Quantitative comparison of nuclear transport inhibition by SARS coronavirus ORF6 reveals the importance of oligomerization

Significance

SARS coronavirus Open Reading Frame 6 (ORF6) proteins block nucleocytoplasmic transport to evade host immunity, but the precise mechanism has been controversial. Solving this problem will help us understand SARS evolution and pathogenesis. To address this question, we made quantitative measurements of the concentration dependence of nuclear transport inhibition by ORF6, and informative mutants, in single living cells. SARS-CoV-2 ORF6 was strikingly more potent than SARS-CoV-1 ORF6, mainly due to its shorter C terminus. The N-terminal region promoted ORF6 oligomerization which was required for multivalent interaction with nuclear pores and could be replaced by synthetic oligomerizers. Our methods could be adapted to measure the concentration dependence of any protein whose effect can be scored in single live cells.

Abstract

Open Reading Frame 6 (ORF6) proteins, which are unique to severe acute respiratory syndrome-related (SARS) coronavirus, inhibit the classical nuclear import pathway to antagonize host antiviral responses. Several alternative models were proposed to explain the inhibitory function of ORF6 [H. Xia et al., Cell Rep. 33, 108234 (2020); L. Miorin et al., Proc. Natl. Acad. Sci. U.S.A. 117, 28344–28354 (2020); and M. Frieman et al., J. Virol. 81, 9812–9824 (2007)]. To distinguish these models and build quantitative understanding of ORF6 function, we developed a method for scoring both ORF6 concentration and functional effect in single living cells. We combined quantification of untagged ORF6 expression level in single cells with optogenetics-based measurement of nuclear transport kinetics, using methods that could be adapted to measure concentration-dependent effects of any untagged protein. We found that SARS-CoV-2 ORF6 is ~15 times more potent than SARS-CoV-1 ORF6 in inhibiting nuclear import and export, due to differences in the C-terminal region that is required for the NUP98–RAE1 binding. The N-terminal region was required for transport inhibition. This region binds membranes but could be replaced by synthetic constructs which forced oligomerization in solution, suggesting its primary function is oligomerization. We propose that the hydrophobic N-terminal region drives oligomerization of ORF6 to multivalently cross-link the NUP98–RAE1 complexes at the nuclear pore complex, and this multivalent binding inhibits bidirectional transport.

Macromolecular transport across the nuclear envelope occurs through nuclear pore complexes (NPCs) which are densely filled with intrinsically disordered regions of the FG-nucleoporins (FG-NUPs) enriched in phenylalanine–glycine (FG) repeats (1). The FG-NUPs in the central scaffold, including NUP54, NUP58, NUP62, and NUP98, constitute a permeability barrier that interacts specifically with transporter proteins to facilitate their diffusion through the NPC with their cargoes bound (1–3). The classical nuclear import pathway uses importin beta-1 (KPNB1) as the transport carrier and adaptor proteins in the importin alpha family (KPNAs) to bind nuclear localization signals (NLSs) (4). The classical nuclear export pathway is mediated by exportin-1 (CRM1/XPO1) which directly binds to leucine-rich nuclear export signals (NESs) (5). Directionality is provided by compartment-specific assembly and disassembly of cargo-carrier complexes mediated by the small GTPase RAN (6).

Nucleocytoplasmic transport is required for several innate immune anti-viral responses; for example, the interferon (IFN) response requires nuclear import of phosphorylated STAT1 and IRF3 (7). SARS coronaviruses, like many other viruses, were reported to inhibit the nucleocytoplasmic transport system of infected host, resulting in reduced innate immune signaling (8–14). A small protein that is unique to SARS coronaviruses, Open Reading Frame 6 (ORF6), was shown to inhibit the KPNB1-mediated classical nuclear import pathway (15). Several inhibition mechanisms were proposed. Early studies proposed that SARS-CoV-1 ORF6 tethers KPNA2 to membranes of the endoplasmic reticulum (ER) and Golgi apparatus to sequester KPNB1 (16). Interaction of SARS-CoV-2 ORF6 with KPNA2 was also demonstrated (8, 9). Contrarily, a recent mutagenesis study showed that membrane association is not required for the IFN antagonism by ORF6 (17). Recent studies on SARS-CoV-2 ORF6 showed that it binds to the NUP98–RAE1 heterodimer and implicated this interaction in transport inhibition (9, 18–20). The N terminus of NUP98 contains a large FG-repeat region and a RAE1-binding domain, making important contributions to the permeability barrier and mRNA export (21–23). Miorin et al. proposed that ORF6 binding to the NUP98–RAE1 complex inhibits the docking of KPNB1 to the NPC (9). Conversely, Kato et al. suggested that aberrant nucleocytoplasmic transport is caused by the ORF6 dislocating NUP98 from the NPCs to the cytoplasm (24). NUP98 is a prominent target of other viral toxins (25–28), and therefore understanding ORF6-NUP98 interaction could shed light on their actions.

