GC cytokines support the latency II transition

To systematically characterize GC cytokine effects on EBV latency gene expression, we incubated the LCL GM12878 with a panel of Tfh-derived cytokines, IL-2, IL-4, IL-10 or IL-21. In parallel, we incubated GM12878 with the FDC-derived cytokines IL-6, IL-15 or IL-27 for 0, 2, 4 or 6 days (Figs 1A and S1A). While it is not known how long EBV+ B-cells reside within the GC, it is likely that they remain present for at least several days, in order to proliferate and differentiate into memory B cells, and GC structures themselves persist for weeks to months. Cytokine effects on EBV latency programs were defined by immunoblot for EBNA2, EBNA3C and LMP1, since this panel of EBV oncoproteins can be used to assign the latency program. Interestingly, most of these cytokines reduced EBNA2 and 3C expression, though results were the most pronounced for IL-21, which rapidly and robustly impaired EBNA2/3C expression (Fig 1B). By contrast, IL-10 and IL-21 upregulated LMP1 expression within 2 days of treatment (Fig 1B). Similar effects were observed in a second LCL, GM12881 (S1B Fig). IL-21 also suppressed EBNA2 and upregulated LMP1 in latency III Jijoye Burkitt cells (S1C Fig), suggesting generalizable effects on the latency III program. Consistent with prior reports, IL-21 did not hyper-induce LMP2A expression in either GM12878 or Kem III LCLs, indicating that IL-21 may fail to induce recruitment of an activator to the LMP2 promoter or to instead dismiss a repressor (S1D Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. GC cytokines support the transition to EBV latency II.

(A) GC schematic, illustrating key T follicular helper cell (Tfh) and follicular dendritic cell (FDC) secreted cytokines and CD40 ligand (CD40L) that signal to GC B-cells. (BioRender was used to create the schematic models.) (B) Immunoblot analysis of whole cell lysates (WCL) from GM12878 treated with the indicated cytokines for two, four, or six days. (C) Volcano plots of EBV gene expression from n = 3 replicates of GM12878 stimulated by IL-15 vs. mock-stimulated for 6 days. (D) Volcano plots of EBV gene expression from n = 3 replicates of GM12878 stimulated by IL-21 vs. mock-stimulated for 6 days. (E) Immunoblot analysis of WCL from GM12878 cells treated with the indicated cytokines for six days. (F) Immunoblot analysis of WCL from Mutu I cells treated with the indicated cytokine for one day or from GM12878 for comparison. (G) Volcano plot of EBV gene expression from n = 3 replicates of Mutu I stimulated by IL-4+CD40L vs. mock-stimulated for one day. (H) Volcano plot of EBV gene expression from n = 3 replicates of Mutu I stimulated by IL-21 vs. mock-stimulated for one day. All cytokines were used at 100 ng/ml and 50 ng/ml in GM12878 vs Mutu I, respectively, and were refreshed every two days. Immunoblots are representative of n = 3 replicates.

https://doi.org/10.1371/journal.ppat.1011939.g001

We next performed RNA-seq analysis to systematically characterize IL-15 and IL-21 effects on EBV genome-wide expression, as representative of FDC vs Tfh cytokine signaling, respectively. After six days of treatment, IL-15 significantly decreased expression of multiple latency III genes, including EBNA2, EBNA3, EBNA-LP, LMP1 and LMP2s, but increased expression of a subset of lytic cycle genes, including immediate early BZLF1 and early BMRF1 (Fig 1C and S1 Table), suggestive of an abortive lytic cycle. Instead, IL-21 significantly increased abundance of LMP1 mRNA but decreased abundances of EBNA2, EBNA3, EBNA-LP, and LMP2 mRNAs (Fig 1D and S1 Table). Consistent with effects on EBNA2 and LMP1 expression, IL-15 downregulated the EBNA2 target gene CD300A, while IL-21 upregulated levels of the LMP1/NF-κB target ICAM-1 and downmodulated CD300A [21,39,40] (S1E and S1F Fig and S2 Table).

GC cytokines alter B-cell gene expression patterns via multiple effectors, including distinct JAK/STAT pathways. As expected, the panel of cytokines differentially activated STATs, including STAT5 activation by IL-2 and IL-15 versus STAT6 activation by IL-4 versus STAT3 activation by IL-6, IL-10, IL-21 and IL-27, as judged by immunoblot, using well characterized phosphorylation marks of STAT activation [26,27] (Figs 1E and S1B–S1C). Consistent with our RNA-seq analyses, IL-15 de-repressed BZLF1 and BMRF1 expression at the protein level, as did IL-2 (Fig 1E), suggesting that it induces an abortive lytic cycle in at least a subset of cells. Notably, these two cytokines share receptor beta and gamma chain subunits, which are transmembrane proteins that activate downstream pathways, including JAK/STAT [41].

EBV-infected B-cells likely also experience combinatorial cytokine signaling in germinal center microenvironments. To model these effects, we next treated GM12878 and GM12881 LCLs with the Tfh cytokines IL-4+IL-21, the FDC cytokine IL-15 together with IL-21 or the FDC cytokines IL-15 and IL-27 (Fig 1A). IL-15/IL-21 and to a lesser extent IL-15/IL-27 suppressed EBNA2 and EBNA3C expression somewhat more strongly than either cytokine alone. Interestingly, while IL-4/IL-21 did not repress either EBNA2 or EBNA3C more than IL-21 alone, IL-4/IL-21 did boost LMP1 levels beyond those observed with IL-21 or IL-4 alone (S2A and S2B Fig), likely due to alterations in JAK/STAT signaling.

