The ER membrane protein complex is a transmembrane domain insertase

See allHide authors and affiliations

Science  14 Dec 2017:
DOI: 10.1126/science.aao3099


Insertion of proteins into membranes is an essential cellular process. The extensive biophysical and topological diversity of membrane proteins necessitates multiple insertion pathways that remain incompletely defined. Here, we found that known membrane insertion pathways fail to effectively engage tail-anchored membrane proteins with moderately hydrophobic transmembrane domains. These proteins are instead shielded in the cytosol by calmodulin. Dynamic release from calmodulin allowed sampling of the endoplasmic reticulum (ER), where the conserved ER membrane protein complex (EMC) was shown to be essential for efficient insertion in vitro and in cells. Purified EMC in synthetic liposomes catalyzed insertion of its substrates in a reconstituted system. Thus, EMC is a transmembrane domain insertase, a function that may explain its widely pleiotropic membrane-associated phenotypes across organisms.

The mammalian genome encodes hundreds of tail anchored (TA) membrane proteins with essential roles in diverse processes such as vesicular trafficking, apoptosis, signal transduction, and lipid biosynthesis (1). A single transmembrane domain (TMD) close to the C terminus mediates post-translational TA protein targeting and membrane insertion. Many TA proteins destined for the endoplasmic reticulum (ER) utilize the conserved TMD recognition complex (TRC) targeting pathway whose central component is the ATPase TRC40 (2). Structural studies of Get3, the yeast homolog of TRC40, have revealed a deep hydrophobic groove that binds and shields the hydrophobic TMD of TA proteins (3) until release at an ER-resident receptor complex (4, 5). The surface properties of the substrate-binding groove in Get3 is consistent with biochemical studies showing a preference for TMDs of high hydrophobicity (6, 7). Yet, the TMDs of ER-targeted TA proteins display a wide range of hydrophobicity and length (1). Whether or how the TRC pathway might handle this diversity is unclear.

The TMDs from eight ER-destined TA proteins of widely varying biophysical properties (fig. S1) were cloned into a standardized TA protein cassette (Fig. 1A) and shown to insert into ER-derived microsomes in vitro (Fig. 1B and fig. S2). However, only the three most hydrophobic TMDs interacted efficiently with TRC40 by native co-immunoprecipitation (Fig. 1B). Competitive inhibitors of TRC40 targeting dose-dependently reduced insertion of only the TA proteins that efficiently engaged TRC40 (Fig. 1C and fig. S3, A to C). The other TA proteins were completely resistant to inhibition. One of these resistant TMDs, from the ER-resident enzyme squalene synthase (SQS), became sensitive to TRC pathway inhibition when the hydrophobicity of its TMD was increased (Fig. 1D and fig. S1). This switch from resistance to sensitivity correlated with TRC40 interaction (Fig. 1E). Even when SQS was assembled with TRC40 in a purified system, the complex dissociated before appreciable insertion into ER microsomes (fig. S4).

Fig. 1 Detection of a non-TRC insertion pathway for TA proteins.

(A) Diagram of the TA protein reporter cassette used for most of the analyses in this study. (B) 35S-methionine labeled TA protein reporters with the indicated TMDs (see fig. S1) were translated in nucleased reticulocyte lysate (RRL) and incubated without or with canine pancreas-derived rough microsomes (cRMs). Glycosylation (‘+ glyc’) indicates successful insertion (see fig. S2). Relative hydrophobicity values for each TMD are shown. In a parallel experiment, reactions lacking microsomes for each protein were affinity purified via the substrate’s FLAG tag and analyzed for TRC40 association (by immunoblot) and substrate (by autoradiography). Identical results were obtained in native RRL and nucleased RRL (shown in figure). (C and D) Relative normalized insertion efficiencies for the indicated TA proteins with increasing amounts of the WRB-coiled coil (WRB-CC), a fragment of the TRC40 receptor at the ER (see fig. S3A). (E) An experiment as in (B) for a set of SQS mutants that successively increase TMD hydrophobicity (fig. S1). (F) Analysis of SQS and VAMP2 insertion using ER microsomes from HEK293 cells (hRM) or trypsin-digested hRM (tRM; see fig. S3D).

These observations indicated that the TRC pathway only handles relatively hydrophobic ER-destined TA proteins. Based on the approximate threshold for TRC40 dependence, we estimate that around half of TA proteins are inserted into the ER via a non-TRC pathway. This conclusion is consistent with variable degrees of insertion defects seen when the TRC pathway is impaired (8). The mechanism of non-TRC pathway insertion remains unclear, although earlier proposals include unassisted and Sec61-mediated insertion (9, 10). In support of a protein mediated process, SQS insertion was impaired into ER microsomes pre-treated with trypsin (Fig. 1F and fig. S3D). We thus used SQS as a model non-TRC substrate to identify cytosolic factor(s) that maintain its insertion competence and ER factor(s) needed for its insertion.

