key: cord-0984149-tqn9y2f7
authors: Baralić, Katarina; Jorgovanović, Dragica; Živančević, Katarina; Miljaković, Evica Antonijević; Antonijević, Biljana; Djordjevic, Aleksandra Buha; Ćurčić, Marijana; Ćosić, Danijela Đukić
title: Safety assessment of drug combinations used in COVID-19 treatment: in silico toxicogenomic data-mining approach
date: 2020-09-11
journal: Toxicol Appl Pharmacol
DOI: 10.1016/j.taap.2020.115237
sha: c7cac72e01e1a2fb8b7baab2e2eec40214a6ea8c
doc_id: 984149
cord_uid: tqn9y2f7

Improvement of COVID-19 clinical condition was seen in studies where combination of antiretroviral drugs, lopinavir and ritonavir, as well as immunomodulant antimalaric, chloroquine/hydroxychloroquine together with the macrolide-type antibiotic, azithromycin, was used for patient's treatment. Although these drugs are “old”, their pharmacological and toxicological profile in SARS-CoV-2 – infected patients are still unknown. Thus, by using in silico toxicogenomic data-mining approach, we aimed to assess both risks and benefits of the COVID-19 treatment with the most promising candidate drugs combinations: lopinavir/ritonavir and chloroquine/hydroxychloroquine + azithromycin. The Comparative Toxicogenomics Database (CTD; http://CTD.mdibl.org), Cytoscape software (https://cytoscape.org) and ToppGene Suite portal (https://toppgene.cchmc.org) served as a foundation in our research. Our results have demonstrated that lopinavir/ritonavir increased the expression of the genes involved in immune response and lipid metabolism (IL6, ICAM1, CCL2, TNF, APOA1, etc.). Chloroquine/hydroxychloroquine + azithromycin interacted with 6 genes (CCL2, CTSB, CXCL8, IL1B, IL6 and TNF), whereas chloroquine and azithromycin affected two additional genes (BCL2L1 and CYP3A4), which might be a reason behind a greater number of consequential diseases. In contrast to lopinavir/ritonavir, chloroquine/hydroxychloroquine + azithromycin downregulated the expression of TNF and IL6. As expected, inflammation, cardiotoxicity, and dyslipidaemias were revealed as the main risks of lopinavir/ritonavir treatment, while chloroquine/hydroxychloroquine + azithromycin therapy was additionally linked to gastrointestinal and skin diseases. According to our results, these drug combinations should be administrated with caution to patients suffering from cardiovascular problems, autoimmune diseases, or acquired and hereditary lipid disorders.

pneumonia, progressive respiratory failure and death (Velavan and Meyer, 2020) . Key steps in the COVID-19 treatment include the prevention of the virus spread, which is achieved by home isolation of suspected cases and those with mild illnesses, along with the strict infection control measures at hospitals that include contact and droplet precautions. Mild illness should be treated at home, managed by maintaining hydration and nutrition, with the avoidance of non-adequate use of antibiotics and antiviral drugs. In severe, hospitalized patients, provision of oxygen through nasal prongs, face mask, high flow nasal cannula or non-invasive ventilation are indicated as well as mechanical ventilation, extra corporeal membrane oxygen support and renal replacement therapy (Singhal, 2020) . Today, there are different recommendations for COVID-19 treatment, including the use of "off-label" drugs, traditional Chinese medicine approach, and intravenous application of immunoglobulins, interferons and plasma of patients recovered from COVID-19 (Abedi et al., 2020; Cao et al., 2020; Ren et al., 2020; Saavedra, 2020) . The most common drugs used in order to ameliorate patient"s condition are oseltamivir, ganciclovir, 5 the molecular pathways and biological processes behind drugs-induced side effects which may be crucial for their management and prevention.

The Comparative Toxicogenomics Database (CTD; http://CTD.mdibl.org), Cytoscape software package version 3.8.0 (https://cytoscape.org) and ToppGene Suite portal (https://toppgene.cchmc.org)

were used as the main in silico data-mining tools in our research.

The Comparative Toxicogenomic Database is a publicly available database that processes and integrates data describing relationships between chemicals/drugs, genes, and diseases (Davis et al., 2019) .

This database is updated on a monthly basis, thus allowing the accuracy, cohesiveness and accessibility of data (Davis et al., 2019 (Davis et al., , 2008 . It primarily focuses on chemicals present in the environment and outcomes relevant to human health, but also provides data from animal studies if the gene of interest is also relevant to humans (Davis et al., 2008; Meng et al., 2013) .

The database contains various tools that can be used for the extraction of toxicogenomic data. For our in silico analysis we used SetAnalyzer and VennViewer CTD tools. SetAnalyzer CTD tool performs set-based enrichment for collections of chemicals or genes, and pathway generation for collections of genes while VennViewer CTD tool forms a diagram to compare datasets of the observed drug/chemical, gene, and disease relationships.