We next asked the extent to which Tfh and FDC cues can alter EBV latency gene expression within the latency I B-cell context. While several Tfh signals, including IL-4+CD40L, IL-10 or IL-21 can each de-repress LMP1 expression in B-cells with the latency I program [30–34], it has remained unknown the extent to which other GC microenvironmental cues more broadly alter EBV latency gene expression within latency I. To gain insights, we treated latency I Mutu I and Kem I Burkitt cells with a panel of Tfh and FDC cytokines, as there is no primary human B-cell latency I models currently available. IL-21 strongly activated STAT3, as judged by tyrosine 705 phosphorylation, and robustly de-repressed LMP1 expression in Mutu I and Kem I (Figs 1F and S2C and S2D). By contrast, IL-4+CD40L or IL-10 treatment also induced STAT3 phosphorylation and LMP1 expression, albeit to a lesser extent (Fig 1F). This did not appear to be a full transition to the latency II program, as neither IL-10 nor IL-21 induced LMP2A to an appreciable degree in Mutu I or Kem I (S2E Fig). Differences between GC cytokine STAT activation in the latency I vs III context may relate to altered expression of receptors versus negative regulators of JAK/STAT signaling.

To then systematically analyze latency I B-cell responses to the Tfh signals IL-4+CD40L vs IL-21, we performed RNA-seq on Mutu I that were mock-stimulated or stimulated by these Tfh cues for 1 day. This early timepoint was chosen since we observed robust effects on LMP1 de-repression by that early timepoint, and as we observed reduced Mutu I viability with longer treatments. Consistent with our immunoblot analysis, IL-4+CD40L only modestly increased LMP1 expression, whereas IL-21 strongly induced LMP1 (Fig 1G and 1H). Notably, these stimuli did not significantly de-repress expression of EBNA or mildly increased LMP2 mRNAs, suggesting a specific effect at the level of the LMP1 promoter.

Analysis of Mutu I host transcriptome responses to either IL-4+CD40L or IL-21 treatment highlighted upregulation of multiple LMP1 target genes [42], including mRNAs encoding the NF-κB subunits RelB and p100/52 (encoded by NFKB2), ICAM-1 and IRF4 (S2F and S2G Fig). The NF-κB pathway signaling pathway and EBV infection were amongst the pathways most highly enriched by either cytokine treatment. While direct effects of the cytokines themselves may account for a subset of these changes, we note that IL-21 is not a strong inducer of NF-κB signaling, suggesting that de-repressed LMP1 may be an important mediator of the observed host transcriptomic changes.

STAT3 and STAT5 mediate GC cytokine effects on the EBV latency III program

We next investigated effects of chemical or CRISPR JAK/STAT blockade to gain further insights into specific STAT roles in control of EBV latency gene expression downstream of IL-15 and IL-21. First, to broadly characterize JAK/STAT roles in LMP and EBNA expression, we treated latency III GM12878 and Jijoye cells with IL-15 or IL-21, in the absence or presence of the pan-JAK ATP-competitive inhibitor CAS 457081-03-7 (also referred to as JAK inhibitor I or JAKi). On-target JAKi effects were confirmed by immunoblot analysis of STAT3 and STAT5 phosphorylation, which demonstrated loss of STAT5 Tyrosine 694 and downmodulation of STAT3 Tyrosine 705 phosphorylation in IL-15 and IL-21 treated cells, respectively (Fig 2A and 2B). JAKi impaired IL-15 downmodulation of EBNA2 and 3C expression (Fig 2A and 2B). Likewise, JAKi treatment partially impaired IL-21 suppression of EBNA2 and EBNA3C expression and reduced the extent to which IL-21 hyper-induced LMP1 (Fig 2A and 2B). Incomplete blockade of IL-21 driven STAT3 phosphorylation may explain the comparatively milder JAKi effects on IL-21 than on IL15 regulation of latency III expression.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. STAT3/5 roles in EBV latency gene regulation by IL-15 or IL-21.

(A-B) Immunoblot analysis of WCL from GM12878 (A) or Jijoye (B) cells pre-treated with DMSO vehicle or JAK inhibitor CAS 457081-03-7 (JAKi, 200 ng/ml) for one hour, followed by treatment with IL-15 or IL-21 for six days. (C-D) Immunoblot analysis of WCL from GM12878 cells expressing control sgRNA versus sgRNA targeting the indicated STAT3 and/or STAT5 genes, which were then mock treated or treated with IL-15 (C) or IL-21 (D) for six days. (E-F) Immunoblot analysis of WCL from GM12878 (E) or Jijoye (F) cells expressing cDNA encoding GFP or constitutive activated STAT3 (STAT-CA), whose expression was induced by 0.5 or 1 μg/ml doxycycline (Dox). (G) Methylated DNA immunoprecipitation and quantitative PCR (MeDIP-qPCR) analysis of GM12878 cells with control or STAT3/5A/5B sgRNA expression, which were mock treated or treated with IL-15 or with IL-21 for six days. Shown is mean ± standard deviation (SD) from n = 3 replicates of Cp qPCR signal. *p < 0.05; NS: not significant. IL-15 and IL-21 were used at 100 ng/ml throughout. Immunoblots are representative of n = 3 replicates.

https://doi.org/10.1371/journal.ppat.1011939.g002

To examine individual STAT transcription factor roles downstream of IL-15 or IL-21, we next used CRISPR/Cas9 editing. Since IL-15 and IL-21 most robustly induced STAT5 and STAT3 phosphorylation (Fig 1E), we tested effects of CRISPR depletion of STAT3, of STAT5A or STAT5B isoforms [43], or of combinations thereof, given their potentially redundant roles. IL-15 repression of EBNA2 or 3C was not significantly perturbed by depletion of STAT3, STAT5A or STAT5B alone. However, concurrent GM12878 and Jijoye STAT5A/5B depletion impaired repression of EBNA2 and 3C by IL-15 and to a lesser extent by IL-21 (Figs 2C, 2D and S3A). In addition, concurrent CRISPR depletion of STAT3, STAT5A and STAT5B more strongly impaired EBNA3C repression by IL-15 (Fig 2C), suggestive of a partially redundant STAT3 and 5 roles, likely at the EBV C promoter.

