Cholesterol 25-hydroxylase suppresses SARS-CoV-2 replication by blocking membrane fusion

Significance The novel severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), the etiological agent of coronavirus disease-2019 (COVID-19), has swept the world in unprecedented speed. In a few months, SARS-CoV-2 has infected millions of people and caused tens of thousands of deaths. There are no Food and Drug Administration-approved antivirals or vaccines yet available and clinical treatments are limited to supportive therapies that help alleviate the symptoms. Thus, there is an urgent need to identify effective antivirals as countermeasures before safe and effective vaccines are developed, tested, and then produced on a large scale. Our approach is to harness the germline-encoded interferon antiviral response to inhibit SARS-CoV-2 replication thereby limiting its pathogenicity.

T he novel severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), the etiological agent of coronavirus disease-2019 (COVID-19) (1, 2), has infected millions of people worldwide and caused hundreds of thousands of deaths. Currently, there are no Food and Drug Administration-approved vaccines available. In most instances, treatment is limited to supportive therapies to help alleviate symptoms. With the concern that monotherapy would rapidly result in the emergence of resistance, there is a pressing need to identify multiple effective antivirals as countermeasures before safe and efficacious vaccines are developed and deployed. Here, we sought to harness the host innate immune responses to inhibit SARS-CoV-2 replication. Interferons (IFNs) are a group of small, secreted proteins (3,4) that potently suppress the replication of many viruses through the action of hundreds of IFNstimulated genes (ISGs) (5). IFN and ISG levels are up-regulated in SARS-CoV-2-infected cells and lung tissues from COVID-19 patients (6,7). Compared to SARS-CoV, SARS-CoV-2 appears to be more sensitive to the antiviral activities of IFNs (8). SARS-CoV-2 replication is inhibited by IFN treatment in both immortalized and primary cells (9)(10)(11). While direct IFN administration often results in adverse effects in humans (12,13), a targeted approach of activating the antiviral effects of specific ISGs holds promise.
To identify potential ISG effector proteins that act to block coronavirus (CoV) at the entry or egress stages of the replication cycle, we utilized replication-competent chimeric vesicular stomatitis virus (VSV) eGFP reporter viruses decorated with either fulllength SARS-CoV spike (S) protein or SARS-CoV-2 S in place of the native glycoprotein (G) (14). We used an HEK293 cell line that stably expresses plasma membrane-localized mCherry-tagged human ACE2, the SARS-CoV and SARS-CoV-2 receptor (2, 15-17) (SI Appendix, Fig. S1A). HEK293-hACE2-mCherry cells supported 100-fold more VSV-SARS-CoV-2 replication than wildtype HEK293 cells (SI Appendix, Fig. S1 B-D). We recently showed robust SARS-CoV-2 infection of primary human intestinal enteroids (18). By RNA-sequencing of these intestinal enteroid cultures, we identified the ISGs most highly and commonly induced by type I IFN (IFN-β) and type III IFN (IFN-λ). We transduced HEK293-hACE2 stable cells with lentiviruses encoding 57 of these individual ISGs and tested their ability to suppress VSV-SARS-CoV and VSV-SARS-CoV-2 replication.

Significance
The novel severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), the etiological agent of coronavirus disease-2019 (COVID- 19), has swept the world in unprecedented speed. In a few months, SARS-CoV-2 has infected millions of people and caused tens of thousands of deaths. There are no Food and Drug Administration-approved antivirals or vaccines yet available and clinical treatments are limited to supportive therapies that help alleviate the symptoms. Thus, there is an urgent need to identify effective antivirals as countermeasures before safe and effective vaccines are developed, tested, and then produced on a large scale. Our approach is to harness the germline-encoded interferon antiviral response to inhibit SARS-CoV-2 replication thereby limiting its pathogenicity.
