HiJAKing the hematopoietic system: a low-frequency JAK2V617F clone drives myeloproliferative neoplasm pathology

Abstract

JAK2V617F is one of the most common mutations in clonal hematopoiesis of indeterminate potential (CHIP) and a major driver of myeloproliferative neoplasms (MPNs). To determine the impact of a low-frequency JAK2V617F clone on both the hematopoietic system and the bone marrow (BM) stroma, we developed a traceable murine MPN model, in which whole BM transplantation (BMT) was performed using CD45.2 5.0 × 10⁶ JAK2V617F donor cells transplanted into unconditioned CD45.1 recipient mice. BMT recipients developed a polycythemia vera–like phenotype (elevated hematocrit and leukocytosis) with a 2.7% average donor cell chimerism in peripheral blood. Eight months after BMT, RNA sequencing (RNA-seq) analysis of BM cells sorted according to CD45.1/CD45.2 expression showed significant upregulation of early erythroblast- and myeloid cell–specific transcripts, and downregulation of lymphoid transcripts in donor-derived cells compared to controls. Surprisingly, recipient-derived cells also showed upregulation of myeloid- and erythroblast-related transcripts, indicating a skewing of the non–JAK2V617F-carrying recipient hematopoietic system toward an MPN-like phenotype. In addition, RNA-seq analysis of the BM stroma from JAK2V617F BMT recipients indicated significant loss of osteomesenchymal transcripts. Consistently, micro–computed tomography imaging indicated loss of trabecular bone. In sum, our results indicate that low-frequency MPN-driving cells in unconditioned recipients not only impact hematopoiesis-supporting stroma but also profoundly influence unmutated cells, uniquely altering their transcriptomic and phenotypic profiles. These observations are challenging our current understanding of the etiology and therapeutic approaches to MPNs and other CHIP-associated diseases.

Introduction

JAK2V617F, the most common driver mutation in classical myeloproliferative neoplasms (MPNs),^1–4 is present in ~70% of all MPN cases, including 98% of polycythemia vera (PV), 55% of essential thrombocythemia (ET), and 60% of primary myelofibrosis. Intriguingly, some individuals with JAK2V617F display no pathology, a state termed clonal hematopoiesis of indeterminate potential (CHIP),⁵ whereas others may present with MPN, or develop secondary acute myeloid leukemia.⁶ Polygenic loci, sex, allelic burden, mutational order, or downstream STAT signaling have all been queried as possible contributors to the pleiotropy of JAK2V617F-dependent disease phenotypes.^6–12 However, the ability of JAK2V617F to drive this spectrum of phenotypes remains a mechanistic puzzle, especially given that JAK2V617F can trigger MPN without secondary mutations.^13,14 In addition, current JAK inhibitors do not show curative effect.¹⁵ Therefore, factors beyond overactive JAK/STAT signaling likely contribute to MPN pathology, and identifying these could lead to improved therapies.

JAK2V617F is among the top 10 mutations associated with CHIP,¹⁴ and, depending on detection sensitivity, shows a prevalence of 0.1% to 3.1% in the general population, with a corresponding MPN diagnosis rate of 2.3% to 76%.^6–12 Unlike other common CHIP-associated genes (eg, DNMT3A or TET2), clonal expansion of JAK2V617F displays high interindividual variability with no correlation to secondary mutations.^16,17 Lineage tracing analyses of hematopoietic stem and progenitor cells (HSPCs) from the bone marrow (BM) of patients with MPN suggest that the initial JAK2V617F mutation occurs decades before diagnosis.^18,19 Therefore, independence from secondary mutations, extensive latency before diagnosis, varied disease phenotypes, and the low therapeutic index of JAK inhibitors suggest that MPN progression is influenced by an unidentified clonally extrinsic factor.

Hematopoiesis primarily occurs within the BM and relies on supportive nonhematopoietic stromal cells, including mesenchymal stem and progenitor cells (MSPC) that have osteogenic, adipogenic, and chondrogenic potential. MSPCs produce a spectrum of cytokines, chemokines, and growth factors to support hematopoietic stem cell (HSC) homing, quiescence, and differentiation.^20–22 Osteosclerosis and BM fibrosis are commonly observed in primary myelofibrosis, whereas osteoporosis and increased fracture incidence are reported in PV and ET. These findings highlight the cross talk between the JAK2V617F clone and BM stroma, although the underlying mechanisms of JAK2V617F-linked phenotypic pleiotropy remain obscure.

MPNs are clonal, inflammatory, aging-related diseases with a 60-year median patient age at diagnosis.^23–28 JAK2V617F, when found within HSPCs, confers a competitive advantage, particularly in environments with high levels of inflammatory cytokines such as interleukin-6 (IL-6) or tumor necrosis factor α.²⁹ Hematopoietic cells carrying JAK2V617F can alter BM microenvironment and promote inflammatory cytokine production by unmutated MSPCs, thereby contributing to the expansion of the diseased clone through a feedback mechanism.^30,31 Obesity,^32,33 smoking,²⁴ and autoimmune diseases^34,35 are all associated with chronic inflammation and increased risk of MPN. However, it remains unclear if CHIP progression or MPN initiation are driven by increased inflammation, or if the diseased clone itself remodels its microenvironment resulting in increased inflammation.

Studying early MPN progression in a low-variant allele frequency (VAF) pathogenic mutation setting requires unique approaches given that patient samples from the very early stages of MPN are rarely available. Herein we present, to our knowledge, a new strategy whereby JAK2V617F MPN-like cells are transplanted into unconditioned recipient mice that as a result develop low-VAF–driven MPN. Absence of conditioning in JAK2V617F bone marrow transplantation (BMT) recipients allowed us to uncover profound direct impact of the JAK2V617F⁺ cells on both the microenvironment and unmutated native hematopoietic system in a clinically relevant setting.

Materials and methods

Mouse models

The parental donor MPN line was generated by crossing B6N.129S6(SJL)-Jak2^tm1.2Ble/AmlyJ (JAK2(VF/+), Jax no. 031658³⁶) with B6.Cg-Tg(Mx1-cre)1Cgn/J (Tg(Mx1-Cre)(+), Jax no. 003556³⁷). Progeny and crosses are described in Figure 1A and supplemental Table 1 (available on the Blood website). All experiments involving mice were approved by the Institutional Animal Care and Use Committee.

