DisCoVering potential candidates of RNAi-based therapy for COVID-19 using computational methods

Datasets

The complete genome sequence of SARS-CoV-2 was obtained with accession No. NC045512.2 from the GenBank database (Benson et al., 2012; Wu et al., 2020b). The complete genome of betacoronavirus sequences from the NCBI database and sequences compiled by Ceraolo & Giorgi (2020) were considered for finding evolutionarily conserved regions.

In this study, we conducted three independent analyses for suggesting candidate miRNAs and one analysis for designing efficient siRNAs. The miRNAs and their interaction information in theses analysis are obtained from the following sources:

• In the first analysis (Analysis #1), the list of miRNAs and their interactions were collected from CLASH (Helwak et al., 2013). This database comprises 7,390 experimentally validated interactions between 270 miRNAs and 7,390 mRNAs (Plotnikova, Baranova & Skoblov, 2019). This database contains the free binding energy of interactions as well as the sequences of miRNAs and mRNAs involving in the interactions.

• In the second analysis (Analysis #2), the sequences of 539 bat-specific miRNAs were obtained from the previously published article by Huang, Jebb & Teeling (2016). Since the interaction between bat-specific miRNAs and human mRNAs are not provided experimentally, we extracted the human mRNA targets for bat-specific miRNA using mirDB (Chen & Wang, 2020) tool.

• In the third analysis (Analysis #3), the known miRNAs of 286 species were downloaded from the miRNA registry, mirBase (Kozomara, Birgaoanu & Griffiths-Jones, 2019). mirBase is the most comprehensive database for miRNAs (Drury, O’Connor & Pollard, 2017) and comprises 6,133 miRNAs. However, it does not contain the miRNA-mRNA interactions. We used mirDB (Chen & Wang, 2020) tool for obtaining the human mRNA targets for candidate miRNAs from mirBase.

• The fourth analysis (Analysis #4) is conducted for proposing new potential siRNAs, for which siDirect 2.0 (Naito et al., 2009) tool is used.

Methodology

All analyses in this study are conducted with the aim of either recommending potential miRNAs or proposing efficient siRNAs. These two procedures are described in the “Recommending Potential miRNA” and “Proposing Efficient siRNA”. The overall workflow is shown in Fig. 1. In recommending efficient miRNAs and siRNAs, we aim to target UCRs. Thus, the UCRs of SARS-CoV-2 must be determined.

Analysis #1: Suggesting candidate miRNAs from the human body for the treatment of COVID-19

To evaluate the efficiency of available miRNAs in the human body for targeting UCRs, we analyzed the CLASH database. CLASH is a gold standard dataset that contains experimentally validated interactions between human miRNAs and human mRNAs.

The goal of this analysis was to find appropriate available miRNAs that have efficient interactions with UCRs and low-affinity to human mRNAs. For each miRNA ( $\mathcal{M}$) in CLASH and each UCR ( $\mathcal{U}$) in SARS-CoV-2, the free binding energy of the hybridization of $\mathcal{M}$ and $\mathcal{U}$ is denoted by $E_{(\mathcal{M},\mathcal{U})}$ and computed by IntaRNA (Mann, Wright & Backofen, 2017). Moreover, the free binding energy of targeting each human mRNA ( $\mathcal{R}$) by miRNA ( $\mathcal{M}$) is denoted by $E_{(\mathcal{M},\mathcal{R})}$ is retrieved from CLASH and computed by IntaRNA. Among all mRNAs in human, the mRNA with the highest affinity to bind with miRNA is considered as the most potent rival in competition with UCR to be bound by miRNA. The efficiency score for targeting UCR by miRNA (denoted by $S_{(\mathcal{M},\mathcal{U})}$) is calculated as (1).

