The hematopoietic stem cell compartment of JAK2V617F-positive myeloproliferative disorders is a reflection of disease heterogeneity

Chloe James, Frederic Mazurier, Sabrina Dupont, Ronan Chaligne, Isabelle Lamrissi-Garcia, Micheline Tulliez, Eric Lippert, François-Xavier Mahon, Jean-Max Pasquet, Gabriel Etienne, François Delhommeau, Stephane Giraudier, William Vainchenker and Hubert de Verneuil


The JAK2V617F somatic point mutation has been described in patients with myeloproliferative disorders (MPDs). Despite this progress, it remains unknown how a single JAK2 mutation causes 3 different MPD phenotypes, polycythemia vera (PV), essential thrombocythemia, and primitive myelofibrosis (PMF). Using an in vivo xenotransplantation assay in nonobese diabetic-severe combined immunodeficient (NOD/SCID) mice, we tested whether disease heterogeneity was associated with quantitative or qualitative differences in the hematopoietic stem cell (HSC) compartment. We show that the HSC compartment of PV and PMF patients contains JAK2V617F-positive long-term, multipotent, and self-renewing cells. However, the proportion of JAK2V617F and JAK2 wild-type SCID repopulating cells was dramatically different in these diseases, without major modifications of the self-renewal and proliferation capacities for JAK2V617F SCID repopulating cells. These experiments provide new insights into the pathogenesis of JAK2V617F MPD and demonstrate that a JAK2 inhibitor needs to target the HSC compartment for optimal disease control in classical MPD.


X-linked clonality studies in myeloproliferative disorders (MPDs), including polycythemia vera (PV), essential thrombocythemia (ET), primitive myelofibrosis (PMF), and chronic myeloid leukemia (CML), suggested that these diseases arose from the transformation of a multipotent hematopoietic stem cell (HSC).1 The presence of a leukemic stem cell was first demonstrated in CML by the detection of the t(9;22) translocation or the BCR-ABL transcript in both myeloid and lymphoid lineages.2 Until recently, little was known about the HSC compartment of BCR-ABL–negative MPD, mainly because of the lack of molecular markers. In 2005, it was shown that PMF peripheral blood (PB) CD34+ cells contained SCID repopulating cells (SRCs) that are skewed toward myeloid differentiation.3 The same year, our group and others reported the presence of an acquired activating mutation of the protein kinase JAK2 (JAK2V617F) in the majority of patients with PV, and about half of those with ET and PMF,47 raising the question of how a unique mutation could explain 3 phenotypically different diseases. Several hypotheses are presently under investigation. One possibility is the presence of additional cooperative mutations that are unique to each of the 3 diseases. Alternatively, disease heterogeneity could be related to the level of differentiation where the oncogenic event occurs. Genotyping of common lympho-myeloid progenitors and cells from all blood lineages in patients showed that the JAK2V617F mutation occurred at least at the level of lympho-myeloid progenitors in both PV and PMF.8 Interestingly, these studies highlighted some biologic differences between PV and PMF patients. The JAK2V617F mutation was detected in B and NK cells in half of the patients with PMF, whereas it was detected in only a minority of patients with PV, suggesting that the burden of mutated HSC was different.

To date, it is still unknown whether the JAK2V617F mutation is present in multipotent long-term HSCs in MPD. In vitro experiments suggest that JAK2V617F is present in HSCs from PV patients, as it could be detected within cells with a HSC phenotype (CD34+CD38LinCD90+),9 and in long-term culture-initiating cells (LTC-IC).10 In vivo experiments in nonobese diabetic-severe combined immunodeficient (NOD/SCID) mice have recently shown that the JAK2V617F mutation could be detected in the human CD45+ cells 8 to 10 weeks after transplantation of PB CD34+ cells from PMF and PV patients with high JAK2V617F burden,11 demonstrating that these patients had some JAK2V617F SRC. These studies, however, do not address the self-renewal, multipotency, and differentiation capacities of JAK2V617F SRC nor the presence, the proportion, and the properties of JAK2 wild-type (WT) SRC from JAK2V617F MPD patients.

Given that JAK2V617F MPDs are phenotypically different diseases characterized by a common molecular event, we hypothesized that disease heterogeneity could be associated with differences in HSC biologic properties. We used an in vivo NOD/SCID xenotransplantation assay to study the properties of JAK2V617F-positive CD34+ cells from PV, PMF, and post-PV MF patients. We show that JAK2V617F SRCs are multipotent though skewed toward myeloid differentiation. In PMF and post-PV MF, JAK2V617F SRCs are predominant in the HSC compartment, at the expense of JAK2WT SRC, with no increase in self-renewal capacities. Together these results give new insights into the pathogenesis of JAK2V617F MPD and provide evidence that anti-JAK2 therapies should target the HSC compartment to eradicate the disease.


