High throughput screening of nucleoside analogs against DENV in different cell models

DENV targets diverse tissues in vivo and also can infect diverse cell types in vitro [18,67–69]. The liver is one of the most affected organs during Dengue hemorrhagic fever, and hepatocytes supports high levels of viral replication [70]. We developed assays in two liver cell models which includes human hepatocellular carcinoma Huh7.5 cells and, HepG2 cells, both of which have been previously used to study DENV infection [71,72]. Epithelial cells are also targets for DENV infection and thus we utilized human adenocarcinoma epithelial cells, A549 to model infection. Lastly fibroblasts are known targets of DENV infection, and we developed assays in non-transformed human lung fibroblast IMR90 cells. For each of these cell types, we optimized the infection conditions including viral MOI, and timing of infection using the prototype strain, New Guinea C of DENV 2 serotype (DENV2). We also optimized infection using previously known nucleoside analogs NITD-008 [48], MK-0608 [73] and AT-527 [59,74] shown to be active against DENV2. The antiviral activity of nucleoside candidates against Dengue was monitored using automated microscopy and immunostaining for DENV envelope (4G2) and nuclei (Fig 1A). Automated image analysis was used to quantify cell number and the percentage of infected cells compared to vehicle control. Dose response studies allowed us to quantify the 50% inhibitory concentration (IC₅₀) and 50% cell toxicity concentration (CC₅₀) for each nucleoside analog as well as the selectivity index (SI = CC₅₀/IC₅₀) (S1 Fig). We confirmed that all three known antivirals MK-0608, NITD-008 and AT-527 were active against DENV2 in all cell types tested (S1 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. High throughput screening of nucleoside analogs identifies antivirals against DENV in different cell models: A.

Representative fluorescence microscopy images of DENV2 infection in the indicated cell models (A549, IMR90, HepG2, and Huh7.5).Cells were treated with vehicle or the indicated compounds and either left uninfected or infected with DENV2 at MOI 2 (Huh7.5 and A549), MOI 1 (HepG2), and MOI 5 (IMR90). At 24 hours post-infection (hpi), cells were fixed and stained for viral antigen (4G2, green) and nuclei (Hoechst 33342, blue). Images were acquired at 10 × magnification. B. Schematic of the screening strategy. 384-well format high-throughput screen was conducted using A549, IMR90, HepG2, and Huh7.5 cells to evaluate a library of 1,101 nucleoside analogs for their ability to inhibit DENV2 infection at 50 μM. Cells were treated with compounds for 2 h then infected with DENV2 at the MOI in A, followed by fixation at 24 hours post-infection. Automated microscopy was used to quantify percent of infection and cell numbers normalized to vehicle control. Percent of control (POC) for DENV2 infection in C. Huh7.5 D. HepG2 cells E. A549 cells and F. IMR90 cells. G. Venn diagram representing 68 non-redundant candidates that showed >80 inhibition of infection; > 60% viability in at least one of the screened cell models (C-F). H. Summary of hit selection and downstream validation showing the progression from 68 initial hits identified in four cell models to 23 candidates with SI > 5 based on dose–response analysis, followed by qPCR validation in selected cell models.

https://doi.org/10.1371/journal.ppat.1013970.g001

We screened an in-house nucleoside analog library which contains 1101 candidates, of which 559 are purines and 542 are pyrimidines at 50 μM. A schematic of the high-throughput screening is shown in Fig 1B. Cells were treated for 1 hour prior to infection and the cell number and percentage of infected cells were quantified 24 hours post infection using automated microscopy and image analysis in each cell model (Fig 1C-1F). Candidates with > 80% reduction in infection and with > 60% cell viability compared with DMSO control were selected in the primary screen. > 80% reduction in infection is a stringent cut-off to ensure that only compounds demonstrating strong antiviral efficacy were considered, similarly a ≥ 60% cell viability threshold (relative to the DMSO control) was applied to exclude compounds exhibiting excessive cytotoxicity that could confound the observed antiviral effect. The Venn diagram is shown for these compounds across the 4 different cell models. Under this threshold criteria, we identified a total 68 non-redundant candidates as active against DENV2 in at least one of the 4 cell models (Fig 1G and S1 Table). 11 compounds were identified across all four cell models and include nucleosides with known activity against DENV (Fig 1G and S1 Table).

Defining nucleotide analogs with broad activity across different cell models

We repurchased these 68 compounds and validated their activity against DENV2 using dose response assays (Fig 1H and S1 Table). We quantified infection using the same assay as the primary screen and we quantified toxicity outside of infection using ATPlite which monitors ATP levels as a surrogate for viability. While most compounds identified from the primary screening were evaluated in subsequent assays, not all compounds exhibiting activity in only one or two cell types were pursued for additional testing (S1 Table). From the tested candidates, 23 exhibited a SI greater than 5 in at least one cell model (Fig 1H and S1 Table). 5 had previously been reported to possess antiviral activity against flaviviruses, including Ribavirin [75], MK-0608 [53], 6-Azauridine, Azaribine (prodrug of 6-azauridine) [76] and azvudine [77]. 12 candidates were previously unstudied (Fig 2 and S1 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Dose response and orthogonal validation of nucleoside analogs: Table presents the comprehensive validation panel of IC₅₀, CC₅₀ (ATPlite), and SI values for indicated nucleoside analogs including the 23 primary hits from the DENV2 primary screen (n = 68) that exhibited a selectivity index (SI) > 5 in at least one cell model. The panel includes 2 positive controls (NITD-008 and AT-527). The serial numbers (S. no.) correspond to compounds listed in S1 Table. Blue SI > 5 and orange SI > 10. ATPlite data shown for each drug in each cell type at 72 h. SI shown for CC₅₀ (ATPlite)/IC₅₀. The log fold change from qPCR assessment in cells pretreated with the indicated compounds (10 μM) vs DMSO vehicle control infected DENV (MOI = 0.05) and subject to RT-qPCR 24 hours post-infection. Data are presented as mean ± SD, showing viral RNA levels relative to the vehicle control (n ≥ 1–3 independent biological replicates). Statistical significance was determined on n ≥ 3 by one-way ANOVA with Dunnett’s correction for multiple comparisons on log₁₀-transformed values (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

https://doi.org/10.1371/journal.ppat.1013970.g002

We found that 14 compounds were 2’-C-methyl adenosine analogs, a modification known to confer antiviral activity (S6 Fig). This includes 2’-C-methyladenosine (UPGNUC129) [36], UPGNUC371 which resembles MK-0608 (UPGNUC120) structurally, and 2’-β-C-Ethynyladenosine (UPGNUC558) had not been previously shown to have antiviral activity against flaviviruses, and has similar features to NITD-008 (Figs 2, S2, and S3 and S1 Table). Azvudine (UPGNUC522) is another 2′-fluorinated nucleoside that has previously shown antiviral activity against HCV and DENV [77–79], and has recently demonstrated therapeutic promise against SARS-CoV-2 receiving conditional approval in China for the treatment of adult patients with COVID-19 [80]. UPGNUC345, a prodrug of Azvudine (UPGNUC522), contains a 2’-deoxy-2’-fluoro modification, a feature shared with UPGNUC125. Finally, gemcitabine (UPGNUC400), a widely used anticancer agent [81], serves as the parent nucleoside for UPGNUC961. Both compounds bear a 2’-deoxy-2’,2’-difluoro substitution and were found to be active against DENV2 in our screening (S6 Fig and S1 Table).

