An ACAT inhibitor regulates SARS-CoV-2 replication and antiviral T cell activity

The severity of disease following infection with SARS-CoV-2 is determined by viral replication kinetics and host immunity, with early T cell responses and/or suppression of viraemia driving a favourable outcome. Recent studies have uncovered a role for cholesterol metabolism in the SARS-CoV-2 life cycle and in T cell function. Here we show that blockade of the enzyme Acyl-CoA:cholesterol acyltransferase (ACAT) with Avasimibe inhibits SARS-CoV-2 entry and fusion independent of transmembrane protease serine 2 expression in multiple cell types. We also demonstrate a role for ACAT in regulating SARS-CoV-2 RNA replication in primary bronchial epithelial cells. Furthermore, Avasimibe boosts the expansion of functional SARS-CoV-2-specific T cells from the blood of patients sampled in the acute phase of infection. Thus, re-purposing of available ACAT inhibitors provides a compelling therapeutic strategy for the treatment of COVID-19 to achieve both antiviral and immunomodulatory effects.


Introduction 42 43
SARS-CoV-2 is a global health issue associated with over 400 million infections and 6 million deaths 44 (WHO, 2022). Preventive vaccines have reduced morbidity and mortality (Gupta et al., 2021b;Sheikh 45 et al., 2021); however, therapeutic strategies for unvaccinated subjects or those with breakthrough 46 infections are needed. Several direct-acting antiviral drugs are now licensed for the treatment of 47 8 membrane peptides, boosting responses in some patients and inducing de novo responses in others 140 ( Fig.3a-c, Supp.Fig.2b,c). The response to AVS was heterogeneous, showing a 50-fold increase in 141 the magnitude of IFNγ-producing T cells in one patient and decreased cytokine production in a 142 minority of patients, as previously reported for other in vitro and in vivo immunotherapeutic 143 approaches (Bengsch et al., 2014;Maini and Pallett, 2018). A similar enhancement was seen for 144 cytokine-producing CD8 + T cells in individual donors but was less consistent than for CD4 + T cells, 145 resulting in no overall significant changes for CD8 + T cell responses across the cohort (Supp. Fig.2d-146 f). CD4 + T cells provide help to activate and differentiate B cells, for example via the interaction of 147 CD40 and CD40L (CD154). AVS increased the SARS-CoV-2-specific expression of CD154 (CD40L) 148 on CD4 + T cells, reflecting an enhanced capacity to co-stimulate CD40 and to activate B cells 149 therapies for viral infections carry the risk of increasing bystander immune responses and cytotoxic 152 tissue damage; however, we did not detect any significant increase of CD107a mobilization to the cell 153 membrane of perforin-producing T cells, markers of degranulation and cytotoxicity respectively 154 (Supp. Fig.2g). COVID-19 severity is associated with male sex (Scully et al., 2020) and increased age 155 (Richardson et al., 2020). We noted that AVS enhancement of SARS-CoV-2-specific T cell responses 156 was seen in both males and females and was independent of age (Supp. Fig.2h), showing the 157 potential of this therapeutic approach for a variety of patients, including those at risk of severe 158

infection. 159
To ascertain whether AVS only boosts virus-specific effector and early memory T cells during or 160 shortly after infection but not memory T cells, we recruited a second cohort of unvaccinated donors 6 161 months after SARS-CoV-2 infection (memory cohort, see methods section). AVS had no consistent 162 effect on SARS-CoV-2-specific memory CD4 + or CD8 + T cell responses 6 months post-infection 163 (Supp. Fig.3a-d). This is in line with our previous findings showing that ACAT inhibition preferentially 164 rescues PD-1 hi T cells and not memory responses to cytomegalovirus (Schmidt et al., 2021). AVS has 165 shown a good safety profile in phase III atherosclerosis studies (Llaverías et al., 2003) and has not 166 been associated with autoimmune responses in murine models (Yang et al., 2016). In line with this, 167 we did not detect any non-specific increase in cytokine production when T cells from the acute cohort 168 were treated with AVS without viral peptides (Supp. Fig.3e). Thus, our data support AVS selectively 9 expanding acutely activated SARS-CoV-2-specific T cells, without affecting memory or non-activated 170 T cells. 171 172

