|Home | About | Journals | Submit | Contact Us | Français|
Quality control pathways such as ER-associated degradation (ERAD) employ a small number of factors to specifically recognize a wide variety of protein substrates. Delineating the mechanisms of substrate selection is a principle goal in studying quality control. The Hrd1p ubiquitin ligase mediates ERAD of numerous misfolded proteins including soluble, lumenal ERAD-L and membrane-anchored ERAD-M substrates. We tested if the Hrd1p multi-spanning membrane domain was involved in ERAD-M specificity. In this work, we have identified site-directed membrane domain mutants of Hrd1p impaired only for ERAD-M and normal for ERAD-L. Furthermore, other Hrd1p variants were specifically deficient for degradation of individual ERAD-M substrates. Thus, the Hrd1p transmembrane region bears determinants of high specificity in the ERAD-M pathway. From in vitro and interaction studies, we suggest a model in which the Hrd1p membrane domain employs intra-membrane residues to evaluate substrate misfolding, leading to selective ubiquitination of appropriate ERAD-M clients.
The endoplasmic reticulum associated degradation (ERAD) pathway mediates the destruction of numerous normal and misfolded ER-localized proteins (Hampton, 2002a; Hampton et al., 1996; Ravid et al., 2006). The ERAD pathway has been implicated in a wide variety of processes, including sterol synthesis, rheumatoid arthritis, fungal differentiation, cystic fibrosis, and several neurodegenerative diseases (Amano et al., 2003; Hampton and Rine, 1994; Liang et al., 2006; Swanson et al., 2001; Zhang et al., 2002). Accordingly, there is great impetus to understand the molecular mechanisms that mediate this important route of protein degradation.
ERAD proceeds by the ubiquitin-proteasome pathway, in which an ER-localized substrate is covalently modified by three enzymes in order to form a multi-ubiquitin chain that is recognized by the cytosolic 26S proteasome (Voges et al., 1999). The E1 ubiquitin activating enzyme uses ATP to covalently activate and then add ubiquitin to an E2 ubiquitin conjugating enzyme. Ubiquitin is then transferred from the ubiquitin-charged E2 to the substrate or the growing ubiquitin chain by the action of an E3 ubiquitin ligase, resulting in a substrate-attached multi-ubiquitin chain. In most cases, ancillary factors participate in substrate recognition and transfer of the ubiquitinated substrate to the proteasome (Carvalho et al., 2006; Denic et al., 2006; Richly et al., 2005).
The HRD pathway is one of the principal routes of ERAD in eukaryotes, being responsible for the degradation of both lumenal and membrane-bound misfolded ER proteins (Hampton et al., 1996; Knop et al., 1996; Vashist and Ng, 2004). The HRD pathway E3 ligase is the highly conserved Hrd1p, which is rate-limiting for degradation (Bays et al., 2001a). Hrd1p is a multi-spanning ER membrane protein, consisting of an N-terminal membrane anchor linked to a soluble C-terminal domain with a RING-H2 domain characteristic of many E3 ligases (Figure 1). The C-terminal region is responsible for catalyzing the transfer of ubiquitin from the appropriate E2s to ERAD substrates (Bays et al., 2001a). However, successful degradation of ERAD substrates requires the presence of the Hrd1p membrane anchor, either as the full-length protein, or when expressed in trans with the active C-terminal region (Gardner et al., 2000). The multi-spanning Hrd1p membrane domain has numerous known functions, including binding to and communication with the lumenal domain of Hrd3p, correct placement of the C-terminal ligase domain, and recruitment of ERAD factors for recognition of misfolded proteins and for later steps in the pathway such as retrotranslocation (Bazirgan et al., 2006; Gardner et al., 2000; Neuber et al., 2005). The namesake substrate of the HRD pathway is Hmg2p, a yeast isozyme of the sterol pathway enzyme HMG-CoA reductase (HMGR). Hmg2p undergoes regulated entry into the HRD pathway, so that when production of sterol pathway products is high, HRD-dependent degradation of Hmg2p is more rapid (Gardner and Hampton, 1999; Gardner et al., 2001a). Regulation of Hmg2 stability appears to occur by pathway signal-induced misfolding of Hmg2p that improves HRD pathway recognition (Gardner et al., 2001b). In this way, ERAD is employed as part of the feedback regulation of sterols, and a similar mechanism operates in mammals (Goldstein and Brown, 1990; Hampton, 2002b; Hampton and Garza, 2009).
