The Histone Demethylase LSD1/ZNF217/CoREST Complex is a Major Restriction Factor of Epstein-Barr Virus Lytic Reactivation

Abstract

Epstein-Barr virus (EBV) contributes to ~1.5% of human cancers, including lymphomas, gastric and nasopharyngeal carcinomas. In most of these, nearly 80 viral lytic genes are silenced by incompletely understood epigenetic mechanisms, precluding use of antiviral agents such as ganciclovir to treat the 200,000 EBV-associated cancers/year. To identify host factors critical for EBV latency, we performed a human genome-wide CRISPR-Cas9 screen in Burkitt B-cells. Top hits included the lysine-specific histone demethylase LSD1 and its co-repressors ZNF217 and CoREST. LSD1 removes histone 3 lysine 4 (H3K4) and histone 3 lysine 9 (H3K9) methylation marks to downmodulate chromatin activation. LSD1, ZNF217 or CoREST knockout triggered EBV reactivation, as did a LSD1 small molecule antagonist, whose effects were additive with histone deacetylase inhibition. LSD1 blockade reactivated EBV in Burkitt lymphoma, gastric carcinoma and nasopharyngeal carcinoma models, sensitized cells to ganciclovir cytotoxicity and induced EBV reactivation in murine xenografts. ZNF217 and LSD1 co-occupied the EBV immediate early gene BZLF1 promoter, which drives B-cell lytic cycle, as well as to the oriLyt enhancer regions recently implicated in EBV reactivation. LSD1 depletion increased levels of activating histone 3 lysine 4 (H3K4) methylation but not repressive histone lysine 9 methylation marks at BZLF1 and oriLyt and induced their interaction by long-range DNA looping. An orthogonal CRISPR screen highlighted a key H3K4 methyltransferase KMT2D role in driving EBV reactivation. Our results highlight H3K4 methylation as a major EBV lytic switch regulator and suggest novel therapeutic approaches.

Keywords: DNA looping; epigenetic; gamma-herpesvirus; histone demethylase; histone methyltransferase; latency; lymphoma; lytic induction therapy; lytic reactivation.

Figure 1.

Human genome-wide CRISPR-Cas9 screen for host factors that restrict EBV lytic reactivation. (A) CRISPR-Cas9 screen workflow. Cas9+ P3HR-1 Burkitt B-cells were transduced with Brunello sgRNA library at multiple of infection (MOI) ~0.3. Nine days post-transduction, cells with de-repressed plasma membrane (PM) EBV lytic gp350 expression were sorted. sgRNA abundances in input vs sorted cells were quantitated and statistically significant hits were identified. (B) Volcano plots visualization of screen hits. Selected hits are highlighted by epigenetic category. (C) Cross-comparison of Brunello versus Avana sgRNA library CRISPR screens for host factors that repress EBV reactivation.

Figure 2.

The LSD1/ZNF217/CoREST complex restricts Burkitt EBV lytic reactivation. (A) Schematic of the LSD1/ZNF217/CoREST complex, which can erase H3K4 methyl and/or H3K9 acetyl marks. (B-D) Immunoblot analysis of whole cell lysates (WCL) from Cas9+ Akata cells expressing control, LSD1 (B), ZNF217 (C), or CoREST (D) sgRNAs for immediate early lytic BZLF1, early BMRF1 and late p18 expression. (E) Fluorescence-activated cell sorting (FACS) analysis of PM gp350 levels on Akata cells expressing control, LSD1, ZNF217, or CoREST sgRNAs. (F) Mean fluorescence intensity (MFI) ± standard deviation (SD) gp350 PM values from n=3 independent replicates. (G-H) Intracellular (G) or extracellular (H) qPCR analysis of EBV genome copy number in Akata cells expressing control, LSD1, ZNF217, or CoREST sgRNA. Shown are mean ± SD values from n=3 replicates. P-values indicate significance of differences relative to sgRNA control cell values. (I) Immunoblot analysis of WCL from Akata cells expressing control, LSD1, ZNF217, or CoREST sgRNA. Blots are representative of n=3 replicates. *p<0.05, **p<0.01, ***p< 0.001.

Figure 3.

LSD1 inhibition induces EBV lytic reactivation in EBV+ cancer cells. (A) Immunoblot analysis of WCL from P3HR-1 or MUTU I Burkitt cells treated with the indicated Corin concentrations for two days. (B) FACS analysis of PM gp350 levels in Corin-treated P3HR-1 or MUTU I cells. (C) Immunoblot analysis of WCL from EBV+ KEM III, Jijoye, Farage, AGS, YCCEL1, C666–1 cells treated with Corin (0, 1, 2, or 5 μM) for two days. (D) Immunoblot analysis of WCL from EBV+/KSHV+ JSC-1 primary effusion lymphoma cells treated with Corin (0, 1, 2, or 5 μM) for two days. (E) Workflow of Corin and ganciclovir (GCV) treatment. Cells were seeded into fresh media on days of Corin treatment. GCV (10 μg/ml) was added twice daily where indicated. (F) EBV+ Burkitt B-cells were treated as described in (E), with the following Corin concentrations: P3HR-1 (0.5 μM), MUTU I (0.5 μM), Rael (0.25 μM) vs vehicle. Shown are mean ± SD live cell number relative to DMSO-treated controls on day 6 from n=3 replicates. Values of DMSO treated control cells were normalized to 1. *p<0.05, ***p< 0.001. (G) MUTU I murine xenograft experiment workflow. Two weeks post Mutu I Burkitt xenograft implantation, mice were treated with vehicle or Corin (30 mg/ml) as indicated. (H) Immunoblot analysis of WCL prepared from xenografts harvested following treatment as in (G). (I) Immunohistochemical analysis of BZLF1 and BMRF1 expression in xenograft tumors following treatment with vehicle vs. Corin. Scale bar = 100 μm. All blots shown are representative images of n=3 replicates.