Tubercidin is a naturally occurring adenosine analog produced by Streptomyces tubercidis, and is known for its broad-spectrum antiviral activity. However, tubercidin’s therapeutic use is limited due to its high cytotoxicity in mammalian cells [64,65,82]. We identified 3’ deoxytubercidin (UPGNUC492), 6-Chloro-7-deazapurine-9-β-D-ribofuranose (UPGNUC255), and a prodrug for UPGNUC255 (UPGNUC855) which have structural similarity to tubercidin but have less cellular toxicity (S7 Fig).

Additionally, we identified several compounds including UPGNUC402, UPGNUC404, UPGNUC697, and UPGNUC782 that share structural similarities with N6-methyladenosine, which has been previously reported to exhibit antiviral effects against HIV, HBV, IAV, and other flaviviruses (S7 Fig) [83,84]. Among them, 6-chloroinosine (UPGNUC697) has been reported as both an antitumor and antiviral agent [85,86]. Another compound, UPGNUC010 (3’-Deoxy-3’-fluoroadenosine), has been reported to possess broad-spectrum antiviral activity against flaviviruses [87]. Lastly, UPGNUC740, UPGNUC338 and UPGNUC337 feature modifications at the 8’ carbon of the adenosine ring, while UPGNUC040 carries a hydroxyl substitution at the 8’ position (S7 Fig). Altogether, primary screen identified 68 initial hits of which 23 validated by dose–response analysis (S1 Table), confirming potency (IC₅₀) and high SIs for the active compounds (Fig 2). A subset of these is active across multiple cell types suggesting efficient utilization by the nucleoside salvage pathways.

Antiviral activity assessment by qPCR

To further quantify the antiviral activity of the identified compounds, we performed an orthogonal assay to quantify the reduction in DENV2 viral RNA across the cell models. Cells were pretreated with vehicle control (DMSO) or the indicated nucleoside analogs at 10 μM for 1hour and then infected with DENV2. NITD-008 and AT-527 were included as positive controls (Figs 2 and S3). As expected, both control compounds demonstrated >10-fold reduction in viral RNA levels across all tested cell types. The qPCR assays were performed for compounds showing a SI > 5 at 10 µM rather than 50 µM since it is challenging to achieve such high concentrations in vivo (S3A, S3B, S3C and S3D Fig). A subset that showed activity at 10 µM were also tested at 2 µM (S3E Fig). We also tested the prodrugs of UPGNUC255 (S3A, S3B, S3C and S3D Fig).

Consistent with results from our microscopy-based assay, UPGNUC255, and UPGNUC558, showed >10-fold reduction in DENV2 RNA in all four cellular models (Figs 2 and S3A-S3E). UPGNUC345 was active in Huh7.5, HepG2 and A549 whereas, UPGNUC028 and UPGNUC492 exhibited >10-fold reduction, but only in Huh7.5 and A549 cells. Additionally, UPGNUC855, UPGNUC010, UPGNUC338, UPGNUC371 and UPGNUC963 significantly reduced viral RNA levels in Huh7.5 and HepG2 cells. UPGNUC961 and UPGNUC402 demonstrated notable antiviral activity in HepG2 cells as well (Figs 2 and S3A-S3D). UPGNUC522 (Azvudine), a known antiviral agent, and UPGNUC400 (Gemcitabine), an anticancer drug with reported antiviral properties, both showed significant reductions in DENV2 RNA levels in all cell types tested (Figs 2 and S3A-S3D). Tubercidin-like compounds sharing structural features include UPGNUC255 and its prodrugs UPGNUC855 and UPGNUC963, as well as UPGNUC492, together with 2′-substituted analogues (UPGNUC558, and UPGNUC371), the 3′-modified compound UPGNUC010 (3′-deoxy-3′-fluoroadenosine), the azvudine-like analogue UPGNUC345, exhibited cell-type–dependent differences in antiviral potency. These variations likely arise from differences in cellular uptake, metabolic activation to the active triphosphate form, or host factor expression across cell models. Together, these findings highlight how specific structural modifications influence antiviral activity across distinct cellular environment and underscoring the role of structure–activity relationships in determining pan cell-type efficacy.

Antiviral assessment across all four dengue serotypes

Dengue circulates in antigenically and genetically distinct serotypes (DENV1–4) [88]. These serotypes do not provide cross protection against each other and secondary infection with a distinct serotype can lead to more severe disease [89,90]. Therefore, it is essential for an antiviral to be effective against all 4 serotypes. We optimized DENV1, DENV3 and DENV4 infection conditions in Huh7.5 cells (Fig 3A) and performed dose response studies on the 12 compounds, which were found to be active for DENV2 confirmed antiviral potency (IC₅₀ ≤ 10 µM) confirmed by both dose response and qPCR assay and favorable selectivity index (SI > 5) in Huh7.5 cells from the primary screen to assess activity across dengue serotypes (Figs 2, S3A and S3B). Notably, all 12 compounds tested, exhibited a selectivity index (SI) > 5 against both DENV1 and DENV3, and 8 of these compounds showed an SI > 5 against DENV4, which is the genetically more distinct serotype (Fig 3B) [91,92].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Nucleoside analog activity across DENV serotypes: A. Representative microscopy images for DENV1, DENV3 and DENV4 infection in the Huh7.5 cells were treated with either DMSO or NITD-008. Cells were fixed 24 hpi and stained for viral infection (4G2, green) and cell number (Hoechst 33342, blue).

10 × magnification. B. Table of IC₅₀, CC₅₀ (ATPlite), and SI values for selected drug candidates active against DENV2, evaluated in Huh7.5 cells infected with DENV1, DENV2, DENV3, or DENV4. Huh7.5 cells were infected with DENV1 (MOI = 1), DENV2 (MOI = 0.05), DENV3 (MOI = 1), or DENV4 (MOI = 0.5). The serial numbers (S. no.) correspond to in S1 Table. ATPlite data shown for each drug in each cell type at 72 h. SI shown for CC₅₀ (ATPlite)/IC₅₀ with blue SI > 5 and orange SI > 10. The log fold change from qRT-PCR analysis of dengue serotypes DENV1, DENV2, DENV3 and DENV4 infection for 24 h in Huh7.5 cells pretreated with the indicated compounds (10 μM) vs DMSO vehicle control. Data are presented as mean ± SD, showing viral RNA levels relative to the vehicle control (n ≥ 3 independent biological replicates). Statistical significance was determined for n ≥ 3 by one-way ANOVA with Dunnett’s correction for multiple comparisons on log₁₀-transformed values (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