(1) $$S_{(\mathcal{M},\mathcal{U})}=E_{(\mathcal{M},\mathcal{U})}-\underset{\mathcal{R}}{min}\{E_{(\mathcal{M},\mathcal{R})}\}$$The negative efficiency score suggests that the affinity of miRNA to target that UCR is greater than targeting any mRNA in the body. Whereas the positive efficiency score denotes that there is at least one mRNA in the body that the miRNA has more propensity to target that mRNA, rather than targeting the UCR. Therefore, the lower the efficiency score is, the more favorable candidate the miRNA is. The analyzed miRNAs were sorted based on their efficiency scores, and the top 10 were considered as the candidate miRNAs. Using either the energies reported in CLASH or energies computed by IntaRNA result in the same list of candidate miRNAs.

Table 1 represents the most favorable candidate miRNAs. This table includes the information of candidate miRNAs, the information about target UCRs in SARS-CoV-2, and the most potent rival mRNAs. It can be seen that the candidate miRNA has more propensity to bind to UCR, rather than the rival mRNA. The entire report of this analysis, containing the interaction information for all miRNA in CLASH and all UCRs are presented in Supplemental Materials.

The efficiency score of top 4 candidate miRNAs are less than −2. Thus, we conducted further analyses on these four miRNAs to investigate their targets in SARS-CoV-2. The hybridization of targeting UCRs by candidate miRNAs are shown in Fig. 2.

Among the potential efficient miRNAs, both hsa-miR-374a-5p and hsa-miR-1-3p have the most affinity to bind nsp11 in SARS-CoV-2. Protein nsp11 is an endoribonuclease that inhibits the production of IFN-β (Shi et al., 2011). The first essential factor of innate immune response to the viral infections is IFN-β (Weber, Kochs & Haller, 2004). Defeating the viral infection by host cells becomes more difficult due to the interferon antagonist property of nsp11 (Weber, Kochs & Haller, 2004; Bowie & Unterholzner, 2008). Hence, targeting nsp11 facilitates the immune response against SARS-CoV-2.

The second candidate miRNA is hsa-miR-548b-3p, which binds to ORF6 with low free binding energy. ORF6 is an accessory protein with key functions in viral pathogenesis (Yoshimoto, 2020; Kumar et al., 2007; Zhao et al., 2009). This protein has the most powerful suppression of interferon production and interferon signaling (Yuen et al., 2020). Moreover, it plays role in promoting RNA polymerase activity by interacting with nsp8 (Kumar et al., 2007). Consequently, silencing ORF6 is a promising approach in designing live-but-attenuated vaccines against SARS-CoV-2 (Yuen et al., 2020).

The fourth candidate miRNA, hsa-miR-224-5p, has the best binding with nsp6 region. Protein nsp6 produces autophagosomes, which assists in the assembly of replicase proteins (Yoshimoto, 2020). In addition, the function of nsp6 in limiting autophagosome/lysosome expansion, which leads to both inducing autophagy of host cells and preventing the autophagy of viral components (Lippi et al., 2020). By this way, nsp6 impedes the degradation of the viral components in lysosomes by prohibiting autophagosomes expansion (Yoshimoto, 2020; Cottam, Whelband & Wileman, 2014). As a result, down-regulating nsp6 improves host immunity by saving host cells from autophagy and degrading viral components.

Moreover, the most probable target genes for these miRNAs were obtained from CLASH (Helwak et al., 2013), and their related gene ontology (GO) and pathways were obtained from Uniprot (Apweiler et al., 2004). Figure 3 represents the GO biological process (BP), molecular function (MF), and cellular component (CC) of the target genes related to the top 4 miRNAs, as well as their corresponding pathways.

Suggesting candidate miRNAs from the bats for the treatment of COVID-19

Bats are one of the most probable sources of SARS-CoV-2, and previous studies have shown that bats are more resistant to CoV infections (Hoffmann et al., 2013). Therefore, this idea comes up that maybe the bat-specific miRNAs have a functional role in their resistance against CoVs. The bat-specific miRNAs were investigated for efficient binding to UCRs to evaluate this hypothesis. The free binding energy for the hybridization of UCRs and bat-specific miRNAs were computed by IntaRNA (Mann, Wright & Backofen, 2017) tool. Since there is no gold standard for the interactions and free binding energy of bat-miRNAs and human mRNAs, the candidate miRNAs were selected based on the energy ratio. The energy ratio for each pair of miRNA $(\mathcal{M})$ and UCR $(\mathcal{U})$ is computed as formula (2).

