Ipom-F selectively inhibits ER translocation of the viral ORF8 and S proteins

To explore the ability of Ipom-F to inhibit the ER translocation of a small, yet structurally diverse, panel of SARS-CoV-2 membrane and secretory-like proteins, we first used a well-established in vitro translation system supplemented with canine pancreatic microsomes (Fig. 2A). To facilitate the detection of ER translocation, we modified the viral ORF8, S, envelope (E), membrane (M) and ORF6 proteins by adding an OPG2 tag – an epitope that supports efficient ER lumenal N-glycosylation and enables product recovery via immunoprecipitation, without affecting Ipom-F sensitivity (Fig. S1A and data not shown). For viral proteins that lack endogenous sites for N-glycosylation, such as the E protein, the ER lumenal OPG2 tag acts as a reporter for ER translocation and enables their recovery by immunoprecipitation. Where viral proteins already contain suitable sites for N-glycosylation (S and M proteins), the cytosolic OPG2 tag is used solely for immunoprecipitation. The identity of the resulting N-glycosylated species for each of these OPG2-tagged viral proteins was confirmed by endoglycosidase H (Endo H) treatment of the radiolabelled products associated with the membrane fraction prior to SDS–PAGE (Fig. 2B, lanes 1 and 2 in each panel).

Using ER lumenal modification of either endogenous N-glycosylation sites (viral S and M proteins) or the appended OPG2 tag (viral E and ORF8 proteins) as a reporter for ER membrane translocation, we found that 1 µM Ipom-F strongly inhibited both the translocation of the soluble, secretory-like protein ORF8–OPG2 and the integration of the type I transmembrane protein (TMP) S–OPG2, and truncated derivatives thereof (Fig. 2B,C; Fig. S1C). Furthermore, membrane insertion of the human type I TMP, ACE2, was inhibited to a similar extent (Fig. 2B,C; ∼70–90% inhibition for these three proteins). These results mirror previous findings showing that precursor proteins bearing N-terminal signal peptides, and which are therefore obligate clients for the Sec61-translocon, are typically very sensitive to Ipom-F-mediated inhibition (Zong et al., 2019). In the context of SARS-CoV-2 infection, wherein ACE2 acts as an important host cell receptor for the SARS-CoV-2 virus via its interaction with the viral S protein (Walls et al., 2020), these data suggest that an Ipom-F-induced antiviral effect might be achieved via selective reductions in the biogenesis of both host and viral proteins (Fig. 1A).

In contrast to the viral S and ORF8 proteins, insertion of the viral E protein was unaffected by Ipom-F (Fig. 2B,C), consistent with its recent classification as a type III TMP (Duart et al., 2020). Type III TMP integration is highly resistant to Ipom-F (Zong et al., 2019), most likely because such proteins exploit a novel pathway for ER insertion (Fig. 3). We therefore conclude that the known substrate-selective inhibitory action of Ipom-F at the Sec61 translocon is directly applicable to viral membrane proteins; whereby the ER translocation of secretory proteins and type I TMPs, but not type III TMPs, is efficiently blocked by Ipom-F.

The viral M protein is a multi-pass TMP with its first transmembrane domain (TMD) oriented so the N-terminus is exoplasmic (N_exo) and hence can be considered ‘type III-like’. Although human multi-pass TMPs of this type typically require both the ER membrane complex (EMC) and Sec61 translocon for their authentic ER insertion (Chitwood et al., 2018), Ipom-F had no significant effect on the ER translocation and insertion of the M protein in vitro, as judged by the efficiency of N-glycosylation of its N-terminal domain (Fig. 2C). We conclude that the integration of its first TMD is unaffected by Ipom-F, consistent with use of the EMC (Chitwood et al., 2018). There was, however, a qualitative reduction in the intensity of both the non- and N-glycosylated forms of the M protein when compared to the control (Fig. 2B; Fig. S1A). We speculate that this decrease may reflect an Ipom-F-induced effect on the Sec61-dependent integration of the second and/or third transmembrane spans of the M protein (Chitwood et al., 2018), and our future studies will aim to resolve this question. Nevertheless, like similar host cell multi-pass TMPs that are resistant to a comparable Sec61 inhibitor, mycolactone (Morel et al., 2018), the M protein appeared to be more resistant to Ipom-F than either the S or ORF8 proteins (Fig. 2; Fig. S1A). In practice, the potential resistance of this highly abundant and functionally diverse class of endogenous multi-spanning membrane proteins (von Heijne, 2007) may limit any Ipom-F-induced cytotoxicity towards host cells.

