Identification of small molecule inhibitors targeting the Zika virus envelope protein

Abstract

The recent emergence of Zika virus, a mosquito-borne flavivirus, in the Americas has shed light on the severe neurological diseases associated with infection, notably congenital microcephaly in newborns and Guillain-Barré syndrome in adults. Despite the recent focus on Zika virus, there are currently no approved vaccines or antiviral therapies available to treat or prevent infection. In this study we established a competitive amplified luminescent proximity homogeneous assay (ALPHAscreen) to identify small molecule inhibitors targeting the envelope protein of Zika virus (Zika E). We utilized this assay to screen two libraries of nearly 27,000 compounds and identified seven novel inhibitors of Zika E. Characterization of these primary screening leads demonstrated that inhibition of Zika virus occurs at non-cytotoxic concentrations for all seven lead compounds. In addition, we found that all seven lead compounds have potent activity against the closely related dengue virus 2 but not vesicular stomatitis virus, an unrelated enveloped virus. Biochemical experiments indicate that these compounds act by preventing E-mediated membrane fusion. This work highlights a new method for the discovery and optimization of direct-acting antivirals targeting the E protein of Zika and other flaviviruses.

Zika virus (ZIKV) is a mosquito-borne virus belonging to the Flavivirus genus, which includes the closely related dengue, Japanese encephalitis, and yellow fever viruses (DENV, JEV, and YFV, respectively). The outbreak of ZIKV in 2015–2016 is estimated to have caused over 500,000 cases throughout Central and South America (PAHO/WHO Zika report). Symptomatic ZIKV infections commonly result in fever with associated rash, conjunctivitis, and joint pain but have also been associated with severe neuropathies including Guillain-Barre syndrome in adults and microcephaly in children born to infected mothers. ZIKV has been observed to establish long-term infections in the testis (Govero et al., 2016; Ma et al., 2016), and continued detection of ZIKV in body fluids (e.g., semen, saliva, breast milk) for weeks to months after the onset of symptoms (Calvet et al., 2018; Paz-Bailey et al., 2017) indicates a need for interventions to reduce viral burden due to replication in these sites and to thereby reduce transmission via alternative mechanisms (e.g., breastfeeding or sexual).

There are effective flaviviral vaccines against YFV, JEV, and tickborne encephalitis virus (TEV), however Dengvaxia, currently the only marketed DENV vaccine, not only fails to elicit protective immunity in those lacking prior DENV exposure but also increases the risk of severe disease in these individuals, likely due to antibody-dependent enhancement of infection (Capeding et al., 2014; Hadinegoro et al., 2015; Sabchareon et al., 2012; Halstead and Russell, 2016; Halstead, 2016; Villar et al., 2015). Although early results towards a ZIKV vaccine are promising (Gaudinski et al., 2018; Modjarrad et al., 2018), the cross-reactivity of ZIKV and DENV antibodies has been shown to enhance infection and disease in cell culture and murine models (Fowler et al., 2018; Sapparapu et al., 2016; Stettler et al., 2016; Zimmerman et al., 2018). Due to the potential challenges this phenomenon may present to vaccine safety and efficacy, antivirals may provide a complementary strategy to limit ZIKV viral burden, viral persistence, and disease severity without the risk of cross-reactive antibodies. The direct-acting antivirals that now cure chronic hepatitis C virus (HCV, Family: Flaviviridae; genus: Hepacivirus) provide precedent for the use of direct-acting antivirals against members of this viral family. Despite this, multiple efforts to develop antivirals against the polymerase and protease proteins of dengue and other flaviviruses have been unsuccessful in advancing candidates to the clinic (Boldescu et al., 2017; Lim et al., 2013). Consequently, complementary efforts against alternative antiviral targets remain of high interest and value.

