Decoupling SARS-CoV-2 ORF6 localization and interferon antagonism

Severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) is a novel coronavirus responsible for the coronavirus disease 2019 (COVID19) pandemic, which continues to be a persistent problem worldwide despite the availability of highly effective vaccines. A major barrier to widespread immunity has been the emergence of variants that exhibit significantly increased transmission rates. Pathogenicity is further enhanced through activities of several viral proteins that suppress type I interferon (IFN) production and signaling. For instance, SARS-CoV-2 NSP1, NSP6, NSP13, ORF6, ORF7b, ORF8 and nucleoprotein antagonize IFN signaling to various degrees, with ORF6 exhibiting the strongest inhibition (Lei et al., 2020; Miorin et al., 2020; Xia et al., 2020).

ORF6 is a 7 kDa accessory protein required for optimal virus replication in vitro and in vivo – likely through its ability to potently suppress IFN signaling (Silvas et al., 2021; Zhao et al., 2009; Zhou et al., 2010). Several studies have demonstrated that ORF6 induces cytoplasmic accumulation of nuclear transport proteins importin-α and importin-β, subsequently inhibiting nuclear translocation of activated STAT1 and IRF3 transcription factors (Frieman et al., 2007; Lei et al., 2020; Miorin et al., 2020). Additional mechanistic clarification was recently provided through proteomics studies that identified the Nup98–Rae1 nuclear pore complex as a high-confidence ORF6 interactor (Gordon et al., 2020; Miorin et al., 2020). Further investigation revealed that a direct interaction between the C-terminal tail of ORF6 and the C-terminal domain of Nup98 impairs docking of cargo–receptor complexes to inhibit nuclear trafficking (Addetia et al., 2021; Kato et al., 2021; Miorin et al., 2020).

Although suppression of innate immune signaling is believed to be its primary function, ORF6 also packages into nascent viral particles and induces membrane structures resembling viral replication compartments (Huang et al., 2007; Zhou et al., 2010). These activities coincide with ORF6 distribution, as it localizes to several membranous compartments during infection and when exogenously expressed in several cell types (Gunalan et al., 2011; Kumar et al., 2007). Both native and epitope-tagged ORF6 proteins colocalize with markers for the Golgi (Kato et al., 2021; Kopecky-Bromberg et al., 2007; Lei et al., 2020; Miorin et al., 2020; Xia et al., 2020; Zhou et al., 2010), endoplasmic reticulum (ER) (Kopecky-Bromberg et al., 2007; Lee et al., 2021; Lei et al., 2020; Zhou et al., 2010), endosomes (Gunalan et al., 2011; Kumar et al., 2007; Lee et al., 2021) and nuclear envelope (Addetia et al., 2021; Kato et al., 2021; Miorin et al., 2020), which is consistent with the current paradigm that ORF6 is a transmembrane protein capable of lateral diffusion within membrane networks (Netland et al., 2007; O'Keefe et al., 2021; Zhou et al., 2010). Taken together, these observations lend weight to an appealing model whereby ORF6 localization at the nuclear envelope facilitates a direct interaction with Nup98–Rae1 to subsequently inhibit nuclear trafficking; however, this direct cause-and-effect relationship has yet to be examined experimentally.

Here, we investigate the link between ORF6 localization and IFN antagonism. Through an extensive panel of truncation and single amino acid substitution mutants in combination with pharmacological experiments, we demonstrate that ORF6 associates with membranous compartments through two distinct structural determinants and is most likely a peripheral membrane protein. The first determinant resides within amino acid residues ¹⁸IMRTFKV²⁴ and is required for Golgi retention, whereas the second determinant encompasses two putative amphipathic helices that maintain steady-state localization and membrane association. Importantly, mis-localized variants induced cytoplasmic accumulation of importin-α (herein referring to importin-α1, also known as KPNA2), inhibited nuclear translocation of activated STAT1, and suppressed IFN signaling. In further support of these observations, a peptide inhibitor that contains the ORF6–Nup98 interaction motif robustly blocked nuclear accumulation of importin-α. Taken together, these results demonstrate that ORF6 membrane association is dispensable for interferon antagonism, raising the possibility that ORF6 might have additional functions within membrane networks.

