Activation of JAK/STAT Signaling in Megakaryocytes Sustains Myeloproliferation In Vivo

The MPNs are clonal hematopoietic stem cell (HSC) disorders characterized by the expansion of mature myeloid elements. The most common genetic alteration is JAK2^V617F, which is found in > 95% of all patients with polycythemia vera, and in 50%–60% of essential thrombocytopenia/primary myelofibrosis cases (1–4). Expression of this mutation in cell lines causes transformation to cytokine-independent growth and constitutive activation of downstream STAT signaling (2, 3), and expression of JAK2^V617F in the hematopoietic compartment is sufficient to cause MPN in mouse models (5–8). These observations, along with gene expression studies in primary patient samples (9), establish aberrant JAK/STAT signaling as a central molecular hallmark of MPN pathogenesis. Clonality studies using MPN patient samples have traced driver mutations, including JAK2^V617F, to the HSC compartment, regardless of clinical phenotype (10). Furthermore, JAK2^V617F can be detected across the hematopoietic ontogeny (11, 12). This implies that JAK2^V617F can have differential effects on discrete cell types, such as progenitor cells versus mature lineage-committed cells (13). However, most studies performed to date have investigated the effects of JAK2^V617F when expressed throughout hematopoiesis.

We initiated this study to explore the contributions of mutant megakaryocytes to MPN pathogenesis. In many patients with essential thrombocytopenia/primary myelofibrosis, JAK2^V617F manifests primarily in the megakaryocyte (Mk) lineage, with less significant involvement of erythroid and myeloid cells as determined by mutant allele burden profiling (14) of specific hematopoietic subsets. Furthermore, aberrant megakaryopoiesis is a pathologic hallmark of MPN, regardless of clinical subtype (15). These abnormal megakaryocytes are known to secrete increased levels of proinflammatory cytokines and other factors (such as TGFβ), which presumably contribute to various MPN-related pathologies, such as bone marrow fibrosis (16, 17). Megakaryocytes are also regulators of HSC quiescence (18, 19), and we previously revealed that Jak2-deficient Mks negatively regulate stem/progenitor cell expansion in vivo (20). These observations raise the possibility that JAK2^V617F-mutant Mks promote MPN pathogenesis, at least in part, by influencing the biology of nonclonal (JAK2 wild-type) cells. Indeed, earlier work has revealed that Jak2^V617F-mutant megakaryocytes promote hyperproliferation of the stem/progenitor cell pool via the TPO/MPL signaling axis (21, 22). However, it remains unclear whether Jak2^V617F-mutant megakaryocytes can affect nonclonal hematopoietic cells, how they mediate these effects, and if Jak2^V617F-mutant megakaryocytes are required for maintenance of the disease state.

To this end, we crossed a conditional Jak2^V617F knock-in mouse where mutant Jak2 is expressed from the endogenous murine locus, with Pf4-Cre transgenic mice to produce a model wherein Jak2^V617F is expressed specifically in megakaryocyte lineage-committed cells (13, 23). While steady-state megakaryopoiesis was significantly elevated as expected, we unexpectedly observed that Jak2-mutant Mks/platelets significantly increased steady-state erythropoiesis and were capable of initiating a myeloproliferative disorder through cell nonautonomous mechanisms.

All mice used in this study were on the C57BL/6 background. Floxed heterozygous Jak^V617F/+ knock-in animals (a gift of Dr. Ann Mullally, Brigham and Women's Hospital of Harvard Medical School, Boston, MA; ref. 13) were bred with Pf4-Cre transgenic mice (a gift of Dr. Radek Skoda, University of Basel, Basel, Switzerland; ref. 23) to induce Jak2^V617F expression in megakaryocyte lineage–committed cells. Vav-Cre (Stock# 008610) and Mx1-Cre (Stock# 003556) transgenic mice were purchased from Jackson Laboratory and bred with Jak2^V617F/+ mice to induce pan-hematopoietic Jak2^V617F expression. For megakaryocyte depletion studies, iDTR mice (which harbor a conditional Cre-inducible diphtheria toxin receptor (DTR); ref. 24) were obtained from Jackson Laboratory (Stock# 007900) and crossed to Pf4-Cre⁺ mice. For lineage tracing studies, mTmG Cre switch reporter mice (25) were obtained from Jackson Laboratory (Stock# 007676) and bred to Jak2^V617F/+; Pf4-Cre⁺ mice to confirm lineage-specific Cre expression. Animal studies were approved by the Institutional Animal Care and Use Committee of both Northwestern University (Chicago, IL) and Memorial Sloan Kettering Cancer Center (New York, NY).