Cytoscape is a free software package used for visualizing, modelling and analysing molecular and genetic interaction networks. It can be used in functional genomics and proteomics experiments for obtaining an interaction network for genes of interest (Cline et al., 2007) . It can also be combined with large databases of protein-protein, protein-DNA, and genetic interactions and numerous plug-ins which provide a way to implement new algorithms, additional network analyses, and biological semantics (Cline et al., 2007; Shannon et al., 2003) .

J o u r n a l P r e -p r o o f 6 ToppGene Suite portal (https://toppgene.cchmc.org) is a freely available online tool which aids in identification of genes sets and their prioritization based on functional annotations and interactions with proteins (Chen et al., 2009) . ToppFun ToppGene tool (https://toppgene.cchmc.org/enrichment.jsp) was applied in this research. This tool can be used for transcriptome, ontology, phenotype, proteome, and pharmacome annotations based gene list functional enrichment analysis (Chen et al., 2009 ).

The flow chart for the different steps of our analyses is shown in the Fig. 1 . In this study, information about the genes affected by the investigated drugs was obtained from the "Genes" data-tabs on the CTD database for each investigated drug. Lopinavir and ritonavir were found in CTD separately and as fixed lopinavir/ritonavir combination under 3 different names: Aluvia, Kaletra, and Lopimune.

Thus, we evaluated the lopinavir/ritonavirgenedisease interactions in two ways. Firstly, we examined fixed lopinavir/ritonavir combination and secondly, we compared it with the results obtained when lopinavir and ritonavir were regarded separately. Since chloroquine/hydroxychloroquine, and azithromycin are present in CTD database only as separate compounds, we investigated their combined effects. Extracted data was further analysed by using freely-available in silico tools: (1) 

In this research, CTD VenViewer tool was used to obtain a list of genes common to the investigated drugs. Next, to generate a tight network of genes/proteins associated with the investigated drug combinations and to obtain the lists of related genes, extracted common genes were inserted into the GeneMANIA Cytoscape plug-in, while Homo Sapiens was selected as target organism. Since constructing a disease specific network and using its topological parameters to rank the hub genes is one of the widely used approaches in biomarker prioritization studies (Ghatge et al., 2018), we used CytoHubba plug-in to calculate the top 15 hub genes out of the constructed GeneMANIA network and ranked them by the Maximal Clique Centrality (MCC) score. MCC, found to be the most accurate of all the methods available in the CytoHubba, was used considering its ability to generate precise predictions of essential proteins (Chin et al., 2014) .

In order to understand the biological importance of the generated hubs and to elucidate biological networks and pathways most relevant to the investigated drug combinations, we performed the gene ontology (GO) enrichment analysis.

ToppGene ToppFun function was used to obtain the lists of the 10 most significant molecular functions and biological processes which may elucidate the connection between the use of the investigated drug combinations and disease development. The default settings of the ToppGene Suite portal were selected for this analysis, meaning that the p-value was set to 0.05 and FDR corrected.

In our research, pathway analysis was performed by Cytoscape ClueGO plug-in version 2.5.6.

KEGG, Reactome, and WikiPathways databases were selected to extract the list of pathways. The two-J o u r n a l P r e -p r o o f Journal Pre-proof 8 sided hypergeometric test was used for the enrichment with a Bonferroni step down correction and a κ score of 0.3 to link the terms.

SetAnalyzer CTD tool was used for the identification of diseases associated with our set of 15 hub genes and thus, the use of investigated drugs. By default, p-value was set at 0.01. It is important to acknowledge that this tool predicts the link based on interacting hub genes without referring to the specific role of the genes in the found disease. Furthermore, we used VenViewer CTD tool to predict additional diseases associated with the data sets connected to the investigated drugs.

Top ten interacting genes reported in CTD for fixed antiviral combination of ritonavir/lopinavir were namely: GLB1, TP53, CCL2, CDKN1A, CXCL8, IL6, APOA1, APOB, APOC3, and LMNA. It was noted that this drug mixture increases the expression, activity or simply secretion of all the mentioned genes/proteins which may explain the development of unwanted effects of its use. This drug combination acts on 171 gene ontology processes and 44 molecular pathways.

On the other hand, when regarded as separate compounds, lopinavir interacted with 16, while ritonavir affected the expression of 114 genes among which 13 were in common ( Fig. 2A) for ritonavir and lopinavir. The total of 655 and 1109 biological processes, and 13 and 107 molecular pathways were detected for lopinavir and ritonavir, respectively.

Notably, only 4 genes were described as interacting genes in both analyses (CCL2, IL6, LMNA,

and TNF) and thus, we further examined the safety of the fixed lopinavir/ritonavir combination separately from the combination of ritonavir and lopinavir, present in CTD as isolated compounds.

