KSHV transactivator-derived small peptide traps coactivators to attenuate MYC function and kills leukemia and lymphoma

In herpesvirus replicating cells, host cell gene transcription is frequently down-regulated because important transcriptional apparatuses are appropriated by viral transcription factors. Here, we identied a small peptide derived from the Kaposi's sarcoma-associated herpesvirus transactivator (K-Rta) sequence, which attenuates cellular c-MYC expression, reduces cell proliferation, and selectively kills cancer cell lines in both tissue culture and a xenograft tumor mouse model. Mechanistically, the peptide functions as a decoy to block the recruitment of coactivator complexes consisting of Nuclear receptor coactivator 2 (NCOA2), p300, and SWI/SNF proteins to the MYC promoter in primary effusion lymphoma cells. Thiol(SH)-linked alkylation for the metabolic sequencing of RNA ( SLAM seq) with target-transcriptional analyses further conrmed that the viral peptide directly attenuates MYC and MYC-target gene expression. This study thus provides a unique tool to control MYC activation, which may have signicant potential as a therapeutic payload to treat MYC-dependent diseases such as cancers and autoimmune diseases. Cell cell The Flag-HA-tagged-K-Rta-inducible, TREx-K-Rta BCBL-1 cell line was generated according to methods previously described (29). The BC-1, BC-3, Ramos, HH, Jurkat, THP-1, U937, A549, and SU-DHL-10 cell lines were obtained from ATCC, expanded, and aliquots stored in order to maintain an inventory of early passage stocks. These cell lines were cultured in RPMI 1640 medium supplemented with 15% FBS, antibiotics, and L-glutamine, or DMEM medium supplemented with 10% FBS, antibiotics, and L-glutamine for A549 cell lines. In vitro cell treatment with VGN50. Leukoreduction system chambers (LRS) from healthy donors were purchased from Vitalant. Peripheral blood mononuclear cells (PBMCs) were prepared by a standard Ficoll gradient method. PBMCs along with THP-1 and BCBL-1 cell lines (2 x 10^6 cells/ml, 0.2 ml 10% FBS/RPMI/DMEM per well of a 96-well plate) were treated with various doses (0-256 mM) of either VGN50 peptide or control mutant peptide for 24 hours. Cell death, apoptosis, and proliferation were analyzed by ow cytometry (BD Acuri) after intracellular staining of Ki67-Alexa488, annexin-V staining, live/dead staining (Invitrogen), or MTT assay. Peptides were commercially synthesized (GenScript) with >90% purity with TFA removal. Peptides were dissolved in PBS and used at various concentration indicated in gure legends.


Introduction
Expression of c-Myc (MYC) is tightly regulated in normal cells but becomes dysregulated and often overexpressed in many types of human cancer cells. Numerous studies suggest that deregulation of MYC expression occurs in more than 70% of cancers overall and contributes to disease progression, metastatic potential, and therapeutic resistance (1). Mechanistically, increased MYC protein concentration in the nucleus facilitates its interaction with E-box transcription factors, which triggers tumorigenesis by promoting cell proliferation; in contrast, these events should not occur in resting cells with tightlyregulated, normal physiological levels of MYC. The ubiquitous nature of MYC deregulation in most types of cancers and its inherent driver function for cell proliferation make MYC a very attractive target for cancer drug development. However, MYC, like other transcription factors, possesses a highly-disordered structure, which facilitates interaction with multiple cofactors and DNA (2,3), but hampers development of speci c small molecule inhibitors. Nonetheless, several MYC-speci c inhibitors have been developed successfully. Omomyc is a MYC-speci c inhibitor, which is a miniprotein 90 amino acids in length derived from the MYC basic helix-loop-helix (bHLH) domain that acts as a dominant negative mutant for the MYC dimerization domain (4). Four mutations in the leucine zipper region still permit Omomyc to dimerize with all MYC family proteins, but effectively prevent the MYC/MAX heterodimers from binding their target promoters thereby inhibiting MYC-dependent gene expression (5). As expected, Omomyc inhibits the cell growth of multiple cancer cell types in vitro and in vivo (5).
In addition to directly targeting MYC, the epigenetic silencing of MYC gene expression by targeting histone modifying enzymes or binding proteins also reduces MYC expression. Epigenetic targets include histone deacetylase, acetylase, demethylase, methylase, and bromodomain and extra-terminal motif (BET) proteins. Small molecule inhibitors targeting the respective enzyme pockets or acetylated histone binding surfaces have all shown some e cacies against MYC, with BET inhibitors being the most wellstudied (6)(7)(8)(9)(10)(11). For example, the small molecule inhibitor JQ1 was one of the rst BET bromodomain characterization of the peptide-mediated gene repression and an example of its therapeutic potential are described in this report.

Results
Identi cation of the KSHV transactivation complex.
