Membrane chaperone Shr3 assists in folding amino acid permeases preventing precocious ERAD

Early in the secretory pathway of eukaryotic cells, integral plasma membrane (PM) proteins are inserted into the lipid bilayer of the ER. The mechanisms enabling polytopic membrane proteins, composed of multiple transmembrane segments (TMSs), to integrate and fold in the ER membrane are not well documented. It is known that hydrophobic amino acids within N-terminally localized TMSs are sufficient to target membrane proteins to the ER, where they interact with and enter a protein-conducting channel, or translocon, with a hydrophilic core. The translocon is a highly conserved heterotrimeric integral membrane complex, i.e., the Sec61 complex in eukaryotes and SecY complex in bacteria (Osborne et al., 2005).

The successful ability to crystallize membrane proteins has greatly enhanced the understanding of two basic processes associated with membrane protein biogenesis. First, the structure of the Methanococcus jannaschii SecY translocon has been elucidated (van den Berg et al., 2004), which revealed that the protein-conducting channel is contained within SecY, the largest translocon subunit and the Sec61 homologue. Despite unresolved questions regarding the structure of an active Sec61/ SecY translocon in vivo (Alder and Johnson, 2004), the protein-conducting channel, although perhaps quite flexible, is too small to accommodate multiple TMSs (Rapoport et al., 2004; Osborne et al., 2005). Second, the structures of bacterial PM transporters LacY and GlpT, each composed of 12 TMSs (Abramson et al., 2003; Huang et al., 2003), have clearly shown that interactions between TMSs within the lipid phase of the membrane are not limited to immediately flanking TMSs. Consequently, to fold properly, N-terminally localized TMSs that partition into the membrane of partially translated proteins must await the insertion of C-terminal TMSs. Together, these advances are consistent with the idea that polytopic membrane protein biogenesis occurs essentially in two discrete stages, i.e., membrane insertion and folding (Popot and Engelman, 1990; Engelman et al., 2003; Bowie, 2005).

To bridge these stages, and to account for the fact that polytopic membrane proteins cannot fold in a strict cotranslational manner, the existence of chaperone-like proteins has been postulated (Lecomte et al., 2003; Alder and Johnson, 2004; Rapoport et al., 2004). The identification of membrane-localized chaperones in bacteria (Nagamori et al., 2004) and yeast (Kota and Ljungdahl, 2005) provided strong support for this notion. Consistent with the two-stage model of biogenesis, these chaperones do not appear to be necessary for the insertion of TMSs (Gilstring and Ljungdahl, 2000; Nagamori et al., 2004) but, nonetheless, they appear to interact early during the translocation process to enable their substrate proteins to obtain native conformations (Nagamori et al., 2004; Kota and Ljungdahl, 2005). Although it remains to be determined, this novel class of chaperones may prevent TMSs of polytopic membrane proteins that do not normally interact in the mature protein from engaging in nonproductive interactions with flanking TMSs, or with other ER components, as they sequentially partition into the membrane.

In the yeast Saccharomyces cerevisiae, the recently characterized membrane-localized chaperones exhibit a striking degree of substrate specificity. The best studied of these chaperones, Shr3, is specifically required for proper folding of amino acid permeases (AAPs; Kota and Ljungdahl, 2005). AAPs comprise a conserved family of 18 proteins with 12 TMSs (Gilstring and Ljungdahl, 2000). AAPs localize to the PM in an Shr3-dependent manner (Ljungdahl et al., 1992), where they function to transport amino acids into cells. In the absence of Shr3 chaperone activity, AAPs aggregate, forming large molecular weight complexes (Kota and Ljungdahl, 2005), which are excluded from coatomer protein II transport vesicles (Kuehn et al., 1996, 1998). Consequently, AAPs accumulate in the ER of shr3-null mutant strains. Shr3 has two well-defined domains, a membrane domain composed of four TMSs and a hydrophilic C-terminal domain. The chaperone activity of Shr3 is associated with the membrane domain; the hydrophilic C-terminal domain is dispensable for function (Kota and Ljungdahl, 2005). The steady-state levels of AAPs are similar in both wild-type and shr3-null mutant cells (Ljungdahl et al., 1992; Gilstring et al., 1999; Kota and Ljungdahl, 2005). In wild-type cells, AAPs are degraded in the vacuole (Roberg et al., 1997; Springael and André, 1998); the fate of aggregated AAPs in shr3 mutants has not been investigated.

PM and secretory proteins are subject to ER quality-control systems that operate to ensure that only properly folded and assembled proteins exit the ER and enter into subsequent stages of the secretory pathway (Ellgaard et al., 1999). The ER lumen contains numerous factors that assist folding reactions (Nishikawa et al., 2005; Bukau et al., 2006). Terminally misfolded, or incompletely assembled soluble or integral membrane proteins are degraded in a process referred to as ER-associated degradation (ERAD; Meusser et al., 2005). ERAD substrates are passed through or extracted from the membrane, ubiquitylated, and ultimately presented to cytoplasmic proteasomes for degradation. Several ERAD pathways have been defined (Huyer et al., 2004; Vashist and Ng, 2004; Ravid et al., 2006). These pathways are composed of multimeric protein complexes organized around two membrane-localized E3 ubiquitin ligases, i.e., Doa10 and Hrd1 (Carvalho et al., 2006; Denic et al., 2006). These pathways appear to differentially recognize ERAD substrates, and thus far, all characterized ERAD substrates exhibit a strong dependence on either Doa10 or Hrd1 pathways. With respect to polytopic membrane proteins, little is known regarding how ERAD surveillance mechanisms differentiate between proteins in the process of membrane insertion and folding from proteins that are terminally misfolded. The potential involvement of membrane-localized chaperones in ERAD has not been examined.

