Methionine Metabolism Controls the B cell EBV Epigenome and Viral Latency

Methionine restriction de-represses EBV lytic and latent genes

The extent to which Burkitt cells require high levels of methionine uptake is unknown. Methionine serum concentrations normally range from 10–30 μM, but can be significantly impacted by diet (Schmidt et al., 2016). By comparison, the methionine concentration in Roswell Park Medical Institute 1640 (RPMI) medium commonly used for in vitro B cell culture is ~100 μM. To gain insights into methionine metabolism roles in control of the EBV epigenome and transcriptome, we investigated methionine restriction (MR) effects on a panel of Burkitt cells with well-defined EBV latency programs. We first tested MR effects Mutu I cells, which were established from a latency I African BL tumor (Gregory et al., 1990). Culture for 72 hours in RPMI with 1 μM methionine induced LMP1 and LMP2 (Figure 1A–B, S1A), whose expression is typically repressed by DNA methylation in EBV latency I (Falk et al., 1998; Guo et al., 2020b; Tao et al., 1998). Unexpectedly, MR failed to de-repress EBNA2, and only modestly induced EBNA3 expression (Figure S1B), perhaps indicating a lower threshold for de-repression of the LMP promoter. We next tested MR effects on P3HR-1 cells, in which a genomic deletion knocks out EBNA2, the major viral activator of the LMP promoter. Despite the absence of EBNA2, MR highly de-repressed LMP1 even when media was refreshed daily, suggesting that methionine metabolism is required for silencing of the Burkitt LMP1 promoter independently of effects on EBNA2 (Figure S1B–C).

We next analyzed Burkitt cell responses to a wider range of methionine concentrations. LMP1 de-repression, and induction of its downstream transcriptional target TRAF1 (Zhao et al., 2014), were observed in all three cell lines, indicating activation of LMP1/NF-κB pathways in a dose-responsive manner (Figure 1C). Similarly, MR induced plasma membrane expression of LMP1/NF-κB target ICAM-1 in all five EBV+ Burkitt cell lines tested, but not in a closely matched EBV negative Akata cell line or in EBV negative Reh acute lymphoblastic leukemia B cells (Figure S1D). MR de-repressed ICAM-1 expression in P3HR-1 in which media was refreshed daily (Figure S1E).

DNA methylation is necessary for EBV lytic gene repression. We therefore analyzed immediate early BZLF1 and early BMRF1 gene expression, both of which were induced by MR in a dose-responsive manner (Figure 1C). Flow cytometry (FACS) analysis of 7-Aminoactinomycin D (7-AAD) vital dye uptake indicated little effect on BL viability at day 3 post MR, while CFSE dye-dilution and cell cycle analyses showed that MR caused a G1 growth arrest, despite EBV oncoprotein induction (Figure S1F–H). Culture in human plasma-like media (HPLM), which has 30μM methionine and which contains physiologically relevant concentrations of additional plasma components, can alter lymphocyte physiology (Cantor et al., 2017; Leney-Greene et al., 2020). Interestingly, acutely shifting P3HR-1 cells from culture in RPMI to HPLM with dialyzed fetal bovine serum de-repressed lytic BZLF1, and to a lesser extent BMRF1, but not LMP1 expression (Fig. S1I). These results may reflect an abortive EBV lytic cycle, perhaps triggered by the reduction in methionine from 100 to 30 μM, together with additional acute metabolite shifts.

To more broadly define MR effects on EBV and host gene expression, RNAseq was performed in P3HR-1 cells grown in 100 vs 10 μM methionine for 72 hours. MR significantly de-repressed 60 EBV lytic genes as well as LMP1, LMP2 and EBNA3 transcripts (Figure 1D, S2A and Supplementary Tables S1). Effects on LMP1, early BMRF1 and late BLLF1 mRNAs were validated by qPCR (Figure S2B). MR may perturb EBNA3 expression at the protein level, given apparently greater increases in EBNA3 mRNA than protein levels. Concordant effects were observed in Rael Burkitt cells, in which DNA methylation enforces a tight latency I phenotype (Dalton et al., 2020; Ernberg et al., 1989; Guo et al., 2020b; Masucci et al., 1989) (Figure S2A and C, Supplementary Tables S1). Notably, while Rael EBV genomes have an intact EBNA2 locus, EBNA2 mRNA abundance was not significantly increased by MR, even though MR increased EBNA3, LMP and lytic gene expression. EBNA-LP mRNA abundance was also only modestly increased, suggesting a methionine metabolism role in control of the mRNA encoding EBNA2 and LP (Figure S2A). Host gene ontology analysis identified that MR significantly upregulated NF-κB pathway targets, suggestive of signaling by de-repressed LMP1 (Figure 1E, S2D). RNAs encoding ribosomal components were strikingly downregulated in both P3HR-1 and Rael cells (Figure 1E, S2D, Supplementary Tables S1), indicating unexpectedly strong cross-talk between methionine and ribosomal metabolism in Burkitt cells.