STAT3 depletion was sufficient to block IL-21 driven LMP1 hyper-induction and impaired IL-21 driven EBNA2/EBNA3C repression (Fig 2D). Nonetheless, combined STAT3/5A/5B editing more strongly impaired EBNA2 and EBNA3C repression by IL-21 (Fig 2D). Despite robust STAT1 activation by IL-21 and to a lesser extent by IL-10 (Fig 1E), CRISPR STAT1 depletion did not alter IL-21 or IL-10 effects on EBNA or LMP1 expression (S3B and S3C Fig). STAT3 KO impaired EBNA2/3C repression and LMP1 hyper-induction downstream of IL-10 (S3B and S3C Fig). Taken together, these results suggest that STAT3 and 5 have partially redundant roles in cytokine mediated EBNA2/3C repression, perhaps through the action of STAT3/5 heterodimers, whereas STAT3 is a major driver of IL-21 driven LMP1 hyper-induction, with relevance to latency III to II reprogramming in the GC microenvironment.

We next tested effects of expressing a STAT3 constitutively active allele (STAT3-CA), in which STAT3 C-terminal loop SH2 domain A662C and N664C point mutations [44,45] drive constitutive STAT3 homodimerization, DNA binding and transcription activation. STAT3-CA expression diminished EBNA2 and increased LMP1 abundance in GM12878 and in Jijoye B-cells (Fig 2E and 2F), further suggesting that STAT3 plays a critical but opposite role in EBNA2/3 vs LMP1 regulation. These observations are consistent with a model in which GC cytokine signaling culminates in assembly of STAT3/5-containing transcriptional repressor complexes at the EBV genomic C promoter, but instead triggers formation of a STAT3 homodimer containing activator complex at the LMP1 promoter.

Since IL-15 and IL-21 upregulated the host transcriptional repressor BCL6 (S1E and S1F Fig) which plays major roles in GC B-cell biology and is critical for GC formation, we tested BCL6 roles in cytokine driven EBV latency gene expression. However, BCL6 CRISPR KO did not appreciably alter IL-21 effects on EBNA2 or LMP1 abundance (S4A Fig). BCL6 KO also did not affect IL-21 effects on LCL plasma membrane CD300A or ICAM-1, which are targets of EBNA2 and LMP1, respectively (S4B Fig).

DNA methylation is critical for suppression of Cp driven EBNA expression in B-cells with latency I, and presumably also in latency II [21,46–49]. We therefore used methylation DNA immunoprecipitation (MeDIP) and qPCR to characterize IL-15 versus IL-21 effects on LCL Cp DNA methylation levels. Notably, IL-21, but not IL-15 significantly increased C, LMP1 and LMP2 promoter methylation levels, and STAT3/5A/5B depletion reversed this effect (Figs 2G and S4C). These results indicate that STAT3/5 promote cross-talk between IL-21, C and LMP promoter DNA methylation.

STAT and DNA methylation role in latency I LMP1 de-repression by GC cytokines

To gain insights into JAK/STAT roles in GC cytokine triggered LMP1 de-repression in latency I B-cells, we treated Mutu I or Kem I Burkitt cells with IL-10, IL-21 or IL-4 together with CD40 ligand, in the absence or presence of JAK inhibition. We tested these GC stimuli since each hyper-induced LMP1 and robustly induced STAT phosphorylation in latency I cells and had previously been reported to de-repress LMP1 expression from latency I [30–36] (Fig 1F). JAKi treatment strongly impaired LMP1 upregulation by each of these stimuli (Figs 3A and S5A), consistent with a key JAK/STAT role in epigenetic regulation at the level of the LMP1 promoter.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. STAT3 roles in GC cytokine mediated LMP1 de-repression in latency I B-cells.

(A) Immunoblot analysis of WCL from latency I Mutu I or Kem I B cells pre-treated with DMSO or JAKi (200 ng/ml) for one hour, followed by treatment with the indicated cytokines for one day. (B) Immunoblot analysis of WCL from Mutu I expressing control sgRNA or sgRNA targeting the indicated STAT transcription factor gene, mock treated or treated with IL-4+CD40L for one day. (C) Immunoblot analysis of WCL from Mutu I expressing control sgRNA or sgRNA targeting the indicated STAT transcription factor gene, treated with IL-10 or IL-21 for one day. (D) Immunoblot analysis of WCL from Mutu I mock treated or treated with IL-21 for one or two days. (E) MeDIP q-PCR analysis of LMP1 promoter methylation in Mutu I (left) or Kem I (right) mock treated or treated with IL-21 for one or two days. Shown are mean ± SD values of % input from n = 3 replicates. (F) MeDIP-qPCR analysis of LMP1p in Mutu I with control or STAT3 targeting sgRNA, mock treated or IL-21 treated for one day. *p < 0.05; **p < 0.01; ***p < 0.001. Cytokines were used at 50 ng/ml. Blots are representative of n = 3 replicates.

https://doi.org/10.1371/journal.ppat.1011939.g003

To next gain mechanistic insights into specific STAT roles, we tested effects of CRISPR depletion of STAT transcription factors that were highly phosphorylated in response to these GC stimuli (Fig 1F). Depletion of either STAT1 or STAT3 blunted LMP1 de-repression by IL-4+CD40L stimulation or even by IL-4 alone (Figs 3B and S5B). By contrast, depletion of STAT3, but not STAT1, impaired IL-10 and IL-21 mediated LMP1 de-repression in Mutu I and Kem I cells (Figs 3C and S5B). Thus, STAT1/3 heterodimers may be important for IL-4 driven LMP1 de-repression, whereas distinct STAT3 heterodimers or homodimers may mediate LMP1 de-repression downstream of IL-10 and IL-21. Consistent with the latter hypothesis, induction of the constitutively active STAT3 allele was sufficient to de-repress LMP1 expression in Mutu I (S5C Fig). Likewise, STAT3 and to a somewhat lesser extent STAT6 over-expression enhanced LMP1 de-repression in response to cytokine treatment (S5D Fig).

