Membrane Topology of S2P, a Protein Required for Intramembranous Cleavage of Sterol Regulatory Element-binding Proteins*

In sterol-depleted mammalian cells, a two-step proteolytic process releases the NH2-terminal domains of sterol regulatory element-binding proteins (SREBPs) from membranes of the endoplasmic reticulum (ER). These domains translocate into the nucleus, where they activate genes of cholesterol and fatty acid biosynthesis. The SREBPs are oriented in the membrane in a hairpin fashion, with the NH2- and COOH-terminal domains facing the cytosol and a single hydrophilic loop projecting into the lumen. The first cleavage occurs at Site-1 within the ER lumen to generate an intermediate that is subsequently released from the membrane by cleavage at Site-2, which lies within the first transmembrane domain. A membrane protein, designated S2P, a putative zinc metalloprotease, is required for this cleavage. Here, we use protease protection and glycosylation site mapping to define the topology of S2P in ER membranes. Both the NH2 and COOH termini of S2P face the cytosol. Most of S2P is hydrophobic and appears to be buried in the membrane. All three of the long hydrophilic sequences of S2P can be glycosylated, indicating that they all project into the lumen. The HEIGH sequence of S2P, which contains two potential zinc-coordinating residues, is contained within a long hydrophobic segment. Aspartic acid 467, located ∼300 residues away from the HEIGH sequence, appears to provide the third coordinating residue for the active site zinc. This residue, too, is located in a hydrophobic sequence. The hydrophobicity of these sequences suggests that the active site of S2P is located within the membrane in an ideal position to cleave its target, a Leu-Cys bond in the first transmembrane helix of SREBPs.

transcription of these genes by interfering with the proteolysis of SREBPs. This feedback regulation allows cells to maintain a steady membrane composition in a changing environment (1).
Our laboratory recently isolated a cDNA encoding a protein that is a candidate for the Site-2 protease. We named this protein S2P. The cDNA was identified by complementation cloning (12) in M19 cells, which are a mutant line of CHO cells that cannot cleave SREBPs at Site-2 (3). As a result of this defect, the M19 cells are unable to transcribe genes encoding cholesterol biosynthetic enzymes or the LDL receptor, and they are therefore cholesterol auxotrophs (3,12,13). M19 cells fail to produce an mRNA encoding S2P, owing to a genomic rearrangement that disrupts the S2P gene (12). When transfected into M19 cells, the cDNA encoding S2P restores Site-2 cleavage of SREBPs and relieves the growth requirement for cholesterol.
The human S2P cDNA encodes a protein of 519 amino acids that contains a consensus metalloprotease zinc-binding site, HEIGH (12). Replacement of either of the two histidines or the glutamic acid destroys the ability of the S2P cDNA to restore SREBP cleavage in M19 cells (12). These findings are compatible with the hypothesis that S2P is indeed a metalloprotease (14), but so far, multiple attempts to demonstrate protease activity of isolated S2P in vitro have failed. S2P differs from known metalloproteases in its extensive hydrophobicity, which suggests the existence of multiple membrane-embedded domains. The protein also contains a stretch of 23 contiguous serine residues beginning at amino acid 114 and a cysteine-rich region containing 10 cysteines (residues 285-377). These two domains constitute the most hydrophilic parts of the protein.
A full understanding of the mechanism of the S2P cleavage reaction requires a more complete structural and functional analysis of S2P. In the current study, we propose a model for the membrane topology of S2P and test it through examination of patterns of N-linked glycosylation and protease susceptibility.

EXPERIMENTAL PROCEDURES
Materials-CHO-7 cells are a clone of CHO-K1 cells selected for growth in lipoprotein-deficient serum (15). M19 cells are a mutant line of CHO-K1 cells auxotrophic for cholesterol and unsaturated fatty acids (13), owing to a deletion in the S2P gene (12). We obtained monoclonal antibody IgG-HSV-Tag TM from Novagen, Inc.; monoclonal anti-BiP antibody from StressGen Biotechnologies Corp.; trypsin (Cat. No. 109827) and soybean trypsin inhibitor from Calbiochem-Novabiochem, Inc.; Triton X-100 from Roche Molecular Biochemicals; and peptide N-glycosidase F (PNGase F), endoglycosidase H, neuraminidase, and restriction enzymes from New England Biolabs, Inc. Other reagents were obtained from sources as described previously (8,9,15).
