Misfolded endoplasmic reticulum (ER) proteins are eliminated by the retrotranslocation pathway in eukaryotes, which is an important physiological adaptation to ER stress. This pathway can be hijacked by certain viruses to destroy folded cellular proteins, such as MHC class I heavy chain. Recent studies have highlighted the importance of the ubiquitin–proteasome system (UPS) in this process.
The endoplasmic reticulum (ER) is the major site of protein biosynthesis in eukaryotes. It receives newly synthesized polypeptides from the cytosol via a narrow channel formed by the Sec61 complex. An elaborate system of ER chaperones subsequently catalyses the folding and assembly of these polypeptides, allowing them to acquire folded, native conformations. However, polypeptides may adopt aberrant conformations, resulting in aggregation-prone, misfolded proteins. The accumulation of misfolded proteins represents a form of ER stress, which has been implicated in the pathogenesis of many human diseases . To preserve ER homoeostasis, eukaryotes have evolved a conserved ER quality-control pathway, termed retrotranslocation or dislocation, which efficiently eliminates unwanted ER proteins by exporting them into the cytosol [2,3]. Polypeptides undergoing retrotranslocation are disposed of by the proteasome in the cytosol. Thus, this pathway is also called ERAD (ER-associated degradation) . The retrotranslocation pathway may be hijacked by certain viruses to destroy folded cellular proteins required for immune defence, allowing the virus to evade host immune surveillance. For example, the human cytomegalovirus (HCMV) encodes two proteins, US2 and US11, each of which can efficiently target newly synthesized MHC class I heavy chain for retrotranslocation and proteasomal degradation. The Vpu protein of human HIV also utilizes this pathway to remove the CD4 receptor selectively from the ER membrane .
Retrotranslocation first requires the recognition of polypeptides and their subsequent targeting to the site of translocation in the ER membrane. Misfolded or unassembled polypeptides are selected by ER chaperones to enter the pathway, whereas viral factors associate with their specific ‘clients’ (e.g. US2 and US11 are ER-localized membrane proteins that specifically bind MHC class I heavy chain), ‘abducting’ them to the translocation site. Polypeptides are then transferred across the membrane, presumably through a protein-conducting channel. Substrates emerging from the ER are polyubiquitinated on the cytosolic face of the ER membrane, and the modified polypeptides are then dislocated into the cytosol and degraded by the proteasome (Figure 1).
ER proteins are degraded in the cytosol by the ubiquitin–proteasome system (UPS)
Ubiquitin is a small polypeptide that can be covalently linked to either the ∊-amino group of a lysine residue or the N-terminal amino group on a substrate. This process requires three types of enzymes: ubiquitin-activating enzyme (E1), ubiquitin-conjugating enzyme (E2), and ubiquitin ligase (E3). Either a single ubiquitin (mono-) or a chain linked by one of the lysines on ubiquitin itself (poly-) can be attached to a substrate. Chains linked by Lys48 specifically target the modified substrates to the proteasome for degradation, while other chain linkages and mono-ubiquitination mediate different cellular processes. Although it was known for many years that misfolded cytosolic proteins are degraded by the UPS, its role in ER protein degradation was not appreciated for a long time. It was initially thought that ER proteins are destroyed in the ER lumen, which lacks the UPS components, such as ubiquitin and the proteasome. However, genetic and biochemical studies in Saccharomyces cerevisiae and in mammalian cells have since convincingly demonstrated the involvement of the UPS in ER protein degradation [2–4]. For example, many misfolded ER proteins were found to undergo polyubiquitination. If polyubiquitination is prevented by the expression of a mutant ubiquitin with Lys48 changed to Ala, the degradation of many substrates is inhibited. In yeast, mutations in many components of the UPS stabilize misfolded ER proteins. Similarly, in mammalian cells expressing a temperature-sensitive mutant of the E1 enzyme, the degradation of ER proteins is blocked at non-permissive temperatures. Finally, inhibitors of the proteasome also affect the degradation of ER proteins, both in yeast and in mammalian cells. Together, these experiments establish a new model in which misfolded ER proteins, both lumenal and transmembrane, are disposed of by the UPS in the cytosol. Obviously, the ER substrates must first be retrotranslocated into the cytosol before encountering the UPS.
