PRMT5 Inhibition Modulates E2F1 Methylation and Gene-Regulatory Networks Leading to Therapeutic Efficacy in JAK2^V617F-Mutant MPN

Protein arginine methylation is a key post-translational protein modification that is catalyzed by the family of protein arginine methyl transferases (PRMT; ref. 1). These enzymes transfer methyl groups, provided by the cofactor S-adenosyl-methionine (SAM), to arginine residues on histones and on non-histone proteins (2). PRMT activity can result either in asymmetric dimethyl arginine (ADMA) or symmetric dimethyl arginine (SDMA). PRMTs are thus classified as either class I (ADMA) or class II (SDMA) methyltransferases (2–4). By methylating arginine residues on histones, PRMTs act as epigenetic writers whose interplay with epigenetic readers (5) and erasers leads to modification of the histone code that contributes to the regulation of transcription, DNA replication, and DNA repair. Arginine methylation of non-histone proteins has also been implicated in cell signaling, ribosome biogenesis, RNA transport, and splicing (6, 7).

PRMT5 is the primary type II PRMT responsible for symmetric dimethylation of arginine residues and has a known role in normal development and an emerging role in different malignant contexts. Prmt5 knockout mice are not viable (8) and conditional Prmt5 knockout mice show alterations in hematopoiesis (9), suggesting that PRMT5 activity has a key role in development and hematopoiesis. Dysregulation of PRMT5 occurs in a variety of malignancies and is associated with increased cancer cell proliferation and dismal prognosis (7, 10–12). The relationship between methylthioadenosine phosphorylase (MTAP) allelic status/function and PRMT5 activity has suggested PRMT5 as a potential therapeutic target in different malignancies (13, 14), and recent data linking PRMT5 function to the spliceosome (15, 16) has further highlighted the potential of PRMT5 inhibition as a cancer therapeutic strategy. As such, therapeutic agents targeting PRMT5 have entered early clinical trials (NCT02783300, NCT03854227, NCT03573310, NCT03886831, and NCT04089449).

Myeloproliferative neoplasms (MPN) are clonal myeloid malignancies initiated by somatic mutations in hematopoietic stem cells (HSC; ref. 17). The most common BCR–ABL-negative MPNs are polycythemia vera, essential thrombocythemia, and myelofibrosis. These diseases are characterized by overproduction of terminally differentiated blood cells along with an increased bleeding and thrombosis risk, and increased risk of transformation to bone marrow (BM) failure and/or acute leukemia. Genetic profiling has shown that the most common mutations in MPN result in dysregulated JAK–STAT signaling, most commonly due to JAK2^V617F mutation, across all three diseases (18–21). On the basis of these discoveries, the JAK1/2 inhibitor ruxolitinib was developed for MPN therapy and approved for the treatment of myelofibrosis in 2011 and polycythemia vera in 2014, and the selective JAK2 inhibitor fedratinib has recently been approved for treatment of myelofibrosis. Although ruxolitinib therapy results in a substantial reduction in constitutive symptoms and spleen size (22–25), JAK inhibitor therapy with ruxolitinib or other agents in clinical development does not reduce mutant allele burden or provide disease modification. This underscores the need to explore different therapeutic targets in MPN either as monotherapy or in combination with JAK inhibitors.

In addition to mutations in the JAK–STAT pathway, genomic studies have identified recurrent disease alleles in epigenetic modifiers in myeloid malignancies, including MPN (26–28). These data underscore the relevance of epigenetic alterations to MPN pathogenesis and suggest a potential role for epigenetic therapies in MPN. Previous studies suggested a functional link between constitutive JAK2 activity, PRMT5 phosphorylation, and MPN pathogenesis (29); however, the therapeutic potential of PRMT5 inhibition in MPN has not been studied. We investigated the role of PRMT5 in MPN pathogenesis through the use of C220, a potent and selective small-molecule PRMT5 inhibitor, and aimed to elucidate key PRMT5 targets contributing to the maintenance of MPN cells.

To assess the requirement for PRMT5 in cells harboring a JAK2^V617F mutation, we performed in vitro proliferation assays with different concentrations of C220—a potent and highly selective small-molecule PRMT5 inhibitor (Supplementary Fig. S1A–S1C). C220 treatment caused a dose-dependent inhibition of proliferation in Ba/F3-EpoR cells expressing JAK2^V617F, with a 2.5-fold lower IC₅₀ compared with cells expressing wild-type JAK (JAK2^WT) or empty vector (Fig. 1A). C220 dose-dependently reduced the proliferation of the JAK2^V617F-mutant cell lines SET2, UKE1, and HEL in vitro (Fig. 1B). We previously developed models of JAK inhibitor “persistent” MPN cells, which survive in the setting of prolonged exposure to ruxolitinib and other JAK inhibitors (30, 31). To address whether ruxolitinib-persistent cells (Supplementary Fig. S2A) could be targeted by PRMT5 inhibition, we performed proliferation assays in ruxolitinib-persistent SET2 cells compared with naïve SET2 cells and found similar sensitivity to C220 (Fig. 1C). C220 exposure for 6 days did not alter phosphorylation of JAK2, STAT5, AKT, or MAPK in SET2 cells at doses that reduced SDMA, indicative of target inhibition (Fig. 1D). These data suggest a distinct mechanism by which PRMT5 inhibition affects the survival of MPN cells. Consistent with these data, combined PRMT5 inhibition with JAK1/2 inhibition showed additive, but not synergistic, effects as quantified by a ZIP synergy score of 5.3 for the JAK2^V617F Ba/F3 and 3.7 for the SET2 cells (Fig. 1E and F; Supplementary Fig. S2B and S2C). No additive effects were observed with the combination in JAK2^WT Ba/F3 cells (Supplementary Fig. S2D and S2E).