HRD pathway substrates fall into two broad categories: soluble lumenal proteins such as CPY*, or integral membrane proteins such as Hmg2p or Pdr5* (Hampton et al., 1996; Knop et al., 1996; Plemper et al., 1998). Substrates are diverse, indicating that misfolded proteins are recognized by structural criteria that transcend the absence of any primary sequence similarity between the various members of each group. In the case of lumenal proteins such as CPY*, a variety of factors have been proposed to mediate the recognition of hallmarks of misfolding required for presentation to the HRD machinery. The classic chaperone Kar2p, the lumenal lectins Htm1p and Yos9p and the ER-anchored lumenal domain of Hrd3p have all been implicated in recognition of lumenal ERAD substrates (Carvalho et al., 2006; Denic et al., 2006; Jakob et al., 2001). However, neither Kar2p or Yos9p is required for degradation of membrane-bound substrates, nor is Hrd3p if sufficient Hrd1p is present (Gardner et al., 2000). Similarly, Der1p is required for lumenal substrate degradation but is dispensible for integral membrane substrates such as Hmg2p or Pdr5* (Plemper et al., 1998; Sato and Hampton, 2006).
Because of these distinctions between lumenal and membrane-anchored substrates, the degradation of each class of proteins has been referred to as ERAD-L for the lumenal substrate pathway, and ERAD-M for the integral membrane pathway. In striking contrast to the success of identifying factors for recognition of ERAD-L substrates, little is known about how membrane proteins are recognized as ERAD substrates (Carvalho et al., 2006; Denic et al., 2006).
In many cases in the ubiquitin pathway, the E3 ubiquitin ligase is the primary mediator of substrate recognition. We wondered if the ubiquitin ligase Hrd1p plays this direct role in ERAD-M. Although less is known about what features an aberrant or “misfolded” membrane protein might possess, those features would likely be present within or near the bilayer. Thus, the multi-spanning transmembrane domain of Hrd1p would be the appropriate region to mediate recognition of ERAD-M substrates. One idea is that a correctly folded and assembled integral membrane protein would be expected to present no free hydrophilic groups within the lipid region of the membrane, while a misfolded membrane protein would expose hydrophobic groups to the bilayer, allowing detection by interacting with similar groups in an integral membrane E3. In fact, the Hrd1p transmembrane anchor has a high proportion of hydrophilic R groups in its six transmembrane spans that might serve such a detection function (Figure 1 and (Deak and Wolf, 2001)).
As part of a systematic analysis of the Hrd1p transmembrane region, we have studied the effects of mutating these hydrophilic groups, along with other residues that are highly conserved in Hrd1p orthologues, to query the mechanisms of specific substrate recognition. We have identified mutants each deficient in recognition of distinct ERAD-M substrates, indicating a role for the membrane domain in substrate detection. Furthermore, a detailed analysis of one of our recognition-deficient mutants indicates a role for the transmembrane domain in regulation of the activity of the ligase upon encountering a misfolded substrate, consistent with our earlier studies on Hrd1p-substrate mechanisms. Thus, Hrd1p bears a code for detection of misfolded proteins in a membrane environment. Unraveling this code will have important implications in understanding the many processes that pertain to management of protein quality in normal and pathological cellular states.
The Hrd1p transmembrane region contains a large number of intra-bilayer hydrophilic amino acids, which we targeted for mutation (Figure 1). We also compared the sequence of the Saccharomyces cerevisiae transmembrane region to that of human Hrd1, human gp78, Schizosaccharomyces pombe Hrd1 and Yarrowia lipolytica Hrd1, to identify conserved residues, as they might be expected to have key roles in Hrd1 action. In total, 77 distinct Hrd1 mutants were created in which a single amino acid codon was changed to alanine. In some cases, tandem codons were altered to two alanines. If Hrd1p participated in the specific detection of substrates, we reasoned it should be possible to find mutants deficient in degradation of distinct classes of substrates, or perhaps even deficient in degradation of individual substrates. In contrast, the C399S RING mutant is unable to degrade all ERAD substrates ((Bays et al., 2001a) and see below).
We tested each Hrd1p mutant to evaluate the altered residue’s importance in Hrd1p-dependent ERAD. Specifically, we evaluated ERAD-M, ERAD-L, and Hrd1p self-catalyzed degradation, since all three modes of HRD-dependent degradation have distinct rules and requirements (Carvalho et al., 2006; Gardner et al., 2000; Vashist and Ng, 2004). To assess ERAD-M, we tested the degradation of the integral ER membrane protein Hmg2p by each Hrd1p variant. We used a non-catalytic Hmg2p-GFP, which allows evaluation of protein stability by flow cytometry or immunoblotting (Gardner and Hampton, 1999). To evaluate ERAD-L, each Hrd1p mutant was screened with KWW (Vashist and Ng, 2004). KWW is an engineered substrate with a misfolded lumenal domain, that enters the HRD pathway by ERAD-L. To evaluate Hrd1p self-degradation, we expressed each in a hrd3Δ null strain. In the absence of Hrd3p, Hrd1p undergoes extremely rapid degradation catalyzed by its own RING domain (Gardner et al., 2000). This self-degradation has been posited to be important for Hrd1p regulation, and appears to be distinct from both ERAD-L and ERAD-M (Carroll and Hampton, manuscript in preparation). Thus, to test these aspects of Hrd1p function, each individual Hrd1p mutant was transformed into hrd1Δ strains expressing either Hmg2p-GFP or KWW, or a hrd3Δ strain, allowing examination of the mutant’s effects on ERAD-M, ERAD-L, and Hrd1p self-degradation, respectively. The effect of each Hrd1p mutant on substrate stability was assayed by cycloheximide chase, in which log phase cultures were treated with cycloheximide to stop protein synthesis, followed by flow cytometry or immunoblotting to determine substrate degradation rate. Interesting mutants were then studied further with other substrates and assays.
