Potentially repurposable drugs for COVID-19 identified from SARS-CoV-2 Host Protein Interactome

We previously presented the protein-protein interaction network - the ‘HoP’ or the host protein interactome - of 332 host proteins that were identified to interact with 27 nCoV19 viral proteins by Gordon et al. Here, we studied drugs targeting the proteins in this interactome to identify whether any of them may potentially be repurposable against SARS-CoV-2. We studied each of the drugs using the BaseSpace Correlation Engine and identified those that induce gene expression profiles negatively correlated with SARS-associated expression profile. This analysis resulted in 20 drugs whose differential gene expression (drug versus normal) had an anti-correlation with differential expression for SARS (viral infection versus normal). These included drugs that were already being tested for their clinical activity against SARS-CoV-2, those with proven activity against SARS-CoV/MERS-CoV, broad-spectrum antiviral drugs, and those identified/prioritized by other computational re-purposing studies. In summary, our integrated computational analysis of the HoP interactome in conjunction with drug-induced transcriptomic data resulted in drugs that may be repurposable for COVID-19.


Introduction
COVID-19 (Coronavirus Disease 2019) is an infectious virus outbreak which rapidly developed into a pandemic health crisis. The novel coronavirus (SARS-CoV-2/nCoV-19) is the causative agent of this disease. 1 Several groups responded to the urgent need for effective therapeutics by leading systems-level efforts to identify drugs repurposable for COVID-19, through the lens of the virus-host protein interactome, 2 and the interactomes of SARS-CoV-2-modulated host proteins 3 and host proteins modulated by other human corona viruses such as SARS-CoV,  Repurposing or nding alternate uses for approved drugs has proved to be a better strategy than de novo identi cation in terms of time and cost effectiveness. [5][6][7] Discovery of therapeutic agents for infectious diseases in the past was largely serendipitous, and focused on screening and prioritizing drugs that target the viral system. 8 Over the last few years, the focus has shifted towards computationally identifying drugs that could counter the virus attack A primary strategy is to repurpose drugs with the ability to revert the genes differentially expressed in the host upon viral infection to their normal levels; i.e. to revert the host transcriptional pro le induced upon viral infection to its normal state. 8 This "inverse genomic signature approach" involves identifying drugs that induce gene expression pro les negatively correlated with host-speci c gene signatures induced by viral infection, and has been used to select candidates repurposable against in uenza viruses and MERS-CoV. 9-11 Availability of disease-associated and drug-induced transcriptomic pro les in online repositories such as NCBI GEO (Gene Expression Omnibus) and CMAP (Connectivity Map), 12,13 allow these pro les to be compared using bioinformatics data analysis software suites such as the BaseSpace Correlation Engine. 14 Changes in the host transcriptome induced by viral infection are also re ected in the host proteome, speci cally as perturbations in the interaction networks of the host proteins. This complex Page 3/12 network of protein-protein interactions (PPIs) called the 'interactome' has the potential to restrict viral replication in host cells, or conversely to be taken over by the virus for its replication.
We had previously presented the Host Protein Interactome (HoP Interactome) of 332 human proteins identi ed to interact with 27 SARS-CoV-2 viral proteins by Gordon et al. 2,15 . This interactome, consisting of 6,076 PPIs of the host proteins including 1,941 novel interactions predicted by HiPPIP, provided an integrated view on how host genes in various high throughput COVID-19 and SARS transcriptomic/proteomic studies are functionally linked. 15 In this study, we identi ed drugs targeting the proteins in this interactome, and studied the correlation of the gene expression pro les induced by these drugs in various cell lines, with SARS/COVID-associated pro les observed in lung-derived (MRC5, Calu-3, NHBE and A549) cell lines, and in peripheral blood mononuclear cells (PBMCs) of SARS patients. Our work differs from previous efforts to identify drugs repurposable for COVID-19 2-4 in that it considers the host protein interactome, and includes computationally predicted novel interactors of the host proteins, which may lead to identi cation of drugs that were hitherto not prioritized.

