Sulfated Small Molecules Targeting EBV in Burkitt Lymphoma: From In Silico Screening to the Evidence of In Vitro Effect on Viral Episomal DNA

viral infection. Thus, using computational molecular docking, a virtual screen of a library of small molecules, including xanthones and ﬂavonoids (described with potential for antiviral activity against EBV), was carried out targeting EBV proteins. The more interesting molecules were selected for further computational analysis, and subsequently, the compounds were tested in the Raji (BL) cell line, to evaluate their activity against latent EBV. This work identiﬁed three novel sulfated small molecules capable of decreasing EBV levels in a BL. Therefore, the in silico screening presents a good approach for the development ofnewanti-EBVagents.

Received 24 May 2012, revised 13 November 2012 and accepted for publication 8 January 2013 Epstein-Barr virus (EBV) infection has been associated with several human malignancies, including Burkitt lymphoma (BL), as EBV is responsible for transforming B cells and contributing to their malignant phenotype.Epstein-Barr virus is a 184-kbp double-stranded DNA linear genome virus which, once in the nucleus of infected cells, circularizes adopting an episomal shape (1).Epstein -Barr virus episome presents two origins of replication, oriP and oriLyt, which are used to replicate the virus genome during its latent or lytic phase, respectively.Epstein-Barr virus may adopt different types of latency (mainly type I, II, and III) during the latent phase of infection, depending on the proteins expressed (2).In type I latency, a very restricted number of latent proteins are expressed, whereas in type III latency, all set of latent proteins are expressed.Epstein-Barr virus nuclear antigen 1 (EBNA1) is the only latent protein that is expressed in all types of latencies.Importantly, the association of this protein with the oriP sequence is required for both initiation and maintenance of replication of EBV episomes in infected cells (3)(4)(5).Moreover, EBNA1 is also able to regulate its own expression (6) and, by binding to oriP, to enhance transcription from several viral promoters, such as the LMP1 promoter, leading to the regulated expression of other latent proteins (7).The replication of EBV episomal DNA also requires cellular proteins, such as DNA polymerase, which is not sensitive to antiviral drugs (8).
The switch from latency to lytic phase may occur following different stimulus (9)(10)(11)(12)(13).The expression of the immediateearly protein Zta was found to trigger viral lytic replication by activating several promoters, therefore leading to the expression of early lytic genes such as the early-antigendiffuse component (EA-D), the EBV DNA polymerase (involved in viral DNA replication), and the BamHI fragment H rightward open reading frame 1 (BHRF1, an anti-apoptotic molecule homologue of cellular Bcl-2).Following this stage, the expression of late EBV proteins involved in the structure of the virus will occur, and viral DNA replication will not depend on the cellular machinery anymore.Completion of the viral lytic cycle results in the production of new infectious viruses that will ultimately lead to cell death and viral dissemination.
As latent EBV is associated with the development of various diseases and because in therapeutics there are no available drugs to target latent EBV, the development of such compounds is necessary.In fact, even though there are some broad-spectrum antiherpes drugs available, which might be used against EBV, they are only active in the lytic phase of EBV infection (8).This is due not only to the fact that EBV episomal replication depends on the host cellular machinery to occur, but also to the fact that some antiviral drugs such as ganciclovir need to be processed by lytic viral proteins to become active (14,15).Therefore, as most EBV+ tumor cells (such as BL cells) have latent EBV, they are resistant to these known antiviral drugs (8).Thus, there is a strong need for the development of new antiviral agents that can target cells that have the virus in a latent state and that do not depend on the reactivation of the virus.Some strategies have been proposed to target latent EBV.For example, hydroxyurea (HU) leads to EBV episome loss from cells and has been efficiently used in vitro (16) as well as in vivo, in the treatment of EBV-associated primary central nervous system lymphoma (17).However, this drug has the disadvantage of leading to the accumulation of additional mutations in the cellular genome (18).Other approaches, using antisense, RNAi, or DNAzymes strategies to downregulate viral oncoproteins such as LMP1 in vitro, have been shown to decrease proliferation and to increase sensitivity to apoptotic stimulus in the treated cells (19,20).In addition, viral lytic replication was shown to be reduced following downregulation of LMP2B by RNAi (21).
Several natural and synthetic compounds, including flavonoids (22)(23)(24)(25), xanthones (26), and other small molecules (27)(28)(29)(30), have been described to have antiviral activity against EBV.The purpose of the current study was to identify compounds with antiviral activity against EBV, by carrying out an in silico screen of molecules from a library of flavonoids and xanthones from the CEQUIMED-UP research center using computational molecular docking and to further investigate their effect in vitro in an EBV+ BL cell line.Our results depict the advantages of strategies combining computational analysis with biological studies in the search for new antiviral agents and show evidence of antiviral activity against EBV for the three sulfated compounds studied.

