Hairpin orientation of sterol regulatory element-binding protein-2 in cell membranes as determined by protease protection.

Sterol regulatory element-binding proteins (SREBP-1 and SREBP-2) are proteins of approximately 1150 amino acids each that are attached to membranes of the endoplasmic reticulum (ER). In sterol-depleted cells, a protease releases an NH2-terminal fragment of approximately 500 amino acids that contains a basic helix-loop-helix leucine zipper motif. This fragment enters the nucleus and stimulates transcription of genes encoding the low density lipoprotein receptor and enzymes of cholesterol biosynthesis. Prior evidence indicates that the SREBPs are attached to membranes by virtue of an 80-residue segment located approximately 80 amino acids to the COOH-terminal side of the leucine zipper. This segment contains two long hydrophobic sequences separated by a short hydrophilic sequence of approximately 30 amino acids. We have proposed a hairpin model in which the two hydrophobic sequences span the membrane, separated by the short hydrophilic sequence which projects into the lumen of the ER (the "lumenal loop"). The model predicts that the NH2- and COOH-terminal segments face the cytosol. To test this model, we constructed a cDNA encoding human SREBP-2 with epitope tags at the NH2 terminus and in the lumenal loop. The COOH-terminal region was visualized with a newly developed monoclonal antibody against this region. Sealed membrane vesicles were isolated from cells expressing the epitope-tagged version of SREBP-2. Trypsin treatment of these vesicles destroyed the NH2- and COOH-terminal segments and reduced the lumenal epitope to a size consistent with protection of the lumenal sequence plus the two membrane-spanning segments. The lumenal epitope tag contained two potential sites for N-linked glycosylation. The size of the trypsin-protected fragment was reduced by treatment with N-Glycanase and endoglycosidase H, indicating that this segment was located in the lumen of the ER where it was glycosylated. These data provide strong support for the hairpin model.

loop-helix leucine zipper (bHLH-Zip) domain that mediates homodimerization and DNA binding. This is followed by a hydrophobic segment of ϳ80 amino acids that attaches the protein to membranes of the endoplasmic reticulum (ER) and a COOH-terminal domain of ϳ550 amino acids whose function has not been assigned.
When cultured cells are depleted of cholesterol, the SREBPs are cleaved by a protease that releases the NH 2 -terminal regions (4). These fragments of ϳ500 amino acids enter the nucleus and bind to a 10-base pair sterol regulatory element (SRE-1) in the promoter of the genes encoding 3-hydroxy-3methylglutaryl-coenzyme A synthase and perhaps other cholesterol biosynthetic enzymes, thereby increasing cholesterol synthesis. They also bind to SRE-1 in the promoter of the gene encoding the low density lipoprotein receptor, thereby increasing cholesterol uptake from plasma lipoproteins. When sterols accumulate in the cell, proteolysis of the SREBPs is suppressed, and the residual nuclear fragments are rapidly degraded by a protease that is sensitive to inhibition by acetyl-leucinal-leucinal-norleucinal (ALLN) (4). As a result, transcription of the SRE-containing genes declines. The fate of the COOH-terminal segments of the SREBPs has not been studied because of the lack of antibodies that react with this fragment.
Cultured cells such as human HeLa cells and hamster fibroblasts produce two SREBPs, designated 1 and 2. The two human proteins are ϳ50% identical to each other, and they share all of the landmark features outlined above (1)(2)(3)5). They bind to the same 10-base pair SRE-1, and they activate transcription of the same genes. The two SREBPs act independently in cultured cells, and there is no evidence that heterodimer formation is required (2). Proteolysis of both proteins is activated in parallel by sterol depletion and inhibited in parallel by overloading with sterols such as 25-hydroxycholesterol (4,5).
Both SREBPs behave biochemically as integral membrane proteins. They are removed from membranes only by detergents and not by treatment with high salt or alkali (3,4). The membrane attachment domain consists of two long hydrophobic segments of at least 20 residues each that are separated by a short hydrophilic sequence of ϳ30 residues. Deletion of this domain markedly reduces the proportion of SREBP-1 that is bound to membranes (3). Based on these observations, we have proposed a hairpin model for the orientation of SREBP in the membrane. The model postulates that the two hydrophobic segments span the membrane bilayer in opposite directions separated by the short hydrophilic sequence of ϳ30 amino acids, which projects into the lumen of the ER or the nuclear envelope. The NH 2 -and COOH-terminal segments both face the cytosol (3,4).
