Mitochondrial Glycerol Phosphate Acyltransferase Contains Two Transmembrane Domains with the Active Site in the N-terminal Domain Facing the Cytosol*

The topography of mitochondrial glycerol-3-phosphate acyltransferase (GPAT) was determined using rat liver mitochondria and mutagenized recombinant rat GPAT (828 aa (amino acids)) expressed in CHO cells. Hydrophobicity analysis of GPAT predicts two transmembrane domains (TMDs), residues 472–493 and 576–592. Residues 224–323 correspond to the active site of the enzyme, which is believed to lie on the cytosolic face of the outer mitochondrial membrane. Protease treatment of rat liver mitochondria revealed that GPAT has a membrane-protected segment of 14 kDa that could correspond to the mass of the two predicted TMDs plus a loop between aa 494 and 575. Recombinant GPAT constructs containing tagged epitopes were transiently expressed in Chinese hamster ovary cells and immunolocalized. Both the C and N termini epitope tags could be detected after selective permeabilization of only the plasma membrane, indicating that both termini face the cytosol. A 6–8-fold increase in GPAT-specific activity in the transfected cells confirmed correct protein folding and orientation. When the C terminus and loop-tagged GPAT construct was immunoassayed, the epitope at the C terminus could be detected when the plasma membrane was permeabilized, but loop-epitope accessibility required disruption of the outer mitochondrial membrane. Similar results were observed when GPAT was truncated before the second TMD, again consistent with an orientation in which the loop faces the mitochondrial intermembrane space. Although protease digestion of the HA-tagged loop resulted in preservation of a 14-kDa fragment, consistent with a membrane protected loop domain, neither the truncated nor loop-tagged enzymes conferred GPAT activity when overexpressed, suggesting that the loop plays a critical structural or regulatory role for GPAT function. Based on these data, we propose a GPAT topography model with two transmembrane domains in which both the N (aa 1–471) and C (aa 593–end) termini face the cytosol and a single loop (aa 494–575) faces the intermembrane space.

catalyzes the first and committed step in de novo cellular glycerolipid synthesis, the formation of 1-acyl-sn-glycerol-3-phosphate (lysophosphatidic acid) from glycerol-3-phosphate and long chain fatty acyl-CoA substrates (1,2). Mammalian cells contain two GPAT isozymes that have different cellular locations, one in the endoplasmic reticulum and the other in mitochondria, that can be distinguished by sensitivity to inactivation by sulfhydryl reagents. The mitochondrial GPAT is an outer mitochondrial membrane (OMM) protein composing about 10% of the total enzymatic activity in most mammalian tissues. In liver, however, GPAT specific activity is similar in the two subcellular organelles.
Only the mitochondrial GPAT isoform has been cloned. Recombinant mouse (3) and rat (4) GPAT open reading frames encode proteins of 827 and 828 amino acids, respectively, with a high degree of homology. Alignment of rat mitochondrial GPAT with other acyltransferases shows four homology blocks (5), and when conserved amino acids within these regions are mutated, GPAT activity is altered or abolished (5)(6)(7)(8). Kinetic studies using recombinant GPAT with site-directed mutations in this region identify specific amino acids that contribute to substrate binding or to catalysis. Thus, it was surprising that a recent report in this journal did not include information on these extensive studies but concluded that the active site region of GPAT lay in a domain on the side of the mitochondrial membrane opposite to the substrate binding and catalytic domains (9).
Identification of the structure and orientation of membrane proteins is essential to elucidate enzyme interactions with substrates and products, especially when these are also hydrophobic. Furthermore, knowledge of the orientation of membrane domains is critical in determining their accessibility to regulatory signals. To clarify the orientation of mitochondrial GPAT and the topography of its active site in the OMM, we used native mitochondria and recombinant epitope-tagged proteins expressed in cultured CHO cells to probe the positions of the different domains. OR). Tissue culture and transfection media and reagents were from Life Technologies-Invitrogen (Princeton, NJ). Proteases and protease inhibitors were purchased from Sigma and Life Technologies.