One limitation of published studies of the ORF6 mechanism is the lack of precise quantitative data. Specifically, it would be informative to measure the relationship between the intracellular expression level of ORF6 and the resulting reduction in the nuclear transport efficiency. Accurate concentration–effect characterization would enable objective comparison of ORF6 activity across viral species, natural genetic variation, and function-probing mutants. Here, we apply a recently developed optogenetic assay (29) to measure inhibition of nuclear import and export by ORF6 in single, living cells. In parallel, we measured the expression level of untagged ORF6 in the same cells using a calibrated method. Cell-to-cell variation in ORF6 expression following transfection allowed us to generate concentration–effect plots from a single population of cells. By applying these methods to natural and engineered ORF6 variants, we generated mechanistic insights into their inhibitory mechanisms.

Results

Pipeline for Live-Cell Concentration–Effect Characterization of Nuclear Transport Inhibition.

To measure nuclear transport kinetics, we applied a recently reported live-cell microscopy assay (29) that uses LOV2-based optogenetic probes that translocate from the cytoplasm to the nucleus or vice versa upon blue-light stimulation (30, 31). The import and export probes respectively contain a photoactivatable NLS (LINuS) and NES (LEXY) that, upon activation, outcompete the counteracting NES and NLS. The fluorescence from mCherry in the probes is quantified in each nucleus to measure the translocation kinetics. The assays are performed using an automated microscopy platform and can readily process hundreds of cells per conditions. ORF6 is a small (MW ~7 kDa), membrane-associated protein, so tagging it with a fluorescent protein (MW ~28 kDa) might perturb its activity. We therefore developed a method for measuring its concentration in single cells without direct tagging. U2OS cells stably expressing the optogenetic nuclear import or export probe in the red channel were transfected with a plasmid encoding GFP-2A-ORF6 (Fig. 1A). In this system, GFP and untagged ORF6 are co-expressed from the same vector using a bicistronic system based on the 2A self-cleaving peptide (32). The self-cleaving activity of the 2A sequence occurs during translation (33), resulting in GFP and ORF6 being produced at the same rate as separate polypeptides. This enabled estimation of ORF6 concentration based on the GFP fluorescence intensity, with a conversion factor that depends only on GFP intensity and relative protein stability (see below). GFP intensity and nuclear transport kinetics were repeatedly measured in hundreds of individual cells at 3- to 6-h intervals for up to 28 h, after which ORF6 became toxic to some cells. The toxicity of ORF6 has been reported previously and is presumed to result from the nuclear transport inhibition (34). GFP expression level per cell varied more than 100-fold due to the combined effects of variable transfection efficiency, stochastic gene expression, and time-dependent increases. This can be a problem for bulk measurements, but here, it was an advantage since it spread cells over the concentration axis of concentration–effect curves. GFP intensity was converted to absolute GFP concentration using a null-contrast method (SI Appendix, Fig. S1 and Supplementary Text). In this method, living cells were surrounded by increasing concentrations of GFP added to culture medium. The internal concentration was given by the external concentration at which the cell becomes dimmer than the medium. This method has the large advantage that it is not sensitive to precise imaging conditions. GFP and ORF6 are co-translated in our assay, but their steady-state concentrations may differ due to difference in stability. To correct for this effect, we measured GFP:ORF6 ratios in bulk populations using calibrated western blotting for wild-type ORF6s (see below).

Fig. 1.