To further address the relevance of PRMT5 function to MPN pathogenesis, we assessed PRMT5 expression in human peripheral blood granulocytes using previously published RNA expression profiling data (32). Ninety-three patients with MPN displayed a median PRMT5 expression that was significantly higher compared with 11 age-matched normal subjects (P = 0.027; Fig. 1G). PRMT5 expression was significantly higher in 28 patients with polycythemia vera compared with healthy controls (P = 0.023) and showed a trend toward higher expression in 18 patients with myelofibrosis and 47 patients with essential thrombocythemia (P = 0.597, P = 0.302, respectively, Fig. 1H).

The presence of a JAK2^V617F mutation was associated with significantly higher PRMT5 expression in MPN (P = 0.027) and specifically in JAK2^V617F-mutated polycythemia vera (P = 0.0064) patient samples, and there was a trend toward higher PRMT5 expression in patients with JAK2^V617F-mutated myelofibrosis compared to patients with myelofibrosis with other mutations (Supplementary Fig. S3A and S3B).

Patients harboring homozygous JAK2^V617F mutations (defined by Taqman value of >50) showed significantly higher PRMT5 expression compared with normal controls (MPN: P < 0.001; polycythemia vera: P < 0.0001; myelofibrosis: P = 0.048), and compared with patients with heterozygous JAK2^V617F [(MPN: P = 0.0015; polycythemia vera: P = 0.0046)/CALR (MPN: P < 0.0001)/MPL (MPN: P = 0.037)] mutations or who were triple negative with respect to JAK2, MPL, or CALR mutations (MPN: P = 0.011; Supplementary Fig. S3C and S3D).

Furthermore, we observed a significant positive correlation between PRMT5 expression in granulocytes or SDMA serum levels and JAK2^V617F expression measured by Taqman (P < 0.0001 and P = 0.012, respectively; Supplementary Fig. S3E and S3F).

SDMA serum levels were significantly increased in 9 patients with JAK2^V617F polycythemia vera and 9 patients with JAK2^V617F primary myelofibrosis as compared with 10 healthy controls (P = 0.031 and P = 0.028, respectively; Fig. 1I). In addition, human CD34⁺ cells from patients with polycythemia vera/primary myelofibrosis demonstrated sensitivity to C220 in proliferation assays ex vivo (Fig. 1J) at concentrations of C220 that inhibited SDMA (Supplementary Fig. S3G and S3H). These data further support PRMT5 as a therapeutic target in MPN. WT CD34⁺ cells also showed similar sensitivity in these assays, such that the impact of PRMT5 inhibition was assessed in MPN cells and in normal hematopoietic cells in subsequent in vivo studies.

We next assessed the efficacy of C220 in the JAK2^V617F-driven murine MPN model of polycythemia vera in vivo. BM of Jak2^V617F CD45.2 Mx1-Cre conditional knock-in mice (33) was transplanted in a 50:50 ratio with CD45.1 Jak2^WT BM into lethally irradiated recipients. C220 administered at a dose of 12.5 mg/kg orally once daily 5 days on, 2 days off for 4 weeks demonstrated potent target inhibition (Fig. 2A), as evidenced by 63% and 59% reduction of SDMA protein levels in the BM and spleen, respectively, compared with vehicle-treated control mice (Fig. 2B). Monotherapy with C220 was well tolerated and did not reduce overall mouse body weights or spleen and liver weights as compared with sex- and age-matched WT transplant recipient controls (Supplementary Fig. S4A–S4C). C220 did not attenuate JAK2–STAT5 signaling in vivo (Fig. 2C). Polycythemia vera mice treated with C220 displayed significantly lower hematocrits (mean 63% vs. 76 %, P < 0.05), white blood cell counts (WBC; mean 13.5 G/L vs. 20.5 G/L, P < 0.01), reticulocytes (mean 276 G/L vs. 1612 G/L, P < 0.001), and splenomegaly (mean weight 95 mg vs. 254 mg, P < 0.001) compared with mice treated with vehicle (Fig. 2D–H). C220 therapy reduced the proportion of CD105⁺CD150⁻ progenitors (P < 0.0001), committed erythroid progenitors (Pre-CFU-E; P = 0.0004), bipotential megakaryocyte–erythroid progenitors (Pre-Meg-E; P = 0.0157), as well as megakaryocyte–erythroid progenitors (MEP; P = 0.0139), Ter119^medCD71^hi proerythroblasts (P < 0.0001), Ter119^hiCD71^hi basophilic erythroblasts (P < 0.0001), and CD11b⁺Gr1⁻ cells (P = 0.0023) relative to vehicle. Although treatment with C220 did not reduce mutant allele chimerism in the peripheral blood (Supplementary Fig. S5A), C220 significantly reduced JAK2^V617F+(CD45.2⁺)/CD45.1 chimerism in Lin⁻Sca1⁻cKit⁺ stem/progenitors (P = 0.0214, P = 0.0003) and MEP (P = 0.0148, P < 0.0001), in BM and spleen, respectively, as well as Pre-CFU-E (P = 0.0132), Pre-Meg-E (P = 0.0477), and pre–granulocyte macrophage progenitors (Pre-GM, P = 0.0421) in the BM and orthochromatophilic erythroblasts in the spleen (P = 0.0238; Supplementary Fig. S5B–S5G). In addition, there was a trend to lower chimerism in CD105⁺CD150⁻ progenitors, proerythroblasts, and chromatophilic erythroblasts as well as in CD11b⁺Gr1⁺ populations (Supplementary Fig. S5H–S5N).

C220 therapy reduced cellularity and normalized the proportion of myeloid lineage cells in the bone marrow of C220-treated mice compared with vehicle-treated mice (Fig. 2I). The splenic architecture was also restored in the setting of C220 therapy, with decreased extramedullary hematopoiesis consistent with reduced myeloproliferation (Fig. 2J). C220 significantly reduced serum cytokines in the JAK2^V617F model, consistent with inhibition of the systemic inflammation characteristic of MPN (34).