ORF6 assumes a lumenal-facing hairpin topology in ER-derived microsomes

Cell-based studies of the ORF6 protein from SARS-CoV-1 suggest it has an unusual hairpin topology, with both its N- and C-termini located on the exoplasmic side of the host cell membrane, to which it binds via an N-terminal amphipathic helix (Netland et al., 2007). To independently determine the membrane topology of SARS-CoV-2 ORF6, we prepared versions with OPG2 tags at both N- and C-termini, and single-tagged equivalents (see Fig. 2B schematics: OPG2–ORF6–OPG2, OPG2–ORF6 and ORF6–OPG2). Following membrane insertion, doubly tagged OPG2–ORF6–OPG2 showed significant amounts of species with 3- and 4-N-linked glycans (Fig. 2B). This pattern confirms that the SARS-CoV-2 ORF6 protein assumes a ‘hairpin’ conformation in the ER membrane with both its N- and C-termini in the lumen (Fig. 2B, OPG2–ORF6–OPG2). These 3- and 4-N-glycan bearing OPG2–ORF6–OPG2 species were also resistant to extraction with alkaline sodium carbonate buffer (Fig. S1D) and protected from added protease (Fig. S1E), further indicating that the majority of the ORF6 protein is stably associated with the ER membrane in a ‘hairpin’ (N_exo/C_exo) topology.

Consistent with this unusual membrane topology, we found no indication that the membrane insertion of any of our OPG2-tagged ORF6 variants was reduced by Ipom-F, strongly suggesting that its association with the inner leaflet of the ER membrane does not require protein translocation via the central channel of the Sec61 translocon (Gérard et al., 2020). We noted a subset of OPG2–ORF6–OPG2 species bearing only a single N-glycan was also clearly present in the membrane-associated fraction with or without Ipom-F treatment (Fig. 2B, OPG2–ORF6–OPG2, see 1Gly). Based on comparison to singly OPG2-tagged variants (Fig. 2B), we conclude that OPG2–ORF6–OPG2-1Gly has its N-terminus in the ER lumen, where only one of its two consensus sites is efficiently N-glycosylated (Nilsson and von Heijne, 1993), whilst its C-terminus is either ER lumenal but non-glycosylated or remains on the cytosolic side of the membrane. In the latter case, it may be that, in addition to its hairpin topology, some fraction of ORF6 may be integrated into ER-derived microsomes as a type III TMP (Fig. S2E; Netland et al., 2007) that is resistant to Ipom-F inhibition (Fig. 2C and Fig. 3D; Zong et al., 2019).

The molecular basis for SARS-CoV-2 protein sensitivity to Ipom-F

Having ascertained that Ipom-F inhibits the ER membrane translocation and/or insertion of the viral ORF8 and S proteins, but not of the ORF6, E or M proteins (Fig. 2C), we next investigated the molecular basis for this selectivity. For these studies we employed semi-permeabilised (SP) mammalian cells, depleted of specific membrane components via siRNA-mediated knockdown, as our source of ER membrane (Fig. 3A; Wilson and High, 2007). Based on the quantitative immunoblotting of target and control gene products (Fig. S2A–C), we selectively depleted HeLa cells for core components of the Sec61 translocon (knockdown of Sec61α, also known as SEC61A1; Sec61α-kd, ∼65% reduction), the EMC (knockdown of EMC5, also known as MMGT1; EMC5-kd, ∼73% reduction) and both together (Double-kd, ∼68% and ∼78% reduction of Sec61α and EMC5, respectively) prior to semi-permeabilisation with digitonin and use for in vitro ER translocation assays.

