**3. Results**

#### *3.1. ProTide Library Design*

To investigate the activity of a set of nucleotide analogue ProTides against ZIKV and obtain basic structure-activity relationships, we selected a set of nucleotides designed to inhibit viral RdRPs by three different mechanisms: obligate chain termination, nonobligate chain termination, and lethal mutagenesis (Figure 1) [40,41]. Past work—and the successful results of drug development against HCV—shows that nucleosides alone lack adequate potency owing to slow phosphorylation in the target cells [42]. McGuigan's nucleotide aryloxy phosphoramidates include amino acid esters and are well established; sofosbuvir includes this functional group, and we chose it as our starting point [9,43]. One potential drawback of the McGuigan design is that unmasking is initiated by the action of carboxyesterases, which are highly active in the hepatocytes targeted by HCV but less active in the neural cell targets of ZIKV [42]. For that reason, we chose to investigate a recently developed methylthioethyl tryptamine ProTide from Wagner, which is unmasked by chemical cleavage of methylthioethanol followed by tryptamine removal by HINT1 [26]. HINT1 also catalyzes P-N bond cleavage in the classic McGuigan ProTide. Our hypothesis was that the carboxyesterase-independent unmasking of the methylthioethyl tryptamine ProTide would provide greater potency against ZIKV in human neural stem cells.

The selection of nucleoside analogues for testing focused on ribose modifications known to induce chain termination in vitro when used as a nucleoside triphosphate analogue with the viral RdRP. Some of the selected compounds, e.g., the 3- -deoxy and the 3- -*O*-methyl, are obligate chain terminators, but phosphorylation of this class of compounds in cells is often inefficient [18,19]. Better success in antiviral drug development has been found when using nonobligate terminators, especially 2- modifications that alter the nucleoside conformation and prevent extension after incorporation during RNA synthesis [14,31]. We included sofosbuvir in this set as a point of reference, along with ProTides of the 2- -*C*-methyl ribosides, which have known potency against the Flaviviridae family [19,31,44]. Also included were new ProTides of 2- -*C*-ethynyl- and 2- -*C*-ethenyluridine. The 2- -*C*-ethynyl ribosyl triphosphates are known sub-μM inhibitors of flavivirus RdRPs [31,45], and we prepared the new 2- -*C*-ethenyluridine ProTide to probe the active site of the RdRP and the metabolic consequences of further modification of the 2- -*C-*β position. A 5-fluorouridine ProTide was included in our set of analogues to test inhibition by lethal mutagenesis, although toxicity of mutagenic nucleoside analogues is typically prohibitive of clinical applications.

**Figure 1.** Design and structure of nucleotide analogue inhibitors tested against Zika virus (ZIKV) in this study.

#### *3.2. Activity against ZIKV and Toxicity of Nucleoside Analogue ProTides in Human Neural Stem Cells*

To profile the activity and toxicity profiles of our library of ProTides, we screened compounds in a 20-point, two-fold dose response with a highest concentration of 50 μM against ZIKV PRVABC59/Human/2015/Puerto Rico and ZIKV H/PAN/2016/BEI-259634 in neural stem cells (MOI of 10) using a luminescent cell-viability assay (CellTiter Glo) and determined EC50 (effective concentration 50) and CC50 (cytotoxic concentration 50) values from these data (Table 1). We observed that the compounds with the best selectivity indexes (SIs) were the 2- -*C*-methyluridine aryloxyl phosphoramidate ProTide (EC50PRVABC59: 1 μM, EC50H/PAN: 2 μM, CC50: 47 μM, SIs: >47, >23, respectively) and the 2- -*C*-ethynyluridine aryloxyl phosphoramidate ProTide (EC50PRVABC59: 0.3 μM, EC50H/PAN: 0.8 μM, CC50: >50 μM, SIs: >166, >62, respectively). These outperformed sofosbuvir in terms of anti-ZIKV activity (EC50PRVABC59: 35 μM, EC50H/PAN: 46 μM, CC50: >50 μM, SIs: >1.4, >1.1, respectively). Activity was observed for the 2- -*C*-methyladenosine aryloxyl phosphoramidate ProTide (EC50PRVABC59: 10 μM, EC50H/PAN: 18 μM, CC50: 42 μM, SIs: >4, >2, respectively) and 2- -*C*-methylcytidine aryloxyl phosphoramidate ProTide against PRVABC59 strain (EC50PRVABC59: 48 μM, CC50PRVABC59: >50 μM, SI: >1), albeit at lower levels. None of the other ProTides tested displayed activity. Interestingly, the 2- -*C*-ethenyluridine aryloxyl phosphoramidate ProTide was completely inactive and quite toxic to the host cells, suggesting that this functional group may be more promiscuous in terms of target binding or reactive with cellular components [46]. Replacement of the classic McGuigan ProTide masking group with methylthioethyl tryptamine phosphoramidate for the 2- -*C*-methyluridine ProTide completely abolished activity. This result was not expected because the classic McGuigan ProTides are unmasked with the involvement of both a carboxyesterase and HINT1, whereas the methylthioethyl tryptamine phosphoramidate requires only HINT1 in addition to spontaneous, chemical steps [26]. Clearly, cell line specificity for ProTide strategies must be considered in the design of nucleotide analogue inhibitors of ZIKV.