Animals and Isolation of Liver Mitochondria-Female (150 -200 g) Sprague-Dawley rats were housed on a 12-h/12-h light/dark cycle with free access to water and fed ad libitum with Purina rat chow. They were then fasted for 48 h and refed for 24 h with a high sucrose (69.5%), low fat (0.5%) diet (Dyets Inc.) to up-regulate mitochondrial GPAT expression (10). Animals were killed by CO 2 narcosis. After dissection, livers were immediately placed on ice and homogenized in Buffer A (0.25 M sucrose, 25 mM Tris-HCl, pH 7.4, 10 mM EDTA, 1 mM dithiothreitol). Mitochondria were isolated by differential centrifugation (11), resuspended in Buffer A, and stored in aliquots at Ϫ80°C. Protein concentrations were determined by the BCA method (Pierce) using bovine serum albumin as the standard.
Limited Proteolysis of Mitochondria-Mitochondria were analyzed for protease-protected fragments after preincubation with buffer (Control) or with 1% Triton X-100 to disrupt the OMM. The integrity of the OMM was determined by measuring the activity of the intermembrane space marker, adenylate kinase (12). To determine the accessibility of GPAT domains to protease, 350 g of mitochondrial protein was then incubated in a final volume of 0.1 ml with selected proteases. For proteinase K, 20 g of protease was added with 50 mM Tris-HCl, pH 7.4, 0.25 M sucrose, and 2 mM CaCl 2 for 30 min on ice. The reaction was stopped by adding 1 mM AEBSF (4-(2-aminoethyl)benzenesulfonyl fluoride), 2 mM phenylmethylsulfonyl fluoride, 100 M leupeptin, and 2 g/ml aprotinin. After 5 min the samples were centrifuged at 11,000 ϫ g for 5 min, and the pellet was used for immunodetection of GPAT fragments. For Glu-C endoproteinase (V8 protease), the mitochondria (350 g of protein) were incubated with 0.062 V8 protease units in 50 mM ammonium acetate buffer, pH 4.0, for 90 min at 22°C. The reaction was stopped by adding 0.1 mM 2,4-diisochlorocoumarin and incubating for 5 min on ice. For chymotrypsin, mitochondria (350 g of protein) were incubated with 20 g of the protease in 0.125 M sucrose, 0.5 mM CaCl 2 , and 25 mM Tris-HCl, pH 8, at 22°C for 30 min. The reaction was stopped by adding 2 mM phenylmethylsulfonyl fluoride. For trypsin, mitochondria (200 g) were incubated with 25 g/ml protease in 0.125 M sucrose and 10 mM Tris-HCl, pH 8, at 22°C for 20 min. The reaction was stopped with the addition of 2 mg/ml soybean trypsin inhibitor.
Construction of Epitope-tagged GPAT-Four GPAT constructs were made (Fig. 3). The cDNA encoding the complete open reading frame of rat liver mitochondrial GPAT (4) was first subcloned in pcDNA3.1 (Invitrogen) digested with BamHI-XhoI (pcDNA3.1-GPAT). A FLAG epitope at the C terminus was added by a two-step polymerase chain reaction procedure in which the 30 nucleotides encoding the epitope, an XbaI restriction site, and a stop codon were added at the 3Ј end of the open reading frame (13). The polymerase chain reaction product was then inserted in the BamHI-XbaI sites of the multicloning site of pcDNA3.1. This construct pcDNA3.1-GPAT-FLAG is referred to as GFLAG.
