A Mechanism-Based Targeted Screen To Identify Epstein-Barr Virus-Directed Antiviral Agents

Epstein-Barr virus, which is nearly ubiquitous in humans, is causal to infectious mononucleosis, chronic active EBV infection, and lymphoid and epithelial cancers. However, EBV-specific antiviral agents are not yet available. To aid in the identification of compounds that may be developed as antivirals, we pursued a mechanism-based approach. Since many of these diseases rely on EBV’s lytic phase, we developed a high-throughput assay that is able to measure a key step that is essential for successful completion of EBV’s lytic cascade. We used this assay to screen a library of small-molecule compounds and identified inhibitors that may be pursued for their anti-EBV and possibly even antiherpesviral potential, as this key mechanism appears to be common to several human herpesviruses. Given the prominent role of this mechanism in both herpesvirus biology and cancer, our screening assay may be used as a platform to identify both antiherpesviral and anticancer drugs.

ABSTRACT

Epstein-Barr virus (EBV) is one of nine human herpesviruses that persist latently to establish permanent residence in their hosts. Periodic activation into the lytic/replicative phase allows such viruses to propagate and spread, but can also cause disease in the host. This lytic phase is also essential for EBV to cause infectious mononucleosis and cancers, including B lymphocyte-derived Burkitt lymphoma and immunocompromise-associated lymphoproliferative diseases/lymphomas as well as epithelial cell-derived nasopharyngeal cell carcinoma. In the absence of anti-EBV agents, however, therapeutic options for EBV-related diseases are limited. In earlier work, we discovered that through the activities of the viral protein kinase conserved across herpesviruses and two cellular proteins, ATM and KAP1, a lytic cycle amplification loop is established, and disruption of this loop disables the EBV lytic cascade. We therefore devised a high-throughput screening assay, screened a small-molecule-compound library, and identified 17 candidates that impair the release of lytically replicated EBV. The identified compounds will (i) serve as lead compounds or may be modified to inhibit EBV and potentially other herpesviruses, and (ii) be developed into anticancer agents, as functions of KAP1 and ATM are tightly linked to cancer. Importantly, our screening strategy may also be used to screen additional compound libraries for antiherpesviral and anticancer drugs.

IMPORTANCE Epstein-Barr virus, which is nearly ubiquitous in humans, is causal to infectious mononucleosis, chronic active EBV infection, and lymphoid and epithelial cancers. However, EBV-specific antiviral agents are not yet available. To aid in the identification of compounds that may be developed as antivirals, we pursued a mechanism-based approach. Since many of these diseases rely on EBV’s lytic phase, we developed a high-throughput assay that is able to measure a key step that is essential for successful completion of EBV’s lytic cascade. We used this assay to screen a library of small-molecule compounds and identified inhibitors that may be pursued for their anti-EBV and possibly even antiherpesviral potential, as this key mechanism appears to be common to several human herpesviruses. Given the prominent role of this mechanism in both herpesvirus biology and cancer, our screening assay may be used as a platform to identify both antiherpesviral and anticancer drugs.

INTRODUCTION

Epstein-Barr virus (EBV), also known as human herpesvirus 4, infects more than 90% of the world. Upon de novo infection, EBV establishes lifelong quiescence/latency within host memory B lymphocytes wherein it expresses a few viral latency genes with minimal virion production. While most individuals never develop EBV-related diseases, latent EBV is linked to several cancers of lymphoid and epithelial origin. These include endemic Burkitt lymphoma (eBL), Hodgkin's lymphoma, immunocompromise-associated lymphoproliferative diseases (LPD) generally seen in transplant recipients, nasopharyngeal cell carcinoma (NPC), and gastric carcinoma; indeed, EBV is thought to be causal to eBL, LPD, and NPC (1, 2). Under conditions that are not well understood, EBV periodically enters the lytic cycle whereupon the majority of viral genes are expressed in a highly regulated order, resulting in replication of viral genomes followed by packaging to produce infectious virions. Such episodic activation of the EBV lytic cycle replenishes the viral reservoir in the infected individual and ensures viral transmission in the population. While the EBV lytic cycle contributes to infectious mononucleosis and oral hairy leukoplakia, it is also necessary for the development of EBV tumors (3–5). First, lytic replication amplifies the pool of latently infected cells from which B lymphomas may arise. Second, studies using animal models have demonstrated that lymphomagenesis is dependent on EBV’s ability to undergo lytic replication (6). Therefore, disrupting the lytic cascade is an attractive strategy for treating EBV-related diseases. That said, while there is a clear need for anti-EBV agents, such drugs currently do not exist.

