Ubiquitin is conjugated by membrane ubiquitin ligase to three sites, including the N terminus, in transmembrane region of mammalian 3-hydroxy-3-methylglutaryl coenzyme A reductase: implications for sterol-regulated enzyme degradation.

The stability of the endoplasmic reticulum (ER) glycoprotein 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMGR), the key enzyme in cholesterol biosynthesis, is negatively regulated by sterols. HMGR is anchored in the ER via its N-terminal region, which spans the membrane eight times and contains a sterol-sensing domain. We have previously established that degradation of mammalian HMGR is mediated by the ubiquitin-proteasome system (Ravid, T., Doolman, R., Avner, R., Harats, D., and Roitelman, J. (2000) J. Biol. Chem. 275, 35840-35847). Here we expressed in HEK-293 cells an HA-tagged-truncated version of HMGR that encompasses all eight transmembrane spans (350 N-terminal residues). Similar to endogenous HMGR, degradation of this HMG(350)-3HA protein was accelerated by sterols, validating it as a model to study HMGR turnover. The degradation of HMG(240)-3HA, which lacks the last two transmembrane spans yet retains an intact sterol-sensing domain, was no longer accelerated by sterols. Using HMG(350)-3HA, we demonstrate that transmembrane region of HMGR is ubiquitinated in a sterol-regulated fashion. Through site-directed Lys --> Arg mutagenesis, we pinpoint Lys(248) and Lys(89) as the internal lysines for ubiquitin attachment, with Lys(248) serving as the major acceptor site for polyubiquitination. Moreover, the data indicate that the N terminus is also ubiquitinated. The degradation rates of the Lys --> Arg mutants correlates with their level of ubiquitination. Notably, lysine-less HMG(350)-3HA is degraded faster than wild-type protein, suggesting that lysines other than Lys(89) and Lys(248) attenuate ubiquitination at the latter residues. The ATP-dependent ubiquitination of HMGR in isolated microsomes requires E1 as the sole cytosolic protein, indicating that ER-bound E2 and E3 enzymes catalyze this modification. Polyubiquitination of HMGR is correlated with its extraction from the ER membrane, a process likely to be assisted by cytosolic p97/VCP/Cdc48p-Ufd1-Npl4 complex, as only ubiquitinated HMGR pulls down p97.

The enzyme 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMGR) 1 catalyzes the rate-limiting production of mevalonate, the committed precursor for the biosynthesis of sterols and a myriad of essential nonsterol isoprenoids. The intracellular levels of HMGR are regulated by the cellular needs for sterol and nonsterol metabolites. This regulation involves changes in the transcription of the HMGR gene and, at the post-translational level, alteration of enzyme stability (1)(2)(3). Thus, when demands for sterols are high, HMGR gene is transcribed at a high rate, and the resulting HMGR protein is relatively stable. When the requirements for mevalonate-derived metabolites have been satisfied, transcription ceases, and the enzyme is rapidly degraded (1)(2)(3).
HMGR gene transcription is controlled by specific factors, designated sterol regulatory element-binding proteins (SREBPs), which bind to sterol regulatory elements in the promoter of the gene and activate transcription (1,4). The nuclear SREBPs are derived from the N terminus of large endoplasmic reticulum (ER) membrane-bound precursors through sequential cleavage by two Golgi proteases (5). In sterol-depleted cells, the membrane SREBP precursors are transported by vesicles to the Golgi where cleavage occurs. In this journey, the SREBP precursors are escorted by SREBP cleavage-activating protein (SCAP), a polytopic ER protein with a sterol-sensing domain (SSD) in its membrane region, which associates with the C terminus of SREBPs precursors soon after their synthesis (6 -8). In sterol-replete cells, the SCAP⅐SREBP complex is retained in the ER through interaction of SCAP's SSD with resident ER membrane proteins, Insig-1 and Insig-2. This sterol-stimulated binding of SCAP to Insigs prevents SCAP⅐SREBP complex from reaching the compartment where the SREBP-cleaving proteases reside (9,10). Mutations in the SSD of SCAP disrupt its interaction with Insigs and allow the transport of the SCAP⅐SREBP complex from the ER to the Golgi even in the presence of sterols (11).
Mammalian HMGR is a "high mannose" resident glycoprotein of the ER (12,13). The enzyme is composed of four 97-kDa subunits, each of which can be divided into a highly conserved C-terminal domain that faces the cytoplasm and constitute the catalytic site (14), and an N-terminal hydrophobic domain that spans the ER membrane eight times (15) and bears the single * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: Institute of Lipid and Atherosclerosis Research, Sheba Medical Center, Tel Hashomer 52621, Israel. Tel.: 972-3-5302124; Fax: 972-3-5343521; E-mail: roitelma@post.tau.ac.il. N-glycan (12). This transmembrane region is dispensable for the enzymatic activity (16) but operates as an independent "degron" that is necessary and sufficient to confer metabolically regulated stability onto HMGR as well as onto a variety of heterologous reporter proteins such as ␤-galactosidase in HM-Gal (17). This transmembrane region of HMGR also contains an SSD that shares considerable homology with the SSD of SCAP (11).