To insert hemagglutinin (HA) epitopes, mutagenesis was performed using the Gene Editor system (Promega). For the mutagenic reactions, GPAT-FLAG (BamHI-XbaI) was subcloned into pGEM-11Zf(ϩ) (Promega), and this construct, pGEM11Z-GPAT-FLAG, was used as a template. The HA epitope was inserted near the N terminus after amino acid 33 using a mutagenic primer (75 nucleotides) that contains the 27 nucleotides (in bold in the nucleotide sequence below) that encode the HA epitope (YPYDVP-DYA) flanked with 22 GPAT cDNA complementary bases at the 5Ј end and 26 bases at the 3Ј end. The sequence of the mutagenic primer was: ATGTAAACACACGAATGAGGACTACCCATATGACGTCCCGGACTA-CGCCTGGGTTGACTGTGGCTTCAAACCTAC. Correct insertion of the epitope was corroborated by DNA sequencing and by restriction digestion with NdeI (the NdeI site is underlined in the mutagenic primer sequences). The mutated plasmid, pGEM11f(ϩ)-GPAT-FLAG-HA33, was then digested with BamHI and XbaI and subcloned into the BamHI-XbaI sites of pcDNA3.1 to generate pcDNA3.1-GPAT-FLAG-HA33 (HA33).
To construct GPAT truncated after amino acid 576, pcDNA3.1-GPAT was used as a template for polymerase chain reaction. A 5Ј primer contained the BamHI site and the first bases of the GPAT open reading frame, and a 3Ј primer contained an XhoI site and 21 complementary bases to the GPAT codons that encode amino acids 570-576. The polymerase chain reaction product was then subcloned in the BamHI-XhoI sites of the plasmid pcDNA3.1-Myc-His (Invitrogen). The construct pcDNA3.1-GPATtr576-Myc (Tr576Myc) was confirmed by DNA sequencing.
Expression of Epitope-tagged GPAT in CHO Cells-CHO K1 cells were grown in minimum essential medium containing 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin at 37°C with 5% CO 2 . Cells were grown in 100-mm dishes to 70 -80% confluence and then transfected with the constructs GFLAG, HA33, HA496, and Tr576Myc or with the empty vector pcDNA3.1. Cationic liposomes (LipofectAMINE, Life technologies) were used following the product instructions. After 27 h, the cells were harvested, homogenized in Buffer A, and centrifuged at 20,000 ϫ g for 20 min. The cellular particulate pellet was resuspended in buffer A, separated into aliquots, and stored at Ϫ80°C. Protein was determined using bovine serum albumin as a standard (14).
Immunocytochemistry Analysis of GPAT Expression-Immunocytochemistry was performed as described (15). CHO K1 cells (10 6 cells) were seeded in sterile 100-mm tissue culture dishes, each containing 4 -5 sterile glass coverslips, incubated at 37°C until they reached 70% confluence, and then transfected with the epitope-tagged GPAT constructs. Immunoblotting-The mitochondrial protease digestion products were separated on a 4 -20% gradient (8% for the cellular particulate of CHO cells expressing recombinant GPAT) polyacrylamide gel containing 1% SDS and transferred to a polyvinylidene difluoride membrane (Bio-Rad). For chemiluminescent detection, immunoreactive bands were visualized by incubating the membrane with horseradish peroxidase-conjugated goat anti-rabbit and anti-mouse IgG and PicoWest reagents (Pierce).
GPAT Assays-GPAT was assayed in rat liver mitochondria (20 -80 g of protein) and in total particulate preparations from CHO cells expressing the recombinant GPAT constructs (40 -80 g of protein). The assay was performed at 23°C with 300 M [ 3 H]glycerol 3-phosphate and 112.5 M palmitoyl-CoA in the presence or absence of 1 mM N-ethylmaleimide to inhibit the microsomal isoform (16). Microsomal GPAT was estimated by subtracting the N-ethylmaleimide-resistant activity (mitochondrial GPAT) from the total. All assays measured initial rates (5). [ 3 H]Glycerol-3-phosphate was synthesized enzymatically (17). GPAT activity was also assayed with immobilized palmitoyl-CoA on agarose beads (180 M) (Sigma) using rat liver mitochondria or CHO cell total particulate preparations under isosmotic conditions (intact) or hyposmotic conditions (10 mM Tris-HCl, pH 7.4), which disrupted the OMM. Disruption of the OMM was monitored by loss of adenylate kinase activity.

