Characterization of protein complexes of the endoplasmic reticulum-associated degradation E3 ubiquitin ligase Hrd1

Hrd1 is the core structural component of a large endoplasmic reticulum membrane-embedded protein complex that coordinates the destruction of folding-defective proteins in the early secretory pathway. Defining the composition, dynamics, and ultimately, the structure of the Hrd1 complex is a crucial step in understanding the molecular basis of glycoprotein quality control but has been hampered by the lack of suitable techniques to interrogate this complex under native conditions. In this study we used genome editing to generate clonal HEK293 (Hrd1.KI) cells harboring a homozygous insertion of a small tandem affinity tag knocked into the endogenous Hrd1 locus. We found that steady-state levels of tagged Hrd1 in these cells are indistinguishable from those of Hrd1 in unmodified cells and that the tagged variant is functional in supporting the degradation of well characterized luminal and membrane substrates. Analysis of detergent-solubilized Hrd1.KI cells indicates that the composition and stoichiometry of Hrd1 complexes are strongly influenced by Hrd1 expression levels. Analysis of affinity-captured Hrd1 complexes from these cells by size-exclusion chromatography, immunodepletion, and absolute quantification mass spectrometry identified two major high-molecular-mass complexes with distinct sets of interacting proteins and variable stoichiometries, suggesting a hitherto unrecognized heterogeneity in the functional units of Hrd1-mediated protein degradation.

Hrd1 is the core structural component of a large endoplasmic reticulum membrane-embedded protein complex that coordinates the destruction of folding-defective proteins in the early secretory pathway. Defining the composition, dynamics, and ultimately, the structure of the Hrd1 complex is a crucial step in understanding the molecular basis of glycoprotein quality control but has been hampered by the lack of suitable techniques to interrogate this complex under native conditions. In this study we used genome editing to generate clonal HEK293 (Hrd1.KI) cells harboring a homozygous insertion of a small tandem affinity tag knocked into the endogenous Hrd1 locus. We found that steady-state levels of tagged Hrd1 in these cells are indistinguishable from those of Hrd1 in unmodified cells and that the tagged variant is functional in supporting the degradation of well characterized luminal and membrane substrates. Analysis of detergent-solubilized Hrd1.KI cells indicates that the composition and stoichiometry of Hrd1 complexes are strongly influenced by Hrd1 expression levels. Analysis of affinity-captured Hrd1 complexes from these cells by size-exclusion chromatography, immunodepletion, and absolute quantification mass spectrometry identified two major high-molecular-mass complexes with distinct sets of interacting proteins and variable stoichiometries, suggesting a hitherto unrecognized heterogeneity in the functional units of Hrd1-mediated protein degradation.
Approximately one-third of all proteins in eukaryotes are targeted to the secretory pathway (1,2). To attain functionality, these proteins must fold into their specific three-dimensional conformation, a process that, although assisted by abundant molecular chaperones and folding enzymes, is under constant challenge by molecular crowding, genetic mutations, protein synthesis errors, and environmental stress (3). Production of folding-defective proteins in the secretory pathway is associated with a broad range of diseases including cystic fibrosis, ␣1-antitrypsin deficiency, Gaucher's disease, neurodegenerative diseases, and cancer (4). To relieve the burden of aberrant proteins in the ER, 4 cells have evolved ER quality control mechanisms that detect and selectively remove these potentially toxic proteins via a process termed ER-associated degradation (ERAD) (5). ERAD is also important for regulating the abundance of some native proteins including metabolic enzymes and ER-resident ion channels (6,7).
ERAD is capable of recognizing and degrading both secreted and membrane-spanning integral membrane proteins of all possible topologies and membrane orientations. Once recognized, substrates must, in the case of secreted proteins, be fully dislocated across the ER membrane to be destroyed in a ubiquitin-dependent fashion by 26S proteasomes in the cytoplasm (5). This enormous substrate diversity, together with the inherent topological constraint of having substrates initially on the opposite side of the ER membrane from the protease that degrades them, is reflected by the existence of a highly complex multiprotein system traversing the ER membrane. Intense genetic and biochemical dissection of the ERAD system in yeast and mammals has identified a large set of proteins organized around Hrd1, a membrane-integrated ubiquitin E3 ligase that forms the main structural element of the channel through which ERAD substrates are dislocated across the ER membrane (8 -10). Hrd1 acts as a membrane hub that, via its obligate partner SEL1L, coordinates substrate recognition in the ER lumen with membrane dislocation and, via its ubiquitin ligase activity, recruits the cytoplasmic AAA ATPase p97/VCP to facilitate substrate extraction and hand-off to 26S proteasomes.
Proteomic interrogation of the Hrd1 interaction network in yeast (11) and mammals (12) suggests that Hrd1 exists within a highly elaborate modular protein network containing over 80 distinct nodes (12). However, despite this progress and recent application of functional genomic analyses (13)(14)(15), scant progress has been achieved toward either a structural understanding of the composition and stoichiometry of Hrd1 complexes or their dynamics, which must be highly coordinated to manage substrate diversity to achieve both fidelity and efficient coupling between substrate recognition and degradation.
Knowledge of the Hrd1 network comes mostly from studies using ectopic expression of various ERAD components by transient or stable gene expression from engineered promoters (12, 16 -22). Importantly, several reports have found that overexpression of wild-type Hrd1 or its interacting partners can dominantly impair the degradation of ERAD substrates (12,17,23,24). In this study we used genome editing to generate a cell line with a small tandem-affinity purification (TAP) tag knocked in to the endogenous Hrd1 locus, replacing all genomic copies of Hrd1 with a tagged variant that is expressed at normal endogenous steady-state levels. We have used this cell line to begin to biochemically dissect the Hrd1 multiprotein complex. Quantitative proteomics, together with multidimensional biochemical fractionation of Hrd1 indicates that Hrd1 complexes can be resolved into two distinct macromolecular species of different compositions and sizes. The implications of these findings in terms of the organization of ERAD are discussed.