Figure 4.

Perturbation of LSD1, ZNF217 or CoREST upregulates activating H3K4 methylation marks at the oriLyt enhancer and BZLF1 immediate early promoter. (A) Chromatin immunoprecipitation sequencing (ChIP-seq) analysis of Akata strain EBV genome-wide ZNF217 and LSD1 occupancy. Shown are input, ZNF217 and LSD1 ChIP-seq tracks from n=2 independent replicates, with track heights set to 5,000 (5K). (B) ChIP-qPCR analysis of ZNF217 occupancy at oriLyt versus BZLF1 promoter (BZLF1p) regions in Akata control versus LSD1 knockout (KO) cells. Mean ± SD percentages of input values from n=3 replicates are shown. (C) ChIP-qPCR analysis of LSD1 occupancy at oriLyt versus BZLF1p regions in Akata control versus ZNF217 KO, as in (B). (D-E) ChIP-qPCR analysis of H3K4 mono (H3K4me1), di (H3K4me2) and tri (H3K4me3) methylation levels in Akata cells expressing control sgRNA or either of two independent screen hit sgRNAs against LSD1, ZNF217 or CoRest. Shown are mean ± SD percentages of input values from n=3 replicates for BZLF1p (D) or OriLyt^L (E). Significance values in D-E refer to comparisons with control cell values. NS, not significant; *p < 0.05; **p < 0.01; ***p < 0.001.

Figure 5.

Depletion of LSD1, ZNF217 or CoREST triggers OriLyt and the BZLF1 promoter region long-range DNA interaction. (A) Schematic diagram highlighting chromatin conformation capture (3C) assay anchor primer (red box) and 12 test (T) primers locations along with the linear EBV genome. For reference, EBV genomic terminal repeat (TR) and origin of plasmid replication (OriP) are shown. (B) 3C assay analysis of DNA looping between the BZLF1 anchor and 12 test primer regions. Shown are the mean ± SD 3C assay signals relative to EBV BACmid negative control from Akata cells expressing control, LSD1, ZNF217, or CoREST sgRNAs. (C) Mean ± SD 3C assay signals from Akata cells 24 hours post-treatment with vehicle, C12 (2.5 μM) or Corin (2.5 μM). *p < 0.05; **p < 0.01; ***p < 0.001.

Figure 6.

Human genome-wide CRISPR screen identifies H3K4 methyltransferase KMT2D as important for EBV reactivation. (A) Workflow of human genome-wide CRISPR-Cas9 screen for host factors that support EBV lytic reactivation. Cas9+ P3HR-1 with stable ZHT/RHT conditional immediate early allele expression were transduced with the Brunello sgRNA library. Transduced cells were selected and then reactivated by 4-hydroxytamoxifen (4HT, 0.4 μM) together with NaB (0.5 mM) for 48 hours. The 5% of cells with the least gp350 expression were sorted. sgRNA abundance in input vs sorted cells was quantitated to identify hits. (B) Volcano plot analysis of screen hits, which are highlighted by category. (C) Immunoblot analysis of WCL from P3HR-1 ZHT/RHT cells expressing control or KMT2D sgRNAs and mock induced or induced by 4HT (0.4 μM) and NaB (0.5 mM) for 24 hours. (D-E) FACS analysis of PM gp350 levels (D) and of PM gp350 MFI ± SD from n=3 replicates (E) in P3HR-1 ZHT/RHT cells with control vs KMT2D sgRNAs and mock induced or induced for lytic replication by 4HT and NaB for 24 hours. (F) qPCR of intracellular EBV genome copy number in P3HR-1 ZHT/RHT cells expressing control or KMT2D sgRNA that were treated with or without 4HT (0.4 μM) and NaB (0.5 mM) for 24 hours. (G) Immunoblot analysis of WCL from P3HR-1 ZHT/RHT cells expressing control or KMT2D sgRNA that were treated with or without Corin (2.5 μM) for 24 hours. All blots shown are representative images of n = 3 replicates. Bar graphs are presented as mean ± SD from three replicates. ***p < 0.001.

Figure 7.

Schematic model. In latency, LSD1/ZNF217/CoREST/HDAC complexes co-occupy EBV oriLyt enhancer regions, where LSD1 and HDAC erase activating H3K4 methylation and H3K4 acetylation marks, respectively. Perturbation of LSD1 and HDAC activity enables KMT2D to deposit activating H3K4 epigenetic marks at both oriLyt enhancers and supports their looping to the immediate early BZLF1 promoter. Newly synthesized BZLF1 drives the lytic cycle by binding to early gene promoters and also to oriLyt enhancers to increase their strength, which further upregulates BZLF1 in a positive feedback loop.