Docking of in-house molecules
Structures of several EBV proteins were collected from Protein Data Bank (PDB) (31), namely EBNA1 (PDB IDs 1b3t and 1vhi), Zta (PDB IDs 2c9l and 2c9n), and BHRF1 (PDB ID 1q59).Virtual screening was carried out on a commodity PC running Linux Ubuntu 6.06.The software eHiTS 2009 from SimBioSys Inc (32,33) was used for active site detection and docking.Hyperchem from Hyper-Cube Inc (34) was used to draw and optimize 97 structures from CEQUIMED-UP library.No special preparation of the 3D structures was carried out because eHiTS automatically evaluates all of the possible protonation states for ligands and receptor.eHiTS was run using the DNAbinding residues as clip files, and the docking is performed, by default, within a 7-A margin around them.The docking accuracy was set to 2, the number of docking poses was set to 5, and SDF was chosen as the output file format.A multiple active site correction (MASC) was applied to the obtained docking scores (35).Open Babel (36) was used to manipulate the various file formats of ligands.PyMol from DeLano (37) and chimera from UCSF (38) were used for visual inspection of results and graphical representations.Pymol 'list_contacts' script (cutoff = 4.0 A) was used to determine possible contacts between the receptor and the ligand.For docking validation, a set of 150 molecules obtained from NCI database (39) were also docked to the several EBV proteins in the same conditions as previously described for the in-house molecules.

Cell culture
The BL cell line, Raji, was used.Cells were cultured in RPMI 1640 with Glutamax supplemented with 10% fetal bovine serum (Gibco, Invitrogen, Paisley, UK) and incubated in a humidified incubator at 37 °C with 5% CO 2 in air.
Cell treatment with the tested compounds Raji cells (5 9 10 5 /mL) were treated with 20 lM of compounds 1-3 for 48 h or 96 h.In the studies to analyze viral reactivation, 5 9 10 5 Raji cells/mL were incubated with 30 ng/mL TPA alone or simultaneously with 20 lM of each compound (1)(2)(3).The effect of the DMSO or H 2 O vehicle solvents (controls) or of medium (blank) was evaluated in all experiments.Viable cell number was counted using the trypan blue exclusion assay.
DNA extraction and EBV DNA analysis DNA was extracted from Raji cells following 96-h incubation with the tested compounds (1-3), as previously described in (43).Analysis of the EBV DNA load was carried out with conventional PCR and with real-time quantitative PCR using the following primers: (i) EBV DNA 5′ CGGTCGCCCAGTCCTACCAG 3′ and 5′ CCTGGAGAGG TCAGGTTACT-3′ ( 44) and (ii) actin 5′-GCCATGGTTGTG CCATTACA-3′ and 5′-GGCCAGGTTCTCTTTTTATTTCTG -3′ (45).For the conventional PCR, conditions were the following: 95 °C for 5 min, 35 cycles of amplification (95 °C for 30 seconds, 56 °C for 30 seconds, and 72 °C for 30 seconds), and a final extension step of 72 °C for 10 min.PCR products were then electrophoresed on 2% agarose gels containing GelStar (Lonza, Rockland, ME, USA).Densitometry analysis was carried out by quantifying the intensity of the bands using the software Quantity One -1D Analysis (Bio-Rad, Hercules, CA, USA).For real-time PCR, the Fast SYBR â Green Master Mix (Applied Biosystems, Paisley, UK) was used according to manufacturer's instructions.A standard curve was prepared using serial dilutions of Raji cells' DNA.A dissociation curve was generated for each set of primers, and a single peak was obtained; additionally, the PCR efficiency was estimated as being always higher than 95%.EBV DNA levels were normalized for cellular DNA (actin), to exclude any possible cytotoxic effect of the molecules.