SARS-CoV-2 ORF6 is more potent than SARS-CoV-1 ORF6 in inhibiting bidirectional nuclear transport. (A) Schematic description of calibrated concentration–effect characterization of nuclear transport alteration by exogenous protein expression. The U2OS cell line stably expressing H2A-Halo and optogenetic nuclear import (NES-mCherry-LINuS) or export probe (NLS-mCherry-NEXY) was transfected with a plasmid encoding GFP-2A-ORF6. Due to the 2A self-cleavage sequence, GFP and ORF6 are produced at the same rate as separate polypeptides. The rate of the light-induced nuclear/cytoplasmic translocation of the transport probe and GFP intensity were simultaneously measured in individual cells. The measurement was repeated at 3- to 6-h interval for ~28 h after the transfection. The GFP intensity was converted to GFP concentration and then to ORF6 concentration via calibrations (SI Appendix, Supplementary Text and Fig. S1). The Hill function was fitted to the relationship between the nuclear transport rate and GFP concentration (“concentration–effect curve”) to obtain quantitative parameters. (Scale bar: 50 µm.) (B) Amino acid sequences of SARS-CoV-1 and SARS-CoV-2 ORF6s (ORF6^CoV1 and ORF6^CoV2, respectively). Dots indicate the mismatches. (C) Concentration–effect curves for the nuclear import (Top) and export (Bottom) inhibition by ORF6^CoV1 (Left) and ORF6^CoV2 (Right). n > 2,900 cells for each combination. Green circles are the raw data points. The black error bar represents the median and interquartile range in each bin. The black line and shaded area represent the best Hill function fit and corresponding 95% CI, respectively. The parameter estimates and SEs are shown. The additional x-axes showing the ORF6 concentration were constructed based on the calibration results shown in SI Appendix, Fig. S5. (D) Images and nuclear import assay data of representative cells transfected with GFP-2A-ORF6^CoV2. (Scale bar: 10 µm.)

Plotting the nuclear transport rate (import or export) against GFP concentration revealed the concentration–effect curve, where each point is data from a single cell. We fit these curves to a Hill function to provide parameters that characterize the degree of inhibition (A), half maximal inhibitory GFP concentration (IC₅₀^GFP), and the steepness of the concentration–effect curve (n_H or Hill coefficient). As in pharmacokinetic–pharmacodynamic modeling (35), this fit should be considered descriptive rather than mechanistic. We tested the pipeline with negative controls (SI Appendix, Fig. S2). When only GFP was expressed, the nuclear import rate was constant across GFP concentration, indicating the absence of artifacts due to GFP overexpression. ORF10 and candidate ORF15 are gene products of SARS-CoV-2 that were reported to not affect the IFN signaling and therefore are not likely to perturb nuclear import (8, 36). As expected, these proteins showed flat concentration–effect curves.

We used enhanced GFP (EGFP) for the concentration measurements throughout this study but also tested two other GFP variants with different maturation times and spectral characteristics, mCerulean and mTurquoise2 (37), and obtained very similar concentration–effect curves for SARS-CoV-2 ORF6 (SI Appendix, Fig. S3). Thus, the concentration measurement is not sensitive to the choice of the fluorescent protein.

Comparison of ORF6 Activity between SARS Coronaviruses.

We next compared the inhibitory effects of SARS-CoV-1 and SARS-CoV-2 ORF6s (denoted by ORF6^CoV1 and ORF6^CoV2, respectively) on nuclear transport. Their amino acid sequences are 69% identical; most of the variations lie in the C-terminal half, including the two amino acid C-terminal extension of ORF6^CoV1 (Fig. 1B). We found that ORF6^CoV2 inhibited ~fivefold more potently than ORF6^CoV1 as measured by IC₅₀^GFP (1.35 ± 0.07 µM vs. 6.31 ± 2.18 µM, P = 0.02) (Fig. 1 C and D). The Hill coefficient for ORF6^CoV2 was also higher (4.36 ± 0.77 vs. 1.50 ± 0.45, P = 0.001) while the amplitude of inhibition was similar (0.43 ± 0.09 min⁻¹ vs. 0.47 ± 0.01 min⁻¹, P = 0.66). Both ORF6s inhibited nuclear export with indistinguishable IC₅₀ and Hill coefficient parameters compared to import (P > 0.1) (Fig. 1C). Quantitatively equal blockade of import and export indicates that ORF6 inhibits some part of the NPC that is used to the same extent by both import and export carriers and thus favors models proposing direct inhibition of the NUP98–RAE1 complex (9, 19) over models proposing specific inhibition of KPNA2 (8, 16). The nuclear import and export rates at each concentration of ORF6^CoV1 and ORF6^CoV2 exhibited a colinear relationship, further supporting this point (SI Appendix, Fig. S4).