**Table 1.** Anti-ZIKV activity and toxicity of nucleoside analogue ProTides in neural stem cells.


**Table 1.** *Cont.*

<sup>1</sup> SI = selectivity index: CC50/EC50 for a given viral strain. Data values are the represented ± standard error (SE), *n* = 4.

#### *3.3. Protection from ZIKV-Induced Cytopathic E*ff*ect by 2*- *-C-Methyluridine Aryloxyl Phosphoramidate ProTide and Sofosbuvir in a ZIKV-Sensitive Glioblastoma Stem Cell Line*

We and others have previously shown that certain lines of brain tumors with stem cell-like genetic programs, such as glioblastomas, are highly susceptible to ZIKV infection [29,47–49]. These cells display a striking phenotypic change upon infection. To highlight the ability of a representative ProTide hit from our library, 2- -*C*-methyluridine aryloxyl phosphoramidate, to protect against ZIKV-induced cytopathic effect, we treated the glioblastoma stem cell line GSC 387 with either 10 or 30 μM 2- -*C*-methyluridine aryloxyl phosphoramidate and compared this to 10 or 30 μM sofosbuvir or the DMSO vehicle control treatment. Samples were prepared in both ZIKV-infected (MOI of 10) and uninfected conditions. After 72 h post-infection, bright-field microscopy images were obtained for each sample and are shown in Figure 2. While both compounds were able to protect against the cytopathic effect at 30 μM (intact cell foci were seen), the 2- -*C*-methyluridine aryloxyl phosphoramidate ProTide was better able to protect against ZIKV-induced cytopathic effect at the lower concentration of 10 μM than sofosbuvir, which is in line with our EC50 data in neural stem cells. These data suggest that monosubstituted 2- -*C*-modified uridylate aryloxyl phosphoramidate ProTides may be an attractive chemical series to pursue for more potent and selective ZIKV polymerase inhibitors than sofosbuvir.

**Figure 2.** Protection from ZIKV-induced cytopathic effect by 2- -*C*-methyluridine aryloxyl phosphoramidate ProTide and sofosbuvir in a ZIKV-sensitive glioblastoma stem cell model. DMSO-treated GSC 387 cells [-ZIKV and +ZIKV, H/PAN multiplicity of infection (MOI) 10], 10 or 30 μM sofosbuvir-treated cells or 10 or 30 μM 2- -*C*-methyluridine aryloxyl phosphoramidate ProTide (2- -CMU ProTide)-treated cells were incubated for 72 h at 37 ◦C and 5% CO2, and cell foci were subsequently imaged in bright-field.

#### *3.4. Repression of ZIKV Titers by 2*- *-C-Methyluridine Aryloxyl Phosphoramidate ProTide and Sofosbuvir in Neural Stem Cells*

To further assess the antiviral activity of the 2- -*C*-methyluridine aryloxyl phosphoramidate ProTide against ZIKV, we performed plaque assays at 48 h post-infection to titer ZIKV (H/PAN, MOI of 0.1) levels during treatment with either 10 or 30 μM of 2- -*C*-methyluridine aryloxyl phosphoramidate ProTide or 10 or 30 μM of sofosbuvir in human fetal neural stem cells (Figure 3). At 30 μM compound, both 2- -*C*-methyluridine aryloxyl phosphoramidate ProTide and sofosbuvir were able to repress viral titers to undetectable levels. However, only 2- -*C*-methyluridine aryloxyl phosphoramidate ProTide was able to reduce ZIKV titers to undetectable levels at the 10 μM compound, again highlighting its superior antiviral activity compared to sofosbuvir.

