|Home | About | Journals | Submit | Contact Us | Français|
Latent infection with Epstein-Barr Virus (EBV) is a carcinogenic cofactor in several lymphoid and epithelial cell malignancies. At present, there are no small molecule inhibitors that specifically target EBV latent infection or latency-associated oncoproteins. EBNA1 is an EBV-encoded sequence-specific DNA-binding protein that is consistently expressed in EBV-associated tumors and required for stable maintenance of the viral genome in proliferating cells. EBNA1 is also thought to provide cell survival function in latently infected cells. In this work we describe the development of a biochemical high-throughput screening (HTS) method using a homogenous fluorescence polarization (FP) assay monitoring EBNA1 binding to its cognate DNA binding site. An FP-based counterscreen was developed using another EBV-encoded DNA binding protein, Zta, and its cognate DNA binding site. We demonstrate that EBNA1 binding to a fluorescent labeled DNA probe provides a robust assay with a Z-factor consistently greater than 0.6. A pilot screen of a small molecule library of ~14,000 compounds identified 3 structurally related molecules that selectively inhibit EBNA1, but not Zta. All three compounds had activity in a cell-based assay specific for the disruption of EBNA1 transcription repression function. One of the compounds was effective in reducing EBV genome copy number in Raji Burkitt lymphoma cells. These experiments provide a proof-of-concept that small molecule inhibitors of EBNA1 can be identified by biochemical high-throughput screening of compound libraries. Further screening in conjunction with medicinal chemistry optimization may provide a selective inhibitor of EBNA1 and EBV latent infection.
Epstein-Barr virus (EBV) is a human gammaherpesvirus that infects over 90% of the adult population world-wide (reviewed in1, 2). Like all herpesviruses, EBV establishes a stable latent infection that can persist for the life of the host. EBV typically establishes a latent infection in long-lived B-lymphocytes3–5, but can also be found in some epithelial cells of the nasopharynx and gastro-intestinal tract6–8. During latent infection, the EBV genome exists as a multicopy episome that expresses a limited set of viral genes important for viral persistence and host-cell survival9, 10. Viral DNA and gene products have been found consistently in a number of human malignancies and are thought to drive cancer cell evolution10. EBV is considered a causative agent in endemic forms of Burkitt's lymphoma, nasopharyngeal carcinoma, and AIDS-associated B-cell lymphoma. EBV is also a causative agent of lymphoproliferative disease in immunosuppressed individuals, especially among recipients of organ transplants11. The viral genes that drive cellular proliferation and carcinogenesis are most commonly associated with EBV latent infection.
At least nine viral proteins and multiple non-coding RNAs have been detected in cells latently infected with EBV. Many of these have oncogenic potential when expressed ectopically in various model systems12, and some are essential for EBV immortalization of primary B-cells in tissue culture. However, many of these viral oncogenes are not consistently expressed in EBV-associated tumor tissue where the viral genome can persist in a more quiescent state relative to the initial transforming process. In most EBV-associated tumor tissue, the viral genome is maintained through the consistent expression of the Epstein-Barr virus encoded nuclear antigen 1 (EBNA1)13. EBNA1 is a sequence-specific DNA binding protein that binds to the EBV origin of plasmid replication (OriP) and facilitates the DNA replication and nuclear persistence of the viral genome in proliferating cells14–16. EBNA1 is consistently expressed in most, if not all, EBV-associated tumors and is required for the efficient establishment of EBV infection during B-cell immortalization. Inhibition of EBNA1 by siRNA depletion17 or by ectopic expression of a dominant negative mutant leads to loss of B-cell survival18. EBNA1 binds to several sites in the EBV genome where it is known to affect both viral chromosome stability and gene expression programs19–22. Recently, EBNA1 was also found to bind to several sites in the cellular genome, suggesting that it might alter host cells functions by directly interacting with chromosomal locations23.