A common approach to identify antiviral or anticancer agents is to screen existing compound libraries for candidates that interfere with phenotypes such as infection, virus production, and cell survival or proliferation. Once identified, derivatives are often generated and further screened for effect and toxicity. Exploration of the mechanism of action generally lags. Although an appealing strategy, this mechanism-blind approach is unable to predict specificity of function, potential off-target effects, or hint at how resistance may develop. Based on our earlier work on the EBV lytic cascade, we have pursued a mechanism-driven approach to identify anti-EBV compounds. As mentioned earlier, the EBV lytic cascade is highly regulated and can be kinetically grouped into expression of immediate early (IE) lytic genes followed by IE-driven transactivation of early (E) lytic genes, viral DNA replication, and expression of late (L) lytic genes. To ensure a robust lytic cascade, EBV intersects with and even repurposes components of cellular machineries. Our studies have demonstrated that the EBV viral protein kinase (vPK), a product of the E lytic gene BGLF4, directly phosphorylates and activates cellular phosphatidylinositol 3-kinase (PI3K)-related kinase ATM; active ATM then phosphorylates the cellular protein KAP1 to deactivate it. KAP1 is a universal corepressor that maintains heterochromatin on both host and viral genomes, including those of the human herpesviruses EBV, KSHV (Kaposi’s sarcoma-associated herpesvirus), and CMV (cytomegalovirus) (7–13). In the context of the cellular genome, phosphorylation of KAP1 at serine 824 results in remodeling and relaxation of bound DNA, a key event necessary for repairing double-strand breaks in heterochromatinized regions of the genome (14, 15). In the context of the EBV genome, we have reported that vPK-ATM-mediated phosphorylation of KAP1 deheterochromatinizes several EBV lytic genes, including the IE gene BZLF1, which encodes the latent-to-lytic switch protein ZEBRA. Thus, a positive feedback loop is established within which ZEBRA transactivates BGLF4 (vPK), vPK phosphorylates ATM, and ATM phosphorylates KAP1, resulting in further derepression of lytic genes and amplification of the lytic cascade (7, 8). We sought to disrupt this lytic amplification loop.

We report here a high-throughput mechanism-based live cell assay that allows screening for compounds able to interrupt the amplification loop. Using this functional assay, we screened a library containing 50,000 small molecules and identified candidates with variable degrees of inhibition of the lytic cascade. We validated several candidates by interrogating them further for their ability to impair key steps in the lytic cascade.

RESULTS

A high-throughput screening assay to identify inhibitors of the ZEBRA-vPK-ATM-pKAP1-ZEBRA loop.

We designed an assay based on our earlier work which described a positive feedback loop in which expression of EBV latent-to-lytic switch protein ZEBRA leads to transactivation of the early lytic gene that encodes vPK; vPK then directly phosphorylates ATM, causing ATM to phosphorylate KAP1 at serine 824 and consequently derepressing ZEBRA (Fig. 1A). By virtue of being a series circuit, inhibition of any component of this loop results in attenuation of phosphorylation of KAP1 and consequent impairment of EBV lytic replication (7–9); thus, the level of p-S824 KAP1 serves as a readout for EBV lytic activation.

FIG 1.

In-Cell Western (ICW) screening assay for identifying chemical inhibitors of EBV lytic cycle. (A) Positive feedback loop comprising EBV lytic proteins ZEBRA and vPK and cellular proteins ATM and KAP1 contributes to amplification of the EBV lytic cycle. This loop involves ZEBRA-mediated transcriptional activation of vPK leading to vPK-mediated phosphorylation of ATM at S2996; ATM then phosphorylates KAP1 at S824, resulting in derepression of BZLF1 (which encodes ZEBRA). Derepression of BZLF1 and other EBV lytic genes (many of which are also transcriptional targets of ZEBRA) results in amplification of the EBV lytic cascade. (B) Workflow of ICW assay that measures p-S824 KAP1 signal normalized to total KAP1 in CLIX-FZ BL cells in which EBV is induced into the lytic phase through addition of doxycycline. (C) An example of an ICW assay showing images of wells in which cells were left untreated, treated with doxycycline, or treated with doxycycline plus an ATM inhibitor (KU55933) and then stained with an antibody to p-S824-KAP1 versus total KAP1. (D) Distribution of SSMD values for initial screen. Assay quality was assessed using strictly standardized mean difference (SSMD; β) using doxycycline-induced (D⁺) and uninduced (D⁻) CLIX-FZ BL cells.