Despite immediate early and early gene expression, MR only weakly induced EBV late genes (Fig 1D, Supplementary Tables S1). For instance, MR more robustly de-repressed transcripts of the immediately early gene BZLF1 and early gene BMRF1 than the late gene BLLF1 (Figure S2E). To further examine this result, we performed FACS analysis of P3HR-1 cells grown in 10 vs 100 μM methionine. MR did not significantly increase late gene gp350 (encoded by BLLF1) plasma membrane abundance, in contrast with a full lytic cycle induced by conditional BZLF1 expression and sodium butyrate (Figure 1F). We therefore asked whether MR induced EBV lytic DNA replication, which is necessary for late gene expression. Culturing P3HR-1 cells with 10 μM methionine only modestly increased EBV DNA copy number, in comparison with positive control activation of a conditional BZLF1 allele together with HDAC inhibitor sodium butyrate treatment (Figure 1G). These results indicate that MR generally results in an abortive lytic cycle.

MR induction of LMP1 and LMP2, in the absence of EBNA2, raised the question of whether EBV lytic reactivation was responsible for their expression. In support, LMP1 is induced in both epithelial and B cells upon lytic reactivation (Webster-Cyriaque et al., 2000; Yuan et al., 2006). To test this hypothesis, we generated CRISPR BZLF1 knockout Mutu I cells to prevent lytic reactivation, and as a control for EBV episome CRISPR editing, also generated Mutu I knocked out for the late gene BLLF1. We confirmed that BZLF1 KO, but not BLLF1 KO, prevented induction of the early BMRF1 antigen in response to immunoglobulin-crosslinking (Figure S2F). Despite being unable to trigger lytic gene expression, as evidenced by lack of BMRF1 induction, MR strongly induced LMP1 in BZLF1 KO cells, suggesting latent cycle LMP promoter activation by a host transcription factor reminiscent of the EBV latency II program (Figure 1A, S2G).

To gain insights into how MR alters Burkitt cell metabolic pathways, we performed liquid chromatography mass spectrometry (LC/MS) metabolomic analyses of P3HR-1 cells grown in RPMI containing 100 vs 10 μM for 72 hours. As expected, methionine abundance was amongst the most strongly downregulated by MR, as were the methionine cycle metabolites SAM, SAH and homocysteine. Intracellular SAM and SAH were reduced about 223 and 11-fold, respectively, indicating that the cellular methylation potential was decreased by ~9-fold (Figure 1H–I, S3A, Supplementary Tables S2). Notably, MR upregulated the abundance of α-ketoglutarate (αKG), which serves as a necessary cofactor for TET family DNA demethylases, but decreased levels of succinate, which instead inhibits TET demethylase activity (Figure 1G, Supplementary Tables S2). MR also downregulated metabolites of Burkitt purine and pyrimidine, TCA and transsulfuration pathways, the latter of which uses methionine for de novo cysteine synthesis (Figure 1J). These results highlight that the Burkitt B cell metabolome is highly influenced by extracellular methionine abundance, which in turn strongly contributes to latency I maintenance.

MR alters the EBV epigenome and reprograms the host epigenetic factor landscape

Given MR effects on the cellular methylation potential, we next investigated how MR alters the EBV-infected cell DNA and histone methylation epigenetic landscapes. 5-methylcytosine (5mC) dot blot analysis identified significant decreases in global DNA methylation in methionine restricted cells (Figure 2A). Similarly, immunoblot for histone methylation epigenetic marks demonstrated reduced histone 3 lysine 9 tri-methyl (H3K9me3) and histone 3 lysine 4 tri-methyl H3K4me3 marks. However, repressive histone 3 lysine 27 tri-methyl (H3K27me3) levels were not significantly changed by 72 hours of MR (Figure 2B), indicating that histone epigenetic marks were not universally diminished.