The flavivirus envelope protein, E, mediates the initial attachment of the virion to the host cell membrane through interaction with cell surface attachment factors (e.g., DC-SIGN, Tim/Tam family phosphatidylserine receptors) (Hamel et al., 2015; Tassaneetrithep et al., 2003). Following internalization of the virion via a clathrin-dependent process, endosomal acidification triggers significant structural changes in E that drive fusion of the viral and endosomal membranes. This creates a fusion pore that releases the viral RNA genome into the cytoplasm where viral gene expression and replication can ensue. We previously identified several classes of small molecule antivirals that bind a pocket between domains I and II of the prefusion, dimeric form of DENV E present on virions and inhibit E-mediated membrane fusion (Clark et al., 2016; Schmidt et al., 2012; de Wispelaere et al., 2018). Due to partial conservation of this site, a subset of DENV E inhibitors also inhibit other flaviviviruses including ZIKV, while others appear to be selective for DENV (Lian et al., 2018; de Wispelaere et al., 2018). We recently established a competitive amplified luminescent proximity homogeneous assay (ALPHAscreen) to enable target-based discovery and optimization of inhibitors of DENV2 E and showed this to be a reliable tool for identifying new scaffolds that inhibit DENV infectivity through the same biochemical mechanism of action (i.e., inhibition of E-mediated fusion) (Lian et al., 2018). Here, we report the development of an analogous competitive ALPHAscreen for Zika E and a pilot high-throughput screening effort leading to discovery and characterization of seven new chemical series that inhibit ZIKV by directly targeting Zika E.

To establish a target-based assay using recombinant, soluble Zika E (ZIKV sE), we first screened biotinylated derivatives of our previously identified pyrimidine inhibitors of DENV E (Clark et al., 2016; de Wispelaere et al., 2018) (Supporting Fig. 1) to identify an appropriate probe for Zika E competitive ALPHAscreen development. Biotinylated inhibitors were immobilized on streptavidin-coated donor beads and incubated with Ni-NTA acceptor beads to which soluble, recombinant Zika E had been bound (Fig. 1A). We found that the probe based on 4,6-disubstituted pyrimidine GNF-2 (“GNF-2-biotin”) consistently produced the highest signal-to-background ratio (Fig. 1B and Supporting Fig. 1). Cross-titration of protein and probe concentrations to maximize signal-to-background and competition in the presence of validated Zika E inhibitor 3–110-22 (de Wispelaere et al., 2018) established optimized conditions of 50 nM ZIKV sE and 100 nM GNF-2-biotin (Fig. 1C), yielding Z’ value of 0.63 with coefficient of variation of 4% for DMSO and 7.5% for 3–110-22, which consistently reduced AlphaScreen signal by ~75%.

Figure 1. Optimization of Competitive ALPHAscreen for Zika virus sE.

(A) Schematic of the competitive ALPHAscreen utilized in the high-throughput screening. Compounds that interfere with the GNF-2-biotin/ZIKV sE interaction cause reduced signal and are identified as “hits.” (B) Candidate biotinylated probes were titrated from 0–100 nM in the presence of 50 nM ZIKV-sE. ALPHAscreen signal is plotted as a function of probe concentration with representative data for n ≥ 2 independent experiments shown. Error bars represent standard deviation of 3 technical replicates within representative experiment. GNF-2-biotin provided the highest signal-to-noise across the titration. (C) Known DENV and ZIKV E inhibitor 3–110-22 served as a positive control and successfully reduces the ALPHAscreen signal across the ZIKV-sE titration (from 0–200 nM) in the presence of 100 nM GNF-2-biotin. The optimal signal reduction is observed at 50 nM ZIKV-sE and 100 nM GNF-2-biotin, and these conditions were utilized for high-throughput screening. Representative data for n ≥ 2 independent experiments shown; error bars represent the standard deviation of 3 technical replicates within the experiment shown.

With these optimized ALPHAscreen conditions, we conducted a high-throughput screen against the commercially available Asinex 2 (macrocycles) and Asinex 3 (nucleoside analogues) libraries (https://iccb.med.harvard.edu/commercial-libraries), totaling 26,954 compounds screened at a final concentration of 30 µM (compound:target ~ 600:1). Plate statistics for the Zika E screen had coefficients of variation ranging from 1.341–11.257, with 90% of the plates between 2.0–4.5. Z’ values ranged from 0.51–0.90, with most plates exhibiting a Z’ > 0.7 and signal-to-background values of ~4.