Results
Three proteins were selected for the docking studies (latent EBNA1, immediate-early Zta, and lytic BHRF1).The reason for choosing these targets was as follows: (i) targeting EBNA1 has been suggested to be a potential strategy to affect latent EBV as this protein is essential in both initiation and maintenance of EBV episome replication and is also the only EBV latent protein present in nearly all EBV-infected cells; (ii) the expression of the immediateearly lytic protein Zta is sufficient for viral lytic reactivation, and therefore, its targeting may result in lytic cycle inhibition with no viral particles being produced; accordingly, the use of antisense and RNAi technologies to reduce Zta expression has been proven to effectively inhibit the EBV lytic cycle in vitro (48)(49)(50), and (iii) the early lytic protein BHRF1 is transiently expressed in early infection and has anti-apoptotic functions (it is an homologue of cellular Bcl-2) (51); in fact, BHRF1 may protect EBV-infected cells by inhibiting apoptosis to maximize viral production; thus, inhibitors directed to this viral target would probably trigger apoptosis of infected cells.

Docking using eHiTS â
Docking procedure resulted in a list of potential EBV inhibitors ranked according to the eHiTS scoring function.The total number of compounds screened was 97.These small molecules correspond mainly to xanthones and flavonoids (90) with diverse substituents such as hydroxy (28), prenyl (15), amine (22), sulfate ( 16), and others.Only the top three molecules (1-3, Table 1) were chosen for further computational analysis.Based on preliminary studies with five of the best molecules in cell lines, in which two of them (non-sulfated derivatives) presented no activity (data not shown), the same top three molecules were then further evaluated in vitro with biochemical assays.To comply with disclosure of information agreement, results from 94 of those compounds cannot be presented here.
As docking scores come from several targets, it is not possible to establish a direct comparison (only possible within the same protein and same binding site).Therefore, a MASC score was also given as previously described (35): MASC score = (S ij Àl i )/r i , where S ij is the original score, l i and r i are the mean and standard deviation of the scores for compound i.Calculation of mean and standard deviation was made using the docking scores obtained, and results are presented in Table 2.This statistical correction greatly increases the accuracy of ligand scoring and reduces the error of docking studies.Multiple active site correction is different for each molecule, and it is specific for the docking program and scoring function used.Multiple active site correction scores (Table 2) help to understand the most probable target protein for each small molecule tested.Compounds 1-3 all dock to Zta proteins (2c9n or 2c9l), whereas compound 3 is the only compound that docks to EBNA1 receptors (1b3t).
Afterward, a careful visual inspection of the SDF files from the best scored ligands on each of the three receptors, EBNA1, Zta, and BHRF1, was performed.As shown in Figure 1A,B, the best fitting molecule (compound 3) adopted several poses within the DNA-binding groove of EBNA1, filling the space supposedly occupied by the DNA doublestrand helix.As far as EBV immediate-early transcription factor Zta is concerned, as shown in Figure 1C, D, the best scoring molecules occupy a cavity formed by the two homologous chains X and Y.This binding pocket is the cavity used by the protein to bind to the EBV gene.As for BHRF1, docking was performed in a cavity flaked by residue ASN-61, described as forming the cavity where other proteins from the apoptotic pathway bind (52,53).However, none of the studied molecules on Table 1 was able to dock to BHRF1, the Bcl-2 homologue from EBV (54).

Docking validation
The NCI database, with more than 250 000 molecules, was used to randomly select 150 compounds.They were   S1.Regarding EBNA1 structures (1b3t and 1vhi), 1b3t was the best one; however, all of the decoy molecules presented a score worse than the score of compound 3 (Table 1).Regarding Zta structures (2c9l and 2c9n), 2c9n was better in distinguishing 'decoy' from active molecules.So, 1b3t and 2c9n target structures were used to analyze important residues involved in the binding.