Patients samples and cell purification

The study was approved by the Local Research Ethics Committees from the Hôtel-Dieu and the Henri Mondor hospitals, and informed consent was obtained from each patient, in accordance with the Declaration of Helsinki (1975, revised in 2000). Patient diagnosis was defined according to the modified Polycythemia Vera Study Group and World Health Organization criteria for PV, and to the Italian criteria for PMF. In this study, we only enrolled patients positive for the JAK2V617F mutation in the granulocytes. Bone marrow (BM) of PV patients was aspirated at the time of diagnosis, to look for the presence of endogenous erythroid colonies. None of the patients received a treatment at the time of BM or PB sampling. Granulocytes and BM mononuclear cells were isolated as previously described.8 BM or PB CD34+ cell enrichment (92%-99%) was achieved by positive selection according to the manufacturer (EasySep; StemCell Technologies, Vancouver, BC).

NOD/SCID mice repopulation assays

CD34+ cells were intravenously injected into 7- to 10-week-old mice. Twenty-four hours before transplantation, mice were sublethally irradiated (3.5 Gy) and injected intraperitoneally with 200 μg of anti-CD122 antibody generated from the TM-b1 hybridoma cell line.12 This protocol allows a high level of engraftment after injection of 105 cord blood CD34+ cells.13 BM aspirates were obtained from the right femur at 6 and 12 weeks after transplantation, and mice were killed at week 15. Human cell engraftment was determined by the sum of human leukocytes (CD45+) and erythroid populations (CD45CD36+ and CD45CD36 Glycophorin A [GpA]+) as assessed by flow cytometry. BM cells were seeded in methylcellulose for colony assays. Fifteen weeks after transplantation, 80% of the total BM cell populations (human and mice cells) was intravenously injected into secondary NOD/SCID mice and 20% was frozen for further analysis.

Flow cytometry

Flow cytometry analysis and sorting were performed using a FACScalibur and a FACSVantage (BD Biosciences, San Jose, CA) as previously described.13 Briefly, cells were stained with antibodies specific for human antigens, including anti–CD45-PC7 (Beckman Coulter, Fullerton, CA), anti–CD19-phycoerythrin (BD Biosciences), anti–CD33-allophycocyanin (BD Biosciences), anti–CD36-phycoerythrin (BD Biosciences), and anti–GpA-allophycocyanin (BD Biosciences), antibodies. CD34+ cells were sorted using an anti–CD34-fluorescein isothiocyanate antibody (BD Biosciences).

Progenitor cells assays

A total of 1.5 × 103 human CD34+ cells or 105 BM cells from engrafted mice were seeded in methylcellulose (Stemα.I, Stem alpha SA) supplemented with SCF (50 ng/mL), interleukin-3 (20 ng/mL), with or without erythropoietin (2 UI/mL) or granulocyte-macrophage colony-stimulating factor (25 ng/mL).10 Burst forming unit-erythroid and colony forming unit-granulocyte macrophage derived colonies were counted and picked on day 14.

Assessment of long-term culture-initiating cells and B, myeloid, and NK differentiation

LTC-IC assays were performed as previously described.10 Briefly, CD34+ cells sorted from the BM of engrafted mice were plated at 20, 50, 100, 200, and 500 cells per well onto a confluent layer of MS-5 cells using standard procedures. After 5 weeks in liquid culture and 2 weeks in methylcellulose, individual colonies were picked and genotyped. Assessment of B, myeloid, and NK differentiation was performed as previously described.8

Nucleic acid extraction, genotyping, and JAK2V617F quantification

DNA from patients granulocytes and mice BM cells was extracted using the QIAGEN DNA extraction kit (QIAamp DNA micro; Qiagen, Valencia, CA). JAK2V617F quantification in granulocytes was performed as previously described.14 DNA from colonies grown in methylcellulose and B/NK/myeloid clones grown in liquid medium was prepared using a 10-mg/mL proteinase K (Invitrogen, Carlsbad, CA) buffer containing 0.2% Tween 20 (Sigma-Aldrich, St Louis, MO). Genotyping was realized using fluorescent competitive probes for quantitative real-time polymerase chain reaction (PCR) on an ABI 7500 (Applied Biosystems, Foster City, CA) as reported previously.10


CD34+ cells from JAK2V617F-positive PV, post-PV MF, and PMF patients contain SRC

To study the HSC compartment of JAK2V617F-positive PV and MF patients, CD34+ cells were collected from the BM of PV patients and from the PB of PMF and post-PV MF patients as these patients' BM is fibrotic and contains very few hematopoietic cells. Samples from PV and PMF patients were collected within 6 months after diagnosis. CD34+ cells isolated from 9 PV, 6 PMF, and 4 post-PV MF patients were transplanted into irradiated NOD/SCID mice depleted in NK activity with an anti-CD122 antibody (Table 1). A number of 0.6 to 10 × 105 CD34+cells were transplanted per mouse. Because of the limited number of CD34+ cells available from each patient, most experiments were performed with a single recipient mouse per human donor. When several mice received transplants with a sample from the same patient, the level and kinetics of human cell engraftment were comparable between recipients.