Small molecule inhibitors of virus infection have been identified for numerous viruses, including herpesviruses24, 25. Among these, acyclovir and phosphonoacetic acid derivatives have proven to be the most effective family of inhibitors of herpesvirus DNA polymerases and lytic cycle DNA replication25. Despite these potent inhibitors of DNA polymerase, there remain few therapeutic agents that are equally effective against latent infection of herpesviruses. EBV is a particularly attractive herpesvirus to target for latent infection because it expresses a large number of well-characterized proteins during latency and can be readily cultured as a latent virus. Since most EBV pathogenesis is associated with latent infection, identification of inhibitors of latent infection are of great clinical significance. Inhibitors of lytic infection are only marginally effective in the treatment of EBV-associated malignancy. A few chemotherapeutic anti-cancer agents have been found to inhibit EBV infections. Hydroxyurea has been shown to cause the loss of EBV genomes from some Burkitt lymphoma tissue, and has been used effectively in the treatment of EBV-associated thymomas26, 27. Since hydroxyurea is thought to act on the cellular enzyme ribonucleotide reductase, its inhibition of EBV is indirect, and therefore subject to potential cellular toxicities and complications for long term use28, 29.
A broadly used method for development of small molecule inhibitors is to use high-throughput screening of compound libraries to identify candidate lead compounds that can be further derivatized to improve medicinal and pharmacological properties. In this work, we describe the development of a robust fluorescence polarization (FP) high-throughput screening assay for small molecule inhibitors of EBNA1 DNA binding. FP assays are ideal for HTS since they are homogeneous real-time assays that can used to rapidly assess the binding properties in solution30. Moreover, FP assays have been used successfully to develop HTS for inhibitors of numerous enzyme-substrate interactions, as well as for identification of inhibitors of DNA binding proteins, like C/EBP and Myc31, 32. EBNA1 is an attractive target for small molecule inhibitors because of its critical and consistent role in EBV-associated tumors. EBNA1 is also an attractive target because it is a viral encoded protein with no known host-cell orthologue. Identification of small molecule inhibitors of the EBNA1 DNA binding function will be a first step in the development of a selective inhibitor that is effective in cell-based and animal models of EBV tumorigenesis.
Amino acids 459 to 607 of EBNA1, encoding the DNA binding domain (DBD) were expressed as a hexa-histidine fusion protein in Escherichia coli. Expression was induced in Rosetta2 cells with 0.3 mM IPTG for 3 hours at 25°C. Soluble protein was recovered using a modified version of the method described by Frangioni & Neel and purified via Ni-NTA agarose33. Bound protein was extensively washed with 20mM HEPES, pH 7.9, 1M NaCl, 5 mM 2-mercaptoethanol, 40 mM imidazole and 10% glycerol to dissociate nonspecific DNA bound to EBNA1 prior to elution in buffer containing 250 mM Imidazole. Peak fractions of the eluted proteins was pooled and dialyzed against 20mM HEPES, pH 7.9, 500 mM NaCl, 5 mM 2-mercaptoethanol, 10% glycerol, and 0.2 mM PMSF.
A reaction mix was prepared containing 200 mM NaCl, 20mM Tris-Cl pH 7.4, 1 mM DTT, 10ug/mL BSA, and 10nM Cy5 Labeled EBNA1 BS Hairpin with or without 246 nM purified recombinant EBNA1 DBD protein. This solution was incubated for 20 minutes at room temperature to promote the establishment of equilibrium binding of EBNA1 to the DNA hairpin prior to dispensing (BioTek MicroFlow Select) 30 uL of this solution to each well of a 384 well microtiter plate containing 0.5 ul of compounds dissolved in DMSO. Fluoresence polarization (EX: 620, EM: 680) was measured using an Envision Xcite Multilabel Reader (Perkin Elmer).