In-Cell Western (ICW) is a quantitative immunofluorescence-based high-throughput assay. We designed such an ICW assay to simultaneously quantitate immunostained phosphorylated versus unphosphorylated KAP1 within each cell to minimize variations between experiments and runs. Furthermore, to avoid potential interactions between candidate inhibitors within a library and chemical triggers commonly used to lytically activate tightly latent but inducible EBV-positive (EBV⁺) cell lines, we used CLIX-FZ cells, a BL-derived cell line that we had previously generated; in CLIX-FZ cells, we induce ZEBRA expression from a stably integrated doxycycline-inducible BZLF1 gene (8). Figure 1B shows a schematic of such an assay in which we induced the EBV lytic cycle in CLIX-FZ cells using doxycycline (Dox). We then measured the ratio of p-S824 KAP1 to total KAP1 in each well of the plate. An example of such an assay is shown in which KU55933, a known inhibitor of ATM, impairs phosphorylation of KAP1 at S824 (Fig. 1C).

To assess assay quality, the strictly standardized mean difference (SSMD; β) was calculated using cells uninduced and induced with Dox (D⁻ and D⁺, respectively). Z′ factor is a common statistical method developed for quality assessment in high-throughput screening assays (16) However, Z′ factor does not account for nonnormal distributions, unequal variance, and/or outliers. On the other hand, SSMD accounts for variability across groups and does not depend on sample size, allowing for a rigorous probability interpretation (17). In addition, SSMD accommodates various effect sizes of the controls through the use of multiple thresholds (18). As such, SSMD has been shown to be an accurate indicator of quality that can address the limitations of moderate controls (19). Although commonly used for RNA interference (RNAi) screens, SSMD is emerging as a reliable statistical parameter for in vitro (20–22) and in vivo high-throughput drug screening assays (23–25). Figure 1D shows that during the initial screen, all plates yielded good or excellent SSMD values for moderate effect size (β = 4.610 ± 0.4943). Furthermore, treatment with a known inhibitor of ATM resulted in maximal inhibition of KAP1 phosphorylation within the parameters of the ICW assay (β = 4.618 ± 0.9055).

Identification of candidate compounds that inhibit phosphorylation of KAP1.

Fifty thousand compounds from a carefully selected library (DIVERSet-CL library, ChemBridge) that provides broad pharmacophore space coverage and chemical structure diversity were tested for their potential effects on EBV lytic activation. Of the 5,000 10-compound mixtures tested, three cocktails showed 30 to 50% inhibition of KAP1 phosphorylation associated with EBV lytic activation (data not shown). The corresponding 30 individual compounds from the 3 cocktails were examined individually for their effects on phosphorylation of KAP1 at serine 824. As shown in Fig. 2, out of 30, 22 compounds inhibited phosphorylation of KAP1 by ≥ 23%. Moving forward, we were able to obtain 21 of these 22 compounds from ChemBridge; these compounds demonstrated minimal impairment of metabolic activity in CLIX-FZ cells (Fig. 3).

FIG 2.

Validation of 30 candidate compounds by measuring their effects on KAP1 phosphorylation. CLIX-FZ BL cells were treated with doxycycline, doxycycline plus 10 μM (each) for 30 candidate compounds, or left untreated. Twenty-four hours later, cells were harvested for ICW as imaged (A) and analyzed for ratios between p-S824 KAP1 and total KAP1 (B). Error bars represent SEM of biological triplicates. Red line, mean of p-S824 KAP1:total KAP1 in doxycycline-treated cells. *, P < 0.05; **, P < 0.01; ns, not significant.

FIG 3.

Evaluating toxicity of 21 candidates on CLIX-FZ BL cells. CLIX-FZ BL cells were treated with doxycycline or doxycycline along with 21 individual compounds at different concentrations (10 nM, 1 μM, or 100 μM). At 24 h posttreatment, cell cultures were mixed with WST-1 assay substrate and incubated for 2 h for measurement. Error bars represent SEM of biological triplicates.

Effects of candidate antiviral agents on the release of virus and EBV genome replication.