Further step of our analysis included constructing a tight networks of lopinavir/ritonavir interacting genes, along with 20 related genes. In the case of fixed antiviral combination, the results have shown that more than half of these genes (52.12%) were in physical interactions, which means that there is an interaction between the protein products of these genes while 10.23% of them were in co-expression, which means that their expression levels are similar under the defined conditions in the gene expression study. 8.31% belonged to the same molecular pathway, 5.95% of the interactions were predicted by the server, 5.70% interactions were genetic, while 0.13% were shared protein domains (Fig. 3A) . Gene network between 13 lopinavir/ritonavir interacting genes along with 20 related genes has shown that the majority of these genes (77.27%) were in co-expression. 12.47% of the investigated genes were in colocalization, 8.82% were predicted by the server, 1.28% were shared protein domains, while 0.16% were genetic interactions (Fig. 3B ).

Top 15 hub genes extracted from the constructed networks are presented in Fig. 4A (for fixed lopinavir/ritonavir combination) and Figure 4B (when lopinavir and ritonavir are regarded separately).

In the two sets, 6 genes were in common: CCL2, CXCL1, CXCL2, CXCL8, ICAM1, and IL6, all highly involved in the regulation of immune system related biological processes. Moreover, even the intensity of correlation for mentioned genes was similar in two sets, with an exception of CXCL1. Fixed lopinavir/ritonavir combination increased its expression to a higher level when compared to lopinavir plus ritonavir treatment.

J o u r n a l P r e -p r o o f

Gene ontology (GO) describes gene products in terms of their associated biological processes, cellular components and molecular functions. The top 10 molecular functions and biological processes for the generated gene sets are shown in the Table 1 . Molecular functions of the hub genes related to fixed lopinavir/ritonavir combination were shown to be highly associated to chemokine activity and lipid metabolism, such as cholesterol and sterol binding, or phosphatidylcholine-sterol O-acyltransferase activity. Similarly, the most important biological processes were related to chylomicron remodeling, lipid metabolism, as well as interleukin signaling and inflammation, as shown below. On the other hand, we noticed that majority of the hub genes linked to lopinavir/ritonavir combination show cytokine related activity and consequently, regulate inflammatory response. Additionally, under the top 10 biological processes, we detected cellular responses to lipopolysaccharides and lipids (Table 1) .

Our pathway enrichment analysis has shown that the hub genes detected for lopinavir/ritonavir fixed combination could be grouped into 2 main molecular pathways, IL-10 signaling and chylomicron assembly,. When exploring the lopinavir/ritonavir interacting hub genes and corresponding pathways, we detected 5 distinct clusters: IL-10 signaling, viral protein interaction with cytokine and cytokine receptor, Legionellosis, RANKL/RANK signaling, and mRNAs involvement in the immune response in sepsis.

Finally, we detected associated diseases (marker/mechanism and/or therapeutic) for both, fixed combination and lopinavir/ritonavir mixture (Table 2) , among which were cardiovascular, respiratory diseases, immune system related disorders, skin diseases, to name a few. While apolipoprotein-coding genes were critical for development of distinct cardiovascular diseases, inflammatory genes have been reported as mechanistically related to, cardiovascular, respiratory, and immune system diseases. COVID-19 was seen in the list of hub genesrelated diseases in both analyzes.

Our further CTD analysis predicted the total of 225 diseases linked to the fixed combination associated hub genes and 295 for the other set of highly connected genes. Among them 191 were in common (58.1%).

Moreover, we constructed a list of diseases to which development might contribute the use of a lopinavir and ritonavir combination. There were 9 such diseases: bipolar disorder, chemical and drug induced liver injury, diabetes insipidus, nephrogenic, Fanconi syndrome, HIV-associated lipodystrophy syndrome, metabolic syndrome, neutropenia, renal insufficiency, status epilepticus.

Obtained results have shown that chloroquine interacted with 2296, hydroxychloroquine with 45 and azithromycin with 14 genes. Among these genes, chloroquine and azithromycin had a common effect shown that more than half of these genes (82.95%) were in co-expression, while 10.82% were in colocalization, which means that they are expressed in the same tissue or their gene products are identified in the same cellular location (Fig. 7A ). The rest of the 6.23% of the interactions between genes were predicted by the server, while none of the genes were in physical interactions.

In the case of hydroxychloroquine and azithromycin combination, the results have shown that more than half of these genes (81.21%) were in co-expression, while 10.98% were in co-localization ( Fig.   7B ). Similarly, to the chloroquine/azithromycin combination, the rest of the 7.81% of the interactions between genes were predicted by the server, while none of the genes were in physical interactions.

The top 15 hub genes for chloroquine/azithromycin and hydroxychloroquine/azithromycin are presented in Fig. 8A and 8B, respectively. Obtained hub genes were identical for both combinations, while the intensity of correlation for some of them was different when compared among the two sets.