In the current experiments we use the classic method of protease protection (6) to test the hairpin model for the orientation of human SREBP-2 in the membrane. The ability to perform this test is based on two advances: 1) the development of a monoclonal antibody against the COOH-terminal domain of human SREBP-2 and 2) the development of an epitope tag containing sites for N-linked glycosylation that can be inserted into the lumenal loop. The results are consistent with the hairpin model.

EXPERIMENTAL PROCEDURES
Materials and Methods-Standard molecular biology techniques were used (7). DNA sequencing was performed with the dideoxy chain termination method on an Applied Biosystems model 373A DNA sequencer. Newborn calf lipoprotein-deficient serum (d Ͼ 1.215 g/ml) was prepared by ultracentrifugation (8). Oligonucleotides were synthesized on a model 380A DNA synthesizer (Applied Biosystems, Inc.). Oligonucleotide-directed mutagenesis was carried out with a Muta-Gene phagemid in vitro mutagenesis version 2 kit (Bio-Rad) as described by Kunkel (9). We obtained trypsin (catalog number L503740) from Worthington; soybean trypsin inhibitor (catalog number 650357) from Calbiochem; anti-BiP monoclonal antibody from StressGen Biotechnologies (Victoria, British Columbia, Canada); N-Glycanase from Genzyme; endoglycosidase H, neuraminidase, Klenow fragment, and restriction enzymes from New England Biolabs; HSV-Tag™ monoclonal antibody from Novagen; Takara DNA ligation kit from Panvera Corp.; and Pfu DNA polymerase from Stratagene. All plasmid DNAs for transfection were prepared with Plasmid Mega or Maxi kits (Qiagen). Other reagents were obtained from sources as reported previously (4).
Construction of Plasmid pTK-HSV-BP2-7D4 -This plasmid was constructed from pTK-HSV-BP2 (the parent plasmid) 2 from pcDNA3 (Invitrogen), pTK␤ (Clontech), and human pSREBP-2 (Ref. 2). Briefly, the parent plasmid contains a herpes simplex virus (HSV) thymidine kinase promoter in front of two tandem copies of the HSV epitope tag (QPELAPEDPED) (10) fused to the coding sequence of human SREBP-2 (amino acids 14 -1141). To construct pTK-HSV-BP2-7D4, we inserted a cDNA segment encoding the 7D4 epitope into the loop region of the SREBP-2 sequence in the parent plasmid. First, we used oligonucleotide site-directed mutagenesis (9) to replace amino acids 505-513 in the loop region of human SREBP-2 in pTK-HSV-BP2 with the epitope YPYDVPDYA derived from amino acids 98 -106 of influenza hemagglutinin (11). The hemagglutinin epitope contains a unique BsiWI site in the codons for amino acids PYD. The nucleotide sequence encoding the 7D4 epitope (residues 33-250 of hamster SREBP-2 (Ref. 5) was amplified by polymerase chain reaction with a pair of primers containing a BsiWI site at each 5Ј end using Pfu DNA polymerase under the following conditions: initial temperature of 80°C, followed by 94°C for 3 min, followed by 20 cycles of 94°C for 1 min, 60°C for 1 min, and 72°C for 2 min. The amplified product was digested with BsiWI and cloned into the unique BsiWI site in the loop region (see Fig. 5). The resulting construct was designated pTK-HSV-BP2-7D4. The sequence of the cloning junctions was confirmed by sequencing. To protect against polymerase chain reaction artifacts, two independent clones were used in each of the transfections and proteolysis experiments, and the results were always consistent.
