Nanoscopic Elucidation of Spontaneous Self-Assembly of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Open Reading Frame 6 (ORF6) Protein

Abstract

Open reading frame 6 (ORF6), the accessory protein of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) that suppresses host type-I interferon signaling, possesses amyloidogenic sequences. ORF6 amyloidogenic peptides self-assemble to produce cytotoxic amyloid fibrils. Currently, the molecular properties of the ORF6 remain elusive. Here, we investigate the structural dynamics of the full-length ORF6 protein in a near-physiological environment using high-speed atomic force microscopy. ORF6 oligomers were ellipsoidal and readily assembled into ORF6 protofilaments in either a circular or a linear pattern. The formation of ORF6 protofilaments was enhanced at higher temperatures or on a lipid substrate. ORF6 filaments were sensitive to aliphatic alcohols, urea, and SDS, indicating that the filaments were predominantly maintained by hydrophobic interactions. In summary, ORF6 self-assembly could be necessary to sequester host factors and causes collateral damage to cells via amyloid aggregates. Nanoscopic imaging unveiled the innate molecular behavior of ORF6 and provides insight into drug repurposing to treat amyloid-related coronavirus disease 2019 complications.

The prevalence of coronavirus disease 2019 (COVID-19) has escalated to unprecedented pandemic levels since 2019. Unlike severe acute respiratory syndrome coronavirus (SARS-CoV) and Middle East respiratory syndrome coronavirus (MERS-CoV), many patients with COVID-19 were asymptomatic. Xia and co-workers reported that open reading frame (ORF) 6 protein was the strongest IFN-I antagonist among SARS-CoV-2 proteins,¹ suggesting that ORF6 protein may be crucial for delaying the onset of COVID-19 symptoms.

ORF6 protein (61 amino acids in length) is a SARS-CoV-2 accessory protein. Similar to other viruses,² SARS-CoV-2 expresses an array of accessory proteins to reprogram the host environment to favor its replication and survival. Comprehensive protein–protein interaction analyses revealed that ORF6 protein could interact with host nucleoporins (Rael and Nup98),^3,4 mitochondrial proteins,^3,4 and cytoskeletal proteins.⁴ We previously reported that ORF6 protein caused cytoplasmic retention of Nup98 and Rae1 in both HEK293T and PC9 cell lines.⁵ In addition, we and others found that the ORF6 protein disrupted host mRNA export.^5,6 Nucleocytoplasmic trafficking is often targeted by viruses during infection.^7,8 ORF6 protein hijacks host nuclear transport to robustly suppress host IFN-I signaling.^1,9,10 Temporal expression analysis indicated that ORF6 protein, nucleocapsid protein, and envelope protein demonstrated efficient protein translation.¹¹ Furthermore, the amino acid sequence of ORF6 protein is highly conserved among SARS-CoV-2 variants.¹² Collectively, these results imply that the ORF6 protein is an important accessory protein for SARS-CoV-2.

To date, structural information on ORF6 protein has been limited to its existence as a complex comprising a C-terminal tail, Rae1, and Nup98.^12,13 According to our earlier study, ORF6 protein existed either in clusters or in large aggregates, localized in the perinuclear region rather than confined in the nuclear rim.⁵ Using counterstaining, Wong et al. demonstrated that ORF6 protein resides on the endoplasmic reticulum (ER), Golgi apparatus, and mitochondria.¹⁴ We hypothesize that ORF6 protein could aggregate and form a shield at the perinuclear region to trap host nucleoporins (Rae1 and Nup98)⁵ or sequester nuclear transport receptors (karyopherins α).^1,9 This hypothesis could explain the cytoplasmic retention of nucleoporins and impaired nuclear import in ORF6-expressing cells. Moreover, a similar mechanism has been reported for SARS-CoV ORF6 protein.¹⁵ Therefore, elucidation of the native structure and conformational dynamics of full-length ORF6 protein is imperative to understand its molecular behavior; the findings from which may be essential for drug development against ORF6 protein to restore host nuclear transport and maintain cell viability.

Atomic force microscopy (AFM) relies on the interaction between cantilever tip and biological samples for nanoimaging and quantitative analysis of biomolecular events or properties such as association and dissociation of biomolecules, binding affinity, and more (reviewed in ref (16)). High-speed AFM (HS-AFM) has a high spatiotemporal resolution and an optimal tip–sample interaction that allows nanoscopic elucidation of dynamic properties of biomolecules¹⁷ and organelles¹⁸⁻²⁰ without destroying sample integrity. HS-AFM solves the technical limitations of cryoelectron microscopy (cryo-EM) and X-ray crystallography for investigating the structural properties of proteins with intrinsically disordered regions (IDRs).²¹ Moreover, the temporal resolution of HS-AFM supersedes other imaging tools in capturing the structural dynamics of biomolecules,²²⁻²⁴ the dynamics of biomolecule interactions,²⁴⁻²⁷ and the dynamics of biomolecule–organelle interactions.²³⁻²⁵ Here, we use HS-AFM to elucidate the full-length ORF6 protein structure and its intrinsic molecular dynamics. ORF6 appeared as oligomers, and these oligomers readily formed short or long protofilaments, especially at higher temperatures or on fluidic lipid surfaces. The formation of ORF6 protofilaments was predominantly mediated by hydrophobic interactions, because the filaments were dissociated after treatment with aliphatic alcohols (1,6-hexanediol [1,6-HD] and trans-1,2-cyclohexanediol [CHD]), urea, or sodium dodecyl sulfate (SDS).