The ubiquitination machinery for retrotranslocation
Many components of the ubiquitination system implicated in retrotranslocation were initially identified in yeast and later studied in mammalian cells. As expected, many of these enzymes are associated with the ER membrane. In yeast, the relevant E2s include Ubc7 (ubiquitin-conjugating enzyme 7), Ubc6, and Ubc1 [6–8]. Ubc7 is bound to the ER membrane by association with its partner Cue1 (coupling of ubiquitin conjugation to ER degradation 1) , and Ubc6 is an ER-localized transmembrane protein . Deletion of these genes individually or in combination prevents the ubiquitination and degradation of many misfolded ER proteins [8,10,11]. The mammalian homologues of Ubc6 and Ubc7 appear to play a similar role [12,13]. In yeast, Hrd1 (hydroxymethylglutaryl-CoA reductase degradation) and Doa10 (degradation of alpha2-10) are the two major E3s dedicated to retrotranslocation [11,14]. Both Hrd1 and Doa10 are multi-spanning membrane proteins containing a RING (really interesting new gene)-finger domain facing the cytosol. These ligases each mediate the retrotranslocation of a distinct class of misfolded substrates [14,15]. In addition, a HECT (homologous to E6-associated protein C-terminus) domain-containing ligase, Rsp5, appears to assist Hrd1 in dealing with misfolded substrates when their accumulation exceeds the capacity that can be handled by Hrd1 . In mammals, an orthologue of yeast Hrd1, and a Hrd1-related E3 called gp78/AMFR (autocrine motility factor receptor), mediate the degradation of some misfolded ER substrates [17,18]. Other E3 enzymes involved in degradation of misfolded ER proteins include CHIP (C-terminus of the heat-shock-protein-70-interacting protein) , a U-box (modified RING motif without the full complement of Zn2+-binding ligand)-containing E3, and Parkin, a RING-finger E3 linked to the juvenile Parkinson's disease . Two F-box-containing proteins, Fbs1 (F-box protein 2) and Fbs2 (F-box protein 6), each of which is part of a multi-subunit SCF [Skp1 (S-phase associated protein-1)–Cdc53 (cell-division cycle 53)/Cul1 (Cullin homologue 1)–F-box protein] ubiquitin ligase, can both recognize carbohydrate chains. The corresponding E3 ligase may be specific for the retrotranslocation of glycoproteins [21,22]. CHIP, Parkin, Fbs1 and Fbs2 are not membrane proteins, and must be recruited to the translocation site by association with a membrane partner.
The role of polyubiquitination in protein retrotranslocation
Polyubiquitination is required for retrotranslocation
When polyubiquitination is prevented, substrates cannot be moved into the cytosol; instead, they accumulate in the ER lumen. Thus, polyubiquitination not only is required for guiding substrates to the proteasome, but also participates in transferring polypeptides across the ER membrane . However, because polyubiquitination occurs at the cytosolic face of the ER membrane, before substrates are completely exported into the cytosol, it cannot account for the initiation of retrotranslocation for lumenal proteins. It is currently unclear how these substrates are inserted into the membrane to initiate the translocation process.
What is the precise role of polyubiquitination in retrotranslocation? Studies on the US11-dependent retrotranslocation of MHC class I heavy chains in a permeabilized cell system have provided some clues. When membranes containing MHC class I heavy chains are incubated with cytosol in the presence of AMPPNP (adenosine 5′-(β,γ-imino)triphosphate), an ATP analogue that supports the polyubiquitination reaction but not other ATP-dependent processes, heavy chains can partially exit the ER and undergo polyubiquitination, but the modified substrates cannot be moved into the cytosol. These membrane-associated retrotranslocation intermediates can be released into the cytosol upon addition of ATP. These experiments indicate that polyubiquitination on its own is insufficient to move substrates from the ER membrane into the cytosol; it probably serves as a recognition signal for a cytosolic ATPase, which utilizes ATP hydrolysis to extract polyubiquitinated substrates and release them into the cytosol .