(2) $$E{R}_{(\mathcal{M},\mathcal{U})}=E_{(\mathcal{M},\mathcal{U})}/E_{(\mathcal{M},\overline{\mathcal{M}})}$$where $E_{(\mathcal{M},\overline{\mathcal{M}})}$ is the free binding energy of the miRNA to its completely complement sequence. The bat-specific miRNAs were sorted based on their energy ratio, and the top 20 were selected as candidate miRNAs. The human mRNA targets for the candidate miRNAs were predicted using mirDB (Chen & Wang, 2020) tool. Table 2 represents the most favorable candidate miRNAs in bat, as well as their target UCR and their most probable predicted human mRNA by mirDB. The complete list of energy ratios for all bat-specific miRNAs and UCRS, as well as the entire list of predicted target genes for the selected bat-specific miRNAs are provided in Supplemental Materials.

Figure 4 displays the hybridization of targeting UCRs by candidate bat-specific miRNAs. The first three suggested bat-specific miRNAs target UCRs in nsp3, nsp4, and nsp6. These three proteins have transmembrane domains (Sakai et al., 2017). There is a fundamental relation between CoV infections and the nsp3 association (Khailany, Safdar & Ozaslan, 2020). Protein nsp3 is the most extended protein in SARS-CoV-2 with multiple domains, which plays crucial roles in forming the replication/transcription complex (RTC) (Lei, Kusov & Hilgenfeld, 2018). Due to essential protease activity for releasing proteins with viral activity, down-regulating nsp3 can be a desirable goal for antiviral activity (Báez-Santos, John & Mesecar, 2015). Protein nsp4 has an important interaction with nsp3, which induces membrane rearrangement and viral replication (Yoshimoto, 2020). This interaction is a critical factor in viral replication via the rearrangements of host-derived membranes. Elucidating this interaction leads to terminate SARS-CoV-2 replication (Sakai et al., 2017). Another transmembrane protein, nsp6, together with nsp3 and nsp4 proteins, configure the organelle-like replicative structures (double-membrane vesicles) (Cárdenas-Conejo et al., 2020). Making use of three first suggested miRNAs that target nsp3, nsp4, and nsp6 may strongly down-regulates the viral replication.

The next three suggested miRNAs have favorable affinities to bind the spike protein (surface glycoprotein). The entry of SARS-CoV-2 into cells is mediated by spike (S) glycoproteins, which binds to human angiotensin-converting enzyme 2 (ACE2) for cell entry (Hoffmann et al., 2020; Yoshimoto, 2020). The spike protein identifies the ACE2 protein on the surface of the host cell (Lan et al., 2020; Shang et al., 2020; Walls et al., 2020). The transmembrane spike (S) glycoprotein forms the homotrimers protruding from the viral surface (Tortorici & Veesler, 2019). This glycoprotein contains two subunits S1 and S2 that are essential for receptor binding and membranes fusion, respectively (Walls et al., 2016; Park et al., 2016; Kirchdoerfer et al., 2016). The S2 subunit is fusion machinery that facilitates the viral and cellular membranes fusion (Gui et al., 2017; Song et al., 2018; Yuan et al., 2017; Hoffmann et al., 2020). Since the surface glycoprotein S intervenes the virus entry into host cells, there is a major focus on this protein in therapeutic strategies and vaccine design (Walls et al., 2020; Hoffmann et al., 2020). Thus, the suggested miRNAs may have great antivirus activities via down-regulating the surface glycoproteins.