https://doi.org/10.1371/journal.ppat.1013970.g003

We next quantified the antiviral activity of seven selected nucleoside analogs against all four DENV serotypes using RT-qPCR. These were more active candidates were chosen based on (i) strong pan-cell and pan-serotype inhibitory activity by dose response assay, (ii) structural diversity representing distinct nucleoside scaffolds, and (iii) consistent and robust inhibition of DENV2 replication in secondary qPCR validation assays. Notably, UPGNUC255, UPGNUC492, and UPGNUC558 reduced DENV1 RNA levels by >10-fold, DENV3 by >5-fold, and DENV4 by >10-fold in Huh7.5 cells, indicating broad-spectrum activity against multiple DENV serotypes (Figs 3B and S4A-S4C). In contrast, UPGNUC010 showed activity against DENV1 and DENV4, while UPGNUC028, UPGNUC345 and UPGNUC371 were active against DENV1 and DENV3 (Figs 3B and S4A-S4C). These results suggest that UPGNUC255, UPGNUC492, and UPGNUC558, reduced viral RNA levels across DENV1, DENV2, DENV3, and DENV4, highlighting their potential as pan-Dengue antiviral candidates. Moreover, compounds with serotype-specific activity, such as UPGNUC010, UPGNUC028, UPGNUC345, and UPGNUC371 may provide valuable insights into serotype-dependent mechanisms of antiviral efficacy. The pan DENV-serotype active compounds have broad-spectrum activity most likely results from targeting conserved viral replication mechanisms rather than serotype-specific or host-specific factors.

Nucleotide analogs show broad activity against different flaviviruses

Next, we set out to identify the activity of nucleoside analogs against additional flaviviruses and thus we optimized HTS assays for WNV (NY2000) and ZIKV (MR766) in Huh7.5 cells, utilizing NITD-008 as a positive control (Fig 4A and S2 Table). We screened the in-house nucleoside analog library at 50 µM as we previously screened against DENV2 using a selection criteria of 80% reduction in viral infection and greater than 60% cell viability compared to the DMSO control (Fig 4B and 4C). We identified 21 compounds active against WNV and 16 compounds active against ZIKV (Fig 4D and S2 Table). Comparison with DENV2 screening data in Huh7.5 cells revealed 7 compounds that were active against all three viruses WNV, ZIKV, and DENV2. Additionally, 12 compounds were shared between DENV2 and WNV, and 8 compounds were common between DENV2 and ZIKV (Fig 4D and S2 Table). We performed dose-response assays for the compounds identified from the primary antiviral screens against WNV, ZIKV and which were active for DENV2 in Huh7.5 cells (S2 Table). We also optimized assays for YFV (vaccine strain17D) in Caco-2 cells using NITD-008 as a positive control (Fig 4A). We also tested the DENV2 active compounds against YFV and found a total of 20 compounds exhibited an SI > 5 against at least one additional flavivirus (S2 Table). Fourteen compounds showing activity against at least two flaviviruses with a selectivity index (SI) > 5 were included as the final panel for cross-flavivirus validation (Fig 4E).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Identification of nucleoside analogs active across diverse flaviviruses: A. Representative microscopy images for Huh7.5 cells infected with WNV (NY2000) or ZIKA (MR766) and Caco-2 cells infected with YFV (17D) pretreated with either DMSO or NITD-008 (10uM). Cells were fixed 24 hpi and stained for viral infection (4G2, green) and cell number (Hoechst 33342, blue).

10 × magnification. Percent of control (POC) for (B.) WNV infection and (C.) ZIKA infection in Huh7.5 cells screened against the library of nucleoside analogs. D. Venn diagram showing 41 non-redundant candidate compounds with >80% inhibition of infection and >60% cell viability against DENV2, WNV or ZIKV, in Huh7.5 cells. E. Table of IC₅₀, CC₅₀ (ATPlite), and SI values for selected drug candidates against WNV, ZIKA, DENV2, infection in Huh7.5 cells and YFV in Caco2 cell with SI > 5 at least against one virus. The serial numbers (S. no.) correspond to S1 Table. ATPlite data shown for each drug in each cell type at 72 h. SI shown for CC₅₀ (ATPlite)/IC₅₀ with blue SI > 5 and orange SI > 10. The log fold change from qRT-PCR analysis of WNV infection; ZIKA infection YFV infection at MOI = 0.5 and DENV2 infection at MOI = 0.05 each in Huh7.5 cells pretreated with the indicated compounds (10 μM) or DMSO vehicle control at 24 hpi. qPCR data are presented as mean ± SD, showing viral RNA levels relative to the vehicle control (n ≥ 1-3 independent biological replicates). Statistical significance was determined for n ≥ 3 by one-way ANOVA with Dunnett’s correction for multiple comparisons on log₁₀-transformed values (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

https://doi.org/10.1371/journal.ppat.1013970.g004

We further quantified the activity of candidates from individual virus primary screens (S2 Table), including those active against DENV2, based on (i) strong antiviral activity across at least two flaviviruses by dose response assay (ii) consistent SI > 5, and (iii) structural diversity using qPCR assays against WNV NY2000 (Figs 4E and S4D), WNV Kunjin (Figs 4E and S4E), two strains of ZIKV (Puerto Rico and MR766 (Figs 4E, S4F and S4G) and YFV 17D (Figs 4E and S4H). We found that in addition to the positive controls, UPGNUC255, UPGNUC492, UPGNUC558, UPGNUC010, UPGNUC338, and UPGNUC400 exhibited potent antiviral activity, resulting in >10-fold reductions in viral RNA across all tested flaviviruses (Figs 4E and S4D-S4H). This suggests a conserved mechanism-of-action either directly on RdRp or a host enzyme. Additionally, UPGNUC371 (structurally similar to UPGNUC558, a 2′-substituted analog) showed antiviral activity against both ZIKV and YFV, and UPGNUC345 exhibited activity specifically against WNV and YFV (Figs 4E and S4D-S4H). These findings demonstrate in vitro pan-antiviral activity for UPGNUC255, UPGNUC492, and UPGNUC558.

Nucleoside analogs show activity in primary cells

Our initial studies leveraged a panel of human cell lines including the cancer cell lines Huh7.5, HepG2, A549 and Caco-2 along with the immortalized but not transformed IMR90. We also set out to develop assays in primary human cells to confirm efficacy in more relevant cell types. During vector transmission, local infection at the skin is required [93,94]. Studies have shown that keratinocytes and fibroblasts are infected during flaviviruses infections; therefore, we tested antiviral activity in these primary cells. We optimized DENV2 infection protocols for primary human keratinocytes and fibroblasts and for WNV infection in primary human keratinocytes. We used NITD-008 as a positive control and found that treatment could block infection of DENV or WNV (Fig 5A and 5B). We also optimized a RT-qPCR assay again testing NITD-008, AT-527 and MK-0608 and found all three compounds resulted in approximately a 100-fold reduction in DENV2 viral RNA levels in primary human keratinocytes (Figs 5C and S5A). Additionally, treatment with NITD-008 achieved a ~ 100-fold decrease in West Nile virus (WNV) RNA levels in primary keratinocytes (Figs 5C and S5C). In primary fibroblasts, NITD-008 led to ~10-fold reduction in DENV2 RNA (Figs 5C and S5B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Nucleoside analog activity in primary cells: A. Representative fluorescence microscopy images of primary human keratinocytes and fibroblasts pretreated with either DMSO or 10 μM NITD-008 for 1 hour prior to infection with DENV2 (MOI = 1 for keratinocytes; MOI = 10 for fibroblasts).