Analysis of best ranked molecules' docking poses
The docking poses of all active compounds and possible interactions established within the active site of the analyzed EBV targets were carefully inspected.Initially, the general shape and polar/apolar main groups from the targets binding site were analyzed, as seen in Figure 2A.Three evident polar evaginations (P1, P2, and P5) and two polar groups (P3 and P4) were present.Two large hydrophobic regions (H1 and H2) also defined the   walls of the binding pocket.In fact, LYS514 represents polar region P1 and ARG469 represents region P5, whereas LEU536 occupies region P2 and ARG522 occupies region P3.GLY450 and ARG469 form the polar ground P4; on the other hand, the hydrocarbon chains on TYR518, LEU536, and ARG521 form the hydrophobic region H1 and ARG469 form region H2.
For better understanding of the most probable and important residues involved in the binding of EBNA1 inhibitors to EBNA DNA-binding pocket, compound 3 interactions were thoroughly analyzed as seen in Figure 2B.Amine groups present in ARG469, LYS514, ARG521, ARG522, GLN471, and GLY470, as well as a phenol group in TYR518, seem to be responsible for the establishment of H-bond interactions between ligand and target.Other type of interactions may occur, such as dipole-induced dipole interactions between polar groups in the ligand and induced dipoles in LYS514, LEU536, ARG521, LEU554, GLY480, GLN471, and ARG469 carbon chains seem to occur.The presence of a scaffold filling the EBNA1 DNA cavity and the presence of polar groups in strategic positions to establish hydrogen interactions with polar nitrogen and oxygen atoms of the amino acids surrounding that cavity would be an interesting strategy for the design of more potent EBV inhibitors.Involvement of some residues, such as TYR518, LYS514, and ARG469, had already been published as being important in the establishment of polar contacts, which reinforces these findings (56).As highlighted in Figure 2B, compound 3 has sulfate groups that can establish hydrogen interactions or even ion-dipole interactions with the polar amino acids on the walls of EBNA1 DNA-binding pocket.
Regarding the immediate-early transcription factor Zta, its structure was hypothesized to allow the design of new agents that block activation of the EBV lytic cycle by occupying the DNA-binding pocket (57).2c9n and 2c9lpdb files present the crystal structure of Zta's DNA-binding domain (residues 175-236, chains X and Z) bound to an EBV lytic gene promoter element.As seen in Figure 1D, the best fitting small molecules occupy exactly the same place as DNA, thus blocking EBV transcription, and therefore blocking the production of new virions.Analysis of the shape and characteristics of that site (Figure 3A) reveal the presence of two pairs of symmetric evaginations, P1-P2 and P3-P4.A polar sulfur region in P5, as well as polar region P6, is also relevant.A large hydrophobic cavity H1 is also easily detected.Comparing Figure 3A,B, it emerges that ARG190 from chains X and Z are represented in regions P1 and P2, whereas LYS194 from chains X and Z are represented in regions P3 and P4.P5 corresponds to CYS189 and P6 to SER186.The large hydrophobic cavity H1 corresponds to aromatic rings in both PHE193 residues.
Compound 1 was used as an example to analyze the possible interactions (Figure 3B).Guanidine group [-NHC(NH) NH-] in ARG190 appears as the principal region for the establishment of H-bond interactions with the ligand.Other polar contacts are established by amino and thiol groups in CYS189 and carbonyl group in SER186.Hydrophobic contacts happen between aromatic ring in PHE193 and hydrocarbon chains in LYS194 and ARG190.Both chains Y and Z contribute to the binding, and several conformations of the ligand inside the binding pocket (with ARG190X and ARG190Y being responsible for H-bonding) are possible and energetically favorable.To understand the binding mode of sulfate within the DNA-binding site of EBNA1 (1b3t and 1vhi), BHRF1 (1q59), and Zta (2c9l and 2c9n), compounds 1-3 in complex with the target protein were analyzed and important binding residues are listed in Table 3.

Analysis of the effect of compounds 1-3 in cellular viability
Antiviral drugs should not present cell toxicity.Therefore, the effect of compounds 1-3 on the number of viable cells was evaluated 48 h following the treatment of Raji cells with 20 lM of each tested compound or with the appropriate solvent control (H 2 O).Results show that, at the concentration analyzed, all the tested compounds (1-3) did not present great cytotoxicity to Raji cells (Figure 4).Indeed, while compound 3 hardly affected the number of viable cells compared with blank treatment (90% of viable cells in relation to blank), compounds 1 and 2 had only a slight effect on the number of viable cells, reducing viable cell number to 82% or 74%, respectively, which indicates a weak cytotoxic effect to these BL cells.In addition, these compounds had previously been tested in three other tumor cell lines (MCF-7, NCI-H460, and A375-C5), using the sulforhodamine assay, and presented no cytotoxicity (unpublished data, in preparation for publication).Chem Biol Drug Des 2013; 81: 631-644

Analysis of the effect of compounds 1-3 in EBV genome copy number
To verify whether the selected compounds 1-3 had an effect on EBV genome copy number in the Raji cells, viral DNA was analyzed 96 h after the incubation of the cells with 20 lM of each tested compound (or with the solvent, as control).By conventional PCR analysis, EBV DNA was detected, and the resulting bands were quantified by densitometry (by detecting cellular DNA through the amplification of actin DNA and further relating EBV DNA to cellular DNA levels).The results from the densitometry analysis of the EBV DNA load (Figure 5A) were indicative of a reduction in EBV DNA levels in Raji cells following the treatment with compounds 1-3, when comparing with the control cells (treated with solvent).To confirm these results, experiments were repeated, and EBV DNA load was quantified by real-time PCR (again normalizing for the amount of cellular DNA).Results confirmed the decrease in the EBV DNA load following treatment with compounds 1-3.Although the levels of decrease were not as high as previously observed by the conventional PCR approach (semiquantitative), the same trend of decrease was observed (with real-time quantitative PCR) for the three compounds (1-3).

