BMP2/SMAD pathway activation in JAK2/p53-mutant megakaryocyte/erythroid progenitors promotes leukemic transformation

BCR-ABL–negative myeloproliferative neoplasms (MPNs) are characterized by somatic mutations (eg, JAK2, CALR, and MPL mutations) that activate the JAK-STAT pathway.^1-4 In a substantial proportion of cases, MPNs undergo leukemic transformation (LT), which carries a dismal prognosis and often does not respond to classical antileukemic therapies.^5,6 The most frequent genomic alterations to occur at the time of LT involve TP53 (mutated in approximately one-third of patients at time of LT^7,8), in marked contrast to the low frequency of TP53 alterations in the chronic phase of MPN (2%-5%).^8,9 In addition, alterations in chromosome 17p, which contains the TP53 locus, predict a high likelihood of LT and are commonly observed in LT.^10,11 These data thus indicate that TP53 alterations are a key regulator of the process of LT.

We have previously shown that the variant allele fractions of TP53 mutations are significantly greater in post-MPN acute myeloid leukemia (AML) than in MPN. In the majority of cases, the TP53 mutated clone had a variant allele fraction >50% at the time of LT, consistent with selection for biallelic TP53 inactivation. By contrast, previous studies have shown that low abundance TP53 mutations do not necessarily influence disease progression in MPNs and that TP53 haploinsufficiency may be insufficient for blast phase transformation.^12,13

The vast majority of patients with myeloid malignancy carry TP53 missense mutations, most of which are located in the DNA-binding domain.^8,14,15 However, the mechanisms by which TP53 mutational status, as well as allelic configuration, drives disease progression remains largely unknown. We have previously shown that in a Trp53 knockout background, overexpressed JAK2^V617F promotes LT, consistent with observations in humans.⁷ Here we seek to investigate the impact of Trp53 mutation status as well as allelic state on the process of LT and to identify the mechanisms of LT mediated by alteration of Trp53.

Jak2^V617F knock-in mice¹⁶ were crossed to Rosa-CreERT2 and Trp53^R172H knock-in¹⁷ and Trp53 knock-out¹⁸ mice. Cre expression was induced by oral gavage of tamoxifen one time at a 4-mg dose. Deletion/recombination was confirmed 2 weeks after tamoxifen oral gavage. Murine transplant procedures are described in the supplemental Methods (available on the Blood Web site). Mouse husbandry, care, and all animal interventions were in strict compliance with the Institutional Animal Care and Use Committee guidelines.

Antibody staining for surface marker was performed as previously described.¹⁹ The antibodies used are listed in the supplemental Methods.

Cytokine-supplemented methylcellulose medium (STEMCELL Technologies, M3434) was used for colony-forming assays as previously described.²⁰ 

For RNA-sequencing (RNA-seq), RNA was isolated by TRIzol extraction from sorted-cell populations as indicated. RNA-seq libraries were generated by 3′ sequencing and SMART-Seq2 amplification and sequenced on an Illumina HiSeq 4000.

Lin^–c-Kit⁺ bone marrow (BM) cells from mice were sorted for viability (4′,6-diamidino-2-phenylindole negative). Single-cell RNA-seq (scRNA-seq) libraries were prepared following 10X Genomics protocols (Chromium Single Cell 3′ Reagent Kit version 3). The final libraries were assessed by using an Agilent Technology 2100 Bioanalyzer and were sequenced on an Illumina NovaSeq 6000 system (S2 flow cell, paired-end).

Western blot was performed as previously described.²⁰ The antibodies used are listed in the supplemental Methods.