Figure 1. Establishment of a murine BMT donor line carrying JAK2MUT.

(A) Crossbreeding strategy for creation of the inducible MPN donor line. (B) Polymerase chain reaction analysis to determine the presence of knockin JAK2V617F genotype using primers J2KI-1 and J2KI-2³⁶ (left) and evidence of recombination post–Cre recombinase induction by poly(I:C) using primers J2KI-4 and Neo-Exc-RT.³⁶ (C) Serial CBCs of PB from the 4 donor genotypes and WT CD45.1 mice. Age ranges from 3 to 26 weeks, with a minimum of 1 data collection point per animal (n = 26–77). Locally estimated scatterplot smoothing (LOESS) trendline displaying 95% confidence interval. Shading defines 90% confidence interval. Outlier values were capped to reduce y-axis skewing for PLTs (maximum, 5.0 × 10⁶/μL), LYMs (maximum, 1.8 × 10⁴/μL), and total WBC (maximum, 2.5 × 10⁴/μL). (D) Kaplan-Meier survival curve of JAK2MUT (n = 73) vs JAK2CTRL (n = 52) genotypes. (E) Representative visual comparison of organs (spleen, femur, and liver)and hind paws between non–JAK2MUT-genotyped (left) and JAK2MUT-genotyped animals (right). Non-JAK2MUT designation includes the genotypes WT, MxCTRL, and JAK2CTRL. (F) Representative total PB collected by cardiac puncture of non–JAK2MUT-genotyped (left) and JAK2MUT-genotyped animals (right). (G) Comparison of isolated liver and spleen weights from 10-week-old non–JAK2MUT-genotyped (n = 9) and JAK2MUT-genotyped animals (n = 12). (H) Representative hematoxylin and eosin (H&E) sections of spleen and femur BM from non–JAK2MUT-genotyped and JAK2MUT-genotyped animals at 10 weeks of age showing disrupted spleen morphology and increased megakaryocyte numbers in both spleen and femur. Scale bar, 125 μm. Zoomed-in boxes (75 × 75 μm) highlight representative megakaryocytes. (I) Representative reticulin-stained sections of spleen and femur BM from non–JAK2MUT- and JAK2MUT-genotyped animals at 10 weeks of age showing no increase in reticulin fibers of JAK2MUT animals. Scale bar, 125 μm. (J) Representative Prussian blue stained sections of spleen and humerus BM from non–JAK2MUT- and JAK2MUT-genotyped animals at 10 weeks of age displaying loss of stored iron in JAK2MUT animals. Scale bar, 125 μm. LYM, lymphocyte; PB, peripheral blood; PLT, platelet; WBC, white blood cell. Panel A was created in BioRender. Bonal, D. (2025) https://BioRender.com/p51l197.

BMT

BM isolated from 10-week-old female wild-type (WT) C57BL/6J (Jax no. 000664) or sex-matched JAK2(VF/+);Tg(Mx1-Cre)(+) (JAK2MUT) parental animals was used for BMT using a single tail vein injection in concentration ranges from 0.05 × 10⁶ to 15.0 × 10⁶ cells per mouse as described, except no recipient conditioning was used.^38,39 B6.SJL-Ptprc^a Pepc^b/BoyJ (Jax no. 002014⁴⁰) mice were used as recipients.

Flow cytometry and cell sorting

Donor chimerism in the peripheral blood (PB) and BM of mice that underwent transplantation and WT mice that did not undergo transplantation was assessed using anti-CD45.1 (A20)-FITC and anti-CD45.2 (104)-PerCP/Cy5.5 antibodies (Bio-Legend). Whole BM was isolated using a protocol developed by Tikhonova et al.⁴¹ Hematopoietic BM cells were gated to isolate CD45.2⁺ donor-derived and CD45.1⁺ recipient-derived cells using a BD Influx cell sorter. Stromal-enriched populations were sorted from phycoerythrin-stained lineage-negative, CD45.1/CD45.2 double-negative cells. The population markers used are described in the supplemental Methods. Gating strategies are presented in supplemental Figures 1–4.

RNA-seq

RNA was isolated from sorted BM (hematopoietic and non-hematopoietic) fractions, and RNA sequencing (RNA-seq) was conducted by Novogene (Sacramento, CA) using NovaSeq 6000 (Illumina), generating 150-bp paired-end reads. Quality control, alignment, and differential expression analyses were performed in-house.

CT and tissue analysis

All organs and tissues were processed and stained using standard Gömöri, Prussian, tartrate-resistant acid phosphatase, immunohistochemistry, and hematoxylin and eosin protocols and visualized using Olympus VS200 system with ASW 4.1.1 software (Evident Scientific). EDTA-decalcified bones were imaged on a Bruker in vivo SKYSCAN 1276 micro–computed tomography (CT) system.

Statistical analysis

All statistical analyses and graphing were performed within the R environment.⁴² Unpaired t tests were used to generate P values, and when required, Holm-adjusted P values (Padj) were applied.

Results

Successful establishment of an MPN donor line harboring JAK2MUT

To establish a murine model harboring inducible, traceable JAK2V617F-expressing donor cells, female B6N.129S6(SJL)-Jak2^tm1.2Ble/AmlyJ (JAK2(VF/+))³⁶ mice were crossed with male B6.Cg-Tg(Mx1-cre)1Cgn/J mice (Tg(Mx1-Cre)(+)),³⁷ yielding 4 genotypes: WT (JAK2(+/+);Tg(Mx1-Cre)(–)), MxCTRL (JAK2 (+/+);Tg(Mx1-Cre)(+)), JAK2CTRL (JAK2(VF/+);Tg(Mx1-Cre)(–)), and JAK2MUT (JAK2(VF/+);Tg(Mx1-Cre)(+)) (Figure 1A). All JAK2MUT mice were heterozygous because of homozygous lethality, as reported previously.³⁶ The survival, crossing patterns, and genotype distribution are summarized in supplemental Table 1.