We have investigated the chaperone activity of Shr3 from two perspectives. Using split Gap1 constructs, we examined the temporal requirement of Shr3 during membrane insertion and folding. Our results indicate that Shr3 interacts with an N-terminal fragment composed of TMSs I–V and maintains it in a conformation that enables it to functionally assemble with a coexpressed C-terminal fragment with TMSs VI–XII. Second, we assessed the potential role of Shr3 in quality-control mechanisms that monitor AAP folding. We report that in the absence of Shr3, Gap1 aggregates are redundantly targeted to Doa10- and Hrd1-dependent ERAD pathways. Cells lacking these ERAD pathways exhibit restored amino acid uptake capacities. Thus, given sufficient time, AAPs are able to attain functional conformations independent of Shr3. These findings highlight the intimate link between folding and degradation during the biogenesis of polytopic membrane proteins.

Previous studies examining the ER retention of AAP in shr3 mutants have relied on microscopic evaluation (Ljungdahl et al., 1992; Kota and Ljungdahl, 2005). The intracellular distribution of Gap1 in wild-type and shr3Δ mutant strains was examined by subcellular fractionation of whole cell lysates. In lysates from wild-type cells (Fig. 1, left) the majority of Gap1 cofractionated with the PM marker protein Pma1 (fractions 2 and 3) and the late Golgi marker Kex2 (peak fraction 6). The distribution of Gap1 was distinct from the ER marker Wbp1, which exhibited the highest concentration in fraction 5. Gap1 migrated as two bands: a top band present in PM-containing fractions and a bottom band that was observed in fractions containing internal membranes. Gap1 is known to be posttranslationally modified by phosphorylation, which correlates with increased amino acid uptake (Stanbrough and Magasanik, 1995). Also, phosphorylation is important for proper targeting of Gap1 from the Golgi to the PM and for preventing Gap1 down-regulation (De Craene et al., 2001). The pattern of mobility we observed likely correlates with Gap1 becoming phosphorylated as it progresses through the secretory pathway on its way to the PM.

In lysates from shr3Δ cells (Fig. 1, left) the bulk of Gap1 (bottom band form) cofractionated with the ER marker Wbp1 (fractions 5 and 6), and there was minimal overlap with the Golgi marker Kex2 (peak fraction 7). Interestingly, a small amount of Gap1 (top band form) cofractionated with Pma1, the PM marker protein (fractions 2 and 3), suggesting that a portion of Gap1 is able to fold independently of Shr3. However, based on our extensive phenotypic analysis of shr3 mutants, including quantitative amino acid uptake assays, microscopic analysis (indirect immunofluorescence and Gap1-GFP), and biochemical approaches, the fraction of folded and correctly localized Gap1 in shr3 mutant cells is too low to confer detectable Gap1- dependent phenotypes (Ljungdahl et al., 1992; Kuehn et al., 1996, 1998; Gilstring et al., 1999; Kota and Ljungdahl, 2005). These results demonstrate that Shr3 is required to enable Gap1 to efficiently exit the ER and correctly localize to the PM.

In early studies regarding polytopic membrane protein biogenesis, it was found that truncated N- and C-terminal fragments of bacteriorhodopsin could incorporate into membranes and assemble to attain catalytic active conformations (Huang et al., 1981). Subsequent to this pioneering work, there have been numerous examples of functional “split” polytopic membrane proteins reported in the literature, including bacterial, yeast, and mammalian proteins. To more fully understand the chaperone-dependent folding of AAPs, and to investigate the regions of Gap1 that require Shr3, we sought to take advantage of the possibility that Gap1 could be functionally expressed as a split protein.

The two-dimensional structure of Gap1 is depicted in Fig. 2 A. The N and C termini and even-numbered hydrophilic loops (L2–L10) are oriented toward the cytoplasm. The odd-numbered loops (L1–L11) are luminally oriented during biogenesis and extracellular when Gap1 is correctly localized to the PM. With the exception of L5 (35 amino acids) and L7 (22 amino acids), the odd-numbered loops are short and are composed of <10 amino acids. Consequently, the bulk of the non–membrane-associated amino acids are on the cytoplasmic side of the membrane.

We constructed alleles encoding two matched pairs of N- and C-terminal Gap1 fragments (Fig. 2 B). The first pair, encoded by the gap1 TM1-5 and gap1 TM6-12 alleles, expresses nonoverlapping Gap1 fragments truncated between TMS V and VI. The gap1 TM1-5 allele encodes an N-terminal Gap1 fragment composed of the first 263 amino acid residues, and the gap1 TM6-12 allele encodes a C-terminal fragment of Gap1 with a methionine residue placed immediately before residues 264–602. The second pair, encoded by the gap1 TM1-6 and gap1 TM7-12 alleles, expresses partially overlapping Gap1 fragments truncated between TMS VI and VII. The gap1 TM1-6 encoded fragment is composed of the N-terminal 320 amino acid residues, and the gap1 TM7-12 allele has a methionine residue placed before residues 311–602. Each of these alleles is expressed under control of the endogenous GAP1 promoter.

The truncated Gap1 alleles were separately introduced into the gap1Δ-null mutant strain FGY15, and the growth of transformants was assessed to determine whether the individually expressed N- and C-terminal fragments exhibited Gap1 transport activity (Fig. 2 C). Gap1 mediates the uptake of toxic d-amino acids (Regenberg and Hansen, 2000); consequently, cells lacking Gap1 activity are able to grow in media containing d-histidine (dilution series 1), whereas cells expressing functional full-length Gap1 cannot (dilution series 2). As expected, cells expressing the truncated N- and C-terminal fragments of Gap1 grew (dilution series 3–6), indicating that the individually expressed truncated proteins are nonfunctional.