Gene ontology connected with our set of 15 hub genes and, thus, potentially linked to the use of the two drug combinations was analysed. Top 10 obtained biological processes and molecular functions are shown in the Table 3 .

In order to understand the biological importance of the hubs, we performed the pathway analysis.

The results have shown that chloroquine/hydroxychloroquine + azithromycin combination interacting hub genes can be grouped into 8 pathways: NOD-like receptor signalling pathway, viral protein interaction with cytokine and cytokine receptor, NF-kappa B signalling pathway, interleukin 10 signalling, Legionellosis, rheumatoid arthritis, IL-17 signalling pathway and IL-18 signalling pathway.

Top 10 diseases (marker/mechanism and/or therapeutic) associated with our set of 15 hub genes are presented in the Table 4 , sorted by the p-value. The listed diseases include both those for which the J o u r n a l P r e -p r o o f investigated drug combinations could be used as a therapy and those which could be induced by the said combinations.

However, results obtained by VenViewer CTD tool have shown that the therapeutic effect of both chloroquine/azithromycin and hydroxychloroquine/azithromycin is connected with their effect on 4 genes (CCL2, IL6, TNF and IL1B1) directly associated with COVID-19 disease. Both combinations reduced the expression of IL-6 and TNF proteins. These genes were also listed in inflammation and immune system diseases.

Further VennViewer CTD analysis was conducted to obtain a list of diseases to which development might contribute the use of a combination of chloroquine and azithromycin or hydroxychloroquine and azithromycin. Results of the CTD analysis also reveal that acute renal failure, depression, headache, nausea, pain, and tachycardia (ventricular) could be linked to simultaneous decreased expression of TNF and IL-6 proteins caused by the investigated drug combinations.

Generally, it has been shown that drug combinations are more effective against certain viruses, such as HIV and SARS-CoV-2, in comparison with the individual ones (Muralidharan et al., 2020) .

J o u r n a l P r e -p r o o f 14 Lopinavir and ritonavir combination has been proposed to control the virulence to a great extent in the COVID-19 affected patients within 48 h. It has even been suggested that these two drugs could be accompanied with other antiviral agents to enhance antiviral effects and improve clinical outcomes . For example, molecular docking and molecular dynamics simulation studies have shown that combination of lopinavir, oseltamivir and ritonavir were found to be highly effective against SARSCoV-2 protease (Muralidharan et al., 2020) . Furthermore, the combination of hydroxychloroquine and azithromycin on SARS-CoV-2 infected cells was tested in vitro, showing that there was a considerable synergy of these two substances applied at doses correlating the concentrations likely to be obtained in humans (Gautret et al., 2020b) . Epidemiological studies have also shown that hydroxychloroquine and azithromycin therapy was far more successful compared to untreated patients or treatment with hydroxychloroquine alone (J. . The benefits of azithromycin in the treatment of viral infections can be supported by a limited number of data indicating inhibitory activity for Zika virus proliferation (Retallack et al., 2016) , as well as a decrease in inflammation in a murine model of viral bronchiolitis after azithromycin administration (Beigelman et al., 2010) .

However, since balance between benefit and negative effects of drugs is sometimes hard to achieve, we considered both risks and benefits of proposed anti-COVID-19 therapy in order to understand the molecular pathways and biological processes behind drugs-induced side effects, which may be crucial for their management and prevention.

We attempted to understand the potential of lopinavir/ritonavir combination to cause side effects or complications of already existing diseases by applying system biology approaches and comprehensive data-mining. Our results obtained from the CTD show that fixed ritonavir/lopinavir combination interacts TNF and IL6 were recognized as lopinavir/ritonavirmodulated genes involved in the COVID19 treatment, leading to a suspicion that these proteins may be critical for some of the therapeutic properties of the given drugs. Increased expression of the pro-inflammatory cytokines, CCL2 and IL-6, triggers the activation of downstream pathways such as NOD-like receptor signaling pathway, involved in the detection of various pathogens and generation of innate immune responses (Davis et al., 2011; Philpott et al., 2014) . Furthermore, IL-17, found in the molecular pathways triggered by the investigated drug combination, is known to connect innate and adaptive immunity. It activates the induction of IL-6 and TNF-alpha, while CCL2 is among its signature genes. On the other hand, IL-10 family cytokine activates many of the same innate inflammatory genes as IL-17 (Gaffen, 2008) . In other words, these lopinavir/ritonavir -IL-17/IL-10/CCL2/IL6 interactions appear as important stimulators of the host immunity and pyroptosis and, thus, viral clearance (Hachim et al., 2020) . However, their over-activation might cause inflammationmediated side effects or induce AGE-RAGE signaling pathway in diabetic complications and might elicit the activation of multiple intracellular signaling pathways, including NF-kappaB activity (Nguyen et al., 2012) . NF-kappa B further promotes the expression of pro-inflammatory cytokines and a variety of atherosclerosis-related genes, such as VCAM-1, tissue factor, VEGF, and RAGE (Santoro et al., 2003) . Thus, it seems that antiretroviral drugs used in SARS-CoV-2 infection may exert both beneficial and adverse effects by interacting with the set of common genes making it harder to distinguish between their pharmacological characteristics and possible toxic effects.

