For the GFP-OSBP and HA-OSBP constructs, the OSBP gene was amplified by PCR using OSBP gateway primers (Table S1) from a pLJM1-FLAG-GFP-OSBP vector [gift from Roberto Zoncu (Addgene plasmid # 134659; http://n2t.net/addgene:134659; RRID: Addgene_134659)] (Lim et al., 2019) and cloned into a pDONR207 gateway vector via BP reaction to generate pDONR207-OSBP. OSBP was further cloned from pDONR207-OSBP into pDEST-GFP and pDEST-N-tag-HA gateway compatible destination vectors via LR reaction as previously described (Ma-Lauer et al., 2016) to generate pDEST-GFP-OSBP and pDEST-HA-OSBP gateway constructs, respectively. For OSBP knockout constructs, the oligo-dimers of OSBP gRNA1 or gRNA2 primers (Table S1) were cloned into a pLentiCrispr v2 vector (gift from Feng Zhang [Addgene plasmid # 52961; http://n2t.net/addgene:52961; RRID: Addgene_52961]) (Sanjana et al., 2014) via BsmBI sites.

For yeast 2-hybrid assays, SARS-CoV-2 genes were amplified from the SARS-CoV-2 genome cDNA using the appropriate primers (Table S1) (Thao et al., 2020) and cloned into pDONR207 vectors via BP reaction to generate constructs of pDONR207-SARS-CoV-2 genes. The transmembrane domain-deleted version of S2 (Δa.a.1206–1254) was generated based on pDONR207-SARS-CoV-2-S2 via Q5 Site-Directed Mutagenesis Kit (New England Biolabs, E0554S) according to the standard protocol provided by the manufacturer (primers used are listed in Table S1). The resulting pDONR207-SARS-CoV-2 genes and pDONR207-OSBP were cloned into yeast gateway destination vectors pGBKT7g and pGADT7g (Stellberger et al., 2012) via LR reaction.

For split YFP constructs, OSBP was cloned from pDONR207-OSBP into a pDEST-c-myc-YFP^N gateway compatible vector via LR reaction to generate YFP^N-OSBP construct as previously described (Ma-Lauer et al., 2016). The VAP-B gene was cloned from pDONR223-VAP-B (Pfefferle et al., 2011) into pDEST-c-myc-YFP^N via LR reaction to generate the YFP^N-VAP-B construct. SARS-CoV-2 Nsp4, Nsp6, orfM, orf3a, orf3b, orf7a, and orf7b genes were cloned from their corresponding pDONR207-SARS-CoV-2 plasmids into a pDEST-HA-YFP^C vector (Ma-Lauer et al., 2016) via LR reaction to generate HA-YFP^C viral gene fusion constructs.

For split-NanoBit assays, the VAP-B gene was PCR amplified using the appropriate primers (Table S1), and the VAP-B PCR product was inserted into pBiT1.1N LgBiT vector (Promega, #N2014) via EcoRI/XbaI cleavage sites to generate pLgBiT-VAP-B constructs. Since no proper cleavage sites provided by pBiT2.1N SmBiT vector (Promega, #2014) could be used to clone OSBP into pBiT2.1N SmBiT vector via restriction enzyme, OSBP was cloned in an alternative way. First, the rfB cassette from the pDEST-GFP Gateway destination vector was amplified using the appropriate primers (Table S1). The resulting rfB PCR fragment was then inserted into pBiT2.1N SmBiT via XhoI/XbaI cleavage sites to generate a pSmBiT gateway compatible destination vector. Finally, the OSBP gene was cloned from pDONR207-OSBP into the newly constructed pSmBiT Gateway destination vector via LR reaction to generate pSmBiT-OSBP.

For other plasmids used in this study, SARS-CoV-2 genes were cloned from the corresponding pDONR207-SARS-CoV-2 gene constructs into a pDEST-RFP gateway destination vector (Ma-Lauer et al., 2016) to generate pDEST-RFP-SARS-CoV-2 genes. The SARS-CoV-2 Nsp3 a.a.1–1363 expressing vector pCG-SARS-CoV-2 Nsp3 a.a.1–1363 was kindly provided by Konstantin Sparrer, Ulm, Germany. For the pSmBiT-GFP construct, GFP was amplified using the appropriate primers (Table S1) and subsequently cloned into the pDONR207 vector via BP reaction. An LR reaction was performed on the resulting pDONR207-GFP and pSmBiT Gateway vector to generate pSmBiT-GFP.