At 5 weeks of age, 2 doses (10.0 mg/kg each) of poly(I:C) (polyinosinic:polycytidylic acid) were administered 48 hours apart, triggering Cre recombinase–driven JAK2V617F expression, confirmed by polymerase chain reaction (Figure 1B). Longitudinal complete blood count (CBC) analysis (between 4 and 26 weeks) indicated disease development exclusively in JAK2MUT mice (Figure 1C). Regardless of sex (supplemental Figure 5A), a PV-like pathology emerged at ~5 weeks of age, with elevated hemoglobin (Hb; >20.0 g/dL), hematocrit (HCT; >50%), and red blood cells (RBCs; >15.0 × 10⁶/μL), along with significantly elevated frequencies of granulocytes (GRANs; >3.0 × 10³/μL), monocytes (MONOs; >1.5 × 10³/μL), and white blood cells (WBCs; >15.0 × 10³/μL). Additionally, thrombocytosis (platelet [PLT] count, >15.0 × 10⁵/μL) was also noted at ~10 weeks of age. Abnormal CBC values observed before poly (I:C) administration suggest some poly(I:C)-independent recombination, as previously reported.⁴³ However, JAK2MUT mice with or without poly(I:C) treatment displayed comparable longitudinal disease manifestations³⁶ (supplemental Figure 5B).

The median survival of JAK2MUT animals was 14.4 weeks (P < .0001, both sexes, with and without poly(I:C); Figure 1D). However, when disregarding poly(I:C) administration, JAK2MUT males compared with females exhibited significantly decreased median survival (12.3 vs 18.2 weeks, respectively; P = .0031; supplemental Figure 6A). Across both sexes, poly(I: C) administration resulted in decreased survival compared with untreated animals (P = .049; supplemental Figure 6B), but this trend did not show significance in sex-specific analyses (supplemental Figure 6C–D). By 10 weeks of age, all JAK2MUT animals displayed classic PV traits, including organ, femur, and extremity reddening (Figure 1E) with increased PB volume (Figure 1F). Splenomegaly was observed with loss of white and expansion of red pulp (Figure 1G; supplemental Figure 7). Hematoxylin and eosin staining of BM sections revealed erythroid hyperplasia (Figure 1H; supplemental Figure 8). JAK2MUT mice displayed no evidence of splenic or BM fibrosis, and Prussian blue staining indicated loss of stored iron (Figure 1I–J; supplemental Figures 9 and 10).

JAK2V617F-mutant donor cells successfully engraft in unconditioned recipients after BMT

To optimize BMT into unconditioned primary recipients, we transplanted whole BM cells isolated from 10-week-old female C57BL/6J (CD45.2⁺) WT donors into WT B6.SJL-Ptprc^a Pepc^b/BoyJ (CD45.1⁺) recipients using a single tail vein injection with 15.0 × 10⁶ cells per recipient (supplemental Figure 11A). PB analysis in unconditioned BMT recipients revealed consistent donor cell chimerism averaging 3.49% (standard deviation [SD], 0.675) at 25 weeks (supplemental Figure 11B). BMT had no observable impact on body weight or induction of a disease phenotype, as indicated by longitudinal CBC comparisons with age-/sex-matched CD45.1 controls (supplemental Figure 11C–D). To assess the functionality of long-term HSCs in unconditioned recipients, a secondary BMT was conducted. CD45.2 cells were isolated from the BM of 4 primary BMT recipients (2.21%, 2.99%, 4.22%, and 4.51% in PB at 25 weeks after BMT). Cells isolated from each of the 4 donors were divided evenly and transplanted into 4 irradiated (950 Gy) secondary recipients at concentrations between 2.5 × 10⁵ and 6.25 × 10⁵ cells per recipient (supplemental Figure 12A). Given the low donor cell concentrations, irradiation ensured secondary engraftment. An average of 83% donor chimerism was observed by 17 weeks (supplemental Figure 12B), and all secondary BMT recipients survived past 26 weeks.

Applying an unconditioned BMT approach, 10-week-old recipients were injected with varying concentrations of JAK2MUT parental BM cells (5.0 × 10⁴, 5.0 × 10⁵, 1.0 × 10⁶, 5.0 × 10⁶, and 15.0 × 10⁶ cells per recipient; Figure 2A; supplemental Figure 13A). Dose-dependent donor chimerism was observed in BMT recipients (CD45.1⁺), but the proportion of circulating donor JAK2MUT (CD45.2⁺) cells never exceeded an average of 10% over ~25 weeks (Figure 2B; supplemental Figure 13A–B). Surprisingly, despite donor chimerism below 5% or 10% among recipients transplanted with 5.0 × 10⁶ and 15.0 × 10⁶ JAK2MUT cells per recipient, respectively, an MPN-like disease phenotype comparable to the JAK2MUT parental line was observed by 10 weeks after BMT, with elevated Hb (>20.0 g/dL), HCT (>50%), and RBCs (>15.0 × 10⁶/μL; Figure 2C). Unlike the parental JAK2MUT genotype, we did not observe aggressive expansion of GRANs, MONOs, WBCs, or PLTs (Figure 2C; supplemental Figure 13C). The median survival for unconditioned BMT recipients receiving 15.0 × 10⁶ JAK2MUT cells was 12.9 weeks, similar to that of the JAK2MUT parental line, whereas survival of unconditioned BMT recipients receiving 5.0 × 10⁶ JAK2MUT cells was comparable to that of WT controls (Figure 2D). No deaths were observed among the WT controls or unconditioned BMT recipients receiving 1.0 × 10⁶ JAK2MUT cells.

Figure 2. Dose-dependent BMT of JAK2MUT into unconditioned recipients.