We previously reported that when KSHV reactivation begins, RNA Polymerase II (RNAPII) molecules are effectively recruited to viral episomes and form a complex with a viral protein, KSHV replication and transactivation (K-Rta), for viral gene expression (36). We examined cellular proteins that are localized in the RNAPII and K-Rta protein complex during KSHV reactivation by utilizing rapid immunoprecipitation mass spectrometry of endogenous protein (RIME) (40). This method is well-suited to comprehensively identify transcription factor complexes on chromatin. We examined interacting protein partners of K-Rta and RNAPII by immunoprecipitation with anti-Flag antibody for K-Rta and anti-RNAPII antibody, respectively, before (K-Rta is absent) and after (K-Rta is present) triggering reactivation (Fig. 1a). Proteins co-precipitated with RNAPII that were further increased in the presence of K-Rta, and also co-precipitated with K-Rta (anti-Flag antibody) were considered to be recruited by K-Rta for viral gene transactivation. Non-reactivated samples served as a negative control for K-Rta RIME experiments. In addition, RIME conducted with non-speci c IgG was used to control for non-speci c interactions for both uninduced and induced samples. After stringent ltering with criteria described in the methods section, RIME identi ed a total of 87 K-Rta interacting proteins (Supplementary Table 1); these include several protein complexes functioning in RNA processing, RNA splicing, chromosome organization, DNA repair, and DNA replication. Furthermore, RNAPII RIME showed that 85 of the K-Rta interacting proteins were among the set of RNAPII interacting proteins (Supplementary Table 1). Fifteen cellular proteins were identi ed that newly interacting with RNAPII during KSHV reactivation with >1.5-fold enrichment with p value less than 0.05 (Fig. 1b). In contrast, proteins that dissociated from RNAPII upon K-Rta induction are known to function in DNA replication and DNA strand elongation (FDR=5.12E-12) ( Supplementary Fig. 1). The 15 RNAPII interacting proteins for which the interaction is induced by K-Rta expression include 8 SWI/SNF components (41), three mediator complex proteins, p300 histone acetylase, nuclear receptor coactivator 2 (NCoA2), and a protein that functions in nuclear export. Analysis with STRING, a bioinformatics tool to analyze protein-protein interaction networks, demonstrated these activator complexes could interact as a single large protein complex (Fig. 1b). Overlapping or shared protein interactions with K-Rta also suggested that there is a pre-assembled large coactivator complex. Co-immunoprecipitation with anti-Flag (for K-Rta) antibody showed that interaction between K-Rta and NCoA2 were increased during KSHV reactivation along with an increasing amount of K-Rta protein expression, which con rmed the RIME studies (Fig. 1c). In addition, our RIME studies agreed well with previous reports, which demonstrated that the SWI/SNF complex, mediator complex, and NCoA2 are K-Rta interacting proteins, and that the interaction plays a signi cant role in K-Rta's transactivation function in KSHV-infected cells (37,38).
Functional annotation of KSHV reactivation-induced K-Rta and RNAPII-interacting proteins predicted that these proteins are primarily involved in chromatin disassembly to trigger transcription elongation [false discovery rate (FDR)=1.00E-14 ( Fig. 1d)].
KSHV K-Rta interacts with coactivators at its transactivation domain.
The KSHV reactivation-induced interaction of coactivators with both RNAPII and K-Rta suggested that K-Rta might physically bring the complex to the RNAPII complex poised on the KSHV genome (42), thereby stimulating KSHV lytic gene expression. To test this hypothesis, we next examined direct-physical interaction between K-Rta and the coactivator complex and mapped the interaction domain(s) in detail. Previous studies with GST-K-Rta fusion proteins showed that K-Rta precipitated SWI/SNF proteins and the mediator complex from cell lysates, and amino acid residues 612-621 of the K-Rta transactivation domain (Fig. 1e) were responsible for the precipitation of the coactivators from cell lysates (38). To rule out indirect interactions, which may result from crude cell lysates, we puri ed components of the SWI/SNF complex from recombinant baculovirus-infected Sf9 cells. The K-Rta transactivation domain (Fig. 1e) and its mutant were expressed as GST-fusion proteins and puri ed from E.coli (Mut1, 2, and 3 in Fig. 1f, g). The GST-pull down studies using wild type (WT) K-Rta domain showed that the SWI/SNF complex containing SMARCA4, SMARCC2, and SMARCB1, directly interacted with K-Rta between residues 551 and 650. Alanine substitutions were introduced to generate three mutant K-Rta proteins referred to as Mut1, 2, and 3 (Fig. 1f), and we determined that residues 619-621 (ILQ), which are mutated in Mut2 and 3, were critical for the interaction (Fig. 1h). On the other hand, mutation of residues 612-613 (DD), which are mutated in Mut1, did not signi cantly impair the interaction with the puri ed proteins (Fig. 1h). Interestingly, additional mutation of negatively charged amino acids (DD) to hydrophobic amino acids (AA) in Mut3 slightly restored interaction of the Mut2 protein with SMARCB1 (Fig. 1h). Mut2 has a mutation reduces hydrophobicity due to alanine substitution (i.e., ILQ > AAA). The result may indicate that the hydrophobicity of this sequence stretch plays an important role in the interaction between K-Rta and SMARCB1. Computer modeling with IUPred2 (43,44) showed that the interaction domain is located within an intrinsically disordered region (IDR) of K-Rta (Fig. 1e, bottom panel), and IDRs are known to be involved with transcriptional activation through exible interactions with both nucleotides and proteins (45). The results suggest that K-Rta utilizes the IDR to directly interact with the large coactivator complex and recruit the complex to K-Rta DNA binding sites for transactivation during lytic replication.
K-Rta peptide inhibits leukemia and lymphoma cell growth in vitro.