Caco-2 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Gibco) containing 20% (v/v) fetal bovine serum (FBS), 1% sodium pyruvate, 1% non-essential amino acids, and 1% penicillin/streptomycin. Huh7, BHK-21, L929, and HEK293 cells were cultured in DMEM containing 10% (v/v) FBS and 1% penicillin/streptomycin. All cell lines were grown in a humid atmosphere with 5% CO2 at 37°C.

For plasmid transfection into HEK293 or Huh7 cells, cells were seeded on appropriate culture plates to reach 70% to 80% confluence for transfection on the second day. Transfection was performed using Lipofectamine 3000 (Thermofisher) according to the manufacturer’s standard protocol.

The successfully cloned pLentiCrispr v2 OSBP gRNA1 or gRNA2 were co-transfected with psPAX2 (gift from Didier Trono [Addgene plasmid # 12260; http://n2t.net/addgene:12260; RRID: Addgene_12260]) and pCMV-VSV-G (gift from Bob Weinberg [Addgene plasmid # 8454; http://n2t.net/addgene:8454; RRID: Addgene_8454]) (Stewart et al., 2003), plasmids into HEK293T cells to produce Crispr OSBP gRNA lentiviruses. Three days post-transfection, the cell culture supernatant containing lentiviruses was harvested and applied to infect Huh7 cells in a 6-well plate. Forty-eight hours after infection, the Huh7 cells were selected with DMEM growth medium containing 10% (v/v) FBS, 1% Penicillin/Streptomycin and 2 µg/ml puromycin for 5 days. Surviving cell pools were used for Western blot as well as viral infection assays.

Cells at approximately 70% confluence in culture plates were washed with phosphate-buffered saline (PBS) and then inoculated with virus-containing serum-free DMEM for 1.5 hours at 33°C with 5% CO₂. The inoculum was then removed. Cells were washed with PBS and re-cultured in growth medium at 33°C with 5% CO₂ for an additional 24, 48, or 72 hours.

A 40 mM ZJ-1 stock solution was prepared by dissolving 1 mg of the sodium salt of ZJ-1 in 28.44 µl dimethyl sulfoxide (DMSO). The stock solution was then further diluted in growth medium at various concentrations for viral inhibition assays.

For the plaque titration assay, SARS-CoV-2 (BetaCoV/Munich/ChVir984/2020 (B.1, EPI_ISL_406862) plaque reduction titers were determined by plaque titration experiments (n=2) from supernatants of Caco-2 cells treated with ZJ-1. 175,000 Vero E6 cells were seeded in a 24-well plate one day before infection. After washing with PBS, the cells were inoculated with 200 µl of serially diluted cell culture supernatants from infected cells at MOI 0.1. After 1 hour of adsorption at 37°C, the virus dilutions were removed and 500 μL of high-viscosity overlay (1:1 mixture of 2.4% Avicel and 2x concentrated DMEM supplemented with 5% FBS, 2% non-essential amino acids (#11140–035, Gibco), and 2% sodium pyruvate (#11360–039, Gibco) was added to each well. The overlay was discarded at 3 dpi and cells were fixed in 6% formaldehyde for 30 minutes, washed with PBS, and stained with crystal violet solution (Sigma Aldrich) for 15 minutes. Plaques were counted from one to two dilutions where clear plaques were detectable (Niemeyer et al., 2022).

Mouse antibody 1H11 (1:400) recognizing HCoV-229E/-NL63 N-protein was purchased from INGENASA, Spain. Rabbit anti-OSBP (HPA039227, 1:1,000) and mouse anti-vinculin (V9264, 1:1,000) were purchased from Sigma-Aldrich. Rat anti-HA antibody (1:1,000) was purchased from Roche (11867423001). Rabbit polyclonal anti-GFP (1:1.000) was purchased from Invitrogen (A6455). Mouse anti-RFP (1:1.000) was purchased from Invitrogen (MA5–15257). Mouse anti-PDI (1:100) was purchased from Invitrogen (MA3–019). Mouse anti-dsRNA (1:1,000) was purchased from SCICONS, Hungary (J2), and anti-β-actin antibody (1:100,000) was provided by Sigma-Aldrich (A3854).