Monoclonal Antibodies-Monoclonal antibody IgG-7D4 against amino acids 32-250 of hamster SREBP-2 has been described previously (12). This antibody recognizes hamster, but not human, SREBP-2. To produce an antibody against the NH 2 -terminal region of human SREBP-2, we produced in Escherichia coli a fusion protein containing amino acids 48 -403 of human SREBP-2 (Ref. 2) with six histidines following the initiator methionine. The protein was injected into Balb/C mice, and spleen cells were used (13) to produce a monoclonal antibody, designated IgG-3H8. A fusion protein containing six histidines after the initiator methionine followed by amino acids 833-1141 of human SREBP-2 (Ref. 2), followed by a Flag™ octapeptide epitope (14), was expressed in E. coli and used to immunize mice to produce a monoclonal antibody, designated IgG-1C6. Prior to immunization, both fusion proteins were purified by Ni ϩ -Sepharose chromatography. All monoclonal antibodies were purified from ascites fluid by protein G chromatography.
Cell Culture, Transfection, and Cell Fractionation-Human HeLa S3 cells were grown in spinner culture as described (4). On day 0, cells were set up in replicate spinner cultures at a density of 2.5 ϫ 10 5 cells/ml in Joklik's minimum essential medium containing 100 units/ml penicillin, 100 g/ml streptomycin sulfate, and 2.5% (v/v) newborn calf serum. On day 1, sterols dissolved in ethanol were added to each flask. The final concentration of ethanol in the medium was 0.2% (v/v). After incubation for 0.5-6 h, cells were harvested and fractionated into a nuclear extract and membrane fraction (10 5 ϫ g pellet) as described previously for HeLa cells (4) with one modification: the extract from the crude nuclear pellet was centrifuged at 55,000 rpm for 30 min at 4°C in a Beckman TLA100.2 rotor, and the supernatant from this spin was used. For proteolysis experiments, the membrane fraction was washed once with buffer A (10 mM Hepes-KOH at pH 7.4, 10 mM KCl, 1.5 mM MgCl 2 , 0.5 mM sodium EDTA, 0.5 mM sodium EGTA, and 100 mM NaCl) and then resuspended in buffer A.
Monolayers of human embryonic kidney 293 cells (3) 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 Dulbecco's modified Eagle's medium supplemented with 100 units/ml penicillin, 100 g/ml streptomycin, and 10% (v/v) fetal calf serum. On day 2, the cells were transfected with 5 g of pTK-HSV-BP2-7D4 using a modified bovine serum transfection kit (Stratagene) according to the manufacturer's instructions, except that the monolayers were not washed prior to addition of 6% (v/v) modified bovine serum in Dulbecco's modified Eagle's medium. The transfection was carried out at 35°C for 3 h in 3% CO 2 , after which each dish of cells was washed once with 5 ml of phosphate-buffered saline and switched to 5 ml of Dulbecco's modified Eagle's medium supplemented with penicillin, streptomycin, 10% (v/v) calf lipoprotein-deficient serum, 1 g/ml 25-hydroxycholesterol, 10 g/ml cholesterol, 50 M compactin, and 50 M sodium mevalonate. On day 3 (20 h later), 5 l of solution containing 25 mg/ml ALLN was added to each dish, and the cells were harvested 3 h later. To prepare the membrane fraction, the cells from 24 -32 dishes were pooled and allowed to swell at 4°C for 10 min in buffer B (buffer A without NaCl and supplemented with 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin, 25 g/ml ALLN, and 1 mM dithiothreitol) and then homogenized with 12 strokes of a tight pestle in a Dounce homogenizer. The suspension was pelleted at 1000 ϫ g for 10 min at 4°C, and the supernatant was centrifuged at 55,000 rpm for 30 min at 4°C in a TLA100.2 rotor to obtain the membrane fraction (10 5 ϫ g pellet). The pellets were washed once with buffer A and resuspended in buffer A for use in trypsin proteolysis experiments.