In eukaryotic cells, the majority of the short-lived proteins are degraded by the ubiquitin proteasome pathway, which selectively conjugates polyubiquitin chain(s) to these proteins and marks them for degradation by the 26 S proteasome. Ubiquitination of specific proteins is executed by distinct combinations of ubiquitin-conjugating enzymes (E2s) and ubiquitinprotein ligases (E3s). Initially, these enzymes attach the 8-kDa ubiquitin molecule to lysine residue(s) in the target protein, forming an isopeptide bond between the carboxyl group of the C-terminal glycine of ubiquitin and the ⑀-NH 2 of that lysine in the target protein (18,19). Polyubiquitin chains are subsequently formed through the processive addition of ubiquitin molecules such that the C terminus of each added ubiquitin unit is linked to a specific lysine residue (usually Lys 48 ) in the previous ubiquitin moiety. However, it remains unclear whether and how the E2/E3 enzymes select specific lysine(s) to which the first ubiquitin in the polyubiquitin chain is conjugated.
In addition to its role in the degradation of proteins in the cytosol and nucleoplasm, the ubiquitin-proteasome system also participates in the ER-associated degradation (ERAD) of misfolded or misassembled membrane and luminal proteins that fail to pass the scrutiny of the quality control of the ER (20). Such aberrant proteins are dislocated back to the cytosol in a process that requires the activity of the cytosolic AAA-ATPase p97/VCP/Cdc48p-Ufd1-Npl4 complex and are degraded via the ubiquitin-proteasome pathway (21)(22)(23)(24). The dislocation of ERAD substrates is linked to their polyubiquitination (22,(25)(26)(27), consistent with the ability of p97/VCP/Cdc48p as well as its partner Ufd1-Npl4 complex to bind polyubiquitin chains (28).
We have previously established that the degradation of mammalian HMGR is mediated by the ubiquitin-proteasome system in what appears to be the metabolically regulated routing of HMGR into the constitutive ERAD pathway (29). Since both HMGR and HMGal are polyubiquitinated under conditions that accelerate HMGR degradation, we hypothesized that ubiquitination occurs in their common transmembrane region. Here, we demonstrate that, indeed, HMGR is ubiquitinated at its membrane region and pinpoint lysines 89 and 248 as the polyubiquitin acceptor sites. Moreover, our data indicate that the N terminus of HMGR is also ubiquitinated. Polyubiquitination of HMGR appears to be carried out by ER E2 and E3 enzymes and facilitates the extraction of this polytopic protein from the membranes, probably with the aid of the cytosolic p97/VCP.

EXPERIMENTAL PROCEDURES
Materials-Oligonucleotide primers were synthesized by Sigma-Genosys. Geneticin and LipofectAMINE Plus reagent were obtained from Invitrogen Life Technologies. 25-Hydroxycholesterol was purchased from Steraloids and immobilized recombinant Protein A was obtained from RepliGen. MicroBCA protein reagent and Super Signal West Pico chemiluminescent substrate were from Pierce. MG-132 was purchased from Calbiochem. Biotinylated ubiquitin (cat. UW8705) and purified rabbit E1 were obtained from Biomol International LP. Compactin was a kind gift from R. Simoni, Stanford University. Unless otherwise noted, all other reagents were from Sigma-Aldrich Co. Fetal bovine lipoprotein-deficient serum (LPDS, d Ն1.25) was prepared by ultracentrifugation, as described (30).
Cell Fractionation-Enriched ER membranes were prepared from LP-90 cells by centrifugation through discontinuous sucrose gradient, as described by McGee et al. (33). To prepare cytosolic fraction, washed UT-2 cells were swollen for 10 min by incubation on ice in hypotonic buffer (20 mM K-HEPES, pH 7.4, 10 mM KCl, 1.5 mM MgCl 2 ). The cells were collected in 0.5 ml of the hypotonic buffer that was supplemented with 250 mM sucrose and homogenized by 20 strokes in a Dounce homogenizer followed by 20 passages through a 25-gauge needle. Cell debris and nuclei were removed by centrifugation (10,000 ϫ g for 15 min at 4°C) and the postnuclear supernatant was spun again (200,000 ϫ g for 30 min at 4°C). The resulting supernatant is designated "cytosol." Construction of HMG 350 -3HA, HMG 240 -3HA and Generation of Mutants-The sequence for triple HA epitope (3HA), with appended SmaI and NotI sites at its 5Ј-and 3Ј-ends, respectively, was amplified from pRH990 (kindly provided by R. Hampton) and ligated in-frame downstream to an EcoRI-DraI-digested Syrian hamster HMGR cDNA (at residue 350). The resulting 1.2 kb HMG 350 -3HA cDNA was subcloned as an EcoRI-NotI fragment in pBlueScript. This plasmid was used for all subsequent mutageneses by the overlap extension method (34), using high fidelity Pfu DNA polymerase (Promega), the appropriate internal mutagenic primers (list of primers is available upon request), and the T3 and T7 primers of pBlueScript. Lysine-less HMG 350 -3HA was created by successive Lys 3 Arg mutagenesis of all 11 lysines within the wild-type construct. The triple HA epitope does not contain any lysines. To construct HMG 240 -3HA, HMGR cDNA was first cut at residue 241 with BstXI and the protruding 3Ј-overhang was blunted with Mung bean nuclease (New England Biolabs). Following digestion with EcoRI, the blunt 3Ј-end of this 0.7 kb fragment was ligated upstream at the SmaI site of the above described 3HA epitope. The authenticity of all constructs and the presence of the desired mutations were verified by DNA sequencing. For mammalian expression, the wild-type and mutant cDNAs were cloned at the EcoRI-NotI sites of pIRESneo2 vector (Clontech) and plasmids were stably transfected into HEK-293 cells using LipofectAMINE Plus reagent (Invitrogen Life Technologies) according to the manufacturer's instructions.