The anti-Nsp3 7G9–1-1 (SARS-CoV-2) monoclonal antibody was generated as described previously (Flaswinkel and Reth, 1994; Yeruva et al., 2020). Briefly, a peptide containing the amino acids DKKIKA-abu-VEEVTTTLEET (orf1a aa pos. 1229–1246; MN908947.2) was synthesized and coupled to OVA (Peps4LS, Heidelberg, Germany). Within this sequence, C was replaced by abu to avoid problems with coupling of the peptide to ovalbumin. C57BL/6 mice were immunized subcutaneously and intraperitoneally with a mixture of 50 µg peptide OVA, 5 nmol CPG oligonucleotide (Tib Molbiol, Berlin, Germany), 100 µl PBS and 100 µl incomplete Freund’s adjuvant. A boost without adjuvant was given six weeks after the primary injection. Fusion was performed according to standard procedures. Hybridoma supernatants were first tested by ELISA. Briefly, supernatants were positively selected for binding to SUD-VE: eBio-DKKIKACVEEVTTTLEET-HHHHHH and negatively selected for binding to SUD-ID: eBio-DKKIKACIDEVTTTLEET-HHHHHHH peptides. In both cases, the respective peptides were bound to avidin-coated ELISA plates from stock. Unbound peptide was removed before hybridoma supernatant was applied. Selected supernatants were further analyzed by Western blot. Anti-Nsp3 7G9–1-1 was cloned based on its ability to bind also in Western blot.

Peroxidase-conjugated secondary anti-mouse (A9917, 1:20,000); and anti-rat (A9037, 1:5,000; A9037) antibodies were obtained from Sigma-Aldrich. Peroxidase-conjugated anti-rabbit secondary antibody (1:1,000) was purchased from DAKO (P0217). Goat anti-rabbit FITC (1:1000; Sigma-Aldrich, F0382) and goat anti-mouse Alexa-555 (1:500; ThermoFisher, A21424) were used as fluorescent secondary antibodies.

The inhibitor ZJ-1, a structurally simplified analog of the natural product OSW-1, was originally designed and synthesized by the research team of Prof. Zhendong Jin at the Department of Pharmaceutical Sciences and Experimental Therapeutics, College of Pharmacy, The University of Iowa, as a novel anticancer agent that exhibits exactly the same potent anticancer activity and therapeutic selectivity as OSW-1 by in vitro evaluation (Zhou et al., 2005). Structurally, the C-17 tertiary alcohol in OSW-1 is sterically hindered and is unlikely to be involved in binding to the drug target. By replacing the C-17 hydroxy group with a hydrogen, the synthesis of ZJ-1 (17-deoxy OSW-1) is simpler than that of OSW-1, and the synthetic process developed by Prof. Jin’s team can be readily scaled up. Currently, ZJ-1 is available from Cfm Oskar Tropitzsch GmbH, Germany.

The Bioline ISOLATE II RNA Mini Kit (BIO-52073) was used to isolate RNA from either supernatant or cells according to its standard protocol. For inhibition assays of ZJ-1 with HCoV-OC43, viral RNA isolated from the supernatant was processed by probe-based qPCR (SensiFAST Probe Hi-ROX One-Step Kit (BIO-77005), BIOLINE) using the appropriate primers and probes listed in Table S1. For Sybr Green qPCR, total RNA isolated from cells was used as a template to synthesize cDNA using oligonucleotide (dT)18 and SuperScript IV reverse transcriptase (Thermofisher Scientific/Invitrogen, 18090050). Quantification was then performed by SYBR Green qPCR using HA-YFP^C- and β-actin-specific primers (Table S1) and AceQ SYBR qPCR Master Mix (#Q111, Absource/Vazyme).