To determine a suitable substrate for ORF6 protein adsorption, we used the Prot Pi Web server to compute its net charge at pH 7.4 (Figure 1a). The results showed that ORF6 protein had a weak negative charge (ζ: −4.7) at pH 7.4, and bare mica was suitable for protein absorption with sufficient mobility under a physiological buffer (50 mM Tris HCl, 150 mM NaCl, pH 7.4). The hydropathicity of amino acids determines the hydrophobic interactions for protein quaternary folding in aqueous solutions. Using a predictor for hydrophobicity,²⁸ we confirmed that ORF6 protein is highly hydrophobic (93%) (Figure 1b), with the N-terminal being more hydrophobic than the C-terminal. To evaluate whether the ORF6 protein is ordered, flexible, or disordered, we used seven predictors (IUPred long, IUPred short, PONDR VL3-BA, PONDR VLXT, PONDR VSL2, PONDR Fit, and PrDOS) to predict the IDRs of the protein. The propensity scores of the amino acid residues computed by each predictor, together with the mean propensity scores of the seven predictors, are illustrated in Figure 1c. The C-terminus of the ORF6 protein was found to be disordered, with an IDR coverage of 22.95%. The flexibility of the C-terminal tail may favor the critical amino acid, methionine at residue 58 (M58),⁹ providing access to the mRNA-binding groove of the Nup98–Rae1 complex to halt mRNA export.^12,13 The full-length ORF6 protein structure has yet to be resolved by cryo-EM or X-ray crystallography. We did not use AlphaFold2 for structural prediction because the accuracy of this predictor is compromised by the IDRs of ORF6 protein.²⁹ Instead, we directly visualized the native structure of full-length ORF6 protein using HS-AFM. Recombinant ORF6, expressed as a GST-fusion protein, was subjected to HRV3C protease cleavage and the removal of GST, and was then further purified by size exclusion chromatography (SEC) (Figure S1a). SEC revealed that the ORF6 protein was in a higher molecular weight form because the elution volume was much lower than that of aprotinin, which has a molecular weight (6.5 kDa) close to the molecular weight of the ORF6 monomer (7.3 kDa) (Figure 1d). Prior to HS-AFM scanning, we performed NMR analysis on the purified ORF6 protein to obtain structural information at the atomic level (reviewed in ref (30)), which is difficult to achieve by HS-AFM. In the ¹H–¹⁵N band-selective optimized flip angle short transient heteronuclear multiple quantum coherence (SOFAST HMQC) spectrum of ¹⁵N-labeled ORF6, most peaks were observed at 7.8–8.4 ppm in the ¹H chemical shift axis (Figure 1e), indicating that the ORF6 oligomer contains a highly flexible and disordered segment. HS-AFM images showed that ORF6 oligomers are ellipsoidal (Figure 1f,g, Figure S1b, Movie S1). Real-time interaction between a polyclonal anti-ORF6 antibody and ORF6 oligomers was observed at the nanoscopic level (Figure S1c). The mean cross-sectional height of ORF6 oligomers was 3.26 nm (Figure 1h). Real-time height fluctuation depicted the intrinsic elasticity of the ORF6 oligomers during HS-AFM scanning (Figure 1i).

Figure 1.

Nanoscopic observation of full-length SARS-CoV-2 ORF6 protein under a near-physiological environment. (a–c) In silico analysis of the molecular properties of ORF6. (a) Prot-Pi Web server was used to determine the ORF6 protein net charge at pH 7.4 (ζ potential: −4.7). (b) Protscale Web server was used to predict the hydropathicity of ORF6 protein. The values are shown in the Eisenberg scale. (c) Seven intrinsically disordered region (IDR) predictors were used to predict the IDRs in ORF6 protein. The output of each predictor is illustrated on the left graph, and the average score for the seven outputs is shown on the right graph. (d) The elution profile of size exclusion chromatography (SEC) shows that ORF6 protein mainly exists as oligomers (HMW, high molecular weight; LMW, low molecular weight). Aprotinin, a protein with a molecular weight close to the molecular weight of ORF6 monomer, is used as a reference. (e) ¹H–¹⁵N SOFAST HMQC spectrum of ORF6 protein. The peaks originating from the side chain are labeled with asterisks. (f) HS-AFM images of ORF6 oligomers (scale bar, group, 20 nm; single, 10 nm). (g) 3D image of ORF6 oligomers. (h) A histogram, together with a Gaussian distribution curve, illustrates the cross-sectional height distribution of ORF6 oligomer (n: 1385). (i) Real-time height fluctuation of ORF6 oligomer (n: 10).