Next, plasmids encoding gap1 TM1-5 and gap1 TM6-12 or gap1 TM1-6 and gap1 TM7-12 were introduced together as pairs into FGY15 (gap1Δ). The growth of both strains carrying the matched pairs of N- and C-terminal fragments was inhibited by d-histidine (Fig. 2 D; dilution series 3 and 4). All strains grew well in the absence of d-histidine (Fig. 2 D, top); thus, the trivial explanation for the lack of growth, e.g., inhibitory secondary effects resulting from the expression of truncated Gap1 constructs, can be ruled out. These results demonstrate that in both instances, the truncated N- and C-terminal fragments assemble into an active split Gap1 that correctly localizes to the PM and facilitates d-histidine uptake. A slight difference in growth of the split Gap1 expressing strains was noted; the growth of the split Gap1TM1-5/TM6-12 expressing strain (dilution series 3) was nearly as poor as the strain expressing the intact full-length GAP1 allele (dilution series 2), whereas the growth of the split Gap1TM1-6/TM7-12 expressing strain was better (dilution series 4). The noticeable difference suggests that split Gap1TM1-5/TM6-12 is more active.

To facilitate subsequent biochemical analysis of the C-terminal fragment, we inserted a thrice-reiterated myc epitope immediately preceding the stop codon of the gap1 TM6-12 allele. We examined the functionality of the C-terminal–tagged split gap1 TM6-12myc construct by coexpressing it with gap1 TM1-5 in strain FGY15 (gap1Δ). Growth of this strain (dilution 12) on media containing d-histidine was poor and similar to the untagged Gap1TM1-5/TM6-12 expressing strain (dilution 11). Because growth was inhibited by d-histidine, the myc-tagged split Gap1 was able to assemble together with the N-terminal fragment into an active Gap1.

Finally, we tested whether the untagged and tagged split Gap1 proteins require Shr3 for functional expression (Fig. 2 D; dilution series 5–8 and 13–16, respectively). The matched plasmid pairs were introduced into an shr3Δ gap1Δ strain (FGY135). Strain FGY135 is unable to take up d-histidine and grows well because of the lack of Shr3 (dilution series 5 and 13), even when full-length GAP1 is reintroduced on a plasmid (dilution series 6 and 14). The transformants expressing either of the functional untagged and tagged split Gap1 proteins exhibited robust growth (dilution series 7–8 and 15–16, respectively). The strict requirement for Shr3 indicates that the split proteins, similar to full-length Gap1, require the chaperone activity of Shr3 to attain functional conformations.

To further investigate the role of Shr3 in Gap1 biogenesis, we examined whether split Gap1TM1-5/TM6-12 and Gap1TM1-5/TM6-12myc aggregate in the absence of Shr3. Cell lysates from strains FGY15 (gap1Δ) and FGY135 (shr3Δ gap1Δ) expressing full-length Gap1, untagged split Gap1 TM1-5/TM6-12, or tagged Gap1 TM1-5/TM6-12myc were solubilized in the presence of dodecyl-β-d-maltopyranoside (DM), and soluble proteins were separated by blue native PAGE (BN-PAGE; Fig. 3). Consistent with our previous results (Kota and Ljungdahl, 2005), monomeric full-length Gap1 was readily extracted from membranes prepared from SHR3 wild-type cells (Fig. 3 A, lanes 1 and 3), but not from membranes prepared from shr3Δ cells (Fig. 3 A, lanes 2 and 4). In the absence of Shr3, Gap1 aggregates form high molecular weight complexes that migrate as a diffuse smear. Similar results were obtained when we examined untagged split Gap1 (Fig. 3 B); monomers of the N-terminal fragment were readily extracted from membranes containing Shr3 (Fig. 3 B, lanes 1 and 3), whereas fewer monomers were extracted from membranes lacking Shr3 (Fig. 3 B, lanes 2 and 4). We were unable to detect intact assembled split Gap1 monomers, indicating that DM extraction disrupts N- and C-terminal fragment association. The decreased efficiency to solubilize monomeric gap1 TM1-5 in shr3 mutant cells was not a consequence of reduced levels of gap1 TM1-5; both SHR3 and shr3Δstrains expressed similar amounts of the N-terminal fragment (Fig. 3 B, bottom).

Antibodies to native Gap1 recognize the N-terminal hydrophilic sequence preceding TMS I (anti-NT-Gap1; De Craene et al., 2001). To investigate the aggregation state of the C-terminal fragment, we coexpressed Gap1 TM1-5/TM6-12myc in SHR3 and shr3Δ cells. In a manner entirely consistent with the untagged split Gap1, the N-terminal fragment was readily extracted from membranes containing Shr3 (Fig. 3 C, lanes 1 and 3), and fewer monomers were extracted from membranes lacking Shr3 (Fig. 3 C, lanes 2 and 4). Next, we assessed whether monomers of the C-terminal fragment could be extracted (Fig. 3 D). Similar to the N-terminal fragment, C-terminal monomers were more readily extracted from membranes with Shr3 (Fig. 3 D, lanes 1 and 3) compared with membranes lacking Shr3 (Fig. 3 D, lanes 2 and 4). However, the difference between gap1 TM6-12myc monomer levels in extracts from SHR3 and shr3Δcells was not as striking as for the N-terminal fragment. The SHR3 and shr3Δ strains expressed similar amounts of the N- and C-terminal fragments (Fig. 3, C and D, bottom). These results indicate that in the absence of Shr3, the N- and C-terminal fragments of split Gap1 are prone to aggregation.

To investigate the temporal requirement of Shr3 during the assembly of split Gap, we examined the aggregation state of individually expressed N- and C-terminal fragments. In cells expressing only the N-terminal fragment, gap1 TM1-5 monomers were readily extracted from membranes with Shr3 (Fig. 4 A, lanes 1 and 3), but fewer monomers were extracted from membranes lacking Shr3 (Fig. 4 A, lanes 2 and 4). These results, analogous to what we observe when membranes with coexpressed N- and C-terminal fragments are solubilized (Fig. 3), indicate that Shr3 interacts directly with the N-terminal fragment and prevents its aggregation.