Generation of clonal cell lines expressing tandem-affinitytagged Hrd1 from its native locus
We used genome editing to generate a clonal HEK293 cell line (Hrd1.KI) harboring a TAP tag "knocked in" to the genomic locus encoding the C terminus of Hrd1 (supplemental Fig.  S1A). This TAP tag consists of a tandem Strep tag and S-epitope tag separated by a PreScission protease cleavage site. Steadystate levels of Hrd1-TAP in Hrd1.KI cells were comparable with those of Hrd1 in unmodified HEK293 cells and ϳ2.7-fold lower than in a transgenic line, Hrd1.OE, that stably expresses Hrd1 bearing a variant tag, TAP*, which differs from the tag in Hrd1.KI cells by the presence of His 6 instead of Strep tag. To determine whether TAP-tagged Hrd1 in Hrd1.KI cells is functional in ERAD, we performed translational shut-off assays to monitor the stabilities of two different ERAD substrates known to have a strong dependence on Hrd1: the core-glycosylated form of CD147, an endogenous type I membrane protein that assembles inefficiently (21) (Fig. 1B) and the null Hong Kong variant of ␣1-antitrypsin (NHK), an ectopically expressed folding-defective luminal protein (20) (Fig. 1C). Knockdown of endogenous Hrd1 (shHrd1) depleted immunodetectable Hrd1 (Fig. 1A) and strongly stabilized both CD147 and NHK. Importantly, no difference was observed in the rate of degradation of either substrate in Hrd1.KI compared with unmodified HEK293 or Hrd1.OE cells. These results establish that Hrd1.KI cells harbor a homozygous, in-frame TAP tag insertion into the endogenous Hrd1 locus and are functional in degrading two known Hrd1-dependent ERAD substrates. shHrd1, Hrd1 knockdown cells; Hrd1.OE, Hrd1-TAP* overexpression cells. The TAP tag consists of Strep and S tags separated by a PreScission protease cleavage site, and the TAP* tag consists of His 6 and S tags separated by a PreScission protease cleavage site. B, effect of Hrd1 levels and tag insertion on degradation of endogenous CD147. The indicated cell lines were treated with 25 M emetine to inhibit protein synthesis for the indicated times. Upper panel, cell lysates were analyzed by immunoblotting with antibodies to the indicated endogenous proteins and the band intensities were quantified. M, mature; CG, core-glycosylated. Lower panel, the amount of core-glycosylated CD147 remaining at each time point was normalized to the tubulin signal, and the ratio was plotted relative to untreated. Two biological replicates for each condition are plotted. C, effect of Hrd1 levels and tag insertion on degradation of ␣1-antitrypsin null Hong Kong variant. Indicated cell lines were transfected with NHK-HA expressing plasmid. After 24 h, the cells were treated with 25 M emetine for the indicated times, and the cell lysates were analyzed by immunoblotting (upper panel) and plotted (lower panel) as in A. The data represent a single experiment.

Hrd1 overexpression alters the composition of the Hrd1 complex
To analyze the effect of Hrd1 overexpression on the relative abundance of associated proteins, we used S-affinity capture to isolate Hrd1-containing protein complexes from Hrd1.OE and Hrd1.KI cells following lysis in 1% digitonin and analysis on SDS-PAGE ( Fig. 2A). Although the overall pattern of major proteins from the two cell lines was grossly similar when examined by silver staining (Fig. 2A), analysis by immunoblotting (Fig. 2B) revealed notable differences in the relative abundance of proteins co-purifying with Hrd1 from Hrd1.KI cells compared with Hrd1.OE cells. Specifically, we found that SEL1L, OS9.1, and XTP3B were reduced relative to Hrd1 by a factor of ϳ2, indicating that modest overexpression can alter the stoichiometry of Hrd1 with its interaction partners. Similar stoichiometries were seen when we compared the relative spectral counts of Hrd1 interaction partners obtained by LC-MS/MS analysis of Hrd1.KI cells with our previously reported proteomic data from a stable transgenic line expressing C-terminal S-tagged Hrd1 (12). This comparison shows that most Hrd1 Hrd1-TAP and associated proteins were isolated by S-affinity capture from 1% digitonin lysates of either Hrd1-TAP* overexpression cells (Hrd1.OE) or Hrd1-TAP knock-in cells (Hrd1.KI). This experiment was repeated three times with similar results. B, immunoblot analysis of affinity captured Hrd1-TAP protein complexes. Hrd1-TAP and associated proteins were isolated by S-affinity capture from 1% digitonin lysates of indicated cell lines and analyzed by immunoblotting for the indicated proteins. The band intensities were quantified by LiCOR and normalized to Hrd1. The dotted graph on the right plots the ratio of Hrd1-normalized band intensity for the indicated interactors in Hrd1.OE cells compared with that in Hrd1.KI cells. Individual dots represent two independent biological replicates, and the line represents the mean. C, Hrd1 complex stoichiometry is altered by overexpression. Hrd1-associated proteins were isolated by S-affinity capture from 1% digitonin lysates of Hrd1.KI cells and analyzed by LC-MS/MS. The graph plots the ratio of bait-normalized TSCs for the indicated interactors in Hrd1-overexpressing cells (12) compared with the bait-normalized TSCs in Hrd1.KI cells. The data are calculated from the averages of three independent experiments for Hrd1.KI and four independent experiments for Hrd1-overexpressing cells. The raw data and calculations are presented in supplemental Table  S1, A and B. Similar LC-MS/MS capture profiles were observed for three other independently isolated Hrd1 knock-in clones (clones 10, 15, and 20 in supplemental Fig. S1B; data not shown). Asterisks indicate proteins that were not detected in Hrd1-overexpressing cells. D, overexpression promotes Hrd1 oligomerization. Immunoblot analysis of Hrd1-TAP complexes captured from 1% digitonin lysates of the indicated cell lines by S-protein affinity capture and analyzed under non-reducing (lanes 1-3) or reducing (lanes 4 -6) conditions. The blot was probed with S-tag antibody. Numerals in blue indicate different oligomeric status. Band intensities were quantified by LiCOR and graphed as a fold increase in Hrd1 level in non-reducing condition in Hrd1.OE cell line compared with the Hrd1.KI cell line. The data plotted are the means of two biological replicates with individual data points displayed as dots.
interactors, including SEL1L, Derlin-2, Ube2g2, Ube2j1, XTP3B, and BiP, were present in affinity-captured material from cells that modestly overexpress Hrd1 at half or less the abundance (relative to Hrd1) than in Hrd1.KI cells ( Fig. 2C and supplemental Table S1). Some interactors, including HERP1 and HERP2, were consistently detected in Hrd1.KI cells but were undetectable in Hrd1 overexpressing cells (Fig. 2C, asterisks). Others, notably Erlin1, Erlin2 (elevated 4ϫ and 26ϫ; supplemental Table S1), and VCP (elevated 2.3ϫ; Fig. 2C and supplemental Table S1), were dramatically elevated in relative abundance upon Hrd1 overexpression. We failed to detect proteasome-derived peptides in Hrd1.KI cell line, even though we robustly captured the entire 26S proteasome in our previous published proteomic Hrd1 interactome (12). Hrd1 also had a 20 -40-fold greater propensity to form high-molecular-mass species on non-reducing SDS-PAGE, suggesting that overexpressed Hrd1 can form non-native disulfide-linked multimers (Fig. 2D). Taken together, these data suggest that modest overexpression of Hrd1 may favor Hrd1 oligomerization over interactions with native binding partners.