Antibodies used to characterize mouse cell surface makers by flow cytometry include mouse PE/Cy7-CD41 (catalog no. 25-0411-80) and APC-Gr1 (catalog no. 17-5931-82) were purchased from eBioscience. Mouse PE-CD42b antibody (catalog no. M040-3) was purchased from Emfret Analytics. Antibodies for mouse APC-Ter119 (catalog no. 116211), PE/Cy7-CD71 (catalog no. 113811), FITC-Mac1 (catalog no. 101205), APC/Cy7-cKit (catalog no. 105825), FITC-Sca1 (catalog no. 108105), PerCP/Cy5.5-FcγR (catalog no. 101323), PE-CD150 (catalog no. 115903), APC-CD105 (catalog no. 120413), and PE-CD34 (catalog no. 128609) were purchased from BioLegend. PE-Phospho-Stat5 (catalog no. 5387) antibody was purchased from Cell Signaling Technology. Rabbit polyclonal Von Willebrand Factor (VWF) antibody (catalog no. A008229-5) used for IHC was purchased from Agilent Technologies. Recombinant mouse stem cell factor (mSCF, catalog no. 250-03), recombinant human EPO (hEPO, catalog no. 100-64), recombinant mouse IL3 (mIL3, catalog no. 213-13), recombinant human TPO (hTPO, catalog no. 300-18), recombinant mouse Cxcl1 (mCxcl1, catalog no. 250-01), and recombinant mouse Cxcl2 (mCxcl2, catalog no. 250-15) were purchased from PeproTech.

Bone marrow and spleen cells were harvested from mice. Surface marker staining for mouse CD41, CD42b, Ter119, CD71, Gr1, and Mac1 was performed by incubating cells in antibodies diluted in PBS + 0.5% BSA for 30 minutes. To characterize the myeloerythroid stem/progenitor cell compartment, bone marrow cells were stained as follows: cells were first stained using a mouse hematopoietic progenitor enrichment kit (StemCell Technologies) containing CD5, CD11b/Mac1, CD19, CD45R, Gr1, and Ter119. Cells were then incubated with Pacific Blue–conjugated streptavidin and simultaneously with antibodies against cKit, Sca1, FcγR, CD41, CD150, and CD105 as described previously (26).

Levels of phosphorylated Stat5 were determined as described previously (8). For intracellular phospho-protein analysis, freshly isolated whole bone marrow cells were first resuspended in 1-mL RPMI + 1% BSA and incubated at 37°C for 1 hour. Cells were then stimulated with hEPO (1 U/mL) or hTPO (10 ng/mL) with or without mIL3 (10 ng/mL) for 10 minutes. Labeling for surface antigens was performed. Cells were then fixed in 16% paraformaldehyde at room temperature for 10 minutes, washed twice with PBS + 2% BSA, and permeabilized in ice-cold 95% methanol for 10 minutes. Cells were then stained immediately with PE-phospho-Stat5 in the dark at room temperature for 20 minutes and immediately analyzed.

FACS data were acquired using an LSR II flow cytometer (BD Biosciences) and analyzed using FlowJo software version 9.9.4 (Tree Star). Cells were sorted using a FACS Aria (BD Biosciences).

Age- and gender-matched symptomatic Jak2^V617F/+; Pf4-Cre⁺ mice were randomized to receive an antibody against mouse IL6 (R&D Systems, catalog no. MAB406) at 0.3 mg/kg or PBS every other day by intraperitoneal injection for 6 weeks. Jak2^V617F/+; Pf4-Cre⁺ mice were classified as symptomatic if peripheral blood analysis indicated an elevated (>60%) hematocrit.

Retroviruses were generated by transfecting Plat-E cells with MIGR1 containing either human MPL^W515L or JAK2^V617F using Fugene VI (Roche Life Sciences) according to the manufacturer's instructions. Viral supernatant was harvested at 48 hours posttransfection. cKit⁺ HSPCs were isolated from mouse bone marrow using microbeads (catalog no. 130-091-224, Miltenyi Biotec), and transduced by mixing cells in viral supernatant with 8 μg/mL polybrene (catalog no. TR-1003-G, Millipore Sigma) and centrifuging the mixture for 2,500 rpm for 90 minutes at 32°C. Transduced cells were cultured overnight, and transduction efficiency was confirmed by assessing GFP expression using flow cytometry. A total of 3 × 10⁵ GFP⁺ cells were transplanted in to lethally irradiated (11 Gy) recipient C57BL/6 mice via tail vein injection.