MR effects on epigenetic reader and writer expression could also affect the EBV epigenome. Indeed, MR highly perturbed levels of mRNAs encoding key epigenetic reader and writers in both P3HR-1 and Rael cells (Figure 2C, Supplementary Tables S1). MR significantly decreased levels of the initiator DNA methyl transferase DNMT3B and maintenance of DNA methylation enzymes UHRF1 and DNMT1. By contrast, MR increased mRNA levels of the TET2 and TET3 demethylases, which have important roles in support of EBV latency III gene expression and in suppression of EBV lytic genes (Guo et al., 2020a; Guo et al., 2020b; Lu et al., 2017; Wille et al., 2017b; Zhang et al., 2020) (Figure 2C). TET demethylase activity requires the substrate αKG and is competitively inhibited by succinate (Xiao et al., 2012). Since MR increased αKG and decreased succinate levels, changes in the αKG/succinate ratio may further increase TET activity and contribute to Burkitt epigenetic reprogramming. Furthermore, EBV C promoter UHRF1 occupancy was significantly reduced in P3HR-1 cells cultured in RPMI with 10μM methionine (Figure 2D). Similarly, expression of the H3K9me2 histone methyltransferase G9a (encoded by EHMT2) was decreased by MR. Effects on G9a protein abundance were also evident on the protein level (Figure 2E). Since H3K9me2 has important roles in propagation of DNA methylation marks by the epigenetic enzyme UHRF1, MR-driven loss of G9a may synergistically or additively hypomethylate Burkitt cell genomes.

MYC is a major regulator of B cell immunometabolism (Stine et al., 2015). MYC expression was strongly suppressed by MR, which we confirmed by immunoblot analysis (Figure 2F). Pronounced MR effects on MYC and G9a expression raised the question of how MR more globally alters EBV-infected B cell translation. To test this, we used a puromycin pulse labeling assay, in which puromycin incorporation into elongating nascent chains is visualized by an anti-puromycin antibody (Schmidt et al., 2009). Whereas MR strongly reduced protein translation levels in multiple diffuse large B cell lymphoma (DLBCL) and an acute lymphocytic leukemia cell line, reduction of media methionine concentration by 90% for 72 hours did not strongly impair translation in the three EBV+ Burkitt cell lines examined (Figure 2G). These results are consistent with the observation that MR induces expression of EBV latency and lytic genes on the protein level, and raise the question of whether EBV may alter SAMTOR sensing in Burkitt cells.

To gain insights into methionine metabolism roles in control of Burkitt EBV genomic methylation, we performed methylated DNA immunoprecipitation (MeDIP) of cells grown in 100 vs 10 μM methionine. As expected, the EBV C and LMP latency promoters, as well as BZLF1 and BMRF1 lytic promoters, were highly methylated under methionine replete conditions. By contrast, incubation in 10 μM methionine for three days significantly decreased DNA methylation levels at each of these key EBV promotors (Figure 2H, S3B). Since EBV genome methylation is critical for maintenance of latency I, these results indicate that EBV+ Burkitt cells are reliant on methionine pathway flux to silence LMPs and to maintain latency.

EBV genomic H3K9me3 and H3K27me3 marks contribute to silencing of lytic and latency III genes (Guo et al., 2020b; Ichikawa et al., 2018; Murata and Tsurumi, 2013). To define MR effects on these repressive chromatin marks, we performed chromatin immunoprecipitation (ChIP)-qPCR analysis on extracts from cells grown in 100 vs 10 μM methionine. MR significantly perturbed repressive H3K9me3 and H3K27me3 signals at the C, LMP, BZLF1 and BMRF1 promoters in both P3HR-1 and Mutu I cells (Figure 2I, S3C). MR more strongly affected H3K9me3 than H3K27me3 levels at these key promotors. Notably, levels of the H3K27 methyltransferase EZH2 were not significantly diminished by MR in multiple B cell lines, including EBV+ BL (Figure 2G).

The methionine cycle is necessary for maintenance of highly restricted EBV latency I

The methionine cycle produces the universal methyl donor SAM and also regenerates methionine from SAH via homocysteine (Cantoni, 1985; Dai et al., 2018; Sanderson et al., 2019) (Figure 3A). To define whether flux through the methionine cycle is necessary for maintenance of EBV latency I in Burkitt cells, we used CRISPR/Cas9 editing to deplete MAT2A, the enzyme that catalyzes SAM production from methionine and ATP. We analyzed cells at an early timepoint following CRISPR editing, prior to the onset of cytotoxic effects of MAT2A inhibition. MAT2A targeting by either of two single guide RNAs (sgRNAs) de-repressed LMP1 expression to levels similar to those observed in Mutu III cells, a subclone of Mutu I with the latency III program (Figure 3B). As observed with MR, MAT2A depletion did not robustly induce latency III EBNA antigens, but did de-repress BZLF1 (Figure 3B). De-repressed LMP1 induced plasma membrane ICAM-1 (Figure 3C–D).