(A) Transplantation strategy from MPN induction in donor at 5 weeks of age to whole BMT into unconditioned recipients at 10 weeks of age. (B) JAK2V617F (CD45.2) chimerism in PB of WT BMT recipients (CD45.1) detected by flow cytometry. BMT cell concentrations ranged between 0.0 × 10⁶ and 15.0 × 10⁶. Evaluated from 0 to 30 weeks after BMT. (C) Serial analysis of CBCs of PB from the JAK2MUT donor line, BMT recipients, and WT controls. Collection points were aligned to display between −1 and 30 weeks of age or weeks after BMT, with a minimum of 1 data collection point per animal (n = 10–78 individuals per group). Shaded regions represent LOESS trendline displaying 95% confidence interval. The dotted vertical line indicates 10 weeks after BMT. Outlier values were capped to reduce y-axis skewing for PLT count (maximum, 5.0 × 10⁶/μL), LYMs (maximum, 1.8 × 10⁴/μL), and WBCs (maximum, 2.5 × 10⁴/μL). (D) Kaplan-Meier analysis comparing JAK2MUT (n = 73), WT CD45.1 (n = 40), 1.0 × 10⁶ BMT (n = 14), 5.0 × 10⁶ BMT (n = 15), and 15.0 × 10⁶ BMT (n = 10) recipients. Shading represents 90% confidence interval. Panel A was created in BioRender. Bonal, D. (2025) https://BioRender.com/s99y894.

Small subset of JAK2V617F⁺ cells drives MPN-like disease in mice

Using our BMT strategy, we assessed the impact of low-frequency JAK2MUT cells in driving the MPN-like phenotype. We focused on unconditioned BMT recipients receiving 5.0 × 10⁶ JAK2MUT cells per recipient, tracking effects over 28 to 32 weeks after BMT, and comparing them with age- and sex-matched WT controls (Figure 2A). We used WT (CD45.1 no-BMT) mice as controls because their CBC profiles did not significantly differ from those of CD45.1 recipients of JAK2CTRL CD45.2 or WT CD45.2 BMT at doses of 5.0 × 10⁶ or 15.0 × 10⁶ cells (supplemental Figure 14). Donor cell chimerism in PB remained <10% for all mice, with female and male chimerism not exceeding 5% and 10%, respectively (Figure 2B; supplemental Figure 15A). Average donor chimerism in BM was at 10.2% (supplemental Figure 15B). Both male and female JAK2MUT BMT recipients showed reduced body weight gain (supplemental Figure 15C). Despite variable donor cell chimerism, both sexes displayed consistent MPN-like features, including elevated HCT, Hb, MONO, and RBC levels, compared with controls (supplemental Figure 15D). No significant increase in PLT count was observed, consistent with previous reports using various JAK2V617F models in BMT experiments.^44–47 Survival was not affected, with 2 mice euthanized before study completion because of ulcerative dermatitis (supplemental Figure 15E).

At 28 to 32 weeks after BMT, gross pathology revealed hepatosplenomegaly (Figure 3A–B) with expansion of spleen red pulp among the JAK2MUT BMT recipients (Figure 3C; supplemental Figure 16), similar to the JAK2MUT parental line (Figure 1H; supplemental Figure 7). A significant increase of megakaryocytes was evident in both BM and spleens of BMT recipients (Figure 3D; supplemental Figures 16 and 17). Livers of unconditioned BMT recipients displayed elevated weights with increased erythrocyte infiltration, but no evidence of extramedullary hematopoiesis (Figure 3A,C; supplemental Figure 18). Prussian blue staining indicated iron depletion in BM and spleen, likely due to increased erythropoiesis (Figure 3E; supplemental Figures 19 and 20; supplemental Table 2). No increase in splenic fibrosis was observed via reticulin staining (Figure 3F; supplemental Figures 21 and 22).

Figure 3. Characterization of PV-like phenotype in unconditioned recipients of JAK2MUT BMT.

(A) Comparison of the raw weight (grams) and relative percentage of body weight (organ weight per final body weight) of livers and spleens between WT no-BMT (n = 8) and 5.0 × 10⁶ BMT (n = 7) animals at 28 to 32 weeks after BMT. Unpaired t test (***P ≤ .001; **P ≤ .01; *P ≤ .05). (B) Visual comparison of isolated liver and spleen sizes of WT no-BMT (n = 8) and 5.0 × 10⁶ BMT (n = 7) at 28 to 32 weeks after BMT. The organs imaged were previously fixed in 10% buffered formalin. (C) Representative H&E sections of spleen, femur BM, and liver of WT no-BMT and 5.0 × 10⁶ BMT at 28 to 32 weeks after BMT showing dysregulation of spleen morphology and increased megakaryocyte numbers in both spleen and femur. Livers display increased infiltration of erythrocytes and no evidence of extramedullary hematopoiesis. Scale bar, 125 μm. Zoomed-in boxes (75 × 75 μm) highlight representative megakaryocytes. (D) The average number of megakaryocytes per 10× high-powered field in spleen and femur BM between WT no-BMT (n = 8) and 5.0 × 10⁶ BMT (n = 7) animals at 28 to 32 weeks after BMT. Unpaired t test (***P ≤ .001; **P ≤ .01; *P ≤ .05). (E) Prussian blue sections of spleen and femur BM of WT no-BMT and 5.0 × 10⁶ BMT at 28 to 32 weeks after BMT showing loss of stored iron within both spleen and BM of BMT recipients. Scale bar, 125 μm. (F) Reticulin-stained sections of spleen and femur BM of WT no-BMT and 5.0 × 10⁶ BMT at 28 to 32 weeks after BMT displaying no increase in concentration of reticulin fibers. Scale bar, 125 μm.

Presence of JAK2V617F clone profoundly impacts nonmutated hematopoietic cells

Unconditioned BMT recipients of 5.0 × 10⁶ JAK2MUT cells developed a PV-like phenotype after 28 weeks, with elevated HCT, Hb, RBC, and WBC counts despite low donor cell chimerism in PB (<5%; mean, 2.74%; SD, 2.07; Figure 4A; supplemental Figure 15A) and BM (~10%; mean, 10.20%; SD, 5.63; supplemental Figure 15B). Therefore, we next evaluated the impact of JAK2MUT clone on BMT recipients’ hematopoietic system using flow cytometry and transcriptomic analysis (supplemental Figure 23A). BM hematopoietic cells were isolated from 28- to 32-week-old unconditioned BMT recipients, and age- and sex-matched WT no-BMT controls. CD45.1 (recipient-derived) and CD45.2 (donor-derived) cells were sorted (supplemental Figure 23B). RNA was isolated from these populations and from WT no-BMT controls, and then subjected to RNA-seq.

