key: cord-0942800-1ebzorvy
authors: Krishnamoorthy, Pandikannan; Raj, Athira S.; Roy, Swagnik; Kumar, Nachimuthu Senthil; Kumar, Himanshu
title: Comparative transcriptome analysis of SARS-CoV, MERS-CoV, and SARS-CoV-2 to identify potential pathways for drug repurposing
date: 2020-11-24
journal: Comput Biol Med
DOI: 10.1016/j.compbiomed.2020.104123
sha: dba2601fec12a6f38925755a6de83cdf5486ef93
doc_id: 942800
cord_uid: 1ebzorvy

The ongoing COVID-19 pandemic caused by the coronavirus, SARS-CoV-2, has already caused in excess of 1.25 million deaths worldwide, and the number is increasing. Knowledge of the host transcriptional response against this virus and how the pathways are activated or suppressed compared to other human coronaviruses (SARS-CoV, MERS-CoV) that caused outbreaks previously can help in the identification of potential drugs for the treatment of COVID-19. Hence, we used time point meta-analysis to investigate available SARS-CoV and MERS-CoV in-vitro transcriptome datasets in order to identify the significant genes and pathways that are dysregulated at each time point. The subsequent over-representation analysis (ORA) revealed that several pathways are significantly dysregulated at each time point after both SARS-CoV and MERS-CoV infection. We also performed gene set enrichment analyses of SARS-CoV and MERS-CoV with that of SARS-CoV-2 at the same time point and cell line, the results of which revealed that common pathways are activated and suppressed in all three coronaviruses. Furthermore, an analysis of an in-vivo transcriptomic dataset of COVID-19 patients showed that similar pathways are enriched to those identified in the earlier analyses. Based on these findings, a drug repurposing analysis was performed to identify potential drug candidates for combating COVID-19.

receptor. 15 Nevertheless, all three viral infections have similar clinical features, including fever, 85 dry cough, and dyspnea. In addition, patients exhibit similar chest radiograph abnormalities, 86 lowered total lymphocytes, increased lactate dehydrogenase, prolonged prothrombin rate, and 87 cytokine storms. [16] [17] [18] [19] Also, early reports revealed that in SARS patients, there is a pulmonary 88 infection and severe lung damage associated with elevated pro-inflammatory cytokines in serum 89 (IL-1β, TNF-α, IL6, IL-8, IL-12, IFN-γ, IP10, MCP1, and many others). 20 Meanwhile, other 90 studies showed that MERS-CoV infection elicits type-I and type-II interferon responses and Th1 91 and Th17 cytokine profiles (IFN-γ, TNF-α, IL-15, and IL-17). 21 A similar cytokine profile has 92 also been reported in COVID-19 patients, where there is increased IL1β, IFN-γ, IP10, and 93 MCP1, probably culminating in activated Th1 responses. These reports suggest that infection 94 followed by a cytokine storm might be the probable reason for disease severity. Moreover, 95 comorbid factors such as hypertension and diabetes are also shared by all three of these 96

infections. 22, 23 Hence, it is essential to find out whether unique or similar pathways are activated 97 or suppressed by SARS-CoV and MERS-CoV in order to relate and compare these findings to 98 those on the current pandemic-causing coronavirus, SARS-CoV-2. As the race for vaccines and 99 therapeutics intensifies, it has become even more important to identify the underlying molecular 100 mechanisms behind SARS-CoV-2 infection in order to design potential drugs or vaccine 101

Transcriptome meta-analysis is a powerful tool for identifying the underlying genes or signaling 103 pathways as well as discovering the biomarkers for various diseases. Due to the increasing 104 availability of microarrays and transcriptome high-throughput sequencing data, it is expected that 105 SARS-CoV-2 infected transcriptome signature provides a ray of hope on the already approved 116 list of potential drugs that can be used for the treatment of In this study, first, we explored the host transcriptome responses upon SARS-CoV-2 infection 118 with those of the two previous outbreak-related pathogenic coronaviruses, SARS-CoV, and 119 MERS-CoV, through conducting a meta-analysis at different time points. The main rationale 120 behind this meta-analysis was to identify the significant genes and associated pathways that are 121 dysregulated at different time points of infection. As mentioned above, SARS-CoV and MERS-122

CoV share many common symptoms with SARS-CoV-2 infections. Hence, the transcriptomic 123 responses of these two viruses were compared with those of SARS-CoV-2 in order to identify 124 the common activated and suppressed pathways across all three of these contagious HCoV 125

infections. Next, we cross-validated our in-vitro data observations with the results of the pathway 126 analysis of in-vivo sequencing data derived from SARS-CoV-2 positive patients. This 127 comparison revealed many pathways to be dysregulated upon all the three HCoVs. Then, we 128 used a pipeline named cogena, which is based on the clustering of the coexpressed genes in the 129 disease state, to identify potential drugs that can be repurposed for COVID-19 treatment. Thus, 130 this study not only reveals potential drug candidates for COVID-19 treatment but also shows that 131 specific pathways are altered similarly by different pathogenic coronaviruses. This study, 132 therefore, provides valuable insights on the pathogenesis of human coronaviruses and novel drug 133 targets to tackle future coronavirus outbreaks. 134 J o u r n a l P r e -p r o o f