The enzyme adenosylhomocysteinase (AHCY) converts SAH to L-homocysteine, which can be metabolized to regenerate methionine or can instead be used by the transsulfuration pathway for cysteine biosynthesis (Figure 3A). CRISPR AHCY targeting de-repressed LMP1 and BZLF1, suggesting that even at a step downstream of SAM generation, methionine cycle flux regulates EBV latency and lytic antigens (Figure 3E). RNAseq identified that whereas both AHCY or MAT2A knockout de-repressed EBV latency and lytic transcripts, effects were stronger on the mRNA level with MAT2A editing (Figure 3F, Supplementary Tables S3).

To gain insights into MAT2A versus AHCY roles in Burkitt methionine metabolism, we performed LC/MS metabolomic analyses of Rael cells expressing control, MAT2A or AHCY sgRNAs. Intracellular metabolites were profiled at day six of sgRNA expression, prior to the onset of cytotoxic effects. MAT2A depletion reduced SAM, SAH and the SAM:SAH methylation potential, but increased methionine and methionine sulfoxide levels (Figure 3G, Supplementary Tables S3). AHCY depletion decreased homocysteine levels, but had little effect on SAM abundance. Interestingly, AHCY depletion significantly increased SAH levels (Figure 3G, Supplementary Tables S3). Thus, we speculate that increased SAH, which is a potent methyltransferase inhibitor in lymphocytes (Yi et al., 2000), together with decreased SAM:SAH, may underlie the observed AHCY knockout effects on the EBV epigenome and transcriptome. However, stronger MAT2A knockout effects on SAM:SAH may more robustly de-repress EBV latency and lytic genes.

Similar to MR, viral load analysis suggested that MAT2A or AHCY KO caused a largely abortive EBV lytic cycle, with only low-grade increases in EBV genome copy number (Figure 3H). MeDIP-qPCR identified that repressive 5mC LMP and BZLF1 promoter marks were significantly decreased in cells depleted for either MAT2A or AHCY (Figure 3I). Collectively, these results support a model in which maintenance of EBV latency I is dependent upon flow of carbon units from methionine through the methionine cycle to the EBV epigenome.

EBV-driven methionine metabolism upregulation supports primary B cell transformation

EBV robustly upregulates MAT2A and AHCY upon primary human B cells infection (Figure 4A) (Wang et al., 2019a; Wang et al., 2019b). To gain insights into how EBV alters methionine metabolism in newly infected cells, we performed LC/MS metabolomic profiling of B cells at rest versus at Days 1 or 2 post-EBV infection, early timepoints in which EBV-causes B cell activation and remodeling, but prior to the onset of B cell proliferation. EBV significantly increased methionine, SAM and SAH abundances by 1 DPI (Figure 4B–C, S4A, Supplementary Tables S4). We next profiled cells at Day 10 post-infection, a timepoint in which EBV has transformed infected B cells into activated lymphoblasts. SAM levels and the SAM:SAH ratio further increased at 10 DPI. Decreased SAH and increased homocysteine levels observed at Day 10 may have resulted from elevated AHCY expression and/or activity. Hierarchical clustering highlighted widespread metabolome remodeling between Days 0 versus 1 post-infection, and also between Days 1/2 versus 10 post-infection (Figure S4B, Supplementary Tables S4). Metabolomic pathway impact analysis identified methionine metabolism as highly rewired by EBV at both 1 and 10 DPI relative to uninfected cells, whereas glutathione metabolism was more highly impacted by EBV at 10 DPI. Purine metabolism was highly upregulated at both timepoints (Figure S4C).

We next defined MR effects on EBV genes in newly infected B cells. Reduction of extracellular methionine concentration from 100 to 10 μM suppressed EBNA2 induction, which is essential for EBV-driven B cell outgrowth (Pich et al., 2019), as well as CD23 upregulation downstream of EBNA2 (Wang et al., 1987) (Figure 4D–E). Unexpectedly, MR also delayed maximal EBNA2 induction to a later timepoint, raising the question of how newly infected B cells survive under MR conditions (Figure 4D). In contrast to Burkitt cells, MR failed to induce EBV lytic gene expression over the first 14 days of infection, perhaps in part because DNA methylation levels are considerably lower in resting and in newly infected B cells than in Burkitt cells. Transformation assays demonstrated that MR strongly impaired LCL outgrowth (Figure 4F). Interestingly, MR effects on primary B cells were stimulus specific. While MR limited outgrowth of B cells activated by anti-immunoglobulin crosslinking together with the Toll-like receptor 9 agonist CpG, it did not block outgrowth of cells stimulated by CD40 ligand (CD40L) in combination with interleukin-4 (IL4) (Figure 4G–H). As CD40L/IL4 is typically provided by T-cells, our results raise the possibility that T-cell dependent versus independent B cell responses may differ in their requirements for extracellular methionine.