48 hpi, cells were fixed and stained for viral antigen (4G2, green) and nuclei (Hoechst 33342, blue). Images captured at 10 × magnification. B. Representative fluorescence microscopy images of primary human keratinocytes pretreated with either DMSO or 10 μM NITD-008 for 1 hour prior to infection with WNV (MOI = 1). 48 hpi, cells were fixed and stained for viral antigen (4G2, green) and nuclei (Hoechst 33342, blue). Images captured at 10 × magnification. C. Table of IC₅₀, CC₅₀ (ATPlite), and SI values for selected drug candidates active against DENV2, and WNV evaluated in Huh7.5 cells. The log fold change from RT-qPCR assay in primary keratinocytes and primary fibroblasts pretreated with the indicated compounds (10 μM) vs DMSO vehicle control were infected with DENV2 (MOI = 0.5) or WNV (MOI = 0.5) for 48hpi. For all qPCR, data are presented as mean ± SD, showing viral RNA levels relative to the vehicle control (n ≥ 1-3 independent biological replicates). Statistical significance was determined for n ≥ 3 by one-way ANOVA with Dunnett’s correction for multiple comparisons on log₁₀ transformed values (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

https://doi.org/10.1371/journal.ppat.1013970.g005

Next, we evaluated the antiviral activity of selected nucleoside analogs from the validated panel chosen based on their activity against DENV2 in multiple cell lines and cross-flavivirus potency, specifically those showing (i) strong antiviral activity across at least two flaviviruses, (ii) consistent SI > 5, and (iii) structural diversity representing distinct chemical scaffolds, against DENV2 in primary keratinocytes using qPCR. In addition to the positive control drugs, UPGNUC028, UPGNUC255, UPGNUC558, UPGNUC855, UPGNUC522, UPGNUC010, UPGNUC338, UPGNUC400, UPGNUC345, UPGNUC961 and UPGNUC371, reduced DENV2 viral load by more than 10-fold (Figs 5C and S5A). Similarly, UPGNUC028, UPGNUC255, UPGNUC558, UPGNUC855, UPGNUC345 and UPGNUC400 reduced DENV2 infection 10-fold in primary fibroblasts (Figs 5C and S5B). In contrast, UPGNUC492 showed no significant reduction in viral load in primary keratinocytes and fibroblasts, indicating its lack of efficacy in primary cells (Figs 5C, S5A and S5B). We also tested activity of these nucleoside analogs against WNV and found that only UPGNUC255, UPGNUC558, UPGNUC345, UPGNUC010, UPGNUC371 and UPGNUC338 demonstrated more than a 10-fold reduction in viral in primary keratinocytes (Figs 5C and S5C), suggesting their potential as broad-spectrum antiviral candidates active in primary human cells.

UPGNUC492 possesses an unusual structural modification featuring a carbon substitution at the N-7 position of the purine ring and lacking the 3’-hydroxyl group on the ribose and exhibited low cytotoxicity in 3 models (Huh7.5, HepG2 and A549). However, its antiviral activity against DENV2 was only evident at higher concentrations across all four tested cell models (Huh7.5, A549, IMR90, and HepG2) by dose response assay, suggesting limited potency and making it a compound of reduced therapeutic interest (Figs 2, S2A and S2B). Similar to microscopy-based observations, RT-qPCR analysis confirmed significant viral RNA reduction at 10 µM only in Huh7.5 and A549 (Figs 2 and S3A-S3D), while no activity was observed at 2 µM (S3E Fig). Interestingly, UPGNUC492 showed broader antiviral activity, achieving ≥10-fold reductions in viral RNA levels for DENV1–4 (Figs 3C and S4A-S4C) and 500-, 10-, 100-, and 500-fold reductions in WNV, Kunjin, ZIKV, and YFV, respectively, at 10 µM tested in Huh7.5 cells (Figs 4E and S4D-S4H). Nevertheless, due to its lack of significant activity at lower concentrations, limited potency in key models with no activity in primary cells, UPGNUC492 is considered a less promising candidate.

UPGNUC255 and UPGNUC558 are broadly active and potent across flaviviruses

Our data show that UPGNUC255 and UPGNUC558 display high potency and broad activity among different cell models including primary cells and were active against a panel of flaviviruses. UPGNUC558 is structurally related to NITD-008, distinguished by the presence of an ethynyl group at the 2’ position of the ribose moiety in the adenosine scaffold. UPGNUC558 shows activity across cell types against DENV2, and across all four DENV serotypes and broad-spectrum antiviral activity against WNV, KUNV, ZIKV, and YFV. Although both qPCR and microscopy assays showed UPGNUC558 and NITD-008 to have comparable efficacy in reducing viral RNA levels, UPGNUC558 consistently displayed lower IC₅₀ values and higher selectivity indices (SI) across multiple viruses, including DENV1, DENV2, DENV3, DENV4, WNV, KUNV, ZIKV, and YFV, in Huh7.5 cells (Figs 2, 3B and 4E). Given these observations, we set out to perform further mechanistic studies.

UPGNUC255, a synthetic adenosine nucleoside analog related to tubercidin (7-deazaadenosine). In addition to the substitution of the nitrogen at the purine N7 position with carbon, there is a replacement of the 6-amino group with a chlorine atom (S2A Fig). In our primary screen, Tubercidin (UPGNUC211, S1 Table) displayed activity which has been previously shown against diverse RNA viruses (e.g., poliovirus, rhinovirus, and Zika virus) but as expected, we also observed pronounced cytotoxicity (S1 Table) [65,82]. Although UPGNUC255 exhibited some cytotoxicity in IMR90 and HepG2 cells, it demonstrated a markedly improved selectivity index (SI) compared to tubercidin and structurally related analogs, primarily due to its higher antiviral potency at lower effective concentrations. UPGNUC255 exhibited broad-spectrum activity, significantly inhibiting replication of all four DENV serotypes, as well as WNV, Kunjin virus, ZIKV, and YFV (Figs 2, 3B, and 4E). These results highlight UPGNUC255 as a promising lead candidate, offering enhanced antiviral efficacy coupled with a more favorable therapeutic window than tubercidin.

We also identified two prodrugs of UPGNUC255, UPGNUC855 and UPGNUC963. UPGNUC855 displays reduced cytotoxicity but requires higher concentrations to achieve antiviral activity, as reflected in its elevated IC₅₀ values across all tested cell models (Fig 2) and reduced efficacy at 2uM (S3E Fig) and other flaviviruses (Figs 4E, S4D, S4H and S4G). In contrast, UPGNUC963 shows comparable IC₅₀ values to UPGNUC255 but exhibits greater cytotoxicity in hepatocytes and lung fibroblasts (S3A-S3D Fig and S1 Table), limiting its potential despite its antiviral potency.