To analyze the possible side effect of using the candidate bat miRNAs in human body, we enriched their predicted targets in GO using UniProt (Apweiler et al., 2004). Figure 5 demonstrates the GO terms and Reactome pathways related to the most probable target mRNAs for the candidate bat-specific miRNAs.

Analysis #3: Suggesting potential miRNAs from 286 species for the treatment of COVID-19

The previous analyses suggested the favorable candidate miRNAs from human and bat that have a potential role in the treatment of COVID-19. Nevertheless, other species may contain miRNAs that reveal promising antiviral activities against SARS-CoV-2. To investigate this idea, we considered the miRNA sequences of all species available in mirBase (Kozomara, Birgaoanu & Griffiths-Jones, 2019). To choose the most effective miRNAs, we conducted the same analysis as the previous section, but on the miRNAs of numerous species. The selected miRNAs from all species are represented in Table 3. The entire list of analyzed miRNAs, as well as the list of the predicted targets for miRNAs are provided in Supplemental Materials.

The first candidate miRNA is for the red flour beetle (Tribolium castaneum) and targets a UCR in ORF9ab, which encodes nucleocapsid protein in SARS-CoV-2. It binds to the viral genome and provides stability for virus (Yoshimoto, 2020). It also has high expression during infection, which enables antibody responses (Peng et al., 2006). This protein is essential in viral RNA transcription and replication (Kang et al., 2020). Therefore, it is a promising target in UCRs for reducing viral transcription and replication.

Other candidate miRNAs are from gray short-tailed opossum (Monodelphis domestica), sea vase (Ciona intestinalis), red junglefowl (Gallus gallus), human (homo sapiens), and Lotus japonicus. These miRNAs target UCRs in nsp3, nsp4, and spike protein. As it is mentioned in “Suggesting Candidate miRNAs from the Human Body for the Treatment of COVID-19”, these regions are of high importance for viral transcription, replication, and binding to host cells. The suggested miRNAs from various species are potentially promising alternatives for targeting these important regions in SARS-CoV-2 and lowering its functionality and spread.

The hybridization of UCRs and selected miRNAs are displayed in Fig. 6. Furthermore, an investigation for possible side effects of using these miRNAs in human body was conducted by checking the GO terms and Reactome pathways of the rival mRNAs. The list of GO terms and Reactome pathways were obtained from Uniprot (Apweiler et al., 2004) and are exhibited in Fig. 7. Among the candidate miRNAs, tca-miR-6014-3 and hsa-miR-4500 are more favorable choices, since their molecular functions and pathways are related to defeating virus and improving immunity.

Analysis #4: Proposing efficient siRNA for the treatment of COVID-19

An efficient RNA-intervention technique is designing new efficient siRNAs. Since the length for siRNA is 21 nt, we considered all UCRs with minimum length of 21 nt. Then, the selected UCR sequences were given to the siDirect tool (Naito et al., 2009) for designing efficient siRNA to target these regions. Five siRNAs targeting four UCRs were found. The information of UCRs that are targetted by the designed siRNA is presented Table 4. The designed siRNA are listed in Table 5

The minimum free energy (MFE) foldings of these five siRNAs were obtained by RNAfold in the ViennaRNA package (Lorenz et al., 2011). The obtained foldings are depicted in Fig. 8. Moreover, the hybridization of the designed siRNA with the target UCR is gained using IntaRNa tool (Mann, Wright & Backofen, 2017) and shown in Fig. 9. In order to investigate the potential side effects, the most potent target mRNAs were calculated using the mirDB tool (Chen & Wang, 2020). The most likely targets for siRNA#1, siRNA#2, siRNA#3, siRNA#4, and siRNA#5 are ENSG00000213047, ENSG00000169241, ENSG00000070882, ENSG00000118263, and ENSG00000157106, respectively. The retrieved genes were inquired into Unitprot (Apweiler et al., 2004) for analyzing the Go terms and Reactome pathways. The GO terms and pathways related to the target genes of each siRNA are presented in Fig. 10.