For the Renilla luciferase assay, HCoV-229E-RLuc-infected cells (Carbajo-Lozoya et al., 2012) were harvested using the Renilla Luciferase Assay System (Promega, #E2820) according to its standard protocol. Luciferase activity was measured using a CLARIOstar microplate reader (BMG LABTECH). For the Gaussia luciferase assay, the supernatant of MHV-GLuc-infected L929 cells was directly harvested for luciferase activity measurement using the Renilla luciferase substrate. For the NanoBiT luciferase assay, cells from each well in a 96-well plate were lysed with 20 µl NP-40 lysis buffer (1%NP-40, 150 mM NaCl, 50 mM Tris-HCl, pH 8.0) and then treated with 20 µl substrate containing 0.5% Nano-Glo^® Luciferase Assay Substrate (Promega, #N113A) in 1x Nano-Glo^® Blotting Buffer (Promega, #N242A) for measurement.

Cell viability was measured using the CellTiter-Glo^® 2.0 Cell Viability Assay (Promega, #G9242). Cell lines were plated in 96-well plates and incubated with inhibitor concentrations corresponding to each inhibition experiment.

HEK293 cells were transfected with plasmids in a 6-well culture plate using Lipofectamine 3000 (Thermofisher). Twenty-four hours after transfection, cells were lysed and processed with GFP-Trap_A bead-based CoIP (ChromoTek) according to its standard protocol. Western blotting protocol was described elsewhere (Carbajo-Lozoya et al., 2014).

The mock and infected cells seeded on coverslips with or without ZJ-1 treatment were washed with 1x PBS (Gibco) and fixed in 4% PFA (Carl Roth GmbH) in PBS for 15 minutes. After a wash step with PBS, cell membranes were permeabilized in 0.1% Triton X-100 (Carl Roth GmbH) in PBS for 10 minutes for 15 minutes. The cells were washed again with PBS and blocked in PBS containing 5% BSA (Carl Roth GmbH) and 0.2% Tween-20 (Sigma-Aldrich Chemie) for one hour at room temperature. The cells were then incubated with primary antibodies of anti-PDI (Invitrogen) at 1:100 or anti-dsRNA (SCICONS) at 1:1000 and anti-OSBP (Sigma-Aldrich) at 1:300 dilution in PBS containing 5% BSA and 0.2% Tween^®20 at 4°C overnight. After three washes with PBS, cells were incubated with secondary antibodies AlexaFluor^®555 anti-mouse (Invitrogen) at 1:500 and FITC anti-rabbit (Invitrogen) at 1:1000 in PBS containing 5% goat serum (Bio&SELL GmbH) and 0.2% Tween^®20 at room temperature in the dark for 1 hour. The cells were then stained with DAPI (Sigma Aldrich) diluted 1:1000 in PBS for 10 minutes in the dark. The cells were washed three times with PBS. Immunostained coverslips were mounted onto glass slides using Roti^® Mount-FluorCare (Carl Roth GmbH). Images were captured using a Leica DM4000 B fluorescence microscope with a 63x objective. Split-YFP slides of live cells were prepared as described elsewhere (Ma-Lauer et al., 2016). Here, images were captured using a 40x objective. GFP, RFP, and DAPI or Hoechst images of fixed cells were captured using a 20x objective.