To investigate the genetic interactions between Trp53 deficiency and recurrent Trp53 mutations and Jak2^V617F signaling in leukemogenesis, we generated an allelic series of genetically engineered mouse models (supplemental Figure 1A). To investigate the cell autonomous interaction between Trp53 loss or mutant Trp53 and Jak2^V617F, we transplanted BM cells from wild type (WT), Jak2^V617F/+ (J^VF), Jak2^V617F/+Trp53^+/− (J^VFP^+/−), Jak2^V617F/+Trp53^−/− (J^VFP^−/−), Jak2^V617F/+Trp53^R172H/+ (J^VFP^R172H/+), and Jak2^V617F/+Trp53^R172H/– (J^VFP^R172H/–) mice into recipient mice. Recipient mice transplanted with J^VF, J^VFP^+/−, and J^VFP^R172H/+ cells developed an MPN-like phenotype (supplemental Figure 1B) with a median survival of 178 days, 191 days, and 162 days, respectively (Figure 1A). By contrast, the recipients of J^VFP^−/− and J^VFP^R172H/– initially developed an MPN phenotype followed by transformation to an acute leukemia phenotype with a median survival of 135 days and 130 days. Survival was significantly shorter in J^VFP^−/− vs J^VFP^+/− mice, and in J^VFP^R172H/– vs J^VFP^R172H/+ mice, and both J^VFP^−/− and J^VFP^R172H/– mice had significantly shorted survival than J^VF mice (with no significant difference between J^VFP^−/− and J^VFP^R172H/–mice). Peripheral blood (PB) counts revealed a significant increase in the white blood cell count and spleen weight of J^VFP^−/− and J^VFP^R172H/– mice compared with J^VF mice, as well as a significant decrease in platelet count and hemoglobin (HGB) in J^VFP^R172H/– vs J^VF mice (Figure 1B). Histopathologic analysis of leukemic mice revealed sheets of immature blast-like cells in the PB, BM, spleen, and liver, which were not observed in J^VF, J^VFP^+/−, and J^VFP^R172H/+ mice (Figure 1C). Finally, to confirm that the process of transplantation of hematopoietic stem/progenitor cells (HSPCs) from primary mice did not contribute to the development of the leukemic phenotype observed in J^VFP^−/− and J^VFP^R172H/– mice, tamoxifen-induced recombination was conducted in primary mice. J^VFP^−/− (n = 3) and J^VFP^R172H/–(n = 2) mice developed an MPN phenotype followed by transformation to an acute leukemia, consistent with the earlier observations (supplemental Figure 1C-F). Thus, LT occurred only in the background of Trp53 biallelic inactivation.

Immunophenotypic analysis revealed CD45^–c-Kit⁺ blast cells, negative for mature myeloid and lymphoid markers (Figure 1D; supplemental Figure 1G), similar to the immunophenotype of human PEL.²¹ Furthermore, these blast cells were CD71⁺Ter119^–, and by immunohistochemical stains, blasts were CD45^–c-Kit⁺Gata1⁺CD71⁺Ter119^–CD41^– blasts (Figure 1E; supplemental Figure 1H), which is consistent with the Bethesda criteria for erythroleukemia in mice.²² 

Functionally, BM cells from J^VFP^−/− and J^VFP^R172H/– exhibited increased serial replating capacity, both at the time of LT (Figure 1F) and at the MPN stage (supplemental Figure 1I). The PEL was transplantable to secondary and tertiary recipients, and all recipients rapidly developed lethal PEL (Figure 1G).

The observed latency between the development of an MPN phenotype and a leukemic phenotype suggested the possibility that the process of LT is associated with the acquisition of additional somatic alterations. Using sparse whole-genome sequencing, we identified that the PEL stage (Figure 1H-I), but not its MPN precursor (supplemental Figure 1J), exhibited a complex karyotype (defined as ≥3 karyotypic alterations), with recurrence of acquired events. Mutational analysis identified acquisition of somatic mutations in J^VFP^−/− and J^VFP^R172H/– PEL mice without evidence for a clear, additional driver of LT (supplemental Table 1). Although there was no consistent pattern of mutation acquisition observed, the most frequent alterations observed were in Ptpn11 (largely subclonal). This finding is consistent with prior observations in humans showing that subclonal RAS pathway mutations can occur in TP53 mutant AML.²³ Thus, Jak2^V617F and Trp53 biallelic inactivation results in a highly penetrant form of leukemia that immunophenotypically resembles human PEL and is associated with the development of genomic instability during the process of LT.