Recombinant Plasmids-All vectors expressing S2P or S2P mutants were driven by the cytomegalovirus (CMV) promoter-enhancer in the pcDNA3 vector (Invitrogen). Expression vectors pCMV-HSV-S2P and pCMV-Myc-S2P have been previously described (12). pCMV-HSV-LDLR-S2P is an expression vector encoding a fusion protein consisting of an initiator methionine, two tandem copies of the HSV epitope (QPELAPEDPED) (16), amino acids 811-860 of the human LDL receptor (17), three novel amino acids (TGD, generated by blunt end ligation of filled-in AgeI and BspDI restriction sites), and amino acids 2-519 of human S2P (12). pCMV-Myc-S2P-LDLR-HSV encodes a fusion protein consisting of an initiator methionine, two tandem copies of the c-Myc epitope (EQKLISEEDLN) (18), two novel amino acids (ID) encoded by the sequence for the BspDI restriction site, amino acids 2-519 of human S2P, two novel amino acids (TG) encoded by the sequence for the AgeI restriction site, amino acids 811-859 of the human LDL receptor, two novel amino acids (TG) encoded by the sequence for the AgeI restriction site, and two tandem copies of the HSV epitope (QPELAPEDPED). pSRE-Luciferase encodes a luciferase reporter cDNA driven by a promoter consisting of three tandem copies of Repeat 2 ϩ 3 of the human LDL receptor promoter plus the adenovirus E1B TATA box (19). pCMV␤ encodes ␤-galactosidase driven by the cytomegalovirus promoter/enhancer (Stratagene). Expression plasmids that encode S2P mutants were derived from the above mentioned plasmids using the QuikChange TM site-directed mutagenesis kit (Stratagene). The open reading frames of all expression vectors were sequenced in their entirety.
Transient Transfection of 293 Cells-Monolayers of human embryonic kidney 293 cells were set up on day 0 (4 ϫ 10 5 cells/60-mm dish) and cultured in 8 -9% CO 2 at 37°C in medium A (Dulbecco's modified Eagle medium containing 100 units/ml penicillin and 100 g/ml streptomycin sulfate) supplemented with 10% (v/v) fetal calf serum. On day 2, the cells were transfected with the indicated plasmids using an MBS kit (Stratagene) as described previously (2). Three h after transfection, the cells were switched to fresh medium A supplemented with 10% fetal calf serum, incubated overnight, and harvested on day 3.
Luciferase Reporter Assays-This assay is an indirect measure of the ability of S2P cDNAs to restore sterol-regulated transcriptional activity in transfected M19 cells using the pSRE-Luciferase reporter plasmid. On day 0, M19 cells were set up at 10 5 cells/22-mm well in medium B (a 1:1 mixture of Ham's F-12 medium and Dulbecco's modified Eagle medium containing 100 units/ml of penicillin, 100 g/ml streptomycin sulfate, and 5% fetal calf lipoprotein-deficient serum) supplemented with 20 M sodium oleate. On day 1, cells were transfected using the MBS kit as described previously (12) with 1 g of an expression vector containing no cDNA insert (pcDNA3.1) or the indicated wild-type or mutant HSV-tagged S2P cDNA together with 0.8 g of pSRE-Luciferase reporter plasmid and 0.05 g of pCMV␤ as a control for transfection efficiency. After incubation for 3 h at 35°C and 3% CO 2 , the cells were switched to medium B supplemented with 50 M sodium mevalonate, 50 M sodium compactin, and 0.2% (v/v) ethanol that contained either no sterols or an amount of sterols that gave a final concentration of 1 g/ml 25-hydroxycholesterol plus 10 g/ml cholesterol. On day 2, whole cell lysates were prepared and assayed for luciferase and ␤-galactosidase activities with substrates that generate chemiluminescence products as described (12).
Cell Fractionation-Duplicate monolayers of cells were scraped into the culture medium, centrifuged at 1000 ϫ g for 5 min at 4°C, resuspended in 1 ml of phosphate-buffered saline, and centrifuged again as above. The cell pellet from each dish was resuspended in 0.4 ml of Buffer A (10 mM Hepes-KOH at pH 7.4, 250 mM sucrose, 10 mM KCl, 1.5 mM MgCl 2 , 1 mM sodium EDTA, 1 mM sodium EGTA, 10 g/ml leupeptin, 5 g/ml pepstatin A, and 1.9 g/ml aprotinin), passed through a 22-gauge needle 30 times, and centrifuged at 1000 ϫ g for 5 min at 4°C. The pellet was then used to prepare a nuclear extract fraction as described (3). The supernatant was centrifuged at 2 ϫ 10 4 ϫ g for 10 min, and the resulting pellet was designated the membrane fraction.
Trypsin Proteolysis-293 cells were harvested, and cell membranes were prepared as described above. The membrane pellets were suspended in 0.1 ml of Buffer B (Buffer A supplemented with 0.1 M NaCl) or in Buffer B containing 1% (v/v) Triton X-100 and rocked at 4°C for 5 min. Varying amounts of trypsin were added in a volume of 2 l, and the samples were incubated at 25°C for 30 min in a final volume of 92 l. The reactions were stopped by the addition (2 l) of 400 units of soybean trypsin inhibitor, followed by the addition of 0.1 ml of Buffer C (125 mM Tris-HCl at pH 6.8, 8 M urea, and 5% (w/v) SDS) and 50 l of Buffer D (150 mM Tris-HCl at pH 6.8, 15% SDS, 12.5% (v/v) 2-mercaptoethanol, 25% (v/v) glycerol, and 0.02% (w/v) bromphenol blue). The samples were then incubated at 37°C for 1 h and subjected to SDS-PAGE on an 8% gel.