The AAA (ATPase associated with various cellular activities) ATPase p97 is required for protein retrotranslocation
The cytosolic ATPase that acts on polyubiquitinated polypeptides at the ER membrane turns out to be p97 [also called VCP (valosin-containing protein), or in yeast, Cdc48], a member of the AAA ATPase family. The function of p97/Cdc48 is as a molecular chaperone in many cellular processes, including retrotranslocation, the activation of a membrane-anchored transcription factor, the formation of nuclear envelope, spindle disassembly, and the homotypic fusion of ER/Golgi membranes [24,25]. The ATPase contains two similar Walker-type ATPase domains (D1 and D2) and an additional N-terminal domain (N-domain) (Figure 2). Through its N-domain, p97 can bind to a variety of different cofactors, which explains the functional diversity of this ATPase. The cofactor that assists p97 in retrotranslocation is a dimer consisting of Ufd1 (ubiquitin fusion degradation protein-1) and Npl4 (nuclear protein localization gene 4). In yeast, mutations in each individual component of the Cdc48 complex prevent the export of various misfolded proteins from the ER . As a consequence, the unfolded-protein response (UPR), a collection of signalling transduction pathways that help rectify the folding problem in the ER, is elicited [27,28]. In yeast mutants of Ufd1 and Npl4, retrotranslocation substrates are polyubiquitinated, but the modified substrates remain associated with the ER membrane, indicating that the ATPase complex functions downstream of polyubiquitination [28,29]. In mammalian cells, p97 associates with retrotranslocating substrates in an ATP-dependent manner . Overexpression of a p97 mutant defective in ATP hydrolysis stabilizes many misfolded ER proteins and triggers UPR ([27,30] and Y. Ye, unpublished work). The mutant p97 appears to be able to ‘grab’ at a retrotranslocating substrate, but is unable to ‘pull’ it out of the ER membrane. These experiments indicate that ATP hydrolysis by p97 is required to move polypeptides into the cytosol during retrotranslocation.
How p97 hydrolyses ATP when acting on substrates is unclear. Several studies showed that mutations in the D2 domain affect the ATPase activity more drastically than those in D1 [31,32], but both D1 and D2 are functionally important in vivo . One possibility is that the two ATPase domains might alternate in ATP binding and hydrolysis , similar to the mechanism proposed for the p97/Cdc48-like ATPase Hsp104 (heat-shock protein 104) .
The p97/Cdc48 ATPase complex selectively acts on polyubiquitinated substrates
Why does p97/Cdc48 selectively act on polyubiquitinated substrates? The question was at least partially answered when several groups found that both the ATPase and its cofactors Ufd1 and Npl4 can interact with ubiquitin (Figure 2). Both p97 and Cdc48 have the capacity to bind polyubiquitin chains [32,34,35], albeit with low affinity. In addition, the UT3 domain of Ufd1 and the Zn-finger (ZF) motif of Npl4 can both recognize ubiquitin [32,36]. The ZF motif of Npl4 does not seem to play a role in retrotranslocation. In contrast, ubiquitin recognition by p97 and Ufd1 is essential because, firstly, p97 and Ufd1 act synergistically to bind ubiquitin chains linked through Lys48, the chain linkage preferred by the proteasome acting downstream of the ATPase complex ; secondly, these proteins do not recognize ubiquitin chains synthesized with a ubiquitin fusion to glutathione S-transferase (GST–ubiquitin), which may explain why poly(GST–ubiquitin)-modified substrates cannot be extracted from the ER membrane ; and thirdly, a cofactor complex lacking the ubiquitin binding domain of Ufd1 inhibits the US11-dependent retrotranslocation .
Nonetheless, ubiquitin binding is not required for the initial substrate recognition by p97. When polyubiquitination is abolished, the ATPase can still associate with retrotranslocating polypeptides in the ER membrane, but the associated substrates cannot be released into the cytosol . In addition, the ATPase complex interacts with substrates modified by poly(GST–ubiquitin), presumably through a non-modified segment, because the attached poly-(GST–ubiquitin) chains on their own do not bind to either p97 or Ufd1 . Based on these observations, a dual recognition model was proposed , in which p97/Cdc48 itself first binds a non-modified, presumably unfolded, substrate emerging from the translocation site. Once ubiquitin chains are attached to the substrate, the conjugates can be recognized by both the ATPase and Ufd1. Since interfering with ubiquitin recognition leads to the same defect as that caused by a mutant p97 lacking ATPase activity, ubiquitin binding by the p97 complex may activate the ATPase, leading to the extraction of substrates from the ER membrane. The dislocated substrates are subsequently delivered to the proteasome, which is probably mediated by a different set of ubiquitin-binding proteins that can also interact with the proteasome (see Chapter 4 for further discussion).
The requirement of ubiquitin for the function of p97 is not a unique feature for retrotranslocation. In fact, many other p97 cofactors also contain ubiquitin-binding motifs that are capable of recognizing ubiquitin. For example, p47, which regulates the function of p97 in the homotypic fusion of Golgi membranes, contains an essential UBA (ubiquitin-associated) domain that binds mono-ubiquitin . During the activation of the membrane-anchored transcription factor Spt23, Cdc48 and its cofactors Ufd1 and Npl4 recognize the endoproteolytically cleaved Spt23, which is at least partially mediated by the ubiquitin conjugates attached to Spt23. The ATPase complex then mobilizes the transcription factor from the ER membrane, allowing its nuclear translocation and activation .
