The DNA loop release factor WAPL suppresses Epstein-Barr virus latent membrane protein expression to maintain the highly restricted latency I program

Epstein-Barr virus (EBV) uses latency programs to colonize the memory B-cell reservoir, and each program is associated with human malignancies. However, knowledge remains incomplete of epigenetic mechanisms that maintain the highly restricted latency I program, present in memory and Burkitt lymphoma cells, in which EBNA1 is the only EBV-encoded protein expressed. Given increasing appreciation that higher order chromatin architecture is an important determinant of viral and host gene expression, we investigated roles of Wings Apart-Like Protein Homolog (WAPL), a host factor that unloads cohesin to control DNA loop size and that was discovered as an EBNA2-associated protein. WAPL knockout (KO) in Burkitt cells de-repressed LMP1 and LMP2A expression, but not other EBV oncogenes, to yield a viral program reminiscent of EBV latency II, which is rarely observed in B-cells. WAPL KO also increased LMP1/2A levels in latency III lymphoblastoid cells. WAPL KO altered EBV genome architecture, triggering formation of DNA loops between the LMP promoter region and the EBV origins of lytic replication (oriLyt). Hi-C analysis further demonstrated that WAPL KO reprogrammed EBV genomic DNA looping. LMP1 and LMP2A de-repression correlated with decreased histone repressive marks at their promoters. We propose that EBV coopts WAPL to negatively regulate latent membrane protein expression to maintain Burkitt latency I.

Here, we tested the hypothesis that EBV utilizes WAPL to regulate viral gene expression.Burkitt WAPL knockout (KO) de-repressed LMP1/2A, but not other EBV latency genes, suggestive of a switch towards latency IIa.WAPL KO altered specific EBV genomic DNA loops, especially at the LMPp and oriLyt enhancers.

WAPL is necessary for maintenance of EBV latency I
To test the role of WAPL in EBV gene regulation, we knocked out WAPL in latency I Burkitt MUTU I or latency III GM12878 lymphoblastoid (LCL) cells (S1A-S1F Fig) .WAPL KO did not significantly alter MUTU I or GM12878 proliferation, even though it dramatically altered nuclear morphology (S1A-S1F Fig) , consistent with prior studies in EBV-negative cancer cells [33,34].
To define how WAPL KO affects human and EBV gene expression, we performed RNA sequencing (RNA-seq) in WAPL KO and control MUTU I and GM12878.While most EBV gene levels were unchanged by WAPL KO, MUTU I LMP1 and LMP2A levels increased (Figs 1C and S1I and S1 Table ).By contrast, EBNA2 levels did not substantially increase, suggesting an alternative, latency IIa-like mechanism of LMP1/2A induction (Figs 1C-1D, S1H, and S1I).WAPL KO did not significantly increase expression of most EBV lytic genes or EBV genome copy number (Figs 1C-1D and S1G-I and S1 Table ).WAPL KO modestly impacted GM12878 LMP1/2A abundances but did not significantly alter EBNA2 or EBNA1 (Figs 1E-1F and S1I and S1 Table ).WAPL mRNA and protein levels were approximately 50% lower in GM12878 than in MUTU I (S2A-S2E Fig), which correlated with a comparatively modest effect of WAPL depletion on GM12878 LMP1/2A abundances.

Discussion
Much remains to be learned about epigenetic mechanisms that maintain latency I.We found the cohesin release factor WAPL represses LMP1/2A in Burkitt latency I and alters higher order EBV genomic architecture.WAPL KO triggered DNA loops between oriLyt and LMPp, decreased LMPp repressive H3K9me3/H3K27me3 marks, and de-repressed LMP1/2A coexpression, even without EBNA2.These results highlight an important WAPL role in preventing reversion to latency II.
WAPL loss permits cohesin to slide beyond human CTCF anchors and enlarges DNA loops [33].Our findings suggest that WAPL likewise regulates EBV genome architecture.EBV genomic structure may be distinct between latency IIa germinal center B-cells and latency I memory B-cells.Future work will determine whether WAPL abundance or activity differs between these states.Since germinal center cytokine IL-21 boosts LMP1 expression in latency I [12], it may alter WAPL activity, potentially in a STAT3-dependent manner.
WAPL KO reduced LMPp histone repressive marks in latency I, suggesting that WAPL supports an EBV genomic configuration that represses LMP1/2A (Fig 4G).While WAPL KO may alter a host factor that itself alters LMPp epigenetic marks, WAPL KO did not alter expression of H3K9me3/H3K27me3 writers or erasers.Instead, WAPL may prevent oriLyt/LMPp loop formation.DNA loops between oriLyt and LMPp regions were described in gastric carcinoma and natural killer cells [25,55], but not previously in B-cells.Instead, in latency III, cohesin and CTCF bind to the LMP1/2A region to drive OriP enhancer and LMPp looping.However, the OriP/LMPp loop is present in MUTU I, where LMP1/2A are silenced [30], suggesting additional mechanisms repress LMP in latency I. Consistent with our finding that WAPL KO diminished LMPp cohesin occupancy, WAPL creates a pool of free cohesin.WAPL knockdown can alter cohesin occupancy and enhancer-promoter looping at human genomic sites [32].We speculate that a similar mechanism accounts for the decrease in LMPp cohesin level.Furthermore, LCL SMC1 or RAD21 knockdown increases LMP1/2A levels [32], consistent with a model in which WAPL supports LMPp cohesin occupancy in latency I.Additional marks and inhibits LMP expression.In the absence of WAPL antagonism, a loop forms between the LMP promoter region and oriLyt regions.Juxtaposition of the oriLyt enhancer reduces repressive H3K9me3 and H3K27me3 marks and supports LMP1 and LMP2A co-expression in the absence of EBNA2 (latency II).For reference, in latency III, an alternative loop forms between the oriP and the Cp to drive expression of all the EBNA genes.https://doi.org/10.1371/journal.ppat.1012525.g004regulators, such as cytokines [12], likely work with DNA looping to regulate latency gene expression.
WAPL was discovered as an EBNA2 binding partner [37].Since EBNA2 is a major inducer of LMP1/2A in EBV latency III, an intriguing possibility is that EBNA2 not only activates LMPp chromatin but also dismisses WAPL from LMPp.EBNA2 may alter EBV genomic architecture to reduce H3K9me3/H3K27me3 repressive marks in support of LMP expression in newly infected cells.In latency III, this mechanism may function with EBNA2-driven TET2 DNA hypomethylation [56,57].Future work will determine whether EBNA2 alters EBV genomic looping.Our studies highlight a correlation between WAPL depletion and altered EBV genomic looping, and further studies are needed to define a direct WAPL role in control of EBV genomic architecture.
In conclusion, our data suggest that EBV coopts WAPL in latency I to regulate higher order EBV genome architecture to restrict LMP1/2A expression.WAPL KO provides a new latency IIa B-cell model and lays the foundation for future studies of how WAPL remodels enhancer/ promoter communication for EBV three-dimensional genome regulation, an area that is of interest to investigate for double stranded DNA viruses more broadly.

