Research ArticleInfectious Disease

Prunin suppresses viral IRES activity and is a potential candidate for treating enterovirus A71 infection

See allHide authors and affiliations

Science Translational Medicine  30 Oct 2019:
Vol. 11, Issue 516, eaar5759
DOI: 10.1126/scitranslmed.aar5759

A translational antiviral

Human enterovirus A71 (HEVA71) is a major cause of hand, foot, and mouth disease, which can lead to severe neurological complications including fatality; currently, however, HEVA71 lacks effective treatment options. Gunaseelan et al. screened for compounds that targeted the HEVA71 internal ribosome entry site and thereby impeded viral, but not human, protein synthesis. They found that the flavonoid prunin effectively inhibited HEVA71 protein synthesis in vitro and decreased viral loads and mortality in a mouse model of infection. Prunin may therefore warrant further investigation as a potential treatment option against hand, foot, and mouth disease.


Human enterovirus A71 (HEVA71) causes hand, foot, and mouth disease (HFMD) in young children and is considered a major neurotropic pathogen but lacks effective antivirals. To identify potential therapeutic agents against HFMD, we screened a 502-compound flavonoid library for compounds targeting the HEVA71 internal ribosome entry site (IRES) that facilitates translation of the HEVA71 genome and is vital for the production of HEVA71 viral particles. We validated hits using cell viability and viral plaque assays and found that prunin was the most potent inhibitor of HEVA71. Downstream assays affirmed that prunin disrupted viral protein and RNA synthesis and acted as a narrow-spectrum antiviral against enteroviruses A and B, but not enterovirus C, rhinovirus A, herpes simplex 1, or chikungunya virus. Continuous HEVA71 passaging with prunin yielded HEVA71-resistant mutants with five mutations that mapped to the viral IRES. Knockdown studies showed that the mutations allowed HEVA71 to overcome treatment-induced suppression by differentially regulating recruitment of the IRES trans-acting factors Sam68 and hnRNPK without affecting the hnRNPA1-IRES interaction required for IRES translation. Furthermore, prunin effectively reduced HEVA71-associated clinical symptoms and mortality in HEVA71-infected BALB/c mice and suppressed hepatitis C virus at higher concentrations, suggesting a similar mechanism of prunin-mediated IRES inhibition for both viruses. These studies establish prunin as a candidate for further development as a HEVA71 therapeutic agent.


Human enterovirus 71 (HEVA71) was discovered in the late 1960s in patients in California who exhibited severe symptoms of the central nervous system diseases (1). Since then, large-scale epidemics of HEVA71 have been observed in Asia-Pacific countries involving millions of young children under 3 years old (24). These children were diagnosed with mild and self-limiting hand, foot, and mouth disease that began with the emergence of fever and blisters on hands, feet, and mouth cavities. Over time, a substantial subset of these patients displayed serious neurological deficits in the form of aseptic meningitis, brain stem encephalitis, and poliomyelitis-like acute flaccid paralysis, which ultimately induced fatality or permanent neurological sequelae (5, 6). Given the present lack of authorized antivirals for HEVA71 (7, 8) along with the eradication of poliovirus through successful vaccination (9), HEVA71 is a major worldwide nonpolio neurotropic virus.

As a representative of the Picornaviridae family of viruses, HEVA71 is a small, nonenveloped enterovirus with a positive-sense single-stranded RNA genome. Its 7.4-kb genome begins with a 5′ untranslated region (5′UTR) and terminates with a 3′UTR, with its open reading frame translated into four capsid proteins (VP1–4) and seven nonstructural proteins (2A–C and 3A–D) (10). Of interest is the internal ribosome entry site (IRES) that lies within stem loops 2 to 6 of the 5′UTR. These IRES elements were initially identified in Picornaviridae RNA viruses with long 5′UTRs lacking a 5′ cap structure (11, 12). By folding into secondary RNA structures, the IRES substitutes some of the functions of host translation initiation factors, thereby permitting cap-independent translation of viral proteins (13). Furthermore, cap-dependent translation of host cell proteins is decreased via HEVA71 IRES–encoded viral proteases, namely, 2Apro and 3Cpro, which cleave host translation factors such as eIF4G to favor IRES-mediated translation (14). Hence, it is crucial for competent antiviral strategies to target the HEVA71 IRES–facilitated protein synthesizing machinery so that host cell translation can proceed normally.

IRES elements have been used in biotechnology to construct bicistronic or polycistronic vectors where the IRES was positioned between two cistrons (1517), thereby facilitating the expression of two proteins concurrently, one from a reporter gene and another from a gene of interest. The use of IRES sites in mediating gene translation aids in establishing simultaneous expression of the pair of cistrons in more than 90% of the cells (17), without affecting the ratio of protein expression by both cistrons positioned before and after IRES elements (17, 18). IRES also allows for the translation of RNAs without the 5′ cap structures synthesized by RNA polymerase I and III (1922). Last, IRES cap-independent translation can be used in circumstances where cap-dependent host cell translation is impaired, such as during programmed cell death and cell cycle checkpoints (23). Overall, these features enable IRES elements to be potentially advantageous in genetic engineering.

In this context, we used reverse genetics to generate a functional bicistronic reporter vector consisting of a HEVA71 IRES where cap-dependent translation could be distinguished from HEVA71 IRES–mediated translation. Recently, natural products such as artemisinin and its derivatives from Artemisia plants were found to be effective against malaria parasites and were successfully introduced into the clinical market as a U.S. Food and Drug Administration (FDA)–approved drug for malaria (24). This prompted us to investigate the effects of natural compounds like flavonoids on HEVA71 IRES–mediated translation to address the demands for HEVA71 antiviral options.


A bicistronic assay system allows screening of a flavonoid library

To discover compounds capable of discriminating host-mediated cap-dependent protein synthesis and IRES-facilitated cap-independent translation, we screened a 502-compound library of flavonoid derivatives using a HEVA71 IRES bicistronic construct for antivirals suppressing HEVA71 IRES translation without affecting host cell translation. Figure 1A illustrates the model from which we generated a HEVA71 IRES–containing bicistronic luciferase reporter vector. This model has shown efficacy in searching for potent antivirals against IRES-mediated protein-synthesizing machinery without affecting host translation (25).

Fig. 1 Experimental design, generation, and relative activities of bicistronic reporter vectors.

(A) Strategy for generating a bicistronic construct to discover antivirals that reduce the synthesis of protein 2 by inhibiting IRES-mediated translation without affecting protein 1 production by cap-dependent translation. (B) Schematic of the HEVA71 IRES bicistronic reporter used in this study. The HEVA71 IRES is located in between luciferase genes R Luc and F Luc, which are, in turn, placed downstream of a CMV promoter. R Luc is translated into R Luc protein (red) by cap-dependent translation, whereas F Luc is translated into F Luc protein (pink) by cap-independent translation via the HEVA71 IRES. (C) Schematic illustration of the HEVA71 IRES bicistronic hairpin used in this study. The HEVA71 IRES hairpin is located in between luciferase genes R Luc and F Luc, which are, in turn, placed downstream of a CMV promoter. R Luc gene is translated into R Luc proteins (red) through cap-dependent translation, whereas F Luc gene is not translated into F Luc proteins (pink) because of the presence of the HEVA71 hairpin structure. Relative (D) IRES (Firefly) and (E) CMV (Renilla) activities of the HEVA71 IRES bicistronic reporter (BICIS) and HEVA71 IRES bicistronic hairpin (BICIS HP) at 12 and 24 hours after transfection. Normalized relative IRES (Firefly), CMV (Renilla), and IRES:CMV (F Luc:R Luc) activities of both (F) HEVA71 IRES bicistronic reporter and (G) HEVA71 IRES bicistronic hairpin at 24 hours after transfection with 1 and 2 mM amantadine hydrochloride. One-way ANOVA with Dunnett’s correction, ****P < 0.0001; ***P < 0.001; ns, not significant.