Trypsin Proteolysis and Glycosidase Sensitivity of Membrane-bound SREBP-2-For proteolysis sensitivity experiments, aliquots of the 10 5 ϫ g membrane fraction of HeLa cells (0.3 mg of protein in 60 l of buffer A) were centrifuged at 10 5 ϫ g at 4°C for 30 min. The resulting pellet was washed with 0.2 ml of buffer A and resuspended in 110 l of the same buffer in the absence or presence of 1% (v/v) Triton X-100. The samples were incubated at room temperature in a model REAX 2000 mixer (Whatman LabSales) with mild shaking for 30 min. Varying amounts of trypsin were then added to each tube in 3 l of Dulbecco's phosphate-buffered saline (catalog number 14190 -144, Life Technologies, Inc.). After incubation at room temperature for 30 min, 300 units of soybean trypsin inhibitor in 6 l of phosphate-buffered saline were added to stop proteolysis, after which 40 l of the suspension was subjected to SDS/8% PAGE, followed by immunoblot analysis.
For proteolysis of 293 cell membranes, aliquots of the 10 5 ϫ g membrane fraction (0.1 mg of protein in 40 l of buffer A) were incubated at room temperature in a model REAX 2000 mixer with mild shaking for 30 min in the absence or presence of 1% Triton X-100. Trypsin and soybean trypsin inhibitor were added as described above, followed by SDS-PAGE and immunoblotting.
To determine the glycosidase sensitivity of the trypsin-resistant fragment, aliquots of the 10 5 ϫ g membrane fraction of transfected 293 cells (0.1 mg of protein in 40 l of buffer A) were treated sequentially as follows: 1) incubation with 0.3 unit of trypsin (added in 3 l) at room temperature for 30 min; 2) addition of 150 units of soybean trypsin inhibitor (added in 3 l); 3) addition of Triton X-100 (final concentration, 1%); 4) boiling for 5 min in the presence (N-Glycanase and endo H reactions) or absence (neuraminidase reaction) of SDS (final concentration, 0.5%) and ␤-mercaptoethanol (final concentration, 0.1%); and 5) addition of the indicated glycosidase (added in 4 l) and incubation at 37°C for 2 h as described in the legend to Fig. 8. Each sample (final volume, 59 l) was mixed with 10 l of 5 ϫ SDS loading buffer (15) and used for SDS-PAGE and immunoblot analysis.
Immunoblot Analysis-Samples were mixed with 5 ϫ SDS loading buffer (15) prior to SDS-PAGE on 8% gels. After electrophoresis, the proteins were transferred to Hybond-C extra nitrocellulose sheets (Amersham Corp.). After incubation with monoclonal antibodies at the concentrations indicated, the sheets were washed, and the antibodies were detected with horseradish peroxidase-conjugated anti-mouse IgG using the enhanced chemiluminescence (ECL) Western blotting detection system kit (Amersham) according to the instructions of the manufacturer with modification as described (4). Gels were calibrated with prestained molecular weight markers (New England Biolabs). Filters were exposed at room temperature to reflection film (DuPont NEN) for the indicated time. Fig. 1 shows hydrophobicity plots for human SREBP-1 and -2 according to the method of Kyte and Doolittle (16) with a window of 17 residues. The two hydrophobic segments, designated 1 and 2, are clearly visible, as is the short hydrophilic segment between them. These segments are located ϳ80 residues to the COOH-terminal side of the bHLH-Zip region, which is underlined in Fig. 1.

RESULTS
To compare the fates of the NH 2 -and COOH-terminal segments of SREBP-2, we took advantage of a newly developed monoclonal antibody, designated IgG-1C6, that was raised against a fragment of human SREBP-2 extending from residue 833 to the COOH terminus. In the experiment of Fig. 2, HeLa cells were partially deprived of sterols by growth in the presence of a low concentration of newborn calf serum (2.5%). We then overloaded the cells with sterols by adding a mixture of 25-hydroxycholesterol and cholesterol to the culture medium. At varying times, cells were harvested, and high salt nuclear extracts and 10 5 ϫ g membrane pellets were subjected to SDS-PAGE and blotted with a monoclonal antibody directed against the NH 2 terminus and with the COOH-terminal monoclonal IgG-1C6. At zero time, when the cells were in a sterol-depleted state, both antibodies stained the full-length precursor form of SREBP-2, which was found in the membrane fraction (designated P in Fig. 2, lane F). As expected, the NH 2 -terminal antibody also stained a fragment whose migration corresponded to a molecular mass of 68 kDa, which was found in the nuclear extracts (lane A). The NH 2 -terminal fragment was shown previously to migrate with anomalous slowness, owing to the acidic NH 2 -terminal activation domain (3). The COOHterminal antibody stained a fragment of ϳ65 kDa that was bound to the membranes (lane F). Two h after sterol addition, the NH 2 -terminal fragment had almost disappeared from the nucleus (lane D). The COOH-terminal fragment was also rapidly degraded. By 2 h, it had been reduced by more than 80% (lane I), but a small amount continued to be detectable at 6 h (lane J). No COOH-terminal fragment was detected in immunoblots of the 10 5 ϫ g supernatant fraction (data not shown).