To identify the leukemia-initiating population, we first examined the proportion of HSPCs pre-LT and post-LT by using flow cytometry. At the MPN stage, the megakaryocyte-erythroid progenitors (MEPs) (Lin^–c-Kit⁺Sca-1^–CD34^–FcgRII/III^–) and LSKs (Lin^–c-Kit⁺Sca-1⁺) were significantly expanded in the BM of both J^VFP^−/− and J^VFP^R172H/– mice, compared with WT, J^VF, J^VFP^+/−, and J^VFP^R172H/+ mice (Figure 2A; supplemental Figure 2A; supplemental Figure 2C). In the spleen, the MEPs, but not the LSKs, were significantly expanded in mice with Trp53 mutations compared with J^VF. No significant change in the size of the granulocyte-macrophage progenitor (GMP) compartment was observed in J^VFP^−/− and J^VFP^R172H/– mice (supplemental Figure 2B-C). Significant expansion of the MEPs, but not GMPs, was noted at the time of LT in the BM (Figure 2B). Expansion of the MEPs, GMPs, and LSKs was observed at the time of LT in the spleen; however, the proportional expansion of the MEPs was the greatest among these HSPCs. Notably, in contrast to MEPs from J^VF mice, MEPs from leukemic mice did not express CD45 (supplemental Figure 2D).

To further assess the biologic differences of J^VFP^−/− and J^VFP^R172H/– HSPCs relative to J^VF HSPCs at the MPN stage, competitive transplantation of BM cells from J^VFP^−/− or J^VFP^R172H/– and J^VF into recipients in a 1:1 ratio was performed (supplemental Figure 2E). PB chimerism revealed that the proportion of J^VFP^−/− or J^VFP^R172H/– cells was significantly greater than J^VF cells (supplemental Figure 2F). Analysis of BM LSK, MEP, and GMP chimerism at 13 weeks post-tamoxifen revealed that the majority of cells in each HSPC compartment were either J^VFP^−/− or J^VFP^R172H/– in origin. Importantly, J^VFP^−/− comprised a significantly greater proportion of MEPs than GMPs (91% vs 82%) or LSKs (91% vs 62%) (Figure 2C), which was also observed in J^VFP^R172H/–. These data indicate that J^VFP^−/− or J^VFP^R172H/– significantly enhances the self-renewal and competitive advantage of HSPCs over J^VF HSPCs, most significantly in the MEP compartment.

Given the particular competitive advantages in the MEPs resulting from biallelic Trp53 inactivation, we hypothesized that the leukemia-initiating population resided in this compartment. To test this hypothesis, we transplanted CD45^– MEPs from J^VFP^−/− or J^VFP^R172H/– PEL mice into recipients. These cells were able to efficiently induce a PEL phenotype (supplemental Figure 2G). CD45⁺ MEPs (early-stage untransformed MEPs relative to CD45^– MEPs) were also able to induce the same PEL phenotype but had a longer latency before onset of PEL (supplemental Figure 2H), suggesting that progression through MEP maturation may be required for leukemic progression. Furthermore, MEPs identified by a distinct set of markers (Lin^–Sca-1^–c-kit⁺CD150⁺CD41^–CD71^– cells, referred to as CD150⁺ MEPs²⁴) also caused PEL with the same immunophenotype as primary recipients. We sought to determine if other HSPC compartments from J^VFP^−/− and J^VFP^R172H/– mice contained leukemia-initiating populations at the time of the MPN phenotype. Recipient mice transplanted with MEPs, but not GMPs, developed a PEL phenotype, without a preceding MPN phenotype. Mice transplanted with LSKs also developed a PEL phenotype but with a preceding MPN phenotype, and they also had significantly longer median survival than mice transplanted with MEPs (Figure 2D), suggesting that transition through the MEPs was required for disease initiation.

As shown in Figure 1I, progression of MPN to PEL is characterized by the acquisition of copy number alterations (CNAs), presumably through a process of genomic instability and selection. We would therefore predict that these events might be enriched in the HSPCs containing the leukemia-initiating population and its progeny. Notably, we identified that karyotypic/CNAs were observed in the MEPs and not identified in the GMPs from leukemic mice (Figure 2E-F), although Jak2 and Trp53 mutations existed in both populations (supplemental Figure 2I). Collectively, these observations indicate that the biologic and genomic events that lead to leukemic transformation are selected for within the MEPs.