174
This study raises a number of areas for future investigation. AVS inhibition of SARS-CoV-2 fusion in 175 VeroE6 and VeroE6-TMPRSS2 is consistent with ACAT regulating cholesterol levels at both the cell 176 surface and within endosomes, highlighting the need to better understand the role of cholesterol in 177 endosomal pathways that are essential in virus internalization and egress (Glitscher and Hildt, 2021). 178 Our observation that AVS inhibited VSV-G pseudoparticle entry suggests a potential role in regulating 179 the entry of other viruses that would be worth investigating. Cholesterol 25-hydroxylase catalyzes the 180 formation of 25-hydroxycholesterol (25HC) from cholesterol and leads to a redistribution of cholesterol 181 limiting the entry of a range of enveloped viruses (Schoggins, 2019) including SARS-CoV-2 (Wang et 182 al., 2020;Zang et al., 2020;Zu et al., 2020). Wang et al reported that 25HC activated ACAT and 183 suggested this as a mechanism to explain 25HC inhibition of SARS-CoV-2 entry. The authors showed 184 that inhibition of ACAT with SZ58-035 partially reversed the antiviral activity of 25HC in Calu-3 cells; 185 however, they observed a negligible effect on basal plasma membrane cholesterol levels or on 186 SARS-CoV-2 pseudoparticle entry. This contrasts with our results and may reflect variable efficacy of 187 SZ58-035 and AVS to modulate cholesterol levels. Our observation that AVS inhibits SARS-CoV-2 188 pseudoparticle infection of a range of cell lines and primary epithelial cells shows its robust antiviral 189

activity. 190
We focused on T cells specific for two of the key structural proteins targeted in acute infection (Peng 191 et al., 2020) and further studies to assess the effect of AVS on other T cell specificities including those 192 against non-structural viral proteins associated with abortive infection would be of interest (Swadling 193 et al., 2021). The potential for AVS to boost acutely activated CD4 + T effector and helper function 194 even in the elderly, suggests they could be tested for their capacity to adjuvant sub-optimal vaccine 195 responses in this vulnerable group (Collier et al., 2021) or others with waning immunity. The lack of T 196 cell boosting in the memory phase is in line with our previous findings (Schmidt et al., 2021) but 197 conceivably could also be related to the younger age of this cohort. 198 We have shown increased antiviral activity following treatment of circulating T cells; however immune 199 responses at the site of disease, the lung and upper respiratory tract, are shaped by the local 200 microenvironment and nutrient availability. The lung is enriched in cholesterol compared to blood 201 (Chamberlain, 1928) with cholesterol constituting the main neutral lipid in surfactant (Keating et al., 202 2007). We previously reported that ACAT inhibition is enhanced in the presence of high cholesterol. T 203 cells isolated from cholesterol-rich liver and tumour tissues were boosted to a greater extent than 204 those from the blood of the same donors (Schmidt et al., 2021); suggesting a similar enhancement 205 may be seen following ACAT inhibition of SARS-CoV-2-specific T cells infiltrating the infected lung. 206 Further studies to address the effect of AVS on other immune cell subsets associated with the 207 inflammatory response in severe and long COVID-19 would also be of interest. 208 Urgent consideration should be given to trials testing the efficacy of re-purposing ACAT inhibitors like 209 AVS, an oral agent that has been shown to have a good safely profile. We show it has the capacity to 210 exert a unique dual effect, directly inhibiting SARS-CoV-2 entry and RNA replication as well as 211 boosting the acute T cell response that can aid viral elimination and provide protection against re-212 infection. 213

Methods 214
Ethics 215 The COVIDsortium cohort was approved by the ethical committee of UK National Research Ethics 216 Service (20/SC/0149) and registered at https://ClinicalTrials.gov (NCT04318314). The Royal Free 217 Biobank (TapB) was approved by the Wales Research Ethics Committee (16/WA/0289; 21/WA/0388; 218 project approval reference: NC2020.11). The PBEC study was reviewed by the Oxford Research 219 Ethics Committee B (18/SC/0361). All study participants gave written informed consent prior to 220 inclusion in the study and all storage of samples obtained complied with the Human Tissue Act 2004. 221