Hrd1 forms heterogeneous high-molecular-mass complexes with distinct sets of interacting proteins
To analyze native Hrd1 complexes, we used size-exclusion chromatography (SEC) to fractionate S-affinity-captured Hrd1-associated proteins isolated from detergent lysates of Hrd1.KI cells (Fig. 3, A and B). Although digitonin has been widely used to solubilize protein complexes from the ER, we found that it strongly inhibited PreScission protease activity, thereby interfering with our primary elution strategy (supplemental Fig. S2A). We therefore screened a panel of zwitterionic or non-ionic detergents and selected N-dodecyl-␤-D-maltoside (DDM) for this study based on its compatibility with PreScission protease (supplemental Fig. S2A), its ability to solubilize Hrd1 from the ER membrane (supplemental Fig. S2B), and its capacity to preserve the interactions between Hrd1 and known binding partners (supplemental Fig. S2C). Hrd1 complexes were isolated from DDM lysates of Hrd1.KI cells by S-affinity capture and eluted with PreScission protease prior to Superose 6 SEC fractionation (Fig. 3B). SEC fractions were concentrated by a second round of affinity capture with strep-Tactin, and elution in SDS was followed by separation on SDS-PAGE and detection by silver staining (Fig. 3C) or immunoblotting (Fig.  3D). Hrd1 eluted in a broad profile ranging from close to the void volume of the column (ϳ6 MDa) to ϳ100 kDa, near the predicted mass of a Hrd1 monomer (ϳ68 kDa), assuming a ϳ50-kDa micelle molecular mass (25). The majority (95%) of Hrd1 eluted in 8.5-16 ml, corresponding roughly to 0.15 to Ͼ2 MDa and could be resolved into three distinct peaks, which we designated peak I (V e ϭ 8.5-12 ml, molecular mass ϭ 0.6 to Ͼ2 MDa), peak II (V e ϭ 12.5-14 ml, molecular mass ϭ 0.4 -0.6 MDa), and peak III (V e ϭ 14.5-16 ml, molecular mass ϭ 0.15-0.4 MDa) (Fig. 3, C and D). Known Hrd1 interactors exhibited distinct elution profiles that corresponded to one or more of these peaks (Fig. 3, C-F). SEL1L co-eluted with OS9 in a broad asymmetric, monophasic profile with a maximum at 13.5 ml, corresponding to peak II in the Hrd1 elution profile, consistent with the well established association among these three pro-teins (20). Reinjection of fractions from peak II (V e ϭ 13 ml) into a second round of SEC gave a nearly identical elution profile with no appreciable loss of protein, confirming the stability of DDM-solublized complexes (data not shown). XTP3B also eluted with peak II, consistent with its interaction with SEL1L (20), but its elution profile was distinctly more symmetric and right-shifted with respect to OS9 by ϳ0.5-1 ml, suggesting that these two lectins participate in distinct complexes with the membrane-integrated Hrd1-SEL1L machinery. Interestingly, Fam8A1 and Derlin-2 exhibited biphasic elution profiles, cofractionating with peaks I and III, complementary to the elution profile observed for SEL1L/OS9/XTP3B. Fam8A1 is a hydrophobic protein of unknown function that we previously identified as a Hrd1 interactor (12). HERP1, an ER-resident protein that is highly regulated by ER stress and known to associate with Hrd1 (26,27), co-eluted with Fam8A1 in peak I, and a minor fraction was present in peaks II-III (Fig. 3, C-G).
The relationships among the different Hrd1 interactors are more clearly visualized by hierarchical cluster analysis of the mean normalized protein levels determined from immunoblots of four biological replicates ( Fig. 3F and supplemental Table  S2). In this analysis, SEL1L, OS9, and XTP3B co-elute with Hrd1 in peak II, corresponding to an apparent molecular mass of ϳ400 -600 kDa, whereas HERP1, Fam8A1, and Derlin-2 elute with distinctly different profiles that are largely distinct from the core complex enriched in SEL1L and the ER lectins.
These elution profiles of the core Hrd1 complex components described above were confirmed by performing LC-MS/MS analysis on SEC fractions, an approach that also enabled assessment of elution profiles for Hrd1 complex components that were not amenable to quantification by immunoblotting. Total ion currents for selected peptides corresponding to previously identified ERAD components were determined (supplemental Table S3A), and their elution profiles were plotted (supplemental Fig. S3A). Hierarchical cluster analysis of these data ( Fig. 3G) revealed excellent concordance with the immunoblot data and permitted analysis of several additional Hrd1 interactors including GRP94, BiP, Ube2j1, UBXD8, ubiquitin, Erlin2, and AUP1, some of which exhibited unique elution patterns. All the known components of the "core" glycan-sensing ERAD machinery, including SEL1L, OS9, XTP3B, the ubiquitinconjugating enzyme Ube2j1, and BiP, co-elute with Hrd1 in a relatively tight cluster (peak II) of ϳ400 -600 kDa. A substantial fraction of Hrd1 also elutes at considerably higher molecular mass (ϳ0.6 -3 MDa; peak I) where it is associated with Fam8A1, Erlin2, and HERP1. Finally, we also noted that the low-molecular-mass fractions (peak III) were strikingly enriched in the core-glycosylated form of the endogenous Hrd1-dependent ERAD substrate, CD147(CG).