To assess disease transplantability, bone marrow cells were collected from Jak2^V617F/+; Pf4-Cre⁺ mice (CD45.2) or wild-type (WT; CD45.1). A total of 2 × 10⁶ Jak2^V617F/+; Pf4-Cre⁺ and WT bone marrow cells were transplanted into lethally irradiated 2-month-old WT mice at a ratio of 90:10 or 10:90.

Relative levels of Jak2^V617F and Jak2^WT were measured using a previously established protocol (13). RNA was isolated from purified cell populations using TRIzol reagent (catalog no. 15596026, Thermo Fisher Scientific), and cDNA was generated using a Verso cDNA synthesis kit (catalog no. AB-1453/B, Thermo Fisher Scientific). Relative quantification of Jak2^V617F and Jak2^WT transcripts was then performed with SYBR Green reagents (catalog no. 4913850001, Sigma) using the Applied Biosystems QuantStudio 7 real-time system. The following primer sequences were used: Jak2^WT forward primer 5′- TTTGAATTATGGTGTCTGCG; Jak2^V617F forward primer 5′- TTTGAATTATGGTGTCTGCT; Jak2 common reverse primer 5′- CAGGTATGTATCCAGTGATCC.

Levels of circulating cytokines were performed using a Luminex-based Mouse 32-Plex Cytokine Kit (catalog no. MCYTMAG-70K-PX32, EMD Millipore). Analysis was performed on serum samples that were prepared according to Millipore's instructions. In brief, peripheral blood was collected and allowed to clot for 30 minutes. Afterwards, blood samples were centrifuged at 1,000 × g for 10 minutes. Serum samples were then aliquoted (∼25 μL) in to clean tubes and frozen at −80°C until analysis. Before analysis, samples were diluted 1:2 in serum matrix and assayed according to the manufacturer's instructions. Data were acquired using the FlexMAP 3D system and xPONENT software (Luminex), and analyzed using MILLIPLEX Analyst software (EMD Millipore) as described previously (27). Single-plex cytokine-specific ELISAs (R&D Systems), including Cxcl1 (catalog no. MKC00B), Cxcl2 (catalog no. MM200), Ccl11 (catalog no. MME00), and IL6 (catalog no. M6000B), were then used to validate the changes observed in individual cytokines.

EPO-dependent colony formation was assessed using cytokine-free methylcellulose-based media (catalog no. m3234, StemCell Technologies) supplemented with mSCF (10 ng/mL), hEPO (10 U/mL), and varying concentrations of mIL6, mCxcl1, or mCxcl2. LSKs or MEPs were sorted from 6- to 8-week-old WT C57Bl/6 mice and seeded at 3,000 cells per well in triplicate. For erythroid colony formation assays (BFU-E), mouse splenocytes were seeded at 50,000 cells per well in triplicate using m3436 media (StemCell Technologies) supplemented with hEPO. In all assays, colony formation was scored at 7–10 days after plating.

Tissues were fixed in 10% neutral buffered formalin, embedded in paraffin, and stained with hematoxylin and eosin to assess gross cellular histology or reticulin to assess fibrosis. For VWF IHC, antigen retrieval was achieved by incubating slides in 1× target retrieval solution (catalog no. S1699, Dako North America) at 98°C for 30 minutes in a water bath. Sections were stained for VWF (catalog no. GA52761-2, Dako North America) overnight at 4°C at a dilution of 1:200, then incubated with rabbit HRP (catalog no. M4U534, Biocare Medical) for 15 minutes before the addition of chromogen DAB (catalog no. BDB2004; Biocare). Sections were then counterstained with hematoxylin. All slide images were obtained on a Leica DM4000B microscope equipped with a Leica DFC320 color digital camera.

Different groups were reported as mean ± SEM or mean ± SD and compared using unpaired two-sided Student t test. When multiple comparisons were necessary, one-way or two-way ANOVA with posttest Bonferroni correction was used. Statistical significance was established when P < 0.05 (labeled as *) and P < 0.01 (labeled as **). All analysis was performed using GraphPad Prism. Statistical analysis of survival was performed via the Kaplan–Meier method, with significance determined using the Mantel–Cox long-rank test. Group sizes were determined by power calculation to determine the sample size needed to achieve an 80% chance of detecting a significant difference (P < 0.05) between groups. The researchers who analyzed the samples for flow cytometry and histology were not aware of the genotypes.