We also synthesized two derivatives of UPGNUC255, mCot849 with a 4’ fluoro, and mCot903 with an additional chloro-group on the purine ring (S9 Fig). mCot849 exhibited antiviral potency comparable to UPGNUC255 (similar IC₅₀ values), along with a slightly improved cytotoxicity profile (higher CC₅₀), suggesting it may be a more favorable candidate than the parent compound. In contrast, mCot903 displayed higher IC₅₀ values compared to UPGNUC255, indicating reduced potency, and although it showed slightly improved cytotoxicity, its overall potential appears limited (S9 Fig).

Effect of exogenous NTPs on antiviral activity of nucleoside analogs

Direct acting nucleoside analogs compete with natural nucleoside triphosphates (NTPs) for binding to the RNA-dependent RNA polymerase (RdRP) for incorporation. Some compete efficiently while others compete less efficiently and require higher concentrations [95]. Furthermore, some nucleoside analogs can act by inhibiting the biosynthesis of natural NTPs, thereby inhibiting infection through this indirect mechanism and these can be readily competed with exogenous nucleosides [96]. Nucleoside analogs that can be readily competed are less likely to show activity in vivo in animals. To determine whether high levels of exogenous nucleosides can reduce the activity of our nucleoside analogs we treated cells with increasing concentrations of purines or pyrimidines in the presence of increasing doses of the nucleoside analogs. As a control, we tested 6-azauridine (UPGNUC028) which is phosphorylated to 6-azauridine-5’-phosphate to act as a competitive inhibitor of the enzyme orotidylate decarboxylase, required for pyrimidine biosynthesis [97,98]. Thus, 6-azauridine is known to inhibit de novo pyrimidine biosynthesis and should be competed with high levels of pyrimidines. In addition to its role as a de novo pyrimidine biosynthesis inhibitor, 6-azauridine can be phosphorylated to 6-azauridine triphosphate and incorporated into viral RNA, causing errors and leading to non-viable viral particles [99,100]. If 6-azauridine inhibited de novo biosynthesis as the mechanism of action we expected that treatment with exogenous pyrimidines would overcome the antiviral effects of 6-azauridine. Indeed, we observed dose-dependent decreases in antiviral activity of 6-azauridine when pyrimidines, but not purines, were added exogenously (Fig 6A). Next, we tested the other known antivirals NITD-008, AT-527, MK-0608, 2’-C-methyladenosine and Azvudine and observed no change in the dose response (d-r) curves upon addition of purines or pyrimidines (Fig 6A). Lastly, we tested our candidates UPGNUC255, UPGNUC492 and UPGNUC558 which were selected based on their novelty, broad antiviral activity across multiple cell models, and demonstrated efficacy against a panel of flaviviruses. These candidates represent the most active profiles for detailed mechanistic evaluation. Interestingly, we observed no significant change in the antiviral activity or shift in the dose response curves of these three nucleoside analogs by addition of either purines or pyrimidines (Fig 6B), suggesting that these nucleoside analogs are not targeting nucleoside biosynthesis pathways. Thus, these compounds are likely direct acting and potentially able to compete with endogenous nucleosides.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Nucleoside analog activity in the presence of exogenous nucleosides: Dose–response analysis of DENV2 infection of Huh7.5 cells with the reference (A.) and test (B.) nucleoside analogs in the presence of increasing concentrations of pyrimidine (C + U) or purine (A + G) nucleosides.

At each concentration of the indicated nucleoside analogs (0.103, 0.309, 0.926, 2.778, 8.333, and 25 µM), corresponding concentrations of purine (A + G) were co-administered at 0.21, 0.62, 1.85, 5.56, 16.67, and 50 µM, respectively, and pyrimidine nucleosides (C + U) at 0.412, 1.23, 3.70, 11.11, 33.33, and 100 µM, respectively.

https://doi.org/10.1371/journal.ppat.1013970.g006

UPGNUC255 and UPGNUC558 inhibit viral RNA replication through NS5

Nucleoside analogs can inhibit viral replication by direct incorporation into the active viral RNA template by the RdRp which can lead to chain termination or delayed chain termination (NITD-008, and AT-527) [48,101], stalling (Remdesivir) [102,103] or misincorporation (Molnupiravir) [104]. Others can impact additional RNA processing activities such as cap methylation (ribavirin and sinofungin) [105]. Since UPGNUC558 is similar to NITD-008 and MK-0608 we expect that the mechanism to inhibit viral RNA synthesis and that the amino acids required for antiviral activity would be similar across these analogs. In contrast, UPGNUC255 is quite distinct, suggesting that the requirements for incorporation may be distinct.

To identify the residues important for antiviral activity, we selected for viruses that were resistant to NITD-008, MK-0608, UPGNUC558 and UPGNUC255 by passaging viruses in the presence of gradually increasing concentrations of the compounds. Kunjin virus (KUNV), a non-pathogenic strain of West Nile virus (WNV), was used for resistance selection studies because it provides higher viral titers and exhibits more robust replication kinetics in cell culture, enabling faster generation and characterization of resistant variants. Importantly, the compounds selected for these experiments demonstrated potent antiviral activity against both KUNV and DENV, underscoring the broader relevance of the resistance findings across flaviviruses. We passaged the virus in the presence of each compound starting at IC₅₀ concentration and DMSO was used as vehicle control across passages. Supernatants were harvested 2–3 days post infection at the onset of >50% CPE. Fig 7A shows the growth profiles of the KUNV resistant mutant selection. KUNV developed resistance to NITD-008, MK-0608, and UPGNUC558 after 22–26 passages (Fig 7A). We then expanded strains and compared the level of viral RNA replication in the wild type and resistant strains by RT-qPCR. We found that the resistant strains are no longer sensitive to the respective treatments (Fig 7B). Interestingly, resistance to UPGNUC255 began after 14 passages but only achieved a ~ 10-fold increase in EC₅₀ concentration and resistance to treatment as measured by RT-qPCR (Fig 7A and 7B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Resistance profiles to nucleoside analogs: A. Growth kinetics of KUNV strains amplified in Huh7.5 cells in the presence of increasing concentrations of the indicated nucleoside analogs. Red numbers = IC₅₀s B. RT-qPCR analysis of KUNV resistant mutants (MOI = 0.5) in Huh7.5 cells pretreated with the indicated compounds (10 μM NITD-008, MK-0608, UPGNUC558; 2 uM UPGNUC255) or DMSO vehicle control 24hpi.