To further de ne a protein sequence within K-Rta responsible for coactivator complex formation, homologous protein sequences were extracted from other gamma-herpesvirus homologs and a bacterial transcription factor with BLAST analysis. The consensus protein sequence was then depicted with the hypothesis that these conserved amino acid residues are important for biological functions (Fig. 2a). Based on the conserved protein sequence, we next synthesized a series of K-Rta peptides, which are tagged with the cell penetrating 10-amino acid transactivator of transcription (TAT) sequence for intracellular delivery (46). To protect from rapid peptide degradation by cellular proteases, we also used an unnatural D-amino acid at the N-terminus in the peptide synthesis (Fig. 2b). Cell penetration of the peptide was con rmed by using the uorescently labelled peptide Pep 1 (Fig. 2c). The results showed that the peptide was e ciently translocated into the cell nucleus within 15 min of incubation.
We next examined the effect of the peptides on cell viability and viral replication. The hypothesis was that the introduced peptide would compete with the coactivator complex (e.g. SWI/SNF complex) as a decoy to inhibit cellular and viral gene transcription. Different concentrations of peptides were incubated with BC-1 cells, a KSHV latently-infected PEL cell line, and cell viability was examined with MTT assays. The results showed that the peptides Pep 1 and Pep 3 reduced cell viability by 90% at 16 µM concentration compared to the untreated control, while a mutant peptide containing a triple alanine substitution (Mt-P) or a peptide with deletion of leucine at position 616 of K-Rta (Pep 2) only showed 20% reduction at 64 mM (Fig. 2d). Also, Pep 4 lacking the C-terminal 4 amino acids including leucine at position 625 of K-Rta was approximately 4-fold less effective compared to Pep 1 and 3 (Fig. 2d). Notably, some substitutions of non-conserved amino acids (618S to T, 628T to E or S) did not affect the peptide's cell killing ability ( Supplementary Fig. 2). In addition, including two additional D-amino acids at the Cterminus, which has been shown to increase peptide stability (47), did not enhance cell killing ability ( Supplementary Fig. 2). Based on these results, we decided to use peptide 1 (Pep 1) and its mutant (Mt-P) for the further experiments.
We next used two PEL cell lines and ow cytometric analyses to con rm that reduced cell viability was due to down-modulated cell proliferation and apoptosis induction, which was seen as early as 6 hrs post incubation (Fig. 2e). Furthermore, we found that cell killing by Pep 1 was not limited to KSHV-latently infected PEL cells and was also seen in KSHV-negative myeloid and lymphoid cancer cell lines. However, the A549 adenocarcinomic human alveolar basal epithelial cell line, was 4-fold resistant to cell killing by Pep 1 (Fig. 2f). Thus, the cell killing activity is independent of KSHV latent infection. Importantly, PBMCs were approximately 10-fold less sensitive to Pep 1-mediated killing compared to myeloid and lymphoid cancer cell lines as assessed by ow cytometry after live/dead staining (Fig. 2g). Based on the oncolytic activity, we renamed the K-Rta Peptide 1 sequence "Virus de Gann wo Naosu ORF50" (VGN50), which means "curing cancers with viral protein(s)" in Japanese.
VGN50 binds SWI/SNF proteins and mimics K-Rta function to induce KSHV lytic gene expression.
Because VGN50 is derived from the KSHV transactivator sequence, we next examined if VGN50 can inhibit KSHV lytic replication by competitively blocking the coactivator binding to K-Rta. To study this, we used the TREx-K-Rta BCBL-1 cell line, in which we can trigger the expression of exogenous Flagtagged K-Rta in a doxycycline (Dox) inducible manner. TREx-K-Rta BCBL-1 cells were treated with 12 mM VGN50 or mutant peptide, with or without Dox induction. We used approximately one-half of the IC50 concentration of VGN50 to treat the TREx-K-Rta BCBL-1 cells in order to avoid cell toxicity effects that would non-speci cally decrease viral gene expression. KSHV gene expression was pro led with a KSHV genome-wide PCR array, using ribosomal rRNA for normalization and as an internal control. Fold activation relative to the uninduced Mut-P-treated sample (not induced: Dox -) was determined and a heatmap for KSHV lytic gene expression was generated (Fig. 3a). The results showed that, in both Mut-Pand VGN50-treated cells, KSHV lytic gene expression was increased with Dox incubation. VGN50-treated cells with Dox showed the highest expression, indicating that contrary to our expectation, VGN50 enhanced the KSHV lytic program. Immunoblotting was used to con rm the induction of lytic replication by Dox based on exogenous K-Rta protein (Flag-K-Rta) and the expression of two other KSHV lytic proteins (ORF57 and K-bZIP) along with actin and KSHV LANA protein as controls for equal loading and similar viral copy numbers in samples, respectively (Fig. 3b). Notably, VGN50 alone weakly triggered PAN RNA expression, an indication of KSHV reactivation, in KSHV-latently infected BCBL-1 and BC-1 cells (Fig.   2c), which is consistent with enhanced lytic gene expression seen in K-Rta inducible cells (Fig. 2a). Taken together, these results indicate that VGN50 is a biologically active molecule that mimics K-Rta function but fails to block K-Rta function under the current conditions.
To con rm the basis of the VGN50 action, the biochemical interactions between VGN50 and SWI/SNF proteins (i.e., putative VGN50 target molecules), were examined by ELISA-based binding assays using puri ed 5 individual SWI/SNF components prepared from baculovirus-infected Sf9 cells (Fig. 3d). Increasing concentrations of biotin-conjugated VGN50 or Mut-P were incubated in an ELISA plate coated with each SWI/SNF component, and the bound peptides were detected by HRP-streptavidin (Fig. 3e). A high concentration (1 mg/mL) of bovine serum albumin was used for blocking and in every incubation step to reduce non-speci c interactions. The results showed that VGN50 bound to the 5 SWI/SNF components at a concentration as low as 50 nM at different e cacies (Fig. 3f). VGN50 had slightly better binding a nity to SMARCA4, SMARCB1, and SMARCE1, whereas the Mut-P control peptide did not bind SWI/SNF components as much as VGN50 at the same peptide concentrations (Fig. 3f). These results suggest that VGN50, a prototype of the K-Rta IDR, can exibly interact with components of the SWI/SNF complex, and that the highly conserved ILQ sequence is required for binding.