The COOH-terminal fragment of SREBP-2 behaved as an integral membrane protein, just like the full-length precursor form (4). In the experiment of Fig. 3, membranes isolated from sterol-deprived cells were treated with various protein-solubi-lizing agents and subjected to centrifugation at 10 5 ϫ g. The pellets and supernatants were subjected to SDS-PAGE and immunoblotted with IgG-1C6, which visualized both the precursor form and the COOH-terminal fragment of SREBP-2. The precursor and the COOH-terminal fragment were both solubilized partially with 1% SDS and 1% Triton X-100. Neither was solubilized with buffer alone, nor with 0.1 M Na 2 CO 3 or 1 M hydroxylamine.
To determine the orientation of the COOH-terminal fragment with respect to the membrane, we homogenized HeLa cells gently with a Dounce homogenizer and isolated mem- On day 0, HeLa cells were divided into five replicate spinner cultures, each containing 1.2 ϫ 10 8 cells in 500 ml. On day 1, each culture received a mixture of sterols (1 g/ml 25-hydroxycholesterol plus 10 g/ml cholesterol) for the indicated time at 37°C. Staggered additions were made so that the cells were harvested at the same time. A high salt nuclear extract and a membrane fraction were prepared as described under "Experimental Procedures," and an aliquot of each fraction (50 g of protein) was subjected to SDS-PAGE and immunoblot analysis. Duplicate filters were blotted with 5 g/ml anti-NH 2 -terminal IgG-3H8 (upper panel) or 5 g/ml anti-COOH-terminal IgG-1C6 (lower panel). The gels were calibrated with prestained protein standards (New England Biolabs). The filters were exposed to film for 60 s (upper panel) or 80 s (lower panel). P, N, and C denote the precursor form and the NH 2terminal and COOH-terminal fragments of SREBP-2, respectively.  1-5). Each pellet was mixed with 160 l of lysis buffer (22) containing 1% SDS, and an aliquot (32 l) was mixed with 5 ϫ SDS loading buffer (lanes 6 -10). After boiling for 5 min, the samples were subjected to SDS-PAGE, and immunoblot analysis was carried out with 5 g/ml anti-COOH-terminal IgG-1C6. The filter was exposed to film for 30 s. P and C denote the precursor form and the COOH-terminal fragment of SREBP-2, respectively. brane vesicles by centrifugation (Fig. 4). The vesicles were incubated with varying concentrations of trypsin in the absence or presence of Triton X-100 followed by SDS-PAGE and immunoblotting with IgG-1C6. Trypsin at a concentration of 1.7 units/ml obliterated the IgG-1C6 epitope, whether it was present on the precursor or on the COOH-terminal fragment (Fig.  4A, upper panel, lane 7). Disruption of the membranes with Triton X-100 increased trypsin sensitivity only slightly (lower panel). These data indicate that the COOH-terminal fragment of SREBP-2 is exposed to the cytoplasmic face of the membrane. As a control for the intactness of the membrane vesicles, we immunoblotted the digested membranes with a monoclonal antibody against chaperone proteins, called BiPs, that are known to reside in the lumen of the ER (17). The antibody visualizes two BiPs, designated BiP94 and BiP78. BiP94 was completely resistant to trypsin in the absence of Triton X-100 (Fig. 4B, upper panel), and it was readily destroyed when Triton X-100 was present (lower panel). This observation confirms that the membrane vesicles were sealed. BiP78 was not informative because the protein was resistant to trypsin either in the presence or absence of Triton X-100, apparently because of an intrinsic trypsin resistance.