In order to collect the transcriptome data of SARS-CoV infection, we queried the gene 138 expression omnibus (GEO) database with the terms "SARS CoV," "SARS CoV AND Homo 139 sapiens". We screened the datasets in a way that the infection was performed in a human cell line 140 at different time points. We screened for the eligible datasets as per preferred reporting items for 141 systematic reviews and meta-analyses (PRISMA) guidelines ( Table S1 ). As of April 10, 2020, 6 142 datasets were eligible for further analysis, as per PRISMA guidelines. (Table 1) . 143

Using the GEOquery R package , the expression data of all these datasets were downloaded 29 . 144

All these studies had corresponding triplicate time-matched mock and infected samples for each 145 time point. The sample groups which were infected other than SARS-CoV were removed. We 146 segregated expression data of each time point of all the datasets, and meta-analysis was 147 performed time point-wise using the web tool NetworkAnalyst 30,31 . All the expression data were 148 log2 transformed, if not performed, followed by quantile normalization. All the probe Ids were 149 converted into their corresponding Entrez gene IDs for uniformity. In the case of multiple probes 150 mapping to the same gene, the mean of their expression values was considered. 151

Limma R package was used to find the differentially expressed genes between the mock and 152 SARS-CoV groups with a cutoff, adjusted p-value 0.05, and logFC 1.0 32 . The integrity of the 153 data was tested. The major challenge in the integration of multiple microarray data is the batch 154 effect that arose due to the different microarray platforms, cell lines, design of the study 33 . The 155 normalized datasets were subjected to the ComBat, an empirical Bayes method in-built in this 156 tool, NetworkAnalyst 34 . It reduced the study-specific batch effects and confounding factors due 157 to non-biological variation. The batch adjustment was visually examined through principal 158 component analysis (PCA) plots. 159

For combining different datasets for the increased statistical power, we combined the p-values 160 from each dataset using -Fisher's method (-2*∑Log(p)) with a significant level threshold of 161 0.05 35,36 . The procedure was repeated for all individual time points-12hr, 24hr, 36hr, 48hr, 60hr, 162 72hr, common across the selected studies. 163 J o u r n a l P r e -p r o o f For MERS-CoV transcriptome data, we queried the GEO database with the terms "MERS CoV", 165 "MERS CoV AND Homo sapiens". We screened the datasets in such a way that the infection 166 was performed in a human cell line at different time points, and screening was performed as per 167 PRISMA guidelines (Table S2) . 12 datasets were eligible for further analysis ( 

The significant differential expressed genes (DEGs) at each meta-analysis were subjected to 181 over-representation analysis, which includes the Kyoto Encyclopedia of Genes and Genomes 182 (KEGG)-pathway analysis and gene ontology enrichment analysis using the R package 183 clusterProfiler 37 with adjusted p-value cutoff less than 0.05. The relation between the DEGs at 184 each time points was visualized through circle plots, venn diagram, upset plots by using 185

Metascape 38 , and Intervene 39 tools. The dot plots were generated using clusterProfiler, and the 186 number of entries to be displayed were based on the maximum visibility and readability of the 187 plots. Geneset enrichment analysis (GSEA) 40 was performed using clusterProfiler by inputting 188 gene names along with the logFC values. The genes were sorted according to their log fold 189 change, and the number of permutations was set as 10000 with minimum gene set size 3, 190 maximum gene set size 800, and p-value cutoff = 0.05 to identify the significant pathways. The (hereinafter, ORA-KEGG pathway analysis) at a cutoff adjusted p-value of less than 0.05 and 254 their relations and differences were visualized through the use of a dot plot (Fig. 2G) , in which 255 the size of the dot represents the gene ratio of the particular pathway, and the color represents the 256 significance level based on the adjusted p-value. The immune-related pathways, such as the 257 tumor necrosis factor (TNF) signaling pathway, legionellosis pathway, and interleukin (IL-17) 258 signaling pathway, only became enriched at 12 hours. These results suggest that there are very 259 few transcriptomic changes at 12 hours of SARS-CoV infection, which is also evident from the 260 PCA plot (Fig. 2A) . The pathways related to the innate immune system such as the TNF 261 signaling pathway, RIG-I-like receptor signaling pathway, Janus kinase/signal transducers and The downregulated genes at each time point were subjected to an ORA-KEGG pathway analysis 280 at a cutoff adjusted p-value of less than 0.05, and their relations and differences were visualized 281 through the dot plot (Fig. 2I) . The results of the ORA showed that none of the pathways were metabolism, valine, leucine, and isoleucine metabolism pathways were significantly 284 downregulated after 24 hours until 72 hours. Most of the metabolic pathways, such as carbon 285 metabolism, oxidative phosphorylation, propoanate metabolism, fatty acid metabolism, and 286 steroid biosynthesis pathways, were significantly downregulated after 36 hours. In addition, 287 pathways such as glycolysis, glyoxylate and dicarboxylate metabolism, TCA cycle, and pentose 288 pathways were also significantly downregulated after 36 hours. These results suggest that most 289 of the metabolic pathways were downregulated upon SARS-CoV infection and that these 290 pathways may, therefore, play a crucial role in the pathogenesis of SARS-CoV infection. Thus, the meta-analysis identified previously known innate immune pathways as well as many 301 metabolic pathways that have not been well studied with respect to coronavirus infections. 302 303