Glycosidase Treatment-Monolayers of 293 cells were set up for experiments and transfected as described above. Three h after transfection, cells were switched to medium A supplemented with 10% fetal calf serum. After incubation at 37°C for 20 h, the cells were harvested, and the pooled cells from four dishes were fractionated to obtain the membrane fraction. The 2 ϫ 10 4 ϫ g pellets were resuspended in 180 l of Buffer E (Buffer A containing 1% Triton X-100 without protease inhibitors). Aliquots of the membrane fractions (40 l) were digested by the indicated glycosidase as described in the figure legends. After digestion, 50 l of Buffer C and 25 l of Buffer D were added to each reaction, and the resulting samples were then incubated at 37°C for 1 h and subjected to SDS-PAGE.
Immunoblot Analysis-Protein concentration was measured with a BCA kit (Pierce). Gels were calibrated with prestained molecular weight markers (New England Biolabs). Following SDS-PAGE, proteins were transferred to Hybond C-extra nitrocellulose sheets (Amersham Pharmacia Biotech) and incubated with monoclonal antibodies at the indicated concentration. Bound antibodies were visualized by chemiluminescence with horseradish peroxidase-conjugated, affinitypurified donkey anti-mouse IgG (Jackson Immunoresearch Laboratories, Inc.) using the ECL reagent (Amersham Pharmacia Biotech) according to the manufacturer's instructions. Filters were exposed to Reflection TM NEF-496 film (NEN Life Science Products) at room temperature for the indicated time. Fig.  1 shows a hydropathy plot of S2P, and the middle panel shows the predicted membrane topology of the protein. The bottom panel shows the amino acid sequence with hydrophobic regions indicated by shading. The three hydrophilic segments that are believed to reside in the ER lumen are designated A-C. Hydrophobic segments that are postulated to lie within the membrane are denoted by dashed lines in the middle panel.

Model for Membrane Topology of S2P-The top panel of
The sequence of S2P contains an uneven distribution of the charged amino acids lysine, arginine, glutamate, and aspartate (indicated by asterisks in the bottom panel of Fig. 1). Amino acids 1-20 are extremely hydrophobic and are predicted to constitute a membrane-spanning helix. Residues 21-70 are relatively hydrophilic with several charged amino acids. This is designated region A. Residues 71-107 are again hydrophobic and are predicted to insert into the membrane. These residues are followed by the second long hydrophilic segment, designated region B, which extends from residues 108 -141. Region B contains 26 serines, 23 of which are contiguous. Residue 141 begins a very long hydrophobic sequence that ends at residue 258. This sequence is interrupted by a short hydrophilic sequence that begins with the HEIGH sequence at position 171 and extends to the arginine at 185. Four out of 15 residues in this segment are charged. The long sequence between residues 258 and 446, designated region C, contains numerous charged and polar amino acids. This sequence also contains multiple cysteines. The sequence between residues 447 and 476 contains only one charged amino acid and is predicted to dip into the membrane. It is followed by a short hydrophilic segment, residues 477-491, the location of which could not be assigned from these studies (see below). The protein terminates with another hydrophobic sequence extending from residues 492 to 519. The sole charged residue in this segment is the COOH-terminal arginine.
N-Linked Glycosylation of S2P-S2P contains two sequences that conform to the Asn-X-Ser/Thr consensus for N-linked glycosylation (asparagines 337 and 508) as indicated in Fig. 2. Asparagine 337 is in the cysteine-rich hydrophilic loop (region C), and asparagine 508 is in the COOH-terminal hydrophobic sequence. In order to determine whether N-linked glycosyla- based on data in this manuscript. Solid lines denote regions of S2P for which the topology has been assigned; dashed lines denote regions of S2P that are of indeterminate location and may be within the membrane. N337 denotes the position of the N-linked glycosylation site that is shown experimentally to carry a carbohydrate chain. HEIGH denotes the putative pentapeptide zinc binding site of S2P (12), and LDG denotes a conserved sequence that may contribute to zinc binding. The cysteine-rich region is denoted by the multiple letters C. Bottom panel, shaded and unshaded sequences denote regions of S2P that are poor or rich in charged residues, respectively. Asterisks denote these charged residues.