Size fractionation and chemical crosslinking were used to compare the cytosolic interactions made by the TMDs of SQS and VAMP2 (vesicle associated membrane protein 2), an established TRC pathway substrate. As documented previously (11, 12), VAMP2 interacted with each of the factors of the TRC targeting pathway: the chaperone SGTA, the Bag6 quality control complex, and TRC40 (Fig. 2A and fig. S5). The heterogeneous native size of VAMP2 by sucrose gradient fractionation reflects these multiple interactions (Fig. 2A). By contrast, SQS migrated as a smaller complex and failed to crosslink efficiently to any TRC pathway component (Fig. 2A and fig. S5). The primary crosslink seen with SQS was a ~20 kD Ca2+-dependent protein (Fig. 2B and fig. S5) that was identified by mass spectrometry as calmodulin (CaM), a factor shown previously to recognize hydrophobic domains in the cytosol (13).

Fig. 2 Identification of cytosolic factors that maintain TA protein insertion competence.

(A) 35S-methionine labeled SQS and VAMP2 were translated in native RRL, separated by size on a sucrose gradient, and subjected to chemical crosslinking of each fraction using amine- or sulfhydryl-reactive crosslinker (see fig. S5 for full gels). The graph shows the densitometry profiles of each substrate across the gradient, and the individual panels show regions of the crosslinking gels for the indicated interaction partners (verified by immunoprecipitation and mass spectrometry). (B) 35S-methionine labeled SQS translated in native RRL was treated with or without 1 mM EGTA before crosslinking and analysis by SDS-PAGE and autoradiography. The major SQS crosslinking partner (xCaM) is not seen with EGTA. Hemoglobin (Hb), its inter-subunit crosslink (Hb-Hb), and an unspecified translation product (*) are indicated. (C) 35S-methionine labeled SQS containing the benzoyl-phenylalanine photo-crosslinker within the TMD was produced as a defined complex with CaM using the PURE system (protein expression using recombinant elements; see fig. S6). The isolated SQS-CaM complex, prepared in 100 nM Ca2+, was incubated with cRM in the absence and presence of excess Ca2+ (either 0.2 or 0.5 mM), and analyzed directly (left) or irradiated with UV light to induce crosslinking before analysis (right). The glycosylated (‘+ glyc’) and CaM-crosslinked (‘xCaM’) products are indicated. (D) Schematic of the SQS insertion pathway, with a hypothetical membrane factor indicated.

Recombinant CaM was sufficient to prevent aggregation of SQS in a chaperone-free E. coli based translation system assembled from purified translation factors (fig. S6). Addition of ER microsomes to the SQS-CaM complex resulted in SQS insertion at efficiencies similar to that observed in total cytosol (Fig. 2C), while SQS synthesized in the absence of CaM was aggregated, and was not insertion competent (fig. S7). SQS insertion occurred concomitantly with release from CaM as monitored by site-specific photo-crosslinking (Fig. 2C). This suggested an insertion model where dynamic substrate release from CaM [at physiologic Ca2+ concentrations in the cytosol (13)] transiently provides opportunities for ER engagement before re-capture by CaM. In support of this model, insertion was precluded if the SQS-CaM complex was stabilized with super-physiologic levels of Ca2+ (Fig. 2C and fig. S8A), but occurred across the entire physiologic range of cytosolic free Ca2+ (fig. S8B). Furthermore, the unrelated TMD chaperone SGTA, which also associates with substrates dynamically (12), behaved similarly to CaM in supporting insertion of SQS in both complete cytosol (fig. S9) and purified systems (fig. S10). In contrast, the VAMP2-SGTA complex is insertion incompetent into ER microsomes unless complemented with TRC40 and the Bag6 complex (12). Thus, there appears to be a non-TRC pathway tuned to TMDs of moderate to low hydrophobicity. Unlike the highly coordinated TRC targeting system (2, 12), the alternate route can utilize any TMD-shielding factor capable of dynamically releasing substrate for attempts at membrane insertion (Fig. 2D). In native cytosol, the primary factor is CaM (fig. S11), although SGTA can substitute in its absence.