To investigate transcriptional networks in the leukemia-initiating context, we performed scRNA-seq of BM Lin^–c-Kit⁺ HSPCs from WT, J^VF, pre-PEL J^VFP^−/−, and J^VFP^−/− PEL mice (supplemental Figure 3A). Clustering analysis of 29 029 single-cell profiles resulted in a detailed map, including 17 transcriptionally homogeneous clusters based on their expression patterns (Figure 3A; supplemental Figure 3B). We further confirmed and defined these HSPC subpopulations and leukemic populations according to canonical surface markers, including CD45 (Ptprc), CD34 (Cd34), CD16 (Fcgr3), CD32 (Fcgr2b), CD71 (Tfrc), and CD41 (Itga2b) (Figure 3B; supplemental Figure 3C). A total of 15 distinct HSPC subpopulations and a leukemic population (consisting of 2 subpopulations, leukemia-1 and leukemia-2) were identified (Figure 3C). Compared with J^VF mice, the megakaryocyte-erythroid progenitor (MkE-P) and erythroid progenitor (EryP-1 and EryP-2) populations were expanded in pre-PEL J^VFP^−/− mice (supplemental Figure 3D). To confirm this finding, we used an alternative fluorescence-activated cell sorting (FACS) gating strategy to isolate bipotent MEPs (Lin^–c-Kit⁺Sca-1^–CD34^–FcgRII/III^–CD71^–CD41^–CD150⁺CD105^–, referred to as CD150⁺CD105^– MEPs) and erythroid-committed progenitors (Lin^–c-Kit⁺Sca-1^–CD34^–FcgRII/III^–CD71⁺CD41^–).²⁵ We observed that CD150⁺CD105^– MEPs and erythroid-committed progenitors were significantly expanded in J^VFP^−/− and J^VFP^R172H/– mice at the MPN stage compared with WT and J^VF mice (supplemental Figure 3E). Correlation analysis in the average gene expression profile (Figure 3D; supplemental Figure 3F) showed that the leukemia population is most similar to erythroid progenitors and erythroblasts. In addition, analysis of genes upregulated in leukemia-1 and leukemia-2 cells exhibited a significant enrichment in gene sets found in erythroid progenitors (Figure 3E). These data suggest that the leukemia cells evolved from the erythroid lineage.

Given the results indicating that CNAs were found largely in MEPs derived from mice with PEL and the fact that such events were generally large in size, we reasoned that copy-number inference from scRNA-seq data would provide further insights into the specific subpopulations within the MEP compartment associated with LT. Indeed, inference of genome copy number revealed the presence of the previously described CNAs (Figure 2E). Strikingly, these CNAs occurred exclusively in leukemia cells (Figure 3F; supplemental Figure 3G) and not in other populations. Collectively, these data indicate that the leukemic population is a distinct population within the FACS-defined MEP compartment and appears to be most closely transcriptionally related to the normal erythroid progenitor and erythroblast populations. These findings suggest that Jak2 mutation and Trp53 deficiency result in transcriptional reprogramming and acquired copy number events that occur in restricted erythroid-committed progenitors.

As described earlier, we observed that both transplanted LSKs and MEPs from J^VFP^−/− and J^VFP^R172H/– MPNs give rise to PEL. This raises the question as to whether the leukemia-initiating population resides solely in the MEPs or also in the LSKs. To address this question, we performed RNA-seq on sorted MEPs and LSKs at both the MPN stage and the PEL stage. Three distinct clusters were segregated by principal component analysis (PCA) of these samples: MEPs from PEL, MEPs from non-PEL, and LSKs (Figure 4A). Importantly, in the LSKs of J^VFP^−/− and J^VFP^R172H/–, no significant differentially expressed genes associated with erythroid differentiation were upregulated (Figure 4B), suggesting that the expansion of the MEPs (Figure 2A) is not due to lineage bias in the LSKs. Furthermore, we did not identify significant differences, relative to WT mice, in the transcriptional signature of the LSK populations in J^VF, J^VFP^+/−, J^VFP^R172H/+, J^VFP^−/−, and J^VFP^R172H/– mice (supplemental Figure 4A). By contrast, evaluation of the MEP populations by gene set enrichment analysis showed an enrichment in stem cell–like signatures in J^VFP^−/− and J^VFP^R172H/– MEPs at the MPN stage (Figure 4C). To biologically assess for the presence of stem cell–like features in the MEPs of MPN-stage J^VFP^−/− and J^VFP^R172H/– mice, we sorted BM MEPs from J^VFP^−/−, J^VFP^R172H/–, and J^VF and assessed for self-renewal capacity by serial replating in methylcellulose medium. Importantly, J^VFP^−/− and J^VFP^R172H/– exhibited a significantly increased replating capacity relative to J^VF (Figure 4D), thus providing biological validation of the gene set enrichment analysis findings.