FIG. 2. Glycosylation of S2P at asparagine 337, but not at asparagine 508. The diagram shows a schematic of the HSV-S2P fusion protein. White circles denote two potential sites of N-linked glycosylation: Asn-337 and Asn-508. On day 0, 293 cells were set up for experiments as described under "Experimental Procedures." A, on day 2, the cells were transfected with 5 g/dish of either the control vector pcDNA3 (lane 1) or the indicated plasmid (lanes 2-6). Three h after transfection, the cells were switched to fresh medium A supplemented with 10% fetal calf serum. After incubation for 20 h at 37°C, membrane fractions were prepared as described under "Experimental Procedures." Aliquots of the membrane fractions (30 g of protein) were subjected to SDS-PAGE and immunoblotted with 0.5 g/ml of IgG-HSV-Tag. The filter was exposed to film for 15 s at room temperature. B, aliquots of the membrane fractions were treated at 37°C for 1 h as described under "Experimental Procedures" with one of the following glycosidases: lane 1, none; lane 2, 0.0038 IU of PNGase F; lane 3, 0.25 IU of endoglycosidase H; lane 4, 0.83 IU of neuraminidase. Aliquots of each reaction (30 g of protein) were subjected to SDS-PAGE and immunoblotted with 0.5 mg/ml of IgG-HSV-Tag. The filter was exposed to film for 30 s at room temperature. tion occurs at either of these sites, we created mutants of S2P in which one or both of the asparagines were replaced with glutamine residues. We prepared expression vectors encoding wild-type or mutant S2P with two copies of an epitope tag from an HSV protein at the NH 2 terminus. The plasmids were transiently transfected into human embryonic kidney 293 cells, membrane fractions were prepared, and these membranes were subjected to SDS-PAGE and blotted with an antibody against the HSV epitope. Membranes from mock-transfected 293 cells showed no immunoreactivity with the anti-HSV antibody ( Fig. 2A, lane 1). Epitope-tagged wild-type S2P migrated on the gel as a smear of bands between 45 and 49 kDa (lanes 2 and 6). When asparagine 337 was replaced with glutamine, only the smallest band (ϳ45 kDa) was observed (lane 3). When asparagine 508 was substituted with glutamine, the mobility was indistinguishable from wild-type (lane 4). Double mutants in which both asparagines were replaced with glutamines behaved in the same fashion as the single N337Q mutant (lane 5).
These results indicate that the asparagine at position 337, but not the one at position 508, is glycosylated. We therefore infer that region C projects into the ER lumen.
When extracts containing HSV-tagged wild-type S2P were treated with PNGase F, the mobility of the protein increased on SDS-PAGE, confirming that the protein contained an N-linked carbohydrate chain (Fig. 2B, lane 2). The same change was seen after treatment with endoglycosidase H (lane 3), but not with neuraminidase (lane 4). These data indicate that the carbohydrate chain of S2P has not been modified by ␣-mannosidase II, which is found in the cis-Golgi compartment, or by sialyltransferase, which is found in the trans-Golgi (20).
Glycosylation Is Not Essential for S2P Function-The experiments shown in Fig. 3 were designed to determine whether the function of S2P requires its glycosylation. For this purpose, we used M19 cells, which lack S2P, owing to a deletion in the S2P gene (12). In Fig. 3A, M19 cells were transfected with an expression plasmid containing either no cDNA insert or a cDNA encoding the indicated HSV-tagged wild-type or mutant S2P. We cotransfected a reporter plasmid encoding luciferase driven by a promoter that contains three copies of sterol regulatory element-1 (SRE-1). After incubation in the presence or absence of sterols, the cells were harvested. The measured values for SRE-1 driven luciferase activity were normalized for transfection efficiency by measurement of ␤-galactosidase activity generated by a cotransfected plasmid encoding ␤-galactosidase driven by a sterol-independent promoter. In the absence of cotransfected S2P, luciferase activity was low in the absence or presence of sterols (empty vector). When wild-type pCMV-HSV-S2P was transfected, luciferase activity was high in the absence of sterols (closed bars) and was reduced to basal levels in their presence (open bars). As presented in detail earlier (12) these data indicate that wild-type S2P restores cleavage of SREBPs at Site-2, an event that occurs only after the sterol-regulated cleavage at Site-1. The single and double glycosylation site mutants of HSV-S2P also led to enhanced sterol-regulated transcription of the reporter gene, although this enhancement was somewhat reduced when compared with that produced by wild-type S2P. All proteins encoded by the transfected mutant constructs were expressed at levels similar to that of the wild-type (data not shown).
To further test the activity of the nonglycosylated versions of S2P, we compared the ability of the cDNAs encoding wild-type and glycosylation-negative S2P to restore growth of M19 cells in cholesterol-depleted medium (Fig. 3B). When transfected with empty vector, the M19 cells grew in medium supplemented with cholesterol, mevalonate, and oleate, but they failed to grow in the absence of these ingredients, owing to the deficiency of nuclear SREBPs. When the M19 cells were transfected with cDNAs encoding either wild-type or glycosylationdefective S2P, the cells grew in the absence or presence of these supplements. These results parallel the transcription data in Fig. 3A and indicate that S2P does not require glycosylation in order to restore Site-2 cleavage in M19 cells.