Hence, the increased expression of all the mentioned genes by lopinavir/ritonacir combination, could subsequently stimulate inflammatory response in the treated patients and cause serious side effects shown in the Table 2 , such as brain ischemia, reperfusion injury, myocardial ischemia or hemorrhagic shock, particularly in those patients who have other risk factors and secondary diseases. Furthermore, we could even speculate that the proposed therapy could increases the risk of severe complications in SARS-CoV-2-infected patients, the excessive production of proinflammatory cytokines called "cytokine storm", which might be associated with the severity and outcome of COVID-19 disease (Conti et al., 2020; Tufan and Matucci-Cerinic, 2020 ).

In the next step of our analysis, after generating gene networks for both treatment options, we predicted hub genes that were then clustered based on their biological functionality into groups. In both data sets, first group of genes was involved in regulation of IL-10 signaling pathways and its subpathways such as NF-kB signaling, IL-17 signaling, and IL-4 and IL-13 signaling, to name a few. While fixed lopinavir/ritonavir combination modified their activity through IL6, CXCL8, CXCL2, CCL2, PTX3, APOA1, ICAM1, CXCL1, CXCL3, CDKN1A, APOB and PTGS2, combined use of these drugs may interact with NFKBIA, CXCL1, ICAM1, IL1B, CXCL2, TNF, CCL3, CCL4, CXCL8 and IL-6 to cause the change in IL-10 signaling.

As explained, IL-10 pathways represent the link between the investigated drugs and possible development of immune-related side effects. Indeed, our pathway enrichment analyzes predicted inflammation, hypersensitivity, glomerulonephritis, and other immune system related diseases among 10 most probable ritonavir/lopinavirinduced diseases. In other words, this drug combination may cause higher risk than benefit in patients suffering from autoimmune diseases such as rheumatoid arthritis, asthma or similar, possibly through identified biological processes such as cytokine-mediated signaling pathway and response to cytokines.

On the other hand, there were some differences between the data reported for fixed ritonavir/lopinavir combination and their mixture. Among the top 10 interacting genes, fixed antiviral combination increased the expression of 3 apolipoprotein coding genes: APOA1, APOB, and APOC3, J o u r n a l P r e -p r o o f while VennViewer CTD tool has shown that ritonavir and lopinavir commonly interact with 2 genes involved in lipid metabolism, ADIPOQ and LEP, by decreasing their activity. Thus, in the first examined set, enriched pathways were grouped under chylomicron assembly category, while the second one clustered around viral protein interaction with cytokine and cytokine receptor, Legionellosis, RANKL/RANK signaling, and mRNAs involvement in the immune response in sepsis. Subsequently, this led to difference in the top 10 predicted drug-induced diseases.

Fixed lopinavir/ritonavir combination was associated with the risk of developing lipid metabolism disorders such as hyperlipidemia, dyslipidemia or hyperlipoproteinemia, most likely by influencing biological processes such as chylomicron remodeling, chylomicron assembly, triglyceriderich lipoprotein particle remodeling, and forming of protein-lipid complex. This could also be seen 

The experimental studies have demonstrated that both chloroquine phosphate and hydroxychloroquine sulfate inhibit SARS-CoV-2 virus (Gautret et al., 2020b) . Apart from the demonstrated ability of chloroquine to inhibit growth of SARS-CoV-2, an early clinical trial conducted in COVID-19 Chinese patients showed that this drug had a significant effect on clinical outcome (Gautret et al., 2020a) , by reducing fever, improving CT imaging, and delaying disease progression (Gautret et al., 2020b) .

However, hydroxychloroquine is considered to have lower toxicity and greater safety than chloroquine, especially when administrated at higher doses (Chatre et al., 2018; .

Our results have demonstrated that both chloroquine/azithromycin and hydroxychloroquine/azithromycin combinations interacted with 6 genes (CCL2, CTSB, CXCL8, IL1B, IL6, and TNF), while chloroquine and azithromycin interacted with two additional genes (BCL2L1 and CYP3A4). Further genes taken into account in our investigations were the hub genes obtained by the Cytoscape Cytohubba plug-in. We noted that the same hub genes were generated for both drug combinations, which is understandable having in mind that chloroquine and hydroxychloroquine share similar chemical structures and mechanisms of acting as a weak bases and immunomodulators (J. Yao et al., 2020) . However, the intensity of correlation for some of them was different when compared among the two sets. For chloroquine/azithromycin combination the correlation was the highest with ICAM1, IL1B, IL6, and HCAR3, while for hydroxychloroquine/azithromycin it was the highest with TNF and PTGS2. This difference might be behind the less pronounced adverse effects observed in the case of hydroxychloroquine treatment compared to chloroquine. However, our investigation has shown that, in the case of chloroquine/hydroxychloroquine + azithromycin combinations, it is difficult to draw a line between adverse and beneficial effects, since some of the genes (CCL2, IL6, TNF, and IL1B1) are involved not only in the ethiology of their toxic effects, but also in therapeutic action against COVID-19.