We next assessed the efficacy of combined C220 and ruxolitinib in vivo. Combination therapy was well tolerated and did not reduce mouse body weights (Supplementary Fig. S6A). Importantly, combination C220–ruxolitinib therapy showed further reduction in hematopoietic parameters in the JAK2^V617F model, including hematocrit (mean = 37.7% vs. 50.2% or 78.9%, P = 0.0089 and P = 0.0009) and WBC (mean = 2.2 G/L vs. 4.7 G/L or 7.6 G/L, P = 0.0015 and P < 0.0001; Fig. 3A and B) compared with single-agent C220 and ruxolitinib. Mutant allele chimerism in the peripheral blood was not affected by combination therapy, but we observed normalization of reticulocyte numbers, hemoglobin levels, and circulating CD11b⁺Gr1⁻ cells with combination therapy (P < 0.0001 each vs. vehicle, Fig. 3C; Supplementary Fig. S6B–S6D). Furthermore, dual targeting of JAK1/2 and PRMT5 was superior to monotherapy with either agent with respect to reducing splenomegaly (mean weight = 39 mg vs. 76 mg or 134 mg; P = 0.0004 and P < 0.0001), BM hypercelluarity (mean femur counts = 4.3 vs. 20.0 vs. 27.2 × 10⁶ total lysed cells, both P < 0.0001) and hepatomegaly (mean = 1,076 mg vs. 1,369 mg or 1,372 mg, P = 0.0006 and P = 0.0031; Fig. 3D–F). Spleen weights of mice receiving combination therapy were lower than WT transplant recipient controls (39 mg vs. 77 mg, P < 0.0001). Liver weights in combination-treated mice were not different compared with WT transplant recipient controls (1,076 mg vs. 1,128 mg, P = n.s.; Supplementary Fig. S4B and S4C).

Ruxolitinib and C220 monotherapy did not reduce the aberrant expansion of megakaryocytes seen in this model, whereas combination treatment significantly reduced megakaryocytes compared with vehicle treatment [counted by ×200 high-power field (HPF), P < 0.001; Fig. 3G and H]. Combination therapy was able to abrogate the aberrant myeloid infiltrate in spleens of JAK2^V617F mice and restore splenic architecture, which was only partially achieved with C220 or ruxolitinib monotherapy (Fig. 3I).

Flow cytometric analysis showed that combination therapy significantly reduced erythroid progenitors such as Ter119^medCD71^hi proerythroblasts [BM/spleen: 0.07%/0.12% vs. 0.20%/0.68% (P = 0.172/P = 0.0003) vs. 1.74%/2.24% (P = 0.0002/P < 0.0001) vs. 2.00%/5.77% (P < 0.0001/<0.0001)] and Ter119^hiCD71^hi basophilic erythroblasts [BM/spleen: 0.3%/0.2% vs. 3.5%/9.2% (P = 0.0258/P = 0.0388) vs. 30.7%/49.9% (P < 0.0001/P < 0.0001) vs. 33.7%/20.9% (P < 0.0001/P < 0.0001)] compared with C220, ruxolitinib, or vehicle treatment, both in the BM (Fig. 3J) and spleen (Supplementary Fig. S6E–S6G). Furthermore, combination treatment was superior to single-agent C220 or ruxolitinib treatment in reducing BM erythroid and megakaryocytic progenitors including CD105⁺CD150⁻ progenitors (0.1% vs. 9.0% vs. 43.8% of FcGR-negative cells, P = 0.0146 and P < 0.0001), Pre-MegE (2.4% vs. 9.7% vs. 7.0% of FcGR-negative cells, P < 0.0001 and P = 0.0011), and Pre-CFU-E (0.7% vs. 3.1% vs. 5.9% of FcGR-negative cells, P = 0.0095 and P < 0.0001; Fig. 3K). MEPs were significantly reduced by dual PRMT5/JAK targeting compared with either monotherapy or vehicle in the BM (6.0% of BM Lin⁻Sca1⁻cKit⁺ vs. 21.1% vs. 31.8% vs. 30.0%, P = 0.0002, P < 0.0001, and P < 0.0001; Fig. 3L; Supplementary Fig. S6H) and spleen (37.4% of spleen Lin⁻Sca1⁻cKit⁺ vs. 54.4% vs. 89.4%, vs. 91.6%, P = 0.11, P < 0.0001, and P < 0.0001; Supplementary Fig. S6I). Combination treatment, but neither single-agent treatment, also significantly decreased the portion of Lin⁻Sca1⁺cKit⁺ CD48⁻CD150⁻ multipotent myeloid progenitors in the BM (P < 0.0001) and Lin⁻Sca1⁻cKit⁺ myeloid progenitors in the spleen (P = 0.0014; Supplementary Fig. S6J and S6K). A significant decrease of CD41⁺ HSCs in the BM and the spleen was achieved with combination treatment (P = 0.0002, P < 0.0001; Supplementary Fig. S6L and S6M) but not with single-agent treatment. C220 monotherapy and combination treatment both significantly reduced serum cytokines [specifically eotaxin (P < 0.0001, P = 0.0031), macrophage colony-stimulating factor (M-CSF; P = 0.0023, P = 0.0031), IL12 (p40) (P = 0.0003, P = 0.0004), LIX (P < 0.0001, P < 0.0001), RANTES (P = 0.0003, P < 0.0001), IP10 (P = 0.0049, P = 0.0008), IL1β (P = 0.0002, P = 0.0075)], respectively, compared with vehicle control (Fig. 3M; Supplementary Fig. S6N), consistent with inhibition of the systemic inflammation characteristic of MPN. Of note, levels of eotaxin, IL12 (p40), LIX, and M-CSF were not reduced by ruxolitinib treatment alone but were attenuated by both C220 monotherapy and combination treatment (Fig. 3M; Supplementary Fig. S6N). In addition, PRMT5 inhibitor monotherapy also significantly lowered KC (P = 0.0068), MCP1 (P = 0.0397), MIP1a (P = 0.0074), and MIP2 (P = 0.0312) serum levels; levels of these cytokines were not altered by ruxolitinib monotherapy nor by combination treatment (Supplementary Fig. S6N). Taken together, this suggests that distinct inflammatory cytokines are regulated downstream of JAK2 and PRMT5 in MPN cells.