To gain further insights into cytokine cross-talk with EBV-genomic CpG methylation, we next analyzed IL-21 effects on the abundance of DNA methyltransferase machinery. IL-21 downregulated expression of the de novo CpG methylation writer DNMT3B (Fig 3D), whose expression counteracts latency III gene expression [21]. Likewise, IL-21 downmodulated UHRF1 expression (Fig 3D), which is important for maintenance of EBV genomic methylation marks, together with DNMT1 [21]. Therefore, to further characterize IL-21 effects on CpG methylation of key EBV genomic promoters, we performed MeDIP-qPCR analysis on Mutu I or Kem I Burkitt cells treated with IL-21 for 1 or 2 days. Interestingly, IL-21 downmodulated the high level of DNA methylation at the LMP1 and C promoters, but not at the LMP2 promoter in either Mutu I or Kem I (Figs 3E and S5E). Additional epigenetic marks may maintain Cp and LMP2p silencing upon IL-21 stimulation in the latency I context, including those driven by STAT-containing repressive complexes. In support of a STAT3 role in modulation of LMP1p methylation downstream of IL-21, we did not observe diminished LMP1p methylation levels in STAT3 depleted Mutu I cells upon IL-21 treatment (Fig 3F).

GC cytokine effects on LMP1 promoter histone epigenetic marks

In addition to DNA methylation, histone epigenetic marks strongly contribute to EBV latency gene expression [14,21,50–57]. We therefore next profiled GC cytokine effects on the LMP1 promoter. Since previous studies identified three LMP1 promoter sites occupied by STAT factors [58] (Fig 4A), we performed chromatin immunoprecipitation (ChIP) and qPCR analyses in LCLs mock treated or treated with FDC-derived IL-15 or Tfh-derived IL-21. IL-21 increased STAT3 occupancy at the S3 site, located at approximately 600 base pairs (bp) upstream of LMP1p, and to a lesser extent at the S2 and S1 sites, located at approximately 500 and 100 bp upstream of LMP1p (Fig 4B). Interestingly, IL-15 instead downmodulated STAT3 occupancy at S2 and S3, consistent with the observation that it does not hyper-induce LMP1 in latency III. By contrast, IL-15 but not IL-21 significantly increased STAT5 occupancy at S1-S3 (Fig 4C). These results further support the hypothesis that IL-21 driven STAT3, and potentially STAT3 homodimers, are major drivers of LMP1 hyper-induction.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. STAT 3 and 5 role in GC cytokine mediated latency III LMP1 promoter epigenetic remodeling.

(A) Schematic diagram of LMP1 promoter STAT binding sites, S1, S2 and S3 [58]. (B-C) Chromatin immunoprecipitation (ChIP) qPCR analysis of STAT3 (B) or STAT5 (C) LMP1 promoter occupancy in GM12878 mock treated or treated with IL-15 or IL-21 for six days. (D-E) ChIP-qPCR analysis of LMP1 promoter H3K27Ac (D) or H2AK119Ub (E) epigenetic mark abundances in GM12878 expressing control versus STAT3/5A/5B targeting sgRNA, mock treated or treated with IL-15 or IL-21 for six days. (F-G) ChIP-qPCR analysis of LMP2 promoter H3K27Ac (F) or H2AK119Ub (G) abundances in GM12878 expressing control versus STAT3/5A/5B targeting sgRNA, mock treated or treated with IL-15 or IL-21 for six days. (H-I) ChIP-qPCR analysis of LMP1 promoter H3K27Ac (H) or H2AK119Ub (I) abundances in Mutu I expressing control sgRNA versus STAT3 targeting sgRNA, mock treated or treated with IL-21 for one day. All ChIP results are presented as % input mean ± SD from n = 3 replicates. *p < 0.05; **p < 0.01; ***p < 0.001.

https://doi.org/10.1371/journal.ppat.1011939.g004

To characterize STAT roles in LMP1 promoter histone epigenetic regulation, we next performed ChIP-qPCR analysis in control versus CRISPR-edited LCLs. In control LCLs, IL-15 and to a greater extent IL-21 increased LMP1p histone 3 lysine 27 acetylation (H3K27Ac), a mark which correlates with promoter activation. By contrast, IL-15 and IL-21 failed to upregulate LMPp H3K27Ac level in LCLs depleted for STAT3, STAT5A and STAT5B (Fig 4D). Since we recently found a role for the polycomb repressive complex (PRC1) I histone 2A lysine 119 ubiquitin (H2AK119Ub) mark in repression of LMP1 expression [21], we next examined GC cytokine and STAT roles on LMP1p H2AK119Ub levels in LCLs. IL-15 and to a greater extent IL-21 significantly diminished H2AK119Ub abundance in control, but not STAT3/5A/5B KO GM12878 (Fig 4E). Despite lack of appreciable LMP2A hyper-induction, IL-21 nonetheless increased H3K27Ac levels in GM12878 control and STAT3/5A/5B edited LCLs (Fig 4F). Interestingly, IL-21 hyper-induced H2AK119Ub repressive marks at LMP2p in both control and STAT3/5A/5B edited cells (Fig 4G). Given PRC1 roles in repression of LMP expression, this result suggests a potential mechanism by which LMP1 but not LMP2A is hyper-induced in IL-21 treated B-cells, and are consistent with a model in which STAT3/5 occupy LMP1 but not LMP2 promoter sites.