Table 1

PV, PMF, and post-PV MF CD34+ cells contain JAK2V617F-positive SCID repopulating cells

To assess whether CD34+ cells from PV, PMF, and post-PV MF patients contain SRCs, we measured the proportion of human cells in the mouse BM 15 weeks after transplantation. Human cell engraftment (> 0.5%) was observed in 23 of 27 mice, corresponding to 15 of 19 patients (Table 1). The level of engraftment varied considerably between each sample and did not correlate with the number of CD34+ cells injected or with the JAK2V617F burden in granulocytes as assessed in the patients' PB. Our results contrast with the data of Ishii et al showing that only CD34+ cells from PV with a high JAK2V617F burden contained NOD/SCID repopulating cells11 and show that PV, PMF, and post-PV MF CD34+ cells contain some SRCs.

JAK2V617F SRC are present in PV, PMF, and post-PV MF, but in different proportions

To assess the presence of JAK2V617F and JAK2WT SRC in MPD samples, mouse BM was collected at week 15 after transplantation and tested in methylcellulose assays. Under the experimental conditions used, almost all the colonies that grew were of human origin, whatever the level of human engraftment. Colonies were counted on day 14, picked, and genotyped using a real-time PCR assay specific for WT and V617F human JAK2. In mice reconstituted with PV CD34+ cells, 149 of 172 (87%) progenitors were JAK2WT, the remaining being heterozygous for the JAK2V617F mutation (Table 1). In contrast, in mice engrafted with PMF CD34+ cells, the majority of the colonies (248 of 259, 96%) were JAK2V617F. In 2 mice reconstituted with CD34+ cells from 2 PMF patients who carried homozygous cells, homozygous colonies were detected (PMF2, PMF4), suggesting that mitotic recombination is an early event that occurred at the level of an SRC (Table 2). For mice where no colony developed in the methylcellulose assay, DNA was extracted from total BM. Real-time PCR for human-specific sequences revealed that almost all human cells were JAK2V617F-positive in mice receiving transplants of PMF CD34+ cells because the JAK2V617F/total JAK2 ratio was 46% to 76% (mean, 58% ± 11%), thus confirming the observations made on individual colonies (Table 1). Moreover, all human cells present in 2 mice reconstituted with post-PV MF CD34+ cells (PPV-MF2 and 3) were homozygous for the JAK2V617F mutation. Together these data show that JAK2V617F SRC have a low contribution to the hematopoietic reconstitution in mice receiving transplants of PV CD34+ cells, whereas human hematopoiesis is mostly derived from JAK2V617F SRCs in mice injected with PMF and post-PV PMF CD34+ cells.

Table 2

JAK2V617F allelic frequency in patient samples and 6, 12, and 15 weeks after transplantation of CD34+ cells

A property of HSCs is their long-term survival. To determine whether PV and PMF JAK2V617F SRCs were capable of long-term survival, CD45+CD34+ cells from the mouse BM were sorted at week 15 after transplantation and assessed for the presence of LTC-ICs (Figure 1). Because of the number of cells required, we performed these experiments with the BM of 5 mice where the level of human engraftment was more than 20%. Genotyping of the 431 LTC-IC–derived colonies showed the presence of JAK2V617F-positive SRCs capable to give rise to LTC-IC, which proves that JAK2V617F SRC have long-term reconstitution activity.

Figure 1

The JAK2V617F mutation is present in LT-SRCs. CD34+CD45+ cells present in the mouse BM were sorted and grown in LTC-IC medium. The LTC-IC frequencies were determined (y-axis), and LTC-IC-derived erythroid and granulocytic colonies were genotyped. Each histogram represents a mouse reconstituted with PV or PMF sample. On each histogram is indicated the proportion and the number of wild-type (▭), JAK2V617F heterozygous (Embedded Image), and homozygous (Embedded Image) colonies.

Altogether, these data show that JAK2V617F long-term SRCs are present in PV, PMF, and post-PV MF. In addition, they also highlight differences in the HSC compartment between these disorders, with most SRC being JAK2V617F-positive in MF whereas only a minority is mutated in PV.