We also investigated the efficacy of combination therapy in an in vivo SET2 xenograft model. Adequate suppression of PRMT5 was achieved by C220 and combination treatment as demonstrated by potent reduction in SDMA levels with in vivo therapy (Fig. 3N); SDMA levels remained unaffected by monotherapy ruxolitinib treatment (Supplementary Fig. S6O). Consistent with the data in our murine model, we observed significant efficacy with C220 monotherapy (71% tumor reduction) and combination therapy (73% tumor reduction) with 14 days of treatment, which was greater than that observed with JAK inhibition (59%; P = 0.0006, P = 0.0039, P = 0.0040 compared with vehicle-treated mice, respectively; P = 0.0545 and P = 0.0464 compared with ruxolitinib-treated mice, respectively; Fig. 3O).

We next investigated the efficacy of PRMT5 inhibition and of combined PRMT5/JAK inhibition in a MPL^W515L BM transplant (BMT) model of myelofibrosis. C220 therapy at 15 mg/kg oral, once daily (5 days on/2 days off) was sufficient to potently suppress SDMA levels in vivo (100% suppression compared with vehicle, P < 0.005; Supplementary Fig. S7A and S7B). C220 monotherapy significantly reduced splenomegaly (mean: 151 mg vs. 423 mg, P = 0.0042), hepatomegaly (mean: 1,156 mg vs. 1,663 mg, P = 0.0025), WBCs (mean: 12.8 G/L vs. 170.6 G/L, P = 0.0225), and platelet counts (mean: 233 G/L vs. 1,868 G/L, P = 0.049; Supplementary Fig. S7C–S7F) compared with vehicle-treated mice.

C220-treated mice showed a significant reduction of peripheral blood mutant allele burden (27.4% vs. 83.6% of GFP⁺ cells, P = 0.0003) and significantly lower neutrophils (1.4% vs. 49%, P = 0.0028) and monocytes (7.7% vs. 23.5%, P = 0.0374; Supplementary Fig. S7G–S7I) compared with vehicle-treated mice. In the BM, there was a trend toward fewer Lin⁻ cells and MEPs (7.6% vs. 12.3%) in the PRMT5 inhibitor–treated mice (Supplementary Fig. S7J and S7K).

We next investigated the efficacy of combined PRMT5/JAK inhibition. Ruxolitinib in combination with the higher dose of C220 (15 mg/kg, once daily 5 days on/two days off), which showed efficacy in this more aggressive model, resulted in significant pancytopenia. Therefore, combination studies were done with ruxolitinib at 60 mg/kg orally twice daily and C220 at 10 mg/kg orally once daily (5 days on/2 days off). Combined PRMT5/JAK inhibition induced more significant suppression of splenomegaly (mean weight = 137 mg vs. 459 mg, P = 0.006), hepatomegaly (mean weight = 1,114 mg vs. 1,651 mg, P = 0.019), leukocytosis (mean WBC = 6.1 G/L vs. 149.3 G/L, P = 0.044), and hematocrits (mean = 31% vs. 51%, P = 0.0296) and resulted in a trend toward lower platelet counts (mean platelet count = 804 G/L vs. 3,225 G/L, P = 0.1502) compared with vehicle-treated mice (Fig. 4A–D; Supplementary Fig. S8A and S8B). Combination treatment reduced reticulocytes significantly more compared with ruxolitinib treatment (P = 0.023; Supplementary Fig. S8C).

In terms of toxicity, spleen weights of mice receiving combination therapy were still larger compared with MIGR1 (MSCV-IRES-GFP empty vector) transplant recipient controls (137 mg vs. 69 mg, P = 0.0079). Liver weights in combination-treated mice were comparable to MIGR1 transplant recipient controls (1,114 mg vs. 1,148 mg, P = n.s.; Supplementary Fig. S8D and S8E).

C220 monotherapy and combination treatment, but not single-agent ruxolitinib treatment, significantly reduced the proportion of CD11b⁺Gr1⁻ cells in peripheral blood (P = 0.0101 and P = 0.0181) and BM (P = 0.0219 and P = 0.0240; Supplementary Fig. S8F). Histopathologic analysis showed that the myeloid infiltrates in the liver seen in vehicle-treated mice were decreased by single-agent PRMT5 or JAK inhibition and more markedly reduced with combination therapy (Fig. 4E). C220 monotherapy, and to a greater extent combination therapy, significantly decreased splenic extramedullary hematopoiesis and partially restored splenic architecture (Fig. 4E). BM hypercellularity was significantly decreased by C220- and combination therapy–treated compared with ruxolitinib-treated mice (P = 0.006 and P = 0.012; Fig. 4E; Supplementary Fig. S8G). The marked increase in megakaryocytes seen in this model was significantly reversed by PRMT5 inhibition, JAK inhibition, or combination therapy (P = 0.0004, P = 0.003, P = 0.004, respectively; Fig. 4F). Reticulin fibrosis in BM and spleen was increased in vehicle-treated mice but significantly reduced in all three treatment cohorts (Fig. 4G–I). C220 monotherapy and combination therapy significantly lowered levels of circulating IL1β (P = 0.0312 and P = 0.0257), IP10 (P = 0.0143 and P = 0.0004), MIP1b (P = 0.0482 and P = 0.0466), RANTES (P = 0.0102 and P = 0.0139), and TNFα (P = 0.0216 and P = 0.0024), which were not attenuated by ruxolitinib monotherapy (Fig. 4J; Supplementary Fig. S8H). Combination-treated mice displayed reduced circulating levels of LIF (P = 0.0323) and IL13 (P = 0.0324), both of which were also reduced by ruxolitinib (P = 0.0474 and P = 0.0497, respectively). Of note, IL10 levels were only significantly reduced by dual PRMT5/JAK inhibition (P = 0.0257), but not by treatment with either agent alone.