Yeast two-hybrid screening was performed by utilizing a Gateway-compatible yeast 2-hybrid system (Stellberger et al., 2012). First, all SARS2-CoV-2 genes were cloned into pGBKT7g bait vector encoding leucine, and OSBP was cloned into pGADT7g prey vector expressing tryptophan. SARS-CoV-2 subdomains, SUD (a.a. 413–745 in Nsp3), PLP (a.a. 746–1064 in Nsp3), Nsp12 NiRAN (a.a. 1–249 in Nsp12), interface (a.a. 250–396 in Nsp12) and RdRp (a.a. 397–932 in Nsp12), S-RBD (a.a. 325–588 in spike), S2 (a.a. 662–1273 in spike) and S2 Δa.a.1206–1254 (transmembrane domain deleted) were also cloned. The gap in Nsp3 between a.a. 1065–1414 and a.a. 1547–1945 containing transmembrane regions was intentionally excluded. The competent yeast (PJ69–7A) cells transformed with both prey and bait plasmids were cultured in double selection dropout medium (without Leu and Trp) in 96- deep well plates for 4 days and then stamped to both double selection and triple selection dropout agar plates (without Leu, Trp and His) for another 3 days. Theoretically, yeast cells transformed with both prey and bait plasmids should grow on triple selection plates only if the SARS-CoV-2 proteins interact with OSBP. Thus, yeast colonies on double selection plates indicated successful transformation of both prey and bait plasmids and growth on triple selection plates indicated interaction between viral and cellular proteins fused to activation and binding domains of the Gal4 transcription factor. On each agar plate, positive (pGBKT7g-SARS-CoV-SUD plus pGADT7g-RCHY1 and negative controls (pGBKT7g vector plus pGADT7g-OSBP) were included. Conversely, control screenings were performed in the same manner between pGADT7g vector (instead of pGADT7g-OSBP) and bait plasmids fused to all SARS-CoV 2 genes or subdomains.

The data from qPCR and luciferase measurements are presented as the mean ± standard deviation of triplicate or quadruplicate values. Statistical analysis is conducted using one-way ANOVA with Dunnett’s multiple comparison test, or student t-test for two-group analysis utilizing GraphPad Prism 10 software with a significance level set at α=0.05. Graphs for all data are created using GraphPad Prism 10, using both the raw data and statistical analyses performed within the same software. The symbols in charts denote the following significance levels of p-values: “ns” (not significant) indicates p>=0.05, one asterisk (*): 0.01<p<0.05, two asterisks (**): 0.001<p<0.01, three asterisks (***): p<0.001, four asterisks (****): p<0.0001.

It has been shown that knockdown of OSBP impairs the replication of HCV and Aichi virus (Amako et al., 2009; Ishikawa-Sasaki et al., 2018), suggesting an auxiliary function of OSBP in the replication of (+) ssRNA viruses. To determine whether OSBP also mediates coronavirus growth, HEK293 cells transiently expressing the HCoV-229E aminopeptidase N (APN) receptor and GFP-OSBP or GFP control were challenged with HCoV-229E carrying a Renilla luciferase reporter (HCoV-229E-RLuc) (Ma-Lauer et al., 2020). As a result, expression of GFP-OSBP resulted in significantly increased luciferase activity of HCoV-229E-RLuc compared to the GFP control ( Figure 1A ). This suggests that OSBP positively mediates HCoV-229E replication.

To confirm that OSBP enhances coronavirus replication, an OSBP knockout was generated in Huh7 cells, which can be naturally infected by several human coronaviruses. Subsequently, OSBP knockout Huh7 cell pools with two different CRISPR guide RNAs (gRNA) were challenged with HCoV-229E-RLuc. As shown in Figure 1B , luciferase activity carried by HCoV-229E-RLuc was significantly reduced in both independent OSBP knockout Huh7 cell pools. Western blot showed that the knockout of OSBP was effective in both pools, although not 100% efficient ( Figure 1B ). Taken together, OSBP positively regulates HCoV-229E-RLuc replication.

We further investigated whether an OSBP-binding compound could inhibit coronavirus growth, as OSBP has stimulated HCoV-229 replication and OSBP-binding compounds have been reported to have antiviral activity against enteroviruses and HCV (Amako et al., 2009; Albulescu et al., 2015; Strating et al., 2015; Albulescu et al., 2017; Roberts et al., 2019). To achieve this goal, a novel OSBP-binding compound, ZJ-1 ( Figure 1C ), was applied to the SARS-CoV-2 replication assay. ZJ-1 is a structurally simplified analog of the oxysterol binding protein (OSBP)-binding compound OSW-1. Similar to the effect of its parent compound OSW-1 (Roberts et al., 2019), treatment with ZJ-1 also reduces endogenous OSBP in several cell lines ( Figure 1D ). Western blot measuring nucleocapsid expression showed that ZJ-1 treatment reduced protein intensity of nucleocapsid of HCoV-229E and HCoV-NL63 in Huh7 cells, while cell viability assay revealed low cytotoxicity of ZJ-1 to Huh7 cells ( Figure 1E ).