GPAT Is Predicted to Have at Least Two Transmembrane Domains-
The GPAT amino acid sequence deduced from the cDNA complete open reading frame was examined using algorithms that predict the possible location of TMDs based on the hydrophobicity of the residues. Four different programs available on the Internet (TMPred (18), TMHMM (19), DAS (20), and SOSUI) strongly predicted the presence of two TMDs between amino acids 472-493 and 576 -592 (Fig. 1A). These two TMDs were the only ones predicted by TMHMM and SOSUI. These programs also identify three non-membrane segments: an N-terminal region (aa 1-471), a loop region (aa 494 -575), and a C-terminal region (aa 593-828).
DAS and TMPred predict two additional TMDs between residues 184 -203 and 235-254. These predicted TMDs have lower scores, and because the region aa 235-254 lies between two highly conserved regions that correspond to the active site of GPAT (5,6,8), this predicted TMD is highly improbable because it would cause portions of the active site to lie on opposite faces of the OMM. The TMD predicted for aa 184 -203 is the only possible one between the N terminus and the active site domain. Specific experiments were carried out to determine the existence of this TMD and the two prime TMD predictions.
The Outer Mitochondrial Membrane Protects Some GPAT Segments from Protease Hydrolysis-When rat liver mitochondria were treated with different specific and nonspecific proteases and the reaction products were blotted with a polyclonal anti-full-length GPAT antibody, several immunoreactive bands were detected (Fig. 2, A-C), suggesting that the OMM had acted as a barrier for the proteolysis. One of the bands (13.5/16 kDa) corresponds approximately to the size of the loop plus the two TMDs (calculated: 13.6 kDa). The higher molecular mass bands (23,26,28, and 31 kDa) might result from incomplete hydrolysis of protected sites within nearby membrane-associated regions. As previously reported (21), trypsin did not hydrolyze GPAT unless the OMM was disrupted (Fig. 2D).
Epitope-tagged GPAT Can Be Expressed in CHO Cells-To determine the orientation of GPAT, several constructs were designed to contain specific epitopes in the domains delimited by the two TMDs or truncations of GPAT proximal to the second TMD (Fig. 1B). CHO K1 cells were transiently transfected with these constructs, and the expression of GPAT was monitored by Western blot both with the specific antibodies for the epitopes (anti-FLAG, anti-HA, and anti-Myc) and with assays for GPAT activity (Fig. 3). Although the four proteins corresponding to the different constructs were expressed with the correct predicted molecular weights, only the GPAT transfectants with epitopes added at the C terminus (GFLAG) or near the N terminus (HA33) were active. Compared with vector-transfected control cells membranes, GFLAG and HA33 cells had 6 -8-fold increases in mitochondrial GPAT specific activity. Cells transfected with GPAT constructs containing the HA epitope in the loop (HA496) or with the truncated construct (Tr576Myc), however, expressed GPAT activity similar to the vector-transfected control. This lack of activity was surprising because neither of the loop constructs was near the active site domain, and in both, the active site domain remained on the opposite, cytosolic side of the OMM (see below).