To determine true IC₅₀ values for ORF6 inhibition of nuclear transport, we sought to convert the IC₅₀^GFP value to the corresponding ORF6 concentration by correcting for differences in protein stability. Quantitative mass spectrometry was not successful due to the small size of the protein. We therefore used western blot with protein standards to measure the steady-state molar ratio of co-expressed GFP and ORF6. Accurate measurement of this ratio required considerable technical optimization (SI Appendix, Supplementary Text). Summarizing, ORF6 was much less stable than GFP, with a molar ratio of GFP to ORF6 of ~70 for ORF6^CoV1 and ~200 for ORF6^CoV2 (SI Appendix, Fig. S5). Consequently, we estimated the half maximal inhibitory ORF6 concentrations (IC₅₀^ORF6) to be ~90 nM for ORF6^CoV1 and ~6 nM for ORF6^CoV2 (Fig. 1C), indicating that the inhibitory potency of ORF6^CoV2 is ~15 times higher, consistent with previous, less quantitative comparisons of their IFN antagonism (15, 36). The high instability of ORF6 relative to GFP points to degradation as a possible cellular defense mechanism against viral inhibitors of nuclear transport.

The Absence of C-Terminal Tyr-Pro Extension Largely Explains the Higher Potency of SARS-CoV-2 ORF6.

ORF6 is characterized by abundance of hydrophobic residues in the N-terminal region and negatively charged residues in the C-terminal region. To identify regions in ORF6 that are accountable for the difference in the concentration–effect characteristics between ORF6^CoV1 and ORF6^CoV2, we performed site-directed mutagenesis to substitute each cluster of mismatches in ORF6^CoV2 with corresponding ORF6^CoV1 residues (Fig. 2A). None of the mutations affected the amplitude of inhibition (A), while some of them significantly altered the IC₅₀^GFP and Hill coefficient (n_H) of ORF6 (Fig. 2B and SI Appendix, Fig. S6). Notably, three mutations, E46K/N47K/K48N (#8), I60L (#12), and C-terminal addition of YP (#13), changed both the IC₅₀^GFP and n_H of the ORF6^CoV2 toward those of ORF6^CoV1 (Fig. 2B and SI Appendix, Fig. S6). Among them, the mutation #13 showed the strongest effect, increasing IC₅₀^GFP by a factor of 2.5 and decreasing Hill coefficient by a factor of 1.7 (Fig. 2C). We verified the effect of this mutation by removing YP at the C terminus of ORF6^CoV1 (#13r). The mutation #13r drastically reduced the IC₅₀^GFP of ORF6^CoV1 by a factor of 7, making it even lower than ORF6^CoV2 (Fig. 2 B and C). This suggests that the lack of C-terminal YP mainly contributes to the higher potency of ORF6^CoV2. Consistent with our results, a previous study observed that mutations #8 and #13 decreased the inhibitory effect of ORF6^CoV2 on the IFN signaling activity (15).

Fig. 2.

The lack of C-terminal Tyr-Pro mainly contributes to the higher potency of SARS-CoV-2 ORF6. (A) Notation of ORF6^CoV1 and ORF6^CoV2 variants. (B) IC₅₀^GFP (Top) and n_H (Bottom) of each ORF6 variant. The error bar represents the SE of the parameter estimate. n > 1,000 cells for each variant. The P-values for the significant difference from the wild type (two-tailed, two-sample z-test) are reported. *P < 0.1, **P < 0.01, ***P < 0.001, ****P < 0.0001. See also SI Appendix, Fig. S6. (C) Concentration–effect curves for the nuclear import inhibition by variants #13r (deletion of YP from the C-terminal end of ORF6^CoV1, Left) and #13 (addition of YP to the C terminus of ORF6^CoV2, Right). Green circles are the raw data points. The black error bar represents the median and interquartile range in each bin. The black line and shaded area represent the best Hill function fit and the corresponding 95% CI, respectively. The parameter estimates and SEs are shown in the plots. Dotted lines are the best Hill function fits for the ORF6^CoV1 (Left, red) and ORF6^CoV2 (Right, blue) wild types shown for comparison. (D and E) The N-terminal and C-terminal regions are simultaneously required for ORF6 activity. Concentration–effect curves for the nuclear import inhibition by (D) ORF6^CoV2 WT (black circles), residues 1 to 37 (yellow upward-pointing triangles), residues 1 to 55 (green squares), and M58R mutant (red downward-pointing triangles); and (E) WT (black circles), residues 38 to 61 (yellow upward-pointing triangles), residues 19 to 61 (green squares), and D22-30 mutant (red downward-pointing triangles). Wild-type-like potency: ++, compromised potency: +, no potency: −. n > 1,100 cells for each variant. The error bars show the median and interquartile range in each bin. The shaded area represents 95% CI of the best Hill function fit. Fits are not shown for the concentration–effect curves that failed to reject the null hypothesis that the inhibition amplitude (A) is zero (α = 0.1).