Next, to gain further insights into the pathogenesis of fatal and contagious human coronaviruses 306 (HCoVs), we investigated the transcriptomic response to MERS-CoV, which caused an outbreak 307 in 2015 and resulted in a higher mortality rate than SARS-CoV. After searching the GEO 308 database for transcriptome datasets related to MERS-CoV infection in human cell lines, we 309 selected 12 datasets for the analysis, as listed in The results of the meta-analysis showed that 1552, 8682, 9301, and 9318 genes were 317 significantly dysregulated in the mock-infected and MERS-CoV infected groups with an 318 adjusted p-value cutoff of less than 0.05, at 12 hours, 24 hours, 36 hours, and 48 hours, 319 respectively (Table S4 ). Among these, 1026, 4037, 4038, and 4116 genes were upregulated, and 320 526, 4645, 5263, and 5202 genes were downregulated at 12 hours, 24 hours, 36 hours, and 48 321 hours, respectively. The significantly upregulated and downregulated genes obtained from each 322 time point were visualized through circos plots (Fig. 3A, 3C ) and venn diagrams (Fig. 3B, 3D) . 323

Then, an ORA-KEGG analysis was performed for the upregulated and downregulated genes at and viral responses were also observed to be upregulated after 24 hours. However, the interferon 347 pathways, which are critical for the antiviral response, were not much upregulated in MERS-348

CoV as compared to the SARS-CoV infection. Furthermore, the electron transport chain and 349 other cellular respiratory processes and energy-generating processes, as well as metabolic 350 processes, were significantly downregulated (Fig. S2) . 351

Overall, we found that similar pathways, including innate immune system-related pathways, and 352 metabolic pathways such as glutathione metabolism were significantly dysregulated in both the 353 SARS-CoV and MERS-CoV infections. Also, there were differences in the time points of 354 activation of several pathways, and the type 1 interferon response also varied between both 355 coronavirus infections. These results revealed the different as well as the commonly disturbed 356 pathways, which may be ideal targets for therapeutics and drug development. 357 358

We also sought to investigate and better understand the SARS-CoV-2 transcriptomic response by 361 identifying its similarities and differences with the transcriptomic responses of SARS-CoV and 362 MERS-CoV. This approach was adopted as it was envisaged that it would help us to identify the 363 common and specific features of SARS-CoV-2 that have enabled it to cause a pandemic as 364 compared to the two previous strains that caused comparatively less spread. To this end, we 365 searched the GEO database for transcriptomic profiles related to SARS-CoV-2. As of April 10, 366 2020, we obtained only the GSE147507 dataset, which is an RNA-sequencing dataset. That only 367 one dataset was identified is not surprising, considering that we are continually developing our 368 knowledge about this novel virus. Using the GSE147507 dataset, we decided to perform 369 experiments on four lung cell lines, namely, Calu-3, NHBE, A549, and ACE2-overexpressed 370 A549, which were infected with SARS-CoV-2 for 24 hours. package, GREP2, and differential expression analysis was performed using the EdgeR package. 385

The subsequent ORA-KEGG pathway analysis showed that similar pathways were enriched to 386 those that were observed to undergo enrichment in the meta-analysis, and there were no 387 pathways that were enriched for the SARS-CoV downregulated genes (Fig. S3) . 388

We also conducted a gene set enrichment analysis (GSEA) as it is a powerful method that can be 389 used to identify whether a set of genes shows statistically concordant differences between two 390 biological states. Hence, GSEA for the KEGG pathway for SARS-CoV (Fig. 4C) , MERS-CoV 391 ( Fig. 4D) , and SARS-CoV-2 ( Fig. 4E ) was performed to identify the significantly enriched 392 pathways that were activated and suppressed in the case of each coronavirus infection (Table  393 S5). From a comparison of the results, we found that 29 pathways were enriched in all three 394

HCoVs infections at 24 hours (Table 3) . Among these, 20 pathways were activated, and nine 395 pathways were suppressed after viral infection. 396

Most of the pathways that were activated upon infection by these three HCoVs were found to be 397 related to the innate immune system, such as the TNF signaling pathway, toll-like receptor 398 pathways, retinoic acid-inducible gene-I(RIG-I)-like receptors pathway, and the nucleotide- We also found that pathways such as butanoate metabolism, D-glutamine and D-glutamate 423 metabolism, arginine, and proline metabolism were suppressed in the case of SARS-CoV-2 424 alone. The meta-analysis results showed that these pathways were downregulated at different 425 time points at later stages in MERS-CoV and SARS-CoV infections (Table S6) . These results 426

suggest that SARS-CoV-2 infection enriches mostly similar pathways to those enriched by 427 SARS-CoV and MERS-CoV, but the transcriptional response of SARS-CoV-2 seems to be 428 higher as compared to SARS-CoV. Hence common pathways would be ideal targets for drug 429 discovery for the COVID-19 pandemic and possibly for any future coronavirus outbreaks or 430 pandemics. 431 drug repurposing analysis identifies potential drug candidates 434

In addition, we sought to extend our observations by analyzing the in-vivo transcriptome data of 435 human COVID-19 patients. We were curious to know whether the common pathways that we 436 was visualized through a PCA plot (Fig. 5A ). In addition, GSEA was performed to identify the 444 activated and suppressed pathways (Fig. 5B, Table S7 ). 445