For the flavonoid library screen, we constructed a HEVA71 IRES bicistronic luciferase reporter (Fig. 1B) consisting of a human cytomegalovirus promoter sequence (CMV promoter) at its start point, with downstream Renilla luciferase (R Luc) and Firefly luciferase (F Luc) genes flanking the HEVA71 IRES site. The human CMV promoter facilitates transcription of R Luc and F Luc genes; these transcripts are subsequently translated into R Luc and F Luc proteins in cap-dependent and cap-independent manners, respectively. Hence, this reporter assay could likely determine compounds that inhibit IRES-mediated F Luc expression without influencing cap-dependent R Luc translation. We also included another strategy of replacing the HEVA71 IRES element of the bicistronic construct with an IRES hairpin structure (Fig. 1C), which was found to interrupt IRES activity (26). This construct serves as a positive control, where cap-independent IRES translation of F Luc is prevented with no effects on cap-facilitated R Luc protein synthesis.

Functional characterization of the HEVA71 IRES bicistronic reporter and HEVA71 IRES bicistronic hairpin

To characterize these constructs, we measured luciferase expression by both bicistronic reporters that we separately transiently transfected into human rhabdomyosarcoma (RD) cells at two different time points (12 and 24 hours after transfection). As predicted, the HEVA71 IRES bicistronic reporter exhibited higher IRES-mediated F Luc production and similar amounts of cap-dependent R Luc expression compared with the HEVA71 IRES bicistronic hairpin at both time points (Fig. 1, D and E). Moreover, we observed significant increases in IRES-facilitated (P = 0.0008) and CMV-facilitated (P = 0.00006) F and R Luc synthesis in HEVA71 IRES bicistronic reporter-transfected RD cells from 12 to 24 hours, indicating high translational efficiency for up to 24 hours: 3.91 to 4.57 log10RLU for F Luc and 3.38 to 3.88 log10RLU for R Luc from 12 to 24 hours (Fig. 1, D and E). The HEVA71 IRES bicistronic hairpin showed only a rise in R Luc expression (3.22 to 3.74 log10RLU) with no significant change (P = 0.28) in F Luc synthesis (2.61 ≈ 2.71 log10RLU) from 12 to 24 hours due to the presence of its IRES hairpin structure, as expected (Fig. 1, D and E).

We further evaluated the functionalities of both constructs using a known anti-HEVA71 IRES compound, amantadine hydrochloride (25). This compound disrupts influenza A replication by binding matrix (M2) protein (27, 28) and interferes with hepatitis A, hepatitis C, and HEVA71 IRES translational activity, possibly by binding to IRES trans-acting factors in humans (25, 2931). We transfected RD cells with either the HEVA71 IRES bicistronic reporter or the HEVA71 IRES bicistronic hairpin and incubated them with amantadine (1 and 2 mM) at 12 hours after transfection for a duration of 24 hours. After normalizing by transfected RD cells without treatment, we found that amantadine reduced HEVA71 IRES activity (F Luc) by 52% (P = 0.00006) and 82% (P = 0.00004) in a dose-dependent manner (Fig. 1F) with minimal effects on CMV activity (R Luc) of the HEVA71 IRES bicistronic reporter. Amantadine led to corresponding declines in F Luc–to–R Luc expression ratios by 47 and 80% with increasing concentrations of amantadine (Fig. 1F). In contrast, we observed no significant changes in F Luc (P = 0.23) and R Luc (P = 0.44) expression and ratios of HEVA71 IRES bicistronic hairpin (P = 0.35) with amantadine (Fig. 1G). These results indicated that amantadine is suitable as another positive control along with the HEVA71 IRES bicistronic hairpin for downstream compound screening. Moreover, amantadine increased the potential of both reporter constructs to accurately pinpoint HEVA71 IRES suppressors during high-throughput compound screening.

Heterogeneous antibiotic selection enhances consistent luciferase expression

Although we validated the functionality of both bicistronic constructs through transient transfection, genes that are transiently transfected tend to lose their expression over time, which could limit the duration of our treatment studies and introduce bias into gene expression profiles. To approach this problem, we studied the luciferase expression of RD cells transfected with either the HEVA71 IRES bicistronic reporter or the HEVA71 IRES bicistronic hairpin and selected for successfully transfected cells with geneticin (G418) selection over time (0, 4, 6, and 8 days) at 24, 48, and 72 hours (Fig. 2A). We used a working G418 concentration of 0.25 mg/ml, at which 50% of RD cells survived after a week with continuous G418 treatment (Fig. 2B).

Fig. 2 Heterogeneous selection and consistency of luciferase expression.

(A) Process of heterogeneous G418 antibiotic selection of HEVA71 IRES bicistronic reporter or HEVA71 IRES bicistronic hairpin. (B) Relative cell viability profiles of RD cells treated with G418, where red dashed lines represent corresponding G418 concentration (0.25 mg/ml) for 50% cell viability (CC50). Relative IRES (Firefly) (C) and CMV (Renilla) (D) activities of HEVA71 IRES bicistronic reporter at 24, 48, and 72 hours after 0, 4, 6, and 8 days of selection with G418 (0.25 mg/ml). Relative IRES (Firefly) (E) and CMV (Renilla) (F) activities of HEVA71 IRES bicistronic hairpin at 24, 48, and 72 hours after 0, 4, 6, and 8 days of selection with G418 (0.25 mg/ml). Relative IRES:CMV (F Luc:R Luc) activities and calculated Z factors to measure luciferase expression consistencies of (G) BICIS and 2 mM amantadine–treated BICIS, and (H) BICIS and BICIS HP at 48 hours after 6 days after selection. Data are expressed as means ± SD from three independent experiments consisting of triplicates. One-way or two-way ANOVA with Dunnett’s or Tukey’s correction. ****P < 0.0001; ***P < 0.001.

The relative IRES and CMV activities of the HEVA71 IRES bicistronic reporter gradually increased from 0 to 6 days after G418 selection and peaked at 5.41 and 4.81 log10RLU, respectively, after 48 hours (Fig. 2, C and D). Beyond 8 days of G418 selection, luciferase expression decreased rapidly until it was significantly lower at all time points than at the onset of G418 treatment: 3.79 < 4.57 log10RLU (24 hours; P = 0.00003), 4.42 < 4.80 log10RLU (48 hours; P = 0.00003), and 3.33 < 4.18 log10RLU (96 hours; P = 0.00005) for F Luc, and 3.21 < 3.89 log10RLU (24 hours; P = 0.00003), 3.92 < 4.26 log10RLU (48 hours; P = 0.00002), and 2.89 < 3.62 log10RLU (96 hours; P = 0.00002) for R Luc (Fig. 2, C and D). The HEVA71 IRES bicistronic hairpin did not increase its IRES activity over the time course of antibiotic selection (Fig. 2E) because of its IRES hairpin structure, although we observed progressive elevation in its CMV activity, which reached a maximum of 4.57 log10RLU at 48 hours after 6 days after G418 selection (Fig. 2F). The CMV and IRES elements of the bicistronic hairpin–facilitated luciferase expression displayed an identical pattern of deterioration after 8 days of G418 selection, where luciferase measurements of R Luc and F Luc hit below the start of G418 treatment at all time points: 3.07 < 3.67 log10RLU (24 hours), 3.85 < 3.97 log10RLU (48 hours), and 2.85 < 3.37 log10RLU (96 hours) for R Luc, whereas 2.48 < 2.81 log10RLU (24 hours), 2.51 < 2.83 log10RLU (48 hours), and 2.20 < 2.53 log10RLU (96 hours) for F Luc (Fig. 2, E and F). This accelerated depletion of luciferase signals by both constructs could potentially be due to the plasmid loss over time, as the bicistronic constructs have large plasmid sizes of more than 7.5 kb. Nevertheless, heterogeneous G418 selection aided in extending the treatment window by 6 days for observation of downstream treatment effects. G418 treatment also reduced the heterogeneity of transfected population of RD cells, thereby increasing the potential use of the assay for compound screening purposes, particularly for detecting HEVA71 IRES inhibitors.