The N-Terminal and C-Terminal Regions Are Simultaneously Required for the Inhibitory Function of SARS-CoV-2 ORF6.

To gain more insights into the inhibitory mechanism, we characterized additional artificial and natural variants of ORF6^CoV2. The hydrophobic N-terminal region of ORF6 interacts with multiple membranous compartments, including the nuclear envelope, ER, and Golgi apparatus (38, 39). The negatively charged C-terminal region is responsible for interactions with NUP98–RAE1 and KPNAs (9, 15, 16, 19, 20). We found that C-terminal truncation of 24 residues resulted in the complete loss of the inhibitory effect of ORF6^CoV2 on the nuclear import, and so did C-terminal six-residue truncation, which had been discovered in COVID-19 patients in Italy (40) (Fig. 2D). Together with the comparative mutational analysis of ORF6^CoV1 and ORF6^CoV2 (Fig. 2B), these results suggest that the C-terminal end is crucial to the inhibitory function of ORF6. We also tested a point mutation, M58R, which has been shown to abolish the ability of ORF6^CoV2 to bind NUP98–RAE1 while maintaining its KPNA binding (9). This point mutation also resulted in the complete loss of the inhibitory function of ORF6^CoV2, suggesting that NUP98–RAE1 binding of the C-terminal region is essential while KPNA binding is not (Fig. 2D).

We then evaluated the influence of mutations in the N-terminal and middle regions on the inhibitory function of ORF6^CoV2 (Fig. 2E). We found that the peptide of residues 38 to 61 did not show a significant inhibitory effect on the nuclear import (Fig. 2E), indicating that NUP98–RAE1 binding alone is not sufficient for the transport inhibition. Alternative translation of ORF6^CoV2 gene results in the lack of the first 18 residues (41), which we found to increase the IC₅₀^GFP of ORF6^CoV2 by a factor of 3.4 (4.57 ± 0.76 µM) (Fig. 2E). A previous study discovered ORF6 having a nine-residue deletion from the central part (residues 22 to 30) in a SARS-CoV-2 strain passaged in vitro in the IFN-deficient Vero E6 cell (42). Although the authors predicted that the deletion would dramatically alter the structure of ORF6 binding to the membrane, the deletion did not affect the concentration–effect characteristics of the nuclear import inhibition (Fig. 2E). This is consistent with K23R/V24I/S25A and Y31V mutations (denoted by #2 and #3) not affecting the concentration–effect characteristics (Fig. 2B and SI Appendix, Fig. S6). Taken together, these mutational analyses suggest that the N-terminal and C-terminal regions are simultaneously required for the ORF6 activity in the nuclear transport inhibition, while the central region is not important.

Localization Requirement for the Inhibitory Function.

Membrane binding by ORF6 was implicated as mechanistically essential in some proposals (8, 9) but not others (17). We therefore localized ORF6 using small epitope tags and immunofluorescence. To test whether tags altered function, we characterized the concentration–effect curves of ORF6^CoV2 N- or C-terminally tagged with three different epitope tags: ALFA-tag (43), Flag-tag, and HA-tag. All three epitope tags weakened inhibition (i.e., increased the IC₅₀^GFP) of ORF6^CoV2 when positioned at the C terminus (SI Appendix, Fig. S7A). N-terminal tagging with ALFA-tag and HA-tag had negligible influence on the concentration–effect characteristics, and N-terminal Flag-tag even reduced the IC₅₀^GFP (Fig. 3A and SI Appendix, Fig. S7A). We chose N-terminal ALFA-tag for subsequent immunofluorescence using a well-characterized, commercially available fluorophore-conjugated anti-ALFA nanobody (43).

Fig. 3.

Membrane binding is not required for the inhibitory function of ORF6. (A) Concentration–effect curves for the nuclear import inhibition by ALFA-ORF6^CoV2. n > 1,000 cells. The dashed line is the Hill function fit for the untagged ORF6^CoV2 for comparison. The green circles are raw data points, and the black error bars are medians and interquartile ranges binned by GFP concentration. The black line represents the best-fitted Hill function, and the black shaded area is the corresponding 95% CI. Red arrows indicate the GFP concentrations of the cells shown in the immunofluorescence images in (B). The same data are also shown in SI Appendix, Fig. S5A for a different purpose. (B) Representative NUP98 and ALFA immunofluorescence images of ALFA-ORF6^CoV2 at low and high concentrations when focused on the basal plane (“bottom”) and midplane (“middle”) of the nucleus. (Scale bars: 10 µm.) Insets: 10× magnification. See also SI Appendix, Fig. S7.