Data are presented as mean ± SD, showing viral RNA levels relative to the vehicle control (n = 3 independent biological replicates). Statistical significance was determined by one-way ANOVA with Dunnett’s correction for multiple comparisons on log₁₀-transformed values (*P < 0.05, **P < 0.01, ***P < 0.001). C. Chemical structures of nucleoside analogs related to UPGNUC558 and UPGNUC255. D. Dose response curves for Huh7.5 cells pretreated with the indicated drugs infected with the KUNV strains (MOI = 1) 24hpi with black showing POC infection and green showing POC cell viability. E. Table of the sensitivity profile of each selected KUNV strain compared to wild type for the indicated nucleoside analogs (IC₅₀ Resistant strain/ Wild type) indicating the extent of cross-resistance to structurally diverse nucleoside inhibitors.

https://doi.org/10.1371/journal.ppat.1013970.g007

We performed dose response studies to the selected nucleoside as well as a larger panel of nucleosides to determine if each strain resistant to the drug it was selected with was cross-resistant to other nucleosides. As expected by their similarity, we found that strains resistant to obtained against NITD-008, MK-0608 and UPGNUC558 exhibited similar patterns of cross-resistance, with more than 10-fold reduced susceptibility to 2’-C-methyl-modified nucleosides, including 2’-C-methyladenosine, 7-deaza-2’-C-methyladenosine (MK-0608), 7-deaza-2’-C-ethynyladenosine (NITD-008) and IDX184 (prodrug of 2′-methylguanosine) (Fig 7C-7E). These data indicate that these structurally related nucleoside analogs act through the same mechanism targeting the RdRp.

The KUNV strains resistant to UPGNUC255 displayed reduced sensitivity to UPGNUC255 and the related nucleoside analog tubercidin, with a ~ 14-fold increase in the IC₅₀ compared to wild-type KUNV (Fig 7C-7E). In contrast, these resistant strains did not show markedly altered sensitivity to the 2’-C-methyl-modified nucleosides, suggesting specificity in the resistance phenotype (Fig 7D-7E).

UPGNUC558 and UPGNUC255 target viral NS5 protein

To determine the amino acid changes that confer resistance we performed whole-genome sequencing (WGS) analysis on resistant mutants and DMSO selected control strains. WGS analysis of the resistant mutants revealed several single nucleotide changes, which were absent in the DMSO-treated control strain (Fig 8A). NITD-008 and MK-0608 are incorporated into the viral RNA, which inhibits RNA chain elongation [48,56]. Sequencing analysis found that all UPGNUC558-, NITD-008-, and MK-0608-resistant strains had a S604T mutation in the RdRp domain of the NS5 protein (Figs 8A and S10C). This mutation has been previously associated with resistance to 2’-C-methylated nucleoside analogues in flaviviruses, including WNV resistant to Sofosbuvir [106]; HCV resistant to 2’-C-methyl ribonucleosides and NITD-008 [47,101] and TBEV resistant to MK-0608 and 2’-C-methylcytidine [66]. This amino acid is conserved across flaviviruses (S10C Fig) and studies suggested that this Serine-to-Threonine substitution reduces analogue interactions and displaces metal ion cofactors [107]. Furthermore, the additional methyl group disrupts water-mediated hydrogen bonding with the 2’-C-methylated analogue, redirecting interactions towards substrate-binding regions [56,107].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. UPGNUC558 and UPGNUC255 interact with RdRp to confer antiviral activity: A. Sequencing of selected KUNV strains showing percent amino acid for each strain, gene and domain. “+” indicates that the mutation was identified by Sanger sequencing B. Modeling shows that the S604T substitution (green, with blue arrow) is localized near the active site (yellow) of the RNA-dependent RNA polymerase (RdRP) domain within the NS5 protein.

The RNA template and incoming nucleoside in the primers site are also shown (grey circle). The A534 residue (green) is located far from the active site. C. Modeling shows that the R355Q mutation (green, blue arrow) is situated in the vicinity of the RdRP active site (yellow) of RdRp region of NS5. The RNA template is also shown with the incoming nucleoside (grey circle). D. Cells transfected with wild-type or specified alleles were treated with the nucleoside analogs as indicated. Mean±SD shown (n ≥ 3) and are normalized either to wild-type controls (indicated in green stars) or to the untreated condition of the respective samples (indicated in red stars). Statistical significance was determined using one-way ANOVA with Dunnett’s correction for multiple comparisons (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

https://doi.org/10.1371/journal.ppat.1013970.g008

In addition, we identified a previously unreported A534S mutation located within a conserved region of the NS5 RNA-dependent RNA polymerase (RdRP) domain (Figs 8A and S10B). This amino acid is conserved across flaviviruses (S10B Fig) and was present in one of the NITD-008-resistant mutants and in both MK-0608-resistant mutants. However, UPGNUC558 resistant strains did not encode this change.

The WGS analysis of UPGNUC255-resistant strains revealed a R355Q mutation in the conserved RdRp domain of NS5 (Figs 8A and S10A), which was present in both UPGNUC255-resistant strains (UPGNUC255 R4, R5 at passage 14) and in one NITD008-resistant strain (NITD008 R2). This amino acid is also conserved across flaviviruses (S10A Fig) and was not been previously reported among flaviviruses strains resistant to nucleoside analogs. We directly sequenced the residues around R355 and found that two additional strains resistant to UPGNUC255 also had this R355Q change (S10D Fig).

In order to visualize the location of the mutations in context, we modeled the Kunjin NS5 structure. The starting point was an AlphaFold 3 model of Dengue Type 2 NS5 in complex with an RNA template, growing strand, and incoming ATP nucleotide. A homology model of Kunjin was made based on this template and used to visualize the relative positions of the mutations with respect to the active site (GDD, yellow) and nucleic acids (Fig 8B-8C). Resistance-associated residues were mapped onto the RdRp domain of NS5 at positions S604, A534 (Fig 8B), and R355 (Fig 8C). The modeling revealed that the active nucleotide binds at the catalytic site, centered around the conserved GDD motif (yellow). Notably, residues S604 (green) and R355 (ball and stick) are located in close proximity to the active site, suggesting that they are gating the incoming nucleosides. In contrast, A534 (green) is positioned more distally and thus is less likely to be playing a role in nucleoside selection.

Replicons confirm single amino acid changes confer resistance

To determine whether these amino acid changes confer resistance to the selected nucleosides we utilized a WNIIrep-GFP (WN 956 D117 3B) subgenomic replicon [108,109]. This subgenomic replicon encodes a genome capable of autonomous viral RNA replication but harbors a 5’ deletion of the structural genes C, prM and E, which are replaced with a GFP reporter. This replicon unable to generate infectious virions, but offers a sensitive approach to measure the impact of amino acid substitutions at specific residues on viral RNA replication [108,109].

As controls, we used the S135A mutant of NS3 and the D668A mutant of NS5 as these are replication-deficient strains. The S135A mutation renders the NS3 protease inactive, thereby inhibiting viral polyprotein processing, while the D668A mutation disrupts the RNA-dependent RNA polymerase (RdRp) activity of NS5, abrogating viral RNA synthesis [109]. We introduced individual A534S, S604T and R355Q mutations into the WNV replicon using site-directed mutagenesis. These mutant replicons were transfected into HEK293T cells to assess the impact of each mutation on viral replication in the presence and absence of nucleoside analogs, as measured by automated microscopy. First, we validated the antiviral activity of our nucleoside analogs in HEK293T cells against KUNV, observing a ≥ 10-fold reduction in viral RNA levels upon treatment with the compounds (S10E Fig). Upon transfection, the wild type (wt) replicon resulted in approximately, 50% of cells expressing the GFP reporter, whereas, polymerase dead (D668A) and protease dead (S135A) mutants exhibited more than 90% reduction in GFP expression compared to the wt replicon, confirming that GFP expression is dependent on RNA replication (Fig 8D). A534S, R355Q, and S604T each exhibited attenuated replication, reducing the percentage of cells expressing GFP by ~25%, 30% and 50%, respectively, compared to wt. This suggests that these mutations negatively impact viral fitness (Fig 8D).