Chemical compounds that pass all Lipinski and drug reactivity filters used to qualify drug-like libraries and confirmed to be at least 80% pure by LC-MS were obtained from the Lankenau Chemical Genomics Center. Compounds were pre-plated (0.5ul) in 100% DMSO in columns 3–22 of 384 well black opaque Optiplates (Perkin Elmer). The 14,000 compounds were screened as a mixture of 10 compounds per well. Each compound in the library was screened at a final concentration of 15 μM and is compressed in two independent dimensional arrays orthogonal to each other, such that each compound is represented twice surrounded by nine different molecules in each assay well. DMSO (0.5 μl) was pre-plated in columns 1,2,23, and 24 of each assay plate. These wells were used for Maximum (DNA probe + protein) and Minimum (probe alone) controls of fluorescence polarization of EBNA1 DNA binding. To compound containing assay plates, 30 μl of a preformed EBNA1: Cy5-DNA hairpin complex was dispensed to wells using a BioTek MicroFlow Select. After a 1 hr incubation at room temperature, Fluoresence polarization (EX: 620, EM: 680) was measured using an Envision Xcite Multilabel Reader (Perkin Elmer). Percent inhibition of EBNA1 DNA binding was calculated for each compound well relative to assay plate control wells (i.e. % inhibition=(mPMax-mPCmpd)/(mPMax-mPmin) × 100). Upon deconvolution, results were stratified into 4 categories: actives (i.e. the bioactive compound displays >75% inhibition of EBNA1 DNA binding and cleanly maps to a unique well in both the horizontal and vertical dimensions), ambiguous (i.e the bioactive compound maps to 2 or more wells in either dimension), orphan (i.e. an orthogonal match can not be identified in the second dimension), and inactive (<74% inhibition of EBNA1 DNA binding activity). This threshold cutoff was determined by averaging by normalized percent inhibition of all data points, plus three standard deviations of that average. This threshold cutoff yielded a hit rate of ~0.3% that were selected for confirmation. Chemicals selected for further analysis were reordered as powders from ChemDiv (www.chemdiv.com) and were designated as LB2 (#3241-0296), LB3 (#3241-0772), LB7 (#1071-0020), and LC7 (#K048-1003).
An EMSA reaction buffer was prepared containing 10% Glycerol, 200 mM NaCl, 20mM Tris-Cl pH 7.4, 1 mM DTT, 10μg/mL BSA, 10nM Cy5 Labeled EBNA1 BS Hairpin and with or without 246 nM purified EBNA1 DBD. This solution was incubated for 20 minutes at room temperature to promote equilibrium binding. 30 μL of this solution was dispensed to eppendorf tubes containing 0.5μL of a test compound in DMSO and mixed. Samples were then loaded onto a 6% polyacrylamide gel and electrophoresed for 90 minutes at 170V in 1× TBE. Nucleic acid migration was visualized using a Typhoon Imager (General Electric).
To determine the relative IC50s of candidate hit compounds, an 11- point, 2 fold titration of each compound in 100% (v/v) DMSO was assessed in duplicate with a 80 uM compound concentration as the upper limit. Percent inhibition of EBNA1 DNA binding at each concentration of a compound was calculated relative to assay plate control wells (i.e. % inhibition=(mPMax-mPCmpd)/(mPMax-mPmin) × 100). Data from duplicate measurements were fitted separately to a 4 parameter fit logistic model in order to determine the relative IC50 concentration for each compound. GraphPad Prism 5.0 software was used to generate a four-parameter fit dose-response curve.