To further understand the mechanism by which C220 leads to therapeutic efficacy in MPN, we performed RNA sequencing (RNA-seq) in sorted Jak2^V617F-mutated MEPs (Supplementary Fig. S9A) from the JAK2^V617F model and from SET2 xenografts after vehicle or C220 in vivo therapy. In both models, principal component analysis (PCA) of RNA-seq data separated C220-treated from control groups (Supplementary Fig. S9B and S9C). The transcriptome of C220 and vehicle-treated SET2 xenografts or JAK2^V617F-mutated MEPs revealed a total of 4,219 and 2,862 genes which were differentially expressed, respectively, of which 2,223 and 874 were significantly downregulated (Fig. 5A and B; Supplementary Fig. S9D). Gene set enrichment analysis (GSEA) revealed a significant enrichment for p53 pathway activation signatures and reduced expression of gene expression signatures associated with E2F targets (including DNA repair and G₂–M checkpoint) and IFNα and IFNγ response (Fig. 5C–F) with PRMT5 inhibition.

GSEA of Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) data revealed a reduced accessibility of gene sets associated with E2F targets, DNA damage repair, and IFNα and IFNγ response with C220 (Supplementary Fig. S9E). Genes with significant reduced accessibility and gene expression during C220 therapy included genes involved in immune system processes (including the genes encoding CCL3 and CCL4), metabolism, and DNA binding (Supplementary Fig. S9F and S9G). Motif enrichment on peaks with lower accessibility in ATAC-seq and concurrently reduced gene expression revealed an enrichment of MYC and NF-E2 targets (Supplementary Fig. S9H).

We next integrated genes that were significantly downregulated by PRMT5 inhibition with genes that were upregulated in gene expression data derived from patients with CD34⁺ JAK2^V617F-mutated polycythemia vera, essential thrombocythemia, or primary myelofibrosis (35, 36). There was an overlap of 452, 387, and 452 genes for polycythemia vera, essential thrombocythemia, and primary myelofibrosis, respectively, which were upregulated in primary MPN cells and significantly attenuated with PRMT5 inhibition in SET2 xenografts (Fig. 5G–I). GSEA of overlapping genes showed a highly significant enrichment for IFN response in all three MPN subtypes and for E2F targets in polycythemia vera and essential thrombocythemia. In concordance with these results, the ranked difference of normalized enrichment scores between SET2 xenografts and CD34⁺ Jak2^V617F-mutated polycythemia vera, essential thrombocythemia, or primary myelofibrosis indicated IFN response and inflammatory response as well as MTORC1 signaling as pathways that were upregulated in these MPNs and downregulated by C220 (Supplementary Fig. S9I–S9K).

E2F targets were one of the two major target gene sets downregulated upon treatment with C220; these included E2F targets involved in a variety of processes including DNA damage repair (e.g., RAD54L), checkpoint control (e.g., BRCA1), and cell-cycle entry (e.g., CDK1/2; ref. 37). We focused on elucidating (i) whether PRMT5 was needed for the survival of SET2 cells and (ii) the mechanism by which PRMT5 inhibition affects E2F function and downstream target gene expression. We performed a CRISPR knockout competition assay in SET2 cells stably transduced with Cas9, and first showed reduced fitness of PRMT5 knockout cells validating our pharmacologic data with C220 (Fig. 6A; Supplementary Fig. S10A). Likewise, knockout of E2F1, E2F2, and E2F3 decreased proliferation in SET2 cells (Fig. 6A; Supplementary Fig. S10A–S10D; Supplementary Tables S1 and S2).

We performed immunoprecipitation studies with E2F1 followed by Western blot analysis of PRMT5 in SET2 and UKE1 cells and observed that PRMT5 coimmunoprecipitated with E2F1 (Fig. 6B; Supplementary Table S3). We also immunoprecipitated PRMT5 and were able to detect E2F1 and E2F2, but E2F3 was detected to a much lesser extent (Supplementary Fig. S10E). To test the hypothesis that PRMT5 affects E2F targets by modulating E2F1 and E2F2 methylation status, SET2 cells were treated with different C220 concentrations, followed by immunoprecipitation for E2F1 or E2F2 and Western blot analysis for SDMA in both the total lysate and the coimmunoprecipitates (Fig. 6C; Supplementary Fig. S10F). C220 exposure resulted in attenuated symmetric dimethylation of E2F1, but not of E2F2 (Fig. 6C; Supplementary Fig. S10F). This suggests that there may be increased interaction between demethylated E2F1 and known cofactors, including the retinoblastoma (RB) tumor suppressor protein. C220 exposure did not affect total RB levels, but increased association between E2F1 and RB (Fig. 6D). These data suggest that PRMT5 inhibition may attenuate E2F target gene activation through differential activation of transcriptional activators and repressors based on E2F methylation status (Supplementary Fig. S10G).

E2F gene-regulatory networks play a major role in regulating checkpoints (e.g., BUB1B, BUB3) and cell-cycle entry (e.g., CDK1,2 CCNA2, CCNE2). Modulation of E2F gene expression by PRMT5 inhibition caused a G₂–M cell-cycle arrest in SET2 (44.8%, 35.5%, 30.0%, and 24.7% of cells in G₂–M phase in 111 nmol/L, 12.4 nmol/L, 4.1 nmol/L. and DMSO-treated SET2 cells, respectively, P < 0.0001 for each vs. DMSO control; Fig. 6E) and UKE1 cells (16.4%, 6.7%, 3.3%, and 1.0% of cells in G₂–M phase in 111 nmol/L, 12.4 nmol/L, 4.1 nmol/L, and DMSO-treated UKE1 cells, respectively, P < 0.0001 for each vs. DMSO control; Supplementary Fig. S10H). We observed increased apoptosis as demonstrated by upregulation of TP53 (100 nmol/L vs. DMSO, P = 0.0013; Fig. 6F and G; Supplementary Table S4) and a significant increase in activated caspase-3 expression (111 nmol/L vs. DMSO control: P = 0.0006 in SET2, P < 0.0001 in UKE1 cells; Fig. 6H; Supplementary Fig. S10I). Moreover, a subset of E2F target genes have known roles in DNA damage repair (e.g., RAD54L, MLH1, BRCA1; Fig. 5B). C220-treated SET2 cells showed a significant increase in radiation-induced double-strand breaks as assessed by γH2AX Western blot analysis (Fig. 6I), γH2AX immunofluorescence staining (100 nmol/L vs. DMSO control: γH2AX foci/cell P < 0.0001; Fig. 6J and K), and a neutral comet assay (100 nmol/L vs. DMSO control: arbitrary units tail moment P = 0.019; Fig. 6L). In addition, repair of single-strand breaks was impaired in the setting of PRMT5 inhibition, with increased PARP expression/cleavage and increased activated caspase-3 expression in C220-treated cells compared with vehicle (Fig. 6I and H). These findings prompted us to explore combination therapy with C220 plus inhibitors targeting DNA damage repair pathways, including ATM (AZD0156) and PARP (olaparib). Indeed, combined PRMT5 and ATM inhibition, as well as combined PRMT5 and PARP inhibition, showed additive effects in SET2 cells (ZIP synergy scores of 2.8 and 2.8, respectively; Supplementary Fig. S10J and S10K) consistent with increased efficacy of the combination in vitro.