Because previous studies indicated that the active site of GPAT faces the cytosol (21-23), we tested GPAT activity in CHO cell total particulate preparations by using immobilized palmitoyl-CoA as a substrate to corroborate the correct insertion of the recombinant proteins (Table I). GPAT was able to access agarose-bound palmitoyl-CoA under isosmotic conditions when the OMM was intact, showing that the active site of the enzyme faces the cytosol and can interact with agarose beads that cannot cross the membrane. Furthermore, disruption of the OMM under hyposmotic conditions did not increase Two additional membrane-associated segments are predicted to lie at residues 184 -203 (a) and 235-254 (b) (gray blocks). The putative membrane-associated segment b is located within the domain containing the acyltransferase homology blocks (dark block), which comprises the active site of the enzyme. B, epitope-tagged GPAT cDNA constructs used to determine the protein topography. The antigenic epitope FLAG (diamonds) was added in the C terminus of the GPAT (GFLAG), HA epitopes (triangles) were inserted within the loop (HA496) or near the N terminus (HA33) before the first predicted transmembrane domain, or a Myc epitope (oval) was added to the truncated protein as shown (Tr576Myc). GPAT activity, indicating a lack of latency of GPAT activity. The integrity of the OMM under each condition was confirmed by assaying the activity of the intermembrane marker, adenylate kinase. Full adenylate kinase activity was present in the membranes treated with isosmotic buffer, whereas 75% of adenylate kinase activity was lost after incubation with the hyposmotic buffer. Similar results were obtained when we tested the GPAT activity in intact and hyposmotic-swelled rat liver mitochondria. GPAT specific activity in intact mitochondria (adenylate kinase 100% active) was 0.62 nmol/min/mg of pro-tein and in OMM-disrupted mitochondria (adenylate kinase 22% active) was 0.58 nmol/min/mg of protein.
Epitopes Present in the Loop Region Are Detected Only after OMM Disruption-In each of the epitope-tagged GPAT transfectants, the location of each epitope relative to the OMM was determined by probing with specific fluorescent antibodies after digitonin or Triton X-100 treatment. Digitonin permeabilizes only the plasma membrane, whereas Triton X-100 permeabilizes intracellular membranes, including the OMM (24,25). We confirmed this differential permeabilization by examining the susceptibility of adenylate kinase to protease digestion. When cells were incubated with 50 g/ml proteinase K and then permeabilized with digitonin, adenylate kinase activity in both total cellular particulate and intact CHO cells was 100% of control activity, showing a lack of protease entry into the intermembrane space. In contrast, cells or total cellular particulate fractions permeabilized with 1% Triton X-100 lost 75% of the adenylate kinase activity, indicating disruption of the OMM that allowed proteinase K to have access to the enzyme.
Fluorescent images for the FLAG, HA, and Myc epitope antibodies (green) or the mitochondrial marker (red) were analyzed either after the plasma membrane was permeabilized with digitonin or after the intracellular membranes were permeabilized with Triton X-100 (Fig. 4). Specificity of the antibody signal was tested by probing control CHO cells transfected with the empty plasmid with both the primary antibody (anti-HA, anti-FLAG, or anti-Myc) and the fluorescein isothiocyanate-labeled secondary antibody. CHO cells expressing the epitope-tagged GPAT constructs were also probed with only the secondary antibody. In both cases, the fluorescent signal re-

FIG. 3. Expression of epitope-tagged-GPAT in CHO-K1 cells.
Expressed GPAT was probed using epitope-specific antibodies, and enzymatic activity was measured. CHO-K1 cells were transiently transfected with the pcDNA3.1-GPAT constructs or pcDNA3.1 empty vector, and total cellular particulates were obtained after 27 h. Total particulate protein (75 g) from cells transfected with GFLAG, HA496, HA33, Tr576Myc, or the empty vector pcDNA3.1 (VEC) were separated by 8% SDS-polyacrylamide gel electrophoresis and probed with the monoclonal antibodies, anti-FLAG, anti-HA and anti-Myc. N-ethylmaleimide-resistant GPAT activity was assayed in the total cellular particulates, and the values are expressed as the percent of the maximal activity (0.54 nmol/min/mg of protein). The results are representative of at least three independent transfection experiments. The molecular mass of the expressed protein was 92 kDa (full-length) and 68 kDa (truncated).