To study the role of megakaryocytes in MPN pathogenesis, we generated Jak2^V617F/+; Pf4-Cre⁺ mice where mutant Jak2 is expressed from the endogenous Jak2 locus specifically in megakaryocyte lineage–committed cells (8, 23). Our model is distinct from prior studies, which employed a transgenic Jak2^V617F model, where Jak2 is expressed at nonphysiologic levels from a nonspecific integration site (21). Prior studies have suggested that Pf4 is expressed at low levels in HSPCs (28, 29), which raises the possibility that the Cre-mediated recombination (and subsequent expression of Jak2^V617F) occurs prior to megakaryocyte lineage commitment in our model. We took several approaches to confirm the specificity of Pf4-Cre in our model. First, we performed allele-specific qPCR on sorted cell populations to assess Jak2^V617F expression across the hematopoietic ontogeny. We sorted long-term HSCs, bivalent Mk/erythroid progenitor cells (Pre-MegE), erythroid progenitor cells (Pro-ERY), Mk progenitors (MkP) as described previously (26), in addition to whole bone marrow mononuclear cells (MNC), CD41⁺ Mks, and platelets from Jak2^V617F/+; Pf4-Cre⁺ mice, and performed allele-specific qPCR for Jak2^V167F. We could not detect significant expression of Jak2^V617F in MNCs, LT-HSCs, Pre-MegEs, or Pro-ERYs, whereas we could detect significant levels of Jak2^V617F in MkPs, and equal levels of Jak2^V617F and Jak2^WT in CD41⁺ Mks and platelets (Fig. 1A). As Jak2^V617F could only be detected in megakaryocyte lineage–committed populations and was not detectable in earlier progenitor compartments or in erythroid lineage–committed cell types, we could thus conclude that the Pf4 promoter drives Cre expression specifically in megakaryocyte lineage–committed cells.

We also employed the mTmG Cre switch reporter mouse (25) to confirm the lineage specificity of Pf4-Cre. In this model, a double-fluorescent Cre reporter cassette is within the Rosa26 locus. Prior to Cre-mediated recombination, tdTomato is expressed. In Cre-positive cells, the locus undergoes a recombination event that switches the cassette to express GFP. Therefore, Cre-negative cells will be tdTomato⁺, while Cre-positive cells will be GFP⁺. We crossed Jak2^V617F/+; Pf4-Cre⁺ mice with mTmG^+/+ mice, which were then subjected to extensive flow cytometric analysis to survey reporter expression across hematopoiesis. No significant population of GFP⁺ cells could be detected in the LSK compartment, in Mk/Erythroid progenitors (MEPs) or in CD71⁺/Ter119⁺ erythroid cells, whereas the CD41⁺ Mk compartment was approximately 99% GFP⁺ (Fig. 1B).

Jak2^V617F causes constitutive downstream activation of Stat5 (3, 8). We therefore sought to further confirm Jak2^V617F expression at the functional level by assessing levels of phosphorylated Stat5 (pStat5) by intracellular flow cytometry. In CD41⁺ Mk cells from Jak2^V617F/+; Pf4-Cre⁺ mice, basal levels of pStat5 were similar to levels observed in hTPO-stimulated cells (with or without mIL3; Fig. 1C). In contrast, basal pStat5 levels in Jak2^V617F/+; Pf4-Cre⁺ CD71⁺ erythroid cells were similar to levels observed in Jak2^+/+; Pf4-Cre⁺ CD71⁺ cells (Fig. 1D). These observations provide compelling evidence that Pf4-Cre is active exclusively in megakaryocytes, and thus drives Jak2^V617F expression in megakaryocyte lineage–committed cells.

Because the Jak2^V617F allele used in our model causes a distinct phenotype from the transgenic allele used by Zhan and colleagues, which was not expressed from the endogenous locus (21), we assessed whether physiologic expression of Jak2^V617F in Mks would induce a MPN phenotype. We monitored the peripheral blood counts of Jak2^V617F/+; Pf4-Cre⁺ mice for 200 days, and observed progressive and significant increases in the platelet counts, as well as an increase in the hematocrit and hemoglobin levels (Fig. 2A–C). We were not able to detect a significant population of Cre⁺ (GFP⁺) cells by flow cytometry (Fig. 1B), and additional allele-specific qPCR on sorted CD71⁺/Ter119⁺ cells confirmed that this population was Jak2^+/+ (Supplementary Fig. S1). Therefore, the aberrant erythropoiesis in our model is due to the interactions of Jak2^V617F-mutant Mks with the Jak2^WT hematopoietic milieu and not due to the outgrowth of a select subset of Pf4⁺ HSPCs.