While our initial screen hinged on measuring phosphorylation of KAP1, to assess antiviral potential of candidates, we tested their effects on readouts that directly measured distinct aspects of the EBV lytic phase. With release of infectious virions from lytically active cells as the most relevant indicator of antiviral effect, we first assayed for encapsidated extracellular viral genomes. We harvested cells treated with doxycycline and each of the 21 candidates after 48 as well as 72 h since we did not know the half-lives of the candidates or where and when in the lytic phase they functioned. As shown in Fig. 4, 17 candidates demonstrated significantly reduced (between 10% and 45%) amounts of encapsidated extracellular EBV genomes compared to cells exposed to doxycycline alone.

FIG 4.

Testing the effects of 21 candidates on release of encapsidated EBV from cells. CLIX-FZ BL cells were treated with doxycycline, doxycycline along with 1-μM or 10-μM concentrations of 21 individual compounds, or left untreated. Released virus particles were collected from culture supernatants and quantified via qPCR at 48 h (A) or 72 h (B) after treatment. An equal amount of supernatant from each sample was used as PCR template. Error bars represent SEM of 3 technical replicates from 2 experiments. Red line, mean encapsidated extracellular virions in doxycycline-treated cells. *, P < 0.05; **, P < 0.01.

A reduction in the amount of released EBV particles could result from blockade of events upstream or downstream of EBV DNA replication. To distinguish between these two possibilities, we measured the amounts of cell-associated EBV genomes after treatment with doxycycline and individual candidate compounds. Based on quantitative PCR (qPCR) results shown in Fig. 5, we were able to divide the candidates into two groups. Group I included 5 compounds (21H10-2, 21H10-4, 21H10-5, 21H10-6, and 21H10-7) which reduced the amounts of released EBV without affecting viral DNA replication, suggesting that they target postreplication events such as virion packaging or egress. In contrast, 12 compounds (21H10-8, 21H10-10, 47A4-1, 47A4-2, 47A4-4, 47A4-5, 47A4-6, 47A4-7, 47A4-8, 47A4-9, 47A4-10, and 47A7-4) belonging to group II reduced the amounts of extracellular EBV particles and EBV DNA replication, indicating that they function upstream of viral DNA replication; however, additional effects downstream of DNA replication cannot be excluded.

FIG 5.

Examination of the effects of 21 candidates on EBV replication. CLIX-FZ BL cells were induced with doxycycline, doxycycline plus 21 individual compounds at two different concentrations (1 μM or 10 μM), or left untreated. Cell-associated total DNA was extracted at 18 h (A) or 24 h (B) after treatment and quantitated by qPCR after normalizing to 18S genomic fragment. Error bars represent SEM of 3 technical replicates from 2 experiments. Red line, mean cell-associated EBV DNA in doxycycline-treated cells. *, P < 0.05; **, P < 0.01.

Effects of candidates on the expression of the EBV latent-to-lytic switch protein ZEBRA.

The EBV lytic cycle is initiated with the expression of the viral latent-to-lytic switch protein ZEBRA, which then triggers the expression of a series of kinetically ordered downstream viral genes, resulting in viral DNA replication followed by virus packaging and egress. We previously reported that KAP1 restrains the EBV immediate early gene BZLF1, which encodes ZEBRA, and EBV lytic activation-associated phosphorylation of KAP1 at serine 824 results in derepression of BZLF1, consequently amplifying the entire lytic cascade. To further dissect the mechanisms of action of the compounds that impaired p-S824 KAP1 levels, we tested their ability to antagonize expression of ZEBRA. Again, we used CLIX-FZ cells to avoid possible interactions between chemical lytic cycle inducers and candidate compounds; furthermore, CLIX-FZ cells allowed us to investigate the effects of candidate inhibitors on inducible FLAG-ZEBRA-driven expression of endogenous ZEBRA from the viral genome. As shown in Fig. 6, nine candidates (21H10-3, 21H10-4, 21H10-5, 21H10-6, 21H10-7, 21H10-8, 21H10-10, 47A4-3, and 47A4-4) demonstrated >25% reduction in endogenous ZEBRA levels following normalization to doxycycline-induced FLAG-ZEBRA. Thus, while 9 of the 21 candidates that impair phosphorylation of KAP1 impair ZEBRA’s ability to autoregulate itself, the others may impair other aspects of the lytic cascade.

FIG 6.

Investigating the effects of 21 candidates on expression of the EBV latent-to-lytic switch protein ZEBRA. CLIX-FZ BL cells were treated with doxycycline, doxycycline plus 10 μM concentrations of 21 individual compounds, or left untreated. After 24 h, cells were harvested for immunoblotting with anti-ZEBRA antibody. Numbers reflect ratios between endogenous ZEBRA and doxycycline-inducible FLAG-ZEBRA. Numbers in green indicate inhibition > 25%. Experiments were performed at least twice.