TABLE I
The active site of GPAT faces the cytosol N-Ethylmaleimide-resistant (mitochondrial) GPAT activity was measured with agarose-immobilized palmitoyl-CoA as a substrate in total particulate preparations from CHO cells expressing GPAT-FLAG and GPAT-HA33 after either disruption by hypotonic swelling or incubation under isotonic conditions. The degree of the OMM disruption was assessed by the percentage of adenylate kinase (AK) activity recovered under each condition. Control adenylate kinase activity was 8.1 M ATP/min. GPAT activity is expressed as nmol/min/mg of protein. The results show that the C terminus of GPAT faces the cytosol, because in the full-length GPAT construct tagged in the C terminus (GFLAG), this epitope was recognized in digitonin-treated cells when only the plasma membrane was permeabilized (Fig. 4). As expected, no change in staining was observed when these cells were permeabilized with Triton X-100. Similarly, the full-length GPAT containing an HA epitope near the N terminus (HA33) could be probed with anti-HA antibody after permeabilization with either digitonin or Triton X-100, showing that in this construct the N terminus is exposed to the cytosol. Both GFLAG and HA33 transfectants expressed high GPAT activity (8-and 6-fold more than control cells) (Fig. 3) and had their active sites exposed on the cytosolic surface, as evidenced by activity with agarose-linked palmitoyl-CoA substrate (Table I). These data indicate that the N and C termini and the active site of GPAT all face the cytosolic face of the OMM.
Two constructs were made to test the presence of the putative loop between residues 472-493 and 576 -592. CHO cells transfected with HA496 were identified with the HA antibody only after permeabilization with Triton X-100 but not after digitonin. This difference indicates that the HA epitope was not accessible to the antibody unless the OMM was disrupted. As was observed in GFLAG and HA33 transfectants, the C-terminal FLAG epitope present in the HA496 transfectants remained accessible to the FLAG antibody in digitonin-permeabilized cells, indicating that the C terminus of this construct was in its usual position on the cytoplasmic face of the OMM. The correct orientation of the GPAT-HA496 construct was further verified in total membrane fractions obtained from CHO cells that expressed HA496. When these membranes were exposed to proteinase K, anti-HA antibody detected a membraneprotected fragment whose molecular mass (ϳ14 kDa) matches the calculated molecular mass of the loop plus the two TMDs (Fig. 5). This fragment did not react with the anti-FLAG antibody, consistent with protease degradation of the FLAG epitope and C terminus on the cytosolic face of the OMM. Adenylate kinase activity corroborated the integrity of the OMM.
Similarly with the second loop construct, Tr576Myc, which is truncated just before the second TMD, cells became stained with the Myc antibody only after intracellular membranes had been disrupted with Triton X-100; the Myc epitope was not accessible to the antibody in cells permeabilized with digitonin, indicating that the shortened C terminus was located in the intermembrane space. Taken as a whole, these results indicate that GPAT C-and N-terminal domains are located in the cytosolic face of the OMM and that the loop lies in the intermembrane space.

DISCUSSION
The mitochondrial isoform of GPAT is an intrinsic membrane protein (1) and requires detergents for isolation and purification (26,27). Studies of intrinsic membrane proteins in internal mammalian membranes have most commonly investigated proteins of the endoplasmic reticulum where the methodological techniques to determine cytosolic and lumenal domains are well established (15,25,28,29). Only a few OMM proteins have been topographically studied (30,31). The fragile OMM contains structural pores and is easily disrupted during isolation and storage. Additionally, the OMM is permeable to molecules with molecular masses less than 6 kDa, which makes it difficult to use conventional labeling or protein modification reagents to determine the sidedness of a residue.
With protease digestion of native intact mitochondria, we found that at least one GPAT segment is protected from hydrolysis by V8 protease or chymotrypsin and that the size of this fragment matches the theoretical molecular mass of the loop plus the two TMDs. Additional higher molecular weight fragments, however, were observed when intact mitochondria were exposed to proteinase K. Some membrane proteins have a tight tertiary structure lying outside the membrane-spanning regions (32), and this folded structure may prevent proteases from cleaving at all their potential recognition sequences. Furthermore, some proteins contain domains that have a compact conformation with helices parallel to the membrane plane, and these may not be accessible to proteases (33). It is likely that GPAT has one or more protease-protected regions in extramembrane domains, especially at its active site, because its acyl-CoA substrate and its lysophosphatidic acid product are both amphipathic molecules and may need to interact with the mitochondrial membrane itself or with another hydrophobic or amphipathic structure.