We previously observed that GFP-tagged ORF6 protein aggregates surrounded the perinuclear area rather than being confined within the nuclear rim.⁵ In addition, protein–protein interaction network analysis indicated that ORF6 protein could interact with host proteins that are involved in protein oligomerization: actinin alpha 2 (ACTN2) and inositol 1, 4, 5-trisphosphate receptor type 3 (ITPR3).⁴ Moreover, Tavassoly and co-workers reported that spike protein-derived peptide is associated with amyloid formation in the brain, which can progress to neurodegenerative disease.³¹ To determine whether ORF6 protein could oligomerize and subsequently form aggregates, we first conducted in silico analysis using two amyloid protein predictors: AGGRESCAN³² and FoldAmyloid.³³ The results showed that ORF6 protein is amyloidogenic (Figure S2a,b). Both the tendency for amyloidogenicity and the percentage of amyloidogenic amino acid residues in ORF6 protein are far greater than those in amyloid proteins found in neurodegenerative diseases (Table S1). In addition, PASTA2.0 analysis³⁴ demonstrated that ORF6 protein has the highest free energy (−21.16) compared with other amyloid proteins for amyloidogenesis (Table S1). These findings suggest that the ORF6 protein may have a strong tendency to aggregate spontaneously in cells. We used HS-AFM to demonstrate ORF6 protein self-assembly into filaments, either in a circular or in a linear conformation (Figure 2a, Figure S2c, Movie S2). At higher concentrations, short ORF6 filaments were packed together on a bare mica surface (Figure S2d). Polyclonal anti-ORF6 antibody not only bound to ORF6 oligomers (Figure S1c) but also bound to ORF6 filaments (Figure S2e).

Figure 2.

HS-AFM reveals spontaneous self-assembly of ORF6 oligomers into protofilaments. (a) ORF6 protein appeared as rings or linear filaments (scale bar, R1, 20 nm; R2, 16 nm; R3, 27 nm; R4, 14 nm; L1, 40 nm; L2, 20 nm; L3, 40 nm; L4, 44 nm). (b) ORF6 oligomers spontaneously assembled after contact (scale bar, b1, 7 nm; b2, 40 nm). (c–e) High temperature promoted longer ORF6 filaments and higher frequency of the circular conformation on bare mica (n, 4 °C: 804, 22 °C: 335, 37 °C: 508; l, length; C, circumference; scale bar: 100 nm). Statistical analysis in (d) was performed using the Mann–Whitney U test. Data are presented in mean ± SEM (n.s., non-significant; **p < 0.01; **** p < 0.0001). (f–h) ORF6 filaments had higher mobility on a cationic lipid bilayer compared with bare mica. (f) Manual tracking was performed to illustrate the movement of ORF6 filaments during HS-AFM scanning (scale bar, 40 nm). (g) Regardless of conformation (circular or linear), real-time distance fluctuations of ORF6 protein were greater on the cationic lipid bilayer. (h) A line graph presents the real-time velocity of ORF6 protein on different substrates (M: mica; L: lipid). (i) Cationic lipid substrate but not bare mica has a fluidic surface. This property allowed ORF6 filament to switch between circular and linear conformations during HS-AFM scanning (scale bar, 30 nm). (j) Proportion of filaments in the circular conformation on the lipid substrate was lower than on that bare mica (blue, linear; pink, circular).

Spontaneous ORF6 assembly of two oligomers (Figure 2b₁, Movie S3) and ORF6 filament elongation (Figure 2b₂, Movie S3) were demonstrated. These results indicated that the mobility of the ORF6 protein could drive protein assembly. We hypothesized that the ORF6 particle collision rate was increased at a higher temperature in which this condition boosted ORF6 oligomerization. We first incubated ORF6 protein at 4 °C, 22 °C, or 37 °C before we loaded it on bare mica for HS-AFM imaging at room temperature. Results showed that both length and circumference of ORF6 filament were the longest at 37 °C and the shortest at 4 °C (Figure 2c,d). Intriguingly, ORF6 filaments preferred to appear in a circular conformation at higher temperature (Figure 2c,e, Movie S4). Collectively, these results supported our hypothesis. Furthermore, the ORF6 protofilaments with higher mobility promptly underwent circular motion.

To further investigate the relationship between ORF6 mobility and ORF6 oligomerization, we modulated the ORF6 mobility by using two different types of substrates: bare mica and cationic lipid bilayer-coated mica. Lipid bilayer is fluidic; hence, ORF6 protein readily moves on the substrate. Highly fluidic biological membrane³⁵ and lipid bilayer substrate containing unsaturated alkyl chain, DOPC (1,2-dioleoyl-sn-glycero-3-phosphocholine) for example (reviewed in ref (36)), are not suitable for HS-AFM imaging of proteins at low density because the samples diffuse rapidly. Therefore, we selected DPPC (dipalmitoylphosphatidylcholine), a phospholipid containing saturated alkyl chain, to formulate the low fluidic lipid substrate³⁵ for our experiments. We found that ORF6 protein moved faster on cationic lipid substrate compared with that on bare mica (Figure 2f–h, Figure S3a,c). Moving trajectories of ORF6 protein were analyzed and are presented as blue lines (Figure 2f, Figure S3c, and Movie S5). During sample incubation, ORF6 oligomers/protofilaments readily diffused on lipid substrate and oligomerized with adjacent molecules to form longer protofilaments. In contrast, ORF6 oligomers and protofilaments were relatively immobile on bare mica, and this greatly limited ORF6 self-assembly. As a result, ORF6 protein mainly appeared as filaments on a lipid substrate but in a mixture of oligomers and filaments on bare mica (Figure S3b, Movie S6). ORF6 filaments rapidly transformed between linear and circular conformations on lipid substrate (Figure 2i,j, Figure S4a,b, Movie S7 and Movie S8). Moreover, complex dynamic patterns of ORF6 protein, such as circular motion, protein assembly, and flipping, were visualized in real-time (Figure S4c, Movie S8). Collectively, our results suggested that mobility of ORF6 molecules is crucial for ORF6 oligomerization in which it can be expedited by both temperature and lipid fluidity. During SARS-CoV-2 infection, fever (high body temperature) could favor ORF6 oligomerization, and ORF6 oligomers/protofilaments subsequently localize on lipid membrane of membranous organelles (ER, Golgi apparatus, mitochondria, nuclear envelope etc.). Lipid fluidity further accelerates ORF6 oligomerization that could establish “ORF6 networks/aggregates” to sequester host factors such as karyopherins, STAT1, nucleoporins (Nup98 and Rae1), and more.