One-carbon folate metabolism is necessary for Burkitt latency I

We hypothesized that folate and one-carbon metabolism enzymes contribute to Burkitt latency I maintenance by supplying methyl groups for the methionine cycle (Figure 5A) (Ducker et al., 2017; Parsa et al., 2020; Sanderson et al., 2019). In support, the one-carbon enzymes MTHFD1 and MTHFD2 were hits in our recent human genome-wide CRISPR screen for host factors that support latency I maintenance (Guo et al., 2020b) (Figure S5A). We therefore blocked the cytosolic and mitochondrial one-carbon metabolism pathways with the highly selective serine hydroxymethyl-transferases (SHMT) antagonist SHIN1 (Ducker et al., 2017) (Figure 5A). SHIN1 is a folate-competitive inhibitor that blocks the cytosolic SHMT1 and mitochondrial SHMT2 isoforms, and their functional redundancy may have precluded either from scoring in our CRISPR screen. SHIN1 de-repressed LMP1 and BZLF1 in all three EBV+ Burkitt cell lines tested. Latency I was maintained when SHIN1 was added together with the one-carbon pathway product formate, suggesting on-target SHIN1 effects (Figure 5B). We also used CRISPR to deplete SHMT1, SHMT2 or both in Rael Burkitt cells. Whole cell lysates were collected at Day 4 post-sgRNA expression. Consistent with our SHIN1 results, LMP1 de-repression was observed in SHMT1/2 double KO cells (Figure S5B). Furthermore, SHIN1 induced LMP1 target gene ICAM1 plasma membrane expression in Mutu I cells, an effect that was stronger when cells were grown with dialyzed fetal bovine serum to reduce extracellular formate (Figure 5C, S5C). Similar results were obtained in Rael Burkitt cells (Figure S5D–E).

To identify SHIN1 effects on global EBV gene expression, we performed RNAseq. Reminiscent of results obtained with MR, SHIN1 de-repressed most EBV genes, which was rescuable by formate (Figure 5D, Supplementary Tables S5). qPCR analysis validated SHIN1 effects on latency gene LMP1, early gene BMRF1 and late gene BLLF1 mRNAs (Figure S5F). Gene ontology analysis indicated that SHIN1 significantly induced NF-κB target genes, consistent with LMP1 de-repression (Figure 5E). Cross-comparison of RNAseq datasets from MR and SHIN1-treated Burkitt P3HR-1 cells identified concordant transcriptome-wide responses (R²= 0.9512) (Figure 5F). Mechanistically, SHIN1 strongly reduced repressive histone H3K9me3 and H3K27me3 marks at the EBV genomic LMP1, C and BZLF1 promoters (Figure S5G).

To gain further insights into SHIN1 effects on the Burkitt cell metabolome, we performed LC-MS analysis of soluble metabolites extracted from Rael cells, treated with vehicle versus SHIN1 for 72 hours. Consistent with on-target blockade of serine catabolism, serine levels were increased ~18-fold in SHIN1 treated cells. Likewise, levels of the de novo purine intermediate aminoimidazole carboxamide ribonucleotide (AICAR), whose conversion by ATIC to 5-formamido-4-imidazolecarboxamide ribonucleotide (FAICAR) requires donation of a methyl group via a 10-formyl-tetrahydrofolate donor, was increased by ~253-fold (Figure 5G, Supplementary Tables S5). ATIC product inosine monophosphate (IMP) levels were instead significantly suppressed by SHIN1, which was partially reversed by formate supplementation. Intracellular methionine levels were significantly increased in SHIN1 treated cells, whereas intracellular SAM and SAH levels were instead significantly reduced. Formate supplementation partially restored intracellular SAM, but not SAH levels in SHIN1-treated cells (Figure 5G, Supplementary Tables S5).

We asked whether restriction of the non-essential amino acid serine, or as a control glycine, alters EBV gene expression. Mutu I were cultured in serine and glycine free RPMI with dialyzed fetal calf serum. Consistent with SHIN1 effects, serine, but not glycine restriction, de-repressed LMP1 and BZLF1, effects that were rescued by serine add back (Figure 5H). Serine, but not glycine restriction, hypomethylated the LMP1, BZLF1 and to a somewhat lesser extent C promoters (Figure 5I, S5H). Similarly, serine restriction significantly reduced H3K9me3 and H3K27me3 repressive chromatin marks at the LMP1, C and BZLF1 promoters (Figure 5J, S5I).