To localize ORF6, U2OS cells were transfected with GFP-2A-ALFA-ORF6^CoV2 for 24 h. Prior to paraformaldehyde fixation, GFP concentration was measured to estimate the co-expressed ORF6 level as in the nuclear transport concentration–effect analysis (SI Appendix, Fig. S1). The fixed cells were permeabilized with the plasma-membrane specific agent digitonin to preserve the intracellular membrane structure and then immunostained using the anti-ALFA nanobody and anti-NUP98 antibody. ALFA-ORF6^CoV2 showed strong NPC staining with negligible staining elsewhere, and this pattern was consistent across different expression levels (Fig. 3B). At high ORF6 concentrations, NUP98 was dislocated from NPCs as scored by co-staining, consistent with a previous study (24). However, NUP98 dislocation was not noticeable at lower ORF6 concentrations that still strongly inhibited nuclear transport. These data suggest that ORF6 inhibits by binding to intact NPCs, and not by displacing NUP98.

We were surprised that the N-terminally tagged ORF6^CoV2 did not exhibit any membrane localization, inconsistent with previous reports (8, 9, 17), while showing the same concentration effect as the untagged protein. We suspected that the tagging terminus could affect the immunostaining pattern, so we examined the localization of the C-terminally tagged ORF6^CoV2. Indeed, ORF6^CoV2-ALFA was strongly stained at the cytoplasmic membranes and the nuclear envelope, confirming the membrane binding of the untagged N-terminal region of ORF6^CoV2 (SI Appendix, Fig. S7B). The lack of ALFA-ORF6^CoV2 immunostaining at the membranes could result from the N-terminal ALFA-tag hindering the membrane binding or from the membrane binding restricting the N-terminal ALFA-tag inaccessible to the anti-ALFA nanobody. To test this possibility, we examined the immunofluorescence localization of ORF6 having ALFA tags at both the N- and C-termini (ALFA-ORF6^CoV2-ALFA) (SI Appendix, Fig. S7C). Similar to ALFA-ORF6^CoV2, this construct showed strong ALFA immunostaining at NPCs with negligible staining elsewhere. Thus, we concluded that the N-terminal ALFA tag inhibits the membrane binding of ORF6^CoV2 without affecting its inhibitory function.

Forced Homo-Oligomerization Rescues the Inhibitory Function of N-Terminally Truncated ORF6.

We next sought to discriminate alternative models for the role of the N-terminal region in transport inhibition. Several other proteins that bind FG-NUPs and inhibit carrier-mediated nuclear transport, including wheat germ agglutinin (44) (WGA), hyperphosphorylated tau (45), C9orf72 dipeptide repeats (46), and vesicular stomatitis virus matrix (VSV M) protein (25, 47) are oligomeric and presumably able to multivalently cross-link FG domains (48–50). A recent computational analysis predicted that ORF6 is also prone to aggregate via the hydrophobic N-terminal region (51). Therefore, we hypothesized that the mechanistic role of the N-terminal region of ORF6 is to mediate homo-oligomerization or aggregation.

We first evaluated the oligomeric state of ORF6 in cells by performing fluorescence correlation spectroscopy (FCS), which quantifies the diffusion constant and concentration of fluorescent molecules (52). As expected, the autocorrelation curve for sfGFP in the cytoplasm was fitted well with a diffusion model of a single species whose diffusion constant was estimated to be 10.1 ± 0.7 µm²/s (mean ± SEM) (Fig. 4 A and B). On the other hand, the autocorrelation curve of sfGFP-ORF6 in the cytoplasm displayed two species with diffusion constants of 5.2 ± 0.7 µm²/s and 0.17 ± 0.05 µm²/s (mean ± SEM) (Fig. 4 A and B and SI Appendix, Fig. S8). The slower diffusion of sfGFP-ORF6 compared to sfGFP (P < 0.001) may result from sfGFP-ORF6 being in large protein complexes, its interactions with intracellular membranous structures, or combination of both. The existence of two diffusive species suggests that the sfGFP-ORF6-containing complex may be heterogenous in size and interaction with intracellular membranes. Assuming free diffusion, Stokes–Einstein’s relation, and a perfectly globular shape of molecules, we estimate the molecular weight of the fast-diffusing sfGFP-ORF6 species to be (10.1/5.2)³ = 7.3 times of that of sfGFP, or 6.2 times of that of the monomeric form of sfGFP-ORF6 (53). The molecular brightness of the cytoplasmic sfGFP-ORF6, calculated based on the concentration measured by FCS, was on average 1.9 times higher than that of sfGFP in cells (P < 0.01) (Fig. 4B). The molecular brightness is expected to underestimate the number of sfGFP-ORF6 monomers in the complexes because a large fraction of sfGFP would be immature due to the fast degradation of ORF6 (SI Appendix, Fig. S5). Together, these results suggest the presence of large complexes containing multiple copies of ORF6 proteins in live cells.