Next, we investigated the interaction forces that mediate ORF6 protein self-assembly to form filaments. ORF6 filaments remained intact at low pH (pH 5) (Figure S5a) or in high ionic strength (2 M NaCl) environments (Figure S5b,e, Movie S9). These results suggested that electrostatic interactions may not be dominant for the assembly of ORF6 oligomers into protofilaments. To determine the role of hydrophobic interactions in the ORF6 protofilament formation, we treated the protein with urea and SDS to disrupt the hydrophobic interactions. HS-AFM imaging clearly showed that ORF6 filaments were dissociated, and the height of filament was significantly reduced after treatment with 2 M urea (Figure S5c,e, Movie S9). Similarly, rapid breakdown of the ORF6 filament was found after the addition of 0.1% SDS during scanning (Figure S5d,e). We also found two peaks in SEC analysis, suggesting that ORF6 filaments were dissociated to oligomers and monomers (Figure S5f). In addition to these strong denaturants, we also visualized the disassembly of ORF6 filaments by aliphatic alcohols (1,6-HD, and trans-1,2-CHD) during HS-AFM scanning (Figure 3b,c, Figure S5g, Movie S10). By contrast, Ivermectin, an importin α/β blocker, did not induce dissociation of ORF6 filaments (Figure 3a, Movie S10). At higher magnification, we were able to calculate the number of breaks produced (Figure 3d). A significant reduction in height was found in ORF6 protein treated with aliphatic alcohols (Figure 3e). Likewise, GFP-ORF6 aggregates in the GFP-ORF6-expressing HeLa cells were diminished after the cells had been treated with 2% 1,6-HD for 10 min (Figure S6). 1,6-HD is commonly used to disrupt liquid–liquid phase separation (LLPS) condensates by targeting weak hydrophobic interactions. Altogether, these findings revealed that ORF6 protein self-assembly could be mainly orchestrated by hydrophobic interactions.

Figure 3.

ORF6 protofilament formation is predominantly mediated by hydrophobic interactions. (a–c) Real-time visualization of ORF6 filaments in response to (a) 5 μM ivermectin, (b) 15% w/v 1,6-HD, or (c) 15% w/v CHD treatment (scale bar, a (−) and (IV), 200 nm; a (1–3), 30 nm; a (4), 40 nm; b (−) and (1,6-HD), 200 nm; b (1), 56 nm; b (2), 36 nm; b (3), 24 nm; b (4), 30 nm, c (−) and (CHD), 200 nm; c (1,2), 44 nm; c (3), 48 nm, c (4), 40 nm). ORF6 filaments rapidly dissolved when treated with 15% aliphatic alcohol (1,6-HD and CHD). (d) The number of breaks that appeared after 1,6-HD, or CHD treatment. (e) Aliphatic alcohols (1,6-HD and CHD) significantly reduced the height of ORF6 as a result of deoligomerization. The height of ORF6 protein before and after exposure to the designated chemicals was measured and compared. Data are presented as the mean ± SEM (n_Ivermectin, 6; n_1,6-HD, 4; n_CHD, 5). Statistical analyses performed by two-tailed Paired t test (Ivermectin and 1,6-HD) or Wilcoxon Signed-Rank test (CHD), ** p < 0.01, *p < 0.05.