Metabolic Labeling, Immunoprecipitation, and Immunoblotting-On day 1, 1.5 ϫ 10 6 cells were plated in 60-mm dishes in Medium A. On day 2 the cells were re-fed with medium B (Dulbecco's modified Eagle's medium containing 10 mM Na-HEPES, pH 7.4, 2 mM glutamine, 100 units/ml penicillin, 100 g/ml streptomycin, and supplemented with 10% (v/v) LPDS, 2 M compactin, and 100 M sodium mevalonate). On day 3, the cells were starved for 1 h in 1 ml of Medium C (methionineand cysteine-free Dulbecco's modified Eagle's medium supplemented with 10% (v/v) LPDS, 2 M compactin and 100 M MVA) and then pulse-labeled for 30 min with 100 Ci of Expre 35 S 35 S protein labeling mix (PerkinElmer Life Sciences; Ͼ1000 Ci/mmol). Chase was commenced by aspirating the radioactive medium and addition of fresh Medium B that was supplemented with 2 mM unlabeled methionine and the indicated additions. At the designated time points, the cells were lysed and processed for immunoprecipitation with the indicated antibodies and immobilized protein A, as described (29,35). Samples were resolved by 5-15% SDS-PAGE, and gels were quantified using Cyclone phosphorimager and Optiquant software package (Packard).
For immunoblotting analyses, after cell lysis and estimation of protein content, samples containing equal amounts of protein were resolved by SDS-PAGE and electroblotted onto Optitran BAS-83 reinforced nitrocellulose membranes (Schleicher & Schuell; Germany). The membranes were probed with the primary antibodies detailed below, followed by horseradish peroxidase (HRP)-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories). HRP was visualized by enhanced chemiluminescence reaction.
Endogenous HMGR and HMG 350 -3HA were immunoprecipitated with specific antiserum directed against the membrane domain of HMGR (15). Alternatively, HMG 350 -3HA and HMG 240 -3HA were precipitated with anti-HA antibody (clone 12CA5; Roche Applied Science). HMGR was immunoblotted with the A9 monoclonal antibody (36). For ubiquitination and pull-down assays, HMG 350 -3HA was first immuno-  2) or with pIRES-HMG 350 -3HA plasmid and G-418-resistant colonies were pooled and expanded. Cells were set up for the experiment, as described under "Experimental Procedures." Following incubation for 16 h in Medium B, the cells were pulse-labeled for 30 min with 35 S protein labeling mix (300 Ci/ml; 100 Ci/dish), and chased in the absence (Ϫ) or presence (ϩ) of sterols (2 g/ml 25-hydroxycholesterol plus 20 g/ml cholesterol). At the indicated time points, cells were lysed, and endogenous HMGR and HMG 350 -3HA were simultaneously immunoprecipitated with anti-HMGR membrane domain antiserum. C, degradation of the truncated HMG 240 -3HA is not regulated by sterols. HMG 240 -3HA, stably expressed in HEK-293 cells, was analyzed by pulse-chase and immunoprecipitation, as described in B, except that HMG 240 -3HA protein was immunoprecipitated with anti-HA-agarose. Lane 1, immunoprecipitation from labeled mock-transfected HEK-293 cells.

RESULTS
To identify the site(s) in HMGR to which ubiquitin is conjugated in a metabolically regulated fashion, we focused on its transmembrane region, which also allowed us to minimize the number of lysines that are potentially modified. For this purpose, we constructed truncated versions of HMGR that consist of the 350 or 240 N-terminal residues of the enzyme, each tagged with 3 copies of the HA epitope, as shown in Fig. 1A. This scheme illustrates the current model for the secondary structure of the HMGR transmembrane region. Previous studies, including in vitro translation and membrane translocation experiments, in vivo analysis of HMGal constructs with engineered N-glycosylation sites and direct immunological examination with anti-peptides antibodies, indicated that this region consists of eight transmembrane spans separated by three cytosolic and four luminal loops (15,37). Yet, the boundaries of these spans have not been precisely determined. Fig. 1B shows that stable transfection of pIRES-HMG 350 -3HA led to the expression of a protein doublet of ϳ44kDa that was immunoprecipitated with antibodies against the membrane domain of HMGR as well as with an anti-HA antibody (see below). This doublet represents the non-glycosylated (lower band) and N-glycosylated (upper band) forms of HMG 350 -3HA, as indicated by digestion with endoglycosidase H. This treatment increased the electrophoretic mobility of the upper band such that it co-migrated with the lower band, which was not affected by this treatment (data not shown). Importantly, although the turnover rate of HMG 350 -3HA in steroldepleted cells was significantly faster than that of endogenous HMGR (t1 ⁄2 3.4 h versus ϳ12 h, respectively; see also Ref. 38), its degradation was markedly accelerated upon addition of sterols to the cells (see Table I), similarly to the effect of sterols on the degradation of native HMGR. Moreover, like endogenous HMGR, degradation of HMG 350 -3HA was inhibited by MG-132 (data not shown; see also Fig. 3A), indicating that this degradation was also mediated by the ubiquitin-proteasome pathway. Fig. 1C shows that the truncated mutant HMG 240 -3HA, which includes the entire putative SSD of the reductase, was degraded with a half-life similar to that of HMG 350 -3HA (3.4 h). Nonetheless, this process was no longer regulated by sterols. Thus, the 350 N-terminal residues of HMGR that encompass all eight transmembrane spans are necessary and sufficient to endow metabolically-regulated turnover, making the HMG 350 -3HA a bona fide model to study the intracellular fate of HMGR.