Translocation of substrates through the central pore in p97/Cdc48
How does p97/Cdc48 extract polypeptides from the membrane? Crystallography studies reveal that p97 forms a hexameric ring with a central channel [37,38]. This structure resembles the AAA ATPase ring connected to the ClpP (caseinolytic protease P) protease in Escherichia coli or the one docked on the 20 S core particle of the eukaryotic 26 S proteasome. In both cases, polypeptides are threaded through the central channel of the ATPase rings to reach the enzymatic sites embedded within the proteolytic particle . Although p97/Cdc48 does not appear to be directly linked to the 20 S proteasome, it might ‘pull’ substrates out of the ER membrane by moving them through its axial channel. One complication with this model is that p97 acts on substrates containing bulky ubiquitin chains that seem to exceed the size limit of its pore. Thus, it may need to co-operate with deubiquitination enzymes to allow the removal of ubiquitin conjugates before substrates enter the channel. No such enzymes have been identified for retrotranslocation, although in p97-mediated Golgi reassembly, a deubiquitination enzyme, VCIP135, is involved . In addition, because ubiquitin chains are also required downstream of p97 for substrate delivery to the proteasome, deubiquitination at p97 would mean that substrates exiting the p97 channel need to be remodified by polyubiquitination. Nonetheless, an elegant study recently demonstrated that ClpB, a p97/Cdc48-like ring ATPase in bacteria, indeed threads its substrates through its central pore . The structural similarity between p97/Cdc48 and ClpB indicates that the central pore of p97 may also be important for its function.
Association of p97/Cdc48 with the ER membrane
A large fraction of the p97 ATPase complex is closely associated with the ER membrane. However, neither p97, nor its cofactors Ufd1 and Npl4, contain any transmembrane segments. Thus, a membrane receptor is required to dock the ATPase complex to the ER membrane. Recent studies have identified an ER–membrane-protein complex in mammals that provides at least one binding site for p97 and its cofactors . One component of the complex, Derlin-1, belongs to a conserved multi-spanning membrane protein family. The other component of the complex, VIMP/SelS (VCP-interacting membrane protein/selenoprotein S), is a single-spanning membrane protein present only in vertebrates. The complex binds p97 via two binding sites, one in the C-terminal tail of Derlin-1, and the other in the cytosolic domain of VIMP (Y. Ye, unpublished work).
Derlin-1 appears to be a central component of the retrotranslocation pathway for many substrates . The homologue of Derlin-1 in yeast, Der1, is required for the degradation of some misfolded ER proteins [15,43]. Inactivation of Derlin-1 in Caenorhabditis elegans induces UPR, presumably because misfolded ER proteins are not efficiently degraded in the absence of Derlin-1 . In mammals, Derlin-1 associates with retrotranslocation substrates as they move across the membrane. It also binds a subset of ubiquitin ligases dedicated to the retrotranslocation pathway (Y. Ye, unpublished work). The interaction between Derlin-1 and E3 ligases is at least partially mediated by p97. This allows the simultaneous recruitment of the ATPase and the ubiquitination machinery to Derlin-1.
The role of Derlin-1 in retrotranslocation is best understood for the US11-dependent retrotranslocation of MHC class I heavy chains [42,44]. Derlin-1 can bind the viral protein US11 as well as p97. The interaction between US11 and Derlin-1 is essential for targeting heavy chains to the retrotranslocation pathway , whereas the interaction between Derlin-1 and p97 presumably docks the ATPase and the associated E3 enzymes to the site of translocation, allowing these enzymes to act synergistically on substrates as they emerge from the ER. Thus substrate recognition in the ER lumen is directly linked to its ubiquitination and p97-mediated dislocation in the cytosol via Derlin-1, suggesting that it may play a direct role in transferring substrates across the ER membrane, perhaps as a component of the protein-conducting channel.
A role for VIMP in retrotranslocation is suggested by its up-regulation upon ER stress , and by its association with retrotranslocation substrates and many components of the pathway, such as Derlin-1, p97 and E3 ligase . Other Derlin-like proteins may also be involved in retrotranslocation, because the human Derlin-2 also binds p97 as well as a retrotranslocation-specific E3 ligase, Hrd1 (B. Lilley and H. Ploegh, personal communication).