After the establishment of a suitable extended therapeutic window for downstream large-scale screening, we assessed the robustness of the assay by calculating the effect size (Z factor) between the HEVA71 IRES bicistronic vector and the associated positive controls, namely, the HEVA71 bicistronic hairpin and HEVA71 IRES bicistronic reporter with 2 mM amantadine. On the basis of the ratios of IRES to CMV activity (n = 60), we calculated Z factors to be 0.859 between the HEVA71 IRES bicistronic reporter and bicistronic hairpin (Fig. 2G), and 0.646 between the HEVA71 IRES bicistronic reporter and its treated counterpart (Fig. 2H). According to published literature, these Z factors fell between robust cutoff values (32) of 0.5 to 1.0, demonstrating that the luciferase-screening assay was a useful platform for downstream treatment studies.

Screening a flavonoid derivative library identifies seven compounds that suppress HEVA71 IRES activity

After 6 days of G418 selection, we exposed HEVA71 IRES bicistronic reporter–transfected RD cells to flavonoids at a concentration of 20 μg/ml and measured their subsequent cell viability and luciferase activity. Normalized ratios of IRES-to-CMV expression were calculated and converted to percentage inhibitions of HEVA71 IRES activity for each flavonoid (table S1). We also included the HEVA71 IRES bicistronic hairpin or amantadine treatment as separate positive controls in this study. We observed that the 25 top flavonoid hits suppressed 55 to 62% of HEVA71 IRES activity, which was more than that in the amantadine treatment (40%), although only 13 flavonoids exhibited low cytotoxicity profiles (more than 80% cell viability; Fig. 3 and Table 1). We were unable to detect prior publications or patents pertaining to HEVA71 for a subset of the hits, namely, ST077105, ST024699, ST024368, ST002086, ST024702, ST066904, and ST024081 (Table 1). We therefore chose these seven compounds for a series of downstream validation assays to confirm their anti-HEVA71 activity.

Fig. 3 Screening profiles of the top 25 flavonoids.

HEVA71 IRES inhibition profiles (brown bars) and cell viability profiles (red dots) of the top 25 flavonoid hits, with 2 mM amantadine hydrochloride and HEVA71 IRES bicistronic hairpin (BICIS HP) as positive controls. Seven flavonoids (underlined in pink) with more than 80% cell viability (blue dashed line) were chosen for further analysis. Data are means ± SD from three independent experiments consisting of triplicates.

Table 1 Thirteen flavonoid hits ranked according to RD cell viability profiles (>80%).

Seven flavonoids (highlighted) were chosen for further evaluation.

View this table:

Functional validation of prunin as a potent inhibitor of HEVA71 with limited spectrum activity

To pinpoint the most suitable candidate for downstream characterization, we explored the effects of the seven chosen flavonoids on infectious HEVA71 by quantifying HEVA71 viral titers with plaque reduction assays at 12 hours postinfection (hpi). Six of the seven hits significantly (P = 0.0008) reduced HEVA71 viral titers. Prunin (ST077105), a compound found in certain immature citrus fruits (33, 34) and in minute amounts in tomatoes (35), most effectively suppressed HEVA71 viral titers [by about 1.8 log10 plaque-forming units (PFU)/ml; Fig. 4A] at 20 μg/ml (34.5 μM). Only one flavonoid, ST002086, showed no significant change (P = 0.52) in HEVA71 viral titer. We next examined and profiled the activity of prunin against HEVA71. We characterized the cytotoxicity of prunin by treating RD cells with a wide range of prunin doses, ranging from 100 nM up to a maximum concentration of 5 μM. With a threshold of 80% cell viability, we found that only concentrations of prunin below 1 μM were well tolerated by the RD cells after 24 hours (Fig. 4B); doses above 1 μM reduced cell viability below 80%. Using these data, we calculated the cytotoxic concentration resulting in 50% cell death (CC50) as 2715 nM.

Fig. 4 Downstream validation and functional studies of flavonoids.

(A) Inhibition profiles of seven chosen flavonoid compounds through HEVA71 viral titer quantification with plaque assay at 12 hpi. (B) Cell viability profile and calculated CC50 of prunin after 24 hours. Concentrations below 1 μM showed more than 80% cell viability (blue dashed line). Red dashed lines represent corresponding prunin dose (2715 nM) for 50% cell viability. (C) Inhibition (bars) and cytotoxicity (gray dots) profiles of prunin (31.25 to 1000 nM) through HEVA71 viral titer quantification with plaque assay at 12 hpi and cell viability assays, respectively. Concentrations of prunin used showed more than 80% cell viability (blue dashed line). (D) Tabulated EC50 of prunin. Red dashed lines represent corresponding prunin dose (115.3 nM) for 50% inhibition of HEVA71 viral titers. Disruption of HEVA71 viral protein production by prunin (31.25 to 1000 nM) measured at (E) 6 hpi and (F) 12 hpi via Western blot with band intensities of VP2 (blue solid lines) and VP0 (black dashed lines). Disruption of HEVA71 viral RNA synthesis by prunin (31.25 to 1000 nM) at (G) 6 hpi (red bars) and (H) 12 hpi (black bars) by quantitative reverse transcription polymerase chain reaction (qRT-PCR). (I) Inhibition profiles of prunin (31.25 nM) against HEVA71 clinical isolates, HEVA71 strain H, HEVA71 strain B5 genotype, HEVA71 strain C4 genotype, CA6, CA16, ECHO7, CB5, CA24, HRV10, HSV, and CHIKV through respective viral titer quantification with plaque assays at 12 hpi. DMSO (1%) was used as vehicle control. Data are means ± SD from three independent experiments consisting of triplicates. Statistical analyses were performed against 1% DMSO. One-way ANOVA with Dunnett’s correction, ****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05.

On the basis of the cytotoxicity profile of prunin, we decided to use a range of prunin concentrations from 31.25 to 1000 nM to determine the effectiveness of prunin on HEVA71 viral titers. At 12 hpi, we noted significant dose-dependent reductions (P = 0.00003 to 0.00008) in HEVA71 viral titers at doses between 62.5 and 1000 nM (Fig. 4C). Cytotoxicity assays showed that all doses resulted in cell viabilities above 80% (Fig. 4C). In addition, 1000 nM prunin significantly reduced HEVA71 viral titers (P = 0.00007) by about 3.5 log10 PFU/ml in comparison to 1% dimethyl sulfoxide (DMSO) (vehicle control), whereas 31.25 nM prunin did not result in any significant changes (P = 0.69) in HEVA71 viral titers. From these data, we calculated EC50 (median effective concentration) of prunin on HEVA71 viral titers as 115.3 nM (Fig. 4D).

We note that the EC50 (115.3 nM) and CC50 of prunin (2715 nM) were lower in comparison with the screening concentration (34.5 μM) used, which could be due to interbatch variability in the production and extraction of prunin from various companies. However, our flavonoid library hits (Table 1) also included compounds such as fisetin (≈69.9 μM), apigenin (≈74.1 μM), and luteolin (≈69.9 μM), which inhibited HEVA71 at 20 μg/ml, indicative of possibly only prunin being affected by interbatch variability. To ensure the reliability of prunin as a HEVA71 inhibitor, we evaluated AKOS02428449, a compound with the same structure as prunin, against HEVA71 (fig. S1), which yielded similar EC50 (145.2 nM) and CC50 (2918 nM) values as prunin.

We next investigated the effects of prunin on HEVA71 viral protein synthesis. RD cells were infected with HEVA71 [multiplicity of infection (MOI), 1] and treated with a range of prunin doses (31.25 to 1000 nM), including 1% DMSO (vehicle control), and then lysed at 6 and 12 hpi. Cell lysates were collected and analyzed for VP2, a known HEVA71 structural protein, along with precursor HEVA71 viral proteins P1, VP0, and (VP4 + VP2 + VP3). In contrast to the vehicle control, we noticed dose-dependent reductions in all probed viral proteins [VP2, P1, VP0, and (VP4 + VP2 + VP3)] from 62.5 to 1000 nM prunin treatment at both 6 hpi (Fig. 4E) and 12 hpi (Fig. 4F). No visible viral proteins could be detected with 1000 nM prunin treatment at 6 and 12 hpi, although prunin at a dose of 31.25 nM did not affect HEVA71 viral protein production at either point. With reference to actin and 1% DMSO, we analyzed the blots further by quantifying the band intensities of VP2 and precursor VP0 at 6 and 12 hpi. The observed dose-dependent decreases in VP2 and VP0 due to prunin (62.5 to 1000 nM) were significant (P = 0.00004 to 0.00008) at both time points compared with vehicle control. Prunin (1 μM) resulted in decreases in VP2 and VP0 band intensities at 6 hpi by 0.99 and 0.97 U and at 12 hpi by 0.99 and 0.98 U, respectively. However, it should be noted that 31.25 nM prunin did not significantly (P = 0.55) reduce band intensities of VP2 or VP0, in comparison with the vehicle control at either time point. These results were also seen with AKOS02428449 treatments (fig. S1) and were consistent with the effects of prunin on HEVA71 viral titers.