Further analysis of the hematopoietic compartment at 6 months revealed evidence of an overt MPN. We observed splenomegaly (Fig. 2D and E), along with marked changes in splenic architecture, coupled with a profound increase in erythroid progenitors, white pulp expansion, increased megakaryocytes, and marked fibrosis (Fig. 2F–H). We also observed impaired survival in Jak2^V617F/+; Pf4-Cre⁺ mice relative to littermate controls (Fig. 2I). We noted that the peripheral phenotype in our model emerged with a significantly longer latency than what is typically seen in models of pan-hematopoietic Jak^V167F expression. We thus compared Jak2^V617F/+; Pf4-Cre⁺ mice to Jak2^V617F/+; Vav-Cre⁺ mice (where Jak2^V617F is expressed in every hematopoietic cell). While fully penetrant and similar, the disease phenotype of Jak2^V617F/+; Pf4-Cre⁺ was less severe (Supplementary Fig. S2), suggesting other cell types expressing JAK2^V617F also contribute to the phenotype in Jak2^V617F/+; Vav-Cre⁺ mice.

As expected, flow cytometry confirmed the histopathologic observation of megakaryocytic expansion in the bone marrow and spleen (Fig. 3A and B; Supplementary Fig. S3A and S3B). Jak2^V617F/+; Pf4-Cre⁺ mice displayed a robust expansion of Jak2^WT immature erythroid cells (Ter119^low/CD71^high R1 and Ter119^high/CD71^high R2 cells; refs. 30, 31) in the spleen (Fig. 3C) and in the R1 population in the bone marrow (Supplementary Fig. S3C). The granulocyte lineage was also expanded in both organs (Fig. 3D; Supplementary Fig. S3D). Analysis of the HSPC compartment confirmed the finding that Jak2^V617F Mks cause HSPC expansion (21). Lin⁻Sca1⁺cKit⁺ (LSK) HSPCs were expanded in bone marrow and spleen (Fig. 3E; Supplementary Fig. S3E). Downstream myeloid progenitor populations were also expanded, and skewed toward the Mk/erythroid lineage (Fig. 3F–H; Supplementary Fig. S3F–S3H). Splenic progenitor cells from Jak2^V617F/+; Pf4-Cre⁺ mice also produced more BFU-E, CFU-Mk, and CFU-Myeloid colonies in vitro (Supplementary Fig. S4A–S4C).

Because overt MPN appeared in the Jak2^V617F/+; Pf4-Cre⁺ model after a relatively long latency period (∼20 weeks), we analyzed the composition of the hematopoietic compartment prior to symptom onset. Young asymptomatic Jak2^V617F/+; Pf4-Cre⁺ showed only enhanced megakaryopoiesis (increased platelet counts; Supplementary Fig. S5A) and Mk expansion in both spleen (Supplementary Fig. S6A and S6B) and bone marrow (Supplementary Fig. S7A and S7B). All other compartments were comparable with littermate controls (Supplementary Figs. S6C–S6K and S7C–S7K). No tissue abnormalities or fibrosis were apparent upon histologic examination (Supplementary Fig. S5G–S5I). These data suggest that the duration of exposure to Jak2^V617F Mks is an important factor to MPN development, and that the effect of mutant Mks on the HSPC niche occurs is exerted over time. Furthermore, LSK and MEP cells sorted from symptomatic Jak2^V617F/+; Pf4-Cre⁺ mice did not show EPO hypersensitivity, which is an established feature of polycythemia vera (Supplementary Fig. S4D and S4E). This observation, combined with the long latency period, strongly suggests that Jak2^V617F-mutant Mks can promote neoplastic growth of Jak2^WT cells in a cell nonautonomous manner.

To investigate whether the Jak2^V617F-mutant Mks is cell nonautonomous, we performed bone marrow transplantation experiments. We transplanted lethally irradiated recipients with either 90% or 10% Jak2^V617F/+; Pf4-Cre⁺–derived bone marrow mixed with wild-type support, and monitored recipients for symptom onset. Only recipients receiving 90% Jak2^V617F/+; Pf4-Cre⁺–derived bone marrow showed signs of the polycythemia vera phenotype (elevated hematocrits and hemoglobin, splenomegaly, expanded erythroid populations, expanded white pulp) observed in primary mice (Fig. 4A, B, F, G, and H). Platelet counts were elevated over time in recipients receiving 90% Jak2^V617F/+; Pf4-Cre⁺–derived bone marrow (Fig. 4C), and neither group showed significant leukocytosis (Fig. 4D). These data support our hypothesis that the observed phenotype is mediated by mutant Mks in a cell nonautonomous manner.