293T cells (human embryonic kidney cells transformed with SV40 T antigen) were added to 24-well plate at a concentration of 50,000 cells/well in DMEM media with 10% FBS. A Qp-Luciferase reporter construct was designed to assay the DNA binding function of EBNA1 in cell-May based assays. The Qp region of EBV (nucleotides 49712-50,250) were amplified by PCR and cloned into pGL3-Basic (Invitrogen) using Asp718-HinDIII restriction sites. Following an 18hr incubation at 37°C to allow cells to adhere, cells were transfected using 2μL Lipofectamine (Invitrogen) per well. All the cells were transfected with 10 ng/well of a Renilla expression plasmid (N1457), 200 ng/well of a QP-Luciferase Reporter Plasmid (N1852), and 6.25 ng/well of a FLAG-EBNA1 (N803) or control FLAG (N799) vector. After 6 hrs the transfection medium was replaced with fresh medium supplemented with test compounds added to achieve final concentrations ranging from 100 – 1.56 μM. Cells were incubated at 37°C for 48hrs and then analyzed for Luciferase activity using the Promega Dual Reporter system.
Raji cells (EBV positive Burkitt lymphoma derived cell line obtained from ATCC) were grown at a density of 2–4 × 106 cells/ ml in 2mL of RPMI media supplemented with 10% FBS, 10 mM Streptomycin, and 10 mM Penicillin in 6-well plates. Test compounds dissolved in 100% DMSO were added to cultures to achieve a final concentration of 10 μM. Cells were grown at 37°C for three days and then passaged 1:10 into fresh media with the same concentration of drug for an additional 72 hours. Genomic DNA was isolated using a ChIP Lysis Buffer/Phenol Chloroform Method. DNA was quantified using Quantitative PCR with primers for cellular Actin and the DS region of the EBV genome.
To develop a high-throughput biochemical assay for EBNA1 function, we expressed and purified the EBNA1 DNA binding domain (DBD) as a hexa-histidine amino-terminal fusion protein. Hexa-histidine tagged EBNA1 DBD (aa 459–607) was expressed in E. coli and purified over Ni-NTA agarose to near homogeneity (Fig. 1A). From four liters of IPTG induced E. coli, we recovered ~20 mg (~5mg/ml) of highly purified EBNA1 protein. As a control to eliminate non-selective compounds that broadly disrupt DNA: protein interactions, we also expressed and purified the hexa-histidine tagged Zta. Zta is another EBV-encoded sequence specific DNA binding protein required for lytic cycle replication. Zta is a member of the b-zip family of DNA binding proteins and shares no structural similarity to EBNA1, but does bind DNA with similar affinity. Zta was therefore selected for use as a biochemical counterscreen for EBNA1. Hexa-His tagged Zta was purified from E. coli to near homogeneity, similar to EBNA1 protein (Fig. 1B).
FP assays monitor changes in the molecular rotational properties of a polarized light in solution. As such, changes in rotation rate of fluorescent molecules can be induced by altering the mass of the fluorescently labeled molecule, which can be readily detected by FP. We therefore attempted to develop an assay utilizing a relatively small fluorescently labeled DNA probe bound by EBNA1. A similar assay was developed for Zta to use as a counterscreen to assess compound selectivity. For EBNA1 binding, a DNA probe was synthesized with a fluorescent Cy5 modification at the 5' terminus. Since the consensus sites for EBNA1 are palindromic, we initially designed a self-annealing oligonucleotdie containing an 18 bp perfect palindrome with the high affinity consensus sequence from the OriP family of repeats (GGGTAGCATATGCTACCC)34, 35. EBNA1 bound this probe with high efficiency in electrophoretic mobility shift assays (EMSA) (Fig. 2A, probe A), but gave suboptimal results in FP due to the formation of a fast migrating species that was incapable of binding EBNA1. Since unbound fluorescent probe reduces FP efficiency, we tested whether synthesizing the binding site as a hairpin would improve the annealing efficiency, and eliminate what might be residual single strand DNA that failed to anneal as duplex DNA. The perfect palindromic hairpin DNA (Figure 1, probe C) bound similar to probe A, but did not eliminate the residual unbound forms. Another possible source of the unbound fast migrating species could be the formation of 9-bp hairpins or cruciforms that are not recognized by EBNA1. To eliminate the possibility of the 9-bp hairpin forming, we generated a non-palindromic site with two high-affinity half sites (GGGTAGCATATGCTATCTagatagcatatgctaccc). This probe bound EBNA1 with similar efficiency in EMSA, but performed significantly better in the FP assay (Fig. 2B, probe B). This probe lacked the unbound faster migrating species presumably due to the inability to form internal hairpins and secondary structures. Probe B was then selected for further assay development.