According to the model in Fig. 1A, among the 11 lysines and the N terminus that may potentially serve as ubiquitin acceptor sites in HMG 350 -3HA, only Lys 89 , Lys 248 , and the N terminus face the cytosol and are therefore readily accessible to the ubiquitination machinery at the cytosolic face of the ER membrane. We replaced these lysine residues with arginines and examined by pulse-chase experiments whether these mutations affected the regulated turnover of HMG 350 -3HA. The results are shown in Fig. 2 and the calculated half-lives of the different mutants are summarized in Table I. In the absence of sterols, the degradation of the K89R mutant was similar to that of wild-type (WT) HMG 350 -3HA (Table I; compare lanes 1-5 in Fig. 2, b and a). However, in the presence of sterols the degradation of K89R was noticeably attenuated relative to the wildtype HMG 350 -3HA (lanes 5-9 in Fig. 2, b and a). The K248R mutant was degraded slightly slower in the absence of sterols (Table I; Fig. 2c, lanes [1][2][3][4][5], but similarly to K89R in steroltreated cells (Table I; Fig. 2c, lanes [5][6][7][8][9]. Substituting both lysines (K 89ϩ248 R; Fig. 2d) resulted in a much more pronounced effect in the absence of sterols (Fig. 2d, lanes 1-5, and Table I) and nearly abolished the sterol-accelerated degradation of this mutant (Fig. 2d, lanes 5-9 and Table I). Replacements with arginines of other lysines in the transmembrane region of HMG 350 -3HA, which are assumed to be buried within the lipid bilayer (such as Lys 142 or Lys 336 ; see Fig. 1A), had no appreciable effect on the rate of degradation, either in the absence or presence of sterols (data not shown). Thus, it appears that both Lys 89 and Lys 248 are necessary to confer sterol-regulated turnover of HMGR.
To examine whether Lys 89 and Lys 248 were also sufficient for regulation of HMGR degradation, we generated a mutant of HMG 350 -3HA in which all 11 lysines were substituted, as well as versions in which only Lys 89 or Lys 248 or both lysines were preserved (Fig. 2, e-h and Table I). Relative to the wild-type protein, the lysine-less version of HMG 350 -3HA (No Lys; e) was degraded significantly faster in the absence of sterols. Nevertheless, sterols did not accelerate the turnover of this mutant (Fig. 2e). Preserving the lysine at position 89 did not significantly alter the degradation of this protein relative to the lysine-less HMG 350 -3HA in the absence or presence of sterols (Fig. 2e, lanes 10 -18 and Table I). The phenotype of the preserved lysine at position 248 was much more pronounced than that of Lys 89 ; not only was this HMG 350 -3HA K248 protein eliminated faster in the absence of sterols, but its degradation was also markedly hastened in the presence of sterols ( Fig. 2g and Table I). Fig. 2h shows that, in the absence of sterols, the degradation of HMG 350 -3HA in which both Lys 89 and Lys 248 were preserved was more rapid than that of HMG 350 -3HA K248 (Table I). Moreover, upon addition of sterols, this HMG 350 -3HA K89ϩK248 mutant was degraded even faster, fully restoring the extent of sterol-accelerated degradation observed with the wild-type protein (Table I). Taken together, these results demonstrate that either Lys 89 or Lys 248 are singly sufficient to permit regulated degradation of HMGR. Moreover, it appears that the effects of these residues are additive such that maximal degree of regulation by sterols is the sum of contributions of each lysine, with Lys 248 playing a more prominent role than Lys 89 in endowing sterol-regulated turnover onto the transmembrane region of HMGR.