Trypsin sensitivity of the SQS insertion reaction (Fig. 1F) suggested that this critical step is protein-mediated. Taking a candidate approach, we considered factors that are conserved across eukaryotes, abundant, and cause pleiotropic membrane-associated phenotypes when deleted. In preliminary experiments, we observed no effect on SQS insertion of Sec61 inhibition or knockdown of Sec62 or Sec63, arguing against these possibilities (fig. S12). Although the SND genes are synthetic lethal with TRC pathway mutants in yeast (14), appreciable impairment of TA protein insertion was not seen in yeast or mammalian cells lacking the SNDs (14, 15). We then considered the ER-Membrane protein Complex (EMC), a widely conserved eight to ten subunit complex of unknown function (1618) (fig. S13A). EMC is genetically implicated in many unrelated membrane-associated processes such as quality control, trafficking, protein maturation, and lipid homeostasis (1722), but its biochemical activity has been elusive.

Using semi-permeabilized cultured cells as the source of ER (fig. S13B), we initially noticed that SQS insertion was partially impaired when EMC5 was depleted with siRNAs (fig. S13C). Ablation of EMC5 or EMC6 expression by gene editing of U2OS cells (fig. S14) reduced insertion of SQS but not VAMP2 (Fig. 3A). This deficiency was rescued by re-expression of EMC5 and EMC6 in the respective knockout cell lines. EMC-dependence was also observed when using ER microsomes isolated from HEK293 cells containing or lacking EMC6 (Fig. 3A). This phenotype was seen regardless of whether the substrates were prepared in crude cytosol (Fig. 3A) or provided as defined complexes with CaM (Fig. 3B) or SGTA (fig. S10B).

Fig. 3 The EMC is essential for TA protein insertion in vitro an in cells.

(A) Semi-permeabilized cells (see fig. S13B) from the indicated cell lines were tested for insertion of SQS and VAMP2 using the glycosylation assay. “-” indicates a control reaction lacking semi-permeabilized cells. (B) The isolated SQS-CaM complex (fig. S6) was tested for insertion into cRM or different amounts of hRM from wild type (WT) or ∆EMC6 (∆6) HEK293 cell lines. (C) Flow cytometry analysis of RFP-SQS and RFP-VAMP2, relative to an internal GFP expression control (see fig. S15A), in wild type (WT, grey), ∆EMC6 (red), or ∆EMC6+EMC6 (rescue, blue) cell lines. While the RFP:GFP ratio remains close to 1 for VAMP2 across a wide range of expression levels in all cell lines, SQS is selectively decreased in ∆EMC6 cells, especially at low expression levels (see fig. S15B for histograms of these data). (D) Tabulated mean RFP:GFP ratios for SQS (grey bars) and VAMP2 (black bars) in the indicated cell lines. The results for each construct were normalized to the value in wild type cells and depict mean ± SD from three independent experiments. (E) Immunoblots for SQS-RFP and VAMP2-RFP in the indicated cell lines. Loading was normalized to equivalent amounts of GFP expression as determined by flow cytometry. An aliquot of the WT sample digested with the glycosidase PNGase is shown as a marker for non-glycosylated substrate. Glycosylation of the ER-resident SQS is limited to the core N-glycan, while VAMP2 acquires complex glycans due to trafficking through the Golgi. (F) Live cell images of GFP-SQS in the indicated cell lines shows altered localization in ∆EMC6 cells. In low expressing cells (yellow arrows), the localization is diffusely cytosolic, while punctae presumably representing aggregates are seen in high expressing cells (red arrows). VAMP2 was unchanged in its localization in ∆EMC6 cells (fig. S15C). (G) Summary of dependence on either TRC40 (as judged by inhibitory effect of WRB-CC in Fig. 1) or EMC (see fig. S16) for the indicated substrates.

We exploited the fact that non-inserted TA proteins are typically degraded (23, 24) to analyze SQS insertion in cells. An RFP-tagged TA protein construct was varied to contain the TMD of either SQS or VAMP2 and analyzed for expression by flow cytometry, membrane insertion by glycosylation, and cellular location by microscopy. Relative to the nearly unimpaired RFP-VAMP2, RFP-SQS showed reduced expression (Fig. 3, C and D, and fig. S15, A and B), impaired glycosylation (Fig. 3E), and altered localization (Fig. 3F and fig. S15C) selectively in EMC knockout cells. Thus, in vitro and in cells, SQS insertion into the ER is dependent on EMC, the absence of which causes SQS mislocalisation, degradation, and aggregation.

