of PSS1 and the other directed against a peptide close to the C terminus (amino acids 447-463). Immunoreactive PSS1 protein with a molecular mass of 42 kDa was detected in membrane fractions of CHO cells using either antibody, suggesting that neither the N nor the C terminus of PSS1 in the membranes had been cleaved. In this study by Saito et al., the immunoreactive protein exhibited an anomalously low apparent molecular mass of 42 kDa upon polyacrylamide gel electrophoresis compared with the size predicted from the cDNA sequence (55.3 kDa). This atypical behavior of PSS1 upon electrophoresis was attributed to the exceptionally high content of hydrophobic amino acids in PSS1 (13). The PSS1 detected in our immunoblotting studies, using an antibody directed against the N terminus of PSS1 (amino acids 1-17), also has an apparent molecular mass of 42 kDa. In addition, data presented in Fig. 6, in which myc-tagged PSS1 was expressed in McArdle hepatoma cells and detected by immunoblotting with anti-myc antibody, also support the conclusion that PSS1 is highly enriched in the MAM but undetectable in the ER. Moreover, the observation that the ER contains little choline exchange activity (Fig. 2) supports the conclusion that PSS1 is highly enriched in MAM and is largely absent from the ER. However, the possibility that an N-terminally truncated isoform of PSS1 is present in the ER, which would not have been detected by our immunoblotting experiments, cannot be discounted.
The C-terminal KK Motif of PSS1 Is Not Required for Targeting to MAM-PSS1 contains two lysine residues at the carboxyl terminus of PSS1 (residues 472 and 473, i.e. -KK-COOH) (14,21). This motif is similar to that of reported ER-targeting consensus sequences (-KXKXX-COOH or -XXKKXX-COOH) (22)(23)(24). Since the results presented in Fig. 5 show that PSS1 is highly enriched in MAM, we hypothesized that the KK motif might be required for targeting PSS1 to MAM. The MAM marker protein, PtdEtn N-methyltransferase, similarly contains a highly positively charged C-terminal sequence (-RR-KATRLHKRS-COOH) (5). We therefore generated a cDNA encoding a mutant form of murine PSS1 in which the two lysine residues at positions 472 and 473 were deleted (this mutant is designated as PSS1:⌬cKK). In addition, a myc epitope tag was appended to the N terminus of PSS1 and PSS1:⌬cKK, and the cDNA constructs were stably expressed in McArdle rat hepatoma cells. To confirm that these modifications did not affect the activity of the expressed PSS1 protein, serine exchange activity was measured in cellular lysates. Fig. 6A shows that the expressed myc-PSS1 and the deletion mutant, myc-PSS1: ⌬cKK, possessed approximately equal serine exchange activities, at levels ϳ5-fold higher than in cells transfected with empty expression vector. Immunoblotting of cellular lysates with anti-myc antibody confirmed that the transfected cells also expressed approximately equal amounts of the two recombinant myc-tagged proteins (apparent molecular mass 42 kDa) (Fig. 6B). To determine whether or not the C-terminal lysine residues at positions 472 and 473 of PSS1 were required for the targeting of PSS1 to MAM, subcellular fractions (MAM, microsomes, and mitochondria) were isolated from the transfected cells, and immunoblotting experiments were performed using anti-myc antibody. In cells expressing either myc-PSS1 or myc-PSS1:⌬cKK, the majority of recombinant PSS1 protein was found in MAM. Both proteins were also present at lower levels in microsomes but were absent from mitochondria (Fig. 6C). Therefore, deletion of the lysines at positions 472 and 473 did not hinder the targeting of PSS1 to MAM. Moreover, the myctagged PSS1 was predominantly located in the MAM.
PSS2 Is Also Localized Specifically to MAM-Since the preceding data demonstrated that PSS1 was undetectable in the ER but was highly enriched in the MAM, we hypothesized that the serine and ethanolamine exchange activities in the ER would be imparted by PSS2. Support for the idea that the ER contains PSS2 comes from the observations that (i) the serine exchange activity in the ER is robustly stimulated by PtdEtn (Fig. 3), (ii) the ER contains ethanolamine exchange activity (Fig. 2), and (iii) serine exchange activity of the ER is inhibited by ethanolamine (Fig. 1). Microsomes, MAM, and mitochondria were isolated from CHO-K1 cells, and the proteins were immunoblotted with an antibody directed against a peptide corresponding to the C-terminal sequence of PSS2 from CHO cells. Fig. 7 shows the unexpected result that although PSS2 was abundant in MAM, little immunoreactive PSS2 (M r ϳ52 kDa) was detectable in microsomes from CHO cells.