We used normalized percent inhibition (NPI) compared to the 3–110-22 positive control to identify 263 active compounds (0.97% of the total library) (Fig. 2). For the nucleoside analogue library, active compounds had an NPI > 60% (“hit” rate of 0.25%) while the macrocycle library cut-off was set at NPI > 80% (“hit” rate of 1.3%). Of these compounds, several had similar chemical structures. This suggested specific activity attributed to shared chemical structure and enabled us to narrow down our follow-up efforts to 106 compounds through selection of only a subset of compounds from each chemical series.

Figure 2. Screening progression and validation.

High-throughput screening and validation of screening “hits” followed the progression outlined in the flow chart. From the nearly 27,000 compounds screened using the competitive ALPHAscreen, 106 compounds were identified as putative inhibitors of Zika E and advanced to antiviral testing. Following successive tests for antiviral activity in the plaque reduction by neutralization test (PRNT₅₀) and single-cycle infectivity experiments, seven lead compounds were advanced for further characterization.

We first sought to validate that the compounds had concentration-dependent activity against recombinant ZIKV sE. IC₅₀ values, defined as the compound concentration at which the ALPHAscreen signal was reduced by 50%, were determined by non-linear regression analysis of dose-response experiments. 102 of the 106 initial hit compounds exhibit IC₅₀ < 30 µM, for a validation percentage of 96% (Supporting Table 1). We next performed plaque reduction neutralization tests (PRNT₅₀) to determine whether activity in the biochemical assay is associated with antiviral activity. To ensure that we identified compounds that inhibit ZIKV by blocking E’s function in viral entry versus compounds that inhibit ZIKV via other mechanisms or targets, the compound was pre-incubated with the viral inoculum (~200 PFU, MOI of 0.005) and present during an initial one-hour infection but was otherwise absent during the assay (Supporting Fig. 2). Notably, 96 of the 106 compounds exhibit significant antiviral activity with PRNT₅₀ values less than 30 µM (compound:target ~ 10⁸:1) (Supporting Table 1), confirming that the target-based assay identifies compounds with anti-ZIKV activity.

To prioritize compounds for follow-up studies, we also assessed antiviral potency by quantifying inhibition of ZIKV at 30 µM (compound:target ~ 10⁴:1) in a single-cycle infectivity assay in which we made the assay conditions more stringent by increasing the viral inoculum to achieve MOI of 1 and also increasing the concentration of serum present during the inhibitor treatment from ~0.005% (PRNT₅₀ ) to ~0.1% (infectivity assay) (Fig. 3A). Only seven compounds were observed to inhibit ZIKV infectivity by more than 90% under these conditions using ~200-fold increased challenge virus (Supporting Table 1). We note that the majority of the other ninety active compounds identified in primary screening exhibit potent and reproducible antiviral activity under less challenging conditions (e.g. PRNT₅₀) in which MOI and target abundance are lower. While we view many of these compounds as candidates for medicinal chemistry optimization, here we initially focused our efforts on the seven compounds with potent antiviral activity in the more stringent infectivity assay conducted at high target abundance. These seven compounds include three structurally related linear compounds and four distinct macrocycles (Fig. 3B–C).

Figure 3. Antiviral activity of lead compounds against Zika virus.

(A) Schematic representation of the single-cycle infectivity assay utilized to determine the antiviral IC₉₀ of the 7 lead compounds against ZIKV. (B) The production and release of infectious virus at 24 hours post infection was determined by plaque formation assay (PFA) and plotted versus log₁₀ inhibitor concentration. The anti-ZIKV IC₉₀ value of each of the lead compounds was determined by non-linear regression analysis. The graph shown is representative of n ≥ 2 independent experiments; error bars represent the standard deviation of the technical replicates within the experiment. The reported IC₉₀ values are the average of n ≥ 2 independent experiments performed for each compound. (C) Structures of the seven most potent inhibitors of ZIKV (and the 3–110-22 positive control). These compounds were selected for further characterization.