Effects of candidates on events downstream of ZEBRA.

ZEBRA is an apical transcription factor that transactivates early and late lytic genes. Once expressed, early genes (together with ZEBRA) are needed for viral DNA replication as well as downstream events. For example, the early gene EA-D is a DNA polymerase processivity factor that is essential for viral genome replication (26). Another early gene vPK phosphorylates viral targets such as EA-D and components of the viral nuclear egress complex (27–29) as well as cellular targets such as ATM (7, 30). Because ZEBRA activates downstream events of the lytic cascade, we tested the nine compounds that inhibited ZEBRA autoregulation for their ability to also impair some of these downstream events. We found that five candidates (21H10-5, 21H10-6, 21H10-8, 21H10-10, and 47A4-3) suppressed EA-D expression (Fig. 7). Three of these (21H10-5, 21H10-8, and 21H10-10) also reduced phosphorylation of the vPK substrate EA-D, whereas two others (21H10-4 and 47A4-4) reduced phosphorylation of EA-D without affecting EA-D levels (Fig. 7).

FIG 7.

Examination of the effects of 9 candidates on expression and phosphorylation of EBV early lytic gene product EA-D. CLIX-FZ BL cells were induced with doxycycline, doxycycline plus 10 μM concentrations of 9 individual compounds, or left untreated. Cells were harvested at 36 h posttreatment for immunoblotting with indicated antibodies. The top row of numbers indicates the ratio between phosphorylated EA-D (higher-molecular-mass band) and unphosphorylated EA-D (lower-molecular-mass band). The bottom row of numbers represents relative amounts of EA-D protein after normalization to FLAG-ZEBRA. Numbers in green indicate inhibition >25%. Experiments were performed at least twice.

Taken together, our high-throughput live cell-based assay identified 22 novel compounds that inhibited phosphorylation of KAP1 in the context of the EBV lytic cascade. Many of these impaired virus release and viral DNA replication and subsets of these negatively impacted ZEBRA’s ability to autoregulate itself, transactivate its downstream target EA-D, and phosphorylate EA-D. Notably, two candidates, 21H10-8 and 21H10-10, inhibited (i) phosphorylation of KAP1, (ii) EBV release, (iii) EBV DNA replication, (iv) ZEBRA autoregulation, (v) expression of EA-D protein, and (vi) phosphorylation of EA-D (Table 1). The chemical structures and 50% inhibitory concentrations (IC₅₀s) of both candidates, in the low micromolar range (2.02 μM for 21H10-8 and 4.57 μM for 21H10-10), are shown in Fig. 8A to C; notably, also, both demonstrated negligible toxicity (Fig. 8D and E). To ensure that the candidate compounds indeed targeted the amplification loop, we used an independent assay, i.e., immunoblotting, to demonstrate that 21H10-8 and 21H10-10 and 2 others, 47A4-9 and 47A4-10, suppressed the levels of pKAP1 (Fig. 9). We also found that 21H10-8 and 21H10-10 inhibited EBV release in a lymphoblastoid cell line (Fig. 10). Thus, both candidate inhibitors were effective in impairing the lytic cascade in EBV⁺ cell lines irrespective of EBV type (type 1 EBV in lymphoblastoid cell line versus type 2 EBV in CLIX-FZ BL cells) or lytic trigger (chemical trigger in lymphoblastoid cell line versus ectopic ZEBRA in CLIX-FZ BL cells).

TABLE 1.

Summary of effects of 21 candidate compounds on the EBV lytic cycle^a

^aGreen, inhibitory effect; red, no effect; white, not tested.

FIG 8.

IC₅₀s and toxicity of compounds 21H10-8 and 21H10-10. (A) Chemical structures of compounds 21H10-8 and 21H10-10. (B and C) CLIX-FZ BL cells were treated with doxycycline or doxycycline plus indicated doses of 21H10-8 (B) or 21H10-10 (C) and harvested 72 h later for qPCR quantification of released virus particles in culture supernatants that were concentrated by centrifugation as described in Materials and Methods. (D and E) A549 cells were exposed to 21H10-8 (D) or 21H10-10 (E) at doses corresponding to those in panels B and C. After 24 h, cell cultures were mixed with WST-1 assay substrate and read after another 2 h. Error bars represent SEM of triplicate technical repeats derived from two biological repeats.