As a next step, we sought to identify the cellular pathways that are impacted by VGN50. For this, RNAsequencing (RNA-seq) analysis was conducted on total RNA isolated from cells harvested after 24 hours of incubation with VGN50. Mutant peptide was used as a comparison to identify differentially regulated genes by VGN50 (Fig.4a). Consistent with the fact that VGN50 is derived from K-Rta, and that one of the recognized K-Rta functions is to counteract host interferon responses, gene set enrichment analyses (GSEA) showed that both a-and g-interferon pathways associated gene sets were signi cantly downregulated by the peptide incubation with zero false discovery rate (FDR) (Fig. 4b). GSEA also showed that MYC target genes, which include enzymes associated with DNA replication, were among the signi cantly down-regulated gene sets (Fig. 4a, b). Furthermore, we con rmed the result by RT-qPCR that the expression of MYC itself was decreased in both BCBL-1 and BC-1 cells after treatment with VGN50, but not mutant peptide (Fig. 4c).
In order to reveal direct targets of VGN50, we next employed the thiol(SH)-linked alkylation for the metabolic sequencing of RNA (SLAM seq) method (48). SLAM-seq is an orthogonal-chemistry-based RNA-Seq technique that detects 4-thiouridine (s4U) incorporation in RNA species at single-nucleotide resolution. We separated newly synthesized RNAs that are transcribed in the presence of s4U in tissue culture media from existing RNAs based on sequence reads containing T>C conversions as a result of the alkylated s4U's. We also replicated the experimental conditions used in comprehensive target pro ling studies of the BET bromodomain inhibitor with SLAM-seq (9) and collected samples at the same time points (Fig. 4d), which allowed us to compare target gene pro les between VGN50 and the BET bromodomain inhibitor, JQ1. The BET bromodomain inhibitor is an epigenetic drug, which is also known to down-regulate MYC expression (49). Consistent with the total RNA-seq results, MYC exhibited the most statistically signi cant down-regulation among 25,000+ cellular genes in BCBL-1 with a logFC of -2.15 (FDR-adjusted p-value of 9.51 E-12). Similarly, BC-1 cells also displayed a log2FC -1.643 (FDR-adjusted pvalue 5.78E-12) in MYC expression (Supplementary Table 2 and 3). The Integrative Genomics Viewer (IGV) snapshot of the MYC 3'-UTR region showed that treatment with VGN50 down-regulated active MYC transcription (i.e., nascent transcripts) in BC-1, while pre-existing MYC mRNA species (i.e., non T->C) showed little changes (Fig. 4e). A primary effect of VGN50 is clearly gene down-regulation, as evidenced by the majority of statistically signi cant gene expression changes (p<0.05, red dots) with negative fold changes (Fig. 4d). The results are in good agreement with a previous study, which demonstrates that the KSHV transactivator takes the endogenous host cellular transcription machinery away for KSHV replication (36). GSEA of the VGN50 direct targets were also performed, and the analyses found that MYC target genes, including MYC itself, were clearly enriched in BC-1 cells with a normalized enrichment score (NES) of 2.90 (Fig. 4f). Several cellular genes were commonly up-regulated between BC-1 and BCBL-1, and those genes were associated with in ammatory responses and immediate-early genes (e.g. JUND, FOS, ATF, Supplementary Table 4). As expected, negligible gene expression was altered by incubation with the mutant peptide in both BCBL-1 and BC-1 lines, attesting to the peptide-sequence speci c regulation ( Fig. 4d left panel). Preferential targeting of MYC by VGN50 also prompted us to compare target genes with the BET bromodomain inhibitor. We previously demonstrated that the BET bromodomain inhibitor JQ1 could induce KSHV reactivation similarly to that of VGN50 (Fig. 3a, c), and also down-regulated MYC and MYCtarget genes in KSHV-positive PEL cell lines (50). Publicly available SLAM-seq data sets for K562 (chronic myelogenous leukemia cells) were used to compare direct target gene sets with VGN50-treated PEL cell lines. The top 100 down-regulated genes were extracted and examined for similarity among downregulated genes. The results showed that very few cellular genes were commonly down-regulated between the BET bromodomain inhibitor JQ1 and VGN50, and target gene sets were mostly dependent on cell lines. However, one of the three common target genes was MYC (Fig. 4g).
Finally, using CSCAN (51), we examined enrichment for transcription factors that are bound to promoters down-regulated by VGN50. CSCAN is a ChIP-seq database to identify common regulators that recognize selected gene promoters (51). For this CSCAN analysis, we used the database for GM12878, a lymphoblastoid cell line, which we considered to be the closest to PEL cells among the available datasets. The results again showed that MYC and also IRF4, an upstream regulator of MYC in PEL cells (52,53), are enriched in VGN50 targeted promoters. The results suggested that VGN50, a peptidyl mimic of K-Rta, may appropriate the speci c coactivator complex, which is also utilized by c-MYC and IRF4 in PEL cells. Targeted transcription factors (TFs) were signi cantly overlapped between BC-1 and BCBL-1 (Fig. 4h), even though only 16 genes were shared amongst the top 100 down-regulated genes in common between BC-1 and BCBL-1 (Fig. 4g). In addition, TFs that are known to localize at "superenhancers" such as MYC, IRF4, MEF2A, and STAT5A are commonly targeted by VGN50 in BCBL-1 and BC-1 cells (Fig. 4h, the complete list is presented in Supplementary Fig. 3). Notably, two of three commonly downregulated targets, MYC and CCND2 (Fig. 4g), were recently found as super-enhancer-regulated genes in PEL cell lines (54).