Figure 4. Presence of JAK2MUT clone profoundly impacts cellular profiles of nonmutated hematopoietic fraction.

(A) CBCs of PB at the final time point (28–32 weeks after BMT) from WT no-BMT (n = 8) and 5.0 × 10⁶ BMT (n = 7) animals. Unpaired t test (***P ≤ .001; **P ≤ .01; *P ≤ .05). (B) Principal component analysis of RNA isolated from flow-sorted BM based on CD45 expression. WT no-BMT (CD45.1⁺, n = 8), 5.0 × 10⁶ BMT CD45.1⁺ (recipient derived, n = 7), and 5.0 × 10⁶ BMT CD45.2⁺ (donor derived, n = 7). Ellipses were generated using multivariate t distribution with a 90% confidence interval. Dots and triangles indicate female or male sex, respectively. (C) Volcano plot displaying DEGs in 2 separate comparisons: CD45.1 recipient derived vs WT no-BMT (upper) and CD45.2 donor derived vs WT no-BMT (lower). Horizontal line indicates Padj of .05. Vertical lines indicate log₂FC of −0.25 and 0.25. (D) Combined gene set enrichment analysis results using the Hay BM cell type gene sets and the complete differential expression tables generated when comparing donor derived (left, red) or recipient derived (right, blue) against WT no-BMT CD45.1. Bars indicate normalized enrichment scores with q values ≤0.05 (lighter color) or ≤0.25 (darker color). (E) Flow cytometry analysis indicating the percentage of cell populations within CD45 gate. Unpaired t test (****Padj ≤ .0001; ***Padj ≤ .001; **Padj ≤ .01; *Padj ≤ .05). The gating strategy is explained in supplemental Figure 24A. WT no-BMT (CD45.1⁺, n = 8), 5.0 × 10⁶ BMT CD45.1⁺ (recipient derived, n = 7), and 5.0 × 10⁶ BMT CD45.2⁺ (donor derived, n = 7). (F) Flow cytometry analysis representing origin/derivation of hematopoietic populations within total CD45⁺ BM of WT no-BMT (left) and BMT recipients (right), presented as a stacked bar graph. Unpaired t test (***P ≤ .001; **P ≤ .01; *P ≤ .05). The gating strategy is explained in supplemental Figure 24B. WT no-BMT (CD45.1⁺, n = 8), 5.0 × 10⁶ BMT CD45.1⁺ (recipient derived, n = 7), and 5.0 × 10⁶ BMT CD45.2⁺ (donor derived, n = 7). B, B cell; EO, eosinophil; ERP, erythrocyte progenitor; LMPP, lymphoid-primed multipotent progenitor; MAST, mast cell; MEP, megakaryocyte/erythroid progenitor; NK, natural killer; PC, plasma cell; POS positive.

To confirm the absence of RNA contamination from CD45.2 donor-isolated cells in the CD45.1 recipient-derived fraction, we examined all raw FASTQ reads covering the JAK2V617F region (Chr19:29 272 328). Only 1 of 7 recipient-derived samples displayed minimal contamination, with 3 out of 255 reads (~1%) carrying the mutation. For comparison, 41% of reads in CD45.2 donor-isolated samples carried the G→T mutation, averaging 188 reads per sample (supplemental Table 3). We further verified contamination levels of CD45.2 and CD45.1 fractions within the Ptprc locus using the single nucleotide polymorphism at Chr1:137 992 172 to distinguish CD45.1 (T) from CD45.2 (G). Average contamination was low, with 0.5% detected in the CD45.1 recipient-derived fraction and 2.2% in the CD45.2 recipient-derived fractions (supplemental Table 3). Overall, cross-fraction contamination in both Jak2 and Ptprc was negligible.

Batch-corrected, variance-stabilized transformed (VST) count tables revealed the clustering of the 3 distinct hematopoietic sample sources (Figure 4B). Differential expression analysis identified 6569 significantly differentially expressed genes (DEGs) (Padj ≤ .05) between either CD45.2 donor-derived or CD45.1 recipient-derived samples and CD45.1 WT no-BMT samples. Comparing the transcriptome between CD45.2 donor-derived and CD45.1 WT no-BMT samples revealed 5521 DEGs, with 3106 downregulated (log₂ fold change [log₂FC] less than −0.25) and 2415 upregulated (log₂FC > 0.25). Comparing CD45.1 recipient-derived with CD45.1 WT no-BMT samples showed 1048 DEGs (Padj ≤ .05), with 630 downregulated (log2FC of −0.5 or less) and 418 upregulated (log2FC ≥ 0.5; Figure 4C; supplemental Table 4).

Gene set enrichment analysis using the Hay BM cell type gene sets⁴⁸ identified 8 significantly enriched sets (Padj ≤ .05) in CD45.2 donor-derived vs CD45.1 WT samples, displaying upregulation of early erythroblast-related cell types and downregulation of lymphocyte-related cell types (including B, T, and natural killer cells; Figure 4D; supplemental Table 5). Surprisingly, comparison of CD45.1 recipient-derived cells with CD45.1 WT controls revealed significant upregulation in myeloid (MONOs, eosinophils, and neutrophils) and erythroblast-related gene sets, suggesting overproduction of these cell types in recipient-derived non-JAK2MUT hematopoietic cells after BMT.

To confirm our transcriptomic results, we analyzed unsorted BM by flow cytometry (supplemental Figure 24A). In donor-derived cell populations (gated on CD45.2), we observed significantly higher levels of CD71⁺/Ter119⁺ (erythroid progenitors), CD42d⁺/CD41⁺ (megakaryocyte progenitors), CD11b⁺ (myeloid cells), and Gr-1⁺/CD66a⁺ (GRAN) cells compared with CD45.1 WT, with significantly lower frequencies of CD3⁺ (T cells) and B220⁺/CD19⁺ (B cells; Figure 4E). Intriguingly, in CD45.1 recipient-derived vs CD45.1 WT populations, we observed significant increases in both F4/80⁺/CD11b⁺ and CD71⁺/Ter119⁺ cells, along with nonsignificant increasing trends in CD11b⁺ and Gr-1⁺/CD66a⁺ cells, and nonsignificant decreasing trends in B220⁺/CD19⁺ cells, and no change in CD3⁺ cells (Figure 4E). These data suggest overproduction of monocytic and erythroblast populations in recipient-derived non-JAK2MUT–carrying hematopoietic fraction.