FIG 9.

Selected candidate compounds inhibit phosphorylation of KAP1 in an independent readout. CLIX-FZ BL cells were induced with doxycycline or doxycycline plus individual compounds at indicated concentrations or left untreated. Cells were harvested at 18 h posttreatment for immunoblotting with indicated antibodies. Compounds 21H10-8 and 21H10-10 were tested using concentrations approximating their IC₅₀ values (2 and 5 μM, respectively) and 10 μM, while 47A4-9 and 47A4-10 were tested at 1-μM and 10-μM concentrations. The experiment was performed twice.

FIG 10.

Compounds 21H10-8 and 21H10-10 inhibit release of encapsidated EBV from a lymphoblastoid cell line. Cells were exposed to the lytic trigger sodium butyrate and compound 21H10-8 or 21H10-10 at indicated concentrations, or left untreated. Culture supernatants were harvested 72 h later, concentrated by centrifugation as described in Materials and Methods, and assayed for released virus particles by qPCR. Error bars represent SEM of 3 technical replicates. *, P < 0.05; **, P < 0.01.

DISCUSSION

We have developed a mechanism-based live cell screening assay in which we use phosphorylation of cellular KAP1 at serine 824 as the primary readout. This important posttranslational modification is essential for the lytic cascade of both type 1 and type 2 EBV, KSHV, and CMV (7, 8, 11, 13). This assay therefore provides a broad platform for identifying lytic cycle inhibitors of several medically relevant herpesviruses; the lytic cycle is responsible for most herpesvirus-mediated pathology. By participating in a positive feedforward loop to amplify the lytic cascade (7, 8), phosphorylation of KAP1 serves as a convenient readout for inhibitors that function at multiple nodes of this amplification loop. By applying this broad strategy, we have identified 17 compounds that diminish the amounts of released EBV.

During our studies of host epigenetic modifiers that regulate the herpesvirus lytic cascade (7–10), we have learned that phosphorylation of KAP1 is a key event that results in EBV lytic cycle amplification and its completion. We have shown that the cellular kinase ATM is exclusively responsible for phosphorylating KAP1 at serine 824 and that vPK (transactivated by ZEBRA) activates ATM by phosphorylating it at S2996. It is therefore reasonable to infer that any compound that blunts KAP1 phosphorylation at S824 (i) blocks ATM directly or indirectly, (ii) blocks the target site on KAP1, (iii) interferes with vPK, (iv) blocks the target site on ATM, or (v) blocks ZEBRA’s ability to transactivate downstream genes, including vPK. While the precise mechanism by which each inhibitor interrupts the EBV lytic cascade is not clear, their behaviors in more targeted readouts provide insights into possible mechanisms of action and molecular targets. Of the 17, 12 appeared to function upstream of or at DNA replication, while 5 impaired postreplication events such as packaging and virus egress. Of the 12, 9 interfered with ZEBRA’s ability to autoregulate its expression. Because a threshold level of ZEBRA may be enough to trigger lytic (re)activation, we not only examined the ability of the compounds to inhibit ZEBRA autoregulation but also to inhibit downstream readouts. Indeed, of the nine that interfered with autoregulation of ZEBRA, five also interfered with expression of EA-D, a transcriptional target of ZEBRA. Three of the five and two others inhibited phosphorylation of EA-D, a modification mediated by vPK, which is itself a transcriptional target of ZEBRA. These five candidates that inhibited phosphorylation of EA-D also impaired virus release. Notably, vPK-phosphorylated EA-D and vPK are also known to regulate virus egress, suggesting that these five candidates may impair mechanisms related to vPK function. While impaired expression of other lytic proteins may have served as another readout, we were unable to do so, as antibodies to most lytic proteins are not readily available. Furthermore, it is not clear which viral proteins other than vPK and EA-D are part of the lytic cycle amplification loop.

Upon closer inspection of compounds 21H10-8 and 21H10-10, we found that the hydrogen bond donor and acceptor groups resembled those of known kinase inhibitors. Comparison of the sequences of the kinases targeted by such similar inhibitors to the EBV vPK using PSI-BLAST with default gap opening and extension penalties of −11 and −1, respectively, showed low (up to 21%) sequence identity to vPK with relatively high E values (lowest, 0.003) (31, 32). These results argue against constructing a homology-based model. Furthermore, when 21H10-8 and 21H10-10 were queried against the PubChem database, no kinase-related assay was found to be reported for similar compounds; however, four similar compounds were reported to be inactive against cyclic GMP-AMP synthase (cGAS) (33). The lack of effect of these similar compounds on the cGAS-related inflammatory pathway makes 21H10-8 and 21H10-10 good starting hits for lead optimization.