Insertion of Glycosylation Sites into S2P-To determine the membrane orientation of the hydrophilic regions of S2P, we transfected 293 cells with expression plasmids encoding epitope-tagged mutant versions of S2P that contained additional Asn-X-Ser/Thr sequences inserted into these regions (Fig. 4). The cells were harvested, and N-linked glycosylation was detected by treatment with PNGase F followed by SDS-PAGE and immunoblotting with an antibody against the epitope tag. As described above, wild-type HSV-S2P migrated as a cluster of bands between 45 and 49 kDa (Fig. 4, lane 2). Treatment with PNGase F increased the mobility of the pro- M19 cells were transfected on day 1 with 1 g of an expression plasmid containing no cDNA insert (pcDNA 3.1) or the indicated wild-type or mutant HSV-tagged S2P cDNA together with 0.8 g of pSRE-luciferase reporter plasmid and 0.05 g of pCMV-␤ reference plasmid as described under "Experimental Procedures." After incubation for 3 h at 35°C in 3% CO 2 , the cells were switched to medium B in the absence (black bars) or presence (gray bars) of sterols (1 g/ml of 25-hydroxycholesterol plus 10 g/ml of cholesterol). After incubation for 16 h at 37°C, the cells were harvested, and luciferase activity was measured and normalized to ␤-galactosidase activity as described previously (12). Each value represents the average of duplicate transfections. The results shown are representative of two independent transfection experiments. B, cell growth. M19 cells were set up on day 0 at 3.5 ϫ 10 5 /60-mm dish in medium C (a 1:1 mixture of Ham's F-12 medium and Dulbecco's modified Eagle medium containing 100 units/ml penicillin, 100 g/ml streptomycin sulfate, 5% fetal calf serum, 5 g/ml cholesterol, 1 mM sodium mevalonate, and 20 M sodium oleate). On day 1, cells were transfected with the MBS kit using 5 g/dish of the CMV-driven pcDNA3 vector containing the indicated wild-type or mutant version of human S2P cDNA. The cells were washed once with phosphate-buffered saline and refed with medium C 3 h after transfection. On day 2 and every 1-2 days thereafter, the cells were refed with fresh medium C containing 750 g/ml G418. On day 14, cells were divided into two groups and received either medium B (devoid of cholesterol, mevalonate, and oleate) containing 500 g/ml G418 or medium C (with cholesterol, mevalonate, and oleate) containing 500 g/ml G418. Fresh medium B or medium C containing G418 was added every 1-3 days. On day 28, the cells were washed with phosphate-buffered saline, fixed in 95% ethanol, and stained with crystal violet. tein, owing to removal of the N-linked sugar chain at asparagine 337 (lane 3). When a pair of N-linked glycosylation sites was introduced into loop A of S2P (construct A), the mobility of S2P was reduced when compared with wild-type, and this difference was abolished by PNGase F, indicating that one or both of the sites in loop A was glycosylated (Fig. 4, lanes 4 and  5). Similar results were observed when the glycosylation sites were introduced into loop B (lanes 6 and 7). These results suggest that loops A and B face the lumen (see Fig. 1).
To test this model more rigorously, we made a mutant version of HSV-S2P in which Asn-X-Ser/Thr sequences were introduced simultaneously into loops A and B (construct D). This protein migrated slower than any of the proteins containing a single extra carbohydrate chain (Fig. 4, lane 10), indicating that both loops can be glycosylated on the same molecule. This can occur only when both of these loops project into the lumen.