VGN50 inhibits coactivator complex recruitment to MYC promoter and putative enhancer regions.
We next examined the molecular mechanism of MYC down-regulation by VGN50. We rst utilized CUT&RUN-sequencing to examine histone modi cations that are known to be associated with enhancers in order to map putative MYC enhancer regions in PEL cells (Fig. 5a). Genome-wide CUT&RUN studies identi ed putative enhancer and promoter regulatory regions that are marked by H3K27Ac, H3K4me1, POLR2A (RNAPII), BRD4, and SMARCC1, a component of the SWI/SNF complex (Fig. 1b, and 5a). The identi ed putative MYC enhancers were also in good agreement with previous ChIP Hi-C (54) and ChIPseq data sets (55). Speci c primer pairs were then designed to amplify the respective enhancer regions and to examine regulation of coactivator complex recruitment by VGN50. The results demonstrated that the occupancies of RNAPII, BRD4, SWI/SNF complex, along with H3K27Ac modi cation were decreased in the presence of VGN50 compared with Mut-peptide (Fig. 5b, c).
Since VGN50 directly binds SWI/SNF components (Fig. 3f), we hypothesize that VGN50 traps SWI/SNF components and blocks functional enhancer complex assembly. To test the hypothesis, in vitro SWI/SNF complex formation was further examined by sucrose gradient with or without VGN50. The results showed that the VGN50 peptide incubation induced the assembly of larger SWI/SNF complexes and moved the complex to heavier bottom fractions in the gradient (Fig. 5d). Immuno uorescence studies with KSHV reactivating BCBL-1 cells also showed that SMARCC1 and SMARCA4 signal intensity were increased at the KSHV transcription/replication compartment and colocalized with K-Rta protein in reactivating cells (Fig. 5e). Based on these results, we suggest a model, in which VGN50 acts as a decoy to trap the coactivator complex and block their binding to cellular enhancers, thereby preventing RNAPII to re-bind highly-inducible genomic regions (Fig. 5f). In the case of PEL cell lines, for which growth is highly dependent on the MYC and IRF4 pathways, the major targets for VGN50 were c-Myc and IRF4 (Fig. 4f-h).

VGN50 inhibits tumor growth in a PEL xenograft model.
Based on strong and selective MYC down-regulation and cell killing activity of cancer cells in vitro, we next studied the therapeutic potential of VGN50 by examining the e cacy of cancer cell killing over toxicities to mice in a PEL xenograft model with BCBL-1 cells. Pilot studies determined that the peptide was well-tolerated in immunocompetent mice when administered as 10 mg/kg peptide via intraperitoneal (i.p.) injection with a 5 day-on 2 day off schedule for two weeks, as shown by no signi cant changes in body weight, complete blood counts, and serum biochemical properties ( Supplementary Fig. 4). Based on the drug schedule, we examined the effects of the peptide on xenograft PEL cell growth. In this established model, BCBL-1 cells grow in the peritoneal cavity resulting in accumulation of ascites, which can be measured as a body weight gain that corresponds to the volume of ascites (56). We inoculated male NRG mice with 2 x 10 7 BCBL-1 cells via i.p. injection followed by i.p. injections of VGN50 (10 mg/kg), mutant control peptide or PBS, daily for 10 days starting from two days after (day 0) tumor inoculation. All mice in the control PBS-treated groups reached the humane end point by 10 days (Fig.  6a). However, in mice that received the VGN50 treatment, tumor growth was inhibited as evidenced by the absence of weight gain due to accumulation of ascites uid containing tumors during the course of therapy on days 0-10 compared with the PBS or mutant control peptide-treated groups (p< 0.05) (Fig. 6a). At the termination point, tumor growth was also validated by ow cytometry analysis of PEL cells in ascites uid (Fig. 6b, c). Notably, cell size (FCS-A) and scatter (SSC-A) parameter analysis revealed that BCBL-1 cells from ascites of mice treated with VGN50 showed signi cantly reduced cell size and scatter geometric mean uorescent intensity (gMFI) compared to the control groups (Fig. 6c). The results suggest that the peptide caused cell atrophy. A similar phenotype is reported in MYC knock-out cardiac myocytes and keratinocytes (57,58).
Since PELs are frequently involved in effusions in multiple body cavities in human patients and the amounts of in ammatory cytokines in the effusion are strong prognostic factors (59), we also extracted ascites uid and pro led it to see if the PEL cell-derived human cytokine pro le was also altered by continued VGN50 treatment. Principal component analysis of the expression of 15 in ammatory proteins that are detected in ascites uids among 92 total proteins (Supplementary Table 5) screened showed signi cant differences between the treatment groups and that clearly separated VGN50-treated cells from those of control groups (i.e. PBS-or mutant peptide-treated) (Fig. 6d). Among 15 cytokines signi cantly differentially regulated, a large fraction of cytokine secretion, except for a few including IL-8, was downregulated in VGN50-treated tumors (Fig. 6e, Supplementary Fig. 5). The cytokine pro le was also in good agreement with the RNA-seq results from PEL cells in xenograft mice, which demonstrated that genes involved in TNFa signaling with the NF-kB pathway were the most down-regulated gene sets in PEL xenograft cells treated with the peptide (NES 1.74, p<0.01) (Supplementary Fig. 6). Although it was not as clear as in vitro tissue culture studies, MYC target genes were again among the enriched down-regulated gene sets (Supplementary Fig. 6). Taken together, these results suggest that the VGN50 should have therapeutic value to control PEL pathogenicity.