We did not observe decreases in the repressive histone 3 lysine lysine 9 dimethyl (H3K9me2) or trimethyl (H3K9me3) marks with either IL-15 or IL-21 treatment in control or STAT KO LCLs (S6A and S6B Fig). However, repressive histone 3 lysine 27 trimethyl (H3K27me3) repressive marks increased somewhat upon IL-15 or IL-21 treatment in STAT3/5A/5B triple edited LCLs (S6C Fig). Similar effects were observed at the LMP2 promoter, though IL-21 increased the repressive H3K9me3 mark in both control and STAT3/5A/5B edited cells (S6D–SF Fig), potentially contributing to the lack of IL-21 driven LMP2A hyper-induction.

We next characterized IL-21 epigenetic effects on the LMP1 promoter in latency I B-cells, given the observation that IL-21 strongly activates STAT3 phosphorylation and de-represses LMP1 expression, whereas other GC cytokine stimuli did so comparatively weakly. As anticipated, IL-21 significantly increased H3K27Ac at the LMP1 promoter in Mutu I cells (Fig 4H). Interestingly, STAT3 was necessary for this IL-21 driven epigenetic remodeling, as STAT3 depletion prevented IL-21 driven H3K27Ac activating mark at the LMP1 promoter (Fig 4H). Similarly, IL-21 significantly diminished the repressive H2AK119Ub and H3K9me2 marks at the LMP1 promoter in a STAT3 dependent manner (Figs 4I and S7A). By contrast, IL-21 did not significantly alter repressive H3K9me3 or H3K27me3 marks at the latency I LMP1 promoter (S7B Fig). Interestingly, IL-21 did not significantly alter activating or repressive histone marks at the Mutu I LMP2 promoter (S7C–S7F Fig). These results indicate that the absence of STAT3 signaling is important for silencing LMP1 expression in latency I, with relevance to the transition from latency II to latency I.

GC cytokines remodel epigenetic status of C promoter

Multiple GC cytokines repressed latency III EBNA expression, suggestive of epigenetic effects at the level of Cp, which drives the large EBV transcript encoding all six EBNAs.

We performed ChIP to characterize how IL-15 and IL-21 alter STAT3 versus STAT5 occupancy at two predicted STAT binding sites using PROMO online tool [59,60], located at 300 and 400 bp upstream of Cp (Fig 5A). Consistent with our observation that IL-15 and IL-21 predominantly activated STAT5 versus STAT3 in latency III cells, respectively (Figs 1B and S1B and S1C), IL-21 but not IL-15 significantly upregulated STAT3 occupancy at both the S1 and S2 sites upstream of Cp (Fig 5B). Conversely, IL-15 significantly induced STAT5 occupancy at both S1 and S2, whereas IL-21 weakly induced STAT5 binding to S2 (Fig 5C). Taken together with our CRISPR and immunoblot analyses, these data are compatible with a model in which a STAT5 or STAT3 homodimer and to a lesser extent a STAT3/5 heterodimer are critical for IL-15 or IL-21 mediated Cp repression.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. STAT3 and 5 role in GC cytokine mediated C promoter epigenetic remodeling.

(A) Schematic diagram of PROMO [59,60] predicted STAT binding sites on C promoter. (B-C) ChIP-qPCR analysis of STAT3 (B) or STAT5 (C) C promoter occupancy in GM12878 mock treated or treated with IL-15 or IL-21 for six days. (D-G) ChIP-qPCR analysis of Cp H3K27Ac (D), H2AK119Ub (E), H3K9me2 (F) and H3K9me3 (G) abundances in GM12878 expressing control versus STAT3/5A/5B targeting sgRNA, mock treated or treated with IL-15 or IL-21 for six days. All ChIP results are presented as % input mean ± SD from n = 3 replicates. Relative fold changes to Mock treated cells are labeled. *p < 0.05; **p < 0.01; ***p < 0.001.

https://doi.org/10.1371/journal.ppat.1011939.g005

At the epigenetic level, ChIP-qPCR assays highlighted that IL-21 more strongly reduced H3K27Ac marks at Cp than IL-15 (Fig 5D). Consistent with key STAT3 and 5 roles in GC cytokine driven epigenetic remodeling at Cp, CRISPR editing of STAT3/5A/5B blocked H3K27Ac loss at Cp in GM12878 stimulated by either IL-15 or IL-21 (Fig 5D). Similarly, IL-15 and to a greater extent IL-21 increased H2AK119Ub repressive marks at Cp, and CRISPR STAT3/5A/5B editing blunted cytokine-driven H2AK119Ub deposition (Fig 5E). Likewise, IL-21 but not IL-15 significantly increased deposition of the H3K9me2 and H3K9me3 repressive marks at Cp, and this increase was blunted by STAT3/5 editing (Fig 5F and 5G). Interestingly, neither IL-15 nor IL-21 increased repressive H3K27me3 marks at Cp, arguing against PRC2 roles in their repression of Cp (S8A Fig). By comparison, Cp is silenced in latency I, and likely related to that, we observed relatively small differences in the Cp epigenetic status in control or STAT3 edited Mutu I at rest or following IL-21 treatment (S8B–S8F Fig). These results are consistent with a model in which STAT3 nucleate transcription co-activator complexes at LMP1p but STAT3 and 5 mediates repressive complexes at Cp in latency III B-cells, and that latency I cells maintain the ability to respond to STAT-dependent epigenetic remodeling at LMP1p.

JAK/STAT signaling roles in newly infected B-cell latency gene expression and transformation

JAK/STAT signaling contributes to EBV latency gene expression in newly infected primary human B-cells [61,62]. Notably, STAT3 tyrosine 705 phosphorylation is elevated within 30 minutes of primary B-cell exposure to EBV, and total STAT3 protein levels are upregulated within 8 hours of EBV infection [62,63]. Such early effects on STAT3 are likely triggered by EBV interaction with B-cell receptors and/or internalization and preceded EBV latency gene expression [62,63]. To build upon these data, we characterized STAT3 and STAT5 phosphorylation levels over multiple subsequent timepoints, in which EBV converts primary human cells into lymphoblastoid cells. Purified CD19+ human peripheral blood B cells were isolated by negative selection and infected with EBV. Timecourse analysis was then performed to measure levels of EBV latency gene, STAT3 and STAT5 expression, as well as of STAT3 and 5 phosphorylation, to indicate their activation status. Interestingly, EBV upregulated STAT3 and STAT5A levels, in particular between days 4 and 21 post-infection, whereas STAT5B levels were relatively constant. Whereas EBV triggered STAT3 phosphorylation, in particular between days 4 and 21 post-infection, a period in which LMP1 levels were markedly elevated and EBNA2 and 3C levels diminished (Fig 6A). Notably, EBV did not trigger STAT5 phosphorylation, as judged by immunoblot of phosphotyrosine 694 (Fig 6A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. IL-15 and IL-21 remodeling of latency III gene expression in newly infected primary human B cells.