Following the analysis of total OPG2-tagged translation products recovered by immunoprecipitation, we found that: (i) the S protein and a truncated derivative were both more strongly affected by the depletion of Sec61α than of EMC5 (Fig. 3B; Fig. S2D); (ii) the ORF8 protein was likewise strongly affected by Sec61α depletion but was also sensitive to EMC5 depletion (Fig. 3C); and (iii) the E protein showed diminished insertion efficiency after knockdown of Sec61α and EMC5, although the latter had a more pronounced effect (Fig. 3D). In each case, the combined knockdown of Sec61α and EMC5 resulted in a reduction of membrane insertion that was either comparable to, or greater than, that achieved following the knockdown of Sec61α alone (Fig. 3B–D). For the ORF6 protein, the total level of N-glycosylated OPG2–ORF6–OPG2 species was unaffected by any knockdown condition tested (Fig. 3E). However, we noted a marked increase in the proportion of potentially mis-inserted OPG2–ORF6–OPG2-1Gly species, particularly after co-depletion of EMC5 and Sec61α (Fig. 3E; Fig. S2E). We speculate that the unusual hairpin topology of the ORF6 protein may be attributed to the EMC and Sec61 complex acting in concert to provide an Ipom-F-insensitive pathway for protein translocation across the ER membrane (data not shown). Perturbation of this pathway seemingly increases the potential for ORF6 to mis-insert (see Chitwood et al., 2018), perhaps as a consequence of disruption to the translocation of its C-terminus (Fig. S2E).

Taken together, our data establish that, analogous to human membrane and secretory proteins, the principal molecular basis for the Ipom-F-sensitivity of the SARS-CoV-2 ORF8 and S proteins is their dependence on Sec61-mediated protein translocation into and across the ER membrane. In contrast, the E, M and ORF6 proteins appear capable of exploiting one or more alternative membrane insertion and/or translocation pathways that can bypass the translocase activity of the Sec61 complex. These alternatives most likely include a route for type III TMP insertion that requires the insertase function of the EMC, which our data suggest is also sufficient to confer Ipom-F resistance to the viral E protein and at least the first transmembrane span of the viral M protein.

We conclude that Sec61-selective protein translocation inhibitors like Ipom-F hold promise as broad-spectrum antivirals that may exert a therapeutic effect by selectively inhibiting the ER translocation of viral and/or host proteins that are crucial to viral infection and propagation (Mast et al., 2020). In the context of SARS-CoV-2, integration of the viral S protein and its host cell receptor, ACE2, into the ER membrane was significantly reduced by Ipom-F (Fig. 2C and Fig. 3B). Likewise, translocation of the viral ORF8 protein across the ER membrane and into the ER lumen was substantially diminished (Fig. 2C and Fig. 3C). The binding of the viral S protein to cell surface ACE2 is a key step in host cell infection (Drew and Janes, 2020), whilst ORF8 may protect SARS-CoV-2-infected cells against host cytotoxic T lymphocytes (Zhang et al., 2020 preprint), making all three of these proteins viable therapeutic targets (Drew and Janes, 2020; Li et al., 2020; Young et al., 2020).

Like other small molecule inhibitors that target fundamental cellular pathways (Bojkova et al., 2020), the broad-ranging effects of Sec61 inhibitors on host cell membrane and secretory protein synthesis (Morel et al., 2018; Zong et al., 2019), including the strong in vitro effect of Ipom-F on ACE2 biogenesis (Groß et al., 2020), present an obvious hurdle to their future use. Nevertheless, given that Ipom-F is a potent inhibitor of Sec61-mediated protein translocation in cell culture models (Zong et al., 2019) and appears well tolerated in mice (Zong et al., 2020), we propose that future studies investigating its effect on SARS-CoV-2 infection and propagation in cellular models are clearly warranted (Bojkova et al., 2020).