We first determined the IC₉₀, defined as the compound concentration at which the infectious virus was decreased 1-log₁₀ in the single-cycle infectivity assay (Fig. 3A). IC₉₀ values ranged from 5.2 µM for LAS 52154459 to 20.3 µM for LAS 52154463 (Fig. 3B and Table 1). Notably, these compounds have antiviral potencies that are equal or greater in potency than previously identified Zika E inhibitors (de Wispelaere et al., 2018). We counter-screened against cytotoxicity using a commercially available assay to determine the concentration-dependent effects of these compounds on cell viability. Although two compounds, LAS 52509955 and LAS 51635112, cause significant loss of cell viability at concentrations below 50 µM (Table 1), all compounds except LAS 52509955 have selectivity indices (IC₉₀/CC₅₀) greater than 5. This, taken with the fact that the conditions used for the infectivity and PRNT₅₀ assays have a limited window of inhibitor treatment of 1-hour during the initial infection, indicate that general cytotoxicity is unlikely to contribute to the observed antiviral activity.

Table 1: Summary of lead compound characterization.

Luminescence data in the Zika sE ALPHAscreen were plotted as a function of compound concentration. IC₅₀ values, defined as the concentration at which the signal is reduced to 50% of the DMSO control, were determined by non-linear regression analysis of the data. ZIKV PRNT₅₀ values are determined by non-linear regression analysis of data from the plaque reduction by neutralization test (PRNT₅₀) (Supp. Fig 1). Antiviral IC₉₀ values were determined by non-linear regression analysis of data generated using the viral infectivity assay as shown in Fig. 3A–B. CC₅₀ values correspond to the concentration of inhibitor at which a 50% reduction in viability of Vero cells is observed. A selectivity index corresponding to the ratio of CC₅₀ and antiviral IC₉₀ values is reported. DENV2 IC₉₀ and VSV PRNT₅₀ values were determined in experiments analogous to the ZIKV single-cycle infectivity and PRNT₅₀ assays, respectively. All values are the average and standard deviation of n ≥ 2 independent experiments.

Molecule Name	ALPHA IC50, µM	ZIKV PRNT50, µM	ZIKV IC90, µM	CC50, µM	Selectivity Index CC50/IC90	Avg. DV2 IC90, µM	VSV PRNT50, µM
LAS 52154459	4.0 ± 4.5	<0.9375	5.1 ± 0.3	98.11 ± 2.7	19	3.2 ± 1.8	5.1 ± 2.5
LAS 52509955	0.9 ± 0.06	1.4 ± 0.5	10.6 ± 0.2	15.9 ± 5.2	1.5	4.1 ± 1.1	3.9 ± 0.07
MAC 51421886	3.3 ± 2.7	1.7 ± 0.3	10.6 ± 0.6	>100	>10	3.8 ± 0.9	6.5 ± 0.7
LAS 52161573	4.2 ± 1.9	1.1 ± 0.2	11.4 ± 2.8	>100	>7.5	7.2 ± 2.5	2.9 ± 0.4
LAS 52154463	19.3 ± 15.3	7.4 ± 0.1	20.3 ± 5.9	>100	>5	7.9 ± 3.2	26.7 ± 4.7
LAS 51635112	3.8 ± 2.0	1.1 ± 0.3	7.8 ± 3.1	30 ± 2.5	5	2.5 ± 0.1	26.3 ± 5.2
LAS 52154474	18.4 ± 12.5	6.7 ± 0.1	11.0 ± 0.9	>100	>9	4.8 ± 2.8	17.9 ± 3.9