Because of the interaction between the HEVA71 RNA and protein synthesis machineries, we also studied the effects of prunin on viral RNA production. Infected RD cells treated with identical doses of prunin (31.25 to 1000 nM) exhibited similar dose-dependent reductions in HEVA71 viral RNA (P = 0.00003 to 0.0004) to HEVA71 protein reductions after treatment with 62.5 to 1000 nM prunin relative to 1% DMSO (vehicle control) at both 6 hpi (Fig. 4G) and 12 hpi (Fig. 4H). Specifically, 1 μM prunin significantly reduced viral RNA abundance from 8.4 to 6.45 log10 viral RNA copy number at 6 hpi (P = 0.00003) and 9.45 to 8 log10 viral RNA copy number at 12 hpi (P = 0.00008). Consistent with our previous analyses, 31.25 nM prunin did not result in any significant changes in HEVA71 RNA amounts at either time point (6 hpi, P = 0.88; 12 hpi, P = 0.67). This phenomenon was also seen with AKOS02428449 treatment (fig. S1).

Given that we identified prunin in a luciferase-based assay targeting IRES elements, we hypothesized that prunin would also suppress other members of the enteroviruses because of similarities in their IRES structures (10). We infected RD or bone marrow muscle (SJ) cells with various enterovirus species, including HEVA71 clinical isolates, HEVA71 strain H, HEVA71 strain B5 genotype, HEVA71 strain C4 genotype, CA6, CA16, ECHO7, CB5, CA24, HRV10, and other viruses such as herpes simplex virus 1 (HSV) and chikungunya virus (CHIKV). After infection, we treated the cells with the EC50 of prunin (115.3 nM) for 24 hours and quantified HEVA71 viral titers for each virus (Fig. 4I). Prunin significantly decreased viral titers of members of enterovirus A (HEVA71, CA6, and CA16) and enterovirus B (ECHO7 and CB5) species by about 3.0 to 3.5 log10 PFU/ml compared with vehicle control (P = 0.00002 to 0.00009). However, neither distantly related enteroviruses such as CA24 and HRV10, enterovirus C, and rhinovirus A species nor HSV or CHIKV was affected by prunin treatment (Fig. 4I). This was expected because CHIKV is an alphavirus of the Togaviridae family consisting of an RNA genome that is capped and expressed only through cap-dependent translation (36), whereas HSV contains a double-stranded DNA genome with no IRES elements (37). As a precaution, we performed cytotoxicity assays to ensure that SJ cells used for CHIKV infections were not vulnerable to prunin at a dose of 115.3 nM (fig. S2). These results classify prunin as a limited-spectrum enterovirus suppressor against enterovirus A (HEVA71, CA6, and CA16) and enterovirus B (ECHO7 and CB5) species.

Administration of prunin in a HEVA71 murine infection model shows therapeutic efficacy with minimal cytotoxicity effects

Having shown the efficacy of prunin against HEVA71 in vitro, we next evaluated the in vivo efficacy of prunin. One-day-old suckling BALB/c mice were injected with 2 × 107 PFU of HEVA71 per mouse and 1 or 6 hpi; prunin (1, 3, or 10 mg/kg) was administered by intraperitoneal injection once daily for 7 days. We monitored the HEVA71-infected mice daily for 14 days postinfection (dpi) and recorded the survival rate (Fig. 5A) and clinical manifestations (Fig. 5B). Infected mice treated with vehicle control [phosphate-buffered saline (PBS)] or prunin (1 mg/kg) exhibited 100% mortality by 7 dpi, whereas those treated with higher doses of prunin (3 and 10 mg/kg) lived to 14 dpi. Moreover, mice infected with HEVA71 presented with severe clinical symptoms such as ruffled hair, huddling, a sedentary appearance, limb weakness, rapid body weight loss, and eventual hindlimb paralysis (table S2). Mean clinical scores in vehicle- and prunin (1 mg/kg)–treated infected mice increased rapidly compared with the mice treated with larger doses of prunin (3 and 10 mg/kg) by 7 dpi. Although we did observe an initial gradual increase in the clinical scores (inactivity, hunched back, ruffled fur, and limb weakness) of the mice treated with 3 and 10 mg/kg of prunin up to 8 and 7 dpi, respectively, these mice completely recovered from HEVA71 infection and were healthy by 13 dpi (3 mg/kg) and 9 dpi (10 mg/kg). We also measured the body weight of mice given daily intraperitoneal injections of prunin (1 to 10 mg/kg) for 14 days. Prunin-treated mice showed increases in body weight (about 5.61 to 5.83 g from 0 to 14 dpi) similar (P > 0.9999 at all points from 0 to 14 dpi) to those of the mice treated with PBS, where a rise in body weight of 5.10 g was seen until 14 days after treatment (Fig. 5C). Overall, these data indicated prunin at all tested doses had negligible cytotoxicity on the suckling BALB/c mice and had protective efficacy against HEVA71 infection in mice.

Fig. 5 In vivo studies of prunin.

(A) Survival rates and (B) mean clinical scores of mice infected with HEVA71 and treated with prunin at various doses (1 to 10 mg/kg). Groups of six mice were infected via intraperitoneal routes with HEVA71 strain 41 at 2 × 107 PFU per mouse, after which prunin was administered from the day of infection (1 or 6 hpi) to 7 dpi and monitored for 14 dpi. (C) Body mass of mice exposed to daily prunin intraperitoneal injections, ranging from 1 to 10 mg/kg from 0 to 14 days. Data are means ± SD from two independent experiments. (D) Quantification of viral loads in hindlimb muscles of HEVA71-infected mice by plaque reduction assays. (E) H&E staining in prunin-treated (3 mg/kg) infected mice at 7 dpi. (F) IHC staining in prunin-treated (3 mg/kg) infected mice at 7 dpi. PBS was used as treatment control. All tests one-way ANOVA with Dunnett’s correction, **P < 0.01 and ***P < 0.0001. Scale bars, 50 μm.

To further examine the in vivo protective efficacy of prunin against HEVA71, we used plaque reduction assays to quantify the viral loads in hindlimb muscles from BALB/c mice treated with prunin or vehicle control and then euthanized at 7 dpi (Fig. 5D). Viral titers in the infected mice treated with prunin (3 mg/kg) were significantly lower by about 104 PFU/g than those treated with the vehicle control (P = 0.0054). Moreover, a trend of dose dependency (P = 0.00006) was observed, where a higher dose of prunin (10 mg/kg) resulted in a 105 PFU/g decrease in viral load in comparison with PBS-treated mice. We confirmed HEVA71 infection in the muscle tissues of hindlimbs from both groups in terms of tissue deterioration and viral antigen presence by hematoxylin and eosin (H&E) and immunohistochemistry (IHC) staining, respectively, at 7 dpi. Mice treated with PBS displayed loss of muscle fibers leading to severe damage of hindlimb muscle tissues (Fig. 5E). In addition, these mice showed signs of inflammation after HEVA71 infection, as evidenced by infiltration of immune cells into their hindlimb muscles (Fig. 5E). This was in line with previous work showing that neutrophils and macrophages accumulate and secrete proinflammatory cytokines in sites of HEVA71 infection, worsening muscle tissue destruction (38). These effects of muscle tissue damage, immune cell infiltration, and subsequent tissue inflammation induced by HEVA71 infection were suppressed by prunin (Fig. 5E). Consistent with the H&E staining, IHC studies detected extensive viral antigen distribution in the hindlimb muscles of infected mice treated with vehicle control, indicating progressive HEVA71 infection (Fig. 5F). However, prunin-treated mice showed a reduction in viral antigens in their hindlimb muscles, emphasizing the in vivo potency of the compound.