During SARS-CoV-2 infection, an array of accessory proteins is expressed to reprogram the host environment to favor viral replication and survival. ORF6 protein plays an important role in suppression of the host IFN-I response, resulting in asymptomatic infection. Therefore, investigation of the structural properties of ORF6 protein in COVID-19 pathogenesis is essential for drug development. However, IDRs in the ORF6 protein impede conventional structural methodologies, such as cryo-EM and X-ray crystallography, to acquire the precise 3D structure of ORF6 protein. To overcome this technical limitation, we employed HS-AFM to directly visualize the full-length ORF6 protein. Interestingly, the SEC results indicated that the full-length ORF6 protein appeared in oligomeric form. Furthermore, NMR analysis revealed that the oligomeric ORF6 protein contains a highly flexible and disordered region, which was deduced from the poorly dispersed NMR spectrum. Considering the results of the IDR prediction (Figure 1c), this segment presumably corresponds to the C-terminal tail, which would be exposed to the solvent and available for interaction with Rae1/Nup98 in the oligomeric ORF6 protein. However, Yoo and colleagues reported that the hydrophobic N-terminal of ORF6 protein could cross-link with the phenylalanine-glycine domain (FG domain) of Nup98 in the nuclear pore complex.³⁷ These results suggested that ORF6 protein may be functional in oligomeric form to hijack host nuclear transport. Several studies, including ours, have demonstrated that the subcellular localization of ORF6 protein is not confined to the nuclear rim.^5,14,38 Instead, ORF6 protein appeared as aggregates that localized on membranous organelles, including ER, Golgi apparatus, mitochondria, lysosomes, and autophagosomes.³⁸

Several studies have reported Alzheimer’s disease (AD)-like neurological complications in COVID-19 patients (reviewed in ref (39)). Some SARS-CoV-2 viral proteins either disrupt the homeostasis of AD hallmark proteins (tau, Aβ42, and α-synuclein) in neurons (for example, spike protein and nucleocapsid protein)⁴⁰⁻⁴² or form aggregates in lung epithelial cells (for example, full-length ORF8 protein).⁴³ Besides these viral proteins, overwhelmed inflammation in severe COVID-19 could also promote amyloidosis. Highly amyloidogenic calprotectin protein (S100A8/S100A9)^44,45 was found elevated excessively in severe COVID-19 patients.⁴⁶ These findings prompted us to investigate whether the ORF6 protein is amyloidogenic. In silico predictions indicated that ORF6 protein has a high tendency for spontaneous self-assembly. These predictions were further supported by our observation of ORF6 protofilaments by using HS-AFM (summarized in Figure 4a). Cryo-EM imaging of ORF6 protofilaments should also be considered in future study even though it is difficult to resolve ORF6 structure at the single molecular level. The α-helix at the N-terminus of ORF6 protein facilitates its localization on the lipid membrane.³⁸ Our findings demonstrated that the fluidity of the lipid bilayer makes the lipid substrate an ideal platform for ORF6 protein diffusion to achieve robust elongation (summarized in Figure 4a). Our results also implied that the high body temperature in COVID-19 patients with fever could further accelerate ORF6 protofilament formation and elongation (summarized in Figure 4a). We previously found that Nup98 and Rae1 colocalized in ORF6 aggregates in GFP-ORF6-expressing cells.⁵ High body temperature and the localization of ORF6 oligomers on various membranous organelles maximize ORF6 intracellular aggregation (summarized in Figure 4b) to effectively sequester a vast numbers of host proteins, particularly transcription factors involved in IFN-I signaling (STAT1, IRF3),^1,9,47 from nucleus translocation.

Figure 4.

Nanoscopic elucidation of full-length ORF6 protein reveals its spontaneous self-assembly that may be related to COVID-19 complications and may have therapeutic implications. (a, b) SARS-CoV-2 ORF6 protein undergoes self-assembly and forms aggregates in infected cells. (c) Exogenous ORF6 protein released from infected cells (living or dead) could be internalized and then aggregate in adjacent cells. (d) Accumulation of ORF6 aggregates induces IL-6 expression in lung cells. The epithelial-derived IL-6 could be associated with lung pathology in COVID-19 patients. (e) Drugs that disrupt hydrophobic interactions should be investigated for their therapeutic value against ORF6 aggregates.

Certain viral proteins, such as influenza A PB1-F2 protein⁴⁸⁻⁵⁰ and nuclear export protein,⁵¹ undergo self-assembly to produce cytotoxic amyloid fibrils during infection. Charnley and co-workers⁵² demonstrated that the amyloidogenic peptide of ORF6 protein, ILLIIM peptide, formed amyloid fibrils that were cytotoxic to a neuroblastoma cell line, SH-SY5Y. Bhardwaj and colleagues⁵³ reported that many SARS-CoV-2 proteins have amyloidogenic sequences, enabling these peptides to form cytotoxic amyloid fibrils. Both teams emphasized the possibility that these amyloidogenic peptides could be etiological factors in amyloid-related complications in the treatment of COVID-19 patients. Further studies are needed to confirm whether host cells can digest SARS-CoV-2 viral proteins, such as spike protein, ORF6 protein, and NSP6 protein, to produce amyloidogenic peptides. In this study, direct nanoscopic visualization of full-length ORF6 showed the innate molecular behavior of ORF6 protein in SARS-CoV-2-infected cells. We managed to capture the ORF6 protofilaments at the ng/μL concentration level, suggesting that full-length ORF6 protein may contribute to amyloid-related complications in COVID-19 patients.