A membrane channel for retrotranslocation
Retrotranslocation substrates are believed to traverse the ER membrane through a protein-conducting channel. Earlier studies suggested that the Sec61 complex, which mediates the translocation of polypeptides into the ER, might also function in retrotranslocation . Although some genetic and biochemical evidence supports this view, studying some earlier works reveals that the interaction between the Sec61 complex and retrotranslocation substrates cannot be reproduced when the co-immunoprecipitation is done under more stringent conditions. In addition, no interaction between the Sec61 channel and the known components of the retrotranslocation pathway can be detected (Y. Ye, unpublished work). Thus, the Sec61 complex is unlikely to serve as a general channel for retrotranslocation.
As suggested above, several lines of evidence suggest that Derlin-1 defines the site of retrotranslocation, and may be part of the retrotranslocation channel. Consistent with this view, the Derlin family members can form homo- or hetero-oligomers in the ER membrane, resulting in a complex that contains a sufficient number of transmembrane segments for a protein channel. Nevertheless, definitive evidence is still lacking. In addition, assuming Derlin-1 can form a channel, there must be other channels for retrotranslocation because both Derlin-1 and its yeast homologue Der1 only account for the degradation of a subset of ER proteins.
Studies on the US11-dependent retrotranslocation of MHC class I heavy chains have outlined a pathway, from the recognition of the substrate in the ER to its degradation by the proteasome in the cytosol (Figure 3). First, US11 recognizes heavy chains and targets them to Derlin-1. Next, heavy chains are inserted into a protein-conducting channel, which may be formed by oligomerization of the Derlin proteins. Once a segment of the heavy chain has emerged into the cytosol, it is captured by p97. At the same time, a retrotranslocation complex containing p97, its cofactors and a ubiquitination enzyme is assembled. The substrate undergoes polyubiquitination and the attached polyubiquitin chains are recognized by both p97 and the cofactor Ufd1. Finally, the substrate is ‘pulled’ out of the ER membrane and delivered to the proteasome in the cytosol.
The retrotranslocation of some misfolded ER proteins probably occurs by the same mechanism, except that these substrates are recognized by ER chaperones, and that their targeting to Derlin-1 may require additional shuttle proteins that can interact with both Derlin-1 and chaperone-associated substrates. Some substrates seem to use a Derlin-1-independent mechanism to cross the membrane. In this case, the initial substrate targeting to the membrane may involve a different set of factors. Despite such diversity in the early steps of retrotranslocation, almost all retrotranslocation substrates undergo polyubiquitination, and their subsequent movement into the cytosol is dependent on this modification as well as on the ubiquitin-specific p97 ATPase complex.
Misfolded ER proteins are eliminated by a pathway termed retrotranslocation, dislocation, or ERAD. In this process, polypeptides are moved from the ER into the cytosol, where they are degraded by the proteasome. This pathway can be hijacked by certain viruses to destroy folded cellular proteins such as MHC class I heavy chain and CD4. Polypeptides are transferred across the membrane, presumably through a protein-conducting channel.
Most retrotranslocation substrates are modified by polyubiquitination on the cytosolic face of the ER membrane, before they are released into the cytosol. Polyubiquitination is required not only for the targeting of a substrate to the proteasome, but also for moving the substrate from the membrane into the cytosol.
Polyubiquitinated polypeptides are released from the membrane by an ATPase complex, consisting of p97 and its cofactors Ufd1 and Npl4. The ATPase itself initially recognizes a non-modified segment of the substrate emerging from the ER. Once the substrate undergoes polyubiquitination, the attached ubiquitin chains are recognized by both p97 and its cofactor Ufd1, leading to the ‘pulling’ of the substrate into the cytosol.
In mammals, p97 is recruited to the ER membrane by association with a membrane-protein complex, consisting of Derlin-1, a conserved multi-spanning membrane protein, and a single-spanning membrane protein called VIMP.
p97 also interacts directly with several E3 enzymes specific for retrotranslocation. This interaction appears to facilitate the association between the E3 ligases and Derlin-1. The simultaneous recruitment of p97 and the E3 ligases to Derlin-1 indicates that it represents the site from which substrates exit the ER.
I thank Tom Rapoport and Andrew Osborne for critical reading of the manuscript, and apologize for not being able to cite all the papers relevant to the development of this field due to space constraints.
- © 2005 The Biochemical Society