Analysis of the effect of compounds 1-3 in LMP1 latent protein expression
Raji cell line has a type III latency EBV infection and may express all the latency viral proteins, such as LMP1.To determine whether the selected compounds (1-3) had an effect on the expression of LMP1 in Raji cells, cells were treated for 48 h with 20 lM of each tested compound or with their solvent as control (water).The protein levels of LMP1 (latent viral protein) or actin (as loading control) were then analyzed, by Western blot.Results indicate that all the tested compounds (1-3) reduced the expression of the latent LMP1 protein, with compound 3 presenting the strongest effect (Figure 6A), which was reproducible in all the independent experiments performed.
To exclude the possibility of these compounds inducing the EBV lytic cycle, as previously observed for some drugs included by some of us (10,13,58), the expression of a lytic protein (EA-D) following treatment with each of the compounds (1-3) or with their solvent was analyzed.There were no alterations in the levels of EA-D protein expression (whose increase would indicate lytic reactivation) following the treatment of the Raji cells with each of the tested compounds (1-3), thereby confirming that  these compounds did not cause viral reactivation (Figure 6B).

The tested sulfated compounds do not prevent viral reactivation
To determine whether the compounds 1-3 were able not only to interfere with latent EBV but also to prevent EBV lytic reactivation, Raji cells were treated for 48 h with the known lytic inducer TPA and simultaneously treated with 20 lM of each tested compound (or with their solvent).Reactivation of the virus was assessed by analyzing the levels of lytic protein EA-D (and of cellular actin, used as loading control) by Western blot.Results confirmed that TPA induced viral reactivation, as there was an increase in the EA-D protein levels following the treatment of Raji cells with this drug.However, none of the tested compounds, when added to the cells at the same time as TPA, was capable of inhibiting viral reactivation (Figure 7), as the levels of EA-D observed in cells treated with both TPA and with these sulfated compounds were similar to the levels observed following the treatment with TPA alone.