One group of mutants showed a clear specificity for integral membrane, or ERAD-M, substrates. Hrd1p variants L74A Hrd1p, E78A Hrd1p and W123A Hrd1p were all impaired in the degradation of Hmg2p-GFP (Figure 2A). The degradation phenotypes of these mutants were not strong like a hrd1Δ null, which causes complete stabilization of Hmg2p-GFP (Bays et al., 2001a). Instead, the steady state levels of the substrate were elevated, with the normalized degradation rates being similar to wild type. This behavior of hypomorphic HRD pathway mutants can be observed in other cases, for instance in ubx2Δ or usa1Δ ((Neuber et al., 2005; Schuberth and Buchberger, 2005) and Figure S5B, S5C). Although there is still degradation, the efficiency of the pathway appears to be lowered so that a higher steady-state pool is required for the same degradation rate. We next examined the degradation of two other ERAD-M substrates, 6myc-Hmg2p-GFP and Pdr5*. 6myc-Hmg2p-GFP is a misfolded version of Hmg2p which does not respond to the degradation signals of the sterol pathway, and thus is constitutively degraded (Hampton et al., 1996). All three Hrd1p mutants also stabilized 6myc-Hmg2p-GFP and Pdr5* to varying degrees (Figure 2B, 2C), with a particularly strong effect on 6myc-Hmg2p. However, their deficiencies were limited to only integral membrane proteins. We examined the degradation of both the prototype substrate CPY* and KWW and found that each ERAD-M deficient mutant was fully competent for degradation of ERAD-L substrates (Figure 2D, 2E), with no change in steady-state level or degradation rate. In all cases, the levels of each Hrd1p variant was identical to wild type, and each underwent normal, rapid degradation in the absence of Hrd3p (data not shown). Thus, the residues mutated in L74A Hrd1p, E78A Hrd1p and W123A Hrd1p were required for optimal ERAD-M, yet were dispensable for the degradation of misfolded lumenal substrates and Hrd1p self-degradation.
The above mutants showed a selective deficiency for membrane-associated substrates, without any effect on lumenal ones. This implied that ERAD-M selectivity might be determined by distinct information in the Hrd1p protein. In the next set of mutants, the specificity was even more striking, revealing distinct Hrd1p transmembrane determinants for recognition of different ERAD-M substrates. Two primary Hrd1p mutants, S97A/S98A, and D199A, were partially defective in Hmg2p-GFP degradation (Figure 3A). When combined, the resulting triple mutant, S97A S98A D199A Hrd1p (3A-Hrd1p) showed a strong Hmg2p-GFP degradation block (Figure 3B), nearly identical to that of the non-functional C399S. 3A-Hrd1p showed a similarly strong defect in degradation of the related substrate 6myc-Hmg2p-GFP (Figure 3C).
Hrd1p is also in complex with components that mediate retrotranslocation (Carvalho et al., 2006; Denic et al., 2006). To evaluate where in the ERAD pathway 3A-Hrd1p-mediated stabilization of Hmg2p occurred, we directly tested 3A-Hrd1p for Hmg2p-GFP ubiquitination (Bays et al., 2001a). Hmg2p-GFP was ubiquitinated by wild type Hrd1p, but not C399S Hrd1p or 3A-Hrd1p (Figure 3D). This defect was not alleviated by addition of zaragozic acid, which increases the physiological signal for Hmg2p degradation (Hampton and Bhakta, 1997).
Alteration of these three specific amino acids may have produced a hypomorphic Hrd1p mutant. To test whether higher protein expression could complement the Hmg2p degradation deficiency, we overexpressed 3A-Hrd1p by placing it behind the strong TDH3 promoter. This resulted in an approximately twenty-fold increase in Hrd1p levels above the native promoter (data not shown). Nevertheless, Hmg2p-GFP degradation by overexpressed 3A-Hrd1p was still greatly impaired (Figure 3E) compared to degradation by wild type Hrd1p. Thus, 3A-Hrd1p seemed to be intrinsically defective in Hmg2p degradation, even at high levels of this variant.