In contrast to the N-terminal fragment, we were unable to find conditions to efficiently extract gap1 TM6-12myc monomers from membranes in cells expressing only the C-terminal fragment (Fig. 4 B). Only small amounts of monomeric gap1 TM6-12myc were extracted, and importantly, the presence (lanes 1 and 3) or absence (lanes 2 and 4) of Shr3 did not affect the levels of soluble C-terminal monomers. These results indicate that in the absence of the first five N-terminal TMSs of Gap1, Shr3 is unable to associate with and prevent aggregation of C-terminal domain of Gap1 containing TMSs VI–XII. The clear Shr3 dependence of the N-terminal fragment is consistent with Shr3 interacting early during the biogenesis of Gap1, an interaction that is likely to occur before the partitioning of all 12 TMSs into the lipid phase of the ER membrane.

The observed aggregation of full-length and truncated fragments of Gap1 is the likely consequence of misfolding, which accounts for Gap1 retention and accumulation in the ER of shr3Δ mutants. Despite the accumulation of misfolded proteins, shr3Δ mutants do not exhibit an activated ER unfolded protein response (Gilstring et al., 1999). This may be explained by the fact that each of the TMSs of Gap1 are correctly inserted in the membrane independently of Shr3 (Gilstring and Ljungdahl, 2000), and that the luminally exposed hydrophilic loops are rather short (Fig. 2 A). Thus, it is likely that Gap1 aggregates do not expose sequences that are recognized by Kar2, the luminal Hsp70 BiP homologue and key regulator of the unfolded protein response (Rutkowski and Kaufman, 2004). Prototrophic shr3Δ mutants grow as well as wild-type cells, even under conditions when AAP expression is induced (Ljungdahl et al., 1992), suggesting that shr3 mutants are not negatively affected by the accumulation of AAP aggregates. To examine how shr3 mutants cope with AAP aggregates, we examined the turnover of Gap1 using pulse-chase analysis (Fig. 5). Similar to previous reports (Roberg et al., 1997; Gilstring et al., 1999), the half-life of Gap1 in wild-type cells was ∼40 min (Fig. 5, A and C, solid circles). The half-life of Gap1 in shr3Δ mutants was similar to wild-type cells (Fig. 5, B and D, solid circles). Thus, despite the fact that Gap1 is differentially localized and present in different forms, i.e., PM and monomeric in SHR3 versus ER and aggregated in shr3Δ, Gap1 is efficiently degraded. The results suggest that AAP aggregates are degraded by ERAD.

To test this possibility, we initially examined the degradation of Gap1 in pep4Δ mutants with an impaired capacity to degrade proteins in the vacuole (Jones et al., 1982). In SHR3 wild-type cells, Gap1 degradation was markedly impaired (Fig. 5 A, open squares), clearly demonstrating that in the presence of Shr3, Gap1 exits the ER and is degraded in the vacuole. This is consistent with the well-characterized trafficking pathways underlying the turnover of PM proteins (Haguenauer-Tsapis and André, 2004). In contrast, Gap1 was efficiently degraded in shr3Δ pep4Δ cells (Fig. 5 B, open squares), indicating that Gap1 aggregates are degraded independently of vacuolar hydrolases. This finding is consistent with AAP aggregates being degraded by ERAD.

We examined this directly by constructing SHR3 and shr3Δ strains carrying single- and double-null alleles of the genes encoding ER-localized E2 ubiquitin-conjugating enzymes Ubc6 and Ubc7. Ubc6 and Ubc7 are required for ubiquitylating most characterized luminal and transmembrane ERAD substrates (Meusser et al., 2005). As expected, in SHR3 cells, the rate of Gap1 degradation was unaffected in ubc6Δ, ubc7Δ, or ubc6Δ ubc7Δ strains (Fig. 5 C). However, in strains lacking Shr3, the rate of Gap1 degradation was diminished in the presence of either ubc6Δ or ubc7Δ mutations (Fig. 5 D). In comparison to ubc6Δ, the ubc7Δ mutation exhibited a more pronounced stabilizing affect. Gap1 exhibited the greatest stability in the strain carrying both ubc6Δ ubc7Δ mutations. The clear dependency on Ubc6 and Ubc7 demonstrates that aggregated Gap1 is an ERAD substrate.

Doa10- and Hrd1-dependent ERAD pathways are able to recognize and ubiquitylate membrane protein substrates. However, to date, all characterized ERAD substrates appear to be preferentially ubiquitylated by either Doa10- or Hrd1-dependent pathways. Thus, these E3 ligases appear to have distinct and nonredundant roles (Ismail and Ng, 2006). To determine which ligase is primarily responsible for ubiquitylating misfolded and aggregated Gap1, we constructed shr3Δ strains carrying doa10Δ- or hrd1Δ-null alleles. Unexpectedly, the stability of Gap1 increased only slightly and to the same extent in both of these strains (Fig. 6). These results suggested that both Doa10- and Hrd1-dependent pathways might function in a redundant manner to ubiquitylate Gap1 aggregates. This possibility was examined by following the degradation of Gap1 in an shr3Δ strain carrying both doa10Δ- and hrd1Δ-null alleles. In this strain, Gap1 was significantly stabilized, indicating that with respect to ubiquitylating Gap1 aggregates, Doa10 and Hrd1 are equally effective and, indeed, functionally redundant. Thus, multiple ERAD pathways are involved in the degradation of Gap1 aggregates.