Unfortunately, the anti-PSS2 antibody, which is directed against a peptide sequence of PSS2 from CHO cells, does not cross-react with murine liver PSS2, and all of our attempts to generate an anti-murine PSS2 antibody have been unsuccessful. Therefore, as an alternative approach to confirm that PSS2 is present primarily in MAM, we generated McArdle cells stably expressing murine PSS2 containing a C-terminal myc tag. In these cells, the serine exchange activity (specific activity 5.1 nmol/h/mg of protein) was ϳ7-fold higher than in control cells transfected with empty vector (specific activity 0.69 nmol/h/mg of protein). Immunoblotting of cellular lysates with anti-myc antibody confirmed that C-myc-PSS2 protein was expressed in the McArdle cells (Fig. 8A), as indicated by the presence of an ϳ52-kDa immunoreactive protein that was absent from cells transfected with empty vector alone. Subcellular fractions (microsomes, MAM, and mitochondria) were prepared from C-myc-PSS2-expressing cells and control cells, and proteins were immunoblotted with anti-myc antibody. Consistent with the results shown in Fig. 7 for CHO cells, C-myc-PSS2 was abundant in MAM but was undetectable in microsomes and mitochondria (Fig. 8B).
We considered the possibility that the apparent absence of PSS2 from the ER according to immunoblotting (Fig. 8) was the result of proteolytic cleavage of the myc epitope from the C terminus of the protein in the ER. This possibility was investigated by expression of a cDNA encoding PSS2 with a myc tag appended to the N terminus in McArdle cells. N-myc-PSS2 protein was also detected in the MAM but not in microsomes by immunoblotting using an anti-myc antibody (data not shown). Since N-myc-PSS2 and C-myc-PSS2 were similarly localized to the MAM, the presence of the myc tag clearly did not affect their subcellular targeting. The lack of immunoreactive PSS2 in the ER was probably not due to proteolytic cleavage of either the N or C terminus to generate a truncated PSS2. The data do not, however, eliminate the possibility that a truncated isoform of PSS2, from which both the N and C termini have been proteolytically cleaved, is present in the ER, but to our knowledge there are no known examples of such proteins in the ER. We therefore conclude that both PSS1 and PSS2 are highly enriched in the MAM but are largely excluded from the ER.
Conclusion-The studies presented herein show that fulllength PSS1 and PSS2 are localized almost exclusively to MAM and are largely excluded from the bulk of the ER. However, the possibility that truncated isoforms of these proteins are present in the ER cannot be completely eliminated. The reason why both PSS1 and PSS2 are so highly enriched in MAM is not clear. Our previous studies have demonstrated that nearly all mitochondrial PtdEtn is derived from imported PtdSer (7). Therefore, one possible explanation for the localization of the two PtdSer synthases in MAM is that a robust synthesis of PtdSer at this site, in close proximity to mitochondrial outer membranes, would provide an efficient mechanism for the import of newly synthesized PtdSer into mitochondria for decarboxylation to PtdEtn. The findings of the present study, however, leave open the question of which protein is responsible for the serine and ethanolamine exchange activities in the ER. Saito et al. (12) have used a CHO-K1 cell line, PSA-3, which is defective in PSS1, to generate a mutant cell line, PSB-2. The PSB-2 cells are defective in both PSS1 and PSS2 and contain FIG. 7. PSS2 is localized to MAM in CHO-K1 cells. Microsomes (Micr.), MAM, and mitochondria (Mito.) were isolated from CHO-K1 cells. Proteins (50 g) were separated by electrophoresis on a 10% polyacrylamide gel containing 0.1% SDS and then transferred to polyvinylidene difluoride membranes and incubated with affinity-purified antibody directed against a synthetic peptide corresponding to a sequence of PSS2 from CHO-K1 cells. PSS2 is indicated by the arrow at ϳ52 kDa.

FIG. 8. myc-PSS2 is localized to MAM in McArdle hepatoma cells.
The myc epitope was appended to the C terminus of murine PSS2 cDNA. A, proteins of cell lysates (50 g of protein) from McArdle cells transfected with empty expression vector (Control) or myc-PSS2 were separated by polyacrylamide gel electrophoresis and then immunoblotted using anti-myc antibody. B, microsomes (Micr.), MAM, and mitochondria (Mito.) were isolated from McArdle cells expressing myc-PSS2. Proteins (50 g) from each fraction were separated by electrophoresis on a 10% polyacrylamide gel containing 0.1% SDS, transferred to a polyvinylidene difluoride membrane, and then incubated with anti-myc antibody. The PSS2 protein is indicated by the arrow at ϳ52 kDa. essentially no choline exchange activity. The ethanolamine and serine exchange activities of PSB-2 cells are reduced to 4.8 and 11.5%, respectively, of those in wild-type CHO-K1 cells. It is possible that the residual base exchange activity detected in PSB-2 cells is contributed by a putative "ER isoform" of PSS, which, as indicated by our trypsin proteolysis experiments, might have a different structure and/or a different topological arrangement in the ER membranes compared with PSS1 and PSS2 in the MAM.