Mks secrete numerous cytokines and chemokines that regulate the behavior of neighboring cells, including HSPCs (18, 19, 32–34). We hypothesized that mutant Mks could secrete factors that influence the differentiation and proliferation of HSPCS to promote the neoplastic growth of Jak2^WT cells. Milliplex cytokine/chemokine profiling of sera revealed increases in the circulating levels of several cytokines and chemokines, including IL6, Ccl11, Cxcl1, and Cxcl2 (Fig. 5A), in Jak2^V617F/+; Pf4-Cre⁺ mice relative to littermate controls. ELISA measurements validated a 4-fold increase in IL6 (Fig. 5B), which is upregulated in Jak2^V617F-mutant Mks isolated from patients with polycythemia vera (35), as well as significant increases in Ccl11, Cxcl1, and Cxcl2 (Supplementary Fig. S8A–S8C). To confirm that Jak2^V617F/+; Pf4-Cre⁺ Mks are a source of IL6, we quantified IL6 mRNA expression in FACS isolated CD41⁺ Mks, and found that IL6 transcripts were markedly elevated in Jak2^V617F/+; Pf4-Cre⁺ bone marrow Mks (Fig. 5C). We next sought to determine whether any of these factors could independently promote increased erythropoiesis, and thus contribute to the polycythemia observed in our model. We cultured Jak2^WT HSPCs (sorted LSKs or MEPs) in methylcellulose under erythroid colony-promoting conditions, and exogenous IL6 treatment (but not Cxcl1 or Cxcl2) increased EPO-dependent colony formation (Fig. 5D and E; Supplementary Fig. S8D and S8E). MEPs only responded 2-fold to increased IL6 levels, while LSKs showed a > 15-fold increase. These data suggest that LSKs are more responsive to IL6, and that (Mk-derived) IL6 acts on uncommitted HSPCs as well as on erythroid progenitors.

To assess a potential requirement for elevated IL6 in our model, we treated Jak2^V617F/+; Pf4-Cre⁺ mice with an antibody against IL6. The antibody was well-tolerated, as treated animals showed so significant weight loss (Supplementary Fig. S9). While treatment did not change peripheral blood counts (Supplementary Fig. S9), IL6 blockade suppressed erythropoiesis (Fig. 5F-H). Treated splenocytes also formed fewer erythroid colonies (BFU-E), indicative of decreased erythroid progenitors (Fig. 5F). Accordingly, analysis of the spleens revealed a reduced frequency of erythroid progenitor cells (Pre-CFU-E and Pre-MegE; Fig. 5G) in antibody-treated mice. Furthermore, IL6 blockade significantly reduced the proportion of immature R1 and R2 erythroid cell populations (Fig. 5H). We also treated age- and gender-matched Jak2^WT littermate controls to assess the requirement of IL6 on steady-state erythropoiesis, and observed a similar effect (Supplementary Fig. S10). These data demonstrate that increased IL6 contributes to the aberrant expansion of Jak2^WT erythroid cells observed in our model.

Our data suggest that Mks are a source of IL6 in our model. However, this does not preclude the possibility that additional cell types could contribute to the increase in IL6. Circulating platelets could also activate mature myeloid cells, or erythroblastic island macrophages, thus potentiating the disease phenotype. To determine whether activated myeloid cells or erythroblastic island macrophages could serve as contributing sources of IL6, we also examined IL6 expression in sorted mature F4/80⁺ macrophages or bulk Gr1⁺/Mac1⁺ myeloid cells (Supplementary Fig. S11A and S11B). IL6 expression was not elevated in these populations, which suggests that Jak2^V617F-mutant Mks are the primary contributor to the systemic increase in IL6 observed in our model.

Given that Jak2^V617F-mutant MKs are sufficient to cause aberrant erythropoiesis (21, 22), we next sought to determine whether the Mk lineage is necessary for disease maintenance. We first crossed a Cre-inducible Diphtheria toxin receptor transgenic mouse (iDTR^+/−) with Pf4-Cre⁺ mice to generate a model for selective Mk depletion (23, 24). We then retrovirally transduced progenitor cells from either iDTR^+/−; Pf4-Cre⁺ mice or iDTR^+/−; Pf4-Cre⁻ controls with Jak2^V617F, and transplanted infected cells into irradiated recipients. As expected, Jak2^V617F-transduced mice developed the characteristic polycythemia vera–like phenotype, which is characterized by elevated hematocrit and hemoglobin (Fig. 6A). Diphtheria toxin (DT) treatment significantly reduced both the platelet counts and the megakaryocytes in iDTR^+/−; Pf4-Cre⁺ mice (Fig. 6A–C, F, and G). Mk/platelet loss was accompanied by a significant decrease in disease indicators, specifically in hematocrit/hemoglobin levels (Fig. 6A). Furthermore, we observed a significant reduction in the terminally differentiated mature erythroid compartment (R3+R4 populations) in both bone marrow and spleen following DT treatment/Mk depletion (Fig. 6D and E). Importantly, DT treatment also resulted in a significant reduction in IL6 levels (Supplementary Fig. S12), further confirming IL6 as a central mediator of the cell nonautonomous erythropoiesis observed in this model. We did not observe any significant effect of Mk depletion on white blood cell counts (Fig. 6A). However, white blood cell counts are only modestly elevated in this MPN model.