Effect of ubiquitin-proteasome system inhibitors on SEC fractionation of substrate-bound Hrd1 complexes
To investigate the dynamics of Hrd1-substrate complexes, we evaluated the effect of inhibitors of the ubiquitin proteasome system on the elution profiles of Hrd1, SEL1L, and CD147(CG) (Fig. 4). Inhibition of either the ubiquitin-activating enzyme E1 (C1 (28)) or VCP/p97 (CB-5083 (29, 30)) but not the proteasome (MG132) led to a substantial decrease in CD147(CG) abundance in peak III and a concomitant increase in its presence in higher-molecular-mass fractions (Fig. 4, A and  B). C1 and CB-5083, but not MG132, also modestly influenced the elution of Hrd1 and SEL1L, reducing the fraction eluting in peak II with corresponding increases in peak I (Fig. 4A). Treatment with MG132 or CB-5083 led to the appearance of a ladder of bands that were detected with CD147 antibodies in both peak II and peak III fractions (Fig. 4B). This pattern, characteristic of ubiquitylation, is also consistent with our detection of ubiquitin by LC-MS/MS in these fractions ( Fig. 3G and supplemental Fig. S3) and with the absence of the ladder pattern fol-lowing inhibition of the ubiquitin-activating enzyme E1 (Fig.  4B). Considering the low molecular masses of protein complexes eluting in peak III (Ͻ400 kDa) and the calculated molecular mass of Hrd1 (ϳ68 kDa), it is likely that peak III is comprised of largely dissociated Hrd1 complexes. The persistence of a type I ERAD-M substrate with Hrd1 in these low-molecular-mass fractions and the shift toward higher molecular masses observed with inhibitors of ubiquitylation and VCP/p97 are consistent with a role for ubiquitin in substrate dissociation from Hrd1 and suggest a model in which VCP/p97 catalyzed substrate dissociation is a rate-limiting step in ERAD.

Hrd1 forms heterogeneous high-molecular-mass complexes with distinct stoichiometries
We used immunodepletion of affinity-purified and eluted Hrd1 from Hrd1.KI cells with antibodies to Fam8A1 and SEL1L to assess the subunit composition of Hrd1 complexes across the SEC spectrum (Fig. 5). Immunoblot analysis confirmed that we were able to deplete ϳ95% of total SEL1L and ϳ99% of Fam8A1 from affinity-purified Hrd1 complex (Fig. 5A). Depletion of SEL1L resulted in a loss of 28 Ϯ 11% of Hrd1, indicating that a majority of Hrd1 is not complexed with SEL1L under these conditions. Consistent with this observation, depletion of SEL1L caused a shift of Hrd1 away from the largest (peak I) and smallest (peak III) fractions in favor of the "core" complex (peak II) (Fig. 5, B and C, and supplemental Table S4). In sharp contrast, depletion of Fam8A1 resulted in a loss of 71 Ϯ 12% of Hrd1, indicating that the majority of Hrd1 is associated with .KI cells were subjected to S-affinity capture, followed by PreScission protease elution and fractionation by Superose 6 SEC into 48 fractions of 0.5 ml. Hrd1 complexes were captured from each fraction using Strep-Tactin affinity purification, eluted in SDS sample buffer, and analyzed by SDS-PAGE and silver staining. Elution volumes of column calibration standards and their theoretical molecular masses are indicated at the top. The arrowhead between 12 and 12.5 ml indicates where images of two separate gels (prepared and run in parallel) were digitally spliced together. Asterisks indicate nonspecific background bands that are also present in the control SEC elution profile obtained with wild-type HEK293 cells harboring unmodified Hrd1 (supplemental Fig. S2D). This experiment was repeated three times with similar results. D, DDM-soluble proteins associated with purified Hrd1-TAP were captured and fractionated as described for A and analyzed by immunoblotting for the indicated proteins. The arrowhead between 12 and 12.5 ml indicates where images of two separate immunoblots prepared in parallel were digitally spliced together. The data are representative of two (OS9, XTP3B, and HERP1) or four (Hrd1, SEL1L, Fam8A1, and Der2) biological replicates. Derlin-2 was abbreviated to Der2 throughout the figures. E, elution profiles of affinity-captured Hrd1-interacting proteins. Band intensities in B were quantified by LiCOR, normalized to the peak fraction, and graphed as percentages of the total for each fraction (n ϭ 4, Ϯ S.E.). The raw data and calculations are in supplemental Table S2. Major peaks are indicated by Roman numerals. F, heat map of hierarchically clustered, normalized protein levels in SEC fractions obtained from data in D and supplemental Table S2. Band intensities in each blot of D were normalized to the peak fraction and are indicated in the color-intensity scale. G, heat map of hierarchically clustered normalized ion currents obtained for SEC fractions from LC-MS/MS analysis of S-protein affinity-captured Hrd1-TAP protein complexes fractionated by Superose 6 SEC (24 fractions of 1 ml) prior to Strep-Tactin affinity purification and digestion with trypsin. Ion currents of peptides mapping to known Hrd1-interacting proteins, normalized to the maximum signal across the SEC fractions, are indicated in the color intensity scale. The data are derived from a single experiment, and the calculations are in supplemental Table S3A. Fam8A1. Accordingly, Fam8A1 depletion resulted in a pronounced shift of Hrd1 toward "core" (peak II) complexes (Fig. 5, B and C, and supplemental Table S4). Simultaneous depletion of both SEL1L and Fam8A1 led to additive losses of Hrd1 from larger toward intermediate sized (peak II) complexes. Together these data suggest that the high-molecular-mass Hrd1 complexes contain both Fam8A1 and SEL1L, a conclusion that is supported by the observation that depletion of SEL1L and The remaining material was analyzed by immunoblotting for the indicated proteins. Right panel, immunoblot band intensities were quantified using LiCOR and normalized relative to the mock sample. The data represent means Ϯ S.E. of four independent experiments (for SEL1L and Fam8A1 immunodepletion) or two experiments (for co-depletion of SEL1L and Fam8A1). B, SEC elution profiles of Hrd1, Fam8A1, and SEL1L following immunodepletion of the indicated proteins. S-affinity-captured Hrd1 complexes isolated from DDM solubilized Hrd1.KI lysates were immunodepleted with antibodies against SEL1L, Fam8A1, or both SEL1L and Fam8A1 prior to SEC fractionation. Hrd1 complexes remaining in each SEC fraction were reisolated using Strep-Tactin affinity purification and analyzed by immunoblotting for the indicated proteins. Lanes marked Pre and Post denote S-protein affinity-captured Hrd1 complexes prior to and following immunodepletion, respectively. The arrowhead between 12 and 12.5 ml indicates where images of two separate immunoblots prepared in parallel were digitally spliced together. C, immunoblot band intensities in B were quantified by LiCOR, normalized to the peak intensity, and plotted as the percentages of the total of each indicated protein for each depletion condition. The raw data and calculations for Fig. 4 are in supplemental Table S4.
Fam8A1, individually, also caused significant rightward shifts of the elution profile of the reciprocal protein (Fig. 5, B and C). We conclude that, under these solubilization conditions, Hrd1 is broadly distributed into heterogeneous protein complexes that contain variable stoichiometries of its major interaction partners, SEL1L and Fam8A1, as well as distinct cohorts of other interaction partners.