The discovery of somatic JAK2^V617F mutations resulting in constitutively active JAK2–STAT5 signaling underscored activated JAK–STAT signaling as a disease-underlying mechanism in polycythemia vera, essential thrombocythemia, and primary myelofibrosis. These findings led to the clinical development of JAK inhibitors and subsequent approval of ruxolitinib for polycythemia vera and myelofibrosis and fedratinib for myelofibrosis. Despite these important genetic and therapeutic insights, current JAK1/2 inhibitors can achieve significant clinical benefits without altering the natural history of the disease, suggesting a need for new therapeutic approaches that show efficacy alone and in combination with JAK inhibition in MPN.

Dysregulated PRMT5 expression occurs in a variety of malignancies. Moreover, previous work has suggested that PRMT5 may be a specific therapeutic target in MTAP-deleted tumors and in malignancies with spliceosome component mutations. As such, PRMT5 inhibition is currently being tested as a novel promising therapy (12) in a variety of solid cancers, lymphoma, and myelodysplastic syndrome in clinical trials (JNJ-64619178, PF-06939999, GSK3326595, PRT543, PRT811; www.clinicaltrials.gov). The role of PRMT5 in the pathogenesis and therapy of MPN has been unclear. Liu and colleagues showed that JAK2^V617F phosphorylates PRMT5, impairing its methylation activity, suggesting that certain aspects of PRMT5 function might be attenuated in MPN cells (29). However, it is not clear whether PRMT5 phosphorylation completely attenuates PRMT5 catalytic activity, or whether PRMT5 phosphorylation is altered in the setting of preclinically/clinically achievable JAK inhibitor doses. We found that PRMT5 is significantly overexpressed in JAK2^V617F-mutated polycythemia vera, and that the transcriptional effects of PRMT5 inhibition in MPN cells had significant overlap with the set of genes that have altered expression in MPN cells. These data suggest that there is an aberrant gene expression signature mediated by PRMT5 in MPN cells, and we demonstrate a potential therapeutic role for PRMT5 inhibition in MPN. PRMT5 inhibition, alone and in combination with ruxolitinib, showed significant efficacy in JAK2^V617F and MPL^W515L MPN in vivo models, with reduced myeloproliferation, pathologic responses, and attenuated inflammatory cytokine secretion. Despite the significant, potent reduction in many hallmark disease features of MPN, the smaller spleens seen with combination PRMT5/JAK inhibitor therapy (Supplementary Fig. S4B) suggest that the impact of dual PRMT5/JAK inhibitor therapy on normal hematopoiesis will need to be investigated in clinical studies.

Chronic inflammation is a hallmark of MPN (38). The increase of several inflammatory serum cytokines has a crucial role in MPN disease maintenance and progression. Therefore, targeting inflammation at early disease stages is a therapeutic pillar to prevent myelofibrosis. PRMT5 inhibition potently reduced the serum levels of multiple cytokines, including TNFα—a cytokine that has been shown to be essential for clonal expansion in a JAK2^V617F mouse model (39).

Levels of circulating IL8, IL2R, IL12, and IL15 have been shown to be independently prognostic in patients with primary myelofibrosis and associated with impaired outcome (40). IL8 level in particular is an independent prognostic factor that correlates inversely with leukemia-free survival. In the JAK2^V617F model, therapy with C220 or C220–ruxolitinib combination therapy significantly lowered cytokine levels of the murine IL8 analogues LIX and MIP1a. Moreover, the combination of C220 and ruxolitinib was superior to either agent as monotherapy, with respect to reducing elevated circulating levels of LIX. This was also evident in the MPL^W515L model in which we observed a trend of lower levels of the IL8 analogues LIX and MIP1a (Supplementary Fig. S8H). The epigenetic mechanisms that lead to aberrant cytokine secretion, including IL8 analogues in preclinical models and IL8 in human MPN cells, require further investigation; our data suggest a key role for PRMT5 in this aberrant inflammatory circuit.

Most importantly, although C220 and ruxolitinib therapy induced coordinate changes in gene expression, such as those involved in the IFN response (Supplementary Fig. S11A and S11B; Supplementary Table S5), C220 exposure downregulated more than 1,808 genes that are not downregulated by ruxolitinib (Supplementary Fig. S11A, S11C, and S11D; Supplementary Table S6). These data suggest that PRMT5 inhibition induces therapeutic efficacy through an orthogonal mechanism not shared with JAK inhibition. We observed increased expression of E2F target genes and of IFN pathway members in MPN cells, which was potently suppressed with PRMT5 inhibitor therapy in vivo. In addition, integrated RNA-seq and ATAC-seq analyses uncovered that C220 potently reduced expression and accessibility of targets of the transcription factor NFE2, which is overexpressed in the majority of patients with MPN and causes an MPN phenotype with a high rate of leukemic transformation when overexpressed in mice (41). The role of IFN signaling and other inflammatory mediators in the BM microenvironment and molecular mechanisms are postulated to drive progression of BM fibrosis in myelofibrosis, although the exact pathophysiology is not well understood. Our data suggest that modulation of IFN signaling by PRMT5 inhibition may have therapeutic importance in MPN cells and in a broader suite of malignant contexts.