Development of HEVA71-resistant mutants against prunin results in five mutations in the viral IRES

To validate the mechanism of action of prunin against the HEVA71, we grew prunin-resistant viruses by repeated culturing of the wild-type (WT) virus for 19 passages with increasing doses of prunin. For every round of passaging, we also included untreated HEVA71 (1% DMSO) and WT HEVA71 controls. We tabulated treated and control viruses using plaque reduction assays. Representative results are plotted for passages 13, 16, and 19 (Fig. 6A). Despite treatment, HEVA71 exhibited significant initial resistance to 115.3 nM prunin at passage 13 in comparison with WT HEVA71 (P = 0.0007). This resistance improved over passages 16 and 19 (P = 0.00006) when prunin concentrations increased to 500 and 1000 nM, respectively. The HEVA71 mutant increased in concentration from about 5.8 to 7 log10 PFU/ml, even though WT HEVA71 exhibited a dose-dependent reduction in viral titers (4.3 to 3.4 log10 PFU/ml) with increasing concentrations of prunin (115.3 to 1000 nM). Significant differences in viral titers that were observed between nontreated and treated HEVA71 at passages 13 (7.0 versus 5.9 log10 PFU/ml, P = 0.00008) and 16 (7.0 versus 6.2 log10 PFU/ml, P = 0.0006) were lost at passage 19, where no significant disparity (P = 0.76) could be seen between viral titers (7.0 log10 PFU/ml), indicating the generation of prunin-resistant HEVA71 mutants.

Fig. 6 HEVA71 prunin-resistant mutant studies.

(A) Characterization of HEVA71-resistant mutant phenotypes with increasing doses of prunin (black) from passages 1 to 19. Quantification of respective viral titers at passages 13, 16, and 19 were performed through plaque reduction assays. DMSO (1%) (gray) was used as vehicle control. (B) Growth kinetics of WT and mutant HEVA71 from 0 to 96 hpi, (C) effects of synergistic treatments of ribavirin (1 mM) with or without prunin (115.3 nM) on WT and mutant HEVA71, and (D) downstream analysis of either single or combined point mutations on WT HEVA71, measured through plaque reduction assays. (E) Relative IRES (Firefly) and CMV (Renilla) activities of various individual or combined point mutations on WT HEVA71 IRES. Data are expressed as means ± SD from three independent experiments consisting of triplicates. (F) Differences in predicted IRES RNA secondary structures of stem loops 2 and 4 between WT and mutant HEVA71. Sites of mutations (T164C, G165C, G368C, T370G, and C177T) and causative variations in structures are marked in red boxes. All statistical tests are one-way ANOVA with Dunnett’s correction, ****P < 0.0001 and ***P < 0.001.

We characterized the prunin-resistant HEVA71 mutants using multiple approaches. First, we examined growth replication kinetics using plaque reduction assays. We observed no significant differences in viral titers of mutant HEVA71 versus WT HEVA71 from 0 to 96 hpi in RD cells (P = 0.73) (Fig. 6B). Both viruses achieved peaks of around 7 log10 PFU/ml by 96 hpi, indicating no replicative variation between them. Second, we investigated whether acquired resistance could be overcome by treatment with a secondary anti-HEVA71 compound. To this end, we used a broad-spectrum nucleoside analog, ribavirin, previously shown to be a HEVA71 3D polymerase suppressor in vitro and in vivo (3942). Ribavirin inhibits viral RNA replication by causing lethal mutagenesis in vitro (43) and shows anti-HEVA71 properties in vivo, where it reduces HEVA71-facilitated paralysis and mortality events (44). Ribavirin exposure reduced viral titers of both WT and mutant HEVA71 to about 1.2 log10 PFU/ml, whereas the antiviral effects of prunin were limited to WT HEVA71 (Fig. 6C). Moreover, treatment studies coadministering ribavirin and prunin resulted in significant declines in both WT and mutant viruses (P = 0.00006 to 0.00008), albeit with a greater inhibitory effect on WT HEVA71 titers. Notably, there were no significant differences seen in mutant HEVA71 viral titers after treatment with either ribavirin or a combination of prunin and ribavirin (P = 0.53 to 0.66) (Fig. 6C), suggesting that ribavirin was the sole contributor of the detected inhibition. Prunin resistance did not result from ribavirin treatment, indicating that the mutant HEVA71 functioned similarly to WT HEVA71 in terms of its replication fidelity, particularly by 3D polymerase.

Last, we sequenced the viral genomes of WT and mutant HEVA71 to find mutations conferring resistance to prunin. We identified numerous silent mutations in polymerase regions VP1, VP2, VP3, 2B, 2C, and 3D (table S3), along with five nucleotide substitutions in its IRES site (Table 2) compared with WT HEVA71. Two of these mutations (T164CIRES and G165CIRES) mapped to stem loop 2, another two were located in stem loop 4 (G368CIRES and T370GIRES), and the last one was confined to a connecting loop between stem loops 2 and 3 (C177TIRES). Functional analysis of these point mutations in the mutant HEVA71 IRES by plaque reduction assays (Fig. 6D) and bicistronic luciferase assays (Fig. 6E) revealed that individual mutations such as C177TIRES, G368CIRES, and T370GIRES did not play active roles in prunin resistance. However, two mutations in stem loop 2 (T164CIRES or G165CIRES) resulted in similar resistance to prunin-resistant mutant HEVA71, suggesting that stem loop 2 might play a greater role than stem loop 4 in prunin resistance. Close examination of the predicted IRES RNA secondary structures between HEVA71 mutant and WT viruses revealed major folding changes (Fig. 6F): Mutations T164CIRES and G165CIRES shortened stem loop 2 with a removal of an irregular five-sided polygon (GTATC), and mutations G368CIRES and T370GIRES eliminated an additional hairpin loop (GCGCTGGC) in stem loop 4 of the IRES. Consistent with our results from viral plaque assays, mutation C177TIRES did not affect the loop attached between stem loop 2 and 3 of HEVA71 IRES, confirming its lack of effect in conferring resistance to prunin. Overall, these data demonstrate that the four mutations described in IRES stem loops 2 and 4 maintain IRES functionality and play roles in resistance to prunin.

Table 2 Nucleotide substitutions in the IRES region of prunin-resistant HEVA71 mutants.

Four substitutions (highlighted) were predicted to affect IRES RNA secondary structure folding.

View this table:

Prunin displays inhibitory activity against hepatitis C virus via a different type of IRES element

We wished to investigate whether the efficacy of prunin against HEVA71 in vitro and in vivo was exclusive to a particular group of IRES RNAs or could be observed against diverse viral IRES structures. The scientific literature has categorized viral IRESs into four major groups according to their requirements for various host factors, hypothesized secondary structures, the position of the start codon relative to the IRES, and the ability of the IRES to function in rabbit reticulocyte extracts with or without supplementations (45). These findings have placed HEVA71 IRES as a member of the group 3 IRES RNAs, which require canonical eukaryotic initiation factors (eIFs) and other IRES trans-acting factors for their translational functions. To test whether prunin exhibited inhibitory activities against other IRES structures, we tested prunin against hepatitis C virus (HCV), which has a group 2 IRES that interacts with the 40S ribosomal subunit and a subset of canonical eIFs (45).

We treated HCV-infected Huh-7.5 cells with doses of prunin ranging from 62.5 to 500 nM and evaluated the cytotoxicity and effectiveness of prunin against these cells. At 3 and 6 dpi with various prunin dosages, we only observed significant declines in HCV viral titers at a dose of 500 nM (P = 0.011 to 0.022; Fig. 7A), as quantified by immunofluorescence assays. Cytotoxicity assays revealed that the selected concentrations resulted in at least 80% cell viability (Fig. 7A). Furthermore, 500 nM prunin significantly reduced HCV viral titers (P = 0.011 to 0.022) by about 0.5 and 0.7 log10 PFU/ml in comparison with 1% DMSO (vehicle control) at 3 and 6 dpi, respectively, although other concentrations of prunin (62.5 to 250 nM) did not inflict any significant changes in HCV viral titers.