Dietary MR alters the Burkitt metabolome and de-represses EBV antigens in vivo

Since methionine is an essential amino acid, dietary methionine restriction can be used to reduce plasma methionine levels, including in murine models (Gao et al., 2019). Given our in vitro results, we hypothesized that Burkitt tumors are dependent upon abundant methionine supply in vivo to regulate the EBV epigenome and viral gene program. As a pilot study, three non-obese diabetic/severe combined immunodeficiency (NOD-SCID) NSG mice were fed either a control diet with a standard 0.86% weight by weight (w/w) methionine or a MR diet with 0.086% w/w methionine, but that contained balanced total amino acid levels (Figure S6A). At one week, dietary MR lead to a nearly 50% reduction in plasma methionine levels (Figure S6B), consistent with recent mouse model MR studies (Epner et al., 2002; Roy et al., 2020). At 2 weeks after initiation of control or MR diets, Mutu I cells were implanted into mouse flanks. Body weight did not differ significantly between mice on control or MR diets over the next three weeks, suggesting that either diet was generally well tolerated (Figure S6C). However, tumor volumes were smaller at each timepoint in mice on the MR diet (Figure S6D).

We next tested MR effects on EBV+ Burkitt cell metabolomes, EBV gene expression and the EBV epigenome. Bilateral flank Mutu I tumors were implanted and allowed to grow for two weeks in twelve NSG mice fed the control diet. At that point, six mice were continued on the control diet and the other six were switched to the MR diet. At weeks four and five, mice were humanely sacrificed and samples were obtained (Figure 6A). Dietary MR significantly reduced tumor and plasma methionine and methionine sulfoxide (the oxidized form of methionine) levels at week five (Figure 6B–C, S6E). Likewise, MR reduced tumor SAM and SAH levels (Figure 6B, S6E). Body weights were similar on control and MR diet fed mice (Figure S6F).

To define dietary effects on EBV gene expression, qPCR was performed on explanted tumors from mice fed control versus MR diets. At both days 7 and 14, tumors from MR-fed mice exhibited significantly increased levels of the lytic immediate early BZFL1, early BMRF1 and interestingly also late BLLF1 genes. LMP1 mRNA levels were similarly de-repressed (Figure 6D). BMRF1 and LMP1 were upregulated at the protein level in tumors taken from five of six dietary methionine restricted mice. EBNA2 and EBNA3C proteins were also de-repressed in a subset of tumors (Figure 6E). Immunohistochemical analysis demonstrated BMRF1 de-repression in many tumor cells (Figure 6F–G, S7A). Mechanistically, the MR diet significantly reduced 5mC levels at the C, LMP and BMRF1 promoters, suggesting that dietary intervention alone can impact tumor methionine supply to the point of altering the EBV epigenome (Figure 6H).

Finally, we tested effects of the SHMT1/2 antagonist SHIN2, a second generation SHIN1 derivative with improved in vivo pharmacokinetic properties (García-Cañaveras et al., 2021), in the Mutu I xenograft model (Figure 6I). We confirmed that SHIN2 had similar effects as SHIN1 on EBV LMP1 and lytic gene de-repression (Figure S7B–C). We validated SHIN2 on-target effects on EBV+ Rael Burkitt cell metabolomes by LC/MS profiling. SHIN2 significantly increased serine and AICAR, but decreased IMP, SAM and SAH levels (Figure S7D, Supplementary Tables S6). Overall, responses to SHIN1 and SHIN2 were highly concordant (Figure S7E).

Tumors were implanted, allowed to expand for three weeks in control diet fed mice, and mice were then given a single intraperitoneal injection of SHIN2 (200mg/kg) or vehicle control. Tumor samples were analyzed at 3 and 7 days post-SHIN2 dosing. Similar to effects observed with MR, SHIN2 de-repressed LMP1, BZLF1 and BMRF1 at the protein level, particularly at day 3 post administration (Figure 6J, S7F–G). SHIN2 effects on LMP1 induction were evident in two of three tumors at Day 7 post-dosing (Figure 6J). Thus, flux through interconnecting folate one-carbon and methionine cycle metabolism supports the latency I maintenance in Burkitt tumors.

Discussion

Despite its position at the nexus of host cell metabolism and epigenetic reactions, little has been known about methionine cycle effects on EBV, or on virus/host interactions more generally. Building upon recent findings that EBV highly remodels host B cell metabolic pathways, our results highlight key roles of interconnecting methionine and serine metabolism (Maddocks et al., 2016) on the double stranded DNA EBV epigenome. Consequently, perturbation of B cell methionine uptake or metabolism altered EBV genome programs in Burkitt cells and impaired EBV-driven primary B cell transformation (Figure 7).