We assessed total hematopoietic cell output in the CD45⁺ compartment (supplemental Figure 24B). The increase in donor-derived CD42d⁺/CD41⁺ cells suggests that the frequency of megakaryocyte progenitors from donor cells may exceed those in a healthy WT animal (Figure 4F). No significant difference in CD71⁺/Ter119⁺ cells between CD45.2 donor-derived and CD45.1 WT populations suggests that engrafted donor cells have erythroid progenitor output comparable to that of WT controls (Figure 4F; supplemental Figure 24C). Strikingly, CD71⁺/Ter119⁺ cells significantly increased in the CD45.1 recipient-derived population among JAK2MUT BMT mice, suggesting that the elevated erythroid output and MPN phenotype observed in CD45.1 cells are influenced by the presence of JAK2MUT cells (Figure 4F; supplemental Figure 24C).

To deconvolute cell types from gene expression data, we used CIBERSORTx.^49,50 Consistent with gene set enrichment analysis and flow cytometry analyses, we noted loss of lymphoid (supplemental Figure 25A; supplemental Table 6) and gain of myeloid transcripts (supplemental Figure 25B; supplemental Table 6) in both JAK2MUT donor (CD45.2) and recipient (CD45.1) fractions. An increase in erythroid and megakaryocytic transcripts was observed in the donor fraction and, to a lesser extent, the recipient fraction (supplemental Figure 25C; supplemental Table 7). General decreases in hematopoietic stem and progenitor transcripts in donor as well as in recipient fractions (supplemental Figure 25D) were observed. Mapping of top unique transcripts specific to erythroid lineages indicated a portion of significantly elevated transcripts in donor-derived cells, with minor elevation in the recipient fraction (cluster 1; supplemental Figure 25E; supplemental Table 7). Although mapping of top unique transcripts specific to HSCs indicated decreased expression in the JAK2MUT donor fraction, consistent with previous reports,^44,51 some transcripts (Prom1, Hopx, Tal1, Aldh1a, and Spink2) were elevated not only in donor-derived but also in recipient-derived fractions (cluster 1; supplemental Figure 25F). In sum, these results suggest that low-frequency JAK2MUT donor–derived cells (≤5% in PB) significantly skew recipient nonmutated hematopoietic output.

JAK2V617F cells pathologically affect nonmutated erythroid and myeloid lineages

To visualize global gene expression patterns between the 5.0 × 10⁶ JAK2MUT recipient- and donor-derived samples and WT no-BMT controls, we analyzed DEGs (Padj < .05; |log₂FC| > 0.25) from the 3 different comparisons (supplemental Table 4). Using these combined DEGs, we created a scaled heat map based on their VST counts to identify 10 unique gene clusters (Figure 5A). Most clusters in the JAK2MUT BMT recipient-derived fraction displayed transitional gene expression profiles between healthy WT and JAK2MUT donor-derived states and vice versa (Figure 5A). However, 2 clusters displayed distinct patterns.

Figure 5. Presence of JAK2MUT clone profoundly affects molecular profiles of nonmutated hematopoietic fraction.

(A) Heat maps using VST hit count tables displaying all significant differentially expressed protein coding, T receptor, and immunoglobulin coding transcripts (Padj ≤ .05; |log₂FC| > 0.25) within any of the 3 comparisons (recipient derived vs no-BMT, donor derived vs no-BMT, or donor derived vs recipient derived). Unsupervised hierarchical clustering was calculated through Euclidean distance. The heat map was divided into 10 columns and 3 rows. Smaller heat maps are regenerated from VST tables filtered to only include transcripts identified with cluster 7 or 9. The color scale displays z score distribution of VST-transformed RNA-seq counts for a given transcript in an individual sample. (B) ORA using all transcripts identified within cluster 9 compared against the Hay BM cell type gene sets with Padj < .1. (C) Tree plot displaying the top 25 ORA annotations using all transcripts identified within cluster 9 compared against the GOBP cell type gene sets with Padj < .25. (D) ORA using all transcripts identified within cluster 7 compared against the murine BM cell type gene sets with Padj < .1.

Cluster 9 was uniquely upregulated in the recipient-derived fraction (Figure 5A; supplemental Table 8). Overrepresentation analysis (ORA) of transcripts in cluster 9 against the Hay BM gene sets⁴⁸ identified significant enrichment in gene sets specific to MONOs and immature neutrophils (Figure 5B; supplemental Table 9). ORA using the Gene Ontology Biological Process (GOBP) database revealed enrichment in pathways related to bone remodeling, inflammation response, macrophage activation, and cell migration (Figure 5C; supplemental Table 10). Additionally, transcripts involved in cholesterol and fatty acid metabolism (eg, Fabp4, Nr1h3, Lsr, Abcg1, and Pparg) were uniquely upregulated. This suggests presence of a host-derived MONO/macrophage population with a distinct transcriptomic signature in BMT recipients. To further characterize this population, we sorted CD45.1⁺/CD11b⁺ cells from JAK2MUT BMT recipients and WT controls for bulk RNA-seq at 52 weeks after BMT (supplemental Figure 26A). Batch-corrected VST counts showed clustering of the 2 groups (supplemental Figure 26B). Transcriptomic analysis identified 259 DEGs (Padj ≤ .05; |log₂FC| > 0.25) (supplemental Figure 26C; supplemental Table 11). ORA revealed reduced expression of sterol biosynthesis and increased expression of transcripts linked to complement activation, cholesterol metabolism and transport, leukocyte migration, and inflammation in JAK2MUT-exposed cells (supplemental Figure 26C–D; supplemental Table 12), suggesting a distinct metabolic/inflammatory signature in this myeloid fraction.

Cluster 7 showed upregulation of transcripts in the donor-derived fraction only (Figure 5A; supplemental Table 8). ORA using GOBPs on transcripts in cluster 7 revealed significant enrichment in early erythrocyte development, which included transcripts such as Gata1, Klf1, Ank1, Tfr2, and Rhag (Padj < .1; supplemental Table 10). In addition, ORA identified significance for erythroid lineage gene sets when using curated murine BM-specific gene sets⁵² (Figure 5D; supplemental Table 13). These data indicate that although both recipient- and donor-derived fractions show an increase in erythrocyte progenitors, the erythroid populations in each fraction have distinct molecular signatures, likely driven by the presence of JAK2V617F.