Potentially Repurposable Drugs
We compared drug-induced versus SARS-associated differential expression using the BaseSpace Correlation Engine (previously called NextBio) (https://www.nextbio.com), 16,17 to identify drugs for nCoV19. We compiled a list of 933 chemical compounds whose differential gene expression pro le (drug versus no drug) were negatively correlated with at least one of the four SARS differential gene expression datasets (infected versus non-infected); the 4 SARS datasets we studied were: Calu-3 epithelial cells infected for 48 hours with SARS coronavirus versus mock infected cells (GSE17400), Calu-3 lung cells infected for 72 hours with SARS CoV Urbani versus mock infected cells (GSE37827), lung broblast MRC5 cells 24 hours post SARS coronavirus infection, high multiplicity of infection MOI versus mock infection (GSE56189) and peripheral blood mononuclear cells (PBMCs) from patients with SARS versus healthy subjects (GSE1739 18 ). We also compiled a list of 381 chemical compounds with gene expression pro les negatively correlated with the pro le induced in human bronchial epithelial (NHBE) and lung cancer (A549) cells infected with the SARS-CoV-2 strain USA-WA1/2020 (GSE147507 19 ) Although in each case, there would be some genes that are differentially expressed in the same direction for both the drug and the disease (i.e., both cause some genes to overexpress, or both cause some genes to under express), the overall effect on the entire transcriptome would be an anti-correlation. A correlation score is generated by NextBio based on the strength of the overlap between the drug and disease datasets.
Statistical criteria such as correction for multiple hypothesis testing are applied and the correlated datasets are then ranked by statistical signi cance. A numerical score of 100 is assigned to the most signi cant result, and the scores of the other results are normalized with respect to this top-ranked result.
Next, we identi ed 1,130 drugs that target at least one protein in the HoP interactome using WebGestalt. 20 We used the 'redundancy reduction' feature provided by WebGestalt to prioritize drugs with highly signi cant overlaps with the interactome, while also capturing all the unique target gene sets. This feature used an a nity propagation algorithm which clusters sets of genes in the interactome targeted by speci c drugs using Jaccard index as the similarity metric and identi es a 'representative' for each cluster (one drug and its targets), having the most signi cant p-value among all the gene sets in that cluster. This resulted in 209 drugs for further consideration. Given a class of drugs targeting the same set of proteins, this method ensures that only those individual drugs that target a statistically signi cant number of proteins in the interactome are prioritized for further analysis.
Fifty-six drugs were found in common to the above two analyses, i.e. these drugs not only targeted genes in the HoP interactome, but also induced gene expression pro les which are negatively correlated with that induced by SARS-CoV (Supplementary Table S1) and SARS-CoV-2 (Supplementary Table S2

Discussion
An integrated computational analysis of the interactome in conjunction with drug-induced transcriptomic data revealed 24 drugs that may be repurposable for COVID-19. These included drugs with proven in vitro activity against SARS-CoV-2, those that were already being tested for their clinical activity against SARS-CoV-2, those with proven activity against SARS-CoV/MERS-CoV, broad-spectrum antiviral drugs, and those identi ed/prioritized by other computational re-purposing studies.
Information from literature supporting the shortlisted drugs Cyclosporine, sorafenib, tamoxifen and anisomycin were identi ed to have inhibitory effects on SARS-CoV-2 in four independent cell-based assays. 22-25 These drugs have also been shown to be effective against viruses similar to SARS-CoV-2 or other viruses. Cyclosporine in combination with interferon alpha reduced MERS-CoV replication and this reduction was associated with greater induction of interferon stimulated genes. 26 Sorafenib has been shown to suppress the gene expression of HBV (hepatitis B virus) by inhibiting the JNK pathway, which constitutes FXR, a transcription factor that promotes HBV replication and gene expression. 27 Tamoxifen has shown inhibitory effect against the vesicular stomatitis virus, whose effect in Vero cells has been correlated with activation of macrophages and an elevated interferon-I response. 28 Anisomycin reduced the viral load of Zika virus in the blood of AG129 mice. 29 Testing of nitric oxide gas inhalation therapy is already underway in COVID-19 patients (see www.ClinicalTrials.gov for ongoing clinical trials for nitric oxide). Nitric oxide is usually produced by phagocytes in response to interferon-γ. However, it is also rapidly produced in primary broblasts in response to viral dsRNA, especially in the absence of an intact interferon system, which has been noted in the case of nCoV19 infection. 30 Resveratrol has shown anti-viral activity in MERS-infected Vero E6 (kidney epithelial) cells. 31 It inhibited viral infection, prolonged the survival of the host cell after infection by downregulating virus-induced apoptosis and reduced the expression of the viral nucleocapsid protein, which is essential for viral replication. 31 The mTOR inhibitor sirolimus reduced MERS-CoV infection by 60% in Huh7, a hepatocyte-derived cell line. 32 Mycophenolic acid has been shown to inhibit papain-like protease of MERS-CoV. 33,34 A signi cant lack of IFN-I and IFN-III (type I and III interferons) expression was noted in nCoV19 infected human bronchial epithelial and lung alveolar carcinoma cells. 19 Prioritizing recombinant interferons (that are exogenously produced under laboratory conditions) as potential therapeutic options for COVID-19 may be essential in this scenario. Interferon-alpha2b in combination with ribavirin reduces viral replication, regulates the host response and improves clinical outcome in rhesus macaques infected with MERS-CoV. 35 SARS-CoV infection in human bronchial epithelial Calu-3 cells has been shown to be inhibited by interferon alfacon-1. 36 Ramipril is an inhibitor of the angiotensin converting enzyme (ACE1) which belongs to the class of entry receptors targeted by nCoV19. Current evidence shows that the spike protein of SARS-CoV-2 binds to ACE2 and not ACE1. 37 Even though the two enzymes are coded by homologous genes, ACE1 inhibitors are incapable of acting on ACE2 and the physiological actions of ACE1 differ from that of ACE2 (vasoconstriction versus vasodilation). However, since these enzymes function in two counterbalancing arms of the renin-angiotensin system, more investigations are necessary to ascertain whether ACE1 has some (as yet) unidenti ed role in viral pathogenesis. 38 Quercetin inhibits the infection of various in uenza viruses including H1N1, H3N2 and H5N1, and it is suspected that this inhibition arises from their interaction with the viral HA2 (hemagglutinin) subunit. 39 The calcium channel blocker verapamil inhibits the in uenza virus late in their replication cycle, in Madin-Darby canine kidney cells and in murine pulmonary macrophages. 40 Progesterone has been shown to promote faster recovery of female mice from in uenza A virus infection by reducing pulmonary in ammation, repairing damaged lung epithelium and generally improving lung function. 41 Clotrimazole inhibits the entry of recombinant vesicular stomatitis virus pseudotypes (arenavirus) into A549 human lung epithelial cells by targeting the membrane fusion mechanism of the virus. 42 Didanosine is a nucleoside analogue and a reverse transcriptase inhibitor that is used to treat HIV/AIDS. In the host cell, didanosine is converted to dideoxyadenosine-5'-triphosphate (ddATP), whose incorporation into the viral DNA terminates DNA chain elongation, by preventing the formation of the 5' to In this study, we shortlisted drugs potentially repurposable for COVID-19 based on the negative correlation of drug-induced versus disease-associated gene expression pro les. Although this approach has been validated in the past, it has several limitations. Drug associated expression pro les analyzed in this study were induced in several types of cell lines (including cancer cell lines) that may not be directly relevant to COVID-19 or SARS-CoV-2 infection. The effect of the proposed repurposable drugs should be studied in human bronchial epithelial cells and/or in human lung cancer cell lines, both of which were recently used to study host transcriptional response upon SARS-CoV-2 infection. 19 In summary, we showed that drugs repurposable for COVID-19 can be identi ed from the host protein interactome based on gene expression pro les induced by the drug versus those associated with the disease. The dissemination of this list of repurposable drugs to the scienti c community will enable clinical translation of these results.