The results indicated that most of the innate immune pathways such as the toll-like receptor 446 signaling, NOD-like receptor signaling, RIG-I-like signaling pathway, JAK-STAT signaling, 447 chemokine signaling, and antiviral pathways as well as other pathways such as the neuroactive 448 ligand-receptor interaction and olfactory transduction pathways were activated in the POS group, 449 as observed in the in-vitro data analysis. In addition, there was activation of pathways such as 450 complement and coagulation cascades and platelet activation in the SARS-CoV-2 positive 451 patients, and the activation was more significant than that observed in the in-vitro data. Also, 452 suppressed pathways such as PD, oxidative phosphorylation, carbon metabolism, glutathione 453 metabolism, and thermogenesis were found to be suppressed in the POS group as in the in-vitro 454 data analysis. The most significant suppression that was identified in the POS group occurred in 455 the ribosome-associated pathways. Thus, we would suggest that these pathways are crucial for 456

HCoV pathogenesis and associated comorbid effects. 457

In light of the foregoing, drugs that can reverse the above-described pathogenic signature of 458 SARS-CoV-2 will be more effective in the treatment of COVID-19 disease. To test this 459 hypothesis, we used the cogena pipeline, which identifies the coexpressed genes in a diseased 460 group that differs from those in a healthy group. The basic assumption behind this tool is that 461 coexpressed genes tend to have a similar function at the cellular level. Therefore, we subjected 462 the normalized expression data of 1214 significant genes at above logFC 1.5 and at less than an same pathways were not enriched for the same cluster. The DEGs were separated into eight 465 clusters (Fig. 5C ) using PAM clustering. Clusters 1, 2, 5, and 8 consisted of the downregulated 466

DEGs in the POS samples, and clusters 3, 4, 6, and 7 contained the upregulated DEGs in the 467 POS samples. A KEGG pathway analysis was performed on each cluster (Fig. 5D) . 468

The results showed that Cluster 1 enriched pathways such as hypertrophic cardiomyopathy, cell Cluster 8 was selected for the drug repurposing analysis because it attained a higher enrichment 478 score as compared to the other seven clusters (Fig. 5E) are promising drugs whose antiviral activity has already been reported 64-67 and therefore show 492 potential for use in COVID-19 treatment. The genes that were coexpressed in cluster 8 are listed

Most of the drugs that we identified through this approach exert antiviral activity. In addition, 496 they are related to the treatment of comorbid factors such as diabetes, hypertension, and 497 thrombosis. Many of these drugs are currently undergoing clinical trials involving COVID-19 498 patients. The drugs identified for the other clusters are listed in Table S8 . showed that metabolic pathways such as PD, glutathione metabolism, TCA cycle, oxidative 536 phosphorylation, thermogenesis, valine, leucine, and isoleucine degradation were some of the 537 pathways that were enriched for downregulated genes in both SARS-CoV and MERS-CoV 538 across most of the time points in our meta-analysis. Most of these pathways were also enriched 539 for downregulated genes upon SARS-CoV-2 infection at 24 hours (Fig. 4C) . RAGE is an inflammatory pattern recognition receptor that can recognize damage-associated 550 molecular patterns (DAMPs) such as HMGB1 and S100s. 78 The AGE, which is formed as a 551 result of non-enzymatic glycation of proteins at prolonged oxidative stress, binds to RAGE and, 552 in turn, activates the extracellular signal-regulated kinase/mitogen-activated protein kinase 553 (ERK/MAPK) signaling pathway, leading to activation of pro-inflammatory nuclear factor 554 (NFκB). 79 Furthermore, AGE-RAGE signaling activation has been associated with 555 cardiovascular diseases such as diabetes 80 and hypertension, 81,82 as well as inflammatory 556 diseases, 83 ARDS, 84 sepsis, 85 and thrombosis, 86 which are all co-morbidities of SARS-CoV-2 557 infection. The RAGE expression seems to be upregulated during inflammation, and knockdown 558 of RAGE has been shown to protect mice from influenza-induced mortality. 87 However, studies 559 related to this pathway and its association with HCoV pathogenesis remain scarce. Nevertheless, 560 a recent work hypothesized that RAGE could act as a biomarker for the severity of COVID-19. 88 561 are needed to understand the mechanism of this pathway activation upon HCoV infection and 564

how it leads to comorbid mortality. 565

Another important observation is that glutathione metabolism was suppressed in all three interesting finding is that GSH levels also drop in patients with ARDS and sepsis. 53 590

Glutathione, the major antioxidant in fighting oxidative stress, has been found to inhibit various 591 viral infections including influenza, 109 dengue, 110 , and HIV. 111 Glutathione has also been 592 reported to enhance the vitamin D regulatory and glucose metabolism genes and to increase 25-derivatives to block NFκB and subsequent cytokine storm syndrome in two patients suffering 595 from COVID-19 indicated that this approach could have potential as a novel therapy to fight the 596 pandemic 113 . 597

Our results also indicated that the olfactory transduction pathway is significantly activated upon 598 MERS-CoV and SARS-CoV-2 infection, as well as in COVID-19 patients. The involvement of 599 this pathway in SARS-CoV-2 pathogenesis needs to be explored further because SARS-CoV-2 600 patients have been found to exhibit anosmia and olfactory dysfunction 114-116 , and those with 601 olfactory dysfunction as an early symptom recover relatively early from COVID-19. 117,118 602