**Figure 3.** Reduction in ZIKV titers by 2- -*C*-methyluridine aryloxyl phosphoramidate ProTide and sofosbuvir in neural stem cells as observed by plaque assay. Viral infections were conducted in human fetal neural stem cells (H/PAN ZIKV, MOI 0.1) using vehicle-untreated (ONLY ZIKV), DMSO vehicle-treated, 10 or 30 μM sofosbuvir (SOF)-treated, and 10 or 30 μM 2- -*C*-methyluridine aryloxyl phosphoramidate ProTide (2- CMU)-treated samples. The dotted line represents the limit of detection of the assay. Virus was titered as described in the methods section and plaque forming units (PFU) per ml were calculated for each sample, ± standard error (SE), *n* = 4.

#### *3.5. Enzymatic Incorporation of the Active Triphosphate Metabolites of 2*- *-C-Methyluridine and Sofosbuvir Over Time Reveals Higher Levels of Incorporation for the Former Compound*

Single nucleotide incorporation assays were used to compare the relative preference of ZIKV RdRP for UTP versus the UTP analogs 2- -*C*-methyluridine triphosphate and 2- -fluoro-2- -*C*-methyluridine triphosphate (the active form of sofosbuvir) (Figure 4). Both analogues appeared to serve as chain terminators of ZIKV RdRP polymerization, as previously reported [19,31]. Unsurprisingly, UTP

was most efficiently incorporated, with 50% incorporation by 5 min and full incorporation by ~30 min, followed by misincorporating extension. The 2- -fluoro-2- -*C*-methyluridine triphosphate had significantly less efficient incorporation by ZIKV RdRP, as previously reported [19,31,50]. Incorporation plateaued at 15%, and half maximal incorporation was only reached after nearly 90 min. In contrast, 2- -*C*-methyluridine triphosphate reached half of its eventual maximal incorporation of 35% after ~45 min, showing improved incorporation over the active form of sofosbuvir. These trends are supported by previous work focusing on biochemical characterization of 2- -*C*-methyluridine triphosphate and 2- -fluoro-2- -*C*-methyluridine triphosphate incorporation [19,20]. Importantly, here we verify the preference of 2- -*C*-methyluridine over sofosbuvir in both biochemical and cell-based assays, suggesting a robust inhibitor testing pipeline that supports the hypothesis that cellular efficacy occurs as a result of RdRP inhibition by nucleoside analogs.

**Figure 4.** Kinetic characterization of nucleotide and nucleotide analogue incorporation by ZIKV RdRP. (**A**) ZIKV RdRP and primer/template (P/T) substrate (n) was mixed with excess UTP, 2- -F-2- -*C*-methyluridine triphosphate (2- -F-2- -C-MeUTP), or 2-*C*-methyluridine triphosphate (2- -C-MeUTP), and the reaction was quenched at the indicated timepoints. The substrate and the single nucleotide extension product (n + 1) were separated by using a 20% polyacrylamide denaturing gel, and bands were quantified to determine percent incorporation and plotted against time. (**B**) A single exponential fit was used to determine an observed rate of incorporation, *k*obs (% incorporation min<sup>−</sup>1) of 0.151 ± 0.004, 0.0085 ± 0.001, and 0.016 ± 0.001, for UTP (dotted black line), 2- -F-2- -C-MeUTP (dashed red line), and 2- -C-MeUTP (solid blue line), respectively. SE is reported as deviation from the fit, *n* = 2.

#### *3.6. Molecular Superpositioning of the ZIKV NS5 Active Site: Consequences for 2*- *-C-Derivatitized Nucleoside Analogues*

A previously prepared model (gc-o3 [36]) of ZIKV RdRP in complex with RNA and ATP generated from PDB 5TFR [37] and PDB 4WTD [34], an apo structure of ZIKV RdRP (PDB 5WZ3 [38]), and an apo structure of HCV RdRP (PDB 3MWV [39]) were aligned with a crystal structure of HCV RdRP complexed with RNA and sofosbuvir diphosphate (PDB 4WTG [34]), shown in Figure 5. ZIKV RdRP is shown in magenta (apo form) or in red (in complex with ligands), and HCV RdRP is shown in dark blue (apo form) or in cyan (in complex). The crystal structure of inhibited HCV RdRP (cyan) shows that residue D225 moved upward to accommodate sofosbuvir diphosphate, while in the ZIKV model (red) [36], the corresponding residue D1141 in ZIKV RdRP pointed down in the opposite direction to accommodate the nucleotide (Figure 5). While the position of the aspartic acid in both the apo forms of the viral RdRPs was similar, it is interesting that the nucleoside ligands were accommodated in different ways, with ZIKV RdRP displaying a larger change in positioning in this residue. While modeling was used to generate the holo-structure of ZIKV RdRP, this suggests some evidence that the ZIKV binding pocket is poised to accommodate a variety of nucleoside analogues.