This could also be seen in the list of top 10 diseases connected with the use of chloroquine/azithromycin and hydroxychloroquine/azithromycin combinations conducted by our CTD analysis. Inflammation was listed in the first place, having in mind that both chloroquine and J o u r n a l P r e -p r o o f hydroxychloroquine are anti-inflammatory agents that, other than their direct antiviral activity, can significantly decrease the production of cytokines, in particular pro-inflammatory factors (J. . Indeed, cytokine and chemokine activity and binding to combined receptors were seen as the top molecular functions impacted by the investigated drug combinations. Consequently, cellular response to cytokine stimulus and cytokine signalling pathways might be dysregulated, which could be regarded as mechanistic basis of their possible adverse reactions. Hypersensitivity in the immune system category was among the top diseases, which is expected having in mind that these two drugs are also used to treat various immune system diseases, such as rheumatoid arthritis and lupus erythematosus (Savarino et al., 2003; Wozniacka et al., 2006) . Hence, rheumatoid arthritis was also present among our enriched investigated combinations. This is also the case with CCL3, another gene closely associated with myocardial ischemia. Nevertheless, in their toxicogenomic data-mining study, Chen et al. (2019) have found that this gene was enriched Toll-like receptor signalling pathway, which includes immune response through circulating blood cells, increasing infarct size and influencing ventricular remodelling.

In the process of CTD analysis, we also generated a list of other diseases to which development might contribute the treatment with chloroquine/hydroxychloroquine + azithromycin combination. It was noted that fewer diseases were present in the case of chloroquine/azithromycin combination than hydroxychloroquine/azithromycin. Diseases common to all the investigated substances included dizziness, gastrointestinal diseases, headache, hearing loss, heart arrest, heart block, long qt syndrome, muscular diseases and torsades de pointes. Additional diseases connected with chloroquine/azithromycin, but not hydroxychloroquine/azithromycin, were acute kidney injury, arthralgia, delirium, depressive disorder, lipidoses, myasthenia gravis, pruritus, psychoses (substance-induced), tachycardia, (ventricular) and vasculitis (leukocytoclastic, cutaneous).

Predictably, our further GO enrichment analysis of the generated hub genes set has demonstrated that the majority of their molecular processes were linked to cytokine and chemokine activation process.

The pathway enrichment analysis has also demonstrated that chloroquine/hydroxychloroquine + azithromycin combination hub genes could be grouped into several pathways, including NOD-like receptor signalling pathway, viral protein interaction with cytokine and cytokine receptor, NF-kappa B signalling pathway, IL-10 signalling pathway, IL-17 signalling pathway and IL-18 signalling pathway.

All of these pathways are mutually linked and connected to the immune system response. For example, upon recognition of pathogen-associated molecular patterns, NOD-like receptors activate NF-κB or MAP kinases to induce the production of inflammatory cytokines such as IL-1β and IL-18 (Kumar et al., 2011) . Results of a study in which cytokines were measured in blood plasma of 12 COVID-19

patients have shown that 2019-nCoV viral load was highly positively associated with the plasma levels of both IL-10 and IL-17 and could be used as markers to predict COVID-19 disease severity (Y. . IL-17 was among the cytokines found abundant in COVID-19 patients and associated with the J o u r n a l P r e -p r o o f severe lung inflammation. This cytokine promotes the induction of other cytokines, such as IL-1β, IL-6

and TNF-α, as well as growth factors, G-CSF; chemokines; and matrix metalloproteinases (Casillo et al., 2020; Tufan and Matucci-Cerinic, 2020) . Chemokines upregulated by IL-17 also include CXCL1, CXCL2, CXCL5, CXCL8 (IL-8), CXCL9, CXCL10, CCL2, and CCL20 (Xie et al., 2011) , many of them listed as hub genes in our study.

In general, studies have shown that immunomodulatory effect of chloroquine and hydroxychloroquine may be useful in controlling the cytokine storm that occurs in the late-phase of critically ill SARS-CoV-2 infected patients (Yao et al., 2020) . Both of the investigated drug combinations decreased the expression of IL-6 and of TNF proteins and thus, suppressed generated immune response.

According to the CTD base, this could be further linked to the therapeutic action against COVID-19.