Discussion
In BL, as in the majority of the diseases associated with EBV infection, EBV DNA is maintained as an episome giving rise to a latent infection where no infectious virus are produced.The existing antiviral drugs do not target latent EBV (8), being effective only following induction of lytic EBV reactivation.Thus, some strategies for the treatment of these pathologies are based on the use of lytic EBV inducers, which convert EBV into the lytic form, which is susceptible to the activity of the existing antivirals.However, the resulting increase in the number of viral particles during lytic infection will end in infection of new cells, increasing the likelihood of development of EBV-associated diseases.In fact, the development of EBV-associated lymphomas has already been described in patients with rheumatoid arthritis and polymyositis, following treatment with drugs that induce EBV lytic cycle (11,59).
Therefore, new strategies aiming at latent EBV are being developed, some of them trying to target the EBV episome, while others trying to aim at the EBV viral products or at cellular products that are EBV associated.Major contributions to the design of new therapeutic strategies arise from the association of laboratory studies with computational analysis.In fact, through docking models, it is possible to choose the best molecules that associate with a specific target and to score this association according to their binding affinities (60).By sorting the best molecules (and groups of molecules) that fit a certain model, it is possible to further modify those molecules to optimize and to improve their binding to the selected target.The aim is to reduce large numbers of compounds to smaller subsets that are more likely to contain biologically active compounds (61).
The results obtained by the in silico screening indicate that some of the molecules from the CEQUIMED-UP library of compounds may act as possible EBV inhibitors.In fact, using three targets (and for some of them, using different regions from their crystallographic structure as targeted region), from the 97 molecules screened, the sulfated  compounds 1-3 were found to possibly target Zta.One of these small molecules, compound 3, was also found to possibly target EBNA1.Regarding BHRF1, none of the investigated molecules showed scores that might indicate this protein as their target, therefore showing a need to expand the search for BHRF1 inhibitors to a larger library of molecules.Based on the docking results, we can postulate that Zta protein is probably the preferred target of these sulfated small molecules, and EBNA1 could be a secondary target for compound 3.
A number of viruses use sites on heparan sulfate (HS) as receptors for binding to cells.Mimics of these sequences are likely to interact with the virus in solution, thereby tying it up and disrupting the viruses interaction with cell surface HS.The HS proteoglycans present in the lymphoid cell surface had already been published as being important in the establishment of interactions with the transcriptional activator EB1/Zta (62,63), which reinforces the selection of these sulfated compounds for further investigation.However, some reports indicate that EBV does not seem to depend on HS for its binding and entry to the host cell.In fact, B-cell lines such as Raji, for which EBV has tropism, are defective for HS expression (64,65).Additionally, sulfated polysaccharides are well known as powerful antiviral drugs (66); nevertheless, sulfated polymeric molecules might only inhibit virus entry and have limited bioavailability.Several studies concerning antiviral activity of sulfated small molecules have been previously described (67)(68)(69).For instance, thalassiolins are flavones inhibitors of HIV integrase, in which the sulfate group was found to be a substituent that imparts increased potency against integrase in biochemical assays (68).Another example is lamellarin a-20-sulfate, an HIV-1 integrase inhibitor with a dual mechanism of antiviral action possibly binding to a site composed of multiple integrase domains and found to inhibit HIV entry (69).Nevertheless, the effect of sulfated small molecules in latent EBV infection still needs to be addressed.As sulfated small molecules have proven their antiviral potential by interaction not only with envelope glycoproteins, but also with critical proteins involved in other stages of the viral cycle, compounds 1-3 were selected over other potential hit ligands obtained in the docking studies.
To validate the results obtained in the computational docking analysis, biological studies were performed using the non-producer BL cell line, Raji.This cell line stably maintains the EBV episome copy number (50-60 copies per cell) during long-term culture and presents a type III EBV latency, therefore expressing all the EBV latent proteins (12,70).However, it may reactivate, that is, switch from latency to lytic infection following different stimulus, and lytic proteins may then be produced.Thus, using Raji cells, the study of the antiviral effect of the three selected compounds was carried out by analyzing their effect in EBV DNA load, in the inhibition of latent viral protein expression, and in the inhibition of viral reactivation induced by TPA.In addition, the fact that these compounds did not present cytotoxicity at the concentrations tested for their antiviral activity is relevant for their possible use as antivirals, as antiviral drugs should be specific and should not affect cell viability.Previous studies have shown the clinical relevance of the monitoring of EBV DNA load for the diagnosis of EBV-associated disorders and for assessment of the efficacy of therapeutic interventions (71).Moreover, following lytic reactivation, a correlation in EBV-infected cells between the levels of lytic proteins and the amount of EBV DNA present has been described (30).By monitoring the EBV viral load following treatment with the three selected small molecules, the objective was to see whether these compounds affected EBV genome copy number, that is, viral DNA load.The fact that all the tested compounds (1-3) led to a decrease in the EBV DNA present in Raji cells was a first indication that these molecules were targeting EBV in its latent state or interfering with episome maintenance.Interestingly, antisensemediated decrease in EBNA1 expression in Raji BL cells was found to concomitantly lead to a reduction in the EBV episome copy number in this cell line (72).Li et al. (73), using a strategy similar to ours, virtually screened compounds using the solved crystal structure of EBNA1 with computational docking programs and identified sulfonamides as new inhibitors of EBNA1 transcription.Moreover, some of these compounds also reduced EBV genome copy number in a BL cell line.In addition, through the development of a high-throughput screen for inhibitors of Epstein-Barr virus EBNA1 using homogeneous fluorescence polarization assay, Thompson et al.( 74) identified three structurally related small molecules that selectively inhibit EBNA1.These three compounds were active in a cell-based assay specific in disrupting EBNA1 transcription repression function, and one was able to reduce EBV genome copy number in Raji cells.Moreover, the use of hydroxyurea (an anticancer drug) to reduce EBV DNA load, although having been successful, has the disadvantage of leading to the accumulation of additional mutations in the cellular genome (18).In this context, the sulfated small molecules revealed herein may serve as a starting point for the development of alternative molecules to hydroxyurea.
Besides affecting EBV DNA load, all the selected compounds 1-3 decreased the levels of LMP1 protein expression (although to different extent).As none of the tested compounds reactivated the virus (did not increase the levels of the lytic EA-D protein), the effects observed resulted from an effect in latent EBV.However, the data obtained so far does not allow drawing conclusions regarding the mechanism of action of these molecules.In fact, although all the investigated (1-3) compounds may have Zta as target, no evidence was found that they reduced viral reactivation induced by TPA.This probably indicates that their mechanism of action may be other than the interference with viral reactivation.In fact, although Zta is known to affect the transcription of several genes and its function as facilitator of viral replication has been thoroughly studied for many years, some new functions for Zta have been published very recently in the literature, namely the ability to stimulate infected resting B cells to proliferate (75).
From the three studied small molecules, compound 3 was the one that presented best results against EBV (observed at the level of decreased expression of LMP1 protein).Interestingly, compound 3 was also the molecule for which the in silico screen results showed that it likely targeted both Zta and EBNA1.Thus, it is possible that this compound 3 targets both proteins (EBNA1 and Zta), therefore being more efficient in its antiviral activity.
Although a decrease in the levels of LMP1 has been observed following treatment with all the compounds, these studies do not allow concluding whether this is a specific effect for this type of molecules.In fact, as a reduction in EBV DNA is observed with the three compounds (1-3), any alterations observed on the levels of latent proteins may be either a cause or an effect of the decrease in DNA.Therefore, further studies are required to clarify how these molecules interfere with EBV infection.Remarkably, the three top small molecules arising from the in silico screening have a sulfate group in their chemical scaffold structure.We could hypothesize an interference of these small molecules (1-3) in the virus entry as described for other sulfated compounds (68,69).Sulfated small molecules are expected to preserve some molecular properties of polymeric compounds but with reduced anionic character, higher hydrophobic nature, and feasible synthesis that could ultimately lead to an orally active drug candidate.
The most promising finding of this work was the discovery of xanthonic/flavone sulfated small molecules with some antiviral activity against EBV.By analyzing the possibilities of these molecules, the xanthonic scaffold provides a good starting point to study new EBV inhibitors.As the exact target for these small molecules cannot be pinpointed among the targets assayed computationally, the best strategy would be to consider the possible interaction with EBNA1 (1b3t and 1vhi) or Zta (2c9l and 2c9n).Regarding EBNA1, compound 3 occupies the same place as DNA.Indeed, its polar groups, the ether groups from xanthonic scaffold, and the sulfur groups from lateral chains occupy the same place as purine and pyrimidine moieties placed in the interior of the double-strand DNA helix.We have also noticed that the DNA-binding pocket is only partially occupied.Therefore, the synthesis of molecules that would extend throughout the DNA-binding pocket, like xanthonic derivatives with larger and bulkier chains, or xanthonic dimers with the same substitution pattern, could be an interesting starting point to develop other potential inhibitors of EBV.In what concerns the Zta protein, the studied xanthonic derivatives occupy also the same binding site as DNA.A xanthonic derivative that would extend throughout chains Y and Z, occupying a larger area, should be taken into consideration.