(A) Immunoblot analysis of WCL from primary human B cells at the indicated days post infection (DPI) by the Akata EBV strain. (B) Immunoblot analysis of WCL from primary human B cells at 7 DPI, which were then mock treated or treated with the indicated cytokines for six days. (C) Immunoblot analysis of WCL from primary B cells that were treated with DMSO or JAKi (200 ng/ml) for two days at 4, 7 or 10 DPI. GM12878 WCL was included as a control. (D) Primary human B-cell transformation assay characterizing effects of DMSO vs JAKi (200 ng/ml) treatment on primary human B-cell outgrowth following infection by Akata EBV. Fitted non-linear regression curves are presented as mean ± SD from n = 3 replicates, *p < 0.05; **p < 0.01. Blots are representative of n = 3 replicates. Cytokines were used at 100 ng/ml.

https://doi.org/10.1371/journal.ppat.1011939.g006

We next investigated the effects of IL-21 on EBV latency gene expression when dosed at day 7 post-infection, the earliest timepoint when B-cells begin to convert to lymphoblastoid physiology [9,64]. IL-21 reduced EBNA2 and 3C expression and hyper-induced LMP1 (S9A Fig), suggesting conserved STAT roles in EBV latency gene expression in newly infected cells and in LCLs. We next tested the effects of IL-15 and IL-21 on EBV latency gene expression at day 10 post-infection. Treatment with either cytokine for 6 days reduces EBNA2 expression (S9B Fig). IL-21 treatment also strongly down-modulated EBNA2 target gene CD23 [65,66] abundance when applied at multiple distinct timepoints between days 2 and 35 post-infection (S9C and S9D Fig).

To then gain insights into germinal center cytokine effects on EBV-infected primary B-cells, we treated cells at 7 days post-infection with vehicle control, IL-4, IL-15, IL-21, IL-27 or multiple combinations thereof. IL-15 or IL-21 were each sufficient to reduce EBNA2 and EBNA3C expression relative to levels in untreated cells, but we did not appreciate additive effects of combinatorial cytokine treatment on their expression (Fig 6B). While IL-4 did not appreciably alter EBNA2 or EBNA3C levels, it was able to boost LMP1 protein expression above control cell levels. Interestingly, the combination of Tfh cytokines IL-4 and IL-21 hyper-induced LMP1 expression, whereas addition of the FDC cytokine IL-15 to IL-21 did not appreciably alter LMP1 levels relative to those observed with IL-21 treatment alone. Combinatorial treatment with the FDC cytokines IL-15 and IL-27 did not boost LMP1 levels relative to untreated cells (Fig 6B). Immunoblot analysis demonstrated the expected phosphorylation of STAT1, 3, 5 and 6 in response to each of these cytokines, suggesting that their pathways were similarly activated by cytokines in the primary cell as well as transformed B cell settings (Fig 6B).

To characterize the roles of JAK/STAT signaling in EBV-mediated B-cell transformation, we treated newly infected primary human B-cells with JAKi at 4, 7 or 10 DPI. Consistent with JAK/STAT downmodulation of EBNA2 and EBNA3 expression at these early times post-infection, JAKi treatment increased EBNA2 and EBNA3C expression (Fig 6C). Surprisingly, JAKi treatment also mildly increased LMP1 expression, which likely occurred secondary to increases in EBNA2 levels. JAKi treatment also impaired outgrowth of EBV-infected B-cells in a transformation assay (Fig 6D), suggesting that EBV-driven JAK/STAT signaling supports B-cell immortalization, potentially by titrating the levels of EBV oncoprotein expression. Taken together, our results support a model in which JAK/STAT signaling exerts control over EBV latency gene expression through epigenetic effects on key EBV latency gene promoters (Fig 7).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Model of EBV latency promoter epigenetic remodeling by GC cytokine driven JAK/STAT signaling.

Cytokines secreted by germinal center FDC and Tfh cells trigger JAK/STAT signaling to drive STAT homodimer or heterodimer nuclear translocation. STAT dimers bind to the C promoter together with a co-repressor to increase levels of repressive 5-methylcytosine (5-mC), H3K9me2/3 and H2AK119Ub epigenetic marks, while decreasing activating H3K27Ac marks. C promoter inactivation decreases latency III EBNA expression. By contrast, STAT dimers bind to the LMP1 promoter together with a co-activator to drive epigenetic remodeling to increase levels of LMP1 promoter H3K27Ac and to decrease levels of H2AK119Ub, which supports LMP1 expression despite concurrently increased 5mC levels. By contrast, GC cytokines increase the abundances of LMP2 promoter repressive H2AK119Ub, H3K9me3 and 5mC marks to downmodulate LMP2 expression from levels observed in latency III. (BioRender was used to create the schematic models).

https://doi.org/10.1371/journal.ppat.1011939.g007

Ethics statement

Primary B-cells were isolated from discarded, de-identified peripheral blood mononuclear cells from the Brigham and Women’s Hospital Blood Bank, obtained following platelet donation, using an Institutional Review Board approved protocol and written consent was obtained from the donors.