To assess the inflammatory signature in JAK2MUT BMT donors and recipient bystander cells, we analyzed the expression of 130 cytokines, chemokines, and growth factors (supplemental Table 14) in our RNA-seq data. As expected, the CD45.2 donor fraction showed increased transcripts of inflammatory IL-6, IL-13, Ccl3, and Cxcl3. Unexpectedly, the CD45.1 recipient fraction showed elevated inflammation-associated transcripts, including Cxcl14, Cxcl16, IL-12b, and Gdf15 (supplemental Figure 27; supplemental Table 15). These findings suggest distinct inflammatory secretome profiles linked to the heterogeneous composition of donor and recipient cell populations.

JAK2V617F⁺ cells significantly alter the stromal microenvironment

To assess how the JAK2MUT cells impact the host BM microenvironment, RNA-seq was performed on stromal cells (negative for hematopoietic lineage markers) isolated from 7- to 8-month-old BMT recipients of 5.0 × 10⁶ JAK2MUT cells (supplemental Figure 23A–B) compared with WT no-BMT controls. Successful isolation of the stromal cells was confirmed by assessing the most abundant read counts, revealing that 8 of the top 10 significantly enriched gene sets were related to bone development or the extracellular matrix (Figure 6A; supplemental Table 16). Limma batch correction was used to generate VST count tables to visualize the clustering of the 2 stromal-enriched sources via principal component analysis (Figure 6B). Similar to the hematopoietic fractions, differential expression involved a multifactor approach with “ashr” log FC shrinkage⁵³ and revealed 878 DEGs (Padj ≤ .05). Among these, 315 were upregulated (log₂FC > 0.25) and 563 downregulated (log₂FC less than −0.25) in BMT recipients compared with WT controls (Figure 6C; supplemental Table 17).

Figure 6. JAK2MUT cells negatively affect osteomesenchymal stroma.

(A) Top 10 GOBPs identified using ORA of the top 75 most abundant protein coding transcripts (most raw reads). (B) Principal component analysis of isolated stromal-enriched RNA between BMT recipients (n = 7) and WT no-BMT controls (n = 8). Ellipses were generated using multivariate t distribution with a 90% confidence interval. The dots and triangles indicate female or male sex, respectively. (C) Volcano plot displaying DEGs from isolated stromal-enriched RNA between BMT recipients and WT no-BMT controls. The horizontal line indicates Padj of .05. Vertical lines indicate log₂FCs of −0.25 and 0.25. (D) Top 15 GOBP identified using ORA when using all significant downregulated differentially expressed transcripts (Padj ≤ .05; log₂FC ≤ 0.25). (E) Top 15 GOBP identified using ORA when using all significant upregulated differentially expressed transcripts (Padj ≤ .05; log₂FC ≥ 0.25). (F) Representative sagittal micro-CT images of isolated femurs from male (top) and female (bottom) WT no-BMT controls (left) and 5.0 × 10⁶ BMT recipients (right). (G) Quantification of CT analysis on 3 mm of femur trabecular bone beginning 0.5 mm from the distal growth plate. WT no-BMT controls (n = 8; 4 males and 4 females) and 5.0 × 10⁶ BMT recipients (n = 7; 4 males and 3 females). ▴ = male, ● = female. (H) Quantification of TRAP⁺ cells per μm of bone surface area in the distal metaphysis of stained femur sections. WT no-BMT controls (n = 7; 3 males and 4 females) and 5.0 × 10⁶ BMT recipients (n = 6; 3 males and 3 females). ▴ = male, ● = female. (I) Quantification of Sp7⁺ cells per μm of bone surface area in the distal metaphysis of stained femur sections. WT no-BMT controls (n = 7; 3 males and 4 females) and 5.0 × 10⁶ BMT recipients (n = 7; 4 males and 3 females). ▴ = male, ● = female. (G-I) Statistical comparisons were performed using a 2-sided Wilcoxon rank sum test. TRAP, tartrate-resistant acid phosphatase.

Minor contamination of hematopoietic cells was detected, likely due to the presence of lineage-negative HSCs or cells that expressed low-level CD45. Additionally, presence of JAK2V617F-specific transcript reads was noted, suggesting that hematopoietic contamination originated from both recipient- and donor-derived cells (supplemental Table 18). ORA of significantly downregulated genes in recipient-derived stromal fractions identified significant pathways (Padj ≤ .05) linked to mesenchymal cell differentiation, including Wnt signaling pathways, ossification, and osteoblast differentiation (Figure 6D; supplemental Table 19). ORA of significantly upregulated transcripts revealed significant pathways (Padj ≤ .05) linked to bone remodeling and erythroid and myeloid expansion (Figure 6E; supplemental Table 20). These results suggest that the presence of JAK2MUT cells induces a retraction of osteomesenchymal cells in the BM stroma.

CT scans of isolated femurs from JAK2MUT BMT recipients revealed a trend toward reduced trabecular bone volume and complexity in unconditioned BMT recipient compared with WT mice, with a more pronounced effect in male than female animals when evaluating the distal diaphysis (Figure 6F–G; supplemental Table 21). Notably, male BMT recipients exhibited nearly double the donor chimerism observed in females (supplemental Figure 15A). Intriguingly, JAK2MUT parental animals at ~10 weeks of age also showed trabecular bone retraction and loss of trabecular complexity relative to non-JAK2MUT littermates when evaluating the distal diaphysis (supplemental Figure 28A–B; supplemental Table 21). Additionally, loss of adipocytes was observed in female BMT and JAK2MUT parental animals relative to their respective controls, whereas overall adipocyte presence was rare in males from both the parental line and in BMT recipients (supplemental Figures 8 and 17). No evidence of fibrosis was noted in the femurs of either parental or JAK2MUT BMT recipients (supplemental Figures 9 and 21). We assessed osteoclast and osteoblast status in JAK2MUT BMT mice. Tartrate-resistant acid phosphatase staining revealed a trend toward an increased number of osteoclasts per square micrometer of bone (Figure 6I; supplemental Figures 29 and 30; supplemental Table 2). Osterix (Sp7) staining showed more Sp7⁺ cells in female and fewer in male JAK2MUT BMT recipients, with opposite trends observed in control mice (Figure 6H; supplemental Figures 30 and 31; supplemental Table 2). In summary, we interpret these data as indicative that JAK2MUT cells induce an irreversible change to bone and stroma homeostasis.