Drug sensitivity assays revealed that treatment with NITD-008 and UPGNUC558 reduce wt and A534S more than 75% indicating that the A534S mutation does not confer resistance to these compounds (Fig 8D). In contrast, the S604T mutation abrogated antiviral activity of both NITD-008 and UPGNUC558, suggesting that this mutation confers resistance to these nucleoside analogs (Fig 8D).

As expected, the R355Q mutation did not alter sensitivity to NITD-008 or UPGNUC558 but conferred a moderate level of resistance to UPGNUC255. While UPGNUC255 effectively inhibited replication in wt replicon, it was significantly less effective against the R355Q mutant (Fig 8D). These results indicate that S604T confers resistance to both NITD-008 and UPGNUC558 while R355Q specifically impairs the efficacy of UPGNUC255. These findings provide critical insights into the antiviral mechanism of UPGNUC255, highlighting NS5 as a primary target and underscoring key resistance-associated mutations that cluster near the RdRp active site.

Cells

A549 obtained from ATCC (CCL-185) and Huh7.5 (C. Rice, Rockefeller) were cultured in DMEM, supplemented with 10% (v/v) fetal bovine serum, 1% (v/v) penicillin/streptomycin (Invitrogen), 1% (v/v) L-Glutamax (Invitrogen) and were maintained at 37°C and 5%. CO₂. HepG2 cells (ATCC HB-8065) were obtained from ATCC and cultured in MEM, supplemented with 10% (v/v) fetal bovine serum, 1% (v/v) penicillin/streptomycin, 1% (v/v) L-glutamine, and were maintained at 37°C and 5% CO₂. IMR90 cells (CCL-186) were obtained from ATCC and cultured in DMEM-20 (4.5g/L glucose with Na-pyruvate) supplemented with 20% (v/v) fetal bovine serum and were maintained at 37°C and 5% CO₂. HEK293T cells were acquired from ATCC and propagated in DMEM supplemented with 10% heat-inactivated FBS and 1% (v/v) penicillin/streptomycin (Invitrogen), 1% (v/v) L-Glutamax (Invitrogen) and were maintained at 37°C and 5% CO₂. Cell lines were validated to be free of Mycoplasma contamination using a Mycostrip kit (Invivogen).

Primary human keratinocytes and dermal fibroblasts were obtained from the UPenn Skin Biology and Diseases Resource-based Center (P30-AR069589). Briefly, neonatal foreskins were collected in DMEM, 20% FBS, antibiotic/antimycotic (Gibco) and gentamicin (Gibco). Foreskins were dissociated overnight in dispase (Stem Cell Technologies) to separate epidermis from dermis. Epidermal sheets were incubated in 0.05% trypsin-EDTA (Invitrogen), and dissociated keratinocytes were collected by centrifugation and resuspended in keratinocyte culture medium and Antibiotic/Antimycotic (Gibco). Keratinocyte culture medium contains a 1:1 mixture of Keratinocyte Serum Free Media supplemented with recombinant epidermal growth factor and bovine pituitary extract (Gibco 17005042) and Medium 154 (Gibco M154500) supplemented with human keratinocyte growth supplement (Gibco S0015) and maintained at 37°C and 5% CO₂. Dermis was dissociated in collagenase followed by incubation in 0.5% trypsin-EDTA (Invitrogen). Dissociated fibroblasts were collected by centrifugation and resuspended in fibroblast culture medium containing DMEM, 10% FBS, 10 mM HEPES, supplemented with penicillin, streptomycin, and 250 µg/ml Fungizone (Gibco) and maintained at 37°C and 5% CO₂. Keratinocytes and fibroblasts were used for experiments between passages 2–4.

Virus stocks

DENV Type 2 (New Guinea C) from ATCC. DENV Type 1 Strain UIS 998 (NR-49713), DENV Type 3 Strain US/BID-V1043/2006 (NR-43282), DENV Type 4 Strain UIS 497 (NR-49724) and YFV 17D from BEI resources (ATCC). ZIKV (MR766) and Kunjin were obtained from Dr. R. Tesh (The World Reference Center for Emerging Viruses and Arboviruses (WRCEVA) at UTMB) and ZIKV (Puerto Rico) C. Coyne (Duke) and WNV (NY2000) M. Diamond (Washington University). All viruses were propagated in C636 and tittered by TCID₅₀ assay performed on BHK-21 cells and sequence verified.

Antibodies

Monoclonal antibodies against flavivirus glycoproteins (4G2) were generated UPenn in-house facility using hybridomas. Fluorescent secondary antibodies were from Life technologies. Hoechst 33342 was from Sigma-Aldrich.

High-throughput screening

Cells were plated in 384 well plates 3,000 cells per well for Huh7.5, A549, HepG2 and IMR90. The next day, 50 μl of drugs were added to the final concentration of 50 μM. The library of nucleoside analogs were purchased from Granlen and Selleckchem. Positive controls NITD-008, AT-527 (10uM) and negative controls (DMSO) were spotted on each plate. One hour after the drug treatment, cells were infected with either DENV-2 (Huh7.5 MOI = 2, A549 MOI = 2, HepG2 MOI = 1 and IMR90 MOI = 5), WNV (Huh7.5 MOI = 5) for 24 h or with ZIKV (Huh7.5 MOI = 0.05) for 48h. Cells were fixed in 4% formaldehyde in PBS for 15 min at room temperature followed by 3x washes with PBS. Cells were blocked with 2% BSA in PBST for 60 minutes and incubated in primary antibody (anti-4G2) overnight at 4°C. Cells were washed 3x in PBST with an automated plate washer (BioTek) and incubated in secondary antibody (anti-mouse Alexa 488, 1:1,000 and Hoescht 33342) for 1 h at room temperature. Cells were washed 3x in PBST and imaged using an automated microscope (ImageXpress Micro, Molecular Devices) with a × 10 objective, and four sites per well were captured. The total number of cells and the number of infected cells were measured using the cell scoring module (MetaXpress 5.3.3), and the percentage of infected cells was calculated. The aggregated percent infection of the DMSO (n = 32) and 10 µM positive control wells (n = 16) on each assay plate were used to calculate z′-factors, as a measure of assay performance and data quality. Sample well infection was normalized to aggregated DMSO plate control wells and expressed as percentage of control (POC = (% infection_sample/average % infection_DMSO) × 100) and Z score (Z = (% Infection_sample − average % Infection_DMSO)/Standard Deviation %infection_DMSO) in Spotfire (PerkinElmer). Candidate hits were selected as compounds with >60% reduction in infection and maintained >80% cell viability in either replicate compared to DMSO control.