In addition, immunofluorescence microscopy using anti-OSBP and anti-dsRNA antibodies suggested antiviral activity of ZJ-1 against SARS-CoV-2 and close association of OSBP with replication organelles ( Figure 1F ). In DMSO-treated mock-infected Huh7 cells, endogenous OSBP is distributed in the cytosol but concentrated in the ER ( Figure 1F ; Supplementary Figure S1 ), which is consistent with previous reports (Mesmin et al., 2013; Albulescu et al., 2017). dsRNA was not detectable as there was no viral infection ( Figure 1F ). In ZJ-1-treated mock-infected cells, the protein level of OSBP is significantly reduced ( Figure 1F ; Supplementary Figure S1 ), which is consistent with the Western blot results ( Figure 1D ). In the DMSO-treated HCoV-229E-infected cells, OSBP clearly co-localizes with dsRNA ( Figure 1F , images in the 3rd row), indicating its close proximity to replication organelles (DMVs). In DMSO-treated SARS-CoV-2 infected cells, the distribution of both dsRNA and OSBP is slightly different, although they still co-localize ( Figure 1F , images on the 5th row). Here, dsRNA is distributed almost throughout the cytosol and weak OSBP signal is also detected in the nucleus for an unknown reason. Notably, both HCoV-229E and SARS-CoV-2 replication were severely impaired by ZJ-1 treatment, as dsRNA was almost undetectable in ZJ-1-treated virus-infected Huh7 cells ( Figure 1F , images in the 4th and 6th rows), suggesting antiviral activity of ZJ-1 not only for HCoV-229E but also for SARS-CoV-2.

To test whether the antiviral effect of ZJ-1 against SARS-CoV-2 is Huh7 specific, Caco-2 cells, which are much more susceptible to infection by SARS-CoV-2, were used (Saccon et al., 2021). Plaque titration experiments from supernatants of Caco-2 cells treated with ZJ-1 showed that nanomolar range of ZJ-1 treatment (62.5 nM) resulted in approximately three log reduction of viral replication at both 24 and 48 h p.i. ( Figure 1G ), while the 72 h ZJ-1 CC₅₀ in Caco-2 cells was found to be 19.1 µM ( Figure 1H ). This result demonstrates that ZJ-1 inhibits SARS-CoV-2 replication in different cell types.

It was further investigated whether ZJ-1 also inhibited the viral growth of HCoV-OC43 and murine hepatitis virus (MHV) in BHK21 and L929 cells, respectively. The qPCR assay measuring OC43 viral RNA ( Figure 1I ) and the luciferase assay measuring the activity of the Gaussia luciferase reporter carried by MHV-Gaussia luciferase (MHV-GLuc) ( Figure 1J ) showed that both viruses were also inhibited by ZJ-1.

Overexpressing ( Figure 1A ) and knocking out ( Figure 1B ) OSBP tends to stimulate and reduce coronavirus replication, respectively. In addition, ZJ-1, which reduces endogenous OSBP, inhibits viral replication ( Figures 1D–J ). Therefore, OSBP may be an auxiliary host factor for coronaviruses. To further investigate how viral proteins use OSBP to promote viral growth, OSBP was used as a bait to screen SARS-CoV-2 viral proteins for viral-host interactions by yeast 2-hybrid (Y2H) assay. As a result, SARS-CoV-2 nonstructural protein 4 (Nsp4), Nsp6, Nsp3 a.a.1–412, envelope protein (orfE), membrane protein (orfM), and accessory proteins orf3a, orf3b, orf7a, orf7b, and orf9c were shown to be binding partners of OSBP ( Figure 2A ). The Nsp3 SUD domain, Nsp3 a.a.1065–1414, and Nsp3 a.a.1547–1945 also appear to interact weakly with OSBP ( Figure 2A ). The full-length Nsp3 protein was not tested in the Y2H assay because it was too large to be successfully cloned into the yeast vectors. The various negative controls are shown in Supplementary Figure S2 .