We examined the degradation of a number of ERAD-L proteins by 3A-Hrd1p. In striking contrast to Hmg2p, the degradation of the ERAD-L substrates CPY*, KHN and KWW were all degraded normally by 3A-Hrd1p, while showing the expected stabilization in hrd1Δ strains (Figure 4A, 4B, 4C). Thus, 3A-Hrd1p was completely competent for the degradation of lumenal ERAD substrates, despite its near-null phenotype with Hmg2p degradation.
We tested 3A-Hrd1p with the ERAD-M substrates Pdr5* and Sec61-2p, anticipating that they would show a similarly strong block in degradation, like the weaker ERAD-M specific mutants (see above, Figure 2). Surprisingly, Pdr5* was degraded identically by wild type or 3A-Hrd1p (Figure 4D), but showed the expected stabilization by C399S Hrd1p. Sec61p is an essential ER protein which mediates protein translocation. Strains with the sec61-2 mutation are temperature sensitive, due to Hrd1p-mediated degradation of Sec61-2p (Biederer et al., 1996). When Hrd1p is non-functional, sec61-2 strains will grow at the normally non-permissive temperature 37°C, so a commonly utilized assay of Sec61-2p degradation is growth of sec61-2 strains at the non-permissive temperature (Biederer et al., 1996; Flury et al., 2005). sec61-2 strains expressing either wild type Hrd1p, C399S Hrd1p or 3A-Hrd1p were grown at 30°C and 37°C. All strains grew at similar rates at 30°C. sec61-2 strains with wild type or 3A-Hrd1p were severely impaired for growth at elevated temperatures (Figure 4E), while those with the non-functional C399S Hrd1p showed robust growth at elevated temperatures (Bordallo et al., 1998). This phenotype was confirmed through biochemical analysis of Sec61-2p stability upon cycloheximide addition. Both wild type Hrd1p and 3A-Hrd1p were capable of Sec61-2p degradation (Figure 4E). Finally, 3A-Hrd1p stability was tested both in the presence and absence of Hrd3p by cycloheximide chase. Like wild type Hrd1p, 3A-Hrd1p was stable in the presence of Hrd3p and underwent rapid degradation in a hrd3Δ strain (Figure 4F) that was RING dependent (Figure S1). Taken together, these data show that 3A-Hrd1p was impaired only in the degradation of Hmg2p-GFP (Figure 3B) or Hmg2p variants like 6myc-Hmg2p-GFP (Figure 3C), but efficiently degraded other ERAD-M and ERAD-L substrates, and itself.
To get a broader sense of the degree to which 3A-Hrd1p functions normally, we evaluated the unfolded protein response (UPR) in strains harboring the 3A mutant. Loss of Hrd1p results in increased signaling through the UPR pathway (Friedlander et al., 2000). Using a GFP reporter for UPR (Bays et al., 2001b), we found that strains with the 3A-Hrd1p mutant had wild type levels of UPR activity, while the C399S mutant strain had the expected increase in this signaling pathway (Figure S2). Thus, by this measure also, the 3A mutant showed normal function in an assay that requires recognition of what is presumably a wide variety of misfolded proteins that are typically generated during the course of normal ER function.
3A-Hrd1p has alterations in three hydrophilic amino acids that make it incapable of recognizing Hmg2p as a misfolded protein while maintaining essentially wild type degradation of itself and all other ERAD substrates tested. One interpretation of this observation is that distinct residues in the Hrd1p transmembrane domain mediate recognition of a given ERAD-M substrate, presumably through interactions with features of the protein that hallmark misfolding or aberrant assembly. If that was the case, we speculated that other Hrd1p transmembrane mutants in our collection that degrade Hmg2p normally would have deficiencies in degradation of a distinct ERAD-M substrate, due to loss of residues needed for specific recognition of that protein. Accordingly, we re-screened our collection of Hrd1p mutants for the inability to degrade the ERAD-M substrate Pdr5*. We found two such candidates. Strains expressing only R128A Hrd1p were impaired for Pdr5* degradation (Figure 5A S3A), yet were fully proficient for Hmg2p-GFP degradation (Figure 5B, S3B). Similarly, L209A showed a strong bias towards Pdr5* with a defect that rivaled C399S (Figure 5C, S3C). In contrast, degradation of Hmg2p-GFP was slightly compromised, showing a small increase in steady-state levels but a wild type degradation rate when quantified by flow cytometry (Figure 5D, S3D). Sec61-2p degradation by L209A was also similar to wild type Hrd1p as measured by the growth phenotype of the sec61-2 strain and cycloheximide chase (Figure 5E, S3E). CPY* degradation and Hrd1p self-degradation were also only slightly impaired in the L209A mutant (Figure S3F, S3G). Thus, the L209A mutant has a specific lesion that is orthogonal to that of the 3A-Hrd1p mutant: Pdr5* is stabilized to the same extent as the C399S RING mutant, while Hmg2p or Sec61-2p degradation was only very slightly affected.