Native and epitope-tagged SARS-CoV-2 ORF6 proteins localize to several membranous organelles; however, determinants that dictate the subcellular distribution have yet to be thoroughly investigated. To identify which protein regions mediate steady-state localization, a panel of ORF6–mCherry truncation mutants were generated based on a computational model to avoid disrupting putative structural elements, as a crystal structure has yet to be solved. Because the C-terminal region of ORF6 interacts with Nup98–Rae1 and is required for inducing cytoplasmic accumulation of importin-α (Frieman et al., 2007; Miorin et al., 2020), we reasoned that any cis-acting localization determinants reside upstream of this domain. As expected, expression of amino acid residues 1–47 had a localization comparable to wild type, whereas expression of residues 48–61 exhibited no discernable pattern (Fig. 1A,B). Unexpectedly, when we further examined the 1–47 segment for cis-acting localization determinants, we were surprised to observe two distinct patterns. Expression of the first half of this segment (amino acid residues 1–23) resulted in partial colocalization with the Golgi marker, while expression of the second half (amino acid residues 24–47) resulted in localization to an organelle distinct from the ER (Fig. 1A,B). When the 1–23 segment was further truncated to residues 1–17, no localization pattern was observed, suggesting that Golgi retention is partially mediated by amino acid residues ¹⁸IMRTFKV²⁴ (Fig. 1A,B).

Because the 1–23 fragment only partially recapitulated Golgi retention, we reasoned that the putative helix within residues 24–47 is required for maintaining steady-state localization, since when expressed alone it non-specifically associated with a membranous compartment. Therefore, we generated a truncation construct that expressed amino acid residues 18–47 anticipating it would mimic wild-type distribution; instead, this mutant localized to the ER-adjacent organelle, suggesting that the putative helix contained within the 1–17 fragment also contributes to steady-state localization (Fig. 1A,B). These observations suggested that ORF6 maintains steady-state localization through at least two distinct determinants, a longer protein component from residues 1–47 that mediates steady-state membrane associations, and a second region within ¹⁸IMRTFKV²⁴ that dictates Golgi retention. To test this model, we initially focused on ¹⁸IMRTFKV²⁴ to determine whether it harbors a Golgi retention motif. Mutation of ¹⁸IMRTFKV²⁴ to alanine residues in full-length ORF6 (denoted 18-24Ala) not only disrupted Golgi accumulation but induced freely diffuse intracellular puncta (Fig. 1C,D). Further investigation into the conservation of this motif revealed that SARS-CoV-1, bat coronavirus (bat-CoV) and pangolin coronavirus (pangolin-CoV) ORF6 proteins also require this motif to facilitate Golgi retention (Fig. 1C,D). Of note, SARS-CoV-1 and bat-CoV 18-24Ala mutants did not form intracellular puncta as compared to SARS-CoV-2 and pangolin-CoV ORF6 but colocalized with the ER marker, suggesting there is an inherent difference in the way these proteins associate with membranes (Fig. 1C,D). It is possible this difference is attributable to the putative helix from residues 24–47, as there is only ∼50% amino acid identity across species.

While no strict consensus motif has been defined for Golgi retention, numerous Golgi-resident proteins maintain steady-state localization through motifs enriched with positively charged amino acid residues (Tu et al., 2012; Wang et al., 2020); and interestingly, ¹⁸IMRTFKV²⁴ contains arginine and lysine residues at positions 20 and 23, respectively. To test these residues for Golgi retention, we generated single and double amino acid substitution mutants and assessed their localization. Independently exchanging each residue for a glutamate residue had minimal impact on localization; however, the RK20,23EE double mutant lost Golgi retention and localized to the ER-adjacent organelle, raising the possibility that electrostatic interactions facilitate ORF6 targeting to the Golgi (Fig. 1E, bottom rows). This speculation is intriguing given that the Golgi membrane is enriched with phosphatidylinositol-4-phosphate lipids, which unlike phosphatidylcholine, ethanolamine and serine, contain negatively charged headgroups.