To explore the larger relevance of aberrant JAK/STAT pathway activation in Mks to MPN maintenance, we repeated the Mk depletion studies using a second MPN transplant model: the MPL^W515L bone marrow transplant model. MPL^W515L mutation is detected in patients with Jak2^WT essential thrombocythemia/primary myelofibrosis, and also acts as a driver of MPN via constitutive activation of JAK2 and downstream STAT signaling. Retroviral overexpression of MPL^W515L in vivo results in a penetrant, rapid, and lethal myeloproliferative disease that is characterized by thrombocytosis, leukocytosis, and splenomegaly (without evident polycythemia; ref. 36; Fig. 6H and I). Mk depletion in MPL^W515L-expressing in iDTR^+/−;Pf4 Cre⁺ mice reduced Mk numbers in the bone marrow and spleen (Fig. 6J and K), as well as reduced platelet counts (Fig. 6H). Leukocyte counts and spleen weight were also significantly reduced (Fig. 6H and I). Mk depletion also caused a significant decrease in Lineage⁻ cKit⁺Sca1⁻ myeloid progenitor cells in the bone marrow (Fig. 6L), as well as restoration of splenic architecture (Fig. 6M). Therefore, Mk depletion significantly reduced disease burden in a model of MPL^W515L-driven MPN. Together with our findings from the Jak2^V617F transplant model, this reveals an unexpected role of Mks in MPN.

In this study, we developed a megakaryocyte lineage-specific Jak2^V617F knock-in mouse model to investigate the role of mutant megakaryocytes in MPN pathogenesis. Our data, consonant with studies of nonphysiologic Jak2^V617F expression in MKs with less clear phenotypes (21, 22), establishes that mutant Mks interact with the wild-type hematopoietic milieu to both initiate and sustain MPN. Our model is characterized by a strong polycythemia vera–like phenotype, including: impaired overall survival, splenomegaly, and aberrant erythropoiesis (elevated hemoglobin and hematocrit, and expansion of the erythroid progenitor compartments). It is worth noting that the phenotypic changes, while characteristic of polycythemia vera in general, were far less pronounced than the phenotype of established panhematopoietic Jak2^V617F models and occurred after a longer latency period. Therefore, we conclude that Jak2^V617F-mutant Mks are sufficient to generate a polycythemia vera phenotype, and that Jak2^V617F-mutant Mks promote cell nonautonomous myeloerythroid expansion that likely contributes to MPN in concert with clonal expansion of mutant cells in different lineages. These observations are consistent with genetic and clinical correlative studies in a cohort of patients with MPN with a low JAK2^V617F allele burden. These studies reported impaired clinical outcome and disease progression in a cohort of patients with essential thrombocythemia/primary myelofibrosis, even though their hematopoietic compartment was comprised of a relatively low fraction of Jak2^V617F-mutant cells (37, 38). Furthermore, additional studies have revealed that JAK2^V617F allele burden shows little to no correlation with disease phenotype.

Proinflammatory cytokines and chemokines are frequently elevated in patients with MPN, and increased levels of circulating cytokines are associated with adverse survival. This cytokine-driven inflammatory state is hypothesized to be a key contributing cause of the constitutional symptoms of MPN. We initially observed that Jak2^V617F-mutant Mks caused low level inflammation by histopathology, including increased bone marrow fibrosis and expansion of the splenic white pulp. This led us to hypothesize that Jak2^V617F-mutant Mks/platelets could potentiate a myeloproliferative state via aberrant production of proinflammatory cytokines/chemokines. Indeed, we observed that Jak2^V617F-mutant Mks significantly elevated circulating levels of four discrete proinflammatory cytokines/chemokines: IL6, Ccl11, Cxcl1, and Cxcl2. Of these factors, only IL6 supported increased erythroid differentiation in vitro, which implicates this inflammatory mediator as a key promoter of MPN pathogenesis. Consistent with our observations, elevated IL6 levels are correlated with splenomegaly, survival, and JAK2 mutation status in patients with MPN (39). Furthermore, the improvement in constitutional symptoms seen with JAK2 inhibition in these patients correlated with reductions in serum IL6. A previous report has implicated IL6 in mediating the myeloproliferative phenotype observed in miR-146a knockout mice (40).