Fig. 7 Suppression activity of prunin against HCV.

(A) Inhibition (black and gray bars) and cell viability (red line) profiles of prunin (62.5 to 500 nM) on HCV at 3 dpi (black bars) and 6 dpi (gray bars) through HCV viral titer quantification via immunofluorescence assays and cell viability assays, respectively. Concentrations of prunin below 500 nM showed more than 80% cell viability (blue dashed line). (B) Disruption of HCV viral RNA synthesis by prunin (62.5 to 500 nM) at 3 dpi (black bars) and 6 dpi (gray bars) measured through qRT-PCR. Only HCV RNA quantifications at 6 dpi were significant relative to the vehicle control. DMSO (1%) was used as vehicle control. Data are means ± SD from three independent experiments consisting of triplicates. Statistical analyses were performed against 1% DMSO. One-way ANOVA corrected using Dunnett’s correction. ****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05.

To confirm the efficacy of prunin against HCV, we studied the effects of prunin on HCV RNA generation. Treatment of infected Huh-7.5 cells with similar doses of prunin (62.5 to 500 nM) resulted in significant (P = 0.005 to 0.00008), dose-dependent fold reductions in HCV RNA expression relative to 1% DMSO, although only at 6 dpi (Fig. 7B). Prunin (500 nM) significantly reduced HCV RNA by 0.7-fold at 6 dpi (P = 0.00008), consistent with the HCV viral titer quantifications. These data show that prunin can suppress HCV infection, possibly by using an anti-IRES strategy against HCV similar to that against HEVA71, because HEVA71 and HCV contain IRES elements. This also suggests that prunin may be a potential broad-spectrum IRES inhibitor, at least for groups 2 and 3 IRES elements.

Prunin-resistant HEVA71 differentially regulates IRES-mediated activity via Sam68 and hnRNPA1

Given that our earlier results demonstrated that prunin induced major folding changes in HEVA71 mutant IRES secondary structures, we investigated the differences in recruitment of IRES trans-acting factors between the mutant and WT HEVA71 IRES. Specifically, as mutant HEVA71 IRES stem loops 2 and 4 showed structural modifications that played roles in prunin resistance, we questioned which of the reported IRES trans-acting factors bind to those regions and also interact with each other.

We found three proteins that interacted with the IRES: heterogeneous nuclear ribonucleoprotein A1 (hnRNPA1), heterogeneous nuclear ribonucleoprotein K (hnRNPK), and Src associated in mitosis 68-kDa protein (Sam68), all previously documented as RNA binding proteins and known to act as nuclear-cytoplasmic shuttlers for various molecular and cellular functions (46). These three IRES trans-acting factors have been classified as positive regulators of HEVA71 IRES–facilitated translation (4749) and were found to interact with different regions of the HEVA71 5′UTR (4749).

To affirm whether prunin decreased or prevented these proteins from binding to the HEVA71 IRES, we performed knockdown studies where RD cells were treated with a range of small interfering RNA (siRNA) concentrations targeting each IRES-trans acting factors (ITAF), namely, Sam68, hnRNPK, and hnRNPA1, along with nontargeting siRNA control for 72 hours. We then infected siRNA-treated RD cells with either WT HEVA71 or mutant HEVA71 at an MOI of 1, in which siRNA-treated RD cells were lysed at 12 hpi for Western blot and SDS–polyacrylamide gel electrophoresis analysis. We observed dose-dependent reductions in protein expression for Sam68 (Fig. 8A), hnRNPK (Fig. 8B), and hnRNPA1 (Fig. 8C) after siRNA treatment in both uninfected and HEVA71-infected cells. Low Sam68, hnRNPA1, and hnRNPK protein expression was observed with 25 nM Sam68 or hnRNPA1 or 35 nM hnRNPK siRNA but was not observed for the nontargeting control (Fig. 8D). Moreover, quantified band intensities of Sam68, hnRNPK, and hnRNPA1 showed significant siRNA dose-dependent declines in comparison to the mock control (0 nM; P = 0.00005 to 0.00009), suggesting that the siRNA dosages used in this study were effective in silencing Sam68, hnRNPK, and hnRNPA1.

Fig. 8 Role of ITAFs in mutant and WT HEVA71 replication and translation.

Verification of siRNA-mediated knockdown of (A) Sam68, (B) hnRNPK, and (C) hnRNPA1 at 12 hpi via Western blot analyses, where band intensities of each ITAF were tabulated across increasing siRNA doses for both mutant and WT HEVA71. Disruption of viral protein synthesis (VP0 and VP2) of mutant and WT HEVA71 by the above three ITAFs was also quantified via Western blots at 12 hpi. (D) Expression of viral protein (VP0 and VP2) and Sam68, hnRNPK, and hnRNPA1 at 12 hpi after mutant or WT HEVA71 infection in cells transfected with nontargeting siRNA (NTC), which served as knockdown controls. β-Actin was used as loading control. (E) Cell viability profiles (brown dots) of siRNA-treated cells using alamarBlue assays after 72 hours of incubation. Concentrations used resulted in cell viabilities above 80% (blue dashed line). Mutant and WT HEVA71 viral titers in the supernatant of siRNA- and NTC-treated cells analyzed via viral plaque assays. (F) Biotinylated (+) or nonbiotinylated (−) RNA pulldown of hnRNPK, Sam68, and hnRNPA1 proteins bound to WT IRES, mutant IRES, or negative controls including green fluorescent protein (GFP) and actin RNAs via Western blots. Flow-through (FT) without biotinylated RNA-protein complexes was also subjected to Western blot analyses. Band intensities of each IRES trans-acting factor were calculated in comparison to cell lysate control for both biotinylated mutant and WT IRES. Data are means ± SD from three independent experiments consisting of triplicates. Statistical analyses were performed against mock-treated cells. Two-way ANOVA with Bonferroni’s correction. ****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05.

In comparison to mock-treated cells, we observed dose-dependent reductions in both VP0 (a HEVA71 precursor protein) and VP2 (a HEVA71 structural protein) after knockdown of Sam68 (Fig. 8A) and hnRNPA1 (Fig. 8C) in RD cells, regardless of WT or mutant HEVA71 infection. Sam68 knockdown had a greater effect on mutant HEVA71 than WT HEVA71, and decreasing hnRNPA1 protein expression caused a down-regulation of HEVA71 viral protein synthesis during both WT and mutant HEVA71 infections. In particular, 5 nM Sam68 siRNA reduced VP0 and VP2 expression in WT HEVA71 by 0.37 and 0.50 U and in mutant HEVA71 by 0.82 and 0.88 U, respectively. However, 15 nM hnRNPA1 siRNA led to similar significant decreases in VP0 and VP2 protein of both viruses by 0.98 and 0.94 U, respectively (P = 0.00006 to 0.00008). Apart from Sam68 and hnRNPA1, hnRNPK knockdown also resulted in dose-dependent decreases in expression of both VP0 and VP2, which, however, were limited to WT HEVA71 (Fig. 8B). Although 35 nM hnRNPK siRNA significantly knocked down production of VP0 and VP2 during WT HEVA71 infection (P = 0.00006 to 0.00008), the relative band intensities of these viral components after hnRNPK knockdown were not significantly affected during mutant HEVA71 progeny virion synthesis (P = 0.55 to 0.78), which were akin to nontargeting control treatments (Fig. 8D).