We next sought to determine the membrane orientation of the hydrophilic loop between the two putative transmembrane segments. Multiple attempts to raise monoclonal or polyclonal antibodies against this short 30-residue sequence failed. We then used recombinant DNA techniques to insert DNA segments encoding short epitope "tags" into this segment followed by expression in transfected cells. We tried DNA segments encoding epitopes for which monoclonal antibodies are commercially available. These included Myc (18), influenza hemagglutinin (11), T7 gene 10 leader peptide (19), and the Flag™ epitope (14). All of these attempts failed. Although the monoclonal antibodies recognized these epitopes when inserted into other proteins or at other locations in SREBP-2, they did not recognize the epitope when inserted into the 30-residue hydrophilic loop, even after the protein had been denatured by SDS-PAGE.
In a final attempt to circumvent this problem, we decided to insert a DNA sequence encoding a much longer segment of protein into the hydrophilic loop. We chose a 218-amino acid segment to which we already had a potent monoclonal antibody. This segment consisted of amino acids 33-250 of hamster SREBP-2 (Ref. 5), and the monoclonal antibody that recognizes it is IgG-7D4 (Ref. 12). The antibody is specific for the hamster sequence, and it does not recognize human SREBP-2. When this hamster sequence is inserted into human SREBP-2, it constitutes a novel epitope. It should also be noted that this segment is part of the NH 2 -terminal region of hamster SREBP-2 that is normally located on the cytoplasmic side of the membrane. Fig. 5 shows the plasmid that we constructed, which is designated pTK-HSV-BP2-7D4. Fig. 6A diagrams the resultant protein. The protein consists of human SREBP-2 with two epitope tags. The first is a short epitope tag from the HSV glycoprotein that is inserted at the NH 2 terminus of the protein. The second is the 33-250-amino acid segment of hamster SREBP-2 that is inserted into the loop between the two transmembrane regions. By chance, this sequence contains two potential N-linked glycosylation sites (Asn-X-Ser or -Thr) that are underlined in Fig. 5. Expression is driven by a promoter from the HSV thymidine kinase.
To confirm the specificity of IgG-7D4, we transfected human 293 cells with pTK-HSV-BP2-7D4 or with a control plasmid (pTK-HSV-BP2) that contains the NH 2 -terminal HSV tag but lacks the 7D4 epitope (Fig. 6). Membranes from the transfected cells were subjected to SDS-PAGE and immunoblotted with an antibody against the HSV tag or with IgG-7D4. Neither antibody visualized any protein in mock-transfected cells (lanes 1  and 4). and lanes 10 -18, respectively. After incubation for 30 min at room temperature, each sample received 300 units of soybean trypsin inhibitor (added in 6 l), after which a 40 l-aliquot from each sample was subjected to SDS-PAGE and immunoblot analysis with 5 g/ml anti-COOH-terminal IgG-1C6 (A). A duplicate filter was blotted with 5 g/ml anti-BiP antibody (B). The filters were exposed to film for 10 min (A) or 10 s (B). P and C in A denote the precursor and COOH-terminal forms of SREBP-2, respectively. B94 and B78 in B denote BiP-94 and BiP-78, respectively.
FIG . 5. Map of pTK-HSV-BP2-7D4. This expression vector was constructed as described under "Experimental Procedures." Expression is driven by the HSV thymidine kinase promoter (pTK). The plasmid encodes a 1378-amino acid fusion protein consisting, from NH 2 terminus to COOH terminus, of the following: an initiator methionine; two tandem copies of the 11-amino acid HSV epitope tag (QPELAPEDPED); six amino acids (IDGTVP) encoded by BspDI and KpnI sites; amino acids 14 -504 from human SREBP-2; four amino acids (YPYD) from the influenza hemagglutinin (HA) epitope; amino acids 33-250 from hamster SREBP-2 including the 7D4 epitope; eight amino acids (PYDVP-DYA) from the HA epitope; and amino acids 514-1141 from human SREBP-2. BGH pA denotes the bovine growth hormone polyadenylation sequence. Two putative N-linked glycosylation signals in the 7D4 epitope are underlined.