Mutagenesis studies performed on the highly homologous Escherichia coli GPAT (5, 6) and mouse GPAT (7) have clearly demonstrated that the active site of GPAT is located in a highly conserved region corresponding to the amino acids 224 -323. Designation of this area as the active site is also supported by studies of mutations in the human acyl-CoA:dihydroxyacetonephosphate acyltransferase, in which the substitution of His or Cys for Arg-211 in the homologous active site region inactivates the enzyme completely (34). It was previously concluded that the GPAT active site faces the cytosol because GPAT activity was similar when either the palmitoyl-CoA substrate or a polymyxin B activator were added in a soluble form or immobilized on agarose beads (23), which cannot pass through the OMM (30). We corroborated the finding that the active site lies on the cytoplasmic surface when we found that agarose-linked palmitoyl-CoA was a substrate for GPAT in intact mitochondria and in the GFLAG and HA33 constructs and that disruption of the OMM did not reveal latent GPAT activity measured with the immobilized substrate. Based on these data, we conclude that the GPAT active site region from aa 224 -323 faces the cytosol.
TMD prediction algorithms varied in their predictions. All predicted two TMDs that followed the active site domain, but some also predicted a TMD that lay between aa 184 and 203 (Fig. 1A) and just before the active site domain (aa 224 -323). If the N terminus were cytosolic, this TMD would place the active site domain in the intermembrane space. To delineate the topography of GPAT and to determine accurately where the cytosolic and intermembrane domains lay, we transfected epitope-tagged GPAT constructs into CHO cells and localized the epitopes by immunofluorescence. Epitope tagging has been used successfully to elucidate the topography of proteins in the plasma membrane (35) and endoplasmic reticulum (15,24,25,36). We designed the mutated constructs so that the epitopes would be located near the N terminus, at the C terminus, and in the putative loop. Although little is known about the targeting of OMM proteins, the first amino acids may provide a targeting signal (22), so we inserted the HA epitope 33 residues after the N terminus. No predicted TMDs are present between the N terminus and aa 33. In each instance, this C terminus HA33 epitope was visualized after digitonin permeabilization of the plasma membrane, indicating that the C terminus is exposed to the cytosol. Antibodies did not interact with the epitopes located in either the beginning or the end of the hydrophilic domain of the loop (HA496 and Tr576Myc) in the cells unless the OMM was permeabilized, whereas epitopes located at the N and C termini could be probed without disrupting the OMM. Furthermore, immunofluorescence studies and protease treatment of the native enzyme and the HA496 construct showed that GPAT contains one domain that is not accessible from the cytosolic side and has a molecular mass consistent with the loop plus two TMDs (aa 472-593). Thus, the N terminus and the active site domain face the cytosolic surface without an additional intervening TMD. The C terminus faces the cytosol in the three full-length constructs expressed in CHO cells regardless of epitopes that were present elsewhere. These data validate the correct orientation of the recombinant GPAT at its C terminus.