CH25H encodes cholesterol 25-hydroxylase that catalyzes the formation of 25-hydroxycholesterol (25HC) from cholesterol (23). In many cell types, 25HC is further converted to 7-α, 25-dihydroxycholesterol , an oxysterol that functions as a chemoattractant for T cells and B cells (24). 25HC exhibits broad inhibitory activities against enveloped viruses of different families (25,26), including two porcine CoVs (27). Within a singlecycle of replication (6 hpi), CH25H expression slightly inhibited the replication of VSV-SARS-CoV and VSV-SARS-CoV-2 viruses, as detected by measurement of eGFP expression using flow cytometry (Fig. 1C). CH25H also decreased wild-type VSV replication (Fig. 1C), as previously reported (28). We interpret the partial reduction of eGFP fluorescence intensity in the infected cells as an indication that CH25H treatment slows down the kinetics of infection. In contrast, rotavirus and adenovirus replication were not affected (Fig. 1D). Un-like IFIH1, CH25H expression or 25HC treatment did not induce type I or type III IFN expression (SI Appendix, Fig. S1G). The replication of a clinical isolate of SARS-CoV-2 (2019-nCoV/USA-WA1/2020 strain) also was suppressed in HEK293-hACE2 cells expressing CH25H compared to control plasmid transfection (Fig. 1E).
To examine the physiological relevance of CH25H antiviral activity, we used CRISPR/Cas9 to edit the endogenous CH25H locus. CH25H knockout was verified by both Sanger sequencing and Western blot (SI Appendix, Fig. S1 H and I). Genetic depletion of CH25H expression resulted in a significant increase of VSV-SARS-CoV-2 replication (Fig. 1F). Taking these data together, our genetic screen revealed CH25H as a host restriction factor of SARS-CoV-2 replication.
Next, we tested whether the antiviral activity of CH25H depends on 25HC synthesis. As compared to the control 7-α, 25-diHC, pretreatment of HEK293-hACE2 cells with 25HC for 1 h prior to VSV-SARS-CoV-2 infection recapitulated the suppressive effect of CH25H overexpression and reduced virus replication ( Fig. 2A). 25HC dose-dependently inhibited VSV-SARS-CoV-2 infection in MA104 cells, with an approximate EC 50 of 1.49 μM (with 95% confidence interval of 1.09 to 2.08 μM) (Fig. 2B). The 50% cytotoxic concentration (CC 50 ) was calculated to be at 239 μM. Thus, 25HC had an excellent selectivity index of 160. When plaque assays were performed in the presence of 25HC, there was a reduction in both plaque numbers and sizes (SI Appendix, Fig. S2 A and B). Similarly, wild-type SARS-CoV-2 virus replication was inhibited by 25HC treatment (Fig. 2C). The infectivity of a lentivirus backbone-based SARS-CoV-2 pseudovirus also was suppressed by 25HC, suggesting that the antiviral effect is restricted to early-stage virus replication (entry, transcription, translation, and so forth). We and others previously showed that primary human intestinal epithelial cells and cardiomyocytes are extrapulmonary SARS-CoV-2 target cells (7,18,29). Importantly, 25HC inhibited VSV-SARS-CoV-2 replication in these primary cell culture systems in a dose-dependent manner ( Fig. 2 E and F and SI Appendix, Fig. S2C).
We next examined the effect of 25HC on SARS-CoV S and SARS-CoV-2 S mediated membrane fusion, since 25HC blocks cell fusion by Nipah F and VSV G proteins (28), which are class I and class III viral fusion proteins, respectively (35). We set up an in vitro cell-to-cell fusion assay based on the expression of S, eGFP, ACE2, and TMPRSS2 in HEK293 cells, independent of virus infection (Fig. 3C). CH25H expression substantially reduced syncytia formation mediated by SARS-CoV-2 S (Fig. 3C). Although IFITM2 and IFITM3 inhibited VSV-SARS-CoV-2 replication ( Fig. 1 A and B), neither prevented S-mediated fusion (Fig. 3C), suggesting a distinct mode of antiviral action. Compared to SARS-CoV-2 S, SARS-CoV S induced weaker cell fusion, as recently reported (36), and this process was also blocked by CH25H expression (Fig. 3C). CH25H also inhibited the syncytia formation induced by Western equine encephalitis virus glycoproteins (class II) and VSV-G (class III) but not reovirus FAST p10 (class IV) fusion protein (Fig. 3D). To mimic the virus-cell membrane fusion, we cotransfected SARS-CoV-2 S and GFP into donor cells and mixed at 1:1 ratio with ACE2 + TMPRSS2 + TdTomato cotransfected target cells. As expected, we observed robust syncytia formation under mock conditions (SI Appendix, Fig. S3B). CH25H expression in "recipient" cells almost completely abolished cell-cell fusion (SI Appendix, Fig. S3B). Exogenous 25HC treatment phenocopied CH25H expression and blocked SARS-CoV-2 S-mediated syncytia formation ( Fig. 3E and SI Appendix, Fig. S3B). Similar to CH25H, 25HC failed to inhibit reovirus FAST p10mediated fusion (SI Appendix, Fig. S3C).