The FP assay was miniaturized to a 384-well microtiter plate format. During this process we optimized the readout relative to a number of variables that included assay volume, fluorescent tracer concentration, EBNA1 concentration, DMSO sensitivity, assay signal stability, and reagent stability as a function of temperature and freeze-thaw cycles. In culmination, we assessed the optimized assay's variation by calculating the Z-factor for two plates performed independently on three successive days, where 192 wells from each plate contained either the Cy5-DNA hairpin probe alone or the EBNA1: Cy5-DNA hairpin complex in 1.5% DMSO36. In general, the replicate plate experiment for the EBNA1 FP assay yielded an average Z factor of 0.6 with all 6 plates scoring >0.55 (Supplemental Figure 1). A similar assessment of the ZTA FP assay was completed in which the average Z factor was calculated to be >0.8 (Supplemental Figure 2). The higher Z-factor reflects some of the better performance parameters of Zta in the FP assay, including a larger difference in polarization values between bound and unbound probe (data not shown). FP assays were also used to calculate the EC50 for EBNA1 (8.35×10−8 M) and Zta (1.05×10−7 M) binding to their cognate DNA probes (Supplemental Figure 3). These findings indicate that EBNA1 and Zta perform comparably, and at satisfactory levels for compound library screening.
A small molecule library of 14,000 highly diverse compounds with pharmacologically properties in compliance with Lipinski's rule of five37 was selected for an initial screen to identify inhibitors of EBNA1 DNA binding in vitro. To increase cost-and time-effectiveness of the screening process, the library was plated in an orthogonally compressed format. The compression consisted of combining 10 compounds per well with each compound appearing in two different wells surrounded by 9 different compounds. This approach has proven to be highly efficient for large scale screening campaigns38. From this screen we identified ~40 compounds that displayed >75% inhibition at 10 μM concentration of compound (Fig. 3A and B). From liquid retests of these compounds from the original source of the library in both the EBNA1 and Zta FP assays. We found four compounds that demonstrated > 85% inhibition for EBNA1 and >10 fold selectivity relative to inhibition of Zta in two independent experiments of this counterscreen assay (Fig. 3C and D). These four compounds were further analyzed for their effective inhibitory concentrations of EBNA1 and Zta.
The chemical structures of the four compounds selected from the HTS and counterscreen are shown in Figure 4A. Compounds designated LB2, LB3, and LB7 are structurally related, while compound LC7 was a distinct class. The effective inhibitory concentration (IC50) was calculated using FP as well as by EMSA, as an independent secondary assay (Fig. 4B –E). IC50 values for inhibition of EBNA1 were between 1–2 μM for all four compounds (Fig. 4B). IC50 values for Zta inhibition were 38 μM for LB2, 85 μM for LB3, 236 μM for LC7, and unmeasurable for LB7 due to its essentially ineffectiveness up to 100 μM. The EMSA analysis corroborated the findings of the FP assay, where LC7 and LB7 had the greatest differential between inhibitory concentrations for EBNA1 relative to control Zta. The compounds LB7 and LC7 were at least 100-fold more selective for inhibition of EBNA1 relative to Zta.