We sought to test whether our findings regarding the turnover of Gap1 can be applied to other AAPs and chose to investigate the turnover of Agp1 (Iraqui et al., 1999). We previously showed that Agp1 aggregates and is retained in the ER of cells lacking Shr3 (Kota and Ljungdahl, 2005). Using a simplified experimental approach relying on cycloheximide, we found that, in wild-type cells, Agp1 is primarily degraded in the vacuole (Fig. 7 A); Agp1 was stabilized in a pep4Δ mutant, whereas degradation was unaffected by the combination of ubc6Δ ubc7Δ mutations. In cells lacking Shr3, the stability of Agp1 was dependent on ERAD (Fig. 7 B). In these cells, the turnover of Agp1 was largely Pep4 independent and, rather, was Ubc6 and Ubc7 dependent. As was found for Gap1, ubc7 mutations alone stabilized Agp1 aggregates almost as efficiently as ubc6 ubc7 double mutations (not depicted). Also, both Doa10- and Hrd1-dependent pathways redundantly targeted Agp1 for degradation; aggregates were degraded in doa10Δ and hrd1Δ single mutants with equal efficiency but were stable in doa10Δ hrd1Δ double mutants. The data demonstrate that Agp1 aggregates are degraded in a manner identical to Gap1 aggregates.

Little is known regarding surveillance mechanisms that monitor the folding of nonglycosylated polytopic membrane proteins, such as AAPs (Gilstring and Ljungdahl, 2000). Although assumed to exist, mechanisms that distinguish between partially folded and misfolded states have not been identified, and the potential role of membrane-localized chaperones in such quality-control mechanisms has not been examined. We suspected that the chaperone activity of Shr3 would have a critical role in discriminating between these two states and perhaps function to shield partially folded AAPs from being prematurely targeted for degradation by ERAD pathways. To evaluate this possibility, we monitored the growth of strains with impaired ERAD using two growth-based assays that provide an extremely sensitive measure of the activity and, hence, the folding state of three AAPs, i.e., Gap1 (Fig. 8 A) and Agp1 and Gnp1 (Fig. 8 B). We reasoned, based on our finding that low levels of Gap1 can apparently fold and exit the ER and reach the PM independently of Shr3 (Fig. 1 B), that if Shr3 normally prevents precocious degradation of AAPs, then mutations that impair ERAD would enable more AAPs to fold properly and lead to increased amino acid uptake.

Consistent with this notion, we observed that in comparison to shr3Δ mutants with intact ERAD pathways (Fig. 8; dilution series 2), shr3Δ strains carrying mutations that inactivate both Doa10- and Hrd1-dependent ERAD pathways display increased sensitivity to d-histidine and the toxic proline analogue azetidine-2-carboxylate (AzC; Fig. 8; combinations of ubc6Δ ubc7Δ or doa10Δ hrd1Δ; dilution series 5 and 8, respectively). As observed in our studies analyzing the degradation of Gap1 and Agp1 (Figs. 6 and 7), we noted that strains carrying only the ubc7Δ mutation were almost as sensitive to amino acid analogues as the ubc6Δ ubc7Δ double mutant strains (Fig. 8; dilution series 4).

The data indicate that under conditions of impaired ERAD, cells have increased levels of properly folded and functional AAPs in their PMs. We tested this possibility directly by examining the aggregation state of Gap1 in the ERAD-defective strains. Membranes from SHR3, shr3Δ, shr3Δ ubc6Δ ubc7Δ, and shr3Δ doa10 hrd1Δ strains were solubilized in the presence of DM, and soluble proteins were separated by BN-PAGE (Fig. 8 C). In comparison to membranes from the Shr3 wild-type strain, we observed diminished levels of Gap1 monomers in membranes lacking Shr3 (Fig. 8 C, compare lanes 1 and 2). Substantially more (two- to threefold) Gap1 monomers were extracted from membranes prepared from shr3Δ strains carrying ubc6Δ ubc7Δ or doa10 hrd1Δ mutations (Fig. 8 C, compare lanes 3 and 4 with lane 2). The increased ability to solubilize Gap1 monomers accounts for the increased amino acid uptake in strains with impaired ERAD (Fig. 8, A and B). In summary, our data support the idea that in addition to facilitating folding of AAPs, Shr3 functions to shield partially folded AAPs from being prematurely targeted for degradation.

Yeast possesses highly specialized ER membrane-localized chaperones (Kota and Ljungdahl, 2005). In this paper, we have experimentally addressed two questions regarding the chaperone activity of Shr3. First, is there a temporal requirement for the chaperone activity of Shr3 during the membrane insertion and folding of AAPs? Second, does the chaperone activity of Shr3 affect surveillance mechanisms that monitor the status of membrane protein folding? In answering these questions, we document that Shr3 has a central role in the biogenesis of AAPs, and our results illuminate the tight coupling between folding and quality-control mechanisms operating during biogenesis of polytopic membrane proteins.

We probed the temporal requirement of Shr3 during AAP folding by exploiting the fact that independently coexpressed truncated N- and C-terminal fragments of Gap1 can fold and assemble forming functional split Gap1 proteins (Fig. 2). Importantly, the functional assembly of N- and C-terminal fragments of Gap1 exhibited a strict dependence on Shr3. This finding enabled us to directly analyze whether Shr3 maintains the N- and/or C-terminal fragments in productive assembly competent conformations. Consistent with our previous data regarding full-length Gap1 (Kota and Ljungdahl, 2005), we found that the N-terminal fragment exhibited an increased propensity to aggregate in membranes isolated from cells lacking Shr3, and importantly, its aggregation was not affected by the presence or absence of the C-terminal fragment (Fig. 3, B and C; and Fig. 4 A). In marked contrast, the ability to solubilize monomers of the C-terminal fragment exhibited not only a dependence on Shr3 (Fig. 3 D) but, more significantly, on the presence of the N-terminal fragment (Fig. 4 B). These data are consistent with a model in which Shr3 interacts early during the membrane insertion of AAPs. Furthermore, as the N- and C-terminal fragments are likely to individually insert into the membrane, Shr3 is able to maintain the N-terminal fragment in a conformation that enables the C-terminal fragment to assemble with it. These observations raise many interesting questions regarding how these individual fragments, perhaps in analogy to subunits of multimeric membrane protein complexes, find each other before their functional assembly.