During AD progression, misfolded tau and α-synuclein in AD neurons are secreted into the extracellular space, and subsequently these proteopathic seeds are internalized by adjacent neurons via micropinocytosis.⁵⁴ The internalized misfolded protein is then either degraded by lysosomes or escapes into the cytosol and finally aggregates.⁵⁵ On the basis of this propagation model, we postulated that ORF6 protein may behave similarly, with the source of ORF6 protein from infected cells (summarized in Figure 4c). To investigate our hypothesis, we treated lung cancer cell lines, PC9 and A549, with either recombinant GFP protein or recombinant GFP-ORF6 protein (Figure S7a) for 24 h. The results showed that GFP-ORF6 protein was internalized and accumulated in cells (Figure S7b). Furthermore, confocal imaging revealed that the localization of GFP-ORF6 aggregates (Figure S7c) resembled that in GFP-ORF6-expressing cells (Figure S6). According to these results, together with our observation of ORF6 self-assembly on a lipid substrate using HS-AFM, we deduced that the ORF6 oligomers that escaped from endosomes could directly undergo self-assembly followed by aggregation in the cytoplasm, preferentially on the lipid surface of membranous organelles. The internalized ORF6 oligomers may be sufficient to achieve protein aggregation in adjacent healthy cells or to increase the ORF6 aggregation burden in adjacent infected cells.

Chan and colleagues found that accumulation of SARS-CoV spike protein increased ER stress, and high ER stress later induced the unfolded protein response.⁵⁶ Activation of the unfolded protein response promotes interleukin (IL)-6 expression and triggers inflammation.^57,58 Our Western blotting results indicated that accumulation of ORF6 protein, either by transfection (Figure S8a) or by internalization of exogenous GFP-ORF6 protein (Figure S8c), promoted IL-6 expression in lung cancer cell lines (Figure S8b,d). Given that only one band was visualized with recombinant GFP-ORF6 protein (Figure S7a), internalized GFP-ORF6 could escape either in the GFP-tagged form or GFP-untagged form because we detected two GFP bands and multiple ORF6 bands (Figure S8c). In this experiment, the cell pellets were washed multiple times to ensure adequate removal of the remaining recombinant GFP protein or recombinant GFP-ORF6 protein in the culture media. Melms and co-workers reported that epithelial cell-derived IL-6 is a unique feature of COVID-19, with a concentration much higher than that of macrophage-derived IL-6 in late COVID-19.⁵⁹ Additional studies are needed to determine whether the progressive accumulation of SARS-CoV-2 proteins, ORF6 in this case, is one of the etiological factors responsible for this unique feature. IL-6 induces immune dysregulation and lung fibrosis in the lung, resulting in decreased pulmonary function and long COVID-19 complications (summarized in Figure 4d). Potential druggable candidates that dissociate ORF6 aggregates by disrupting hydrophobic interactions (summarized in Figure 4e) should be considered and tested in the near future to evaluate their therapeutic value in COVID-19 management and treatment.

In conclusion, we can directly visualize full-length ORF6 oligomers under near-physiological conditions using HS-AFM to elucidate their innate molecular properties related to the COVID-19 pathology. HS-AFM also provides a nanoscopic assessment platform to investigate the potential chemicals that disrupt ORF6 aggregates. The results can then be further translated into various cell line models.

Methods

Materials. 1,6-Hexanediol and trans-1,2-cyclohexanediol were purchased from Tokyo Chemical Industry Co., Ltd. (Tokyo, Japan).

Plasmids. The pEGFP-ORF6 plasmid was previously described.⁵ The pGEX-ORF6 plasmid and pGEX-GFP-ORF6 for purification of ORF6 and GFP-ORF6 proteins were generated by the In-Fusion cloning method (TaKaRa, Shiga, Japan) of synthetic DNA into the pGEX-6P-1 vector (Cytiva, MA, US) using SmaI restriction site.

Cell Culture and Transfection. HeLa, PC-9, and A549 cell lines were bought from the Japanese Collection of Research Bioresources Cell Bank (Osaka, Japan). Cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM, Thermo Fisher Scientific, MA, US) supplemented with 10% (v/v) fetal bovine serum (Life Technologies, CA, USA) and 1% (v/v) penicillin-streptomycin (Nacalai Tesque, Kyoto, Japan). Cells were grown at 37 °C in a humidified CO₂ incubator. DNA transfection was conducted in six-well plates using Lipofectamine 2000 (Invitrogen, MA, US) with 1000 ng of pEGFP-N1 (Clontech, CA, USA) or pEGFP-ORF6 plasmid DNA according to the manufacturer’s manual, as previously reported.⁵ Transfected cells were used for downstream experiments, such as confocal imaging and Western blotting.

Recombinant ORF6 Protein Synthesis and Purification. pGEX-ORF6 and pGEX-GFP-ORF6 plasmids were transformed into E. coli BL21-CodonPlus competent cells (Agilent technologies, CA, US). The GST fusion ORF6 or GFP-ORF6 protein expression was induced with 1 mM isopropyl-beta-d-thiogalactopyranoside (IPTG; Nacalai Tesque, Kyoto, Japan) and purified on glutathione-conjugated agarose beads (Glutathione Sepharose 4B; Cytiva, MA, US). GST-ORF6 fusion protein attached to glutathione-conjugated agarose beads was mixed with HRV3C protease (Sigma-Aldrich, MO, US) and allowed to digest at 4 °C for 18 h. Digested and eluted free ORF6 and GFP-ORF6 proteins were then applied to nanosep centrifugal devices with omega membrane 10K (PALL Corporation, CA, US) for sample concentration and buffer exchanging. SDS-PAGE and Coomassie Blue staining were then performed to check the Orf6 and GFP-ORF6 protein purities. The purified ORF6 and GFP-ORF6 were further analyzed by Superdex 75 increase 10/300 size exclusion column (Cytiva, MA, USA) connected to an ÄKTA go system (Cytiva, MA, USA).