The HEIGH consensus sequence for the putative zinc bind-ing site of S2P is located in a short hydrophilic sequence that interrupts the long hydrophobic sequence located between loops B and C (see Fig. 1). If the HEIGH faced the ER lumen, it might be subject to glycosylation. To test this hypothesis, we inserted one Asn-X-Ser sequence between the isoleucine and glycine of the HEIGH sequence and a second Asn-X-Ser/Thr sequence at a nearby site (construct C in Fig. 4). When this construct was expressed in the transfected 293 cells, the mobility was the same as that of the wild-type protease, and it increased following PNGase F treatment just like the wild-type (Fig. 4, lanes 8 and 9). These data indicate that the inserted asparagines were not glycosylated even though the protein was inserted into the ER in a proper fashion, as shown by the apparently normal glycosylation at asparagine 337. These negative data suggest that the short hydrophilic segment containing the HEIGH sequence does not project into the ER lumen. We cannot rule out the possibility that this sequence faces the ER lumen and is simply too short to be glycosylated. We have drawn this sequence as a dashed line in Fig. 1 to indicate this uncertainty. Fig. 1 predicts that the NH 2 and COOH termini of S2P face the cytosol. To test this hypothesis, we performed a series of protease protection experiments. In preliminary studies, we transfected 293 cells with expression vectors encoding S2P with two copies of the HSV epitope tag either at the NH 2 terminus or the COOH terminus. Sealed membrane vesicles were prepared from these cells and digested with trypsin in the absence or presence of Triton X-100. The proteolytic reaction was stopped, and the proteins were subjected to SDS-PAGE and blotted with an antibody against the HSV epitope tag. The results of these experiments were not interpretable because the tags were protected from trypsin in the presence as well as in the absence of Triton X-100 (data not shown). We hypothesized that the inaccessibility of the epitope to the protease was attributable to the very short sequence between the HSV tag and the NH 2 -and COOH-terminal hydrophobic sequences of S2P. To circumvent this problem, we prepared expression vectors in which the HSV epitope was separated from the NH 2 or COOH terminus by a 50-amino acid spacer derived from the cytoplasmically disposed COOH-terminal domain of the human LDL receptor (Fig. 5A). The trypsin protection assays were then repeated with membrane vesicles from 293 cells transfected with these constructs (Fig. 5B). Vesicles from mock-transfected 293 cells did not show any reactivity with the anti-HSV antibody (Fig. 5B, lanes 1 and 2). In the transfected cells, the Myc-S2P-LDLR-HSV and HSV-LDLR-S2P proteins appeared on the gel as clusters of bands in the region between 50 and 57 kDa (lanes 3 and 11). Trypsin (0.2 g) completely destroyed the epitope tag in the absence of Triton X-100, and the results were similar when Triton X-100 was added (lanes 8 and 16). As a control for the integrity of the membrane vesicles, we blotted duplicate filters with an antibody against an intralumenal ER resident protein BiP (grp78) (21). The anti-BiP antibody is directed against the COOHterminal KSGKDEL sequence (StressGene Biotechnologies Corp.). It also recognizes another ER resident protein, grp94 (21,22). Both of these proteins were resistant to trypsin (1.8 g) in the absence of detergent (lanes 6 and 14). In the presence of detergent, grp94 was completely digested by the same amount of trypsin (lanes 10 and 18). This result confirms that the membrane vesicles were impermeant to trypsin and that they were permeabilized only in the presence of Triton X-100. The grp78 protein was not fully digested in the presence of trypsin, presumably because this protein is intrinsically trypsin-resistant. Three h after transfection, the cells were switched to fresh medium A supplemented with 10% fetal calf serum. After incubation for 20 h at 37°C, membrane fractions were prepared and incubated for 1 h at 37°C in the absence or presence of 0.0038 IU of PNGase F as described in the legend to Fig. 2. Each sample (30 g of protein) was then subjected to SDS-PAGE and immunoblot analysis with 0.5 mg/ml of IgG-HSV-Tag. Filters were exposed to film for 15 s.

Protease Protection Experiments to Map the NH 2 and COOH Termini of S2P-The model in
To confirm that the Myc-S2P-LDLR-HSV and HSV-LDLR-S2P proteins insert correctly into the membrane, we created versions with additional sites for N-linked glycosylation. These proteins, designated Myc-S2P-LDLR-HSV-GLYCO and HSV-LDLR-S2P-GLYCO, are analogous to the proteins that were used in the protease protection experiment, except that they contain a pair of N-linked glycosylation sites in the serine-rich hydrophilic loop (region B). We transfected 293 cells with the four constructs containing the 50-amino acid LDLR spacer, subjected membrane fractions to SDS-PAGE, and blotted with an antibody against the HSV tag (Fig. 5C). Proteins that contained additional sites of N-linked glycosylated (lanes 2 and 4) migrated on the gel more slowly than their counterparts without these sites (lanes 1 and 3), indicating that Myc-S2P-LDLR-HSV-GLYCO and HSV-LDLR-S2P-GLYCO had become glycosylated at the newly introduced sites. These results indicate that introduction of the 50-amino acid LDLR spacer at either the NH 2 or COOH terminus does not alter the overall topology of S2P.