AQUA-MS analysis
We used AQUA-MS (31) to determine the absolute levels of Hrd1 and its major interaction partners. In unfractionated tandem affinity-purified complexes, Hrd1 was present in stoichiometric excess over its binding partners ranging from ϳ2-fold for Fam8A1 to ϳ5and ϳ8-fold for OS9 and SEL1L, respectively (Tables 1 and 2). The profile of absolute levels of Hrd1 and its interactors following SEC fractionation (Fig. 6) closely paralleled the distributions observed in immunoblotting (Fig. 3E), spectral counting (supplemental Fig. S3B and Table S3B) or ion current ( Fig. 3G and (supplemental Fig. S3A and Table S3A) measurements. However, we found that the absolute molar ratio of Hrd1 to its interactors varies considerably across the elution profile, with SEL1L being strikingly disenriched in the highest-(peak I) and lowest-molecular-mass (peak III) SEC fractions. OS9, which is known to bind directly to SEL1L (20), exhibits a similar elution profile and for the most part is present throughout the SEC fractions at a level approximately twice that of SEL1L, reflecting the possibility that SEL1L has two binding sites for OS9. By contrast, Fam8A1 is strongly disenriched in complexes isolated from peak II and peak III, reflecting that the vast majority of this protein is associated with highest-molecular-mass Hrd1 complexes where it is present at approximately half the level of Hrd1.

Discussion
Hrd1 is the central element of a large protein complex that coordinates the processes of quality-control protein degradation in the ER. In addition to functioning to scaffold and organize substrate recognition and ubiquitylation at the luminal and cytoplasmic sides of the ER membrane, respectively, Hrd1 is likely to constitute the key structural member of the transmembrane pore though which glycoproteins are conducted to be destroyed by cytoplasmic proteasomes (5, 8 -10, 32-36). Understanding the structure and composition of native Hrd1 protein complexes is therefore central to our understanding of the molecular mechanism of ERAD. Although a large number of Hrd1 interactors have been identified from high-throughput (12) and conventional (37-39) biochemical analyses that have relied largely on protein overexpression, the number, size, dynamics, and stoichiometries of mammalian Hrd complexes and the effect of overexpression on these parameters have remained largely uninvestigated.
In this study we used genome editing to generate a human cell line in which we have replaced all copies of endogenous Hrd1 with a variant containing a TAP tag at the C terminus. This knock-in does not alter the genomic organization of the endogenous Hrd1 locus. We demonstrate that steady-state levels of tagged Hrd1 in these cells are indistinguishable from those of Hrd1 in unmodified cells and that the tagged variant is fully functional as judged by observing normal kinetics of degradation of two different ERAD substrates. These cells allow Hrd1 to be efficiently isolated from detergent lysates by sequential affinity capture, establishing them as a useful resource for biochemical and structural analysis of Hrd1 protein complexes.
Because of its central role in mediating ERAD, Hrd1 has been the focus of several prior biochemical and proteomic analyses (11, 40 -42), including our systematic analysis of the Hrd1 interactome (12). However, nearly all previous interaction studies have relied on overexpression of affinity-tagged Hrd1 and often overexpression of the interaction partner(s) as well. The data reported herein show that even modest (2-3-fold) overexpression alters the stoichiometry with nearly all of the interaction partners of Hrd1. We find that SEL1L and its partners, XTP3B and OS9, are reduced relative to Hrd1 by a factor of ϳ2. Other established members of the Hrd1 complex, Table 1 Analysis

of Hrd1 complex stoichiometry by AQUA-MS to determine the protein amount in tandem affinity-captured Hrd1 complexes
The absolute amounts of the indicated proteins within unfractionated Hrd1 complexes and the ratio between interactors are shown. The data represent the averages Ϯ S.E., combining measurements of all peptide standards per target across three biological replicates.