Fig. 4.

Forced homo-oligomerization rescues the inhibitory function of the N-terminally truncated ORF6. (A) Representative fluorescence correlation spectroscopy (FCS) data for U2OS cells transfected with sfGFP (Left) and sfGFP-ORF6 (Right). The Inset shows the cell image, and the red x mark indicates the location where the autocorrelation was measured. (Scale bar: 10 µm.) Gray circles represent the average of 10 to 50 autocorrelation curves, and the black lines are 3D diffusion models fitted to the data. Weighted residuals are shown below. One- and two-component models were fitted to the sfGFP and sfGFP-ORF6 data, respectively. The best-fit diffusion constants (D, D_fast, and D_slow) are shown. See also SI Appendix, Fig. S8. (B) Diffusion constants and molecular brightness of sfGFP (n = 6 cells) and sfGFP-ORF6 (n = 5 cells) measured by FCS. **P < 0.01, ***P < 0.001, ****P < 0.0001 (Welch’s t test, two-tailed). (C) FKBP homo-dimerizes in the presence of the B/B homodimerizer. Concentration–effect curves for the full-length ORF6^CoV2 (Top Left), ALFA-CT (Top Right), ALFA-FKBP-CT (Bottom Left), and ALFA-(FKBP)₂-CT (Bottom Right), where the CT refers to the C-terminal region (residue 38 to 61) of ORF6^CoV2. Green and red colors correspond to the absence and presence of 500 nM B/B homodimerizer, respectively. n > 1,500 cells for each condition. The error bar shows the median and interquartile range in each bin, and the solid line and shaded area represent the best Hill function fit and the corresponding 95% CI, respectively. (D and E) NUP98 and ALFA immunofluorescence images of cells transfected with (D) ALFA-CT and (E) ALFA-(FKBP)₂-CT in the absence (Top) and presence (Bottom) of 500 nM B/B homodimerizer. Samples were focused on the midplane of the nucleus. (Scale bar: 10 µm.) Insets: 10× magnification. (F) Proposed model. ORF6 multivalently cross-links NUP98–RAE1 complexes at the NPC via oligomerization of the N-terminal region (NT) and binding of the C-terminal region (CT). The cross-linking of NUP98–RAE1 reduces the mobility of the NUP98 FG domain as well as other FG-NUPs interacting with NUP98 via cohesive FG–FG interactions, thereby affecting the overall function of the permeability barrier.

We next investigated whether the N-terminal region can be functionally replaced with synthetic oligomerization domains derived from two different proteins that do not bind membranes. We first used a chemically inducible oligomerization system based on the F36V mutant of FK506-binding protein 12 (FKBP) and the B/B Homodimerizer ligand (equivalent to AP20187) (54) (Fig. 4C). Addition of the homodimerizer did not affect the concentration–effect characteristics of the full-length ORF6^CoV2 or the N-terminal truncation of ORF6^CoV2 (residues 38 to 61; CT), confirming the absence of unspecific effects of the homodimerizer on the concentration effect (Fig. 4C). When the CT was fused to FKBP, addition of the homodimerizer partially recovered the inhibitory function of the CT (Fig. 4C). The homodimerizer-dependent recovery further increased when the CT was fused to two FKBPs in tandem, which presumably induce higher-order oligomerization (Fig. 4C). We also tested the tetramerization domain of p53 (residues 326 to 356; p53TD) (55) and found that fusing p53TD also rescues the inhibitory function of the CT (SI Appendix, Fig. S9). We confirmed that the CT and the fusion proteins were localized at NPCs with no membrane localization (Fig. 4 D and E and SI Appendix, Fig. S9). The recovery of the inhibitory function via forced homo-oligomerization suggests that the mechanistic role of the N-terminal region is to drive oligomerization of ORF6, leading to high avidity binding to NUP98–RAE1, and that membrane binding is dispensable (Fig. 4F).