Recently, a report was published that showed that olfactory receptor neurons initiate ultra-rapid 603 antiviral innate immunity against rhabdovirus after binding with its surface glycoprotein in a 604 zebra-fish model. 119 This raises the prospect that those who experience olfactory dysfunction 605 during the early stages of SARS-CoV-2 infection may have better antiviral immunity, and thus, 606 their recovery rate is high from COVID-19. People in the older age group who have fewer 607 olfactory receptor neurons may have a suppressed early response, which may lead to this group 608 experiencing higher disease severity. However, more studies will be needed to justify this 609 speculation. 610

Our analysis also showed that a similar neuroactive ligand-receptor interaction pathway was 611 activated in all three pathogenic HCoVs, and is in line with the neurological implications of 612 SARS-CoV-2 infection that have recently been raised. 120 We also found that the osteoclast 613 differentiation pathway was activated in these three HCoVs. It has already been shown that 614 Our results also indicated that the PD pathway was suppressed in the case of all three HCoVs. It 637 is known that PD is associated with mitochondrial dysfunction. 131 Besides, viral infections are 638 considered to be a risk factor for PD. 132 Although there are no studies that prove the association 639 between PD and HCoV infections, one study detected antibodies against four coronavirus 640 antigens, including two HCoVs, in the cerebrospinal fluid of PD patients and a possible 641 correlation between HCoV infection and PD. 133 Following the proposed dual-hit theory for 642 PD, 134 , it may be that SARS-CoV-2 may also have the potential to increase PD risk in the future, 643 especially considering its neuro-invasive potential. 135 It has already been shown in a comorbid 644 factor analysis that SARS-CoV infection is correlated with PD. 136 A controversial hypothesis has 645 also been put forward that the 1918 influenza pandemic increased the risk of PD and those who 646 were born or who were young during the pandemic had a higher risk of developing PD as 647 compared to those born before 1888 or after 1924. 137-141 In accordance with our analysis 648 findings, it was recently reported that the COVID-19 patients developed PD after infection 142 and 649 the speculation that the infected individuals may be at the increased risk of developing PD in the 650 future 143 . Although these findings seem dubious, future studies should focus on the continuous 651 monitoring of COVID-19 recovered patients to determine the long-term deleterious effects of 652

In an attempt to validate our in-vitro observations by comparing them with in-vivo COVID-19 656 infected patient samples, we reanalyzed the transcriptome datasets derived from the nasal swabs 657 taken from SARS-CoV-2 positive patients and compared them with their corresponding controls. 658

The GSEA of both datasets revealed that most of the pathways that were activated or suppressed 659 were similar to those observed in our analysis using the Calu-3 cell line data. 660

We also employed the cogena workflow to evaluate potential drugs based on the coexpressed 661 genes among the DEGs. Several studies have utilized this workflow to predict potential drug 662 candidates from patient transcriptome datasets. 144,145 In a recently published study, 145 after 663 performing differential analysis, drug repurposing was performed based on the cogena workflow 664 in a reanalysis of transcriptome data of the BALF-fluid obtained from 10 COVID-19 patients and 665 20 healthy subjects. Although that study employed the same pipeline for drug repurposing 666 analysis as ours, we selected a study with a higher number of COVID-19 positive samples (430 667 COVID-19 positive patients) and healthy controls (54 COVID-19 negative healthy subjects), 668 which allows more reliable prediction. The potential drug candidates that were identified in each 669 cluster were expected to reverse the pathogenic disease state and can be further recommended for 670

Among the predicted drugs identified in our drug repurposing analysis, Verapamil and its R-672 enantiomer, Dexverapamil, are calcium channel blockers that are already approved for 673 therapeutic use in the treatment of high blood pressure and heart diseases, which are among the 674 comorbid factors of COVID-19. It has already been shown that Verapamil at higher doses can 675 inhibit filoviruses such as Ebola virus in-vitro. 146 Also, it has been reported that Verapamil can 676 inhibit the influenza virus by disrupting the calmodulin-dependent intracellular activities 677 necessary for viral assembly. 147,148 There is an ongoing clinical trial to investigate its role in 678 treating COVID-19 patients. 60 Another calcium channel blocker in the same group, Diltiazem, 679 has also been reported to exert anti-viral activity 148 and is also reported to be an inhibitor of 680

NFκB signaling that can reduce excessive cytokine production 149 which is the most significant 681 cause of COVID-19 associated deaths. The use of Diltiazem for COVID-19 treatment is being 682 under clinical trials. 62 drug that has been shown to inhibit HIV replication 150 , and its effect on COVID-19 patients is 685 being tested in an ongoing clinical trial. Another drug with potential is colchicine, which is used 686 to treat gout and Bechet's disease, and a clinical trial is underway to test its effect on COVID-19 687 patients. 63 Paclitaxel, which is basically a chemotherapy agent, is another drug that our analysis 688 predicted could be useful in the fight against COVID-19. It has also been reported to have 689 antiviral activity against HIV. 151 However, its use against HCoV infections is yet to be explored. 690

Valproic acid, an epilepsy drug, is another promising drug for the treatment of COVID-19 691 infection that was predicted by the drug repurposing analysis. Its antiviral activity has already 692 been reported 64 , and it can be used in the treatment of ARDS and can reduce cytokine storm by 693 inhibiting the activation of NFκB signaling. 152,153 Moreover, a recent unpublished study showed 694 that the metabolite of valproic acid Co-enzyme A (CoA) has the potential to bind with the non-695 structural protein-12 (nsp-12) of SARS-CoV-2. 154 However, further studies are needed to 696 evaluate its potential in treating COVID-19 infection. 697