Conclusions
Although further studies are needed to validate these molecules as antiviral compounds against EBV, this study indicates that the in silico virtual screening of different compounds may be of added value in the identification of new antiviral molecules.Importantly, as these three possible EBV inhibitors studied are sulfated small molecules, this may be a starting point for further synthetic structural modifications and optimization, as well as for additional computer-assisted methods like 2D and 3D similarity searches.In addition, future studies will be carried out using the molecules presented in this study as well as some of their analogues, to further confirm and increase their potency.Furthermore, we intend to broaden this study to other BL cell lines, including BL virus producer cells, and also to other EBV-associated tumor models, such as NPC.Ultimately, this strategy could result in new antiviral agents against EBV.

Table 1 :
Docking scores (kcal/mol) for the three top molecules (1-3) against EBV target proteins: EBNA1 (1b3t and 1vhi), BHRF1(1q59), and Zta(2c9l and 2c9n) Ligand Docking score on receptor (kcal/mol) Epstein-Barr nuclear antigen 1 (EBNA1) binds to four recognition sites in the minimal origin of latent DNA replication of Epstein-Barr virus and activates latentphase replication of the viral genomes.Both 1b3t and 1q59pdb files are composed of two identical chains, X and Y, and the crystallographed residues 459-607 or 470-607 (in 1b3t and 1hiv, respectively) correspond to the DNA-binding and dimerization domains.Epstein-Barr virus nuclear antigen 1 appears to bind DNA via two independent regions termed the core and the flanking DNA-binding domains.The core DNA-binding domain comprises both the dimerization domain and a helix predicted to bind the inner portion of the EBNA1 DNA recognition element.The flanking DNA-binding domain consists in part of an a-helix whose N-terminus contacts the outer regions of the EBNA1 DNA recognition element(55).