Because several of the ORF6 mutants localized to an ER-adjacent organelle, we co-expressed relevant mutants with markers for the most likely candidates to identify this compartment (i.e. mitochondria and the ER-Golgi intermediate compartment). Interestingly, all ER-adjacent mutants exhibited significant colocalization with mitochondria, but not with the ER-Golgi intermediate compartment (Fig. 1F,G and data not shown).

While fine-mapping localization determinants, several of the mutants exhibited colocalization with the mitochondria (SARS-CoV-2 18–24, 24–47 and RK-EE) or ER (SARS-CoV-1 and bat-CoV 18-24Ala), raising the possibility that a second determinant drives association with membranous compartments (Fig. 1). Because it has been reported that transmembrane domains of Golgi-resident proteins contribute to steady-state localization (Banfield, 2011; Hu et al., 2011; Wang et al., 2013), we reasoned that the putative α-helices contribute to localization. To explore this possibility, we closely examined the amino acid composition of these helices looking for clues as to how they might associate with membranes. We were surprised to discover that ORF6 exhibits a biased hydrophobic index and is predicted to be an amphipathic protein (Fig. 2A,B). From these observations, we postulated two models to explain how ORF6 could be amphipathic and localize to membranous compartments. First, ORF6 is a transmembrane protein that forms higher order homo- or hetero-oligomers that shield the hydrophilic surface from the hydrophobic membrane environment. Second, ORF6 is a peripheral membrane protein that orients hydrophilic helical surfaces toward the cytoplasm and buries hydrophobic portions within membrane surfaces. To explore these models, we generated and tested single amino acid substitution mutants that exchanged a wild-type residue for one with opposing biophysical properties on each respective helical surface. As depicted in Fig. 2, polar residues were replaced with tryptophan, charged residues were exchanged with residues with opposing charge and hydrophobic residues were replaced with glutamate residues, and localization was assessed. If ORF6 were a transmembrane protein that underwent intramembrane oligomerization, we postulated that substitutions made to either surface would disrupt localization. However, if ORF6 were a peripheral membrane protein, only substitutions made to the hydrophobic membrane-interacting surface would disrupt localization. In support of the latter scenario, proteins with substitutions on the hydrophilic surface exhibited normal localization; however, all substitutions on the hydrophobic surface caused mis-localization (Fig. 2C,D). Taken together, these results suggest that SARS-CoV-2 ORF6 is likely a peripheral membrane protein that maintains steady-state localization through the ¹⁸IMRTFKV²⁴ region and this contiguous hydrophobic surface.