As expected, all except one candidate that inhibited DNA replication also inhibited virus release. While it is difficult to explain the reason for the single exception, overall, this consistency between the two readouts was reassuring. On the other hand, while several candidates did not affect DNA replication, they nonetheless affected virus release. This was not surprising, as KAP1 simultaneously represses multiple EBV lytic genes with roles in diverse functions within the lytic cycle (9). That candidates may inhibit more than one aspect of the lytic cascade is also a distinct possibility.

A screening tool such as ours is important, as there is a need for drugs that treat EBV-mediated diseases. Acyclovir is not particularly effective against EBV, while ganciclovir has significant toxicity. Maribavir targets EBV vPK and the UL97 protein kinase of human cytomegalovirus, but point mutations in UL97 render the virus resistant to maribavir (34, 35). In addition to viral proteins, our screening tool is likely to identify inhibitors that target cellular proteins such as ATM and KAP1, both members of the lytic amplification loop. Because KAP1 is important to the lytic cascade of several human herpesviruses, such inhibitors may also function against other herpesviruses. Targeting cellular proteins also limits the likelihood of developing drug resistance. Lastly, the possibility of identifying agents that enhance the lytic cascade and may be developed into oncolytic agents should not be overlooked.

Other groups have used different strategies to identify EBV-specific inhibitors. In two studies, experimental and in silico screening identified inhibitors to EBV proteins SM and EBNA1, respectively; SM is a lytic protein, while EBNA1 is predominantly a latency protein (36, 37). In a third study, agents known to block B cell signaling were found to also impair B cell signaling-mediated activation of the EBV lytic cascade (38). The main difference between the first two studies and ours is that instead of targeting a single EBV protein, we used a broader screen to identify inhibitors that may target both viral and cellular proteins that contribute to activation and amplification of the lytic cascade. By not using chemical agents to turn the lytic cascade on, our assay avoids any potential interference between such chemical agents and candidate inhibitors. The second reason to use ectopic ZEBRA was that while physiologic lytic triggers (other than B cell differentiation) are not well-defined, a common upstream event is the expression of ZEBRA. That said, both 21H10-8 and 21H10-10 were also able to inhibit the lytic cascade when a chemical trigger activated the lytic cycle.

Our assay may be used to screen other compound libraries. While pooling compounds in a library increases the feasibility of screening large libraries, there could be disadvantages to pooling. For instance, individual compounds may exert additive, synergistic, and sometimes even antagonistic effects. As a result, compounds with mild effects may be selected for further studies, while others with substantial inhibitory effects may not be pursued further. Indeed, it is possible that had we selected additional wells that demonstrated <30% inhibition of phospho-KAP1/KAP1 signal during the initial screen, we may have identified additional compounds with substantial antilytic activity.

The lytic phase of EBV contributes to and is even needed for development of EBV cancers. In particular, EBV loads are monitored in transplant recipients to preempt the development of EBV-LPD. Because anti-EBV agents do not exist, such patients are currently treated with monoclonal antibodies that target all B cells. Such indiscriminate killing of infected and uninfected B cells often causes prolonged B cell aplasia, resulting in increased susceptibility to infections. The availability of EBV-specific agents would eliminate the need for therapy with such antibodies. Instead, transplant recipients could be treated or even prophylaxed with an anti(lytic) EBV agent to reduce the production of infectious particles. In a broader context, targeting KAP1 phosphorylation has the added benefit of potentially identifying anticancer agents that may target ATM or KAP1 in non-EBV-related cancers.

MATERIALS AND METHODS

Cell culture and chemical treatment.

A lymphoblastoid cell line (bearing type 1 EBV) and CLIX-FZ cells, an endemic Burkitt lymphoma (eBL) cell line bearing type 2 EBV and containing an inducible BZLF1 cassette, were cultured in RPMI 1640 supplemented with 10% fetal bovine serum (FBS) (Gibco) and 1% penicillin-streptomycin (Gibco). A549 cells were cultured in F-12K medium supplemented with 10% FBS (Gibco) and 1% penicillin-streptomycin (Gibco). Expression of ZEBRA in CLIX-FZ cells was induced with doxycycline (5 μg/ml; Sigma; product no. D9891), and ATM was inhibited using KU-55933 (10 μM, Sigma; catalog no. SML1109). EBV lytic cycle in the lymphoblastoid cell line was activated using 3 mM sodium butyrate as previously described (39). Effects of the DIVERSet-CL (ChemBridge) library containing 50,000 compounds formatted into 10 compounds per well at final concentrations of 1 μM were tested on CLIX-FZ cells using an In-Cell Western assay described below.