Protein
Absolute   including UBXD8, Derlin-2, Ube2J1, and HERP1, are reduced by more than 3-fold upon Hrd1 overexpression. It is possible that this apparent discrepancy could be trivially explained as a simple artifact of the fact that the data were normalized to Hrd1 levels. A simple dilution model would predict that all interactors would be subject to a similar reduction in level upon overexpression. However, we found that several proteins, notably Erlin1, Erlin2, and VCP, are strongly over-represented in Hrd1 captured material from overexpressing cells. Finally the complete absence of 26S proteasome subunits, which were prominently identified in our previously published Hrd1 interactome study (12) and were not detectably associated with Hrd1.KI at endogenous levels, does not support a trivial dilution argument. Our finding of non-native, disulfide-linked Hrd1 oligomers in overexpressing cells leads us to propose that homo-oligomerization of overexpressed Hrd1 displaces other interactors and promotes autoubiquitylation of the cytoplasmically exposed portions of Hrd1, thereby recruiting polyubiquitin binding proteins such as the 26S proteasome and VCP. These findings underscore the need to avoid even modest overexpression to preserve native subunit stoichiometries. SEC fractionation of affinity-purified Hrd1 complexes identified two distinct high molecular complexes centered at ϳ2 and 0.5 MDa that differ most notably in the relative content of the two major Hrd1 interaction partners, Fam8A1 and SEL1L (along with proteins such as OS9 and XTP3B, which bind to the complex via SEL1L). Previous studies using sucrose density gradient fractionation to analyze the size distribution of digitonin-solubilized, overexpressed mammalian Hrd1 complexes reported the identification of a 27S fraction containing Hrd1, OS9, SEL1L, and VCP, along with overexpressed XTP3B (18). However, in addition to differences caused by overexpression and detergent solubilization, SEC and density gradient sedimentation are influenced by fundamentally different physical properties including size, shape, and matrix interactions, thereby limiting direct comparisons of estimated size distributions obtained in previous studies. Comparisons between our data and previous analyses (26,43,44) of mammalian Hrd1 complexes are further complicated because previous work interrogated different subsets of Hrd1-interacting proteins (aside from the aforementioned SEL1L and associated proteins). Ours is the first study to use unbiased mass spectrometric identification to analyze the composition of SEC fractionated mammalian Hrd1 complexes. Of note, one previous study (11) applied similar unbiased metrics to analyze sucrose density gradient-fractionated, chromosomally tagged Hrd1 in Saccharomyces cerevisiae, identifying a complex with estimated molecular mass of ϳ500 kDa composed of the yeast orthologs of the same core components (Hrd3/SEL1L, Yos9/OS9) present in peak II in this study.
The finding in the present study of two high-molecular-mass peaks of Hrd1 in our SEC fractionation experiment could reflect an intrinsic heterogeneity in the number and size of stable endogenous Hrd1 complexes (perhaps with different substrate ranges or localization), a dynamic balance between multiple complexes, or the consequence of dissociation of a single, large complex upon cell lysis in DDM. The strong enrichment in peak I of Fam8A1, a very hydrophobic protein of unknown function that has no apparent ortholog in yeast, and its 1:2 stoichiometric relationship with Hrd1 suggests that Fam8A1 likely contributes to the structure and organization of a highmolecular-mass Hrd1 complex.
SEL1L (and its ortholog in yeast, Hrd3) is an obligate partner of Hrd1 and is required for its stability (45). It is essential for Hrd1-mediated degradation of proteins with luminal conformational lesions (i.e. ERAD-L) in both fungi and mammals (46,47). The very strong disenrichment of SEL1L from peak I is therefore difficult to reconcile with the conclusion that this high-molecular-mass complex is functional, at least for ERAD-L. Our data do not exclude the possibility that Hrd1 may participate, together with Fam8A1, Erlin2, and HERP, in a SEL1L-deficient high-molecular-mass complex dedicated to a specific class of ERAD substrates. For example, Hrd3 is not required for degradation of some integral membrane proteins in yeast upon Hrd1 overexpression and is dispensable for retrotranslocation of ERAD-L substrates in vitro (10). Perhaps the ϳ500-kDa (peak II) Hrd1 complex, which contains a complement of proteins similar to those present in the core yeast complex of similar molecular mass (11), is specialized for degradation of ERAD-M type glycoprotein substrates. Finally, our data do not exclude the possibility that mammalian Hrd1 normally exists as an ensemble of very-high-molecular-mass complex that mediates both ERAD-L and ERAD-M but dissociates upon cell lysis into the smaller complexes that we observe upon SEC fractionation in this study. Further study will be needed to distinguish between these possibilities.

TALEN and targeting vector construction
TALEN genomic binding sites were chosen to be 19 bp in length separated by a 16-bp spacer. The Hrd1 genomic sequences for TALEN binding are as follows: TALEN-L binding site: 5Ј-CCTGTTGCCCACTGACACT-3Ј and TALEN-R binding site: 5Ј-GCTCAAAAGAGCAGAGGCT-3Ј. TALE repeats were constructed using the two-step Golden Gate assembly method described previously (48). The final arrays of repeat variable di-residues were assembled into the MR015 plasmid (a kind gift from Dr. Matthew Porteus, Stanford University), which enables expression of FLAG-tagged TALE fused to a nuclear localization signal and the FokI nuclease domain. pDTA-TK vector (49) was used as the backbone to construct the Hrd1 targeting vector for homologous recombination. The vector was modified to contain a TAP tag and a puromycinresistant gene under the control of the phosphoglycerate kinase promoter. The amino-acid sequence for the TAP tag, designed as immediately preceding the C terminus of Hrd1, is as follows: linker (LE), Strep tag II (WSHPQFEK), PreScission protease cleavage site (LEVLFQGP), linker (LTGRT), and S tag (KETA-AAKFERQHMDS). A 519-bp 5Ј homologous arm (using the primer pairs, 5Ј-GTCGACATCATTTGGCCCTTGAG-TCC-3Ј and 5Ј-CTCGAGGTGGGCAACAGGAGACTCCA-3Ј) and a 516-bp 3Ј homologous arm (using the primer pairs, 5Ј-ATCGATTGACACTGCCCCAGCCCA-3Ј and 5Ј-GCTAGC-CCACCACCACAAGAGGCTTC-3Ј) were amplified by PCR from HEK293 genomic DNA and then cloned into the modified pDTA-TK vector.