To verify that inhibition of viral entry is the mode of action of these compounds, we performed a series of time-of-addition experiments. Using final concentrations corresponding to the previously determined IC₉₀ values, inhibitors were 1) pre-incubated with viral inoculum and present during infection (repeating the infectivity assay), 2) only present during the one-hour infection with cells, or 3) added at either one or four hours post infection (Fig. 4A). We quantified viral yield for all samples at 24-hours post-infection by viral plaque formation assay. As expected, compounds reduced viral yield ~1-log₁₀ after pre-incubation with the inhibitor, while all compounds except LAS 51635112 had minimal activity when added concurrent with infection, presumably due to limited time for target engagement. We observed additional antiviral activity for four of the compounds when they were delivered post-infection (Fig. 4A), indicating that these compounds may additionally inhibit Zika E’s function during virion morphogenesis and/or egress or may have antiviral activity mediated by other targets. In contrast, the 3 linear compounds (LAS 52154459, LAS 52154463, and LAS 52154474) exhibit only very modest inhibition when added post-infection, consistent with inhibition of E during viral entry as the primary antiviral mode of action of these compounds.

Figure 4. Validation of ZIKV sE as the target of compound mediated anti-viral activity.

(A) (left) Schematic of the time-of-addition study with boxes indicating when inhibitor (at the IC₉₀ concentration) was incubated with virus for each of the conditions: pre-co infection, co (concurrent) alone, 1-hour post-infection, or 4-hours post-infection. Inhibitor concentrations corresponded to the antiviral IC₉₀ measured in the infectivity assay as depicted in Fig. 3A and reported in Table 1. (right) Single-cycle viral yield was determined for all samples in the time-of-addition study by viral plaque formation assay of the supernatants at 24 hours post-infection. Data are normalized to the DMSO-control. All compounds exhibit approximately 1-log₁₀ unit decrease when present prior to and during the infection (“pre-co” conditions) as expected. Except for LAS 51635112, this antiviral activity is lost when inhibitor treatment is further limited to the one-hour infection period (“co” condition). Additional antiviral activity is observed for many of the compounds, which may reflect inhibition of E during virion morphogenesis or antiviral activity mediated by a target other than Zika sE. The graph shown is an average of 2 biological replicates, each performed with 2 technical duplicates. Error bars represent standard deviation of the biological replicates. (B) (left) Schematic of capsid protection assay. Acidification of mixtures of DENV with trypsin-loaded liposomes results in fusion of the two membranes, which enables trypsin-mediated digestion of the DENV2 capsid protein (C) while E on the exterior of the viral lipid bilayer remains intact. Inhibition of DENV E-mediated fusion results in protection of the C protein. (right) Western blot analysis of the DENV2 E and capsid (C) proteins is shown for fusion reactions performed as described in the schematic in the absence or presence of 10 µM or 50 µM the inhibitor LAS 52154459. Reactions performed at pH 5.5 in the absence of inhibitor show digestion of C, indicating viral fusion with the trypsin-loaded liposome. LAS 52154459 exhibits concentration-dependent protection of DENV2 C indicating inhibition of fusion. Blots shown are representative of 2 independent experiments. Note that the ratio of target:inhibitor in this assay represents > 10-fold increase relative to that in the viral infectivity assay (Fig. 3) and > 200-fold increase in target abundance relative to the PRNT₅₀ assay.

To assess the selectivity of these ZIKV inhibitors, we examined them for activity against DENV2, a related flavivirus, as well as against vesicular stomatitis virus (VSV), an unrelated, enveloped virus. Using the more stringent infectivity assay (MOI 1) to quantify anti-DENV2 activity, we observed IC₉₀ values < 5 µM. This is surprisingly on average 3- to 4-fold more potent inhibition of DENV2 over ZIKV (Table 1). We hypothesize that this may reflect increased “breathing” activity of the viral particle of DENV as compared to ZIKV (Dowd et al., 2015; Kostyuchenko et al., 2016; Sirohi et al., 2016), which likely affects accessibility of the pocket between domains I and II of E in which these compounds likely bind. A second potential reason for the reduced activity against ZIKV is suggested by the recent report that ZIKV entry can be mediated by alternative pathways independent of receptor binding and acidification (Rawle et al., 2018). Although all seven compounds exhibit modest inhibition of VSV in PRNT₅₀ assays, the antiviral potency is significantly less than observed against ZIKV. Together, these data indicate some level of specificity for ZIKV, DENV2, and perhaps other related flaviviruses.