To define the underlying antiviral mechanisms of the IFN-CH25H-25HC axis further, we investigated whether 25HC acts on viral or host membranes. Preincubation of VSV-SARS-CoV-2 with 10 μM of 25HC for 20 min had no effect on infectivity, as opposed to the pretreatment of host cells (SI Appendix, Fig.  S4A). The timing of 25HC addition suggests it primarily acted at the entry stage of VSV-SARS-CoV-2 replication (SI Appendix, Fig. S4B). We examined a series of early events and excluded possible effects of 25HC on: 1) ACE2 surface levels; 2) S cleavage by TMPRSS2; 3) lipid raft localization, stained by a fluorophoreconjugated cholera toxin subunit B; 4) plasma membrane fluidity, stained by 6-dodecanoyl-2-dimethylamino naphthalene (Laurdan) (37); 5) endosomal pH; and 6) its ability to directly bind to recombinant SARS-CoV-2 S protein (SI Appendix,  . At 24 hpi, the mRNA level of SARS-CoV-2 N was measured by qRT-PCR and normalized to GAPDH expression. (F) HEK293-hACE2 cells expressing CH25H were transduced with mock or CH25H CRISPR/Cas9 targeting lentiviruses and infected with VSV-SARS-CoV-2 (MOI = 1). At 24 hpi, the mRNA level of VSV N was measured by qRT-PCR and normalized to GAPDH expression. For all panels except A, experiments were repeated at least three times with similar results. The experiment in A was performed twice with average numbers indicated on the graph. Raw data are listed in Dataset S1. Data are represented as mean ± SEM. Statistical significance is from pooled data of the multiple independent experiments (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001). Fig. S4F), enabling us to use it as a tool to probe the antiviral mechanism of 25HC. Upon host cell uptake, C4 TopFluor-25HC exhibited punctate patterns and partially colocalized with lysobisphosphatidic acid (LBPA)-positive late endosomes and LAMP1 + lysosomes but not Rab4 + recycling endosomes (Fig. 4A). Thus, we hypothesized that SARS-CoV-2 depends on endosomal trafficking to establish active replication. Consistent with this hypothesis, ectopic expression of Rab5 and Rab7 dominant-negative mutants but not the wild-type proteins significantly decreased VSV-SARS-CoV-2 infection (Fig. 4B and SI Appendix, Fig. S4G). However, Rab5 and Rab7 mutants did not have an additive effect with 25HC treatment (Fig. 4B), further suggesting that 25HC may act at these endosomal vesicles.

Fig. S4E) and blocked SARS-CoV
25HC is capable of binding Niemann-Pick C1 (NPC1) in vitro (39), which is responsible for the egress of cholesterol from the endosomal/lysosomal compartment (40). 25HC treatment led to an accumulation of intracellular TopFluorcholesterol (Fig. 4C). We also observed that in cells stained with filipin, which labels unesterified cholesterol and preferentially recognized cholesterol over 25HC (SI Appendix, Fig.  S5A), a cholesterol accumulation by 25HC and NPC1 inhibitors itraconazole (ICZ) and U18666A (Fig. 4C). ICZ treatment significantly reduced VSV-SARS-CoV-2 titers (SI Appendix, Fig. S5B). The antiviral activity of 25HC and ICZ depended on cholesterol accumulation in the endosomal/lysosomal compartment because the inhibition was markedly diminished in serum-free media that is devoid of cholesterol (SI Appendix, Fig. S5C). Consistent with our model, pharmacological inhibition of NPC1 by U18666A or genetic depletion of NPC1 by CRISPR/Cas9 potently restricted VSV-SARS-CoV-2 replication (Fig. 4 D and E and SI Appendix, Fig. S5D). In contrast to chloroquine and camostat, both of which are antiviral but through different mechanisms, 25HC, ICZ, or U18666A efficiently reduced SARS-CoV-2 S-mediated cell-cell fusion (SI Appendix, Fig.  S5E). Finally, all three compounds suppressed the replication of a recombinant SARS-CoV-2 virus that encodes an mNeon-Green reporter (41) in Vero E6 cells (Fig. 4F). Collectively, our data support a model that 25HC suppresses SARS-CoV-2 S protein-mediated fusion, which inhibits virus replication, possibly by altering cholesterol levels.