Figure 1 :
Figure 1: Surface (A) and ribbon and transparent surface (B) representation of Epstein-Barr virus (EBV) nuclear antigen 1 (EBNA1; pdb: 1b3t) and best ranking poses of compound 3 (several tones of pink).Surface (C) and ribbon and transparent surface (D) representation of Zta protein (pdb: 2c9n) and best ranking poses of compound 1 (purple), 2 (orange), and 3 (pink).Surfaces are colored by element: gray for carbons, blue for oxygen, and yellow for sulfur.

Figure 2 :
Figure 2: (A) Detail of the Epstein-Barr virus (EBV) nuclear antigen 1 (EBNA1; pdb: 1b3t) DNA-binding pocket.Compound 3 is docked in the cavity as an example.Regions P1 to P5 represent polar groups important for the establishment of hydrogen interactions; regions H1 and H2 represent hydrophobic regions.Surface is colored by element (red for oxygen, blue for nitrogen, yellow for sulfur, and gray for carbon).(B) Detail of compound 3 bound to EBNA1 (pdb: 1b3t).Red dashes represent hydrogen interactions, yellow dashes represent other interactions.

Figure 3 :
Figure 3: (A) Detail of Zta (pdb: 2c9n) DNA-binding pocket.Compound 1 is docked in the pocket as an example.Regions P1 to P6 represent polar groups; region H1 represents hydrophobic cavity.Surface is colored by element (red for oxygen, blue for nitrogen, yellow for sulfur, and gray for carbons).(B) Detail of compound 1 bound to Zta (pdb: 2c9n).Red dashes represent hydrogen interactions, yellow dashes represent other non-covalent interactions.

Figure 5 :
Figure 5: Compounds 1-3 reduced Epstein-Barr virus (EBV) DNA levels in Raji cells.DNA was extracted from Raji cells following 96-h treatment with 20 lM of compounds 1-3 or with their vehicle (H 2 O).(A) Densitometry analysis of EBV DNA load following conventional PCR (visualized in agarose gel, data not shown).The intensities of the bands from EBV DNA PCR products were normalized to the intensities of the bands from the respective actin DNA (as internal cellular control).Results, expressed after normalization of the EBV DNA values with the values obtained for cellular actin DNA, were analyzed as percentage of control cells (treated with H 2 O) and are the mean of two independent experiments.(B) Real-time quantitative PCR (qPCR) analysis of EBV DNA load.Results, expressed after normalization of the EBV DNA values with the values obtained for cellular actin DNA, were analyzed as percentage of blank cells and represent the mean AE SE of three independent experiments performed in triplicate (except for compound 2, which is the mean of only two experiments).

Figure 4 :
Figure 4: Compounds 1-3 do not greatly affect Raji viable cell number.Raji cells were treated with 20 lM of compounds 1-3, with their vehicle (H 2 O) or with cell culture medium (blank).The number of viable cells was counted with trypan blue, 48 h after treatment with the molecules.Results were analyzed as a percentage of the blank cell number (considering this as 100%) and are the mean AE SE of three independent experiments.

Figure 6 :
Figure 6: Compounds 1-3 decreased LMP1 viral protein expression in the Raji cell line, but had no effect on viral EA-D expression.Raji cells were treated with 20 lM of compounds 1-3, with their solvent (H 2 O), or with medium only (blank).The expression of latent LMP1 (A) or lytic EA-D (B) proteins was analyzed 48 h after treatment, by Western blot.Actin expression was also analyzed as loading control.Blots are representative of three independent experiments.In (A), the densitometry analysis of the Western blots is shown on the lower panel of the figure.Results are expressed after normalization of the values obtained for the LMP1 with the values obtained for actin and also expressed as percentage of the values obtained for the blank control.Each bar represents the mean AE SE from three independent experiments.

Figure 7 :
Figure 7: Compounds 1-3 had no effect on viral reactivation induced by 12-O-Tetradecanoylphorbol-13-acetate (TPA).Raji cells were treated with the Epstein-Barr virus (EBV) lytic inducer TPA or simultaneously treated with TPA and 20 lM of compounds 1-3.Control treatments included the vehicle (H 2 O) or cell culture medium (blank).The expression of the lytic protein EA-D was analyzed by Western blot.Actin expression was analyzed as loading control.Blot is representative of two independent experiments.

Table 2 :
Multiple active site correction (MASC) scores A B