Discussion

We developed a broadly applicable pipeline for concentration–effect characterization of nuclear transport inhibitions. This combines the optogenetics-based measurement of nuclear transport kinetics (29) with measurement of protein expression level in individual cells. The protein level measurement does not rely on direct tagging and therefore is suitable for small proteins like SARS coronavirus ORF6. Both concentration and effect are calibrated in physical units, µM and min⁻¹, making the data reproducible between experiments and directly comparable across different conditions and microscopes. Here, we fit the widely used Hill function to the concentration–effect data to generate quantitative metrics, noting that more mechanistic models might be used to extract mechanistically justified biophysical parameters in future studies.

Our data revealed key mechanistic features of the inhibitory function of ORF6. First, ORF6 inhibits both KPNB1-mediated nuclear import and CRM1/XPO1-mediated nuclear export, showing the same concentration–effect characteristics. Therefore, the nuclear transport inhibition likely arises from a carrier unspecific perturbation of the nuclear transport machinery, e.g., the NPC impairment, rather than from specific interactions with KPNAs. Second, NUP98 binding of the C-terminal region is a critical determinant of the inhibitory potency but not sufficient for the inhibitory function of ORF6. Compared to ORF6^CoV1, ORF6^CoV2 showed a higher potency in inhibiting the nuclear transport in our study (Fig. 1C), primarily due to differences in the C-terminal end (Fig. 2B). In other studies, ORF6^CoV2 showed a stronger NUP98–RAE1 binding than ORF6^CoV1 (19, 20). This correlation suggests that NUP98 binding of the C-terminal region may determine the potency of ORF6. Finally, ORF6 oligomerizes in cells, and the N-terminal region can be functionally replaced with oligomerization domains that do not bind membranes. Therefore, we argue that the major role of the N-terminal region in the nuclear transport inhibition is to drive oligomerization, while the membrane binding is dispensable. The oligomerization may increase the avidity of ORF6-NPC interaction, cross-link the NUP98–RAE1 complexes, and reduce the flexibility of the FG domain of NUP98 (Fig. 4F). Given that NUP98 interacts with many other FG-NUPs via cohesive FG–FG interactions (56) and plays an essential role in the passive and active nuclear transport (21), the cross-linking could have a strong influence on the overall permeability of the FG barrier to cargo-carrier complexes. Future studies may investigate how the ORF6 binding affects the phase separation of the NUP98 FG domain, which is implicated not in the NPC functionality (21) but also in the development of various leukemia subtypes (57).

From a viral perspective, blocking nuclear transport is a potentially risky strategy for RNA viruses to evade innate immunity. If this block is too rapid or too potent, viral replication might be compromised by depletion of host cytoplasmic factors needed for translation and replication. SARS-CoV-1 and SARS-CoV-2 appear to have evolved to different optima, where the blocking function of ORF6 is much stronger in SARS-CoV-2. This presumably increases the ability of SARS-CoV-2 to evade interferon signaling, possibly at some cost in replication efficiency that might be compensated by other adaptations. Such adaptations could involve changes in other viral factors perturbing the nuclear transport, e.g., Nsp1 (58). How these changes contribute to the overall pathogenicity of SARS-related viruses is an important topic for future research. From a cellular perspective, viral inhibitory proteins are one of many factors with the potential to block nuclear pore and cause cellular pathology. Others include protein aggregates implicated in neurodegenerative diseases (45, 46). One potential defense mechanism is recognition and proteolysis of abnormal pore-adherent proteins. In this light, it is interesting that ORF6 turns over much faster than GFP (SI Appendix, Fig. S5) and that proteasomes have been observed clustering around NPCs in recent micrographs (59). Future studies might reveal an NPC-specific surveillance mechanism.

Materials and Methods

U2OS cell lines stably expressing the optogenetic nuclear transport probes were generated in the previous study (29). The engineered U2OS cells were transiently transfected with variants of EGFP-T2A-ORF6 in pSecTag2 mammalian expression vector. Spinning disk confocal microscopy was used in the concentration–effect and immunofluorescence assays. Two-photon FCS was performed as described previously. Custom python and MATLAB codes were used for quantitative image analyses and statistical analyses.

Detailed information on cell culture, plasmids and recombinant proteins, concentration–effect characterization, western blot, immunofluorescence, FCS measurement, and statistical and computational analyses is provided in SI Appendix, Materials and Methods.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Key analysis codes have been deposited in Github (doi: 10.5281/zenodo.10431821), and key plasmids have been deposited in Addgene (ID 204974, 204975, and 204976) (60).