LB2, LB3, LB7, and LC7 were tested in a cell-based assay for their ability to disrupt EBNA1 binding and function in live cells. EBNA1 is a potent transcriptional repressor of the Q promoter (Qp) because it binds directly to two sites positioned over the transcription initiation region (Fig. 5)20, 39. EBNA1 expression plasmid was cotransfected with a luciferase reporter plasmid fused to Qp. In the absence of EBNA1, Qp expression was 1.4 fold relative to the Renilla internal control expression vector. When EBNA1 was cotransfected at 1 μg (++) or 0.1 μg (+)/106 cells, Qp expression was reduced to values of 0.2 or 0.1, respectively, relative to Renilla. EBNA1 expression vector was set at 0.1 μg/106 cells and then assayed in the presence of varying concentrations of LB2, LB3, LB7, or LC7 ranging from 1.6 to 100 μM. We observed that LC7 was too cytotoxic for analysis in cell-based studies and was not further characterized (data not shown). However, LB2, LB3, and LB7 demonstrated some capacity to inhibit EBNA1 function. LB2, LB3, and LB7 were capable of derepressing Qp-Luciferase in the presence of EBNA1, with LB3 demonstrating the greatest inhibitory activity in this assay (Fig. 5). None of the compounds stimulated Qp-luciferase in the absence of EBNA1 or stimulated Renilla expression (data not shown). In these assays, LB2 and LB3 were most effective at derepressing Qp, suggesting that these two related compounds may be effective in vivo at disrupting EBNA1 function.
Candidate compounds were tested for their ability to deplete EBV genome copy number from latently infected Burkitt lymphoma cells. Raji Burkitt lymphoma cell line carries ~100 copies of EBV per cell, but the copy number can be reduced by treatment with a known EBV inhibitor, like hydroxyurea (HU)29. We compared the ability of the three compounds LB2, LB3, and LB7, which had some anti-EBNA1 activity in cell-based assays, to function similar to HU in causing the elimination of EBV episomes (Fig. 6). We found that LB7 caused a loss of EBV episomes similar or better than HU treatment. After six days in culture with 5 μM LB7, EBV DNA copy number was reduced to ~30% of its original content relative to cellular actin. In contrast, 50 μM HU reduced EBV DNA copy number to ~48% of its original content relative to cellular actin. In these assays, LB2 and LB3 were not significantly effective at reducing EBV copy number, but this may have been a partial consequence of their higher toxicity in Raji cells. The failure to observe selective inhibition of EBV positive cells by LB2 and LB3 may be due to their relatively high non-specific cytotoxicity.
In this work we have developed methods to identify small molecule inhibitors of EBNA1-DNA binding using biochemical assays. We screened a small compound library of ~14,000 molecules and identified ~4 candidates that showed selective inhibition of EBNA1 relative to a control DNA binding protein, Zta. Three of the compounds showed some activity in cell based assays, and one compound was comparable to HU in the elimination of EBV genomes during long-term (six day) treatment of EBV positive Burkitt lymphoma cells in tissue culture. While these findings suggests that methods are available for discovery of potential inhibitors of EBNA1, the candidate compounds identified are unlikely to be clinically relevant without significant modifications that enhance target selectivity and reduce cellular toxicity. Future campaigns with more extensive compound libraries would benefit from a combination of both biochemical and cell-based screening approaches. Furthermore, validation of the mechanism of inhibition using X-ray crystallography of the EBNA1-inhibitor complex will also provide critical information for enhanced inhibitor activity. Nevertheless, we demonstrate a proof-of-concept that small molecule inhibitors can be identified for EBNA1 DNA binding, which could be of great value in treatment of EBV-associated disease, as well as a research tool to control the functional binding of a high-affinity sequence specific DNA binding protein. We anticipate that additional screening, combined with structure-activity relationship and medicinal chemistry, may provide an effective small molecule inhibitor of EBNA1 for cellular and animal based assays.
We thank Andreas Wiedmer and other members of the Lieberman lab for technical instruction and support. We acknowledge the Wistar Institute Cancer Center Core Facility for Protein Expression, Libraries, and Molecular Screening and the Core Facilities for Genomics and Flow Cytometry. This work was funded in part by grants from NIH (3R21NS063906) to PML.