Kim and colleagues suggest a link between RB/E2F and TPO signaling via JAK2 in hematopoietic stem cell homeostasis; however, the role of E2F and E2F target gene expression in MPN pathogenesis has not been studied (42). Our data illuminate a potential biological and therapeutic role for this pathway in MPN. Arginine methylation of E2F1 has been discovered in solid cancer cell lines, but never previously studied in hematologic malignancies (43). We provide evidence that expression of PRMT5, E2F1, E2F2, and E2F3 is needed for proliferation and survival of JAK2-mutant MPN cells. Consistent with the potent suppression of E2F target genes in vivo, we found that E2F1 and E2F2 physically interact with PRMT5 in JAK2-mutant MPN cells. PRMT5 inhibition reduced dimethylated E2F1 leading to a higher affinity between E2F1 and the tumor suppressor RB protein and resultant reduced expression of E2F1 downstream targets. Among those targets were proteins involved in DNA damage repair, and we observed increased DNA damage in C220-treated MPN cells and cooperativity between PRMT5 inhibition and either ATM or PARP inhibition in MPN cells. Previous work by Mullally and colleagues (44) and by Skorski and colleagues (45) have implicated aberrant DNA damage responses in MPN pathogenesis, and our data suggest PRMT5 as a therapeutically tractable mediator of these responses with pathogenetic relevance. Further mechanistic studies of the interaction between PRMT5 and E2F1 in the context of MPN will be of importance to delineate the link between aberrant JAK–STAT signaling, MPN pathogenesis, and PRMT5/E2F1–mediated gene regulation.

Taken together, our data provide evidence that PRMT5 is a therapeutic target in MPN, alone and in combination with JAK inhibition. These data demonstrate novel links between PRMT5, E2F1 gene-regulatory function, and the survival of MPN cells and provide a strong mechanism-based rationale for therapeutic studies of PRMT5 inhibitors in MPN. On the basis of these studies, PRT543, a novel and selective PRMT5 inhibitor, is currently being evaluated in a phase I clinical trial, including in MPNs (NCT03886831).

C220 is a derivative of a published PRMT5 inhibitor (46) and was synthesized by Wuxi Apptec. For in vitro assays, it was stored as a 10 mmol/L solution in DMSO. For in vivo assays, C220 was dissolved in a 0.5% methylcellulose, 0.1% Tween-20 solution. C220 was administered as oral gavage once daily for 5 days a week, for a total of 4 to 6 weeks. AZD0156 and olaparib were purchased from MedChemExpress and stored as 10 mmol/L DMSO solution.

Ba/F3 cells stably expressing EPOR were transduced with viral supernatants containing MSCV-JAK2^V617F-IRES-GFP, MSCV-JAK2^WT-IRES-GFP, MSCV-IRES-GFP empty vector (EV), flow sorted for GFP, and maintained in RPMI 1640/10% FCS supplemented with 10 ng/mL IL3 in the EPOR/JAK2^WT and EPOR/EV cells. HEL and SET2 cells were grown in RPMI 1640 medium containing 10% or 20% FCS, respectively. UKE1 cells were cultured in RPMI 1640 supplemented with 10% horse serum (StemCell Technologies) and 50 μmol/L hydrocortisone (Sigma-Aldrich). SET2 cells transduced with Cas9 were selected for in the presence of blasticidin at a concentration of 10 μg/mL. Ruxolitinib-persistent SET2 cells were generated by incubating SET2 cells at increasing doses of ruxolitinib over 10 weeks. SET2 cells were authenticated by and obtained from DSMZ <6 months before use. All cell lines underwent Mycoplasma testing (Lonza, MycoAlert Mycoplasma Detection Kit, #LT07–318) before use.

For proliferation assays, 10,000 cells/200 μL medium were plated in triplicate and supplemented with increasing doses of inhibitor using 9-point, 3-fold dilutions with a top concentration of 1,000 nmol/L for C220. After 3 and 6 days, cells were split 1:4 and supplemented with fresh medium and inhibitor. Proliferation was assessed at 3, 6, and 10 days using the CellTiter-Glo luminescent cell viability assay (Promega) and normalized proliferation in media with an equivalent volume of DMSO. IC₅₀ was determined using Graph Pad Prism 8.0.

Mononuclear cells were isolated using Ficoll from peripheral blood of deidentified patients with untreated JAK2^V617F-mutant polycythemia vera and primary myelofibrosis. CD34⁺ cells were enriched for using AUTOMACS with the CD34 MicroBead Kit human (Miltenyi Biotec, catalog no. 130-046-702). CD34⁺ cells were plated at a density of 2,500 cells in cytokine-supplemented methylcellulose medium (H4435; StemCell Technologies) with increasing concentrations of C220 or DMSO control. Colonies were scored at day 10. All experiments were performed in triplicate using 5 different patient samples (3 polycythemia vera, 2 primary myelofibrosis).

SET2 cell–derived mouse xenograft studies were conducted in 8-week-old female SCID Beige mice at Charles River Discovery Sciences. Charles River Discovery Services specifically complies with the recommendations of the Guide for Care and Use of Laboratory Animals with respect to restraint, husbandry, surgical procedures, feed and fluid regulation, and veterinary care. The animal care and use program at Charles River Discovery Services is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC), which assures compliance with accepted standards for the care and use of laboratory animals.

SET2 cells used for implantation were harvested during log-phase growth and resuspended in PBS containing 50% Matrigel (BD Biosciences). Each mouse was injected subcutaneously in the right flank with 1 × 10⁷ tumor cells (0.1 mL cell suspension) while under isoflurane anesthesia to reduce ulcerations. Tumor growth was monitored as the average size approached the target range of 100–150 mm³. Twenty-one days later, designated as day 1 of the study, mice were sorted according to tumor size and treated with C220 at 15 mg/kg orally once daily and ruxolitinib at 60 mg/kg orally twice daily as single agent or in combination for 2 weeks. Tumors were measured in two dimensions using calipers and volume was calculated.