To further explore the possibility that SARS-CoV-2 is a peripheral membrane protein, we examined localization patterns in the presence of brefeldin A (BFA) and cycloheximide (CHX). BFA is a well-characterized fungal toxin that promotes Golgi disassembly by inhibiting trafficking between Golgi and ER compartments, subsequently resulting in redistribution of Golgi proteins to the ER (Lippincott-Schwartz et al., 1989). We reasoned that if ORF6 peripherally associates with Golgi membranes, BFA treatment would induce cytoplasmic accumulation, whereas if it is a transmembrane protein, it would relocalize to the ER. The inclusion of the protein translation inhibitor CHX allows for tracking steady-state ORF6 rather than observing newly synthesized protein that may artificially accumulate in the cytoplasm due to loss of the Golgi. As depicted in Fig. 3A, cells treated with BFA exhibited loss of Golgi stacks and subsequent redistribution of the Golgi marker to the ER. Importantly, BFA treatment induced SARS-CoV-2 ORF6 puncta accumulation in the cytoplasm, which resembled what was seen with the 18-24Ala mutant lacking the Golgi retention motif (Figs 1C,D, and 3A). Interestingly, we did not observe this phenomenon for SARS-CoV-1 ORF6, which fully redistributed to the ER, also mimicking its cognate 18-24Ala mutant (Figs 1C,D, and 3A). Next, we wanted to determine whether SARS-CoV-2 ORF6 puncta could re-associate with nascent Golgi membranes following BFA washout. As shown in Fig. 3B, BFA washout resulted in redistribution of the Golgi marker from the ER to discrete puncta in the cytoplasm, which were most likely reassembling Golgi stacks. Interestingly, SARS-CoV-2 ORF6 colocalized with these Golgi puncta following washout, whereas SARS-CoV-1 ORF6 remained colocalized with the ER, suggesting that the cytoplasmic SARS-CoV-2 ORF6 proteins could more readily associate with nascent Golgi stacks (Fig. 3B). To further confirm that BFA treatment induced redistribution of SARS-CoV-2 ORF6 proteins to the cytoplasm, we treated cells with BFA and then with digitonin. Digitonin is a non-ionic detergent that selectively permeabilizes the plasma membrane while leaving other compartments intact. If SARS-CoV-2 ORF6 is indeed released from the Golgi membrane following BFA treatment, we hypothesized that it would deplete from cells following digitonin treatment. To ensure that digitonin treatment would not impact Golgi distribution, we examined cells expressing mCherry and the Golgi marker. As expected, digitonin treatment, but not BFA treatment, resulted in a significant depletion of mCherry fluorescence (Fig. 3C). Furthermore, the Golgi marker only exhibited differential distribution in the presence of BFA, not digitonin, indicating that Golgi structures were not grossly disrupted. When cells expressing SARS-CoV-2 ORF6 were treated with digitonin alone, no impact on protein abundance or localization was observed (Fig. 3C). However, when cells were treated with BFA and then with digitonin, we observed significantly decreased fluorescence intensity for ORF6 but not the Golgi marker (Fig. 3C). Taken together, these observations combined with the mutational data in Figs 1 and 2 strongly support the model that SARS-CoV-2 ORF6 is a peripheral membrane protein, and likely explains how some variants can non-specifically associate with membranous compartments (i.e. 18–24, 24–47 and RK-EE mutants).

ORF6 inhibits interferon synthesis and signaling by blocking nuclear translocation of activated STAT1 and IRF3 transcription factors through a direct interaction with Nup98–Rae1 (Frieman et al., 2007; Kato et al., 2021; Kopecky-Bromberg et al., 2007; Lei et al., 2020; Miorin et al., 2020; Xia et al., 2020). The most likely scenario is that ORF6 accumulation at the nuclear envelop facilitates a direct interaction with Nup98–Rae1 to block nuclear trafficking; however, this relationship has yet to be explored experimentally. To determine whether membrane association is required for inhibiting nuclear trafficking, we examined a panel of localization disrupted mutants for their ability to block nuclear accumulation of eGFP–importin-α. To ensure that ORF6 function was not altered due to the presence of the C-terminal mCherry tag, we generated a mCherry–T2A-ORF6 ‘self-cleaving’ expression cassette, which allows for efficient detection of transfected cells without having to add an epitope tag to ORF6 (Salamango et al., 2019). For these experiments, we stably introduced eGFP–importin-α in the type II alveolar lung epithelial cell line A549 to test ORF6 activity in a more physiologically relevant cell model. Remarkably, all mutants tested induced cytoplasmic accumulation of eGFP–importin-α at efficiencies comparable to wild type (Fig. 4A,C). Next, we wanted to confirm that these mutants could also block nuclear accumulation of activated STAT1. To activate STAT1 and induce cytoplasmic-to-nuclear translocation, A549 cells were treated with type I IFN in the presence or absence of the indicated ORF6 proteins. As anticipated, all mutants could inhibit nuclear accumulation of STAT1 at efficiencies comparable to wild type (Fig. 4B,D). As a complementary approach, we also tested these mutants for inhibition of an IFNβ–eGFP reporter construct. Reporter activation was assessed by quantifying eGFP abundance in the presence and absence of the indicated ORF6 proteins following co-expression with a RIG-I (also known as DDX58) activating mutant (Mibayashi et al., 2007). This RIG-I mutant lacks its autoinhibitory domain and can constitutively active the IFNβ reporter in the absence of an RNA substrate. As depicted in Fig. 4E, both wild-type and mutant ORF6 proteins efficiently suppressed eGFP expression following stimulation with the RIG-I mutant. Finally, to further confirm that membrane association is not required for inhibition of nuclear trafficking, we generated a peptide inhibitor construct by fusing three copies of the ORF6–Nup98 interaction domain to mCherry. As depicted in Fig. 4F,G, expression of this construct was sufficient to drive cytoplasmic accumulation of eGFP–importin-α.