Cell culture

EBV+ latency I cells, P3HR1 (A gift from Dr. Elliott Kieff), Akata (A gift from Dr. Elliott Kieff), Mutu I (A gift from Dr. Jeffrey Sample) and Kem I (A gift from Dr. Jeffrey Sample), and latency III cells, GM12878 (purchased from Coriell Institute), GM12881 (purchased from Coriell Institute), Kem III (A gift from Dr. Jeffrey Sample) and Jijoye (purchased from American Type Culture Collection, ATCC), were all grown in Roswell Park Memorial Institute (RPMI) 1640 medium with 10% fetal bovine serum (FBS). 293T cells (purchased from ATCC) were grown in Dulbecco’s Modified Eagle’s Medium (DMEM) with 10% FBS. All B cell lines used in this study stably express Streptococcus pyogenes Cas9, which were generated by lentiviral transduction followed by blasticidin selection [109]. All cells were grown in a humidified chamber with 5% carbon dioxide at 37°C.

Cytokines and JAKi treatment

Latency I and III B cells were seeded at 500,000 cells/ml in 12-well plates and mock treated with PBS or cytokines (S3 Table) at 50 ng/ml and 100 ng/ml, respectively. EBV infected human primary B cells were treated with IL-15 or IL-21 at 100 ng/ml. For long-term treatment, cells were re-seeded with fresh culture medium supplemented with indicated cytokines, which were refreshed every 48 hours. For JAK inhibitor treatment, cells were pre-treated with JAK inhibitor I (JAKi) at indicated doses for one hour at 37 °C, followed by cytokine treatment, refreshed every 48 hours.

CRISPR/Cas9 editing

CRISPR/Cas9 editing was performed as previously described [110,111]. In brief, Brunello library [112] single guide RNA (sgRNA) were cloned into pLentiGuide-puro (a gift from Feng Zhang, Addgene plasmid #52963), pLenti-spBsmBI-sgRNA-Hygro (a gift from Rene Maehr, Addgene plasmid #62205), or pLentiGuide-zeo (a gift from Rizwan Haq, Addgene plasmid #160091). sgRNA sequences were verified by Sanger sequencing. All sgRNAs used in this study are listed in S4 Table. Target cells were transduced with lentivirus expressing sgRNAs against the target gene, or as a control, against GFP (pXPR-011, a gift from John Doench). Lentivirus were produced by transfection of 293T cells with pCMV-VSV-G (a gift from Bob Weinberg, Addgene plasmid #8454), psPAX2 (a gift from Didier Trono, Addgene plasmid #12260), and the sgRNA expression vector using the TransIT-LT1 transfection reagent. 293T supernatants were added to target B-cells at 48 and 72 hours post-293T transfection. Transduced cells were then selected with puromycin (3 μg/ml) for 3 days.

For STAT5A and STAT5B combinatorial editing, GM12878 or Jijoye Cas9+ cells were initially transduced with lentivirus expressing STAT5A sgRNA and selected with hygromycin (200 μg/ml) for 7 days, followed by transduction with lentiviruses expressing STAT5B sgRNA and selected with puromycin (3 μg/ml) for 3 days. For experiments with STAT5A/5B/3 editing, STAT3 was depleted in STAT5A/5B edited GM12878 cells by transduction with lentivirus expressing STAT3 sgRNA, and transduced cells were selected by zeomycin (200 μg/ml) for 7 days. On-target CRISPR effects were validated by immunoblotting.

cDNA cloning and transduction

cDNA entry vectors used in this study are listed in S4 Table, which were purchased from DNASU and Addgene. STAT1, 3, and 6 cDNA were sub-cloned into the destination vector pLX-TRC313 (a gift from John Doench) and STAT3_p.A662C_N664C (constitutively active STAT3 with A662C and N664C mutations) [44] was cloned into pLIX-402 (a gift from John Doench) by Gateway LR recombination. As described previously [113], the destination vector and donor vector containing the gene of interest were co-incubated with 1x LR Clonase Enzyme Mix overnight at room temperature. The reaction mixture was then transformed into Stbl3 competent cells and plated on LB agar plate containing ampicillin. Destination vectors were used to make lentiviruses, which were used to transduce target B-cells. Transduced cells were selected by puromycin or hygromycin for pLIX-402 or pLX-TRC313 vectors, respectively.

Immunoblotting

Immunoblotting analysis was performed as previously described [21], Cells were lysed in 1x Laemmli Sample Buffer and sonicated briefly. For detection of LMP2A, cells were lysed with M-PER Mammalian Protein Extraction Reagent and incubated on ice for 30 minutes. Lysates were centrifuged at 15,000 x g for 15 minutes. 2x Laemmli Sample Buffer was added into supernatant and boiled at 70 °C for 10 minutes. Lysates were resolved by SDS-PAGE and transferred onto nitrocellulose membranes, which were blocked with 5% nonfat milk in TBST buffer for 1 hour and then incubated with primary antibodies at 4 °C overnight. Blots were then washed 3 times with TBST, followed by secondary antibody incubation for 1 hour at room temperature. Blots were washed 3 times in TBST buffer and were developed with the ECL chemiluminescence substrate. Images were captured by a LI-COR Fc platform. All antibodies used in this study are listed in S3 Table.

Flow cytometry assay

Cells were washed once with FACS buffer (2% FBS v/v, PBS), followed by incubation with primary antibodies in FACS buffer for 30 minutes at room temperature in the dark. Labeled cells were palleted, washed twice and resuspended in FACS buffer into flow cytometry-compatible tubes and processed immediately. Flow cytometry data was recorded with a BD FACSCalibur instrument and analyzed with FlowJo X software.

Akata virus production, primary B cells isolation and infection

EBV was produced from EBV+ Akata cells. In brief, EBV+ Akata cells were resuspended in FBS-free RPMI media at 2–3 million cells/mL and induced with 0.25% (v/v) goat anti-human immunoglobulin G serum for 6 hours at 37 °C. Cells were then pelleted and resuspended in 4% FBS RPMI media and cultured in 37 °C for 3 days. Supernatant were then collected and filtered through 0.45 μM filter. Viruses were 50-fold concentrated by ultracentrifugation and stored at -80 °C until use.