Mks could promote aberrant erythropoiesis either directly, via increased production of these factors, or indirectly, by recruiting and activating myeloid cells. We found that Jak2^V617F-mutant Mks show increased expression of IL6, and that IL6 blockade using a neutralizing antibody significantly impeded erythropoiesis in vivo.

In their transgenic model, Zhan and colleagues reported an essential thrombocythemia–like MPN phenotype with expansion of the stem/progenitor cell pool (21). Our model displays several notable differences–primarily the polycythemia vera–like and inflammatory aspects of the observed phenotype. These studies provide compelling in vivo evidence that Jak2^V617F-mutant Mks can promote the expansion of the wild-type hematopoietic compartment. We expanded upon this to show that Mk-derived factors, such as IL6, mediate (at least partially) the cell nonautonomous role of Mks in MPN pathogenesis. We also established a broader role for Mks in MPN maintenance by employing Mk ablation in two established MPN models. Mk depletion in both the JAK2^V617F and MPL^W515L bone marrow transplant MPN models significantly blunted the polycythemia and leukocytosis phenotypes of either model, respectively. This demonstrates that Mks are necessary for MPN maintenance in vivo. It also suggests that the contributions of MPN Mks are not driver mutation-specific, and are a broader consequence of JAK/STAT pathway hyperactivation with a critical role for JAK-STAT activation in Mks.

Collectively, our findings confirm and provide further mechanistic insight into the role of the megakaryocyte lineage in MPN pathogenesis. We demonstrate that mutant Mks are directly involved in MPN initiation and maintenance in vivo, and that Mks can promote the heterotypic expansion of wild-type HSPCs via aberrant cytokine production. These data support a novel model for MPN pathogenesis, in which MPN manifests as a genetically and phenotypically heterogeneous hematopoietic disorder, with both mutant and wild-type cells contributing to the pathogenic ecosystem. We expect that similar mechanisms underlie the development of other myeloid neoplasms as well as in nonhematopoietic malignancies, where malignant transformation is driven by a complex interplay between discrete clonal cell populations. Our findings also contribute to the collective understanding of how patients with MPN with a low frequency of mutant cells can present with fulminant disease, and highlight the therapeutic potential of targeting these cellular interactions (such as Mks and the HSPC niche) to achieve improved therapeutic benefit in MPN.

R.L. Levine has ownership interests (including patents) in Qiagen, Loxo, Isoplexis, Imago, and Auron; is a consultant/advisory board member for Novartis, Incyte, Roche, Morphosys, Janssen, and AbbVie; and reports receiving commercial research grants from Roche, Prelude, and Celgene. J.D. Crispino is a consultant/advisory board member for Sierra Oncology and MPN Research Foundation, and reports receiving commercial research grants from Forma Therapeutics and Scholar Rock. No potential conflicts of interest were disclosed by the other authors.

Conception and design: B. Woods, Q. Yang, R.L. Levine, J.D. Crispino, Q.J. Wen

Development of methodology: B. Woods, M. Bulic, Q. Yang, Q.J. Wen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Woods, W. Chen, S. Chiu, C. Marinaccio, C. Fu, L. Gu, M. Bulic, A. Zouak, S. Jia, P.K. Suraneni, Q.J. Wen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B. Woods, W. Chen, C. Marinaccio, L. Gu, M. Bulic, Q. Yang, P.K. Suraneni, K. Xu, R.L. Levine, J.D. Crispino, Q.J. Wen

Writing, review, and/or revision of the manuscript: B. Woods, W. Chen, S. Chiu, L. Gu, Q. Yang, R.L. Levine, J.D. Crispino, Q.J. Wen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Bulic, Q. Yang, A. Zouak, Q.J. Wen

Study supervision: Q. Yang, R.L. Levine, Q.J. Wen

This work was supported by grants from the NIH (1F99CA212481-01, to B. Woods; CA108671, to R.L. Levine and J.D. Crispino; HL112792, to J.D. Crispino; NCI R50CA211534, to Q.J. Wen) and the Samuel Waxman Cancer Research Foundation. This work was supported in part by MSKCC Support Grant/Core Grant P30 CA008748. We thank the Pathology Core Facility of the Robert Lurie Comprehensive Cancer Center of Northwestern University for processing mouse tissues and performing H&E and reticulin staining.

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.