To understand the transcriptional programs that distinguish J^VFP^−/− and J^VFP^R172H/– erythroleukemic blasts from non-PEL MEPs, we performed RNA-seq and differentially expressed gene analysis (supplemental Figure 4B). Because J^VFP^−/− and J^VFP^R172H/– result in an identical phenotype and similar transcriptional programs (supplemental Figure 4C), and to explore the common mechanism by which Trp53^−/− and Trp53^R172H/– induce Jak2^V617F/+ MEP transformation, we surveyed pathways associated with the 435 genes upregulated in both J^VFP^−/− and J^VFP^R172H/– PEL MEPs. We identified enrichment of the transforming growth factor β/bone morphogenetic protein (BMP) pathway, which was uniquely consistent across 3 data sets among upregulated pathways identified (Figure 4E-G). This pathway is linked to stem cell self-renewal and proliferation.^26,27 Thus, both J^VFP^−/− and J^VFP^R172H/– PEL converge on upregulation of the BMP pathway.

We identified that Bmp2, Bmp4, Smad1, Smad9, and Id1 (a transcriptional target of the BMP/SMAD pathway) were upregulated in J^VFP^−/− and J^VFP^R172H/– PEL MEPs relative to WT and J^VF (supplemental Figure 5A); this was confirmed via quantitative reverse transcription polymerase chain reaction, western blot, and/or immunohistochemistry (Figure 5A-B; supplemental Figure 5B-C). Furthermore, phosphorylated-SMAD1/5/9 levels (the main SMAD complex through which BMP2 and BMP4 signal) were significantly higher in PEL MEPs compared with J^VF (Figure 5C; supplemental Figure 5D) but not in GMPs from PEL (supplemental Figure 5E- F), thus confirming the specific activation of BMP/SMAD pathway in PEL MEPs.

Previous data have shown that BMP2 expression is not enriched in the erythroid lineage in either the human or murine context (https://servers.binf.ku.dk/bloodspot/). This led us to hypothesize that the observed aberrant expression of BMP2 we observed in PEL MEPs contributed to the process of leukemogenesis. To investigate this possibility, we isolated erythroleukemic blasts (Lin^–CD45^–c-Kit⁺ cells) from PEL J^VFP^−/− and J^VFP^R172H/– mice and transduced them with viral vectors expressing a short hairpin RNA (shRNA)-targeting Bmp2 or a control, both with green fluorescent protein (GFP). Bmp2 knockdown significantly decreased colony formation capacity of erythroleukemic blasts (Figure 5D); upon replating, Bmp2 knockdown resulted in reduced colony formation at each replating relative to control and almost complete loss of colony formation by the fourth replating for MPN J^VFP^−/− and J^VFP^R172H/– MEPs (Figure 5E). Overexpression of Bmp2 in c-Kit⁺ WT BM cells resulted in significantly increased colony formation and increased replating capacity vs controls (Figure 5F). By contrast, Bmp4 knockdown did not decrease colony formation capacity of erythroleukemic blasts (supplemental Figure 5G-H).