Among the coronaviral proteins, the non-structural proteins Nsp3, Nsp4, and Nsp6 are involved in DMV formation (Angelini et al., 2013; Lundin et al., 2014; Wolff et al., 2020). In the Y2H assay, Nsp3 fragments, Nsp4 and Nsp6 appeared as interacting proteins of OSBP ( Figure 2A ). OSBP is known to mediate the exchange of cholesterol and PI4P between the membrane contacts of the ER and DMVs (Nagy et al., 2016). Therefore, it was necessary to examine the binding of OSBP to Nsp4 and Nsp6 in human cells and to investigate whether OSBP interacts with full-length Nsp3.

The ER-localized vesicle-associated membrane protein (VAMP)-associated proteins (VAPs), including VAP-A and VAP-B, anchor OSBP to the ER membrane (Furuita et al., 2010; Baron et al., 2014) and are targeted by some viral proteins (Gao et al., 2004; Hamamoto et al., 2005). Therefore, it was also interesting to analyze the binding between one of the VAPs and the viral proteins involved in DMV formation. OSBP, VAP-B, SARS-CoV-2 Nsp4 and Nsp6 were cloned into split-YFP vectors (Lei et al., 2021) to study their binding and to visualize the binding sites. Nsp3 could not be cloned into split-YFP vectors due to its large size. The split-YFP assay in HEK293 cells indicates that Nsp4 and Nsp6 interact with both OSBP and VAP-B in the cytosol with accumulation at ER-like structures ( Figure 2B ).

Since the full-length Nsp3 always proved to be too large for successful cloning, a large fragment of Nsp3 (a.a.1–1363), covering 70% of the full-length ORF, was alternatively examined for interaction with OSBP in human cells. The plasmid pCG-SARS-CoV-2 Nsp3 a.a.1–1363 was co-transfected with either pDEST-GFP control or pDEST-GFP-OSBP into HEK293 cells for a GFP-trap based co-immunoprecipitation (CoIP) assay. As shown in Figure 2C , SARS-CoV-2 Nsp3 a.a.1–1363 could be co-immunoprecipitated with GFP-OSBP but not with GFP control, indicating the affinity of OSBP for this Nsp3 fragment. In addition, Nsp3 a.a.1–1363 also showed binding activity to VAP-B ( Figure 2D ). Thus, all SARS-CoV-2 viral proteins involved in DMV formation (Nsp4, Nsp6, and Nsp3) target OSBP, which mediates cholesterol transfer from the ER to DMVs, and VAP-B, which is known to anchor OSBP to the ER membrane.

OSBP has previously been reported to interact with VAP-B (Baron et al., 2014). In addition, both proteins are targeted by viral Nsp3, Nsp4 and Nsp6 ( Figures 2B–D ). Therefore, it was also interesting to investigate whether the viral proteins could enhance or disrupt the binding between VAP-B and OSBP. For this purpose, VAP-B and OSBP were cloned into the split-NanoBiT vector p1.1N LgBiT (large BiT) and p2.1N SmBiT (small BiT), respectively. The plasmids expressing LgBiT-VAP-B and SmBiT-OSBP fusion were then co-transfected into HEK293 cells for measurement of split-NanoBiT luciferase activity in the absence or presence of viral proteins. As shown in Figure 2E , a clear interaction was detected when both LgBiT-VAP-B and SmBiT-OSBP were co-transfected into cells in the absence of SARS-CoV-2 Nsp3 a.a.1–1363. However, in the presence of SARS-CoV-2 Nsp3 a.a.1–1363, the split-NanoBiT signal resulting from the interaction between LgBiT-VAP-B and SmBiT-OSBP was almost completely abolished. Unlike the large Nsp3 fragment, Nsp4 does not alter the binding between VAP-B and OSBP, whereas Nsp6 seems to interfere with the binding signal to some extent ( Figure 2F ). It was then investigated whether Nsp3 a.a.1–1363 and Nsp6 could down-regulate the HSV-TK promoter carried by the split-NanoBiT vectors to cause a reduced split-NanoBiT signal. For this purpose, GFP was cloned into a split-NanoBiT vector generating a SmBiT-GFP fusion construct under the control of the HSV TK promoter. This plasmid expressing SmBiT-GFP was co-transfected with control or viral protein expressing vectors. As shown in Supplementary Figure S3 , none of the SARS-CoV-2 Nsp3 large fragment, Nsp4, and Nsp6 downregulated the HSV TK promoter, as the SmBiT-GFP signal intensity was comparable in all samples, excluding the influences of the Nsps on the TK promoter.