Based on our findings here, it is evident that ORF6 localization is independent from IFN antagonism and raises the possibility that ORF6 may have additional functions within membrane networks to enhance viral replication.

A549, HEK293FT and HeLa cells (American Type Culture Collection) were maintained in RPMI medium (Hyclone, South Logan, UT, USA) supplemented with 10% fetal bovine serum (FBS; Gibco, Gaithersburg, MD) and 0.5% penicillin-streptomycin (50 units; Gibco, Gaithersburg, MD, USA). Cells were transfected with polyethylenimine (PEI) using a ratio of 3 µl per 1 µg of DNA. To generate the A549 eGFP–importin-α stable cell line, roughly 250,000 HEK293FT cells in a six-well plate were transfected with a pQCXIH retroviral vector (Clontech) containing the eGFP–importin-α expression cassette, an MLV-GagPol packaging vector, and a VSV-G vector (Addgene). Medium was harvested at 48 h post transfection and frozen at −80°C for 4–6 h, thawed and centrifuged at 1500 g, and overlaid on A549 cells. To generate a pure cell population, cells were treated with hygromycin B (Sigma-Aldrich, 500 µg/ml) at 48 h post transduction. All ORF6 mutants were generated by PCR amplification using Phusion high fidelity DNA polymerase (NEB, Ipswich, MA, USA) and overlapping PCR to introduce the indicated mutations (primer sequences available upon request). UniProt accession numbers for bat-CoV and pangolin-CoV ORF6 proteins are A0A166ZLC5 and A0A6G9KMH4, respectively. To generate the peptide inhibitor construct, we fused 3× tandem copies of the ORF6–Nup98 interaction domain separated by 12-amino-acid glycine-serine-threonine linkers (repeat peptide sequence: KYSQLDEEQPMEID) to the C-terminus of mCherry. The IFNβ reporter construct was generated by cloning an IFN responsive promoter element upstream of eGFP in the pcDNA-5TO expression vector (Addgene) and the constitutively active RIG-I mutant vector was generated by cloning RIG-I amino acids 1–242 into an mTagBFP2-T2A expression cassette in a lentiviral vector (modified from Salamango et al., 2019). All constructs were confirmed by restriction digestion and Sanger sequencing.

For experiments using the IFNβ and RIG-I expression constructs, roughly 250,000 A549 cells were seeded into 12-well culture plates and allowed to adhere overnight. The next day, cells were co-transfected with the indicated combinations of IFNβ–eGFP reporter, activating RIG-I, and ORF6 expression plasmids at 100 ng DNA per construct. At 48 h post transfection, eGFP fluorescence was quantified (ImageJ software) and then graphed with Prism 6.0 (GraphPad Software).