The protein topography deduced from our results (Fig. 6) is opposite that previously proposed (9), which postulated a model of GPAT with N and C termini facing the intermembrane space and a loop that faces the cytosol. In the previous report, peptide antibodies were raised against an internal N-terminal domain distal to the active site (aa 420 -435), against a loop sequence (aa 543-559), and against a sequence near the C terminus (aa 726 -740) (see Fig. 1). These antibodies were tested on native GPAT in rat liver mitochondria. Only the antibody raised against the loop abolished GPAT activity. None of these three peptide antibodies was directed against the GPAT active site region confirmed by site-directed mutagenesis to be responsible for catalysis and substrate binding (5)(6)(7)(8). This active site region is further substantiated by a mutation in a closely related human gene (34). Thus, it is not entirely surprising that the antibodies to the N and C regions had no apparent effect on GPAT activity in intact mitochondria. Although the antibody directed to the loop region did inhibit GPAT activity, no marker was used to prove continued OMM integrity, and it is possible that GPAT inactivation occurred because of structure-function changes similar to our observations with the two loop constructs. These investigators also made two GPAT-GFP fusion constructs and transfected these into Cos-7 cells. One 120-kDa construct contained a GFP sequence inserted at aa 115, and the second expressed GFP at the C terminus. Trypsin did not cleave the fusion protein in intact mitochondria but digested the GFP domains after the membrane had been solubilized with CHAPS. Like an earlier report by these investigators (21), we also observed that trypsin does not digest GPAT efficiently (Fig. 2D), indicating that this protease is uninformative for any investigation of GPAT topography (37). Susceptibility to a protease only in the presence of a detergent is not informative because the detergent may expose residues that are not ordinarily accessible (38), or with recombinant proteins, the lack of digestion may occur because the protein is mis-oriented in the membrane. Furthermore, these investigators did not show that the transfected GFP-GPAT fusion constructs express GPAT activity or that the enzyme could use palmitoyl-CoA provided to the cytosolic face of intact mitochondria, as they had previously reported (21,39) and as would be expected if GPAT were correctly oriented. Thus, we conclude that the topography inferred by the previous report was in error.
Human and bacterial lysophosphatidic acid acyltransferases (LPAATs), which catalyze the second step in glycerolipid synthesis, share with GPAT structurally similar substrates and products but are much smaller proteins (ϳ285 amino acids). Although sequence alignment between GPAT and LPAAT shows that a homologous active site region is present in the LPAAT N-terminal domain, LPAAT lacks the ϳ600 -828 aa C-terminal domain of GPAT, suggesting that the C-terminal domain of GPAT is not involved in catalysis but, instead, might play a regulatory role. Because both 5-amino-4-imidazolecarboxamide riboside, an activator of AMP-activated kinase and recombinant AMP-activated kinase itself decrease GPAT activity (40), it is possible that one of the two consensus sequences for AMP-activated kinase present in the GPAT C-terminal domain might regulate GPAT activity.
The GPAT-specific activities in cells transfected with recombinant proteins tagged at the N and C termini were 6 -8-fold higher than in vector-transfected cells. Surprisingly, the GPAT-specific activities of the two protein constructs containing loop modifications were similar to that of the vector control. This observation indicates expression of an inactive enzyme even though its orientation in the OMM was correct. The loop region faced the intermembrane space, and the C terminus faced the cytosolic face, as determined by immunofluorescence. This interpretation is supported by the presence of the HA loop epitope in a fragment that was protected from protease digestion by the OMM and that had a mass consistent with the theoretical size of the loop. Although the HA insertion did not change the orientation of the active site, lack of an increase in GPAT activity after protein overexpression suggests that the loop is important for activity. Although a number of receptors and transport proteins contain critical functional domains that lie on both sides of the plasma membrane, we are unaware of another intrinsic membrane enzyme whose activity is disturbed by altering a domain that lies on a membrane surface opposite that of the active site. Loops can stabilize membrane proteins to an extent comparable with ligand binding (41) and can also destabilize proteins if interactions with other proteins cause an altered relationship of the associated TMDs (42). Thus, some loop residues might interact with other molecules that stabilize GPAT structure and/or activity. Alternatively, the modified loop in our constructs could alter the spatial conformation of the cytosol-facing domains that are required for interaction of the active site region and the C-terminal domain. This interpretation is consistent with the fact that the LPAAT proteins lack a predicted TMD after their active site region, instead terminating only about 70 aa beyond it. These studies have raised an interesting question about the functions of the GPAT loop and C-terminal domains which will require future experiments.