**Figure 5.** Structural superpositioning of apo and bound ZIKV RdRP and hepatitis C virus (HCV) RdRP. The ZIKV RdRP model gc-o3 [34] in complex with two manganese ions (green), ATP (red/CPK coloring), and RNA (orange/CPK coloring) is shown in red, and a crystal structure of apo ZIKV RdRP is shown in magenta (PDB 5WZ3 [38]). A crystal structure of HCV RdRP complexed with sofosbuvir diphosphate (cyan/CPK coloring), RNA (orange/CPK coloring), and a manganese ion (purple) is shown in cyan (PDB 4WTG [36]), and an apo structure of HCV RdRP is shown in blue (PDB 3MWV [39]). A comparison of the homologous residue D225 (HCV RdRP) and D1141 (ZIKV RdRP) is highlighted in stick form.

#### **4. Discussion**

In this work, we demonstrated that ProTides other than sofosbuvir have promise as potential antiviral drugs to treat ZIKV infections. Interestingly, an apparent bias towards 2- -*C*-modified uridylate ProTides (with the exception of the ethenyl derivative) was observed, suggesting a possible nucleobase bias to compound activity. Recent biochemical data examining levels of off-target incorporation of nucleoside analogue triphosphates by the human mitochondrial RNA polymerase may help explain the phenotypic relationship we observed; both 2- -*C*-methyluridine triphosphate and 2- -*C*-methyl-2- -*C*-fluorouridine triphosphate (the active metabolite of sofosbuvir) show much lower levels of incorporation by this host enzyme than 2- -*C*-modified analogues with other nucleobases [51]. We propose that this may account for the better safety profile observed for sofosbuvir compared to other 2- -*C*-modified compounds that were tested in clinical trials. This may also suggest that there is greater selectivity for 2- -*C*-modified uridylate incorporation by flavivirus polymerases over host polymerases, though the exact mechanism by which this occurs should be examined in future studies.

We also observed that changing the ProTide functionality of 2- -*C*-methyluridine from an aryloxyl phosphoramidate masking group to a 2-(methylthio)ethyl phosphoramidate resulted in a complete loss of activity, suggesting that the local metabolic environment of the host cell is critical for processing of the ProTide functionality. Further studies examining the relationship between different ProTide functionalities and the cell type dependence of their antiviral activity as well as testing the impact of these effects using in vivo animal model testing are planned in our research groups.

We have demonstrated here for the first time that ProTide technology can be broadly applied to anti-ZIKV drug development with nucleobase and ProTide group selectivity observed. These data will guide future work to design highly selective, safe, and bioavailable compounds for the treatment of ZIKV infections.

**Author Contributions:** J.A.B., M.C., S.B., G.A.W., L.A.L., A.E.C., Z.Z., J.N.R., C.D.S., B.W.P., and J.L.S.-N. conceived and designed the experiments; J.A.B., M.C., S.B., G.A.W., L.A.L., D.H., A.E.C., and Z.Z. performed experiments; J.A.B., M.C., S.B., G.A.W., C.D.S., B.W.P., and J.L.S.N. analyzed the data; J.A.B., M.C., S.B., G.A.W., C.D.S., B.W.P., and J.L.S.-N. wrote the original draft of manuscript; J.A.B., M.C., S.B., G.A.W., L.A.L., A.E.C., Z.Z., D.H., J.N.R., C.D.S., B.W.P., and J.L.S.-N. reviewed and edited the manuscript.

**Funding:** This research was supported by Clinical and Translational (CTRI) pilot grant UL1TR001442 (to J.L.S.-N.), and a CSUPERB New Investigator grant (to C.D.S). B.W.P., M.C., G.A.W., L.A.L., and C.D.S. thank San Diego State University and the California Metabolic Research Foundation for financial support.

**Acknowledgments:** The authors acknowledge the access and use of the UCSD Screening Core to perform part of the experiments presented in this manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.