The identification of ISGs against different virus families have provided invaluable insights into both virus entry pathways and host innate immune system evolution (42)(43)(44)(45). To date, few ISGs have been identified that restrict SARS-CoV replication [IFITMs (33) and GILT (46)] and SARS-CoV-2 replication [LY6E (20,21), IFITMs (21,22), and ZAP (47)]. Here, we present evidence that IFN-inducible CH25H and its natural product 25HC restrict S-mediated membrane fusion and block SARS-CoV-2 entry into host cells. 25HC has shown broad antiviral activity against a wide range of enveloped viruses (26,28,30,48) and nonenveloped viruses, such as reovirus (49) and murine norovirus (50). 25HC treatment did not inhibit the infectivity of rhesus rotavirus RRV strain (Fig. 1D), in contrast to the inhibition previously seen with the human rotavirus Wa strain (51), possibly because they differ in the endosomal compartment from which they exit into the cytosol (52). There seem to be two modes of inhibitory mechanisms involved. One requires a high micromolar concentration and more than 6 h of preincubation time to be effective, in the case of reovirus (49), pseudorabies virus (53), and human papillomavirus-16 (54), suggesting an indirect metabolic/cellular pathway-mediated mechanism, whereas the other-which includes influenza A virus (26), Lassa fever virus (55), hepatitis C virus (56), and SARS-CoV-2 ( Fig. 2)-functions at a low-micromolar/high-nanomolar range.
Combined with the recent report that apilimod, a PIKfyve kinase inhibitor, effectively inhibits SARS-CoV-2 infection (57), we confirm that this virus reaches late endosomal compartment for membrane fusion and access to the cytosol, at least in ACE2 + TMPRSS2 − cells. Recently, two independent studies also reported the anti-SARS-CoV-2 activity of 25HC (58,59). In the second study, Wang et al. (59) observed depletion of cholesterol at the plasma membrane by 25HC treatment, which is consistent with our findings of SARS-CoV-2 S protein-mediated fusion blockade (Fig. 3 C and E). Taken together, our data suggest a unifying model in which 25HC results in a redistribution of cholesterol and inhibits both endosomal entry and plasma membrane fusion, which potentially explains the CH25H inhibitory activity against wild-type SARS-CoV-2 in both TMPRSS2 − and TMPRSS2 + cells (Fig. 1 C and E and SI Appendix, Fig. S3 A and B).
One point worth noting is that wild-type VSV is sensitive to 25HC antiviral effect (Fig. 1C), while its entry is not dependent on NPC1 (60). Therefore, in the case of SARS-CoV-2, it is plausible that the reduced virus infection in the absence of NPC1 is uncoupled from 25HC antiviral mechanism of action. The latter is likely multifactorial and involves other cholesterol-binding proteins.
Another question that warrants further investigation is whether 25HC interferes with other steps in the SARS-CoV-2 replication cycle. While we observed a relatively weak inhibition of infection at early time points, such as 6 hpi ( Fig. 1C and SI Appendix, Figs. S3A and S4A), the antiviral effect is magnified by 24 hpi (Figs. 1 D and E, 2C, and 4F). It will be of interest to examine whether this reflects an antagonism of viral particle maturation or the release or infectivity of virus progeny. Our data instruct potential drug combinations of 25HC in conjunction with those targeting the cytoplasmic steps of the SARS-CoV-2 replication cycle, such as its main protease (61,62) or polymerase (63). Further in vivo studies in animal models of SARS-CoV-2 infection and pathogenesis are required to establish the physiological impact of 25HC-based drugs or compounds that modulate antiviral activities.