To correlate between the degradation phenotype of the various mutants and their state of polyubiquitination, we immunoprecipitated the different HMG 350 -3HA proteins and immu-  3, lower panel). When MG-132 was added, the intensity of the ubiquitin-containing material was markedly enhanced (Fig. 3A, lane 2, upper panel) and treatment with sterols and MG-132 resulted in a further increase in ubiquitinated material that was precipitated with the anti-HA antibody (Fig. 3A, lane 4). As expected, MG-132 inhibited the sterol-induced decrease in HMG 350 -3HA levels (Fig. 3A, lane 4, lower panel). In Fig. 3B, we calculated the "specific ubiquitination" (ratio between the intensity of the ubiquitinated material associated with HMG 350 -3HA and the intensity of HMG 350 -3HA band) under conditions of maximal modification (i.e. in the presence of sterols and MG-132; see Fig. 3A, lane 4). This value of specific ubiquitination for the various HMG 350 -3HA mutants was compared with that of the wild-type, which was set as 100%. As shown, elimination of Lys 89 reduced the ubiquitination by 22%, whereas abolishing Lys 248 gave rise to 32% decline in polyubiquitin decoration of HMG 350 -3HA. Replacement of both lysines produced an additive effect such that the specific ubiquitination of this protein decreased by 56%. These results demonstrate that the ubiquitination of Lys 89 and Lys 248 is independent of one another. Importantly, the K89R/K248R double mutant was nonetheless ubiquitinated to a measurable extent, indicating that other amino groups, such as the N-terminal ␣-NH 2 or other lysines within the membrane spans or in the luminal loops, are accessible to the E2/E3 enzymes. Indeed, possibility that the Nterminal ␣-NH 2 could be modified with ubiquitin was confirmed by the finding that the lysine-less HMG 350 -3HA (No Lys) was modified even to a greater extent than the K89R/ K248R double mutant (Fig. 3B, K89ϩ248R). On this background, the presence of lysine at position 89 (K89) had no effect. Yet, restoring a single lysine at position 248 (K248) resulted in ubiquitin decoration that equaled that of the wildtype HMG 350 -3HA. Moreover, the protein in which both lysines were preserved (K89 ϩ 248) was modified slightly more than the wild-type HMG 350 -3HA (Fig. 3B). Interestingly, plotting the relative stability of the various HMG 350 -3HA mutants as a function of their specific ubiquitination clearly yielded an inverse relationship for both types of mutants (Fig. 3C). This indicates that polyubiquitin conjugation is rate-limiting in HMGR degradation. However, polyubiquitination is not the only factor contributing to degradation since mutants that were modified to a similar extent as WT HMG 350 -3HA (e.g. K248) were degraded faster that the WT protein ( Fig. 3C and Table I). Taken together, these results pinpoint Lys 248 and Lys 89 as the primary and secondary sites, respectively, for polyubiquitin attachment to the membrane region of HMGR. Moreover, the presence of lysines other than Lys 89 and Lys 248 appears to attenuate the ubiquitination at these sites and partially stabilize the protein. Finally, the data indicate that the cytosol-facing N terminus of HMGR also serves as an acceptor site for ubiquitin conjugation. According to the topological model shown in Fig. 1A, Lys 89 and Lys 248 lie at the interface between the 2 nd and 4 th cytoplasmic loops and the 3 rd and 7 th membrane spans, respectively. Since there is no consensus as to the specificity of the internal lysines to which ubiquitin is attached, we next tested whether the position of such lysines might affect the degradation of HMG 350 -3HA. Using the lysine-less HMG 350 -3HA as a backbone, we individually mutated to lysines the cytosol-oriented arginines at positions 7, 84, 154, and 240, and examined the phenotype of these mutants. By pulse-chase analysis, none of these substitutions appreciably affected the half-lives of these proteins nor did they restore sterol-regulated degradation ( Fig. 4A and Table I). Interestingly, substituting Arg 240 to lysine in the context of HMG 350 -3HA K89ϩ248R double mutant partially restored sterol-regulated turnover onto this HMG 350 -3HA K89ϩ248R/R240K triple mutant ( Fig. 4B and Table I). These results indicate that the position of the lysines, at least those in the 3 rd loop that includes residues 240 and 248, may be somewhat flexible as long as the other lysines in the membrane region are present. Otherwise, in the context of the lysine-less protein, the position of cytoplasm-exposed lysines is critical for controlling HMGR degradation.
The predicted proximity of Lys 89 and Lys 248 to the cytosolic face of the ER raises the possibility that HMGR is ubiquitinated by membrane ubiquitin-conjugating enzymes and ubiquitin-protein ligases. This possibility was explored by looking for cytosolic factors that are required to ubiquitinate HMGR in vitro. Purified ER membranes from the HMGR-overexpressing LP-90 cells were incubated with biotinylated ubiquitin, ATP, and either crude cytosol form HMGR-deficient UT-2 cells or purified E1. After 2 h of incubation, HMGR was immunoprecipitated and polyubiquitin chains conjugated to the reductase were detected with HRP-strepavidin. As shown in Fig. 5, in the presence of cytosol, HMGR was extensively polyubiquitinated at 37°C (lane 2) but not at 0°C (lane 1). Remarkably, a similar  Table I. b-d, Lys 3 Arg substitutions on wild-type background (a); f-h, restored Lys on a "No Lys" background (e). modification with polyubiquitin chains also occurred upon addition of solely E1 (lanes 4 -6). This reaction depended on ATP (compare lane 7 to lane 5) but the extent of the ubiquitin smear immunoprecipitated along with HMGR indicates that ubiquitination was not as robust as the one carried out in the presence of crude cytosol (compare lane 6 to lane 2). This suggests that, in addition to E1, the processive assembly of high molecular weight polyubiquitin chains on HMGR may require other cytosolic factors. Nevertheless, these results demonstrate that the E2 and E3 enzymes responsible for ubiquitination for HMGR are fractionated with the membrane and that E1 is the only cytosolic protein necessary for this reaction to occur.