The notion that the Hrd1p transmembrane domain mediates the recognition of ERAD-M substrates was further strengthened by a third mutant with a strong bias towards the final test substrate, Sec61-2p. Upon screening the collection of variants, a single point mutant, L61A, was found that partially stabilized only this substrate as measured through both growth of sec61-2 strains at the non-permissive temperature and biochemical analysis (Figure 5F, S4A). There was no effect of this mutant on Pdr5* or Hmg2p-GFP degradation (Figure 5G, 5H, S4B). To verify that the Sec61-2p phenotype was not due to a temperature-specific defect of L61A Hrd1p, we examined L61A Hrd1p stability, Hmg2p-GFP degradation, and Pdr5* degradation at 37°C. In all cases, L61A Hrd1p behaved like wild type Hrd1p (Figure S4C, S4D, S4E). Taken together, the unique substrate specificities of 3A-Hrd1p, L209A Hrd1p, and L61A Hrd1p, indicate that the Hrd1p transmembrane domain mediates recognition of ERAD-M substrates.
The above experiments indicate that ERAD-M substrate specificity requires information in the Hrd1p transmembrane domain. We focused our attention on the 3A-Hrd1p mutant to further understand the mechanism of ERAD-M substrate recognition mediated by the Hrd1p transmembrane domain.
In studies concerning the recognition of misfolded ER proteins, a number of ERAD complex members have been implicated in the degradation of lumenal substrates. These include Yos9p, Der1p, Hrd3p, Usa1p, and Ubx2p (Carvalho et al., 2006; Denic et al., 2006). In contrast, Hrd1p-dependent degradation of membrane proteins can proceed, often at wild type rates, in the absence of these proteins. For example, Hmg2p degradation proceeds normally in a yos9Δ null (Figure 6A), and in the hrd3Δ null if levels of Hrd1p are sufficiently elevated to overcome its rapid degradation (Gardner et al., 2000). As a test of the autonomy of Hmg2p recognition mediated by the mutated 3A-Hrd1p residues, we compared the 3A mutant to wild type Hrd1p in a number of ERAD component null strains, using cycloheximide chases as above. The nulls included yos9Δ, usa1Δ, ubx2Δ, der1Δ, and hrd3Δ. In all cases, the strong stabilizing phenotype of 3A-Hrd1p was unaffected, while Hmg2p-GFP degradation occurred in wild type Hrd1p expressing strains. The normalized data are presented in Figure 6(A-D) and the western blots are shown in Figure S5 (A-D). In the usa1Δand ubx2Δstrains, a mild hypomorphic HRD phenotype was observed as elevated Hmg2p-GFP steady state levels in the blots (Figure S5B, S5C). Nevertheless, the 3A-Hrd1p mutant had strong effects on degradation in all cases. In the hrd3Δ null, we overexpressed Hrd1p to overcome the drastic loss of the protein that occurs in the absence of Hrd3p, and used flow cytometry to analyze the effects on the Hmg2p-GFP substrate (Figure 6E). In all cases, the striking difference between Hrd1p and 3A-Hrd1p in Hmg2p-GFP degradation was evident, and not dependent on any of the HRD components tested.
The Hrd1p ligase is in complex with a number of proteins including the factors tested as nulls above (Carvalho et al., 2006; Denic et al., 2006). The ability of 3A-Hrd1p to efficiently degrade all non-Hmg2p substrates implied that the HRD complex was intact. We directly examined documented interactions between Hrd1p and a number of HRD complex members. Using a native co-immunoprecipitation (co-IP) derived from the previous studies of the HRD complex (Gardner et al., 2000), we tested the interactions of 3HA-tagged native or 3A-Hrd1p with Yos9p-Flag, Hrd3p-3HA, Ubx2p-3HA, Usa1p-6HA, and Protein A-Cdc48p (Denic et al., 2006; Gardner et al., 2000; Sato and Hampton, 2006; Schuberth and Buchberger, 2005). Lysates of strains co-expressing each tagged pair were prepared and then microsomes were isolated and immunoprecipitated with polyclonal anti-Hrd1p antibodies, or in the case of Cdc48p, IgG-Sepharose beads. This was followed by SDS-PAGE and immunoblotting for the Hrd1p or test proteins indicated (Figure 6F-J). For each experiment, a strain containing an empty vector (instead of a Hrd1p expressing plasmid) was included as a control. The intensity of each band was measured by a Typhoon 9400 and ImageQuant 5.2 software and the values for these measurements are depicted. The interaction of Hrd1p and 3A-Hrd1p with the ERAD factors tested was similar when the slightly lower steady-state levels of 3A-Hrd1p were normalized for, by calculating the ratio of Hrd1p and the indicated ERAD component. As negative controls, we tested the binding of Hrd1p to two ER membrane proteins, that are neither HRD complex members, nor HRD substrates. These were the E2, 3HA-Ubc6p, and Ste6-166p-3HA-GFP, which are both Doa10p substrates (Huyer et al., 2004; Swanson et al., 2001). Neither interacted with Hrd1p (Figure 6K, 6L), demonstrating that the association between wild type or 3A-Hrd1p and the Hrd1p-associated complex members was specific.