Burkitt lymphomas are the fastest growing human tumor (Gewurz et al., 2021; Thorley-Lawson and Allday, 2008). With each cell cycle, host factors must resynthesize and epigenetically reprogram the 10–400 EBV genomes per cell. DNA methylation has critical roles in silencing latency III and lytic antigens, which enables EBV+ Burkitt tumors to evade antiviral immune responses (Ambinder et al., 1999; Bergbauer et al., 2010; Gewurz et al., 2021; Guo et al., 2020b). Our observations suggest that continual methionine and folate methyl cycle flux is required to meet high Burkitt demand for methyl donor groups. Superimposed effects of methionine restriction on host epigenetic enzyme expression, including down-modulation of UHRF1 and DNMT1 and upregulation of TET demethylases, likely further contribute to EBV genomic hypomethylation.

MR or CRISPR methionine cycle perturbation similarly altered 5mC and histone methylation marks at the EBV C and LMP promoters. Interestingly, at the mRNA level, we observed upregulation of messages encoding EBNA3A, 3B and 3C, but not EBNA2 or LP mRNAs, even in Rael cells which have intact EBNA2 and EBNA-LP genes. Since EBNA2, LP and 3 are expressed from a common C or W promoter driven transcript, we speculate that methionine metabolism also has important roles in EBNA post-transcriptional regulation, perhaps at the level of splicing. In support, methionine-derived SAM is the methyl donor for RNA methylation reactions, and RNA methylation is increasingly recognized as a key regulator of EBV-encoded mRNAs (Dai et al., 2021; Lang et al., 2019; Xia et al., 2021; Zheng et al., 2021). EBNA2 mRNAs are modified by N6-methyladenosine (m6A) marks, and knockdown of the m6A writer METTL3 decreases EBNA2 expression levels (Zheng et al., 2021). Dietary MR variably induced EBNA2 and 3C in Burkitt xenografts, albeit less robustly than LMP1, suggesting that these immunogenic EBNA antigens may be more inducible by MR in vivo, perhaps due to additional microenvironmental effects.

MR robustly induced EBV immediate early and early lytic genes, but only modestly induced late genes. MYC repression may have contributed to EBV lytic gene de-repression (Guo et al., 2020a), together with hypomethylation. These findings raise the possibility that methionine metabolism supports EBV lytic DNA replication, which is required for EBV late gene expression (Chakravorty et al., 2019; Kenney and Mertz, 2014; Li et al., 2018). An abundant supply of methionine and methionine-derived SAM may be required for modification of host or viral targets with obligatory roles in EBV lytic DNA replication. Metabolomic pathway impact analysis identified significant MR effects on purine and pyrimidine metabolism, which may perturb EBV lytic DNA synthesis and late gene expression. It will be of interest to test whether serine metabolism is similarly important for EBV lytic DNA synthesis, such as to provide one carbon units and ATP for nucleotide synthesis (Fan et al., 2014; Huhtanen and Pensack, 1967; Maddocks et al., 2016).

Cancer cells alter methionine metabolism and increase their reliance on extracellular methionine, known as the Hoffman effect (Hoffman et al., 2019). Our findings suggest that EBV likewise subverts methionine metabolism pathways to support both the lytic and latent cycles in B cells. Given their metabolism regulatory roles, we speculate that EBNA2 and MYC are important methionine metabolism regulators in newly infected cells. In support, EBNA2 is the only EBV protein coding gene required for primary B cell growth transformation over the first 8 days of infection (Pich et al., 2019). Aberrantly elevated MYC levels may similarly drive Hoffman metabolism in Burkitt cells with the latency I program.

Epigenetic enzymes are regulated by substrate and product concentrations (Dai et al., 2020; Lu and Thompson, 2012). Dietary methionine restriction achieved approximately 50% reduction in plasma methionine, significantly reduced xenograft methionine abundance and tumor volumes, and diminished EBV genomic 5mC levels at the C, LMP1 and BMRF1 promoters. Yet, MR diets for up to eight weeks have minimal impact on spleen or lymph node T, B or NK cell numbers, despite similar reduction in plasma methionine to what we observed (Epner et al., 2002; Roy et al., 2020). MR also did not significantly affect CD4+ T-helper 1 (Th1) lymphocyte proliferation or histone epigenetic activation marks. This may relate at least in part to the observation that antigen activated T-cells upregulate plasma membrane methionine transporter expression (Sinclair et al., 2019). These observations raise the possibility that EBV latently infected B cells are particularly sensitive to MR, and that a therapeutic window may exist for implementing MR together with anti-EBV T-cell immunotherapies.