Methods
The list of chemical compounds whose gene expression pro les correlated negatively with four SARS datatsets and one COVID-19 dataset were compiled using the BaseSpace correlation software (https://www.nextbio.com) following the protocol described in prior work 17 (List 1). The datasets considered were human bronchial epithelial (NHBE) and lung cancer (A549) cells infected with the SARS-CoV-2 strain USA-WA1/2020 (GSE147507 19 ), Calu-3 epithelial cells infected for 48 hours with SARS coronavirus versus mock infected cells (GSE17400), Calu-3 lung cells infected for 72 hours with SARS CoV Urbani versus mock infected cells (GSE37827), lung broblast MRC5 cells 24 hours post SARS coronavirus infection, high MOI (3) versus mock infection (GSE56189) and peripheral blood mononuclear cells (PBMCs) from patients with SARS versus healthy subjects (GSE1739 18 ). Next, we identi ed drugs that targeted at least one gene in in the HoP interactome using WebGestalt. 20 After employing the 'redundancy reduction' feature in WebGestalt to reduce the search space of drugs, we were left with a fewer number of drugs (List 2). In this feature, an a nity propagation algorithm clusters gene sets in the interactome targeted by speci c drugs using Jaccard index as the similarity metric, and identi es a 'representative' for each cluster (one drug and its targets), having the most signi cant p-value among all the gene sets in that cluster. We then compared list 1 and list 2 to identify the drugs that not only target proteins in the interactome but are also negatively correlated with SARS/COVID-19.
List of drugs validated to be effective against SARS-CoV-2 in cell-based assays were obtained from the COVID-19 Gene and Drug Set Library (https://amp.pharm.mssm.edu/covid19/). 53 The drug-protein interactome gure was created using Cytoscape. 54

Declarations Data Availability
The list of drugs negatively correlated with SARS and COVID-19 are available as Supplementary Table  S1 and Supplementary Table S2. Author Contributions KBK has carried out the analyses with supervision from NB and MKG. MKG constructed the protein interactome. Manuscript has been drafted by KBK and edited and approved by all authors.

Declarations
Competing interests: The authors declare no competing interests.