To verify whether infectious viral particle production was affected by knockdown of Sam68, hnRNPA1, or hnRNPK, we quantified both mutant and WT HEVA71 viral titers in siRNA-treated RD cells using plaque reduction assays (Fig. 8E). Consistent with the knockdown Western blots, 55 nM Sam68 siRNA decreased mutant HEVA71 viral titers by about 5.8 log10 PFU/ml, significantly greater than the 3.5 log10 PFU/ml reduction in WT HEVA71 titers (P = 0.008). Moreover, knockdown of hnRNPA1 resulted in a reduction in both WT and mutant HEVA71 viral titers by about 4.0 log10 PFU/ml. By contrast, hnRNPK siRNA–mediated knockdown only caused declines in mutant HEVA71 titers of about 3.8 log10 PFU/ml (P = 0.0006), without exerting any effects on WT HEVA71 titers. These observed effects were likely not due to cellular toxicity because siRNA-treated cells displayed cell viabilities of above 80% (Fig. 8E). Overall, as knockdown of Sam68 and hnRNPK had varying impacts on mutant and WT HEVA71 production, we hypothesized that prunin might be preventing Sam68 or hnRNPK from binding to the WT HEVA71 IRES during IRES-facilitated translation of WT HEVA71.

To test the above, we performed biotin-RNA pulldown assays using either the WT or mutant HEVA71 IRES and probed for protein interactions with either hnRNPK, Sam68, or hnRNPA1 and their respective biotinylated RNAs. All three ITAFs interacted with both biotinylated IRESs, but not control RNAs [nonbiotinylated IRES, green fluorescent protein (GFP), and actin]. However, in comparison to WT IRES, hnRNPK associated with the mutant IRES to a lesser extent, which allowed Sam68 to compensate for the above lack of interaction by increasing its association with mutant IRES RNA (Fig. 8F). In particular, band intensities of hnRNPK significantly decreased by 0.67 U (65 kDa) and 0.71 U (51 kDa) (P = 0.00097), whereas the band intensity of Sam68 increased by 0.59 U for the pulldown of mutant HEVA71 IRES versus WT HEVA71 IRES. Overall, these results establish that prunin is able to substantially suppress the binding of hnRNPK to the WT HEVA71 IRES, providing selection pressure for the WT IRES to evolve into a mutant IRES form that can bind Sam68 with greater affinity, in turn facilitating the IRES translation of mutant HEVA71 in the absence of hnRNPK activity.


With recent increases in HEVA71-related morbidity and mortality observed in humans (50), there are pressing demands for the development of effective anti-HEVA71 therapies. We targeted the IRES of HEVA71 to reduce viral protein synthesis without affecting host cell translation. We identified prunin as an effective suppressor of HEVA71 in cell culture and murine models, suggesting its potential for development into HEVA71 therapeutics. Prunin is not the first flavonoid to exhibit inhibition properties against HEVA71 IRES. Research in past years has revealed potent flavonoids against HEVA71 IRES, such as apigenin (51) and kaempferol (52), which functioned at EC50 values in the micromolar ranges of 30.5 and 31.8 μM, respectively. Other than flavonoids, there are also other non–naturally occurring substances that act against HEVA71 IRES, for instance, amantadine, which acts at 2 mM (25). Compared with the concentrations of known compounds that have been tested against HEVA71 IRES, prunin has the lowest EC50, within the nanomolar range. This feature of prunin could allow for its administration at lower dosages to achieve full anti-IRES activity, potentially reducing the probability and severity of side effects.

Many compounds have been designed or discovered to target either HEVA71 structural or nonstructural proteins independently (7, 8). For instance, pleconaril was found to perturb the functions of VP1 capsid structural proteins by stabilizing them, thereby preventing the uncapping and discharge of HEVA71 genome into host cells (9, 53, 54). Other examples include fistein and rutin, which interfere with HEVA71 polyprotein processing and subsequent virion assembly (55). However, by using IRES as its target, prunin is able to simultaneously prevent cap-independent translation of both HEVA71 structural and nonstructural proteins. We estimated the decrease in expression of nonstructural proteins indirectly by measuring HEVA71 RNA because such proteins play critical roles in catalyzing HEVA71 RNA replication. For example, HEVA71 2C protein has been shown to aid in negative RNA strand synthesis (56) and the development of replication complexes (57); HEVA71 3D protein codes for an RNA polymerase that produces positive RNA strands (58); and HEVA71 3A protein has been reported to induce RNA polymerase activity (59). The inhibition of HEVA71 by prunin could potentially offer the opportunity to avoid combination therapy with HEVA71 structural and nonstructural suppressors, which could force the effective dosages of treatments to be reduced to avoid adverse and cytotoxic effects. This was observed in combinatorial therapies comprising well-validated compounds such as NITD008, ALD, and 1-acetyllycorine (suppressors of 3D polymerase, VP1, and 2Apro) or gemcitabine and ribavirin (suppressors of 3D polymerase) (6062).

In addition to prunin, our screen picked up five other hits (ST024699, ST024368, ST024702, ST066904, and ST024081) that reduced infectious HEVA71 viral titers, although ST002086 from the initial screen was later found not to exhibit any effects on HEVA71. Further understanding of the individual structures and classifications of these flavonoids aided us in determining the possibility of the above-mentioned difference (table S4). There are four main classifications of flavonoids (fig. S3), namely, isoflavones, neoflavonoids, chalcones, and another large group (F2) consisting of flavones, flavonols, flavanones, flavanonols, flavanols, and anthocyanins (63). ST002086 was placed in the chalcones group of flavonoids that consisted of open carbon ring structures, whereas the other six flavonoids belonged to the large F2 group, comprising B rings attached to position 2 of the carbon on another C ring. This tells us the importance of having closed carbon ring structures with carbon backbones similar to the F2 group to effectively inhibit HEVA71 IRES activity. These observations can therefore serve as criteria for lead molecules in future antiviral design or as effective tools in helping to understand HEVA71 IRES inhibition mechanisms better. Moreover, HEVA71 structural and nonstructural proteins not present in the initial screening assay could play various roles in sequestering open carbon ring structures like ST002086, thus rendering ST002086 ineffective against HEVA71.

Prunin could possibly suppress HEVA71 IRES activity by directly intercalating between the bases of the IRES RNA at certain regions or stem loops, thereby interfering with IRES secondary RNA structure and indirectly inhibiting one or more ITAFs from interacting with the IRES (64). Combining the results from the limited spectral activity of prunin and four resistance mutations (T164CIRES, G165CIRES, G368CIRES, and T370GIRES) seen in HEVA71 IRES against prunin, we confirmed whether this notion was correct, through comparing the IRES sequences of enteroviruses such as HEVA71, CA6, CA16, ECHO7, and CB5 that were affected by prunin, to identify whether those four nonmutated nucleotides (T164, G165, G368, and T370) were shared commonly among them. Although this was shown to be negative (3), it should be noted that those mutated nucleotides were analyzed from only a single mutated genome (one agarose plug) instead of many genomes. More full-genome sequencing of two or more purified plugs could aid in identifying more mutations or affirming the current mutations present in prunin-resistant HEVA71. Furthermore, recent reports have suggested that IRES elements contain evolutionary conserved motifs, which have the tendency to only preserve sequences affecting their structure and subsequent protein interactions for their functions, despite the lack of conservations in their primary sequence and secondary RNA structure (6567). This supports the notion of prunin directly inhibiting conserved motifs of IRES that could vary in sequence and structure. Moreover, we observed notable declines in HCV RNA amounts and viral titers at 500 nM prunin, thereby suggesting that the mechanism of prunin could be affecting common conserved regions shared between distinct HEVA71 and HCV IRESs.

To understand whether the effect of prunin on IRES structures, in turn, differentially regulates ITAF recruitments, we investigated on three ITAFs (Sam68, hnRNPA1, and hnRNPK) that have been reported in current literature, which positively regulate HEVA71 IRES–mediated translation (4749) along with HCV replication (46, 68, 69). Using biotin RNA pulldown assays, we found that hnRNPA1 bound to stem loop 2 of IRES (47), whereas Sam68 was recently discovered to specifically interact with stem loop 4 of IRES (48). The KH2 domain and proline-rich domain with a neighboring KH domain of hnRNPK were found to maintain contacts with both stem loops 2 and 4 of HEVA71 IRES (49). Further functional interaction studies involving these three ITAFs conducted over the past decades have demonstrated Sam68 as a multifunctional Src homology 3 (SH3) and SH2 adaptor protein (70) that can link with hnRNPs via its five proline-rich motifs (71, 72), which, in turn, can form an ITAF complex that drives IRES-mediated cap-independent translation. hnRNPA1, hnRNPK, and Sam68 were also established as proviral factors for HCV replication (46, 68, 69), either by their direct interactions with viral core proteins (69) or via indirect associations with other cellular scaffold proteins (46, 68). Altogether, this information suggested that the activities of prunin toward WT HEVA71 and HCV IRES might decrease or prevent one or more ITAFs, particularly Sam68, hnRNPK, and/or hnRNPA1, from binding to those respective ITAFs.