However, it should also be considered that both IL-6 and TNF are components of a complex system which consists of many positive and negative regulators. IL-6 has both pro-inflammatory and antiinflammatory properties. It is produced not only by immune cells, as well as monocytes and macrophages, but also by cardiovascular components, like endothelial cells, vascular smooth-muscle cells, and ischemic myocytes. Thus, IL-6 is involved not only in inflammation but also in the regulation of cardiac metabolism (Kanda and Akahashi, 2004) . TNF is also considered a double-natured cytokine which may express protective and adaptive effects in cases of hemodynamic overload and ischemia-reperfusion (Bellisarii et al., 2001) , since TNF signalling pathway has exceedingly been associated with cardiac remodelling following myocardial infarction (Chen et al., 2019) . Accordingly, decreasing IL6 and TNF protective activity might also lead to the possible adverse cardiovascular effects present in the case of chloroquine/hydroxychloroquine + azithromycin. Furthermore, PTSH2 gene, enriched in TNF signalling pathway, was also listed as one of the genes that might influence the occurrence of possible cardiotoxic effects of the investigated combinations, especially since it has been found to be associated with the decreasing risk of stroke (Chen et al., 2019) .

Additionally, our results have shown that both chloroquine and azithromycin decreased the activity of CYP3A4 gene. CYP3A4 is the most abundant metabolic enzyme in human liver, regulating J o u r n a l P r e -p r o o f Journal Pre-proof 22 more than 50% drug metabolism . CYP3A4 and CYP3A5 play important roles in chloroquine metabolism and are the major enzymes responsible for the chloroquine N-demethylation into N-desethylchloroquine in human liver microsomes (Kim et al., 2003) . Inhibitory effect on the activity of this enzyme would mean the inhibition of metabolism and elimination of these drugs from the organism, thus increasing the risk of their possible toxic effects. However, CYP3A4 was not listed among the interactions for hydroxychloroquine. This could also be a reason why hydroxychloroquine/azithromycin combination could be regarded safer than chloroquine/azithromycin.

Another adverse event observed in our investigation included skin changes and muscular diseases. However, these effects are associated with the long-term use of chloroquine/hydroxychloroquine therapy (Chatre et al., 2018) and, again, may be explained through over-activation of cytokine activity and immune response.

Our toxicogenomic data mining approach has proven useful for exploring and gaining better understanding of the safety profile of the selected COVID19 drug combinations. However, freely available tools such as CTD, various Cytoscape plugins and ToppGene Suite portal, have some boundaries which confirm that in silico toxicogenomic investigations could be used only as an addition to other in vitro and in vivo methods in the overall toxicity testing process. Having in mind that the data in such functional annotation-based prioritization processes is extracted by drawing statistical associations between genes and drugs of interest, it is not possible to directly address the dose-response relationship (Harris et al., 2020) . Furthermore, the analysis deeply relies on the online sources the annotations were obtained from (in this case, CTD database), as well as the quality of interaction data present in them (Chen et al., 2009) . It is often necessary to omit certain data due to the insufficient number of gene-drug annotations, and sometimes the analysis is not even possible, which is the case with some other drugs used in COVID-19 treatment, such as remdesivir and interferon beta. It should also be recognized that not only dose, but many other factors might influence the manifestation of drug toxic effects, such as route of J o u r n a l P r e -p r o o f administration, duration, individual drug metabolism rates, etc. (Davis et al., 2008) , which should be explored in further throughput investigations.

In this study we used toxicogenomic data-mining approach to understand the safety profile of two J o u r n a l P r e -p r o o f 

Plausibility of therapeutic effects of Rho kinase inhibitors against Severe Acute Respiratory Syndrome Coronavirus 2 (COVID-19)

Azithromycin attenuates airway inflammation in a mouse model of viral bronchiolitis

Tumor necrosis factor-α and cardiovascular diseases. Ital

Effect of High vs Low Doses of Chloroquine Diphosphate as Adjunctive Therapy for Patients Hospitalized With Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Infection: A Randomized Clinical Trial

Toxicogenomics in risk assessment: Applications and needs

Epigenetic mechanisms underlying arsenic -associated lung carcinogenesis

A Trial of Lopinavir

Could IL-17 represent a new therapeutic target for the treatment and/or management of COVID-19-syndrome?