Primary B-cells were isolated by negative selection from discarded, de-identified peripheral blood mononuclear cells from the Brigham and Women’s Hospital Blood Bank, obtained following platelet donation, using an Institutional Review Board approved protocol and written consent was obtained from the donors. RosetteSep and EasySep negative isolation kits were used according to the manufacturer’s instructions to isolate CD19+ B-cells. B cells were then cultured with RPMI containing 10% FBS. Primary B cells were seeded at 500,000 cells/ml and infected by the Akata EBV strain at multiplicity of infection (MOI) of 0.1, as determined by the Green Daudi assay.

Primary human B cell EBV transformation assay

EBV transformation assays were performed as described previously [114]. Briefly, purified human primary B cells were infected with Akata EBV using serial 10-fold dilutions. Cells were cultured with media containing DMSO or JAKi (200 ng/ml) and were seeded in 96-wells plates at 500,000 cells/ml (30 wells per condition). Media containing DMSO or JAKi was refreshed every three to four days. The percentage of wells positive for B-cell outgrowth at four weeks post infection was calculated and plotted relative to the dilution of virus.

Chromatin immunoprecipitation (ChIP) assay

After cytokine treatment, 10 million cells were cross-linked with 1% formaldehyde in 10 ml growth medium for 10 minutes, followed by quenching with 2.5M glycine in distilled water for 5 minutes. Cells were washed with ice-cold PBS three times and then lysed in 0.5 ml 1% SDS lysis buffer (50 mM Tris, 10 mM EDTA, 1% SDS), supplemented with 1x cOmplete, EDTA-free Protease Inhibitor Cocktail. Chromatin was fragmented using a Bioruptor Pico sonication device with 30s on/ 30s off (20 cycles for GM12878 cells, 12 cycles for Mutu I cells), and centrifuged at 13,200 rpm for 10 mins at 4 °C. This protocol resulted in fragments of average length 100–200 bp, to enable differentiation of STAT occupancy at closely spaced EBV genomic STAT binding sites. Supernatants were removed and then diluted 1:10 in ChIP dilution buffer (1.2 mM EDTA, 16.7 mM Tris, 167 mM NaCl, 0.01% SDS, 1.1% Triton X-100) supplemented with protease inhibitor cocktail. Chromatin from one million cells was used for each ChIP reaction. 1% of sonicated chromatin was saved as input and stored at -80 °C until use. Diluted chromatin was rotated overnight at 4 °C with the indicated antibody and 20 μl protein A+G magnetic beads. Next day, beads were pelleted, washed twice with a lower salt buffer (150 mM NaCl, 2 mM EDTA, 20 mM Tris, 0.1% SDS, 1% Triton X-100) and then a high-salt buffer (500 mM NaCl, 2 mM EDTA, 20 mM Tris, 0.1% SDS, 1% Triton X-100), and once with LiCl buffer (0.25 M LiCl, 1% NP-40, 1% sodium deoxycholate, 1 mM EDTA, 10 mM Tris) and finally TE buffer (10 mM Tris, 1 mM EDTA). Chromatin was eluted in Elution buffer (100 mM NaHCO3, 1% SDS) and reverse cross-linked at 65 °C for 2 hours. QIAquick PCR purification kits were used to purify the immunoprecipitated DNA, followed by qPCR with PowerUp SYBR green PCR master mix on a CFX Connect Real-Time PCR Detection System (Bio-Rad). All reagents, antibodies and primers used for ChIP are listed in S3 and S4 Tables.

Methylated DNA immunoprecipitation (MeDIP) assay

Genomic DNA was extracted using DNeasy Blood& Tissue Kit, followed with MeDIP assay with MagMeDIP kit, following the manufacturer’s protocol. qPCR was then performed with primers specifically target EBV promoters. All reagents and primers used for MeDIP are listed in S3 and S4 Tables.

RNA-seq and data analysis

mRNA was isolated via the RNeasy Mini kit with in-column genomic DNA digestion protocol was followed, according to the manufacturer’s instructions. To construct indexed libraries, 1 μg of total RNA was used for polyA mRNA purification, using the NEBNext Poly(A) mRNA Magnetic Isolation Module, followed by library preparation using the NEBNext Ultra RNA Library Prep with Sample Purification Beads. Each experimental treatment was performed in biological triplicate. Libraries were multi-indexed, pooled and sequenced on an Illumina NovaSeq 6000 using PE150 Sequencing Strategy by Novogene Corporation. Adaptor-trimmed reads were mapped to Akata EBV genome (Accession#: KC207813.1) or human GRCh37.83 transcriptome assembly using salmon (v1.10.0). Quality control was performed using fastqc. Differentially expressed genes were identified in R (v4.0.3) using DESeq2 [115] under default settings with the apeglm shrinkage estimator (https://doi.org/10.1093/bioinformatics/bty895) and annotations derived from the hg19 build from Ensembl release 75 [116] and accessed via biomaRt. Volcano plots were generated in GraphPad Prism 8, using Log₂ (Fold Change) and −Log₁₀ (p value) data. Differentially expressed genes from each condition were subjected to Enrichr analysis and top 10 KEGG pathways with adjusted p value < 0.05 cutoff were visualized. All RNA-seq datasets were deposited to the NIH GEO omnibus with accession number GSE252480. All reagents and kits used for RNA-seq are listed in S3 Table.

Quantification and statistical analysis

All immunoblots were performed with three independent experiments and qPCR was performed in three independent experiments. Statistical significance was assessed with Student’s t test using GraphPad Prism 8 software, where NS = not significant, p > 0.05; * p < 0.05; ** p < 0.01; *** p < 0.001. Biorender was used to create the schematic models.