HSV-BP2-7D4, both antibodies visualized a protein that migrated with an apparent molecular mass of 175 kDa (lanes 3 and 6), which is much larger than native SREBP-2, owing to the presence of the long 7D4 epitope. Fig. 7 shows a trypsin protection experiment similar to that of Fig. 4, but performed with membrane vesicles from 293 cells that were transfected with pTK-HSV-BP2-7D4. In the absence of trypsin, the membranes contained a protein of 175 kDa that was visualized by the anti-HSV antibody which reacts with the NH 2 -terminal epitope tag (upper panel), IgG-1C6 which reacts with the COOH-terminal segment (middle panel), and IgG-7D4 which reacts with the epitope inserted into the putative lumenal loop (lower panel). Treatment with trypsin abolished both the NH 2 -and COOH-terminal segments (lanes 2-4). The 7D4 epitope was preserved on a fragment of ϳ47 kDa that resisted trypsin at a concentration that was 10-fold above the concentration that destroyed the other two epitopes (lane 4). When the digestion was conducted in the presence of Triton X-100, the NH 2 -and COOH-terminal segments were destroyed as before, and the 7D4 epitope became trypsin-sensitive. These results are consistent with the hairpin model for the orientation of SREBP-2, in which the NH 2 terminus and COOH terminus are exposed to the cytosol, and the hydrophilic loop between the transmembrane segments projects into the ER lumen.
The apparent size of the trypsin-resistant fragment on SDS-PAGE, as visualized with IgG-7D4 (47 kDa) was greater than would be predicted if the protected fragment consisted of the 218 amino acid peptide containing the epitope plus the remaining lumenal amino acids and the two transmembrane regions (ϳ35 kDa). We suspected that this might be attributable to glycosylation of the epitope tag at one or both of the N-linked sites shown in Fig. 5. To test this possibility, we digested intact membrane vesicles with trypsin as before, then solubilized the membranes and denatured the proteins by treatment with detergent and boiling. The trypsin-resistant fragment was then digested with one of three glycosidases, and its size was estimated by SDS-PAGE and immunoblotting with IgG-7D4. As shown in Fig. 8, in the absence of glycosidases the trypsinresistant fragment migrated as a 47-kDa protein. The apparent molecular mass was reduced to about 35 kDa by treatment with N-Glycanase and endo H, but not neuraminidase. This pattern of glycosidase sensitivity indicates that the N-linked sugars remained in their high mannose unprocessed forms and that the protein had not been transported to the Golgi complex (20). DISCUSSION The current manuscript provides evidence to support the hairpin model for the orientation of SREBP-2 in the membrane. According to this model, the NH 2 -and COOH-terminal regions of SREBP-2 are oriented toward the cytosol. They are separated by a membrane attachment domain that consists of two membrane-spanning segments separated by a short loop that projects into the lumen of the ER.
The evidence in support of this model comes from trypsin sensitivity experiments performed with membrane vesicles. The NH 2 -and COOH-terminal regions were readily destroyed by trypsin, and the lumenal loop epitope was reduced to a size consistent with protection of the glycosylated loop plus the two  4 -6). The filters were exposed to film for 2 min (lanes 1-3) or 3 min (lanes 4 -6).

FIG. 7.
Resistance of epitope-tagged loop region of SREBP-2 to trypsin proteolysis. 293 cells were transfected with pTK-HSV-BP2-7D4, the 10 5 ϫ g membrane fractions were prepared, and aliquots of these membranes (0.1 mg of protein) were resuspended in 40 l of buffer A in the absence (lanes 1-4) or presence (lanes 5-8) of 1% Triton X-100 as described under "Experimental Procedures." After incubation at room temperature for 30 min, each sample received increasing amounts of trypsin. The final trypsin concentrations (units/ml) were as follows: lanes 1 and 5, 0; lanes 2 and 6, 2.5; lanes 3 and 7, 7.5; and lanes 4 and 8, 25. Triplicate samples for each reaction were incubated and processed for SDS-PAGE as described in the legend to Fig. 4. Immunoblot analysis was carried out with one of the following antibodies: upper panel, 0.5 g/ml anti-NH 2 -terminal antibody (HSV-Tag™ antibody); middle panel, 10 g/ml anti-COOH-terminal antibody (IgG-1C6); and lower panel, 8 g/ml of IgG-7D4. The filters were exposed to film for 10 s (upper and lower panels) or 3 min (middle panel). Arrows denote the position of migration of the endogenous and transfected precursor (P) forms of SREBP-2 and of the trypsin-resistant fragment.