Antibodies.

Antibodies included mouse anti-FLAG Ab (Sigma; product no. F3165), goat anti-KAP1 Ab (Bethyl Laboratories; catalog no. A303-838A), rabbit anti-KAP1 Ab (Bethyl Laboratories; catalog no. A300-274A), rabbit anti-p-S824-KAP1 Ab (Bethyl Laboratories; catalog no. A300-767A), mouse anti-EA-D Ab (EMD; catalog no. MAB8186), IRDye 800CW-conjugated donkey anti-rabbit IgG (Li-Cor Bioscience; catalog no. 926-32213), and IRDye 680RD-conjugated donkey anti-goat IgG (Li-Cor Bioscience; catalog no. 926-68047).

Immunoblotting and In-Cell Western assay.

Cell lysates were harvested and analyzed via immunoblotting with indicated antibodies as previously described (10). For In-Cell Western assay, cells were fixed with Cytofix/Cytoperm solution (BD Bioscience; catalog no. 554722) at room temperature for 15 min, washed with 1× BD Perm/Wash buffer (BD Bioscience; catalog no. 554723), and incubated with goat anti-KAP1 and rabbit anti-p-S824-KAP1 antibodies for 1 h at room temperature. After washing, cells were further incubated with IRDye-conjugated donkey anti-rabbit and anti-goat IgG secondary antibodies for another hour at room temperature, washed, and subjected to image capture using Odyssey 9120 imaging system (Li-Cor Bioscience).

Cell viability assay.

Three hundred thousand CLIX-FZ cells in 300 μl of RPMI medium were mixed with 33 μl of WST-1 substrate (Sigma; catalog no. 5015944001) and incubated at 37°C for 2 h, and absorbance was measured and quantified from each well at a 450-nm wavelength. We seeded 150,000 A549 cells in a 96-well plate; 24 h later, the culture medium was changed with 300 μl of fresh F-12K medium mixed with 33 μl of WST-1 substrate. Two hours later, absorbance was measured at a 450-nm wavelength.

qPCR to quantitate EBV load.

Cell-associated EBV DNA was prepared as previously described (8), and relative amounts of viral DNA were quantified using quantitative PCR by amplifying the EBV BamW locus with forward primer, AGGCTTAGTATACATGCTTCTTGCTTT, and reverse primer, CCCTGGCTGATGCAACTTG. To quantify relative amounts of released EBV particles, equal amounts of supernatants from CLIX-FZ cell cultures were subjected to DNase treatment followed by qPCR using primers targeting the EBV the BamW locus as above. The relative EBV genome copy numbers were calculated using a standard curve qPCR with BACmid p2089 serving as a template. For Fig. 8B and C and Fig. 9, released EBV particles from 1 ml of culture supernatant were pelleted by centrifuging at 14,000 rpm for 2 h at 4°C, washed with 1× phosphate-buffered saline (PBS) (and recentrifuged as above), and treated with DNase in 20 μl of 1× DNase buffer followed by inactivation of DNase; 1 μl was used for qPCR analysis using BamW primers as described above or BALF5 primers as described previously (39).

Statistical analysis and SSMD calculation.

For all statistical analysis, GraphPad Prism software (La Jolla, CA, USA) was used. Data are represented as mean ± standard error of the mean (SEM) unless noted otherwise. The α level (type 1 error) was set at 0.05. Differences were considered significant when the probability of type 1 error was less than 5% (P < 0.05). Means of two groups were compared using the two-sided unpaired Student's t test. To assess ICW assay quality, SSMD (β) was calculated using the formula $\frac{\text{mean}(C_{\text{positive}})-\text{mean}(C_{\text{negative}})}{\sqrt{[\text{SD}{(C_{\text{positive}})}^2]+[\text{SD}{(C_{\text{negative}})}^2]}}$. Cutoff values for poor, inferior, good, and excellent assays were set as follows: poor assay when β < 0.5, inferior assay when 1 > β ≥ 0.5, good assay when 2 > β ≥ 1, and excellent assay when 2 ≥ β.