Identification and Quantification of Oxysterols. Oxysterols were extracted from pellets (3 × To extract secreted oxysterols from the culture medium, medium (1 mL) was added drop-wise, with sonication in an ultrasonic bath (5 min), to 2.3 mL of absolute ethanol containing the same amount of internal standards, as above, to give a solution of 3.3 mL of 70% ethanol. The oxysterols were then separated from cholesterol, derivatized with Girard P reagent and analyzed by LC-MS(MS n ) on a Orbitrap Elite, as described previously (70). Identification was performed by comparing retention time, exact mass, and MS n spectra to authentic standards and quantification was performed by stable isotope dilution using reconstructed ion-chromatograms generated in the Orbitrap.
Cardiomyocyte Infection. Human induced pluripotent stem cell-derived cardiomyocytes (R1117) were purchased from Cellular Dynamics International and seeded onto a 24-well plate coated with 0.1% gelatin (07903, Stemcell Technologies) at 1.36 × 10 5 cells per well, following manufacturer's recommendations. The cells were seeded in culture (37°C, 5% CO 2 ) for 48 h in plating medium and then changed to maintenance media, which was replaced every 2 d. VSV-SARS-CoV-2 infections were performed at day 6 after cell seeding.
RNA Extraction and Quantitative PCR. Total RNA was extracted from cells using RNeasy Mini kit (Qiagen) and reverse transcription was performed with High Capacity RT kit and random hexamers, as previously described (73). Quantitative PCR was performed using the AriaMX (Agilent) with a 25-μL reaction, composed of 50 ng of cDNA, 12.5 μL of Power SYBR Green master mix or Taqman master mix (Applied Biosystems), and 200 nM both forward and reverse primers. All SYBR Green primers and Taqman probes used in this study are listed in SI Appendix, Table S1.
Brightfield and Immunofluorescence Microscopy. For brightfield and epifluorescence, cultured cells were imaged by REVOLVE4 microscope (ECHO) with a 10× objective. For confocal microscopy, samples in eight-well chamber slides were fixed in 4% paraformaldehyde for 10 min at room temperature and stained as previously described (74). Cells were permeabilized and stained with antibodies against DAPI (P36962, Thermo Fisher), LAMP1 (9091S, Cell Signaling), LBPA (MABT837, Sigma), and Rab4 (ab13252, Abcam) at room temperature for 1 h. For filipin staining, cells were fixed and incubated with 50 μg/mL filipin (SAE0087, Sigma-Aldrich) at 37°C for 30 min. Stained cells were washed with PBS, whole-mounted with Antifade Mountant, and imaged with a Zeiss LSM880 Confocal Microscope at the Molecular Microbiology imaging core facility at Washington University in St. Louis. Images were visualized by Volocity v6.3 and quantification was determined by ImageJ (NIH).
Western Blotting. Cell lysates were harvest in RIPA buffer supplemented with protease inhibitor mixture and phosphatase inhibitor. Proteins were resolved in SDS/PAGE and analyzed by antibody as described previously (75) using the following antibodies and dilutions: ACE2 (MAB933, R&D Systems), GAPDH (631402, BioLegend), GFP (2555S, Cell Signaling), NPC1 (ab124801, Abcam), SARS-CoV-2 S1 (40590-T62, Sino Biological), SARS-CoV-2 S2 (40592-T62, Sino Biological), and V5 (13202S, Cell Signaling). Secondary antibodies were antirabbit (7074, Cell Signaling) or anti-mouse (7076, Cell Signaling) IgG horseradish peroxidase-linked antibodies. Protein bands were visualized with Clarity ECL substrate and a Bio-Rad Gel Doc XR system. Statistical Analysis. All bar graphs were displayed as means ± SEM. Statistical significance in the data in Figs. 1 E and F, 2C, 3B, and 4E and SI Appendix, Figs. S1 C and D, S2B, and S5B was calculated by Student's t test using Prism 8.4.2 (GraphPad). Statistical significance in the data in Figs. 1 B and D, 2 D, F-H, 3 C-E, and 4 A and B and SI Appendix, Figs. S1G, S2E, S3 B and C, S4A, and S5E was calculated by pairwise ANOVA using Prism 8. Nonlinear regression (curve fit) was performed to calculate EC 50 and CC 50 values for Figs. 2B and 4 D and F and SI Appendix, Fig. S4E. All data are presented as *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001. All experiments other than Fig. 1A were repeated at least three times. Fig. 1A was performed twice with average numbers indicated on the graph. The raw data are included in Dataset S1.
Data Availability. All study data are included in the article and supporting information.