Confocal Imaging. PC9 and A549 cells transfected with the GFP vector or GFP-ORF6 were cultured on glass coverslips in six-well plates. Treatments with exogenous GFP protein or GFP-ORF6 protein to PC9 and A549 cells were performed on eight-well chamber slides. Cells on coverslips or slides were washed with PBS and fixed with 4% PFA (paraformaldehyde; Nacalai Tesque, Kyoto, Japan)/PBS. After being washed, cells were permeabilized with 0.3% Triton X-100 (Nacalai Tesque, Kyoto, Japan)/PBS for 3 min and washed with PBS again. After that, cells were blocked with 4% bovine serum albumin (BSA; FUJIFILM Wako Pure Chemical Corp., Osaka, Japan). Cells on coverslips or slides were washed with PBS and then mounted on slides or with a coverslip using ProLong Diamond Antifade reagent with DAPI (Invitrogen, MA, US).

In silico analysis of ORF6 protein properties. ORF6 protein net charge at pH 7.4 was measured using the ProtPi Web server (https://www.protpi.ch/). Hydropathicity of ORF6 protein was determined using the Eisenberg scale available in the Expasy ProtScale Web server (https://web.expasy.org/protscale/). Intrinsically disordered regions (IDRs) of ORF6 protein was predicted using seven IDR predictors including IUPRED series (long and short), PONDR (VLS2, VL3, FIT, and VLXT) series, and PrDOS as previously described.²⁴ Outputs of seven IDR predictors were averaged and presented in different graph. Amyloidogenicity of ORF6 was predicted using two predictors known as AGGRESCAN (http://bioinf.uab.es/aggrescan/) and FoldAmyloid (http://bioinfo.protres.ru/fold-amyloid/). Additionally, the spontaneity of ORF6 amyloidogenesis was evaluated according to free energy status using PASTA 2.0 (http://old.protein.bio.unipd.it/pasta2/).

Western Blotting. After transfection (GFP vector or GFP-ORF6) or treatment (GFP protein or GFP-ORF6 protein), PC9 and A549 were treated with Brefeldin A (5 μg/mL) at final 8 h to enhance IL-6 detection. Briefly, cells were harvested and lysed with NP40 lysis buffer. After that, the lysates were added with loading buffer and heated at 90 °C for 10 min to denature proteins. Samples were then subjected to SDS-PAGE, and proteins were transferred to a PVDF membrane. The membrane was incubated with primary antibodies, including anti-ORF6 (1:1000; Abnova, Taipei, Taiwan), anti-GFP (1:2000; Wako, Osaka, Japan), anti-β-actin (1:2000, Cell Signaling Technology, MA, USA), or anti-IL-6 (1:2000; Cell Signaling Technology, MA, USA) at 4 °C overnight. PVDF membrane was washed and then incubated with secondary HRP-conjugated antibodies (1:5000; Cell Signaling Technology, MA, USA). HRP substrate, Immobilon Western Chemiluminescent HRP Substrate (Millipore, MA, USA), was used for detection. Images were captured with a LAS-4000 image analyzer (Fujifilm, Tokyo, Japan).

HS-AFM Imaging. HS-AFM images were recorded using a laboratory-built tapping mode HS-AFM, as previously described.²²⁻²⁶ A small cantilever (BL-AC10-DS-A2; Olympus, Tokyo, Japan) with a spring constant (k) 0.1 N/m and resonance frequency (f) 0.6 MHz (water)/1.5 MHz (air) was used. Amorphous carbon was deposited on the cantilever tip with an apical radius ∼8 nm using electron-beam deposition with a field emission scanning electron microscopy (ELS-7500, Elionix Inc., Tokyo, Japan) to improve image quality.¹⁹ A laser beam (670 nm wavelength) was focused on an EBD-processed cantilever tip using a 20×-objective lens (CFI S Plan Fluor ELWD, Nikon, Tokyo, Japan). Dynamic deflection of the cantilever was detected by a position-sensing two-segmented PIN photodiode. To achieve optimal tip–sample interaction force, the free oscillation amplitude of the cantilever (A₀) was adjusted to 1.5–2.5 nm, and the set-point was tuned to 80–90% of the free amplitude. A glass stage glued with muscovite mica layers (∼0.1 mm total thickness) was affixed on a HS-AFM scanner.