According to the Y2H screening results, OSBP should also interact with SARS-CoV-2 orf7b, orf7a, orfM, orfE, orf3a, orf3b, and orf9c ( Figure 2A ). The original idea was to confirm the interaction between OSBP and these viral proteins in human cells. To achieve this goal, SARS-CoV-2 orf7b was first cloned into the pDEST-HA tag vector to generate the HA-orf7b construct under the control of the CMV promoter for further co-immunoprecipitation (CoIP) assays. However, for an unknown reason, despite the correct nucleotide sequences, there was no expression of HA-orf7b. Therefore, SARS-CoV-2 orf7b was alternatively cloned into a split YFP vector pDEST-HA-YFP^C (Lei et al., 2021) to generate a pDEST-HA-YFP^C-orf7b construct. YFP^C consists of the C-terminal fragment of YFP (a.a. 156–239). Identical amounts of plasmid DNA were transfected into HEK293 cells for CoIP assay. Interestingly, a strong increase in HA-YFP^C-orf7b protein accumulation was observed in the presence of OSBP ( Supplementary Figure S4 , lane 2, input) compared to the control ( Supplementary Figure S4 , lane 1). Although HA-YFP^C-orf7b was successfully co-immunoprecipitated by GFP-OSBP (lane 2), a firm conclusion from this assay for the binding of OSBP to orf7b cannot be drawn from this assay due to the strong increase in orf7b expression in the presence of OSBP. The size of the HA-YFP^C control protein is too small to be detected by Western blot (lane 3). A further Western blot confirmed that OSBP affects the protein level of SARS-CoV-2 orf7b ( Figure 3A ). Orf7a and orfM were used as random controls. By chance it was found that OSBP stimulated protein accumulation of orf7a, but it showed no effect on orfM. Since orf7a, orf7b, and orfM were cloned into the same plasmid pDEST‐HA‐YFP^C under the control of the CMV promoter, we expected that the OSBP-mediated protein accumulation of orf7a and orf7b was a result of the protein level rather than the transcription level. This assumption was confirmed by qPCR analysis ( Supplementary Figure S5 ). To determine whether the protein accumulation of orf7a and orf7b was due to increased protein biosynthesis or decreased protein degradation, cycloheximide (CHX) chase assays were performed to prevent protein synthesis and determine protein stability. As shown in Figures 3B, C , co-expression of OSBP resulted in prolonged half-lives of orf7a and orf7b, respectively. Thus, OSBP stabilizes SARS-CoV-2 orf7a and orf7b at the protein level.

Whether OSBP could stabilize the other binding partners was also tested. pDEST plasmids expressing RFP-fused viral proteins under the control of the CMV promoter were co-transfected with either pDEST-GFP or pDEST-GFP-OSBP vectors into HEK293 cells. The transfected cells were observed 24 hours after transfection. The results showed that OSBP also increased the protein level of orf3a ( Figure 3D ), but had no effect on Nsp4, Nsp6, orfE, orf3b and orf9c ( Supplementary Figure S6 ). In addition, the protein level of Nsp3 a.a.1–1363 was not altered by OSBP, as shown by Western blot ( Supplementary Figure S6 ). The CHX-chase assay showed that orf3a can also be stabilized by OSBP ( Figure 3E ).

Orf7a, orf7b and orf3a are binding candidates of OSBP according to the Y2H screening ( Figure 2A ) and they can be stabilized by OSBP ( Figures 3A–E ). Therefore, it is necessary to investigate their interactions in human cells. CoIP results confirmed the interactions ( Figures 3F, G ), suggesting that these accessory proteins are stabilized by OSBP probably through direct interaction. In addition, OSBP also appears to bind orf3b ( Figure 3G ), which is consistent with the Y2H screening result ( Figure 2A ).