Sulfasalazine and Proguanil reduced SARS-CoV-2 viral replication in cells

0
752

Cambridge scientists have identified 200 approved drugs predicted to work against COVID-19 – of which only 40 are currently being tested in COVID-19 clinical trials.

In a study published today in Science Advances, a team led by researchers at the University of Cambridge’s Milner Therapeutics Institute and Gurdon Institute used a combination of computational biology and machine learning to create a comprehensive map of proteins that are involved in SARS-CoV-2 infection – from proteins that help the virus break into the host cell to those generated as a consequence of infection.

By examining this network using artificial intelligence (AI) approaches, they were able to identify key proteins involved in infection as well as biological pathways that might be targeted by drugs.

To date, the majority of small molecule and antibody approaches for treating COVID-19 are drugs that are either currently the subject of clinical trials or have already been through clinical trials and been approved. Much of the focus has been on several key virus or host targets, or on pathways – such as inflammation – where a drug treatment could be used as an intervention.

The team used computer modeling to carry out a ‘virtual screen’ of almost 2,000 approved drugs and identified 200 approved drugs that could be effective against COVID-19. Forty of these drugs have already entered clinical trials, which the researchers argue supports the approach they have taken.

Data-driven computational approaches for identifying drug repurposing targets for COVID-19. Credit: Winnie Lei

When the researchers tested a subset of those drugs implicated in viral replication, they found that two in particular – an antimalarial drug and a type of medicine used to treat rheumatoid arthritis – were able to inhibit the virus, providing initial validation of their data-driven approach.

Professor Tony Kouzarides, Director of the Milner Therapeutics Institute, who led the study, said: “By looking across the board at the thousands of proteins that play some role in SARS-CoV-2 infection – whether actively or as a consequence of infections—we’ve been able to create a network uncovering the relationship between these proteins.

“We then used the latest machine learning and computer modeling techniques to identify 200 approved drugs that might help us treat COVID-19. Of these, 160 had not been linked to this infection before. This could give us many more weapons in our armory to fight back against the virus.”

Using artificial neural network analysis, the team classified the drugs depending on the overarching role of their targets in SARS-CoV-2 infection: those that targeted viral replication and those that targeted the immune response. They then took a subset of those involved in viral replication and tested them using cell lines derived from humans and from non-human primates.

Artificial Neural Network learned relationships between drugs and their target proteins in the training dataset to predict important mechanism of action, Credit: Winnie Lei

Of particular note were two drugs, sulfasalazine (used to treat conditions such as rheumatoid arthritis and Crohn’s disease) and proguanil (and antimalarial drug), which the team showed reduced SARS-CoV-2 viral replication in cells, raising the possibility of their potential use to prevent infection or to treat COVID-19.

Dr. Namshik Han, Head of Computational Research and AI at the Milner Therapeutics Institute, added: “Our study has provided us with unexpected information about the mechanisms underlying COVID-19 and has provided us with some promising drugs that might be repurposed for either treating or preventing infection.

While we took a data-driven approach – essentially allowing artificially intelligent algorithms to interrogate datasets – we then validated our findings in the laboratory, confirming the power of our approach.

“We hope this resource of potential drugs will accelerate the development of new drugs against COVID-19. We believe our approach will be useful for responding rapidly to new variants of SARS-CoV2 and other new pathogens that could drive future pandemics.”


The global outbreak of SARS-CoV-2 necessitates the rapid development of new therapies against COVID-19 infection. Here, we present the identification of 200 approved drugs, appropriate for repurposing against COVID-19. We constructed a SARS-CoV-2-induced protein (SIP) network, based on disease signatures defined by COVID-19 multi-omic datasets(Bojkova et al., 2020; Gordon et al., 2020), and cross-examined these pathways against approved drugs.

This analysis identified 200 drugs predicted to target SARS-CoV-2-induced pathways, 40 of which are already in COVID-19 clinical trials(Clinicaltrials.gov, 2020) testifying to the validity of the approach. Using artificial neural network analysis we classified these 200 drugs into 9 distinct pathways, within two overarching mechanisms of action (MoAs): viral replication (130) and immune response (70).

A subset of drugs implicated in viral replication were tested in cellular assays and two (proguanil and sulfasalazine) were shown to inhibit replication. This unbiased and validated analysis opens new avenues for the rapid repurposing of approved drugs into clinical trials.

Artificial neural network analysis uncovers drug mechanisms of action

To investigate the mechanism of action (MoA) for the 200 drugs in the context of COVID-19, we used Self-Organizing Map (SOM), a type of artificial neural network, to analyse the relationship between the 200 drugs and the 148 key pathways (termed drug-pathway association).