We next turned our attention to the role the residues defined by the 3A mutation played in Hmg2p recognition. Clearly, E3 ligases must bind a targeted substrate to transfer ubiquitin. However, it is not generally known if substrates need only to bind to ligases, or if in addition, the substrate must activate or transmit information to the ligase to bring about robust poly-ubiquitination. Our cross-linking studies indicate that for Hrd1p, the latter model might be the case (Gardner et al., 2001b). Hrd1p cross-links degraded substrates Hmg2p and 6myc-Hmg2p, but also non-degraded K6R-Hmg2p, or the highly stable homologue Hmg1p with similar efficiencies (Gardner et al., 2001b). Likewise, improving the folding of Hmg2p with the chemical chaperone glycerol prohibits Hrd1p-dependent ubiquitination (Garza and Hampton, manuscript in preparation), but not Hmg2p-Hrd1p cross-linking (Gardner et al., 2001b). Thus, substrate interaction by this criterion appears to be insufficient for Hrd1p-mediated ubiquitination.
We tested whether Hmg2p-GFP was capable of interacting with 3A-Hrd1p by two approaches. We first utilized an in vitro cross-linking assay, in which ER-enriched microsomes were harvested from cells expressing Hmg2p-GFP with wild type or 3A-Hrd1p tagged with triple HA. The lipid-soluble cross-linker DSP was added to the microsomes, followed by an anti-GFP immunoprecipitation. The precipitated protein mixture was then immunoblotted for Hmg2p-GFP or Hrd1p-3HA after SDS-PAGE. Both wild type and 3A-Hrd1p associated with Hmg2p-GFP in a cross-linker-dependent manner (Figure 7A).
We also evaluated the Hmg2p-Hrd1p interaction with a non-denaturing co-IP assay. Microsomes were isolated and added to a 1.5% Tween-20 lysis buffer. An anti-GFP antibody was then added to the lysates in order to immunoprecipitate Hmg2p-GFP, and co-precipitated Hrd1p was then detected with anti-HA immunoblotting. Under these conditions, Hmg2p-GFP bound to wild type or 3A-Hrd1p with equal efficiency (Figure 7B). This interaction was specific, as a control immunoprecipitation with a strain lacking Hmg2p-GFP was unable to pull down Hrd1p. Conversely, two integral membrane proteins (Ubc6p, Ste6-166p) that are substrates of the Doa10 ERAD pathway, failed to co-precipitate with Hrd1p (Figure 6K, 6L). Thus, 3A-Hrd1p was capable of binding to Hmg2p-GFP. Identical results were observed in both UBC7 (data not shown) and ubc7Δ strains (Figure 7B). The ubc7Δ strains were necessary in order to test for in vitro substrate ubiquitination as described next.
We used the same microsomes prepared for the native co-IP experiment to directly evaluate the ability of the 3A-Hrd1p to ubiquitinate Hmg2p-GFP in an in vitro ubiquitination assay. In this assay, microsome strains lacking Ubc7p and cytosol strains with or without Ubc7p, but lacking Hrd1p and Hmg2p-GFP, were utilized as described ((Flury et al., 2005) and in methods). Microsomes, cytosol, and ATP were incubated for 1 hour at 30°C, allowing substrate ubiquitination followed by IP and immunoblotting for ubiquitin and Hmg2p-GFP. Wild type Hrd1p ubiquitinated Hmg2p-GFP in vitro while 3A-Hrd1p did not (Figure 7C). This defect was specific for Hmg2p, as 3A-Hrd1p demonstrated self-ubiquitination in the same in vitro reaction and was also capable of transferring ubiquitin to Pdr5* (Figure 7C). In vitro ubiquitination was Hrd1p, Ubc7p, and Hrd1p RING dependent (Figure S6, Figure 7C). Although Hrd1p and Hmg2p-GFP binding was unaffected by the introduction of the 3A mutations, 3A-Hrd1p could not ubiquitinate Hmg2p-GFP, indicating that substrate binding alone is insufficient to trigger Hrd1p-dependent ubiquitination.
The recognition of misfolded proteins is a central and unresolved process in all protein quality control pathways. The diversity of substrates indicates that structural features serve as criteria for substrate recognition. Since E3 ligases are critical definers of specificity, misfolded substrate detection must include, in some manner, the ligase.