To scan the native conformation of ORF6 protein, the sample was first diluted in a scanning buffer (50 mM Tris HCl, 150 mM NaCl, pH 7.4) to 5 ng/μL. Next, the sample was loaded and incubated on mica for 10 min. Finally, the substrate was rinsed twice with scanning buffer to remove free ORF6 protein and then proceeded to HS-AFM scanning. To prepare cationic lipid bilayer substrate, a mixture containing 0.2 mg/mL DPPC/DPTAP/biotin-cap-DPPE mixture (90:5:5 mass ratio) in 10 mM MgCl₂ was sonicated for 5 min in a bath sonicator (7-5027-01, AS ONE, Osaka, Japan) before deposited on bare mica surface. To study the effect of temperature on ORF6 oligomerization, ORF6 protein was incubated at 4 °C, 22 °C, or 37 °C in sample buffer for 3 h prior to HS-AFM scanning. To investigate the role of electrostatic attraction or hydrophobic interaction in ORF6 polymerization in real-time, 5 ng/μL of ORF6 protein was first loaded on bare mica, and the NaCl or urea concentration was adjusted to 2 M during HS-AFM scanning. Likewise, concentration of 1,6-HD, or CHD in scanning buffer was adjusted to a final concentration of 15% w/v during HS-AFM scanning. For premix setting, ORF6 proteins were incubated in either 1,6-HD or CHD at 0%, 5%, 10%, 15%, and 20% respectively for 12 h prior to HS-AFM scanning. All experiments were repeated at least three times to ensure reproducibility.

HS-AFM Image Processing and Analysis. The ImageJ software was used to process and analyze all HS-AFM images (https://imagej.nih.gov/ij/). A first-order polynomial fit in both x- and y-directions was implemented, followed by convolution with a Gaussian smoothing function (“Gaussian Blur” plugin) to improve image quality. Three-dimensional images of selected HS-AFM images were generated using the “3D Surface Plot plugin”. Spatial parameter, the cross-sectional height, was measured using the same software. To measure the velocity (nm per second) of ORF6 filament movement, the “Manual Tracking” plugin in Fiji ImageJ (https://imagej.net/software/fiji/) was used. Furthermore, the movement track was colored dark blue line. The processed image sequences were saved as videos (AVI format), then arranged, edited, and compiled using the Adobe Creative Cloud suite (https://www.adobe.com/creativecloud.html).

NMR Analysis. For NMR experiments, 100 μM ORF6 was dissolved in an NMR buffer (20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 10% D₂O). All NMR experiments were performed at 25 °C using Bruker Avance 600 NEO spectrometer equipped with a cryoprobe. ¹H–¹⁵N SOFAST HMQC spectrum was recorded with 512 (¹H) and 128 (¹⁵N) complex points using a band-selective excitation pulse (¹H pulse offset 7.6 ppm, width 2.0 ppm). The experiment time was about 30 min. All spectra were processed by TopSpin 4.1 (Bruker, MA, USA).

Statistical Analysis. Graphs were computed using GraphPad Prism version 9 (GraphPad, CA, USA) and R software (R Development Core Team). Comparative analyses were performed using SPSS version 28 (IBM Group, NY, USA). Independent samples were compared using the Student t test or Mann–Whitney U test. Related samples (before and after treatment) were compared using the paired t test or Wilcoxon-signed rank text. Statistical significance was set as P < 0.05.

Data Availability Statement

All data are available in the main text or the Supporting Information.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jpclett.3c01440.

• Native conformation of full-length SARS-CoV-2 ORF6 protein (Figure S1), nanotopology of ORF6 oligomers and protofilaments (Figure S2), comparison of ORF6 mobility on two different substrates (Figure S3), distinct dynamic patterns of ORF6 on lipid substrate (Figure S4), disruption of hydrophobic interactions triggered dissociation of ORF6 protofilaments (Figure S5), dissolution of ORF6 aggregates in HeLa cells after 1,6-HD treatment (Figure S6), internalization and accumulation of exogenous GFP-ORF6 protein in lung cancer cell lines (Figure S7), ORF6 aggregates induced IL-6 expression in lung cancer cell lines (Figure S8), in silico analysis of ORF6 amyloidogenic properties (Table S1) (PDF)

• ORF6 oligomers captured using HS-AFM (Movie S1) (MP4)

• Two distinct conformations of ORF6 protofilament (Movie S2) (MP4)

• Spontaneous self-assembly of ORF6 protein (Movie S3) (MP4)

• Higher temperature favored ORF6 oligomerization and circular conformation (Movie S4) (MP4)

• ORF6 protein showed higher mobility on lipid substrate (Movie S5) (MP4)

• Distinct appearances of ORF6 protein on two different substrates (Movie S6) (MP4)

• ORF6 filaments demonstrated rapid interchangeable linear and circular conformation transition on lipid substrate (Movie S7) (MP4)

• ORF6 protein showed complex structural dynamics on lipid substrate (Movie S8) (MP4)

• ORF6 protofilaments were sensitive to urea but not to concentrated sodium chloride (Movie S9) (MP4)

• Disruption of hydrophobic interactions dissociated ORF6 protofilaments (Movie S10) (MP4)

• Transparent Peer Review report available (PDF)

Author Contributions

^◆ G.N., K.L., and M.T. made an equal contribution. Conceptualization: K.L., R.W. Methodology: G.N., K.L., A.K., A.T. Investigation: G.N., K.L., A.K., T.M., Q.Z., H.M. Visualization: G.N. Funding acquisition: K.L., R.W., Q.Z., N.N. Project administration: K.L., R.W., N.N. Supervision: K.L., R.W., N.N. Writing–original draft: K.L., R.W. Writing–review and editing: K.L., R.W., N.N. indicates equal contribution.

The authors declare no competing financial interest.