Discussion

Given the high interindividual variability in JAK2V617F clonal expansion in humans, with reported BM VAFs ranging from 0.9% to 46%,⁵⁴ decades-long disease latency, and the fact that some patients never experience progression of the disease, a model replicating sustained low-VAF is needed to study early disease progression and assess therapeutic interventions. The successful transplantation of our JAK2 mutation-carrying donor line into unconditioned WT recipients demonstrated several key findings: (1) dose-dependent engraftment in an unaltered BM microenvironment, (2) disease phenotype induction from a low (<3%) frequency of circulating MPN-driving cells, and (3) the capability to track and define the malignant cell’s effects on the recipient system.

The generated model effectively replicates an MPN driven by low-frequency JAK2V617F⁺ cells over time. Given the crucial role that extrinsic factors play in MPN clonal expansion^55–57 and variable interindividual clonal expansion rates,^16,17 our model provides a platform for testing different extrinsic and intrinsic factors that promote or repress MPN progression, helping to elucidate triggers for clonal expansion. Our transplantation model introduces JAK2V617F⁺ cells in a dose-dependent manner, inducing pathology similar to the parental line but with a more PV-like phenotype, unlike the mixed PV/ET phenotype in the donors. Additionally, our method permits very low engraftment levels, replicating CHIP. Although not explored further in this study, this CHIP-like phenotype offers a valuable platform to investigate triggers of mutant clone expansion and systemic CHIP effects. We believe that this model is well-suited to study the silent, decades-long progression of MPN and its impact on the BM microenvironment and native hematopoiesis.

Our BMT model uniquely enabled investigation of 2 key aspects of MPN progression. First, we examined how JAK2V617F-mutant cells affect nonmutated hematopoietic cells. We found that JAK2MUT cells disrupt homeostasis, driving expansion of unmutated erythroid and myeloid cells and repression of lymphoid cells, mimicking MPN-like skewing. CD45.1/CD45.2 clonal tracking confirmed this, although limited CD45 expression on HSPCs prevented definitive characterization of donor vs recipient stem/progenitor populations. Studying such bystander effects in humans is challenging because of the difficulties in isolating JAK2V617F clones. Our findings suggest that MPN phenotypes can emerge even with low VAFs, driven by nonmutated but pathologically influenced bystander cells, an unexpected insight with potentially significant translational implications. Second, our BMT model allowed us to examine the disease impact on the BM stroma. BM cells, such as osteogenic and adipogenic mesenchymal cells, play pivotal roles in supporting hematopoiesis.^58–60 Recent research indicates that mutations in these cells not only drive MPN-like phenotypes^61–63 but also confer chemotherapeutic resistance or support clonal expansion in leukemia.^64–67 Therefore, using a model that tracks MPN progression and its effects on supportive nonhematopoietic stroma, without external perturbations, such as irradiation, can yield valuable, translatable information. We observed trabecular bone loss in animals that underwent transplantation, along with a transcriptomic signature indicating downregulation of BM mesenchymal pathways. Population studies show that patients with PV have an increased risk of bone fractures, suggesting underlying bone density loss, which is consistent with our observations.⁶⁸

We also noted possible sex-linked differences in BM stroma in response to the disease. Male BMT femurs showed trabecular bone loss, whereas female BMT and JAK2MUT parental femurs displayed minimal trabecular bone loss, along with loss of adipocytes in the distal region. Since both adipocytes and osteoblasts originate from a common mesenchymal stem cell (MSC) origin, MPN may dysregulate differentiation of MSCs within the BM. These observed sex-based differences warrant further exploration.

Our data indicate disease-driven remodeling of the BM stroma, reflected by shifts in population frequencies. On the basis of our findings, we propose a mechanistic model whereby diseased cells alter the BM microenvironment, disrupting MSC differentiation and expansion, ultimately leading to a loss of supportive osteomesenchymal cells. This microenvironmental change influences healthy nonmutated cells, shifting hematopoiesis toward increased myeloid and erythroid output. Over time, this imbalance and inflamed microenvironment promote disease expansion, eventually allowing mutant cells to outcompete healthy clones. By this stage, the microenvironment may be permanently altered and unable to support healthy hematopoiesis. Our data suggest that this change may be irreversible given that bone loss has not been shown to be reversed by therapeutic interventions. Although reducing VAFs theoretically allows for self-correction, prolonged clonal presence may remodel the BM stroma too profoundly to achieve recovery. This may explain why current treatments relieve symptoms but do not cure the disease. Our results also support the speculation that patients with early JAK2V617F-driven MPN stages, typically not considered for therapy, might benefit from early interventions before stromal alterations become irreversible. Recent clinical trials with ropeginterferon alfa-2b show promise in reducing JAK2V617F VAF, but its effect on the stroma remains unexplored. Our work underscores the potential critical importance of early reduction of JAK2V617F VAF.

Modeling the progression of MPN from a rare mutant cell to a widespread disease-causing clonal expansion offers opportunities to identify critical vulnerabilities for developing new treatments and early interventions. JAK2MUT likely arises decades before disease onset^18,19,69 and is influenced by clonal expansion rates, which vary widely among individuals.^16,17 This variability highlights the need to understand factors driving clonal expansion toward a clinical phenotype. Therefore, investigating interactions between the mutated cell and the healthy system, and identifying factors that disrupt this balance, may uncover new therapeutic targets, further highlighting the clinical relevance of our low-VAF MPN model.

KEYPOINTS.

• Low-frequency JAK2V617F clone drives MPN-like disease in unconditioned BMT recipients.

• Transplanted JAK2V617F clones profoundly affect nonmutated hematopoietic and stromal bystander cells.