Due to the increased potency of the compounds against DENV, we sought to further establish the mechanism of action with the representative compound LAS 52154459 on DENV E-mediated membrane fusion using a biochemical assay in which acidic pH triggers E-mediated fusion of virions with trypsin-encapsulating liposomes, leading to proteolytic digestion of the viral capsid protein (Fig. 4B). Inhibition of E-mediated fusion results in protection of DENV2 core from digestion in this assay. Following incubation of DENV2 virions with LAS 52154459, we observed concentration-dependent (compound:target ~ 10³:1 at 50 µM) protection of viral capsids from proteolysis after acidification of the solution to pH 5.5 (Fig. 4B). This demonstrates that LAS 52154459 prevents virion-liposome membrane fusion and formation of a functional pore in this assay, consistent with the idea that the ALPHAscreen identifies direct-acting antivirals that inhibit membrane fusion mediated by flavivirus E.

In summary, we have established and validated the first target-based assay to identify direct-acting antivirals of Zika E. Our results suggest the feasibility of developing analogous assays for additional flavivirus E proteins. Such parallel assays could be very useful in identification and development of antivirals with broad-spectrum activity. In the pilot high-throughput screen conducted for this study, we identified seven new lead compounds and characterized their antiviral activity against ZIKV and further against DENV2. The most promising of these seven compounds are a set of three structurally related linear compounds, LAS 52154459, LAS 52154463, and LAS 52154474 (Fig. 3C), all of which exhibit minimal toxicity, and act predominantly on the entry process based upon our time-of-addition experiments. We note that several additional structurally related linear compounds were present in the screen, and most were also were identified as hit compounds in the ALPHAscreen. Follow up work on these additional compounds and medicinal chemistry efforts to define structure-activity-relationships (SAR) driving antiviral potency and selectivity against ZIKV, DENV, and potentially other flaviviruses will be an essential part in efforts to evaluate and optimize direct-acting antivirals against this target class.

Supplementary Material

Figure S1. Supporting Fig. 1: Structures of candidate ALPHAscreen probes.

Structures of candidate ALPHAscreen probes. These candidates were derived from prior assay development and screening efforts against DENV2 E (Lian, 2018).

Figure S2. Supporting Fig. 2: Plaque reduction by neutralization test (PRNT₅₀).

(top panel) Schematic of the PRNT₅₀ indicating the timing of pre-incubation, infection, washes, and fixation of the cells for staining and counting of plaques. (bottom panel) The number of plaque-forming units (PFUs) were normalized relative to the DMSO control and are plotted versus log₁₀ of the compound concentration. Data for each of the compounds were fit by non-linear regression curves to establish PRNT₅₀ values, defined as the concentration at which 50% of the PFU are inhibited. The graph is representative of n ≥ 2 independent experiments performed for each compound. Error bars represent the standard deviation for 2 technical replicates within the experiment shown.

Table S1. Supplementary Table 1: Summary of high-throughput screening.

The initial high-throughput screen was performed at final compound concentration of 30 µM. Average NPI (Normalized Percent Inhibition) is the percent of signal inhibition relative to the positive control, 3–110-22, that was observed in the initial screening efforts. The ALPHA IC₅₀ value is the inhibitor concentration at which signal is reduced to 50% of the DMSO control in the Zika E ALPHAscreen. ZIKV PRNT₅₀ values were determined by non-linear regression analysis of the plaque reduction by neutralization test (PRNT₅₀) as described in Supp. Fig 2. Single-cycle infectivity inhibition is displayed as the potency of inhibition of compounds at 30 µM. (−) < 30% inhibition, + = 30–50% inhibition, ++ = 50–90% inhibition, +++ > 90% inhibition compared to DMSO control. All values are the average of n ≥ 2 independent replicates.