Degradation of ERAD substrates mandates that the proteins should be first extracted from the membrane before proteolysis by the cytoplasmic proteasome ensues. However, it is not clear if such proteins must dislocate fully before proteolysis commences. This question is especially relevant for polytopic membrane proteins with multiple transmembrane spans such as HMGR. To gain insight into the topological status of HMGR as it is targeted for degradation, we examined whether HMGR remains membrane-bound under conditions that favor its degradation. For that, LP-90 cells that were continuously grown in lovastatin were washed with phosphate-buffered saline and transferred either back to fresh medium supplemented with lovastatin or to medium lacking lovastatin but containing MG-132. After 2 h of incubation, the cells were homogenized, and membranes were prepared. The membranes were resuspended in 100 mM either sodium chloride at pH 7.4 or sodium carbon-  Table I) is plotted against their specific ubiquitination (expressed as percentage of that of wild-type from B).
ate at pH 11. The treated membranes were spun again to separate between supernatant and pellet fractions, from which HMGR was immunoprecipitated and immunoblotted sequentially with anti-ubiquitin and anti-HMGR antibodies. As can be seen in Fig. 6, in cells that were maintained in lovastatin, where the rate of HMGR degradation is slow, HMGR in the membrane pellet was decorated with polyubiquitin chains rather poorly (lanes 3 and 4). Moreover, no ubiquitin-reactive material was immunoprecipitated by the anti-HMGR antiserum from the salt wash supernatants, either at neutral pH or basic pH (lanes 1 and 2). The situation was strikingly different in cells that were relieved from lovastatin inhibition, where the intracellular surge in MVA promotes rapid degradation of HMGR (29). Under these conditions, the anti HMGR antiserum precipitated heavily polyubiquitinated material that originated from the 97-kDa HMGR band from the salt wash supernatant only at pH 11 (lane 5) but not pH 7.4 (lane 6). These results indicate that the modification with ubiquitins correlated with alterations in the properties of HMGR from a fully integral protein of the ER membrane to one that is peripherally associated with these membranes. It should be noted, however, that despite the visible amounts of polyubiquitinated reductase in the carbonate wash, the HMGR protein that was immunoprecipitated was barely at the level of detection (lane 5), suggesting that only a very small fraction of HMGR is heavily modified by ubiquitin, and only this fraction is peripherally associated with the ER membrane.
The higher susceptibility of the polyubiquitinated HMGR to extraction from the microsomes suggested that this protein had already (or at least partially) dislocated but remained associated with the ER membrane. Recent studies demonstrated that the dislocation from the ER of both membrane and luminal ERAD substrates requires the activity of the cytosolic p97/VCP/ Cdc48p-Ufd1-Npl4 complex (21,23,24). This complex was also implicated in the metabolically regulated degradation of  Fig. 1A). HEK-293 cells that stably express these mutants were pulse-labeled with 35 S protein labeling mix and chased in the absence (Ϫ) or presence (ϩ) of sterols, as described in the legend to Fig. 2. The HMG 350 -3HA proteins were immunoprecipitated with an anti-HA antibody. B, Arg 3 Lys substitution was introduced at position 240 of the K89R/K248R double mutant, and this K89R/K248R/R240K triple mutant was analyzed as described in Fig. 2 (lower panel). a and d from Fig. 2 (upper and middle panels, respectively) are presented for comparative visualization. Half-life values of these mutants are summarized in Table I.

Ubiquitination of Transmembrane Region of Mammalian HMGR
HMGR in yeast since the hrd-4 mutation was traced to Npl4 (39). It was suggested that p97/VCP/Cdc48p-Ufd1-Npl4 plays a role in a step subsequent to HMGR ubiquitination but prior to its recognition by the 26 S proteasome (39). Therefore, it was of interest to examine whether and under which conditions does p97/VCP interact with mammalian HMGR. As shown in Fig. 7, p97 was preferentially co-precipitated with ubiquitinated HMG 350 -3HA (lane 6) as well as with endogenous HMGR (data not shown), while only minute amounts were pulled down with the unmodified reductase (lane 5). These results are in accord with the reported ability of p97/VCP/Cdc48p and its cofactors Ufd1 and Npl4 to bind ubiquitin (40,41) and indicate that also in mammalian cells p97, presumably complexed with the mammalian Ufd1 and Npl4, operates in the regulated degradation of HMGR most likely at the dislocation of this polytopic protein from the ER membrane. DISCUSSION In the current study, we demonstrate that polyubiquitin chains are covalently attached to the transmembrane region of HMGR prior to reductase degradation by the 26 S proteasome. Of the 11 lysines within the 350 N-terminal residues that constitute this region, lysines 89 and 248 serve as the acceptor sites for ubiquitin conjugation. According to the prevailing topological model for the membrane region of HMGR (Fig. 1A), both Lys 89 and Lys 248 face the cytoplasm and are thus accessible to ubiquitin conjugating enzymes and protein-ubiquitin ligases. When both residues are substituted with arginine, ubiquitination of HMG 350 -3HA is significantly reduced and sterol-accelerated degradation of this model protein is abolished. Quantitatively, although the effects of Lys 89 and Lys 