Our knockdown studies and biotin-RNA affinity assays revealed distinct recruitment patterns of ITAFs by mutant HEVA71, where associations with hnRNPK were forfeited so as to gain higher reliance on Sam68 for maintaining its IRES-facilitated translation. It was also shown that prunin suppresses hnRNPK interaction with the HEVA71 IRES, compromising the essential functions and interactions of hnRNPK during IRES-mediated translation (73) and contributing to selection pressure for the mutation of the WT HEVA71 IRES. We found that Sam68 compensated for the loss of hnRNPK activity in mutant HEVA71, revealing that the conformational change by prunin in mutated stem loop 4 allowed for a greater binding affinity to Sam68. Given that Sam68 interacts with hnRNPA1 during WT HEVA71 infection (74), similar interactions can be maintained without hnRNPK amid mutant virus production, so as to form the ITAF complex necessary to drive mutant IRES-dependent translation. However, the current literature has not identified many ITAFs that interact with various IRES elements, suggesting a need to uncover more ITAFs for a better understanding of IRES-mediated translation and treatment-associated suppression mechanisms.

In line with its in vitro properties, prunin also established high in vivo efficacies in mice models of HEVA71 infection, which could be further characterized for preclinical development. Currently, only a small population of compounds has been effective against HEVA71 in murine models, namely, pleconaril (75) and lactoferrin (76) against HEVA71 VP1 protein, rupintrivir (77) and chrysin (78) targeting HEVA71 3Cpro, lycorine and 1-acetyllycorine (79) inhibiting HEVA71 2Apro, and, last, ribavirin (44) and NITD008 (80) suppressing HEVA71 3D polymerase. On the basis of the above list, it should be noted that no antiviral against HEVA71 IRES has been effective in murine models, placing prunin as a potent anti-HEVA71 IRES compound in vivo. However, a critical caveat to futuristic prunin administration for HEVA71 therapy would be the development of resistance as shown in our study. Given that prunin resistance was surmounted with ribavirin, combinatorial therapies of prunin with other known anti-HEVA71 compounds that target distinct points in the HEVA71 life cycle could be ultimately used. Nevertheless, accurate and careful preclinical profiling should be carried out to define the feasibility of prunin for clinical development against HEVA71 infections. Hopefully, the lag time for the occurrence of prunin resistance will be used to develop a potential vaccine or an improved drug design from lead molecules like prunin.

Other than its potential clinical applications, prunin together with the cell-based bicistronic vector systems can be used as a gene translational regulation system in vitro via the inhibitory effects of prunin on IRES-mediated translation. Traditional inducible gene reporter systems consist of three components: a promoter that can be easily activated akin to a lac promoter (81) or a TRE-CMVmin promoter (82); a protein activator and/or repressor of transcription such as the tTa transactivator, which aids in TRE-CMVmin activation or the lacI repressor that inhibits the activity of the lac promoter; and small-molecule regulators including IPTG (isopropyl-β-d-thiogalactopyranoside) or tetracycline, which mediate interactions among protein transcription activators or repressors and their associated promoters. The problem with this system is its stringent requirement for unique promoters along with associated transcription factors and regulators. This can be resolved by exploiting the simplified IRES repressor assay, which takes into account the exclusive properties of prunin, the HEVA71 bicistronic reporter, and the HEVA71 IRES bicistronic hairpin, where no definitive promoters or protein factors are needed. Moreover, this bicistronic gene regulation system can concurrently express reporter and antibiotic selection genes via cap-dependent translation, which can be used for reaffirming transfection efficiencies and selecting heterogeneous cell lines, respectively. The IRES repressor system can also function as a high-throughput screening platform against numerous compound libraries, where more HEVA71 antivirals like prunin can be identified to meet the demands for potent HEVA71 therapies.

Our work demonstrated that prunin serves to inhibit HEVA71 production in vitro and in vivo through suppressing HEVA71 activity and its subsequent interactions with ITAFs such as Sam68 and hnRNPK. However, our work on ITAFs was based on limited current scientific literature. Hence, there could be other ITAFs also affected by prunin that were not investigated in this study. Moreover, we only demonstrated that prunin directly affected structures using structural predictions; hence, the true conformation of the mutated RNA is still unknown. Once known, more 3D RNA docking studies should be carried out to affirm the mechanism of action of prunin. Another vital point is that HEVA71 elicits resistance to prunin after a few rounds of exposure, which can be overcome by ribavirin. Hence, the combination of ribavirin and prunin should be investigated as a potential approach against HEVA71, which this study did not investigate.


Study design

We aimed to discover flavonoid-based antiviral treatments against HEVA71 IRES. We generated bicistronic constructs consisting of the HEVA71 IRES or the HEVA71 IRES hairpin and introduced these constructs into RD cells. The HEVA71 IRES bicistronic construct allows for simultaneous cap-dependent and cap-independent translations, whereas the HEVA71 IRES hairpin construct attenuates cap-independent translation, mimicking HEVA71 IRES inhibition. We then screened the flavonoid derivatives with the bicistronic assay system to identify antivirals with activity against HEVA71. The antiviral prunin was subjected to downstream validation studies such as protein and RNA studies in vitro and tested for its efficacy in murine models. We investigated the mechanistic action of prunin using a treatment-resistant HEVA71 strain obtained from extensive passaging of HEVA71 with prunin. Biotin-RNA pulldown assays were carried out to affirm the mechanism of prunin against ITAFs and IRES, where mutant HEVA71 was compared with the WT HEVA71 strain. All data were collected from three independent experiments consisting of triplicates.

Statistical analysis

Robustness of the compound screening assay was determined by Z factors (45), which measure the intervals among the SDs of the signals against the background noise of an assay. Z factors of 60 samples were calculated according to the expression 1 − [(3 × SD of positive control + 3 × SD of negative control)/|(mean positive control − mean negative control)|]. For studies involving heterogeneous selection, we conducted two-way analyses of variance (ANOVAs) corrected with Tukey’s posttests and performed one-way ANOVA tests adjusted with Dunnett’s posttests for all other analyses. P ≤ 0.05 was considered significant.


Materials and Methods

Fig. S1. AKOS02428449 activity against HEVA71.

Fig. S2. Cytotoxicity profiling of bone marrow muscle (SJ) cells with prunin.

Fig. S3. Classification of flavonoids.

Table S1. Hit list of flavonoids generated from luciferase-screening assay, which are ranked according to their percentage inhibitions of IRES activity.

Table S2. Clinical scoring system for the EV71 BALB/c mouse model.

Table S3. Silent mutations found in the genomes of HEVA71 prunin-resistant mutants.

Table S4. Structures and classification of the seven chosen flavonoids.

Table S5. The primers (HEV1–16) used for DNA sequencing of the modified HEVA71 IRES bicistronic reporter construct.


Funding: This work was supported by the National Medical Research Council (NMRC)–IRG grant (CBRG13nov02 to J.J.H.C.). Author contributions: J.J.H.C. conceptualized the study. S.G., K.Z.W., and N.M. designed the research and experiments. S.G. conducted all experiments except histopathology studies, CHIKV infection studies, and biotin-RNA pulldown assays, which were performed by J.S., R.C.H.L., and K.Z.W., respectively. HCV studies were performed by N.K.B.M.I. and Y.J.T. S.G. analyzed all data and statistics. J.J.H.C. supervised the study. S.G. and J.J.H.C. wrote the manuscript. Funding (CBRG13nov02 to J.J.H.C.) for the study was acquired by J.J.H.C. Competing interests: S.G. and J.J.H.C. hold a patent entitled “Inhibitors of IRES portion of an enterovirus” (10201806280R; A*STAR reference: IMC/P/10686/01/PCT). Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials.

Stay Connected to Science Translational Medicine

Navigate This Article