During translocation, exclusively hydrophobic TMSs rapidly partition into the lipid phase of the membrane (Heinrich et al., 2000). In contrast, less hydrophobic TMSs containing charged or polar residues partition into the membrane less readily and are retained in proximity to the translocon or to translocon-associated proteins, e.g., TRAMs (Heinrich and Rapoport, 2003). Recent results regarding the membrane insertion of aquaporin-4 are consistent with the view that individual TMSs exhibit distinct requirements during translocation (Sadlish et al., 2005). After individually passing through a single entry site in Sec61, several TMSs of aquaporin-4 were found to interact with secondary peripheral sites on Sec61. These results suggest that the translocon may transiently retain certain TMSs to facilitate early folding events and to control their release into the membrane (Sadlish et al., 2005). The intrinsic chaperone-like activity of Sec61 translocon may thus suffice to facilitate correct folding of many polytopic membrane proteins.

However, our results regarding membrane-localized chaperones indicate that complex, or larger, polytopic proteins require the assistance of additional chaperones to fold. An appealing hypothesis, consistent with AAP aggregation observed in the absence of Shr3, is that Shr3 facilitates the partitioning of TMSs of AAPs containing charged or polar amino acid residues as they emerge from the translocon. According to this hypothesis, Shr3 may physically shield charged or polar residues within TMSs, enabling them to more rapidly partition into the membrane and fold, thereby preventing them from engaging in nonproductive interactions. Gap1 possesses five TMSs that have a single charged residue (Fig. 2 A). Although we have not systematically investigated the role of these charged residues, in several instances they are conserved, and mutations that exchange noncharged residues into their positions abolish catalytic activity (unpublished data). Consistently, a mutant Gap1 protein with a lysine replacing the conserved glutamate residue (E300) in TMS VI is inactive and retained in the ER (Lauwers and André, 2006).

In cells lacking Shr3, both Gap1 and Agp1 aggregates are efficiently recognized by ERAD surveillance systems that redundantly activate Doa10- and Hrd1-dependent pathways (Figs. 5, 6, and 7). In general, our results are consistent with the recently proposed unifying concept for ERAD (Carvalho et al., 2006). However, to our knowledge, AAP aggregates are the first substrates that target to both Doa10- (ERAD-C) and Hrd1- (ERAD-M/L) dependent pathways with equal efficacy. Our data indicate that these pathways function in parallel to degrade AAP aggregates. However, it is important to note that AAP aggregates may in fact be composed of a nonhomogeneous mix of kinetically linked folding intermediates, which raises the possibility that ERAD pathways preferentially degrade discrete folding intermediates. Previous reports regarding the organization of ERAD pathways suggested that membrane proteins with misfolded cytoplasmic domains are substrates of the Doa10 pathway, whereas soluble secretory proteins and membrane proteins with misfolded luminal or membrane domains are substrates of the Hrd1-dependent pathway (Huyer et al., 2004; Vashist and Ng, 2004; Ismail and Ng, 2006; Ravid et al., 2006). Although it remains to be determined, the involvement of the Doa10 pathway can be explained by the fact that the bulk of the non–membrane-associated amino acids of AAPs are on the cytoplasmic side of the membrane; presumably the folding of cytoplasmic domains are dependent on the proper folding of TMSs. With respect to the Hrd1 pathway, we suspect that in the absence of Shr3, misfolded TMSs of AAPs are recognized. Given the high level of sequence homology and dependence on Shr3, the other members of the AAP protein family are likely to be similarly degraded. Consequently, our findings substantially augment the number of known substrates of the Doa10- and Hrd1-dependent pathways.

We have found that AAPs can attain native structures, albeit inefficiently, independent of the chaperone activity of Shr3 (Figs. 1 and 8). Our finding that mutations that impair ERAD enhance functional expression of AAPs highlights the intimate link between folding and degradation. There are two possible explanations for the ability of AAPs to fold independently of Shr3. First, a small portion of AAPs may fortuitously fold and attain native conformation as they insert in the membrane. In this case, other chaperones present in the ER membrane may substitute for the lack of Shr3, or the inherent chaperone-like functions associated with the Sec61 translocon may suffice. Alternatively, AAP aggregates may not be solely composed of terminally misfolded proteins. This raises the possibility that low levels of aggregated AAPs are able to reenter a productive folding pathway. Because the topological orientation of each TMS of Gap1 is correctly fixed as they insert into the membrane (Gilstring and Ljungdahl, 2000), AAPs initiate folding at a point that is already quite constrained (Bowie, 2005). Thus, if provided sufficient time (e.g., in cells with impaired ERAD), aggregated AAPs may spontaneously rearrange from a kinetically trapped conformation into their native lower energy conformations.

The idea that Shr3 interacts with TMSs containing charged residues is consistent with it having a role in surveillance mechanisms that monitor folding. Previous work has demonstrated that charged residues within membrane-spanning segments provide key signals for targeting misfolded proteins for ERAD. In the case of unassembled α-subunits of the T cell receptor, which are retained and degraded in the ER, the degradation determinant was mapped to two basic amino acid residues in the TMS (Bonifacino et al., 1990), and the placement of a single charged residue in a TMS of a cell surface protein caused it to be retained and degraded in the ER (Bonifacino et al., 1991). Thus, by shielding charged amino acid residues, Shr3 not only facilitates folding but also prevents incompletely folded AAPs from being targeted for degradation. Furthermore, the fact that Shr3 does not itself exit the ER (Kuehn et al., 1996), and newly synthesized AAPs copurify with Shr3 (Gilstring et al., 1999), suggests that Shr3 retains actively folding AAPs in the ER. Thus, Shr3 may function analogously to the well-characterized N-linked glycosylation-based quality-control system, which provides temporal cues regarding folding of glycoproteins, preventing their premature exit out of the ER and targeting of folding intermediates to ERAD (Helenius and Aebi, 2004).