membrane-spanning segments. The addition of N-linked sugars to the loop was confirmed by the observation that the protected fragment was reduced in size by treatment with N-Glycanase and endo H.
In order to visualize the lumenal loop in immunoblots, we had to resort to the unorthodox procedure of inserting a long 218-residue segment of protein containing an epitope. This was necessary because short epitopes inserted into this region failed to react with their cognate monoclonal antibodies even after the protein was denatured by SDS-PAGE. It seems likely that this region of the protein must refold during transfer to nitrocellulose, perhaps through hydrophobic interactions between the two membrane-spanning segments, thereby occluding the lumenal epitope. When a long protein segment was inserted, the epitope was no longer occluded.
It is possible that the insertion of such a long epitope changed the orientation of the SREBP in the membrane. We believe that this is unlikely for two reasons: 1) the NH 2 -terminal fragment and COOH-terminal fragment were trypsin-sensitive in the native protein as well as in the protein bearing the epitope tag (Figs. 4 and 7); and 2) the sequence that we inserted does not have lumenal targeting properties, since it is normally on the cytoplasmic side of the membrane.
The current studies also provide the first glimpse of the fate of the COOH-terminal segment of SREBP-2 after the NH 2terminal domain has been released by the sterol-regulated protease. The data suggest that this fragment remains attached to membranes as an integral protein. When sterols are added, most of the COOH-terminal fragment is degraded within 2 h, but a small amount remains detectable for as long as 6 h. We do not know whether this represents a slowly degraded pool or whether it represents a small amount of COOH-terminal fragment that is continuously produced, even in the presence of sterols.
Although the current study defines the orientation of the two transmembrane segments and the lumenal loop, the data do not make any statement about the mechanism by which more distal elements of the COOH-terminal segment interact with the membrane. This region has some affinity for membranes since a fraction of SREBP-2 remains associated with membranes even when the transmembrane segments are deleted (3). Inspection of the hydrophobicity plots in Fig. 1 reveals several moderately hydrophobic sequences in the COOH-terminal region. However, none of these is long enough nor of sufficient hydrophobicity to meet the criteria (16) of true transmembrane segments. It is possible that some of these segments dip into the lipid bilayer, but we do not believe that any of them span the membrane.
Although all of the current experiments were performed with SREBP-2, we believe that the results apply to SREBP-1, since the hydrophobicity profiles of the two proteins are nearly identical (Fig. 1) and because proteolysis of the two proteins is regulated in parallel (4,5,12). Knowledge of the membrane orientation of these proteins is essential if we are to understand the site at which they are cleaved by the sterol-regulated protease and the mechanism of its regulation. FIG. 8. Sensitivity of the protease-resistant, epitope-tagged loop region of SREBP-2 to treatment with glycosidases. Aliquots of the 10 5 ϫ g membrane fraction from pTK-HSV-BP2-7D4 transfected 293 cells (0.1 mg of protein) were digested with trypsin at room temperature for 30 min as described under "Experimental Procedures." After addition of soybean trypsin inhibitor, the membrane fractions were treated with detergent, boiled for 5 min, and then incubated for 2 h at 37°C with one of the following glycosidases: lane 1, none; lane 2, 1.25 IU of N-Glycanase; lane 3, 2.5 IU of endo H; and lane 4, 0.75 IU of neuraminidase. The samples were processed for SDS-PAGE and immunoblot analysis with 8 g/ml IgG-7D4. The filter was exposed to film for 1 min. The arrow denotes trypsin-resistant fragment, and the asterisk denotes the deglycosylated form of the trypsin-resistant fragment.