For Hrd1p, both lumenal (ERAD-L) and integral membrane (ERAD-M) substrates are targets for degradation. Recognition of ERAD-L substrates relies heavily on Hrd1p-associated factors such as Yos9p, Kar2p, and the lumenal domain of Hrd3p (Carvalho et al., 2006; Denic et al., 2006).
In contrast, Hrd1p appears to mediate the recognition of integral membrane substrates, as is demonstrated by the discovery of mutants highly selective for individual ERAD-M substrates. Thus, extreme specificity for ERAD-M substrate recognition lies in the transmembrane domain of Hrd1p. This ability of the Hrd1p transmembrane domain to discern substrates was an autonomous feature of the protein: the high selectivity of the residues altered in the 3A mutant was not dependent on any ERAD factors tested.
Interestingly, none of our mutants had selective defects in ERAD-L. This implies that distinct rules govern recognition of ERAD-M substrates, although a more complete analysis of Hrd1p is needed to fully test the idea that ERAD-L recognition lies outside of the membrane anchor. It is worth noting that the ERAD-C pathway responsible for degrading proteins with misfolded cytosolic domains requires cytosolic chaperones for Doa10p-dependent ubiquitination (Nakatsukasa et al., 2008). Thus, the lack of involvement of such factors, and the autonomous requirement for the Hrd1p membrane domain indicates that ERAD-M employs recognition rules distinct from those used for soluble determinants of misfolding.
3A-Hrd1p is essentially a phenocopy of a C399S RING mutant, but only for Hmg2p-related substrates. Both cross-linking and co-IP assay showed that the 3A-Hrd1p-Hmg2p interaction was intact. Importantly, Hrd1p was unable to co-precipitate two integral membrane DOA-pathway substrates. Thus, although the co-IP assay showed the appropriate specificity for Hmg2p-GFP, it was not affected in the strong 3A mutant. However, in vitro 3A-Hrd1p did not support ubiquitination of Hmg2p, yet still functioned in vitro as a ligase, showing normal self and Pdr5* ubiquitination. Thus, the high specificity of the 3A mutation is not due to any measurable loss of interaction with Hmg2p, but rather to an inability of the still-active 3A-Hrd1p mutant to transfer ubiquitin to Hmg2p that is in its proximity.
In our earlier interaction studies, we had noted that Hrd1p was able to associate with potential substrates in a fairly indiscriminant manner. We proposed that Hrd1p queries a variety of proteins by low-specificity interactions, and only when a given substrate has the appropriate structural features does ubiquitination occur. Likewise, while 3A-Hrd1p binds to Hmg2p, it appears to be incapable of transmitting the structural information to the RING domain to promote ubiquitination, thus supporting an “allosteric” model for Hrd1p selectivity, in which the specificity of substrate ubiquitination lies downstream of low affinity ligase-substrate interactions (Figure S7).
Alternatively, it may be that the residues altered in the 3A mutant do indeed inhibit a high affinity interaction with Hmg2p not detected by our assays, although the same techniques have been used successfully to detect substrate-E3 interactions that do determine selectivity (Denic et al., 2006; Schuberth and Buchberger, 2005). However, we favor the model of structural evaluation determined subsequent to low specificity engagement of substrate with the HRD complex. At least for quality control substrates, this strategy of “general interaction-specific response” makes some teleological sense. A quality control ligase that is over-dedicated to interacting with only a particular type of substrate would not be efficient in the general detection of the very large number of possible misfolded proteins that it might encounter.
The detection of ERAD-M substrates, that is, misfolded or unassembled membrane proteins, might be expected to follow rules distinct from those used to detect ERAD-L substrates. A misfolded aqueous protein would display a larger-than-normal proportion of surface hydrophobic residues, and indeed, proteins that detect misfolded soluble proteins, such as chaperones or UGGT (Dejgaard et al., 2004), have regions that can interact with exposed hydrophobic regions of their clients. Conversely, misfolded integral membrane proteins perhaps expose normally buried hydrophilic residues to the lipid region of the bilayer. Detection of these inappropriate residues could be accomplished by interaction with membrane-embedded hydrophilic residues on the ligase. Consistent with this, the 3A mutant of Hrd1p has three intramembrane residues changed from S, S and D to alanine. Extensive Hmg2p structural analysis has also uncovered conserved hydrophilic residues within the transmembrane spans that, when mutated, result in a completely stable protein (Davis and Hampton, manuscript in preparation). Similarly, it has been suggested that the TUL1 ubiquitin ligase recognizes unassembled membrane protein clients through interactions between hydrophilic intramembrane residues (Reggiori and Pelham, 2002). It is interesting to note that the quality control ligases mammalian Hrd1, gp78, and yeast Doa10p each have a high density of intramembrane hydrophilic residues, as would be expected if hydrophilic scanning was a general strategy for membrane substrate evaluation. These examples indicate that recognition of hydrophilic intramembrane residues on a misfolded protein by similar residues within a ligase may be a broadly employed strategy.