248 on ubiquitination and degradation appear additive, the phenotype of the single K248R mutation is more severe than that of K89R substitution (Fig. 3). This indicates that Lys 248 is the principal site for polyubiquitin attachment while Lys 89 plays a lesser role. While our work was in progress, Sever et al. (42) presented similar results except that in their hands the K89R mutation had no effect on reductase ubiquitination, and degradation was slightly attenuated only when Lys 248 was also mutated. Significantly, when all 11 lysines are replaced, HMG 350 -3HA is still efficiently ubiquitinated and rapidly degraded, yet in a constitutive manner that is not responsive to sterols. Moreover, re-introducing lysines at positions 89 and 248 fully restores sterol regulation. Unexpectedly, however, this lysine-less protein and Lys 89/248 double mutant derived from it are degraded even faster than the K89R/K248R double mutant or the wild-type proteins, respectively, either in the absence or presence of sterols ( Fig. 2 and Table I). These results suggest that the N terminus of HMGR, which is exposed to the cytoplasm (Fig. 1A) and does not appear to be blocked (43,44), is susceptible to ubiquitination in a sterol-independent manner. N terminus ubiquitination is a novel mode of modification that was thus far demonstrated or implicated in the degradation of only a handful of other proteins, both membrane and soluble (Ref. 45; for a recent review see Ref. 46 and references therein). In some cases, substitution of the internal lysines slightly inhibited ubiquitin conjugation as well as degradation. In contrast, the faster turnover of the lysine-less HMG 350 -3HA and, in particular, the increased level of ubiquitination of the Lys 89/248 mutant and its exceptionally high rate of degradation, suggest that, in the wild-type situation, lysines other than Lys 89  none of which is mutually exclusive, may explain these observations: (i) eradicating all internal lysines forces ubiquitination at the N terminus, and N-terminal polyubiquitin serves as a better molecular tag for reductase degradation; (ii) internal lysines serve as alternate acceptor sites, although less efficient than Lys 89 or Lys 248 , thus practically act as "competitive inhibitors" in the ubiquitination of the latter; (iii) substituting the internal lysines with arginines may have rendered the conformation of the transmembrane region non-native so as to be recognized by the ERAD machinery as an unfolded protein.
Nevertheless, such changes in conformation are likely to be subtle since re-introduction of a single lysine at position 248 restored nearly maximal sterol regulation (Table I). It should be noted that there are 39 lysines in the entire 97-kDa (887 residues) subunit of the hamster HMGR. However, if some of these residues could also be ubiquitinated and contribute to reductase elimination by the proteasome, only Lys 89 and Lys 248 are necessary and sufficient for the regulation of degradation by metabolic signals, pointing to the membrane region as the sole regulatory domain of HMGR.
Recent studies by Sever et al. (42,47) have demonstrated that the ubiquitination of HMGR is accelerated by the sterolinduced binding of its SSD to the ER proteins Insig-1 and/or Insig-2. The K248R substitution does not interfere with this binding, but replacement with alanines of the sequence YIYF at positions 75-78 diminishes the interaction between HMGR and Insig-1. This tetrapeptide in the 2 nd transmembrane span (Fig. 1A), which constitutes the N-terminal helix of the HMGR SSD, is also found in the SSD of SCAP (positions 298 -301) (6). Mutant SCAP in which the first tyrosine of this tract has been changed to cysteine (Y298C) fails to bind Insig-1 and is resistant to sterol-mediated retention in the ER (9). However, the homologous site-directed mutation (Y75C) in the membrane domain of HMGR had no effect on its sterol-regulated degradation, 2 and similar results were obtained by Sever et al. (42) with the Y75A substitution. Instead, the isoleucine and the second tyrosine in this tetrapeptide (IY sequence) were found to be required for sterol-stimulated binding of HMGR to Insig (42). Association between HMGR and Insig also appears to depend on four conserved Phe residues in the 6 th transmembrane span, the most C-terminal helix of the HMGR SSD (48). As Lys 248 and Lys 89 are predicted to lie close to the cytosolic surface of the ER membrane, it is possible that binding of Insig to HMGR primes the latter to ubiquitin conjugation by exposing its Lys 248 and Lys 89 . Alternatively, the HMGR⅐Insig complex may recruit/activate a specific membrane E3 that attaches ubiquitins to these lysines. Inasmuch as the re-introduction of ubiquitin acceptor lysines at positions 248 and 89 fully restored sterol-stimulated ubiquitination (Fig. 3B) as well as degradation of the otherwise lysine-less HMG 350 -3HA (Fig. 2, g and h), and Lys 248 is not required for sterol-dependent association of Insig with HMGR (42), all lysines in the transmembrane region are, thus, dispensable for the interaction of HMGR with Insig. Therefore, ubiquitination at the N terminus may be carried out by a different E3 than the ligase(s) engaged in Lys 89/248 modification. Alternatively, the same E3 may attach ubiquitin to the constitutively accessible N terminus as well as to Lys 89/248 , yet the latter are exposed only in response to sterol-induced conformational changes and association with Insig. The current data cannot distinguish between these possibilities.