After the unsupervised training of SOM, the distance between the adjacent neurons (pathways) was calculated and presented in different coloured hexagons, which illustrates the probability density distribution of data vectors (drug-pathway association score) (Vesanto and Alhoniemi, 2000) (Figures 3A and S5). Based on the distance, we applied the Davies-Bouldin (DB) index to separate the key pathways into 9 clusters (Figure 3B).

These clusters of pathways and drugs identified two MoA categories: (1) virus replication (VR) and (2) immune response (IR) (Figure 3C). The SOM also mapped 200 drugs into each neuron (the number of drugs per neuron is shown in Figure 3D and drug names are shown in Figure 3E). Notably, 30 out of the 40 drugs that are in COVID-19 clinical trials(Clinicaltrials.gov, 2020) were in the VR MoA category while only 10 drugs were in the IR (Figure 3D).

Finally, we identified mechanistic roles and connections for the 200 drugs and their target proteins, and mapping the drugs into 9 pathway clusters (Figure 3E). A more extensive analysis of information about each drug is given in Table S2.

Figure 3.Machine learning predicts mechanisms of actions for drug repurposing candidates.(A) A unified distance matrix (U-matrix) is shown of the trained unsupervised SOM used to analyse the relationship between the 200 drugs and the 148 key pathways. This contains the distance (similarity) between the neighbouring SOM neurons (pathways) and shows data density (drug-pathway association scores) in input space. Each subunit is coloured according to distance between corresponding data vectors of neighbour neurons, with low distances areas (dark blue) indicating high data density (clusters).
(B) Cluster solution chosen based on U-matrix and Davies-Bouldin (DB) index to separate the key pathways into 9 clusters. Clusters of each SOM neuron are distinguishable by colour. The size of the black hexagon in each neuron indicates distance. Larger hexagons have a low distance to neighbouring neurons, hence forming a stronger cluster with neighbours.
(C) Two MoA categories identified based on the pathway clustering and the drug mapping.
(D) Mapping of the 200 identified drugs to each neuron (pathway) based on matching rates and inspection of examples from each cluster.
(E) A SOM component map shows mapping results of the 200 drugs into 9 pathway clusters. The names of 9 clusters are shown in the figure, and the drugs with asterisk are already in COVID-19 clinical trials.

We next sought to identify the precise proteins, within the SIP network, targeted by each of the 200 drugs. We found that of the 1,573 proteins targeted by the 200 drugs, most (66%) are targeted by a single drug (Figure S6A). However, there are 30 proteins (0.19%) that are targeted by 8 or more drugs (Figure S6A). To establish whether there is a pathway relationship between these 30 proteins, we interrogated their molecular function.

Figure S6B shows that the most enriched categories of function for these proteins were heme, microsome, oxidoreductase and monooxygenase, all of which are related to nicotinamide adenine dinucleotide phosphate (NADP) and nitric oxide (NO) synthesis. As NO is important for viral synthesis (and because NADP affects NO production), this could provide a potential mechanism by which these drugs might alter viral infection(Kwiecien et al., 2014; Lind et al., 2017; Wang et al., 2006). Based on these findings we decided to validate in cellular assays, five drugs (Ademetionine, Alogliptin, Flucytosine, Proguanil and Sulfasalazine) with good safety profiles which are functioning within this pathway.

Two drugs that target NO production reduce SARS-CoV-2 replication
To assess whether these five drugs are able to reduce SARS-CoV-2 infection, we performed an initial screening using the Vero E6 cell line, where we observed that 2 of the 5 drugs, Proguanil and Sulfasalazine, showed significant antiviral effects without any noticeable cellular toxicity at the indicated doses (Figures 4A and S7A). We then focused on these two drugs, expanding our validation using 2 different cellular models (Vero E6 and Calu-3).

Treatment of Vero E6 and Calu-3 cells with Proguanil and Sulfasalazine illustrated strong anti-SARS-CoV-2 effects (represented by reductions of the envelope and nucleocapsid gene RNAs) in a dose dependent manner, mirroring the results of the initial screen (Figures 4B-E, S7B-E). Importantly, no significant effect on cellular viability was observed at any tested dose (Figures S7F-H). The effective concentration of sulfasalazine is comparable to maximal plasma concentrations achieved routinely in patients with rheumatoid arthritis or inflammatory bowel disease(IARC working Group on the Evaluation of Carcinogenic Risk to Humans, 2016).