Our findings underscore the importance of ER membrane-localized chaperones in governing the stability of membrane proteins. In mammalian cells, insulin-induced gene (INSIG) proteins appear to have chaperone-like activity that influences the sterol-dependent degradation of 3-hydroxy-3-methylglutaryl coenzyme A (HMG CoA) reductase and ER retention of Scap (for review see Goldstein et al., 2006). When sterol levels are high, INSIG proteins bind sterol-sensing membrane domains of HMG CoA reductase and Scap. As a consequence of binding, HMG CoA reductase is targeted to ERAD and Scap is prevented from exiting the ER, two events that lead to decreased cholesterol biosynthesis and uptake. In yeast, the INSIG homologues Nsg1 and Nsg2 associate with HMG CoA reductase, facilitating the folding of the sterol-sensing membrane domain and inhibiting Hrd1-dependent degradation (Flury et al., 2005).

Finally, we note striking similarities between the inefficient folding of AAPs in cells lacking Shr3 and the folding of the mammalian Cl⁻ channel cystic fibrosis transmembrane conductance regulator (CFTR). The folding of native CFTR is inefficient (Ward and Kopito, 1994): only 25% of synthesized CFTR folds properly, and the remaining 75% is degraded by multiple ERAD mechanisms (Younger et al., 2006). The similarities to the situation in yeast suggest that CFTR folding occurs in the absence of dedicated Shr3-like chaperones; consequently, CFTR folding intermediates are more or less constitutively and prematurely targeted for degradation. The ability of small molecules to act as chaperones, and the temperature-dependent characteristics of CFTR folding supports this notion (Denning et al., 1992; Brown et al., 1996; Loo et al., 2005). Also, the folding intermediate that accumulates as a result of the disease-causing ΔF508 mutation is kinetically trapped. Interactions with Hsp90, which are positively and negatively modulated by cochaperones, determine the fate of this misfolded intermediate (Wang et al., 2006). The down-regulation of cochaperone Aha1 enables ΔF508 CFTR to fold, exit the ER, and become functionally expressed in the PM. Presumably, reduced levels of Aha1 circumvent surveillance mechanisms that normally target ΔF508 CFTR for ERAD. It will be interesting to continue to exploit the yeast system described here to dissect out further mechanistic details regarding the quality-control surveillance systems that monitor the folding of polytopic membrane proteins.

Yeast strains and plasmids are listed in Table I. FGY206 was created by transforming PLY123 with a PstI–BamHI DNA fragment containing ubc7Δ∷LEU2 from pGR172 (Vassal et al., 1992). FGY205 was created by transforming PLY123 with a HindIII–HindIII DNA fragment containing ubc6Δ∷LEU2 from pTX33 (Sommer and Jentsch, 1993). PLY127 was transformed with a EcoRI–XhoI DNA fragment containing pep4Δ∷hisG-URA3-neo-hisG from pAS173 (provided by A. Sachs, University of California, Berkeley, Berkeley, CA), and a Ura⁺ transformant was propagated on medium containing 5-fluoroorotic acid (FOA), resulting in the unmarked pep4Δ∷hisG strain FGY217. PLY127, FGY205, FGY206, and FGY217 were transformed with a linear EcoRI–SalI DNA fragment containing shr3Δ5∷hisG-URA3-neo-hisG from pPL288 (Kuehn et al., 1996); Ura⁺ transformants were propagated on FOA-containing medium; and strains FGY212, FGY209, FGY210, and FGY219, each carrying the unmarked shr3Δ6 deletion, were obtained. FGY257 was constructed by transforming FGY212 with a SphI–SalI DNA fragment containing hrd1∷URA3 (Bays et al., 2001). Strains JKY29 and JKY39 were constructed by deleting the entire sequence of DOA10 in FGY212 and FGY257 with a PCR-amplified natMX4 cassette (primers F-doa10D and R-doa10D). JKY36 and JKY37 were constructed from FGY205 and FGY209 by deleting the coding sequence of UBC7 with a PCR-amplified natMX4 cassette (primers F-ubc7D and R-ubc7D).

Plasmid pJK92 was created by inserting a 3.4-kb SalI–NotI fragment encoding Gap1 from pPL247 into SalI–NotI–restricted pRS317 (Sikorski and Boeke, 1991). Plasmids expressing split Gap1 constructs under the control of the endogenous GAP1 promoter were created by homologous recombination in yeast. Plasmids pJK97 (gap1 TM1-5) and pJK96 (gap1 TM1-6) expressing N-terminal fragments of Gap1 were created by introducing Bsu36I–BglII–restricted pPL247 together with primer pairs F:1-5TM/R:1-5TM and F:1-6TM/R:1-6TM, respectively. Plasmids pJK98 (gap1 TM7-12) and pJK99 (gap1 TM6-12) expressing C-terminal fragments of Gap1 were created by cotransforming BsiWI–BsaBI restricted pJK92 with primer pairs F:7-12TM/R:7-12TM and F:6-12TM/R:6-12 TM, respectively. A thrice-reiterated myc epitope was inserted at the C terminus, immediately before the stop codon, creating plasmid pJK100; 3×myc was PCR amplified from plasmid pPL329 (primers F-Gap1TM6-12–3×MYC/R-Gap1TM6-12–3×MYC) and cotransformed together with SphI–SalI–restricted pJK99.

Standard media, YPD (yeast extract, peptone, dextrose) and SD (synthetic complete), were prepared as described previously (Burke et al., 2000). Ammonium-based synthetic complex (Andréasson and Ljungdahl, 2002) and minimal media containing urea (SUD), proline (SPD), and allantoin (SAD) as the sole nitrogen source were prepared as described previously (Ljungdahl et al., 1992; Kota and Ljungdahl, 2005). Minimal media were supplemented as required. Media were made solid with 2% (wt/vol) bacto Agar (Difco).