Discussion
We are very excited to report the identi cation of VGN50 derived from a viral protein sequence, which has signi cant potential as a therapeutic module to regulate MYC expression. The peptide sequence is derived from the intrinsically disordered domain of the extremely potent viral transactivator protein, K-Rta (29,60,61). Even though the transactivation domain is predicted to be highly disordered, the 13amino acid sequence stretch is well-conserved among other gamma-herpesviruses and a bacterial transcription factor, suggesting that the sequence is intended and well-designed to be "non-structured" evolutionarily for transactivation function (Fig. 2a, b).
Previous studies demonstrated that K-Rta inhibits IFN-mediated cellular responses by degrading IRF7 (62,63). Here we show that introduction of a small fragment of the K-Rta transactivation domain attenuated IFN target gene transcription (Fig. 4b), suggesting that K-Rta regulates IFN activation both transcriptionally and post-translationally. Enrichment of IFNg target genes also suggests an interesting possibility that K-Rta may take advantage of host IFN responses for its own gene expression by hijacking actively-assembling transcription coactivators that are specialized for transient and robust IFN target gene expression. We wonder if primarily targeting IFN pathways by K-Rta is the reason for VGN50 being more e cacious against myeloid and lymphoid cells. In these myeloid and lymphoid cell types, regulation of cell growth is more directly associated with host immune responses for replication.
Recent studies showed that another KSHV protein, vIRF3 activates IRF4 "super-enhancer" and activates downstream targets of MYC expression during latency (53), while the lytic KSHV protein, vIRF4, inhibits cellular IRF4 binding to the MYC enhancers (55,64), therefore, inhibiting MYC expression during lytic replication. It is known that KSHV reactivation has a "Yin and Yang" relationship with cellular MYC; MYC expression inhibits KSHV reactivation while knock-down of MYC or IRF4 reactivates KSHV in latently infected PEL cells [ Fig. 3c and (65, 66)]. A similar relationship was also reported recently for Epstein-Barr virus infected B-cells (67). The authors elegantly performed genome-wide CRIPR-Cas9 screening and identi ed that MYC and MYC-activating proteins are necessary for maintaining EBV latency in Bcells (67). These studies indicated that latent gamma-herpesviruses sense the amount of MYC in the host cell nucleus for reactivation. Our study hypothesizes that it is, in part, due to competition for the NCoA2-SWI/SNF chromatin remodeling complex. We hypothesize that overexpression of MYC associates with a limited supply of co-activator enzymes that are needed for K-Rta to activate viral promoters, while a decrease in the total amount of MYC may increase available resources and chances for K-Rta to assemble coactivator complex on KSHV gene promoters. Interestingly, incubation of cells with VGN50 at one-half the IC50 dose did not inhibit KSHV lytic gene expression, but instead further increased the overall KSHV transcripts. We speculate that such concentration of VGN50 may still inhibit coactivator binding to cellular transcription factors (i.e. c-MYC, IRF4), but not KSHV K-Rta. K-Rta is known to form oligomer (68) and may require higher concentration of VGN50 in order to successfully compete with K-Rta protein.
MYC expression is tightly regulated in normal cells, and tumor cells are often addicted to MYC expression for their growth due to a signi cant demand for continued generation of metabolites, thus being more sensitive to MYC inhibition (Fig. 2e, f). Transgenic expression of Omomyc, a dominant-interfering MYC mutant, indeed showed remarkably low toxicity to mice despite well-documented effects of systemic Myc inhibition on normal regenerating tissues (69). Consistent with this, we observed that the EC50 of VGN50 for PBMCs from healthy donors was approximately 10-fold higher than that for cancer cell lines (Fig. 2g), and VGN50 administration to immune competent mice at 10 mg/kg with a 5 day-on 2 day off schedule for two weeks was well-tolerated without signi cant loss of weight ( Supplementary Fig. 4).
Human SWI/SNF complexes are remarkably large, including the products of the 29 genes encoding mSWI/SNF subunits that assemble into three distinct mSWI/SNF complexes, termed canonical BRG1/BRM associated factor (cBAF), polybromo-associated BAF (PBAF), and noncanonical BAF (ncBAF), each of which comprises common as well as complex-speci c subunits (70). Recent studies have revealed a high prevalence of mutations in genes encoding subunits of the SWI/SNF complexes in cancers and neurological diseases (70,71), indicating the importance of this complex as a therapeutic target. In this context, our current study showed that the oncogenic KSHV utilizes the SWI/SNF complex for the viral genome replication through the interaction with the viral transactivator protein K-Rta.