Figure 4.
Figure 4.Proguanil and Sulfasalazine reduce SARS-CoV-2 replication and the p38/MAPK Signalling Activity.(A) qRT-PCR analysis of the indicated mRNA (Envelope, E-protein) from Vero E6 cells pre-treated with the indicated drugs and concentrations for three hours prior to infection with SARS-CoV-2 for 24 hours. Statistical test: Student’s t test. Mean + S.D. of three independent replicates is shown.
(B, C) RT-qPCR analysis of indicated mRNA (Envelope, E-protein) from Vero E6 cells pre-treated with Proguanil or Sulfasalazine at indicated concentrations for three hours prior to infection with SARS-CoV-2 for 24 hours. Statistical test: Student’s t test. Mean + S.D. of three independent replicates is shown. (D, E) RT-qPCR analysis of indicated mRNA (Envelope, E-protein) from Calu-3 cells pre-treated with Proguanil or Sulfasalazine at indicated concentrations for three hours prior to infection with SARS-CoV-2 for 24 hours. Statistical test: Student’s t test. Mean + S.D. of three independent replicates is shown.
(F) Western blot analysis of phosphorylated MAPKAPK-2 (Thr334) in Mock, DMSO, Sulfasalazine or Proguanil-treated Vero E6 cells at indicated concentrations for three hours prior to infection with SARS-CoV-2 for 24 hours.
(G-J) RT-qPCR analysis of the indicated mRNAs from Calu-3 cells pre-treated with Proguanil or Sulfasalazine at indicated concentrations for three hours prior to infection with SARS-CoV-2 for 24 hours. Statistical test: Student’s t test. Mean + S.D. of three independent replicates is shown.

To further investigate the anti-SARS-CoV-2 impact of these two drugs, we examined the status of recently discovered intracellular pathways directly associated with SARS-CoV-2 infection and cytokine production(Bouhaddou et al., 2020). Indeed, treatment with either Proguanil or Sulfasalazine significantly reduced the phosphorylation of MAPKAPK2 (p-MK2, T334) (Figure 4F), an important component of the p38/mitogen-activated protein kinase (MAPK) signalling pathway, which has been shown to be activated via SARS-CoV-2 infection and stimulate cytokine response(Bouhaddou et al., 2020).

Importantly, treatment of Calu-3 and Vero E6 cell lines with Proguanil and Sulfasalazine led to a significant downregulation of the mRNA of key cytokines (Figures 4G-J and S8), which are dictated by the p38/MAPK signalling pathway and shown to become elevated during SARS-CoV-2 infection and replication (CXCL3, IFNB1 and TNF-A). Hence, the above results solidify the promising anti-SARS-CoV-2 effects of the two drugs, both at the viral as well as the molecular level.

To understand why Sulfasalazine and Proguanil are effective against SARS-CoV-2 infection, but others functioning in the same pathway were not (Figure 4A), we looked more closely at the targets of each drug. Figure 5 shows that SARS-CoV-2 orf8 binds to gamma-glutamyl hydrolase (GGH) and regulates the synthesis of NO, which is necessary for viral synthesis.

An additional auxiliary pathway, mediating the synthesis of NADP, can also affect NO production, although indirectly. Sulfasalazine and Proguanil impinge on both of these pathways: Sulfasalazine targets the NFKB inhibitors NFKBIA and IKBKB as well as CYP450 enzymes, whereas Proguanil targets DHFR and CYP450 enzymes plus interacting partners. In this way these two drugs might more effectively target NO production and thus disrupt viral replication.

By contrast, the three drugs that were not effective against SARS-CoV-2 infection (Flucytosine, Alogliptin and Ademetionine) only affect one of the two pathways. This analysis thereby highlights the possibility that targeting NO production through multiple pathways may be the reason for the efficacy of Sulfasalazine and Proguanil in reducing viral replication.

Figure 5.
Figure 5.A schematics depicting the pathways mediating NO production that are targeted by the five tested drugsThe black boxes indicate key proteins in SIP network, and those targeted by the five drugs are highlighted in red colour. Sulfasalazine and Proguanil target proteins in both pathways that directly and indirectly (via NADP production) affect NO production (Choi et al., 2017; Corpas and Barroso, 2014; Hiscott et al., 2001; Wink et al., 2011).

More information: N. Han el al., “Identification of SARS-CoV-2–induced pathways reveals drug repurposing strategies,” Science Advances (2021). advances.sciencemag.org/lookup … .1126/sciadv.abh3032

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Questo sito utilizza Akismet per ridurre lo spam. Scopri come vengono elaborati i dati derivati dai commenti.