*3.3. Mutations in 2C Protein Associated with Resistance to Fluoxetine*

Since the 2C viral protein was reported as the target of fluoxetine antiviral activity, we investigated whether the resistance was associated with mutant variants. The sequence of the 2C region was determined in the stock virus, virus obtained from persistently infected cultures before treatment, and virus collected from persistently infected cultures treated with fluoxetine or DMSO treated for 4 and 10 weeks.

All the positions with amino-acid substitution in the different sequences as compared to CVB4 E2 reference strain (NCBI, accession: AF311939.1) are presented in Table 2.



ND: Not done, virus undetectable; I3 and I4 are resistant to fluoxetine treatment. AA changes are underlined.

Figure 3 focuses on mutations that appeared in the sequences of virus obtained from the fluoxetine-resistant persistently infected cultures I3-F and I4-F.

**Figure 3.** Amino-acid substitutions in fluoxetine-resistant virus. Virus suspensions were collected from persistent CVB4 E2-infected cultures at baseline, and after 4 and 10 weeks of treatment with DMSO or fluoxetine. The sequence of the whole CVB4 E2 2C region (from nt 4039 to nt 5025, 329 aa) was determined using Sanger method. The amino-acid substitutions in the sequences of fluoxetine-resistant viruses (I3 and I4) are shown.

Before fluoxetine treatment, the R296G mutation was observed in I3 and I4, but not in I1, I2, and the initial stock virus. This substitution was present during all follow-ups in the sequences from fluoxetine or DMSO-treated I3 and I4 infections.

During fluoxetine treatment, the I227V mutation was observed in I3-F at week 4 o.t., and in both I3-F and I4-F at week 10 o.t. The A133T emerged in both infections at week 10 o.t. As for mutations R188G and A229V, they were only observed in I3-F infection at week 10 o.t.

Interestingly, these mutations were not observed in the virus obtained from infected cultures treated with DMSO (I1-D, I2-D, I3-D, and I4-D).

#### **4. Discussion**

The investigation of existing drugs with well-established safety profiles for new indications is a cheaper and faster strategy to discover new antiviral agents. Indeed, the screening of approved molecule libraries allowed to identify previously unrecognized inhibitors of enterovirus replication, including fluoxetine [11,15].

We have previously shown that fluoxetine can successfully clear persistent CVB4 E2 infection within a month when cultures were treated at 5.48 μM, twice a week [12]. Further experiments described in this study revealed a failure to clear the virus in some persistent CVB4 E2 infections despite a long-term fluoxetine treatment. The lack of susceptibility to fluoxetine in these "resistant" isolates was confirmed in a model of acute infection using HEp-2 cells.

Fluoxetine and other described enterovirus inhibitors were shown to exert antiviral activity by targeting 2C protein that is one of the most conserved and complex nonstructural viral proteins among picornaviruses [11,15]. This protein was reported to be involved in several key events throughout the virus life cycle (different steps of replication, immune evasion ... ), but its precise role is not fully understood. 2C harbors an N-terminal membrane-binding motif, an adenosine triphosphatase (ATPase) domain, a cysteine-rich motif, and RNA binding sites [16]. The ATPase domain, which belongs to SF3 helicases of the AAA+ ATPase superfamily, contains Walker motifs (motifs A and B) and motif C [17]. ATPase activity has been earlier demonstrated for 2C protein [8]. However, every attempt to determine the helicase activity associated with 2C ATPase has failed until recently when a study provided evidence of this helicase activity in the 2C protein of EV-A71 and CV-A16 [18].

One of the major challenges to overcome during the investigation or the development of an antiviral agent, especially a direct acting agent, is the emergence of resistant mutants. This is particularly true for RNA viruses that usually generate a significant number of mutations during the replication process due to a poor proofreading activity of RNA polymerase [19].

Our model of persistent CVB4 E2 infection in pancreatic cell cultures is attractive to investigate the resistance to antiviral molecules, because it allows multiple and long-term exposition of virus to molecules. This drug pressure promotes the emergence of resistant mutants, which more or less quickly results in failure of virus clearance, depending on the resistance barrier of the drug.

Previous studies have described three residues substitutions in 2C protein (A224V, I227V, A229V), that confer CVB3 resistance to fluoxetine, TBZE-029, GuHCl, and other recently identified 2C targeting inhibitors [8,15]. These substitutions are present in a short stretch of amino acids 224AGSINA229 located immediately at the C terminal of ATPase motif C. The AGSINA motif was found to be conserved in *Enterovirus B* (such as CVB4) and *D* species but is not present in other enteroviruses [10].

In this study, we found two of these substitutions in the fluoxetine-resistant viral suspensions (both I227V and A229V (double population mutant/wild type) in I3-F, and only I227V in I4-F). These changes were not observed before treatment, therefore, have probably emerged under drug-selection pressure. However, given the limit of Sanger sequencing, it cannot be excluded that these mutants preexisted as minor variants. The fluoxetine-resistant mutant used in reported studies (obtained by site-directed mutagenesis) harbored all these three mutations, and the impact of each mutation could not be clearly assessed [10,15]. In addition, the fact that, resistant viruses obtained from persistent infections treated with fluoxetine (I3-F and I4-F) were still susceptible to GuHCl, shows that the effect of these mutations (alone or combined) might depend on the 2C-inhibitor tested. Indeed, de Palma et al. [20]. previously reported a detailed impact of these mutations on the susceptibility to GuHCl. I227V alone did not induce resistance while the combination of I227V + A229V was associated with low-level resistance. Thus, the mutations observed in I3-F and I4-F seems to be insufficient to confer resistance to GuHCl in vitro.

In this report, other mutations not previously described were observed in the 2C ATPase domain of the resistant viruses (A133T and R188G in I3-F, and A133T in I4-F). I3-F, the viral suspension with the most mutations, appears to be the most resistant one to fluoxetine.

Interestingly, the R296G mutation was observed before treatment only in the virus obtained from persistent infections, which would later display resistance to fluoxetine treatment (I3 and I4). Even if viral suspensions from all of the untreated persistently infected cultures (I1, I2, I3, and I4) were susceptible to fluoxetine, the role of R296G mutation in the predisposition of resistance to fluoxetine cannot be excluded. This substitution is located in the zinc finger domain, downstream of the cysteine-rich motif which forms a zinc-binding site [21]. In addition, this residue seems to be conserved in *Enterovirus B* and *D* species (the EV species reported to be susceptible to fluoxetine), and variable in others species. We hypothesized that this mutation might favor other compensatory mutations including those associated with resistance to fluoxetine.

In this study, we focused on the 2C protein; however, the role of changes in other parts of the viral genome cannot be excluded. Site-directed mutagenesis studies are needed to precisely analyze the impact of these new substitutions on the susceptibility of CVB to fluoxetine and other 2C targeting enterovirus inhibitors.

Currently, fluoxetine is not available as a treatment for EV in humans. Nevertheless, the understanding of the inhibition mechanism and resistance profiles can be useful for the design of new compounds [10,22].