RNA-seq
RNA poly-A enrichment was performed prior to library preparation and NGS.Reads were mapped to the GRCh37 human and Akata EBV genomes.Transcripts were quantified with Salmon [58].A log 2 FC of > 0.6 and adjusted p-value < 0.05 were significant.DEGs were determined by DESeq2 [59].

HiChIP
HiChIP was performed as described [51].Read loops between EBV genomic bins (1.5kb) were quantified and normalized using loops per 10k total read pairs.Differences were evaluated by Wilcoxon Rank Sum.Differential loops (p-value < 0.1, difference > 3 normalized read pairs, mean read pairs � 2 in at least one condition) were visualized by circlize [60].

Fig 2 .
Fig 2. Subcellular distribution of LMP1 and LMP2A de-repressed by WAPL KO. (A) Representative immunofluorescence images from n = 3 biological replicates of anti-LMP1 (green) vs. nuclear DAPI (blue) staining of Cas9+ MUTU I cells that expressed control or WAPL sgRNAs, as indicated.Shown at right are zoomed images of a representative cell (indicated by the white box).(B) Mean ± standard deviation (SD) percentage of LMP1+ cells per field of view, from n = 3 fields of view from each of three biological replicates.P-values shown as calculated by one-way ANOVA.(C) Representative immunofluorescence images from n = 3 biological replicates of anti-LMP2A (green) vs. nuclear DAPI (blue) staining of Cas9+ MUTU I cells that expressed control or WAPL sgRNAs with zoomed images presented to the right, as in (A).(D) Mean ± SD percentage of LMP2A+ cells per field of view, from n = 3 fields of view from each of three biological replicates.P-values shown as calculated by one-way ANOVA.https://doi.org/10.1371/journal.ppat.1012525.g002

Fig 3 .
Fig 3. WAPL KO alters higher order latency I Burkitt EBV genome conformation.(A) Schematic of Hi-C workflow and output.Exposed DNA ends were biotinylated and then ligated to capture close DNA contacts.Ligated DNA was sheared, and biotinylated DNA was captured via streptavidin.EBV DNA was captured to enhance viral DNA Hi-C signal.(B) Hi-C maps of EBV genomic loops that were enriched in Cas9+ MUTU I cells expressing WAPL vs. control sgRNAs, from n = 2 biological replicates.LMPp and oriLyt regions are indicated.(C) Hi-C maps of EBV genomic loops that were depleted in Cas9+ MUTU I cells expressing WAPL vs. control sgRNAs, from n = 2 biological replicates, as in (B).https://doi.org/10.1371/journal.ppat.1012525.g003

Fig 4 .
Fig 4. WAPL KO alters latency I Burkitt EBV genomic activated chromatin loops and represses LMP promoter epigenetic marks.(A) Schematic of H3K27ac HiChIP sample preparation and output.Chromatin was formaldehyde crosslinked and fragmented.Exposed DNA ends were biotinylated and then ligated to capture close DNA contacts.Ligated DNA was sheared, DNA was immunopurified by α-H3K27ac antibody, and biotinylated DNA was captured via streptavidin.(B) EBV genomic H3K27ac HiChIP map depicting loops enriched (red) or depleted (blue) in Cas9+ MUTU I cells expressing WAPL vs. control sgRNAs, from n = 3 biological replicates.(C-D) Normalized (C) LMP region-oriLyt R loop and (D) LMP region-oriLyt L loop read counts from n = 3 replicates, as in (B).EBV genome kilobase coordinates for each looping region are indicated at top.* P � 0.05, ** P � 0.01, as calculated by a two-tailed Student's t-test.(E-F) ChIP-qPCR analysis of H3K9me3 and H3K27me3 abundances at the (E) LMP1 promoter and (F) LMP2A promoter in Cas9+ MUTU I cells expressing control or WAPL sgRNAs.Shown are the mean fold change of ChIP-qPCR values relative to input values ± SD from n = 3 biological replicates.Values from sgControl expressing cells were normalized to 1. ** P � 0.01, *** P � 0.001, as calculated by a two-tailed Welch's t-test.(G) Model of WAPL effects on EBV genomic architecture.When present, WAPL releases cohesin (latency I), which enables the accumulation of repressive histone