Cells were grown overnight in SD to an OD₆₀₀ of 2–3, harvested, washed once in water, and resuspended in SUD to a starting OD₆₀₀ of 0.1. Cells were grown at 30°C and harvested when cultures reached an OD₆₀₀ of 0.8, and protein extracts were prepared and fractionated on 12–60% sucrose gradients essentially as described previously (Egner et al., 1995). 1-ml fractions were collected from the bottom of the gradients using a Fraction Recovery System (Beckman Coulter). Fractions 1 and 2 and 10 and 11 were separately pooled, and proteins from equal aliquots of the nine resulting fractions were concentrated by TCA precipitation. Proteins were separated by SDS-PAGE and blotted onto nitrocellulose membranes.

Immunoblots were incubated as indicated with primary antibody (rabbit polyclonal) in blocking buffer (phosphate-buffered saline and 0.1% Tween 20) diluted as follows: α-Agp1p, 1:10,000; α-Dap2, 1:2,000; α-Gap1, 1:20,000; α-Kex2, 1:1,000; α-Pma1, 1:3,000; α-Shr3p, 1:1,000; α-Wbp1, 1:1,000. Blots were washed three times for 15 min with wash buffer (blocking buffer plus 5% milk) and incubated with horseradish peroxidase–coupled secondary antibody, donkey α-rabbit Ig (GE Healthcare), diluted 1:5,000 in wash buffer. Blots were washed three times for 15 min with wash buffer, and immunoreactive proteins were visualized by chemiluminescence detection reagents (ECL-PLUS Western Blotting Detection System; GE Healthcare) and the LAS1000 camera system (Fuji). The α-Gap1 and α-Agp1 antisera were provided by B. André (Universite Libre de Bruxelles, Brussels, Belgium). The α-Pma1, α-Wbp1, and α-Kex2 antibodies were obtained from C. Slayman (Yale University, New Haven, CT), S. te Heesen (Swiss Federal Institute of Technology, Zürich, Switzerland), and R. Fuller (University of Michigan, Ann Arbor, MIC), respectively.

Radiolabeling and immunoprecipitations were conducted essentially as described previously (Silve et al., 1991; Volland et al., 1994). In brief, cells grown in SAD (plus uracil) to an OD₆₀₀ of 0.5–0.8 were used to inoculate fresh SAD (plus uracil) at an OD₆₀₀ of 3. Cultures were incubated for 30 min before the addition of ³⁵S-methionine (25 μCi/OD₆₀₀ cells; GE Healthcare). Cells were labeled for 30 min, after which an aliquot of 100× chase solution (25 mM l-methionine and 25 mM l-cysteine) was added. At the times indicated, 1-ml aliquots of culture were placed in Eppendorf tubes containing 60 μl ice-cold lysis solution (1.85 M NaOH and 7% β-mercaptoethanol), rapidly mixed by rigorous vortexing, and after a 10-min incubation on ice, 60 μl of 50% TCA was added. Precipitated proteins, pelleted by centrifugation for 15 min at 12,000 g, were washed once in 60 μl of 1 M Tris, pH 7.5, and solubilized in 0.5% SDS for 10 min at 37°C. Proteins were diluted with TNET buffer (1% Triton X-100, 50 mM Tris, pH 7.4, 150 mM NaCl, and 5 mM EDTA) to a final volume of 0.6 ml. Insoluble material was removed by centrifugation for 30 min at 12,000 g. Rabbit anti-Gap1 antibodies were added (1:1,500), and samples, continuously mixed by inversion, were incubated overnight at 4°C. 40 μl of a 12.5% (vol/vol) suspension of protein A–Sepharose CL-4B beads (GE Healthcare) was added to each sample, and incubations were continued for 3 h at 4°C. Immunoprecipitated material was collected by centrifugation and washed three times with TNET buffer and once with TNET without Triton X-100. Precipitated proteins were eluted by incubation for 10 min at 39°C in 2× SDS-PAGE sample buffer. Eluted proteins were separated by SDS-PAGE in a 10% gel. Gels were fixed in glacial acetic acid/methanol/H₂O (10:20:70), rinsed briefly in water, and dried. Radiolabeled proteins were detected and quantified by phosphorimaging (Fujix Bio-Image Analyzer BAS1500; Fuji).

Cultures grown at 23°C in SAD (plus uracil) to an OD₆₀₀ of 0.8–1 were treated with 400 μg/ml cycloheximide. At the times indicated, 1-ml aliquots were removed to Eppendorf tubes containing 250 μl ice-cold lysis solution (1.85 M NaOH), rapidly mixing by rigorous vortexing, and after a 10-min incubation on ice, 250 μl of 50% TCA was added. Precipitated proteins, pelleted by centrifugation for 15 min at 12,000 g, were washed once in 100 μl of 1 M Tris, pH 7.5, and solubilized in 2× SDS-PAGE sample buffer. Proteins were resolved by SDS-PAGE and analyzed by immunoblotting.

BN gels and whole cell protein extracts were prepared as previously described (Kota and Ljungdahl, 2005). Protein extracts from cells grown in SAD at 25°C were solubilized at 4°C for 35 min in the presence of DM at the concentrations indicated. Extracts containing full-length Gap1 and split Gap1 fragments were separated using 4–15% and 4–20% gradient gels, respectively. High molecular weight marker proteins (GE Healthcare) were used as standards.

Members of Ljungdahl laboratory are acknowledged for constructive comments. We are indebted to Bruno André for his generous gifts of α-Gap1 and α-Agp1 antisera. We thank Carolyn Slayman, Stephan te Heesen, and Robert Fuller for α-Pma1, α-Wbp1, and α-Kex2 antibodies, respectively.

This work was supported by the Ludwig Institute for Cancer Research and a grant from the European Union (EFFEXPORT project; QLK3-CT-2001-00533).