Cell culture and transfection
HEK293 cells were cultured in DMEM containing 4.5 g/liter glucose and L-glutamine (Mediatech, Manassas, VA), and supplemented with 10% FetalPlex animal serum complex (Gemini Bio-Products, West Sacramento, CA) at 37°C and 5% CO 2 . The pair of TALENs recognizing Hrd1 loci and the targeting vector of Hrd1 fused to TAP tag were co-transfected into HEK293 cells using a standard calcium-phosphate co-precipitation technique (50). Stable cell lines were selected in 10 g/ml puromycin (Sigma-Aldrich), cloned by limiting dilution, and thereafter maintained in 1 g/ml puromycin. pSUPER shRNA expression construct targeting Hrd1 has been described previously (12). HEK293 cells overexpressing Hrd1-TAP or Hrd1-TAP* were generated using a standard calcium-phosphate co-precipitation technique followed by puromycin or G418 selection, respectively. Stable expression of Hrd1-TAP* was driven by a CMV promoter in pcDNA3.1. The TAP* sequence immediately preceding the C terminus of Hrd1 is as follows: linker (GTG), His 6 tag (HHHHHH), PreScission protease cleavage site (LEVLFQGP), linker (GGGT), and S tag (KETA-AAKFERQHMDS). The plasmid encoding HA-tagged NHK was previously described (20) and was transfected and expressed for 48 h prior to performing emetine-chase assays.

Emetine-chase assay
Cells were treated with emetine (Sigma-Aldrich) at a final concentration of 25 M for the indicated time. The cells were collected and washed in PBS before lysis in 1% SDS lysis buffer (50 mM Tris, pH 7.4, 150 mM NaCI, 1% SDS). Lysates were analyzed by immunoblotting.

Immunoblotting
Samples suspended in Laemmli buffer containing 2-mercaptoethanol 2% (v/v) were heated at 65°C for 10 min, separated by SDS-PAGE and transferred to PVDF (GE Healthcare Life Sciences). Laemmli buffer lacking 2-mercaptoethanol was used for non-reducing SDS-PAGE presented in Fig. 2D. PVDF membranes were blocked with 5% nonfat milk in PBS to reduce nonspecific antibody binding and incubated with primary antibodies diluted in PBS containing 0.2% Tween 20 and 3% BSA. Membranes were washed in PBS containing 0.2% Tween 20 and incubated with HRP-conjugated secondary antibodies. Immunoreactivity was detected using ECLϩ chemiluminescence reagents (GE Healthcare Life Sciences) for immunoblots presented in Fig. 1B and supplemental Figs. S1 and S2. Fluorescent IRDye secondary antibodies were used for all the remaining immunoblots and scanned by Odyssey imaging (LI-COR Biosciences). Band intensities were quantified by Image Studio Life software (LI-COR Biosciences).

Cell lysis and affinity purifications
HEK293 cells were harvested by scraping and washed in icecold PBS before lysis in buffer containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 2.5 mM EDTA, protease inhibitor cOmplete EDTA-free tablet (Roche Applied Science), and 1% (w/v) DDM (Anatrace, Maumee, OH) for 30 min at 4°C. The lysates were cleared by centrifugation at 20,000 ϫ g for 15 min at 4°C. Protein concentration was measured by BCA protein assay (Pierce). Normalized lysates were incubated with S proteinagarose (Novagen) for 4 -12 h at 4°C with mixing and washed four to five times in lysis buffer containing 0.2% DDM.
For tandem-affinity purification, protein complexes bound to S protein-agarose were incubated with GST-tagged Prescission protease (GE Healthcare Life Sciences) in 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.2% DDM, and 2 mM DTT for 4 -12 h at 4°C. Prescission protease was removed from the cleavage reaction by incubation with glutathione-Sepharose 4B (GE Healthcare Life Sciences) for 2-3 h at 4°C. Unbound supernatant was filtered with a 0.45-m PVDF syringe filter (Thermo Fisher Scientific). The resulting eluted affinity-purified protein complexes were analyzed by immunoblotting or subjected to SEC.
For antibody depletion assays, after removal of GST-tagged prescission protease, the affinity-purified protein complexes were incubated with anti-SEL1L, anti-Fam8A1, or no antibody with gentle rocking for 3 h at 4°C, followed by incubation with protein A/G Plus agarose bead (Pierce) for 1 h at 4°C. The unbound material was then analyzed by immunoblotting or subjected to size-exclusion chromatography.