Direct loss of such “hydrophilic scanning” residues is probably not the only lesion in some of our mutants. L209A Hrd1p, which has slight general ERAD defects but is completely incapable of Pdr5* degradation has a missing intramembrane leucine, and presumably this alteration creates a structural change that specifically alters Pdr5* binding or evaluation. Similarly the Sec61-2-selective L61A Hrd1p mutant has an intramembrane hydrophobic residue replaced by alanine. It may be that the changes of hydrophobic residues alter the position of key hydrophilic residues, or that other structural features that assist in substrate detection are being altered. With regard to the question of the functions of these residues, it will be most revealing when the structure of the Hrd1p transmembrane domain is solved by high resolution techniques in the future.
Taken together, these studies imply that the Hrd1p transmembrane domain specifically mediates ERAD-M. It appears that the transmembrane domain bears a structural code for detection of features that hallmark a degradation substrate, and this information appears to be multi-faceted, so that the loss of recognition of a single substrate class can be observed in the appropriate mutant. Eventually, unraveling this code will help us understand the rules by which misfolded proteins are led to their destruction.
Escherichia coli DH5α were grown in LB media with ampicillin. Yeast strains were grown at 30°C unless noted in minimal media supplemented with dextrose and amino acids (Hampton and Rine, 1994). A complete description of strain construction and a list of all parent strains, the plasmids transformed into them, and the figures in which they were utilized are listed in the supplemental materials.
Cycloheximide chase degradation assays and flow cytometry were performed as previously described (Sato and Hampton, 2006). All non-GFP strain quantitation was performed using a Typhoon 9400 and ImageQuant 5.2 software (GE Healthcare, Piscataway, NJ). In these cases, a representative western blot was quantified and the data points were graphed.
Ubiquitination of Hmg2p-GFP was examined in log phase cells as previously described (Bays et al., 2001a) by immunoprecipitation (IP) followed by ubiquitin or substrate immunoblotting. A more detailed description can be found in the supplemental materials.
Cross-linking was modified from that used by Gardner et al. (Gardner et al., 2001b; Gardner et al., 2000) and is described in detail in supplemental materials. Microsomes were harvested in B88 buffer and incubated with the cross-linker DSP for 40 minutes at 22°C. The cross-linker was quenched in 50mM Tris pH 7.5, the microsomes were centrifuged, and lysed in SUME lysis buffer plus 1% Triton-X100 and 0.5% DOC. IP buffer was then added to each sample along with anti-GFP antibody. The remainder of the IP was performed as described.
The native co-IP assay was adapted from Gardner et al., 2000 and described in supplemental materials. Briefly, microsomes were harvested as in the cross-linking protocol, except it was performed in MF buffer. Pelleted microsomes were resuspended in 1ml of Tween buffer and incubated on ice for 15 minutes. Lysates were then centrifuged for 30 minutes at 14,000 x g. The remainder of the IP was performed as described except the Tween-20 buffer was utilized for washes.
Growth of sec61-2 strains was measure by dilution assay, performed as described (Sato and Hampton, 2006) and in the supplement.
In vitro ubiquitination assays were performed as described (Flury et al., 2005). Briefly, ubc7Δ microsome strains, containing TDH3-Hrd1-3HA (wild type or 3A) and the indicated substrate were utilized. Microsomes were prepared identically as in the native co-IP experiments and were resuspended in B88 buffer. Cytosol strains underwent freeze-thaw lysis in B88 buffer and ultracentrifuged. For each ubiquitination reaction, a microsome strain was combined with 30mM ATP and cytosol that either did or did not contain TDH3-Ubc7-2HA for 1 hour at 30°C. The IP was then performed as described.
We thank Alexander Buchberger (Max Planck Institute of Biochemistry), Davis Ng (National University of Singapore), Randy Schekman (University of California, Berkeley), Thomas Sommer (Max-Delbruck Center for Molecular Medicine), Jonathan Weissman (University of California, San Francisco) and Dieter Wolf (University of Stuttgart) for providing plasmids, strains and antibodies, Maho Niwa (University of California, San Diego) for use of the Typhoon 9400, and Michael David (University of California, San Diego) for use of the flow microfluorimeter, despite our preference for avid yeast. We also thank the Hampton laboratory for enthusiastic and incisive discussions and technical assistance. Finally, RYH would like to thank his brothers Ron and Bob for lifelong sibling support, spiritual growth, and genetic connection. These studies were supported by NIH grant 5 R01 DK051996-15 to RYH. BKS was a trainee under the NIH Genetic Training grant # 5T32GM008666.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.