SARS-CoV-2 propagation and infection 283
Naïve VeroE6 cells were infected with SARS-CoV-2 at an MOI of 0.003 and incubated for 48-72h until 284 visible cytopathic effect was observed. At this point, cultures were harvested, clarified by 285 centrifugation to remove residual cell debris and stored at -80°C. Viral titre was determined by plaque 286 assay. Briefly, VeroE6 cells were inoculated with serial dilutions of SARS-CoV-2 viral stocks for 2h 287 followed by addition of a semi-solid overlay consisting of 1.5% carboxymethyl cellulose (Sigma-288 Aldrich). Cells were incubated for 72h, visible plaques enumerated by fixing cells using amido black 289 stain and plaque-forming units (PFU) per mL calculated. For infection of Calu-3 cells with SARS-CoV-290 2, cells were plated 24h before infection with the stated MOI. Cells were inoculated for 2h after which 291 the residual inoculum was removed with three PBS washes. Unless otherwise stated, infected cells 292 were maintained for 24h before harvesting for downstream applications. VeroE6 cells were grown to 30% confluence in EMEM+10%FBS. Cells were incubated with fresh 317 EMEM+10%FBS for 1h followed by 1h of incubation in 100μL EMEM+10%FBS with 5μM AVS or 318 equivalent concentrations of DMSO. Cells were rinsed with PBS and then fixed with 3% 319 paraformaldehyde and 0.1% glutaraldehyde for 15min to fix both proteins and lipids. Fixative 320 chemicals were reduced by incubating with 0.1% NaBH4 for 7min with shaking followed by three 321 times 10min washes with PBS. Cells were permeabilized with 0.2% Triton X-100 for 15min and then 322 blocked with a standard blocking buffer (10% bovine serum albumin (BSA) / 0.05% Triton in PBS) for 323 90min at room temperature. For labelling, cells were incubated with Alexa Fluor 647-CTB (Sigma-324 Aldrich) for 60min in 5% BSA / 0.05% Triton / PBS at room temperature followed by 5 washes with 325 1% BSA / 0.05% Triton / PBS for 15min each. Cells were then washed with PBS for 5min. Cell 326 labelling and washing steps were performed while shaking. Labelled cells were then post-fixed with 327 fixing solution, as above, for 10min without shaking followed by three 5min washes with PBS and two 328 3min washes with deionized distilled water. 329 Images were recorded with a Bruker Vutara 352 with a 60X Olympus Silicone objective. Frames with 330 an exposure time of 20ms were collected for each acquisition. Excitation of the Alexa Fluor 647 dye 331 was achieved using 640nm lasers and Cy3B was achieved using 561nm lasers. Laser power was set 332 to provide isolated blinking of individual fluorophores. Cells were imaged in a photo-switching buffer 333 comprising of 1% β-mercaptoethanol (Sigma-Aldrich), oxygen scavengers (glucose oxidase and 334 catalase; (Sigma-Aldrich) in 50mM Tris (Affymetrix) + 10mM NaCl (Sigma-Aldrich) + 10% glucose 335 (Sigma) at pH 8.0. Axial sample drift was corrected during acquisition through the Vutara 352's 336 vFocus system. Images were constructed using the default modules in the Zen software. Each 337 detected event was fitted to a 2D Gaussian distribution to determine the centre of each point spread 338 function plus the localization precision. The Zen software also has many rendering options including 339 removing localization errors and outliers based on brightness and size of fluorescent signals. Pair 340 correlation and cluster analysis was performed using the Statistical Analysis package in the Vutara 341 SRX software. Pair Correlation analysis is a statistical method used to determine the strength of 342 correlation between two objects by counting the number of points of probe 2 within a certain donut-343 radius of each point of probe 1. This allows for localization to be determined without overlapping 344 pixels as done in traditional diffraction-limited microscopy. Cluster size estimation and cluster density 345 were calculated through cluster analysis by measuring the length and density of the clusters 346 comprising of more than 10 particles with a maximum particle distance of 0.1μm. 347

cells in acute infection
(a-e) Human PBMC from donors with acute SARS-CoV-2 infection were stimulated with SARS-CoV-2 peptide pools (Spike and Membrane, Mem) and treated with Avasimibe (AVS) or DMSO for 8d. SARS-CoV-2-specific cytokine production by CD4 + T cells was detected via flow cytometry. The cytokine production/CD154 expression in wells without peptide stimulation was subtracted to determine SARS-CoV-2-specific cytokine production/CD154 expression in summary data. Example plots and summary data for SARS-CoV-2 specific IFNγ (a), TNF (b), MIP1β (c) production and CD154 expression (d) by CD4 + T cells (n=19). (e) Assessment of SARS-CoV-2-specific proliferation determined by CFSE dilution gated on IFNγ + CD4 + T cells (Spike n=10; Mem n=11). Bars mean. Doughnut charts indicate fraction of donors with response to AVS (red). Response defined as de novo or increased cytokine production/CD154 expression. P values determined by Wilcoxon matched-pairs signed rank test.