Analysis of six other TA proteins and the five SQS TMD mutants showed that each TRC40-independent substrate is strongly EMC-dependent (Fig. 3G and fig. S16). Sec61β, a protein of moderate hydrophobicity, showed partial dependence on both EMC and TRC40, identifying the approximate point of overlap between these two pathways. Thus, the TRC- and EMC-dependent pathways are mostly tuned for TMDs of high and low hydrophobicity, respectively, although other features such as TMD length or helicity may also influence pathway choice. The lower hydrophobicity of clients for the EMC pathway presumably explains why a dedicated targeting pathway with constant TMD shielding is not needed, instead relying on temporary release from general TMD binding proteins to engage the membrane.

To determine whether the EMC is sufficient for TA protein insertion, we purified the intact 10-protein complex (Fig. 4A and fig. S17) and optimized conditions for its reconstitution into liposomes. The reconstituted EMC remained fully intact (fig. S18A), with approximately one-third of the complex oriented correctly (fig. S18B). In a protease-protection assay (Fig. 4B), SQS synthesized in native cytosol inserted into EMC-proteoliposomes with approximately half the efficiency observed in native ER microsomes (Fig. 4C). By contrast, VAMP2 insertion, which is efficient into ER microsomes from both wild type and ∆EMC6 cells, was poor in EMC6-proteoliposomes. EMC-proteoliposomes also supported insertion of the recombinant SQS-CaM complex at near-native levels of insertion relative to ER microsomes (Fig. 4D) when the amount of correctly oriented EMC was matched (fig. S18, B and C). As expected, SQS insertion was minimal into liposomes (Fig. 4D) or EMC-proteoliposomes pre-treated with trypsin (fig. S19).

Fig. 4 Reconstitution of EMC-dependent TA protein insertion with purified factors.

(A) Sypro Ruby stained gel of anti-FLAG affinity purification from HEK293 cells expressing untagged or FLAG-tagged EMC5. (B) Diagram of the protease-protection assay for TA protein insertion using a C-terminal epitope tag (red) to selectively recover the protected fragment (PF) diagnostic of successful insertion. (C) Liposomes reconstituted without or with purified EMC were analyzed for insertion of SQS or VAMP2 synthesized in native RRL. For comparison, native ER microsomes (hRM) from WT or ∆EMC6 HEK293 cells were tested in parallel. Immunoblot for EMC2 indicates the relative levels of EMC. As shown in fig. S18, roughly one-third of EMC in the proteliposomes is in the correct orientation. The graph represents four experiments (mean ± SD), normalized to insertion in wild type hRM. (D) Liposomes reconstituted with a constant amount of lipids and varying amounts of purified EMC were analyzed by protease-protection for insertion relative to wild type and ∆EMC6 hRM. The isolated SQS-CaM complex, an aliquot of which is shown in the last lane, was the substrate for these assays. The samples were also immunoblotted for EMC2 to visualize relative EMC levels. The graph represents four experiments (mean ± SD) normalized to insertion in wild type hRM.

The requirement for EMC in microsomes and in cells for SQS insertion, together with SQS insertion into liposomes at near-native efficiencies by purified EMC, rigorously establishes EMC as an insertase for moderately hydrophobic TMDs into the ER membrane. Bioinformatic analyses indicate that EMC3 is a distant homolog of Get1 (25) a subunit of the insertase for the TRC pathway (26). Both Get1 and EMC3 seem to have evolved from an ancestral prokaryotic insertase of the YidC family (25) apparently having acquired different substrate specificities in the process. The substrates that fail insertion without EMC probably contribute to many of its reported phenotypes such as ER stress (17), aberrant membrane protein trafficking or degradation (1821), altered lipid homeostasis (22), or altered viral replication (27).

Supplementary Materials

Materials and Methods

Figs. S1 to S19

References (2844)

This is an open-access article distributed under the terms of the Creative Commons Attribution license, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

References and Notes

Acknowledgments: We are grateful to M. Skehel and his team for mass spectrometry analysis, P. Chitwood and B. Phillips for discussions and help with microsome and EMC preparation, R. Voorhees and S. Shao for useful discussions, S. Juszkiewicz for help with flow cytometry analysis, M. Kivlen for help preparing the PURE system, and R. Keenan, E. Miller, and S. Juszkiewicz for critical reading of this manuscript. This work was supported by the UK Medical Research Council (MC_UP_A022_1007 to RSH, and MR/L001209/1 to JCC) and the Ludwig Institute for Cancer Research (JCC). AG was supported by a Gates Cambridge Scholarship from the Gates Foundation. Additional data can be found in the supplementary materials. This work is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) license, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. To view a copy of this license, visit This license does not apply to figures/photos/artwork or other content included in the article that is credited to a third party; obtain authorization from the rights holder before using such material.
View Abstract

Navigate This Article