Moreover, we showed that VGN50, a K-Rta-derived peptide, can be used to prevent the SWI/SNF complex from forming the MYC transactivation machinery in cancer cells. The limitation of the current study is that we could not pinpoint the speci c molecular target and structural mechanism through which VGN50 interacts with this large SWI/SNF complex. By ELISA-based assays, we attempted to show that VGN50, but not a control mutant peptide which lacks the highly conserved ILQ sequence motif, biochemically binds to several isolated shared subunits of SWI/SNF complexes. However, further studies are required using other techniques and approaches to fully understand the structural mechanism in which this small viral-derived peptide can exibly interact with several SWI/SNF components and capture subunits of SWI/SNF, thus preventing it from assembling a transactivation complex on the MYC promoter. The latter is especially important to better understanding SWI/SNF biology and to identify new approaches for targeting SWI/SNF complexes in cancer therapy.
In summary, our study demonstrated that a viral protein is a unique starting material as the basis for designing therapeutics directed at attenuating cellular protein function(s). We hypothesize that viral proteins continuously evolved structure possessing enhanced e ciencies to hijack cellular protein function, and resulting in the current conserved amino acid sequence. Future studies will likely increase utilities of the VGN50 peptide by targeting speci c cell types as an antibody dependent conjugate or a ligand fusion protein, which may enhance the e cacy of existing antibody drugs. We sincerely hope that VGN50 can contribute to efforts to attenuate the progression of diseases, in which MYC activation plays a pathogenic role.

Materials And Methods
Cell culture. The BCBL-1 cell line was obtained from Dr. Ganem (University of California San Francisco). The Flag-HA-tagged-K-Rta-inducible, TREx-K-Rta BCBL-1 cell line was generated according to methods previously described (29). The BC-1, BC-3, Ramos, HH, Jurkat, THP-1, U937, A549, and SU-DHL-10 cell lines were obtained from ATCC, expanded, and aliquots stored in order to maintain an inventory of early passage stocks. These cell lines were cultured in RPMI 1640 medium supplemented with 15% FBS, antibiotics, and L-glutamine, or DMEM medium supplemented with 10% FBS, antibiotics, and L-glutamine for A549 cell lines.
In vitro cell treatment with VGN50. Leukoreduction system chambers (LRS) from healthy donors were purchased from Vitalant. Peripheral blood mononuclear cells (PBMCs) were prepared by a standard Ficoll gradient method. PBMCs along with THP-1 and BCBL-1 cell lines (2 x 10^6 cells/ml, 0.2 ml 10% FBS/RPMI/DMEM per well of a 96-well plate) were treated with various doses (0-256 mM) of either VGN50 peptide or control mutant peptide for 24 hours. Cell death, apoptosis, and proliferation were analyzed by ow cytometry (BD Acuri) after intracellular staining of Ki67-Alexa488, annexin-V staining, live/dead staining (Invitrogen), or MTT assay. Peptides were commercially synthesized (GenScript) with >90% purity with TFA removal. Peptides were dissolved in PBS and used at various concentration indicated in gure legends.
Cytokine measurement. PEL-derived cytokine pro les in ascites were measured with Olink analysis service with the Olink INFLAMMATION panel using proximity extension technology, a high-throughput multiplex proteomic immunoassay (72). In short, the panel includes 92 immune-related proteins, primarily cytokines and chemokines. The assay utilizes epitope-speci c binding and hybridization of a set of paired oligonucleotide antibody probes, which is subsequently ampli ed using quantitative PCR, resulting in log base 2-normalized protein expression (NPX) values. The data was processed and analyzed with Olink Insight Stat Analysis software.
Rapid immunoprecipitation mass spectrometry of endogenous protein (RIME). TREx-Flag-tagx3-HA-tagx3 (F3H3)-K-Rta BCBL-1 cells were left untreated or KSHV reactivation was induced by adding doxycycline (1 mg/mL) and TPA (20 ng/mL) in the culture media for 28 hours. Cells were xed according to the company's recommendation (Active Motif) by adding one tenth volume of formaldehyde solution into the culture medium [11% methanol-free formaldehyde (ThermoFisher), 100 mM NaCl, 1mM EDTA, 50 mM Hepes (pH 7.9)] and incubation for 8 min. The reaction was stopped by adding glycine to the culture medium to a 100 mM nal concentration for 5 min. Cells were washed three times with cold 0.5% NP-40/PBS and cell pellets were snap-frozen on dry ice. Samples were then shipped to Active Motif (Carlsbad, CA) for their Interactome Pro ling Service. Antibodies used for RIME assays were anti-Flag M2 for K-Rta (Sigma), anti-RNAPII (4H8, Millipore Sigma), and control IgG provided by the company. Each immunoprecipitation was performed in duplicate. The following criteria were used to designate speci c interactions: (i) more than one peptide count in each replicate sample, (ii) more than 5 total peptide counts by combining two replicates, (iii) more than 5-fold enrichment over background noise (IgG). (iv) In addition, scaffold software was used to determine the statistical signi cance of the interactions and a pvalue less than 0.05 was considered signi cant. Filtered data sets from RIME with RNAPII and K-Rta were then combined and increased or decreased interactions with commonly interacting proteins between RNAPII and K-Rta by KSHV reactivation was determined using 1.5-fold as a cutoff (S- Table 1). A total 87 proteins were identi ed as K-Rta interacting proteins of which 85 were also identi ed as RNAPII interacting proteins. Non-processed raw data sets are provided upon request. mM EGTA, 50 μg/ml RNase A, 50 mg/ml glycogen). The beads were incubated with shaking at 37°C for 10 min in a tube shaker at 500 rpm to release digested DNA fragments from the insoluble nuclear chromatin. The supernatant was collected after centrifugation (16,000 x g for 5 min at 4°C) and placed on a magnetic stand. DNA was extracted using the NucleoSpin Gel & PCR kit (Takara Bio, Kusatsu, Shiga, Japan). Sequencing libraries were then prepared from 3 ng DNA with the Kapa HyperPrep Kit (Roche) according to the manufacturer's standard protocol. Libraries were multiplex sequenced (2 x 150bp, pairedend) on an Illumina HiSeq 4000 sequencing system to yield ~15 million mapped reads per sample. With separate replicated experiments, qPCR was used to examine enrichment at selected genomic regions. Primer sequences are provided in Supplementary Table 6. CUT&RUN sequence reads were aligned to the human GRCh38/hg38 reference genome assembly and reference KSHV genome sequence (Human herpesvirus 8 strain: GQ994935.1) with Bowtie2 (74).