EBV is a major trigger for multiple sclerosis (MS), in which autoreactive immune responses attack nerve fiber myelin sheaths (Bjornevik et al., 2022; Lanz et al., 2022; Robinson and Steinman, 2022). Interestingly, dietary methionine restriction ameliorates neuroinflammation and disease onset in the mouse experimental autoimmune encephalomyelitis (EAE) model of MS. 33% of mice on an MR diet exhibited symptoms at two weeks post EAE induction challenge, as compared with 90% in animals fed a control diet (Roy et al., 2020). Mechanistically, dietary MR reduces inflammatory T helper 17 (Th17) cell H3K4me3 activating promoter marks, proliferation and cytokine production (Roy et al., 2020). By contrast, administration of LCLs, or even an EBNA1 peptide implicated in molecular mimicry with the adhesion molecule GlialCAM, exacerbates EAE (Lanz et al., 2022; Polepole et al., 2021). It will therefore be of interest to investigate dietary MR effects on EAE in humanized mouse models of EBV infection. Use of humanized mouse models carrying MS genetic risk factor HLA-DRB1*15:01 (HLA-DR15) allele may be particularly informative (Zdimerova et al., 2021). Given our observation that MR impairs the outgrowth of newly EBV infected B cells, it is plausible that MR could synergistically reduce B cell EBNA1 abundance as well as pathologic Th17 responses.

Current Burkitt lymphoma therapies do not harness the presence of large viral genomes in EBV+ tumors. Our results raise the possibility that lowering extracellular methionine levels through dietary MR, or perhaps by orally-administered methioninase (Yamamoto et al., 2022), may sensitize tumor cells to attack by anti-EBV cytotoxic T-cells. Indeed, B cell LMP1 expression can elicit potent antitumor T-cell responses (Choi et al., 2021). Similarly, MR mediated de-repression of EBV lytic antigens may sensitize tumor cells to ganciclovir, whose cytotoxicity is activated by an EBV-encoded lytic cycle early gene kinase (Meng et al., 2010). Given our observation that MR restrains EBV-driven B cell immortalization, it will also be of interest to investigate whether dietary MR can be used in prophylaxis approaches for the prevention of EBV+ post-transplant lymphoproliferative diseases.

In summary, our results suggest that EBV-infected B cells are highly sensitive to perturbation of extracellular methionine or serine concentrations, or blockade of interconnecting methionine and folate one-carbon metabolism cycles. Disruption of the flow of methyl groups from methionine onto the EBV epigenome de-represses the Burkitt B cell latency I state. MR de-repressed EBV latent membrane proteins and an abortive lytic cycle, largely comprised of highly immunogenic immediate early and early antigens (Taylor et al., 2015). MR also impaired EBV-mediated primary B cell growth transformation.

Limitations of Study

A caveat is that many of our studies used well-characterized but passaged human tumor derived-Burkitt lymphoma cell lines. It remains possible that differences in responses to MR exist in freshly isolated Burkitt tumors or in patient-derived xenograft models. Similarly, our xenograft models utilized mice with severe combined immunodeficiency, and therefore did not capture potential superimposed effects of adaptive immune anti-tumor responses that may affect the microenvironment. These may be better captured with use of human leukocyte antigen matched humanized mouse xenograft models. Relatedly, it remains to be defined whether MR de-repression of EBV lytic antigens sensitizes tumors to killing by EBV-epitope specific T-cells or to treatment with the antiviral ganciclovir, which is activated by an early lytic EBV kinase (Meng et al., 2010). While EBV infects epithelial cells, B cells and more rarely other cell types including T and NK cells, our studies focused on B cell infection models. It will be of interest to define whether methionine and folate metabolism have similarly important roles in other EBV disease contexts, including in nasopharyngeal and gastric carcinomas, which exhibit extreme hypermethylation and highly restricted forms of EBV latency. We identified that Burkitt cell culture in HPLM, which contains 30 μM methionine, de-repressed BZLF1 and to a lesser extent BMRF1. It will therefore be of interest to pinpoint additional nutritional differences between RPMI1640 and HPLM that contribute to this phenotype, as well as to test the effects of further reduction of methionine concentrations in the HPLM system. Finally, single cell RNAseq analyses could provide additional insights into the heterogeneity of MR effects on Burkitt host cell and viral transcriptomes.

RESOURCE AVAILABLITIY

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Rui Guo(rguo2@bwh.harvard.edu), Benjamin Gewurz (bgewurz@bwh.harvard.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

METHOD DETAILS

QUANTIFICATION AND STATISTICAL ANALYSIS