The results of our analysis also identified metformin as a potential drug candidate. This drug is 698 widely used to treat type 2 diabetes, which is also one of the important comorbid factors 699 associated with COVID-19 disease. It has recently been shown that Metformin can reduce the 700 complications and severity of COVID-19 in patients through its ability to reduce the C-reactive 701 protein (CRP) and TNF-A levels. 65, 155, 156 Troglitazone is another antidiabetic drug that can be 702 used for COVID-19 treatment. Troglitazone has been shown to inhibit the hepatitis-B virus 66 and 703 has immunomodulatory effects on cytokine production. 157 It has also been shown to significantly 704 reduce IL-6 levels. 158 705

Among the other drugs predicted by the same cluster in our analysis was Mesalazine. This drug 706 is mainly used in the treatment of inflammatory bowel diseases, 159 , and it has been reported to 707 reduce mortality when used in combination with neuraminidase inhibitor and results in better 708 recovery from H5N1 influenza in infected mice. It would, therefore, be interesting to study its 709 role in treating COVID-19 infection. 160 Another drug was Pyrantel, which is used to treat 710 parasitic worm infection. It has been reported that pyrantel can inhibit Chikungunya virus nsp1 711 cap protein. 161 A further drug, oxymetazoline, which is basically a decongestant used in nasal 712 sprays, has been reported to resolve inflammatory effects mediated by neutrophils 162 and has also 713 been shown to reduce replication of the human rhinovirus. 67 Hence, it may be a promising drug spectrum antibiotic whose antiviral role has not been much studied. However, a recent study 716 identified the potential of cefotiam to target TMPRSS2, and hence it may have an important role 717 to play in interfering with SARS-CoV-2 binding. 163 Lastly, oxalinic acid is a quinolone antibiotic 718 that can inhibit bacterial DNA gyrase enzyme. It has been shown that oxalinic acid can inhibit 719 human polyomavirus (BKV) 164 , but its association with HCoV pathogenesis is yet to be 720

In sum, through our multipronged approach, we identified potential drugs for COVID-19 722 treatment as well as the crucial pathways that are dysregulated upon infection by contagious 723

HCoVs, which cause outbreaks. However, this study has its limitations. First, the time point results are forthcoming, we expect that this pandemic will come to an end soon. 745

Through time point meta-analysis, this study revealed that crucial pathways were dysregulated 748 upon SARS-CoV and MERS-CoV infection. Most of these pathways were also found to be 749 dysregulated by SARS-CoV-2, the cause of the current COVID-19 pandemic. We validated these 750 findings by means of an in-vivo analysis of the transcriptome response of COVID-19 patients. 751

Subsequently, we identified potential pathways and the drugs to be targeting these pathways for 752 the treatment using co-gene expression and drug repurposing analysis. The common datasets for each time point were selected among the screened datasets for meta-1206 analysis. After normalization and differential expression analysis of each datasets using limma 1207 with significance threshold logFC greater than 1 and adjusted p-value less than 0.05, the batch 1208 adjustment was performed. A random effect model was chosen to integrate the data after 1209 Table 1 : SARS-CoV datasets used for meta-analysis 1286 Table 2 : MERS-CoV datasets used for meta-analysis 1287 Table 3 : Common pathways dysregulated upon SARS-CoV, MERS-CoV, and SARS-CoV-2 1288 infection 1289 Table S1 : PRISMA flow chart for SARS-CoV meta-analysis 1299 Table S2 : PRISMA flow chart for MERS-CoV meta-analysis 1300 Table S3 : Significant genes of SARS-CoV at each time point after meta-analysis. 1301 Table S4 : Significant genes of MERS-CoV at each time point after meta-analysis. 1302 Table S5 : Common pathways enriched by GSEA KEGG analysis for SARS-CoV, MERS-CoV, 1303 SARS-CoV-2 Calu-3 -24 hours datasets. 1304 Table S6 : Potential pathways common for SARS-CoV, MERS-CoV, and SARS-CoV-2. 1305 Table S7 : GSEA KEGG analysis results for SARS-CoV-2 positive patients compared with the 1306 healthy patient samples. 1307 Table S8 : Drugs enriched in other clusters in cogena analysis. 1308 

Candidate Agents for the Next Global Pandemic? A Review

Why are RNA virus mutation rates so damn high?

Human Coronaviruses OC43 (HCoV-OC43s) Circulating in France from

Reveals a High Intra-Specific Diversity with New Recombinant Genotypes

SARS and other coronaviruses as causes of pneumonia

Identification of a novel coronavirus in patients 766 with severe acute respiratory syndrome

Koch's postulates fulfilled for SARS 768 virus

A novel coronavirus associated with severe 770 acute respiratory syndrome

Epidemiology and cause of severe acute respiratory 772 syndrome (SARS) in Guangdong, People's Republic of China

Isolation of a 777 novel coronavirus from a man with pneumonia in Saudi Arabia

A new coronavirus associated with human respiratory 781 disease in China

Early Transmission Dynamics in Wuhan, China

14. (WHO) WHO. Coronavirus disease (COVID-2019) situation reports

SARS-CoV-2 Cell Entry Depends on 786 ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor

Radiological findings from 81 patients with COVID-19 789 pneumonia in Wuhan, China: a descriptive study

A major outbreak of severe acute respiratory syndrome in

Clinical features of patients infected with 2019 novel 793 coronavirus in Wuhan

COVID-19) Outbreak in China: Summary of a Report of 72314 Cases 796 From the Chinese Center for Disease Control and Prevention

Transport and direct utilization of gamma-glutamylcyst(e)ine 979 for glutathione synthesis

Glutathione and infection

The role of oxidative stress in viral infections

Redox rescues virus from ER trap

Recent advances towards 989 understanding redox mechanisms in the activation of nuclear factor kappaB

Role of glutathione in cancer progression and 992 chemoresistance

The emerging role of glutathione in Alzheimer's disease

Reduced and oxidized glutathione in the 998 substantia nigra of patients with Parkinson's disease

Decreased production of glutathione in patients with 1001 cirrhosis

Depressed hepatic glutathione and increased diene 1003 conjugates in alcoholic liver disease. Evidence of lipid peroxidation

HIV-seropositive individuals

Influenza A virus replication is dependent on an 1011 antioxidant pathway that involves GSH and Bcl-2

Evidence for antiviral activity of glutathione: in 1013 vitro inhibition of herpes simplex virus type 1 replication

Role of xanthine oxidase activation and reduced 1016 glutathione depletion in rhinovirus induction of inflammation in respiratory epithelial cells

Inhibition of influenza 1019 infection by glutathione

Inhibitory effects of glutathione on dengue virus 1021 production

Glutathione inhibits HIV 1023 replication by acting at late stages of the virus life cycle

Glutathione 1026 Stimulates Vitamin D Regulatory and Glucose-Metabolism Genes, Lowers Oxidative 1027 Stress and Inflammation, and Increases 25-Hydroxy-Vitamin D Levels in Blood: A Novel 1028 Approach to Treat 25-Hydroxyvitamin D Deficiency

Efficacy of glutathione therapy in relieving 1031 dyspnea associated with COVID-19 pneumonia: A report of 2 cases

Gustatory Dysfunction in COVID-19 Patients: A Systematic Review and Meta-analysis

SARS-CoV-2: Olfaction, Brain Infection, and the Urgent Need for 1037

Clinical Samples Allowing Earlier Virus Detection

Self-reported olfactory and taste disorders in 1040 SARS-CoV-2 patients: a cross-sectional study

Self-reported olfactory loss 1042 associates with outpatient clinical course in COVID-19

Olfactory and gustatory 1045 dysfunctions as a clinical presentation of mild-to-moderate forms of the coronavirus 1046 disease (COVID-19): a multicenter European study

Olfactory sensory neurons mediate ultrarapid 1049 antiviral immune responses in a TrkA-dependent manner

SARS-CoV-2 Infection Leads 1052 to Neurological Dysfunction

Potential enhancement of osteoclastogenesis by 1054 severe acute respiratory syndrome coronavirus 3a/X1 protein

Immune checkpoint inhibitors: a 1062 physiology-driven approach to the treatment of coronavirus disease 2019

Melatonin Inhibits COVID-19-induced Cytokine Storm by Reversing Aerobic Glycolysis in 1066 Immune Cells: A Mechanistic Analysis

Proteins Nsp12 and 13 of SARS-CoV-2 Have Mitochondrial 1068 Recognition Signal: A Connection with Cellular Mitochondrial Dysfunction and Disease 1069 Manifestation. Preprints

Decoding SARS-CoV-2 Hijacking of

Host Mitochondria in Pathogenesis of COVID-19

Mitochondria and Microbiota dysfunction 1073 in COVID-19 pathogenesis. Mitochondrion

SARS-coronavirus open reading frame-9b suppresses 1075 innate immunity by targeting mitochondria and the MAVS/TRAF3/TRAF6 signalosome

Upregulation of mitochondrial gene expression in PBMC 1078 from convalescent SARS patients

Mitochondrial Dysfunction in Parkinson's Disease: New 1080

Mechanistic Insights and Therapeutic Perspectives

McKernan Declan P. A role for viral infections in Parkinson's 1083 etiology?

Cerebrospinal fluid antibodies to coronavirus in patients 1085 with Parkinson's disease

Viremic attack explains the dual-hit theory of Parkinson's disease

Severe Acute Respiratory 1089

SARS-CoV-2) and the Central Nervous System

Helmich RC, Bloem BR. The Impact of the COVID-19 Pandemic on Parkinson's 1094 Disease: Hidden Sorrows and Emerging Opportunities

Infection in childhood and neurological diseases in adult life

Parkinson's disease and the environment in early life

Viral etiology for Parkinson's disease--a possible role of 1101 influenza A virus infection

Parkinson's Disease: Are We Dealing with Short-term Impacts or Something Worse? J 1104

A case of probable Parkinson's disease 1106 after SARS-CoV-2 infection

The clinically approved drugs amiodarone, 1115 dronedarone and verapamil inhibit filovirus cell entry

Verapamil inhibits influenza A virus replication

A Sialylated Voltage-Dependent Ca(2+) Channel

Binds Hemagglutinin and Mediates Influenza A Virus Entry into Mammalian Cells

Inhibitors of NF-kappaB signaling: 785 and counting

Inhibition of human immunodeficiency virus type 1 replication in human mononuclear 1126 blood cells by the iron chelators deferoxamine, deferiprone, and bleomycin