We next investigated the role of Bmp2 in erythroleukemic blast maintenance in vivo. We transduced erythroleukemic blasts with control shRNA or Bmp2-shRNA and injected 5000 GFP⁺ transduced erythroleukemic blasts with CD45.1⁺ WT support BM cells into recipient mice. Recipient mice injected with Bmp2-shRNA erythroleukemic blasts showed evidence of restoration of normal hematopoiesis, as evidenced by significant reductions in leukocytosis, significant increases in HGB, and an increase in platelet count (Figure 5G; supplemental Figure 5I). Furthermore, mice that received Bmp2-shRNA erythroleukemic blasts exhibited a significant reduction in the representation of GFP⁺ transduced erythroleukemic blasts in the PB (Figure 5H; supplemental Figure 5J) and a concomitant increase in the proportion of CD45.1 WT cells in the PB (supplemental Figure 5K), indicating that Bmp2 silencing abrogated the fitness advantage of erythroleukemic blasts. Finally, compared with mice injected with erythroleukemic blasts with control shRNA, mice injected with Bmp2-shRNA erythroleukemic blasts had significantly longer survival (Figure 5I; supplemental Figure 5L).

To determine if the upregulation of the BMP protein 2 (BMP2)/SMAD pathway also occurs in human post-MPN AML, we examined the transcriptional profiles of human post-MPN AML patient samples (either M6/7 or non-M6/7 AML) and identified significant enrichment of BMP2/SMAD pathway expression in post-MPN AML cases relative to cases of polycythemia vera (Figure 5J; supplemental Figure 5M-N). Thus, consistent with our observations in our murine model, the BMP2/SMAD pathway is upregulated in post-MPN AML.

Prior observations in patients with polycythemia vera have shown that JAK2^V617F increases replication stress, causing stalling of the replication fork that can be converted to double-strand breaks (DSBs) and impaired intra-S checkpoint in erythroblasts.²⁸ Moreover, inactivation of p53 results in more DSBs and exacerbates genomic instability in JAK2^V617F mutant erythroblasts. Based on these collective observations, we hypothesized that PEL cells have increased activation and dependency on DNA damage response (DDR) pathways for survival. Western blot analysis showed that J^VFP^−/− PEL c-Kit⁺ spleen cells (compared with WT, J^VF, and J^VFP^+/−) had increased expression of phosphorylated RB and CHK1 (associated with DSBs) and phosphorylated RPA32, RAD51, and phosphorylated CDC2 (Y15) (associated with homologous recombination repair) (Figure 6A). Thus, increased activation of DDR was evident in, and largely restricted to, J^VFP^−/− PEL.

Poly(ADP-ribose) polymerase (PARP) inhibition can result in single-strand breaks that can progress to DSBs if not repaired.^29,30 Inhibition of WEE1 can abrogate G2/M arrest, compromise DNA repair, and allow early mitotic entry of cells harboring unresolved DNA damage, which can result in mitotic catastrophe.^31-33 Prior observations have shown that combined inhibition of WEE1 and PARP results in a synergistic increase in DNA damage and reduction in cellular proliferation in different tumor contexts.^34-37 Given the prior observations regarding the impact of JAK2^V617F on DNA damage fidelity, the engagement of DDR pathways (as discussed earlier), and the high proliferative rate of erythroleukemic blasts, we investigated the possibility that targeting both PARP and WEE1 could lead to biologically intolerable DNA damage and increased apoptosis in J^VFP^−/− PEL. We used the WEE1 inhibitor adavosertib and the PARP inhibitor olaparib to treat J^VFP^−/− leukemic mice. Mice were randomized to treatment with vehicle, olaparib (100 mg/kg daily), adavosertib (60 mg/kg daily), or combination (olaparib 100 mg/kg + adavosertib 60 mg/kg daily) for 2 weeks (pretreatment blood counts at the time of randomization of mice are listed in supplemental Table 2). Compared with vehicle and single agents, combination therapy significantly reduced white blood cell count, leukemic burden in PB and BM, and spleen size (Figure 6B-D). Importantly, the overall survival of mice treated with combination therapy was significantly prolonged compared with controls (Figure 6E). WT mice treated with the combination exhibited a modest decrease in HGB compared with vehicle or single-agent olaparib, but no other hematologic toxicities were observed (supplemental Figure 6A-F).