Candidate for Third Zinc-coordinating Residue-Most HEXXH-containing zinc metalloproteases have a third residue that acts together with the two histidines to coordinate the active site zinc (14,23). This residue can be histidine, tyrosine, glutamate, or aspartate. In the prototypic example, thermolysin, the third coordinating residue is a glutamate that is located on a helix that lies adjacent to the helix that contains the HELTH sequence (23). In an attempt to locate the postulated third zinc ligand in S2P, we created point mutations in several FIG. 5. Membrane orientation of NH 2 and COOH termini of S2P as determined by trypsin proteolysis. A, schematic illustration of the fusion proteins used in this experiment. The fusion protein Myc-S2P-LDLR-HSV consists sequentially of two copies of the Myc epitope tag, amino acids 2-519 of S2P, the COOH-terminal 50-amino acids of the human LDL receptor, and two copies of the HSV epitope tag. The fusion protein Myc-S2P-LDLR-HSV-Glyco is an engineered version of Myc-S2P-LDLR-HSV in which two N-linked glycosylation sites have been inserted immediately after serine 121 and serine 131 in the S2P sequence. The fusion protein HSV-LDLR-S2P consists sequentially of two copies of the HSV epitope tag, the COOH-terminal 50 amino acids of the human LDL receptor, and amino acids 2-519 of S2P. The fusion protein HSV-LDLR-S2P-Glyco is an engineered version of HSV-LDLR-S2P in which two N-linked glycosylation sites have been inserted immediately after serine 121 and serine 131 in the S2P sequence. B, 293 cells were transfected with 5 g/dish of control plasmid pcDNA3 (lanes 1 and 2), pCMV-Myc-S2P-LDLR-HSV (lanes 3-10), or pCMV-HSV-LDLR-S2P (lanes [11][12][13][14][15][16][17][18]. Three h after transfection, cells were switched to fresh medium A supplemented with 10% fetal calf serum. After incubation at 37°C for 20 h, cells were harvested, and membrane fractions were prepared as described under "Experimental Procedures." Aliquots of membranes were treated with the indicated amount of trypsin in the absence or presence of 1% Triton X-100 as indicated. After incubation for 30 min at 25°C, reactions were stopped, subjected to SDS-PAGE, and transferred to nitrocellulose. Duplicate filters were immunoblotted with either 0.5 g/ml of IgG-HSV-Tag antibody (upper panel) or 2 g/ml anti-BiP antibody (lower panel) and exposed to film for 5 s (lanes 1-10, upper panel), 45 s (lanes 11-18, upper panel), or 10 s (lanes 1-18, lower panel). C, 293 cells were transfected with 5 g/dish of the indicated plasmid. Three h after transfection, cells were switched to fresh medium A supplemented with 10% fetal calf serum. After incubation for 20 h, cells were harvested, and membrane fractions were prepared as described under "Experimental Procedures." Aliquots of the membrane fractions (30 g of protein) were subjected to SDS-PAGE and immunoblotted with 0.5 g/ml of IgG-HSV-Tag. Lanes 1 and 2 were exposed to film for 30 s, and lanes 3-5 were exposed for 5 s. of the candidate residues that are conserved in the eukaryotic S2P-related proteins (12). The mutated cDNAs were transfected into M19 cells, and the restoration of Site-2 protease activity was assayed with the SRE-luciferase reporter gene described above. As shown in Fig. 6A, all of the tested proteins restored S2P activity, with the single exception of the D467N mutant, which lost all activity. Immunoblotting experiments showed that the D467N protein was expressed at levels similar to wild-type, and it had a similar migration on SDS-PAGE, indicating that it inserted properly into the ER and had become glycosylated (data not shown). In a growth rescue experiment, all of the mutant S2P plasmids except D467N were able to rescue the growth of M19 cells in the absence of cholesterol (Fig. 6B). DISCUSSION The current results support the model presented in Fig. 1 for the membrane topology of S2P. The protease protection assays indicate that the NH 2 and COOH termini of S2P face the cytosol. Potential N-linked glycosylation sites introduced into regions A, B, and C become glycosylated in vivo, suggesting that all three hydrophilic loops project into the lumen. Most of the remaining regions of S2P are hydrophobic and are likely to be associated with the membrane as indicated by the dashed lines in Fig. 1. The only predicted luminal sequence that could not be located unambiguously is the segment of 14 amino acids that immediately precedes the COOH-terminal membranespanning sequence. When a potential N-linked glycosylation site was inserted into this loop, it failed to become glycosylated (data not shown). We believe that this hydrophilic loop is too short to be accessible to glycosyltransferases. We infer that this loop is in the lumen because it immediately precedes the final membrane-spanning sequence, the other end of which is in the cytosol, as revealed by the protease protection experiments.
Sequences resembling S2P were originally noted in the genomes of diverse species, including human, hamster, Drosophila melanogaster, and the Archaea Sulfolobus solfataricus (12). Each of these proteins contains the HEXXH motif, emphasizing the importance of this putative zinc binding site. The length and relative positions of the hydrophobic sequences are conserved, and in each case, the HEXXH sequence occurs in the context of an otherwise hydrophobic segment. The hydrophilic sequences differ, however. The hamster protein contains fewer contiguous serines in the polyserine stretch (17 versus 23 in the human). This number is further reduced in Drosophila to 4. Remarkably, the Sulfolobus sequence lacks both the serinerich region (region B) and the cysteine-rich region (region C) (12). The conservation of sequence from Archaea to humans suggests that the ancestor of S2P played a housekeeping role. Whether mammalian S2P retains any general housekeeping role or whether it is devoted only to SREBP cleavage is unknown. If S2P does have other roles, they are not required for growth of hamster cells in tissue culture. Hamster M19 cells, which lack S2P, grow normally in culture as long as they are supplied with the end products of the SREBP pathway.