Lopinavir/ritonavir in the treatment of HIV-1 infection: A review

Cardiac Complications Attributed to Chloroquine and Hydroxychloroquine : A Systematic Review of the Literature

Identification of Differentially Expressed Genes and Signaling Pathways in Acute Myocardial Infarction Based on Integrated Bioinformatics Analysis

ToppGene Suite for gene list enrichment analysis and candidate gene prioritization

cytoHubba: Identifying hub objects and sub-networks from complex interactome

Role of lopinavir/ritonavir in the treatment of SARS: Initial virological and clinical findings

Integration of biological networks and gene expression data using cytoscape

Induction of proinflammatory cytokines (IL-1 and IL-6) and lung inflammation by Coronavirus-19 (COVI-19 or SARS-CoV-2): anti-inflammatory strategies

A systematic review on the efficacy and safety of chloroquine for the treatment of COVID-19

The Comparative Toxicogenomics Database: Update

The Comparative Toxicogenomics Database facilitates identification and understanding of chemical-gene-disease associations: arsenic as a case study

Nod2 sensing of lysozyme-digested peptidoglycan promotes macrophage recruitment and clearance of S. pneumoniae colonization in mice

Cardiovascular disease risk factors in HIV patientsassoci

An overview of IL-17 function and signaling

Breakthrough: Chloroquine phosphate has shown apparent efficacy in treatment of COVID-19 associated pneumonia in clinical studies

Hydroxychloroquine and azithromycin as a treatment of COVID-19: results of an open-label non-randomized clinical trial

Clinical and microbiological effect of a combination of hydroxychloroquine and azithromycin in 80 COVID-19 patients with at least a six-day follow up: A pilot observational study

Pyroptosis: The missing puzzle among innate and adaptive immunity crosstalk

Identification of environmental chemicals targeting miscarriage genes and pathways using the comparative toxicogenomics database

Interleukin-6 and Cardiovascular Diseases

Regulation of CCL2 expression in human vascular endothelial cells by a neighboring divergently transcribed long noncoding RNA

Cytochrome P450 2C8 and CYP3A4 / 5 are Involved in Chloroquine Metabolism in Human Liver Microsomes

International Reviews of Immunology Pathogen Recognition by the Innate Immune System

HIV protease inhibitors protect apolipoprotein B from degradation by the proteasome: A potential mechanism for protease inhibitor-induced hyperlipidemia

The Author"s response: Case of the index patient who caused tertiary transmission of coronavirus disease 2019 in Korea: The Application of lopinavir/ritonavir for the treatment of COVID-19 pneumonia monitored by quantitative RT-PCR

Hydroxychloroquine , a less toxic derivative of chloroquine , is effective in inhibiting SARS-CoV-2 infection in vitro

Elevated plasma level of selective cytokines in COVID-19 patients reflect viral load and lung injury

Cardiovascular risk and dyslipidemia among persons living with HIV: A review

cardiovascular outcomes and the physical and chemical properties of metal ions found in particulate matter air pollution: A QICAR study

Genetic variations in E-selectin and ICAM-1 : Relation to atherosclerosis

Computational studies of drug repurposing and synergism of lopinavir, oseltamivir and ritonavir binding with SARS-CoV-2 protease against COVID-19

Inflammation in the pathogenesis of microvascular complications in diabetes

Hydroxychloroquine and Chloroquine for Treatment of SARS-CoV-2 (COVID-19)

NOD proteins: Regulators of inflammation in health and disease

Traditional Chinese medicine for COVID-19 treatment

Zika virus cell tropism in the developing human brain and inhibition by azithromycin

Angiotensin receptor blockers and COVID-19

NF-κB and virus infection: Who controls whom

Personal view Antiviral effects of chloroquine Effects of chloroquine on viral infections : an old drug against today " s diseases ? J o u r n a l P r e -p r o o f Personal view

Cytoscape: A Software Environment for Integrated Models of Biomolecular Interaction Networks

A Review of Coronavirus Disease-2019 (COVID-19)

COVID-19 , immune system response , hyperinflammation and repurposing antirheumatic drugs

Leveraging complementary computational models for prioritizing chemicals of developmental and reproductive toxicity concern : an example of food contact materials

The COVID-19 epidemic

Association between CYP3A4 gene rs4646437 polymorphism and the risk of hypertension in Chinese population : a case -control study

Up-regulation of IL-6, IL-8 and CCL2 gene expression after acute inflammation: correlation to clinical pain

Toxicogenomics and systems toxicology: Aims and prospects

WHO, 2020a. Coronavirus disease 2019. World Heal. Organ. 2019, 2633

World Health Organization): Coronavirus disease (COVID-2019) situation reports

Chloroquine treatment influences proinflammatory cytokine levels in systemic lupus erythematosus patients

The outbreak of COVID-19: An overview

COVID-19 -what should anaethesiologists and intensivists know about it?

IL-17 activates the canonical NF-κB signaling pathway in autoimmune B cells of BXD2 mice to upregulate the expression of regulators of G-protein signaling 16

In vitro antiviral activity and projection of optimized dosing design of hydroxychloroquine for the treatment of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)

This work is supported by the Ministry of Education, Science and Technological Development, Republic of Serbia under the call for COVID-19 investigation (No. 451-03-68/2020-14/200161).

This research did not receive any specific grant from funding agencies in the public, commercial, or notfor-profit sectors.

J o u r n a l P r e -p r o o f 

☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.☐The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: J o u r n a l P r e -p r o o f