After 3 days and 6 days of treatment, we evaluated pharmacodynamic markers of WEE1 and PARP inhibitor activity (inhibition of phosphorylation of CDC2, and PAR, respectively) by western blot, which showed inhibition of these pathways in vivo (Figure 6F). Combination therapy significantly increased apoptosis as indicated by cleaved caspase-3 and Annexin V (Figure 6G; supplemental Figure 6G ), as well as DNA damage, indicated by phosphorylated RPA32 and phosphorylated γ-H2AX, compared with vehicle and single-agent treatment. The elevated levels of histone H3 (S10) phosphorylation indicated that compared with controls, the combination resulted in a greater proportion of cells in early mitotic entry. These data indicate that combined PARP inhibitor and WEE1 inhibitor treatment promotes DNA damage and induction of apoptosis in J^VFP^−/− leukemic cells, thus leading to leukemia-specific combination therapeutic efficacy.

Acute leukemias characterized by TP53 mutations with complex karyotypes, including post-MPN AMLs, are associated with lack of therapeutic response to conventional antileukemic therapies and an overall poor prognosis.^38,39TP53 alterations reportedly occur in up to 92% of cases of PEL,^40,41 and co-occurring TP53 mutations and alterations in the 17p locus are frequently observed.^11,40 Collectively, these observations suggest that TP53 alterations, in distinct genomic contexts such as co-occurring JAK2 mutations, may result in a convergent propensity toward erythroleukemia characterized by complex cytogenetic alterations. Here, consistent with these clinical observations, we have developed a genomically and biologically accurate model of erythroleukemic transformation of MPNs characterized by CNAs and extensive DNA damage.

Data in solid tumor models have suggested that certain missense mutant TP53 alleles have gain-of-function (GOF) properties that produce phenotypes distinct from a TP53 null state.^42,43 In terms of leukemia, Loizou et al⁴⁴ identified a GOF Trp53^R172H in murine AML. By contrast, in the context of Jak2^V617F/+, we identified no differences in phenotype, karyotype, origin of leukemia, or survival between Trp53^R172H/– and Trp53 null PEL. Furthermore, the transcriptional networks of J^VFP^−/− and J^VFP^R172H/– PEL share many common molecular signatures. These data provide clear genetic evidence that LT from MPN to acute erythroleukemia requires homozygous inactivation of p53, and that GOF of TP53 mutations are not responsible for LT in the context of activated JAK2. It provides further evidence for context-specific effects of TP53 mutations/allelic status, which will need to be investigated in other tumor types using similar approaches.

Our data suggest a distinct lineage bias induced by Trp53 mutations, including in concert with JAK2 activation in HSPCs evolving to acute leukemia. Lu et al⁴⁵ found that TP53 plays a pivotal role in erythroid progenitor fate specification. Furthermore, in conditions of stress erythropoiesis, the absence of p53 increases the proliferation of erythroid progenitors and immature precursors.⁴⁶ FACS-defined MEPs represent a heterogeneous population, and transformation in our models occurs within this compartment but cannot be further resolved by flow cytometric analysis alone.⁴⁷ Our scRNA-seq data suggest that p53 inactivation expands bipotent MkE-P, promotes erythrocyte-committed expansion, and facilitates the transformation of erythroid progenitors in the context of Jak2 constitutive activation. These data suggest a key regulatory role for p53 in erythroid lineage maturation in the context of certain physiologic stresses, providing a key barrier to transformation that is overcome by concurrent JAK2 activation plus biallelic Trp53 loss.

We identified that the BMP/SMAD pathway has a key role in the pathogenesis of p53 mutant LT. BMPs have been implicated in the self-renewal and maintenance of HSCs, as well as regulating the expansion of progenitor cells.⁴⁸ In adult erythropoiesis, exposure to exogenous BMP2 increases immature erythroid colonies.⁴⁹ Moreover, upregulation of the BMP2/4 pathway also plays a role in enhancing self-renewal of megakaryocytic progenitors in pediatric acute megakaryoblastic leukemia with inv(16)(p13.3q24.3).⁵⁰ Interestingly, the BMP2/SMAD pathway appears to play a role in DDR.⁵¹ Whether the BMP2/SMAD pathway becomes upregulated in response to the increased DNA damage incurred during leukemic progression remains to be investigated.