All localization and immunostaining experiments were repeated at least three independent times and representative images are depicted from surveying at least five fields of view from each condition with at least 25–30 cells exhibiting similar phenotypes (all scale bars shown are at 10 µm). In addition, localization experiments were carried out using HeLa cells, as they are routinely used for microscopy experiments and localization studies due to their morphological characteristics. Roughly 6500 HeLa cells were seeded into an eight-well #1.5 glass bottom chamber slide (Ibidi #80826) and transfected with 100 ng of the indicated ORF6 expression construct along with either 50 ng of an ER marker (eGFP-Calnexin; Addgene 57122), 50 ng of a Golgi marker [mTag-β-galactosidase (1-61); Addgene], or 50 ng of a mitochondrial marker (eBFP2-Mito, Addgene 55248). The next day, cells were imaged using a 60× oil immersion objective on an EVOS M5000 fluorescence microscope.

Prior to immunolabeling and imaging of STAT1, A549 cells were treated with 3000 U/ml type I interferon for 45 min. Immunolabeling of STAT1 was performed as follows. At 24 h post transfection, cells were washed with PBS and fixed in 4% paraformaldehyde (PFA) at room temperature for 10 min. Following fixation, cells were washed using PBS plus 0.3% Triton X-100 (PBST) three times in 5-min intervals and then blocked using PBST supplemented with 5% BSA, 10% goat serum, and 0.3 M glycine for 1 h at room temperature. After blocking, samples were incubated with primary anti-STAT1 antibody (1:300; Cell Signaling Technology, 9172) in PBST with 5% BSA overnight at 4°C. The next day, samples were washed three times with PBS and then incubated with secondary anti-rabbit-IgG conjugated to Alexa Fluor 488 (1:750; Cell Signaling Technology, 4412) and anti-mCherry conjugated to Alexa Fluor 594 (1:750; Invitrogen #M11240) in PBS with 5% BSA for 1 h at room temperature. Finally, cells were washed three times with PBS and then imaged at with a 60× objective on an EVOS M5000 fluorescence microscope. For quantification of nuclear fluorescence, individual cells expressing the indicated ORF6 proteins were scored for importin-α or STAT1 localization by dividing the nuclear fluorescence intensity by the cytoplasmic fluorescence intensity (n=50 for importin-α and n=25 for STAT1), and then graphed with Prism 6.0 (GraphPad Software). Subcellular compartments were defined based on DAPI staining (nucleus).

For BFA and CHX inhibitor treatments, culture medium was replaced with complete medium (RPMI with 10% FBS and 0.5% penicillin-streptomycin) supplemented with 5 µM BFA and 2 µM CHX and allowed to incubate for 30 min prior to imaging. For inhibitor washout, samples were rinsed with PBS following an initial 30-min BFA and CHX incubation and then complete medium supplemented with 2 µM CHX was added. Cells were allowed to recover for 30 min before imaging reconstitution of the Golgi. For digitonin treatments, transfected cells were either left untreated, treated with 50 µg digitonin for 10 min, treated with 5 µM BFA and 2 µM CHX for 30 min, or, treated with 5 µM BFA and 2 µM CHX for 30 min and then with 50 µg digitonin for 10 min.

All flow cytometry experiments were repeated three independent times, and representative histograms are depicted from one experiment. Quantification of fluorescence intensity was performed using a Becton Dickinson FACScan flow cytometer. Briefly, roughly 125,000 HeLa cells were seeded into a 12-well plate and transfected 24 h after plating with 300 ng of the indicated ORF6 and/or fluorescent protein DNA and 200 ng of DNA for the Golgi marker. The next day, cells were treated as described above with the following exception: digitonin treatment was not performed until after cells were removed from culture plates to better preserve cell integrity. Following inhibitor treatment, cells were removed from plates using 0.025% Trypsin-EDTA solution, centrifuged at 300 g for 5 min, and then re-suspended in 2% FBS+5 µM BFA/2 µM CHX in PBS. At this point, 50 µg of digitonin was added for 10 min and samples were subjected to flow cytometry and analyzed using FloJo software.

Statistical analyses were performed using an unpaired two-tailed Student's t-test in the GraphPad Prism software. Pearson's correlation coefficients were determined using the colocalization plugin contained within ImageJ software.

We thank Drs Erich Mackow, Nancy Reich-Marshall and Patrick Hearing for intellectual discussions and constructive feedback.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.259666.