The sites for ubiquitin attachment vary among different substrates and no rules can be formulated as to which protein amine groups may be tagged. In some cases, ⑀-NH 2 group of only distinct lysines are modified, while in others there is little or no specificity. Moreover, the free ␣-NH 2 terminus may also be subject to ubiquitination (19,46). Indeed, here we show that the three NH 2 groups in HMGR membrane domain that face the cytoplasm (N terminus, Lys 89 , Lys 248 ; Fig. 1A) are all available for modification by the ubiquitination machinery. Consistent with the predicted proximity of Lys 89 and Lys 248 to the surface of the bilayer, the ubiquitination of their ⑀-NH 2 groups is carried out by membrane E2 and E3 enzyme(s) (Fig.  5; see also Ref. 49). Yet, replacing arginines in the lysine-less HMG 350 -3HA with lysine residues at the same cytosolic topology, but at alternate positions, did not restore sterol-regulated degradation ( Fig. 4A and Table I), indicating that the ubiquitination of these replacement mutants was not enhanced by sterols. These results suggest that considerable spatial constrains are imposed on these E2/E3s such that only Lys 248 and/or Lys 89 lie in a favorable position. Moreover, it is possible that replacing all 11 lysines in the membrane region of HMGR "rigidifies" its cytosolic loops such that the artificially implanted lysines are no longer accessible to the E2/E3 enzymes. In this context, it is interesting to note that the R240K substitution in the HMG 350 -3HA K89ϩ248R double mutant, but not in the lysine-less protein, partially restored sterol-regulated degradation ( Fig. 4B and Table I).
Hampton and co-workers (50 -52) have identified and characterized Hrd1p as a membrane-anchored E3 ubiquitin ligase that participates, with Ubc7 and/or Ubc1, in the constitutive degradation of many ERAD substrates, as well as in the metabolically regulated turnover of the yeast HMGR isozyme Hmg2p. Our recent studies on the human homolog of Hrd1 demonstrated that, similar to the yeast protein, the human Hrd1p is an ER membrane protein with multiple transmembrane spans and a cytoplasmic RING-H2 finger motif characteristic of many E3s, which cooperates with mammalian Ubc7 to forms Lys 48 -specific polyubiquitin linkages (53). We observed that overexpression of the human Hrd1p in mouse NIH-3T3 cells hastened the degradation of endogenous HMGR only in sterol-depleted cells (so called "basal" degradation) but did not affect the sterol-stimulated degradation of the reductase. Moreover, overexpression of Hrd1p with a RING finger mutation, which abolished its ligase activity, affected neither basal nor sterol-accelerated turnover of HMGR. Nonetheless, consistent with its expected dominant-negative effect, this RINGfinger mutant Hrd1p slowed the constitutive degradation of two classical ERAD substrates TCR␣ and CD3␦ (53). These results suggest that, depending on whether cellular sterols are scarce or abundant, at least two different E3 ubiquitin ligases engage HMGR. Interestingly, transient transfection of the RING-finger mutant Hrd1p into cells that express the lysineless HMG 350 -3HA exerted no dominant-negative effect also on the rapid constitutive degradation of this sterol-insensitive protein (data not shown), lending further support to the notion that the lysine-less HMG 350 -3HA is not grossly misfolded (see above). Thus, the current data suggest that membrane E3s, such as gp78 (54) or mammalian Doa10p (55), are involved in the metabolically regulated degradation of HMGR.
Proteolysis of ERAD substrates by the proteasome mandates their prior extraction out of ER membrane. For some membrane substrates with a single transmembrane span, such as the heavy chain of major histocompatibility complex (MHC) class I and TCR␣, complete dislocation precedes degradation, as the solubilized protein is detected in the cytosol (56,57). However, extraction before degradation has never been reported for polytopic membrane proteins such as HMGR. Here we show that ubiquitination of HMGR is correlated with a change in its properties from a membrane embedded protein to one that could be removed by basic pH treatment, a property characteristic to peripheral membrane proteins (58). Moreover, only the ubiquitinated species of HMG 350 -3HA co-precipitated the cytosolic AAA-ATPase p97/VCP. Taken together, these results suggest that polyubiquitinated HMGR accumulates at the cytoplasmic face of the ER in proteasome-arrested cells. Thus, similar to luminal ERAD substrates (24), and unlike certain membrane ERAD substrates in yeast (59), the dislocation of the polytopic membrane protein HMGR may be accomplished without the participation of proteolytically active proteasome, but probably with the aid of the p97/VCP/Cdc48p-Ufd1-Npl4 complex. Nevertheless, unlike the MHC class I heavy chain, polyubiquitinated HMGR is not released to the cytosol but remains associated with the cytosolic face of the ER membrane.
In conclusion, the current experiments extend our initial observations on the involvement of the ubiquitin-proteasome pathway in eliminating HMGR. We mapped the acceptor sites for ubiquitin conjugation to specific lysine residues in the transmembrane region of the reductase and also provided evidence for ubiquitination of its N terminus, implicating different sets of E2/E3s in HMGR ubiquitination. These results shed light on the complexity of the metabolically regulated turnover of HMGR and should prove useful in identifying the components that are involved in this process.