In summary, we have developed murine models of Jak2-Trp53–altered erythroleukemia characterized by recurrent CNAs and DNA damage that faithfully recapitulate the key features of human erythroleukemia. The process of LT is characterized by upregulation of the BMP/SMAD pathway, which enhances self-renewal and HSPC expansion. Importantly, we identify that the resulting increase in DNA damage can be exploited by exposure to DDR pathway inhibitors, suggesting that combined WEE1/PARP inhibition as a therapeutic approach to treat this high risk, refractory leukemia subtype is urgently warranted in the preclinical to clinical context.

This study was supported by a National Institutes of Health, National Cancer Institute (NCI), Cancer Center Support Grant/Core Grant to Memorial Sloan Kettering Cancer Center (P30 CA008748), NCI grants (1K08CA188529-01, R.K.R.; R35 CA197594-01A1, R.L.L.; and P01 CA108671 11, R.L.L. and R.K.R.), the Kattamuri S. Sarma Leukemia Research Fund, and a grant from Cycle for Survival (R.K.R). B.L. is supported by National Natural Science Foundation of China (NSFC) (82070134) and Tianjin Natural Science Funds (19JCQNJC09400), and NSFC 81530008 (Z.X.). H.W. and K.H. are supported by a Tri-Institutional Stem Cell grant (2019-035). T.B. is supported by the William C. and Joyce C. O’Neil Charitable Trust, Memorial Sloan-Kettering Single Cell Sequencing Initiative. C.G.M. and I.I. are supported by the American Lebanese Syrian Associated Charities of St. Jude Children’s Research Hospital. C.G.M. is supported by NCI R35 Outstanding Investigator Award CA197695, the Henry Schueler 41&9 Foundation, and a Leukemia and Lymphoma Society Translational Research Program. I.I. is supported by the St. Jude Children’s Research Hospital Hematological Malignancies Program Garwood Fellowship. S.W.L. is supported by NCI 1 P50 CA254838-01.

Contribution: R.K.R., R.L.L., B.L., W.A., and H.W. conceived the project; B.L., W.A., H.W., T.B., S.M., A.K., W.X., and Y.L. performed and analyzed experiments; R.P.K. and H.W. performed bioinformatics analyses; I.I. and C.G.M. provided patient RNA-seq data; B.L., W.A., H.W., and R.K.R. wrote the original draft; S.F.C., Z.X., A.D., K.H., S.W.L., R.L.L., and R.K.R. reviewed and edited the original draft; and R.L.L. and R.K.R. supervised the project. All authors reviewed and approved the manuscript.

Conflict-of-interest disclosure: R.K.R. has received consulting fees from Constellation, Incyte, Celgene/BMS, Novartis, Promedior, CTI, Jazz Pharmaceuticals, Blueprint, Stemline, Galecto, PharmaEssentia, AbbVie, Sierra Oncology, and Disc Medicines; and research funding from Incyte, Constellation, and Stemline. R.L.L. is on the supervisory board of QIAGEN; is a scientific advisor to Loxo (until February 2019), Imago, C4 Therapeutics, and IsoPlexis; receives research support from and consulted for Celgene and Roche; has consulted for Lilly, Jubilant, Janssen, Astellas, MorphoSys, and Novartis; and has received honoraria from Roche, Lilly, and Amgen for invited lectures and from Celgene and Gilead for grant reviews. C.G.M. has received research funding from AbbVie and Pfizer, speaking fees from Amgen, and advisory board fees from Illumina; and holds stock in Amgen. I.I. received honoraria from Amgen and Mission Bio. K.H. is a founder of Dania Therapeutics; a consultant for Inthera Bioscience AG; and a scientific advisor for MetaboMed Inc. and Hannibal Health Innovation. S.W.L. is a founder and member of the scientific advisory board of Blueprint Medicines, Mirimus, ORIC Pharmaceuticals, and Faeth Therapeutics; and is on the scientific advisory board of Constellation Pharmaceuticals and PMV Pharmaceuticals. The remaining authors declare no competing financial interests.

The current affiliation for K.H. is Institute of Cancer Research, London, United Kingdom.

Correspondence: Raajit K. Rampal, Leukemia Service, Department of Medicine, 530 East 74th st, New York, NY 10021; e-mail: rampalr@mskcc.org.