Dose–response studies

Candidate drugs were repurchased as powders from Granlen, Selleckchem, MedChemExpress, and MedKoo and suspended in DMSO. Drugs were arrayed in 8-pt dose-response in 384 well plates. Infections were performed using the screening conditions. DMSO (n = 32) and 10 μM positive control drugs (n = 16) were included on each validation plate as controls for normalization. Infection at each drug concentration was normalized to aggregated DMSO plate control wells and expressed as percentage-of-control (POC = % Infection _sample/Avg % Infection DMSO control). Cytotoxicity CC₅₀ in the absence of viral infection was assessed using the ATPlite luminescence assay (PerkinElmer), which quantifies intracellular ATP levels as an indicator of metabolically active (viable) cells. Luminescence intensity, directly proportional to ATP content, served as a surrogate measure of cell viability.

A non-linear regression curve fit analysis (GraphPad Prism 9) was performed on the aggregated average POC Infection and cell viability from ≥ 2 independent experimental replicates versus the log₁₀ transformed concentration values to calculate IC₅₀ values for Infection and CC₅₀ values for cell viability for each drug/cell line combination. Error bars represent the standard deviation of replicate data for each drug concentration tested in independent experiments. Selectivity index (SI) was calculated as a ratio of drug’s CC₅₀ and IC₅₀ values (SI = CC₅₀/IC₅₀).

RT-qPCR

Huh7.5 (200,000 cells/well), A549 cells (200,000 cells/well), HepG2 (400,000 cells/well), IMR90 (200,000 cells/well), were plated in 12 well plates. Next day, cells were treated with indicated drugs to the final concentrations of 2 μM and 10 μM including positive controls in each experiment. One hour later cells were infected with DENV2 (Huh7.5 MOI = 0.05; A549 MOI = 0.05; HepG2 MOI = 0.5 and IMR90 MOI = 0.5). For Primary keratinocytes and fibroblasts 200,000 cells/well and 150,000 cells/well were plated in 12 well plates, respectively. 48 hours after plating cells were treated with indicated drugs to the final concentration of 10 μM including positive controls in each experiment. Cells were infected with DENV2 at MOI = 0.5 with spin at 1250 rcf for 1 hr. For other viruses, Huh7.5 cells were plated at 200,000 cells/well density and next day were infected with WNV (MOI = 0.5), ZIKA (MOI = 0.5), Kunjin (MOI = 0.5), YFV (MOI = 0.5), DENV1 (MOI = 1), DENV3 (MOI = 0.5), DENV4 (MOI = 0.5). Total RNA was purified using Trizol (Invitrogen) followed by RNA Clean and Concentrate kit (Zymo Researc) 24 hpi for Huh7.5, A549, HepG2 and IMR90 cells, whereas, 48 hpi for primary keratinocytes and fibroblasts cells. For cDNA synthesis, reverse transcription was performed with random hexamers and Moloney murine leukemia virus (M-MLV) reverse transcriptase (Invitrogen). Gene specific primers to each virus (listed in S3 Table) and SYBR green master mix (Applied Biosystems) were used to amplify viral RNA and 18S rRNA primers were used to amplify cellular RNA in the QuantStudio 6 Flex RT-PCR system (Applied Biosystems). Relative quantities of viral and cellular RNA were calculated using the standard curve method (Larionov et al., 2005). Viral RNA was normalized to human 18S RNA for each sample.

Selection of resistance mutants

KUNV was sequentially passaged in Huh7.5 cells in gradually increasing concentration of nucleoside analogs to select for drug resistance. In brief, Huh7.5 cells were plated at 100,000 cells/well monolayer in 24 well plates. 24h later the cells are infected with KUNV in media for 2h at 37°C and washed with cold PBS and media replenished with vehicle or indicated nucleoside analog. 48 hours later the supernatants were harvested from each well after the occurrence of cytopathic effect (CPE) from day 2 onwards. 100 µl of supernatant is used to reinfect fresh cell monolayers for subsequent passages, gradually increasing drug concentration up to 10 times the IC₅₀ value. Virus was expanded in C6/36 cells and titered.

Whole Genome Sequencing data analysis

Whole-genome sequence analysis was performed on virus stocks, including mutants and DMSO control virus strains. Short read single-end viral whole genome sequencing was performed on a NextSeq 2000 P2 100 Cycle kit with a read length of 118 bp. Fastq files were generated and then compared to a reference sequence for KUNV (AY274504). Reads were then mapped to the reference genome by breseq 0.36.0 (generated by the Barrick lab) using bowtie2 2.4.1, and subsequent identification of mutations relative to the reference sequence were done using breseq 0.36.0 in polymorphism mode (assuming mixed sample populations of variants sequenced) with the number of processors to use in multithreaded steps set to 8 and a 5% mutation frequency cutoff. The breseq gdtools compare command was then used to generate html files reporting mutation frequency across samples relative to the reference genome [123].

Plasmids and generation of mutants

WNV lineage II Replicon encoding GFP was a generous gift from Dr. Ted Pierson at the NIH. We used wild type and S135A (protease dead) mutant of NS3 and the D668A (polymerase dead) mutant of NS5 were used as replication-deficient controls, as these mutations render the virus incapable of RNA replication. To generate mutants, primers were ordered from Integrated DNA Technologies (IDT) encoding R → 355Q, A → 534S or S → 604T mutations in NS5. Site-directed mutagenesis was performed using InFusion cloning (Takara). The sequence of all replicon plasmids was validated using Oxford nanopore sequencing (Plasmidsaurus).

Cell transfection and replicon assays

HEK293T cells were seeded into 6-well plates and transfected with replicon plasmid using Lipofectamine 3000 reagent and treated with the indicated nucleoside analogs and vehicle (DMSO). 24 h post-transfection, cells were replated in 96-well plates with the addition of indicated nucleoside analogs for automated immunofluorescence analysis. 48 hours post-transfection cells were fixed with 4% paraformaldehyde and processed for automated imaging in an automated microscope (ImageXpress Micro, Molecular Devices). The resulting images were analyzed using MetaXpress software.

Modeling

Molecular modeling of the Kunjin structure was based on an AlphaFold3 model of Dengue2 (WLD15668.1). We used the DENV2 sequence because the KUNV sequence was rejected by the biosecurity sequence filters and could not be directly modeled. Nevertheless, the Dengue polymerase model (66% identity to Kunjin), with included RNA strands and incoming ATP nucleotide, provided a template for modeling of the Kunjin enzyme. Homology modeling was performed in the MOE environment (version 2022, Chemical Computing Group). Locations of mutations and active site residues were visualized on the homology model.

Quantification and statistical analysis

P-values for RT–qPCR experiments were obtained by performing one-way ANOVA, assuming equal SD, with multiple comparisons using Dunnett’s correction for multiple comparisons on log₁₀-transformed values on relative copy number values from at least three independent experiments. For TCID50 experiments, statistical significance was determined using one-way ANOVA with Dunnett’s correction for multiple comparisons was performed on titers from at least three independent experiments. For multiple comparisons, each condition was compared with vehicle control (DMSO). Visualization of data was performed using GraphPad Prism 9. The statistical parameters for experiments are described in the figure legends. n indicates the number of independent experiments performed, and significance is defined as P < 0.05.