Anti-HIV, antitumor and immunomodulatory 1129 activities of paclitaxel from fermentation broth using molecular imprinting technique

Sodium 1132 valproate inhibits production of TNF-alpha and IL-6 and activation of NF-kappaB

Valproic acid attenuates the risk of acute 1135 respiratory failure in patients with subarachnoid hemorrhage

Virtual Screening and Molecular Dynamics Simulation Suggest 1137

Valproic Acid Co-A could Bind to SARS-CoV2 RNA Depended RNA Polymerase

Metformin in COVID-19: A possible role beyond 1140 diabetes

Observational Study of Metformin and Risk of 1142 Mortality in Patients Hospitalized with Covid-19. medRxiv

Troglitazone exhibits 1145 immunomodulatory activity on the cytokine production of activated human lymphocytes

Comparison of 1148 cytokine modulation by natural peroxisome proliferator-activated receptor gamma 1149 ligands with synthetic ligands in intestinal-like Caco-2 cells and human dendritic cells--1150 potential for dietary modulation of peroxisome proliferator-activated receptor gamma in 1151 intestinal inflammation

Mesalazine in inflammatory bowel disease: a trendy 1153 topic once again?

Delayed antiviral plus immunomodulator treatment 1155 still reduces mortality in mice infected by high inoculum of influenza A/H5N1 virus

resolves inflammatory reactions in human neutrophils

Advanced bioinformatics rapidly identifies existing 1164 therapeutics for patients with coronavirus disease-2019 (COVID-19)

Suppression of BK virus replication and 1167 cytopathic effect by inhibitors of prokaryotic DNA gyrase

The common datasets for each time point were selected among screened datasets for meta-1190 analysis. After normalization and bath adjustment using combat, the separation of samples at 1191 each time point 12hours

Differentially expressed genes in each dataset were calculated using the limma R package with a 1193 cutoff of logFC greater than 1, and the adjusted p-value less than 0.05. All the datasets at each 1194 time point were integrated using Fisher's method. The total number of significantly upregulated 1195 and downregulated genes common and unique between different time points were visualized 1196 through upset plots (2H,2J). The Over-representation analysis (ORA) for the KEGG pathway for 1197 upregulated (2G) and downregulated genes

PCA -Principal component 1200 analysis; KEGG-Kyoto Encyclopedia of Genes and Genomes 1201 was performed at 12 hours, 24 hours, 36 hours, 48 hours time points. Significant upregulated and 1211 downregulated genes between different time points obtained through meta-analysis were 1212 visualized through circos plot (3A and 3C) and venn diagram (3B and 3D), respectively. The 1213 Overrepresentation analysis for the KEGG

genes for enriched pathways at each time points was visualized through dot plots

Middle East Respiratory Syndrome coronavirus; KEGG-Kyoto Encyclopedia of 1216

The transcriptome datasets of the Calu-3 cell line infected 1223 by SARS-CoV, MERS-CoV, and SARS-CoV-2 at 24 hrs time point were analyzed for 1224 comparison. GSE33267 for SARS-CoV, which is a microarray-based transcriptome dataset, 1225 GSE65574 for MERS-CoV, which is a microarray dataset, and GSE147507 for SARS-CoV-2, an 1226 RNA-seq transcriptomic data

Gene set enrichment analysis of three different virus-infected profiles in the same Calu-3 cell 1228 line is shown in (4B)

CoV (4D), and SARS-CoV-2 (4E) were visualized through dot plots for significantly activated 1230 and suppressed pathways in each of virus infections. Respiratory Syndrome coronavirus; SARS-CoV-2 -Severe Acute Respiratory Syndrome 1233 coronavirus; KEGG-Kyoto Encyclopedia of Genes and Genomes 1234

Figure 5: GSEA and Coexpression Gene enrichment (Cogena) analysis of SARS-CoV-2 1237 positive patients' transcriptome identify potential pathways and drug candidates 1238

GSE152075, high throughput transcriptome sequencing data obtained from SARS-CoV-2

positive (POS) patients' nasal swabs were re-analyzed. The Segregation of Healthy and SARS

The Differential 1241 expression analysis was performed using EdgeR with significance threshold logFC>1.5 and 1242 adjusted p-value less than 0.05. GSEA analysis identified the activated and suppressed pathways 1243 in SARS-CoV-2 positive patients and visualized through the dot plot (5B). The Normalized 1244 expression of significant genes was subjected to cogena analysis, and different clusters were 1245 visualized through heatmap (5C). KEGG gene enrichment analysis for each cluster was 1246 visualized through the dot plot (5D). Drug repurposing analysis of the selected cluster identified 1247 potential drug candidates

PCA-Principal component analysis; SARS-CoV-2 -Severe Acute Respiratory Syndrome 1249 coronavirus; KEGG-Kyoto Encyclopedia of Genes and Genomes

IISERB) Bhopal. P.K. is supported by (IISER)-Bhopal Institutional fellowship

is 1315 supported by CSIR fellowship

Writing -review & editing

Writing -review & editing

Nachimuthu Senthilkumar: 1341 Writing -review & editing 1342 1343 1344 1345 1346 Conflict of interests: The authors declare no conflict of interests

We acknowledge all the contributors of Gene expression profiling datasets 1311 we used in this study for making it available public for reuse. 1312

Oxolinic acid