Recently, Lewis and Thomas (24) used the Kyte-Doolittle method and other algorithms to identify conserved potential membrane-spanning regions in human and hamster S2P and related sequences from other species that were obtained from data base searches. The computer model predicted that the hydrophobic regions are organized into six helices, each of which completely spans the membrane. In order to place the serine-rich and cysteine-rich loops in the ER lumen, the computer model specified that the NH 2 and COOH termini must both face the lumen (24). This prediction is not fulfilled in the current experiments. Indeed, the protease protection studies indicated that the NH 2 and COOH termini of S2P both face the cytosol. This finding can be explained if some of the hydrophobic sequences dip into the membrane but do not cross it, as shown in Fig. 1. One of these dipping sequences is located between luminal loops A and B, both of which can be glycosylated. If the hydrophobic sequence between them crossed the membrane, then only one of these sequences could be in the lumen. The other dipping sequence is located between loop C and the final membrane-spanning sequence at the COOH terminus. If this sequence spanned the membrane, it would reverse the orientation of the COOH-terminal hydrophobic sequence, which would then place the COOH terminus of the protein in the lumen. Such a topology would be inconsistent with the protease protection data in Fig. 5B.
Among the 26 S2P-related sequences compiled by Lewis and FIG. 6. Activity of wild-type and mutant S2P cDNAs in transfected M19 cells. A, stimulation of SRE-luciferase reporter. B, cell growth. This experiment was carried out using the same protocol as described in the legend to Fig. 3. Thomas (24) is a protein from Bacillus subtilis that they designated SP4G. This appears to be the same protein that is more generally called SpoIVFB (25,26). Lewis and Thomas pointed out that this protein is known to be required for the proteolytic processing of Pro-K , a transcription factor that is activated in mother cells that are producing endospores in response to nutrient deprivation. The cleavage site in Pro-K lies within a hydrophobic segment that may be inserted into a membrane, and expression studies suggest that SpoIVFB is the protease that cleaves this sequence (26). Thus S2P and SpoIVFB are both independently postulated to be proteases that cleave hydrophobic segments of proteins. The other 25 S2P-related proteins have no assigned function (24).
The classic example of an HEXXH-containing zinc endopeptidase is the enzyme thermolysin from Bacillus thermoproteolyticus. The crystal structure of this enzyme was originally deduced in 1972 (27), and it has been studied extensively ever since (23). The catalytic zinc atom is located in the interface between two alpha helices. One helix contains the HEXXH sequence, and the other contains a glutamic acid that also coordinates with the zinc atom. Among the residues that are conserved in all 26 S2P-related proteins is the sequence ⌽DG, where ⌽ is almost always a hydrophobic residue, most frequently leucine. In S2P, the aspartate of the LDG sequence is found at residue 467. In the current studies, we changed this aspartate to asparagine and observed that the protein lost the ability to restore Site-2 cleavage of SREBPs in M19 cells (Fig.  6). This finding, together with the observed conservation of this sequence, strongly suggests that aspartate 467 is the third residue that coordinates the zinc at the active site of S2P. Aspartate 467 is contained within the hydrophobic segment that separates the cysteine-rich region (loop C) and the final transmembrane helix (Fig. 1). If this residue does contribute to the active site, then this hydrophobic segment (indicated by dashed lines in Fig. 1) must be adjacent to the hydrophobic helix that contains the HEIGH sequence. Further experiments will be required to test this hypothesis more fully. 2 With regard to the function of S2P in SREBP processing, two crucial questions remain to be answered. First, is S2P indeed a protease? Second, does it directly cleave the Leu-Cys bond in the transmembrane region of SREBPs? The evidence that S2P is a protease is based upon the retention of the HEXXH sequence and the ⌽DG sequence in species as distant from humans as Sulfolobus and the demonstration that the two histidines and the glutamate, as well as aspartate 467, are all required in order for S2P to carry out its function in facilitating the cleavage of SREBPs (12). The second question is more difficult to answer from available data. Even if S2P is a protease, it is possible that its role is to activate another protease that in turn cuts the Leu-Cys bond in SREBPs. Alternatively, it is possible that S2P cleaves SREBPs at an intermediate site that is between leucine 523 (the site of cleavage by S1P) and leucine 484 (the final cleavage that releases the NH 2 -terminal fragment of SREBPs from membranes). If so, this intermediate cleavage must be required in order for a third protease to cleave the Leu 484 -Cys 485 bond. These questions cannot be resolved until conditions are found in which to assay the proteolytic activity of S2P in vitro.
Further progress requires the establishment of an in vitro assay of S2P function in cleaving SREBP or in activating another protease to do so. This is not a straightforward task in view of the complexity of the natural substrate. This substrate is an intramembranous peptide bond that is not exposed until a prior cleavage has taken place in a short intralumenal loop, i.e. at Site-1. Recreating this substrate in vitro has so far proven difficult, and the difficulty is not lessened by the fact that S2P is a highly hydrophobic protein that requires detergent for solubilization.