Model-based Analysis of ChIP-seq (MACS2) was used for peak detection (75) utilizing the parameters described in the developer's manual. Peaks and read alignments were visualized using the Integrated Genome Browser (IGB) (76).
RNA-sequencing (RNA-seq). Indexed, stranded mRNA-seq libraries were prepared from total RNA (100 ng) using the KAPA Stranded mRNA-Seq kit (Roche) according to the manufacturer's standard protocol.
RNA-Seq data was analyzed using a Salmon-tximport-DESeq2 pipeline. Raw sequence reads (FASTQ format) were mapped to the reference human genome assembly (GRCh38/hg38, GENCODE release 36) and quanti ed with Salmon (77). Gene-level counts were imported with tximport (78) and differential expression analysis including Volcano plot were performed with DESeq2 (79).
SLAM-seq. SLAM-seq (48) was performed using the SLAMseq Kinetics Kit (Lexogen GmbH, Vienna, Austria) according to the manufacturer's standard protocol. Brie y, biological replicate cultures of BCBL-1 cells or BC-1 cells were incubated with VGN50 peptide (24 mM) for 30 min. Subsequently, 4-Thiouridine (s4U; 300 mM) was added to the culture media and the cells incubated for 1 hour in order to label newly synthesized RNA. Total RNA was isolated and then the 4-thiol groups in the s4Uracil-labeled transcripts were alkylated with iodoacetamide (IAA). QuantSeq 3' mRNA-Seq (FWD) (Lexogen, Inc.) Illuminacompatible, indexed sequencing libraries were prepared from alkylated RNA samples (100 ng) according to the manufacturer's protocol for oligo(dT)-primed rst strand cDNA synthesis, random-primed second strand synthesis, and library ampli cation. Libraries were multiplex sequenced (1 x 100 bp, single read) on an Illumina HiSeq 4000 sequencing system. SLAM-Seq datasets were analyzed using the T>C conversion-aware SLAM-DUNK (Digital Unmasking of Nucleotide conversion-containing k-mers) pipeline utilizing the default parameters (48,80). Brie y, nucleotide conversion-aware read mapping of adapter-and poly(A)-trimmed sequences to the human GRCh38/hg38 reference genome assembly was performed with NextGenMap (81). Alignments were ltered for those with a minimum identity of 95% and minimum of 50% of the read bases mapped. For multi-mappers, ambiguous reads and non-3' UTR alignments were discarded, while one read was randomly selected from multimappers aligned to the same 3' UTR. SNP calling (coverage cut-off of 10X and variant fraction cut-off of 0.8) with VarScan2 (82) was performed in order to mask actual T>C SNPs.
Non-SNP T>C conversion events were then counted and the fraction of labeled transcripts determined. All results were exported (i.e., tcount le) and used for downstream analyses, such as differential expression and nascent transcript analysis.
Bioinformatics analysis of SLAM-seq data. The UCSC Genome Browser was used to convert RefSeq IDs to gene symbols (refGene). The resulting data were rst ltered by Log2FC < 0 and sorted by Padj from lowest to highest (<0.05). The top 100 differentially-expressed genes in BC-1 and BCBL-1 were illustrated in a Venn diagram. In addition, the top 100 genes in publicly available SLAM-seq data sets derived from JQ1-treated cells (9) ltered in the same manner, were also depicted in Venn diagrams. In addition, the ltered data of BC-1 cells was analyzed using GSEA (83,84) with default parameters and adjustment of Min size = 5. For transcription factor analysis, the RefSeq IDs ltered by Log2FC < 0 and Padj < 0.05 were submitted to CSCAN (51) to analyze common regulators and to predict transcription factor binding. The following parameters were used; Organism (annotation): Homo sapiens (RefSeq), region: -450/+50, and Cell Line: GM12878. The outputs of BC-1 and BCBL-1 were then ltered with Benjamini-Hochberg adjusted p-value (Pbenj) < 0.05 and plotted in a Venn diagram.
Puri cation of recombinant protein. Spodoptera frugiperda Sf9 cells (Millipore) were maintained in Ex-Cell 420 medium (Sigma), and recombinant baculoviruses were generated with the BAC-to-Bac system as previously described (61,85).   KSHV transactivator interacts with cellular coactivator complexes with its intrinsically-disordered transactivation domain. (a) Scheme for RIME (Rapid immunoprecipitation Mass spectrometry of Endogenous protein) assays. Two antibodies anti-Flag (for K-Rta) and anti-RNAPII (POLR2A) were used to identify cellular proteins that exhibited increased interactions with POLR2A during KSHV reactivation.
(b) STRING Interaction map. List of protein names among RNAPII interacting proteins whose peptide