The median tumor volume for the number of animals [MTV (n)] on day 14 was determined for each group. Percent tumor growth inhibition (%TGI) was defined as the difference between the MTV of the designated control group (group 1) and the MTV of the drug-treated group, expressed as a percentage of the MTV of the control group: % × 100.

BM of 4-month-old male JAK2^V617F (33) CD45.2 primary mice was mixed in a 50:50 ratio with BM of male CD45.1 C57BL/6 mice (obtained from Charles River) and transplanted into 6-to-8-week-old lethally irradiated female recipients (obtained from JAX). After engraftment, mice were randomized by chimerism and blood counts into different treatment arms: vehicle, C220 (12.5 mg/kg, once daily, 5/2 days), ruxolitinib (60 mg/kg twice daily), and combination of C220 (12.5 mg/kg, once daily, 5 days on/2 days off) and ruxolitinib (60 mg/kg twice daily). Treatment was administered by oral gavage for 4 to 6 weeks.

In the conditional MPL^W515L model of myelofibrosis, male BALB/c mice received one dose of 5-fluorouracil (0.150 mg/g mouse, intraperitoneal injection). Five days later, BM was isolated, transduced with retroviral supernatant containing MSCV-hMPL^W515L-IRES-GFP or MIGR1 (MSCV-IRES-GFP-empty vector), and injected into lethally irradiated female 6-to-8-week-old BALB/c recipients (47). After engraftment, MSCV-hMPL^W515L-IRES-GFP mice were randomized according to GFP and blood counts into the different treatment arms: vehicle, C220 (10 mg/kg, once daily, 5 days on/2 days off when used in combination trial; 15 mg/kg, once daily, 5 days on/2 days off when used vs. vehicle only), ruxolitinib (60 mg/kg twice daily, 7 days), and combination of C220 (10 mg/kg, once daily, 5 days on/2 days off) plus ruxolitinib (60 mg/kg twice daily, 7 days).

The ruxolitinib dose of 60 mg/kg twice daily instead of 90 mg/kg twice daily in the MPL^W515L model was chosen to allow for assessment of activity of the ruxolitinib–C220 combination treatment with tolerable BM toxicity with the combination. Mouse husbandry, care, and all animal interventions were in strict compliance with the Institutional Animal Care and Use Committee guidelines.

For details on Methods, see Supplementary Data.

F. Pastore reports other from Prelude Therapeutics and grants from Deutsche Forschungsgemeinschaft during the conduct of the study. N. Bhagwat reports personal fees from Prelude Therapeutics Inc. (employment and ownership interest) outside the submitted work. W. Xiao reports grants from Stemline therapeutics outside the submitted work. J.L.V Maag currently works at Inzen Therapeutics. A. Grego reports personal fees from Prelude Therapeutics Inc. (employment and ownership interest) outside the submitted work. J. Mehta reports personal fees from Prelude Therapeutics Inc. (employment) outside the submitted work. M. Wang reports personal fees from Prelude Therapeutics Inc. outside the submitted work. B.H. Durham reports grants from NCI (salary) and American Society of Hematology (salary) during the conduct of the study, and Erdheim-Chester Disease Global Alliance Foundation (salary and lab supplies) outside the submitted work. R.K. Rampal reports personal fees from Incyte, Celgene, CTI, Jazz, Blueprint, Pharmaessentia, AbbVie, and Galecto outside the submitted work; and grants and personal fees from Constellation and Stemline. P. Scherle reports personal fees from Prelude Therapeutics (employment and ownership interest) outside the submitted work. K. Vaddi reports personal fees from Prelude Therapeutics outside the submitted work. R.L. Levine reports grants from Prelude during the conduct of the study; and personal fees from Qiagen outside the submitted work. R.L. Levine is on the supervisory board of Qiagen and is a scientific advisor to Loxo (until February 2019), Imago, Ajax, Zentalis, Auron, Mana, C4 Therapeutics, and Isoplexis, which each include an equity interest. R.L. Levine also receives research support from Celgene, Roche, and Prelude Therapeutics and is a consultant for Celgene, Roche, Prelude Therapeutics, Lilly, Incyte, Novartis, Astellas, Morphosys, and Janssen. No potential conflicts of interest were disclosed by the other authors.

F. Pastore: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. N. Bhagwat: Conceptualization, investigation, writing-review. A. Pastore: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology. A. Radzisheuskaya: Formal analysis, investigation, methodology. A. Karzai: Methodology. A. Krishnan: Methodology. B. Li: Methodology. R.L. Bowman: Methodology. W. Xiao: Methodology. A.D. Viny: Methodology. A. Zouak: Methodology. Y.C. Park: Methodology. K.B. Cordner: Methodology. S. Braunstein: Methodology. J.L. Maag: Data curation. A. Grego: Methodology.J. Mehta: Methodology. M. Wang: Methodology. H. Lin: Methodology. B.H. Durham: Methodology. R.P. Koche: Data curation, investigation. R.K. Rampal: methodology. K. Helin: Conceptualization, supervision. P. Scherle: Conceptualization, investigation. K. Vaddi: Conceptualization, investigation. R.L. Levine: Conceptualization, resources, supervision, funding acquisition, investigation, writing-original draft, project administration, writing-review and editing.

R.L. Levine has received research support from Prelude Therapeutics. W. Xiao has received research support from StemLine Therapeutics. R.K. Rampal has received research funding from Incyte, StemLine, and Constellation. F. Pastore has received research support from Prelude Therapeutics and the German research foundation PA 2541/1-1. We acknowledge the support of Memorial Sloan Kettering Cancer Center Support Grant NIH P30 CA008748. This work was supported by National Cancer Institute R35 CA197594-01A1 (R.L. Levine) and National Cancer Institute P01 CA108671 11 (R.L. Levine). R.L. Bowman is supported by the Sohn Foundation Fellowship of the Damon Runyon Cancer Research Foundation (DRG 22-17) and by National Cancer Institute K99 CA248460.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.