Sample preparation for MS and AQUA-MS
For LC-MS/MS analysis, the affinity-purified protein complexes bound to MagStrep type 2 beads were washed three times with 50 mM ammonium bicarbonate (pH 8.0), eluted in 0.1% RapiGest SF (Waters, Milford, MA) in 50 mM ammonium bicarbonate (pH 8.0) at 37°C, and incubated with sequencing grade trypsin (Promega) for 16 h at 37°C. Samples were acidified with hydrochloric acid to pH 2, centrifuged at 20,000 ϫ g for 10 min, and analyzed using a linear ion-trap mass spectrometer as previously described (20).
AQUA peptide standards were synthesized and quantified by JPT (Berlin, Germany) or Sigma-Aldrich corresponding to tryptic peptides (stable isotope-labeled amino acids indicated by asterisks) for Hrd1 (ALEGHER*, HQFYPTV*VYLTK), SEL1L (LMTAYNSYK*, LTEEGSPK*), Fam8A1 (TAAGIS-TPAPVALGL*GPR, APHVQASVR*), and OS9 (IVRPWAE-GTEEGAR*). Lyophilized peptide standards were resuspended in 10% formic acid, vortexed, and diluted to 5 pmol/l in 0.1% formic acid before being combined as a master mix. 125 fmol of each peptide standard were spiked into affinity-purified protein complexes subjected to SEC (as described above) and bound to MagStrep type-2 beads in the presence of 50 mM ammonium bicarbonate, 6.8 M urea, and 5 mM tris-(2-carboxyethyl) phosphine (Pierce) and incubated for 30 min at room temperature. Samples were alkylated with 10 mM 2-chloroacetamide for 30 min in the dark at room temperature. 300 ng of mass spectrometry grade lysyl endopeptidase (Lys-C; Wako Chemicals) was added, and samples were incubated for 4 h at 37°C before dilution to 1 M urea with 50 mM ammonium bicarbonate and addition of 300 ng of sequencing grade trypsin (Sigma-Aldrich). The samples were further incubated for 12 h at 37°C before acidification with HCl to pH 2 and desalted with the stage-tip method (51).

Combined discovery and targeted (AQUA-MS) instrumentation and data analysis
Samples were analyzed by online capillary nano-LC-MS/MS. Peptide mixtures were separated on an in-house made 20-cm reversed-phase column (100-m inner diameter, packed with ReproSil-Pur C18-AQ 3.0-m resin (Dr. Maisch GmbH)) equipped with a laser-pulled nanoelectrospray emitter tip. Peptides were eluted at a flow rate of 400 nl/min using a two-step linear gradient including 3-25% buffer B in 70 min and 25-40% B in 20 min (buffer A: 0.2% formic acid and 5% DMSO in water; buffer B: 0.2% formic acid and 5% DMSO in acetonitrile) in a Dionex UltiMate 3000 HPLC system (Thermo Fisher Scientific). Peptides were then analyzed using a LTQ Orbitrap Velos mass spectrometer (Thermo Fisher Scientific). Data acquisition was executed in data-dependent mode with full MS scans acquired in the Orbitrap mass analyzer with a resolution of 60000 and m/z scan range of 340 -2000. A global parent mass list was enabled including the m/z of all the AQUA peptides and their corresponding light peptides, as well as the retention time windows (the time window was set to 4 min and centered around the retention time obtained from a previous regular data-dependent run). The top two most abundant ions from the list and top eight most abundant ions from MS1 with intensity threshold above 500 counts and charge states 2 and above were selected for fragmentation using collision-induced dissociation with an isolation window of 2 m/z, a collision energy of 35%, an activation Q of 0.25, and an activation time of 5 ms. The collision-induced dissociation fragments were analyzed in the ion trap with rapid scan rate. Dynamic exclusion was enabled with repeat count of 3 in 30 s and exclusion duration of 20 s. The AGC target was set to 500,000 and 1000 for full FTMS scans and ITMSn scans. The maximum injection time was set to 500 and 100 s for full FTMS scans and ITMSn scans.
The resulting spectra were searched against a "target-decoy" sequence database (52) consisting of the Uniprot human database (downloaded June 11, 2013 and containing 88,902 entries) and the corresponding reversed sequences using the SEQUEST algorithm (SEQUEST version 28, revision 12). The parent mass tolerance was set to 20 ppm, and the fragment mass tolerance was set to 0.6 Da. Enzyme specificity was set to trypsin, and the maximum number of internal cleavages was set to 3. Oxidation of methionines and acetylation of the protein N terminus were set as variable modification. Carbamidomethylation of cysteines was set as a static modification. The data were filtered to a 1% peptide and 5% protein false discovery rate using a linear discriminator analysis (53). The AQUA peptides were manually identified using Xcalibur. Peak areas were generated for each pair of heavy and light peptides by extracting-ion chromatograms within 10 ppm of the observed m/z for the monoisotopic peaks, as previously described (54).

MS instrumentation and data analysis
Other LC-MS/MS analyses were performed on a LTQ-Orbitrap model XL (ThermoFisher) using identical or similar settings to those described above. Peptide chromatographic separations were performed either in nanoflow mode on a 750-m inner diameter ϫ 10-cm BioBasic C18 column (New Objective, Inc.) or in microflow mode on a 300-m inner diameter ϫ 15-cm C18 column (CVI, Inc.). Peptides were eluted with a linear gradient of 5 to 45% acetonitrile in the presence of 0.1% formic acid. The data were analyzed manually in Xcalibur Qual browser software by extracting parent ion masses of standards and analytes using a mass window of 6 mTh after assuring that the mass measurements were obtained with adequate mass accuracy. Relative peak intensities for standards and analytes were measured on extracted, smoothed chromatographic peaks. Database searches were performed using Byonic software (Protein Metrics Inc.) using a variable modification set consisting of methionine oxidation, deamidation of asparagine and glutamine, pyroglutamate for N-terminal glutamine and glutamic acid, carbamidomethylation of cysteine, mono-and di-methylation of lysine residues on trypsin, and the appropri-ate heavy-isotope modifications on the proteins that had isotope-labeled standards. Searches were against a recent human protein database containing reverse-sequence decoys. Relative intensities of relevant peptides from proteins of interest were extracted from extracted-ion chromatograms as described previously (55).

Heat map analysis
The hierarchically clustered heat maps representing immunoblot and MS data were generated using MultiExperimental Viewer v4.7. Quantified protein levels from Immunoblot ( Fig.  3D and supplemental Table S2) or ion current intensity (supplemental Table S3A), and TSCs (supplemental Table S3B) from MS were normalized to the peak SEC fraction. The heat map scale of normalized intensities is from 0 (white) to 1 (red).