JAK2V617F-positive SRCs are multipotent, irrespective of the disease

Another important property of HSCs is their ability to give rise to multilineage clones comprising differentiating and maturing blood cells.15 To determine whether disease heterogeneity could be related to distinct differentiation properties of HSCs, we analyzed the differentiation potential of SRCs in the 3 diseases. Because PV HSCs were reported to be skewed toward erythroid differentiation,9 we analyzed the proportion of human erythroid cells (Glycophorin A+) in NOD/SCID mice at 15 weeks after transplantation. Surprisingly, we detected only a small proportion of human erythroid cells in a few number of mice, irrespective of the disease origin of CD34+ cells (Table 1). Further analysis of lineage distribution of the human CD45+ cells present in the BM of mice reconstituted with PV CD34+ cells (Table 1) showed a predominant B-lymphoid population (Figure 2). This phenotype was not surprising because the majority of SRCs in PV are JAK2WT. Conversely, myeloid cells represented the predominant population in the BM of mice receiving transplants of PMF and post-PV MF CD34+ cells, as previously described by Xu et al.3 Because most PMF and post-PV MF SRCs are JAK2V617F-positive, these data demonstrate that PMF and post-PV MF JAK2V617F SRCs are skewed toward myeloid differentiation and can hardly support B lymphopoiesis. Because of the weak contribution of JAK2V617F-positive SRCs in the hematopoietic reconstitution of mice engrafted with PV CD34+ cells, it was not possible to assess whether PV JAK2V617F SRCs are also skewed toward the myeloid lineages.

Figure 2

The SRCs from patients with PV and JAK2V617F-positive myelofibrosis have a different differentiation program. (A) The percentages of myeloid (▭) and B-lymphoid (Embedded Image) cells within the human graft of PV (n = 8), PMF (n = 4), and PPV-MF3 reconstituted mice. (B) Flow cytometric analysis of the BM of 3 representative mice, receiving a transplant of PV, PMF, or PPV-MF CD34+ cells. (Top panel) SSC versus CD45 analysis showing human cells (CD45+). (Bottom panel) CD45+ cells analyzed for CD33 and CD19 expression, showing human myeloid (CD45+CD33+) and B-lymphoid (CD45+CD19+) cells.

Having shown that PMF and post-PV MF JAK2V617F SRCs are skewed toward myeloid differentiation, we wanted to establish whether these cells were still capable of undergoing B-cell differentiation. We thus sorted mature B and myeloid cells that had differentiated into NOD/SCID mice at 15 weeks after transplantation and genotyped them. These experiments were performed in 6 mice reconstituted with PV CD34+ cells, 3 with PMF and one with post-PV MF cells. The JAK2V617F mutation was detected in the CD33+ and the CD19+ cell populations of 5 mice (Figure 3A), demonstrating that JAK2V617F SRCs are multipotent. To further confirm this result, we tested the ability of single JAK2V617F CD34+CD38 cells from patients to differentiate into B, myeloid, and NK cells in vitro. These experiments were performed with cells from 6 patients (PV8, 9, PMF1, 2, 4, PPV-MF3). CD34+CD38 cells were grown at one cell per well on the murine stromal MS5 cell line with a combination of 7 cytokines allowing the detection of single clones with B, myeloid, and NK differentiation potential. Each proliferating clone was subjected to immunophenotypic and genotypic analyses (Figure 3B). We detected the presence of JAK2V617F lympho-myeloid progenitors (MNK and BMNK) in all patients. Altogether, our results demonstrate that JAK2V617F SRCs from PV, PMF, and post-PV SRC are multipotent, although a clear skewing toward myeloid differentiation is observed for MF SRCs.

Figure 3

JAK2V617F SRCs are capable to give rise to multilineage differentiated cells. (A) Human B-lymphoid (CD45+CD19+) and myeloid (CD45+CD33+) cells engrafted in 10 NOD/SCID mice 15 weeks after transplantation were sorted and genotyped for JAK2V617F. The percentage of JAK2V617F over total JAK2 is reported. (B) Single CD34+CD38 cells from 2 PV, 3 PMF, and one post-PV MF were cultured in B/NK/myeloid differentiation conditions. Clones with lympho-myeloid potentialities (MNK and BMNK) were genotyped. The histograms represent the numbers of lympho-myeloid clones from each patient. ▭ represents wild-type JAK2; Embedded Imagebars, heterozygous JAK2V617F; Embedded Image, homozygous JAK2V617F.

JAK2V617F SRCs have neither proliferative nor self-renewal advantages in the NOD/SCID mouse model

Two putative mechanisms might explain how a molecular event occurring in an HSC can give rise to clonal mature hematopoiesis. First, only a few HSCs are malignant, but because of an acquired proliferative advantage either at the HSC level or downstream in the hematopoietic differentiation, the majority of mature cells belong to the malignant clone. Second, the molecular event modifies the self-renewal properties of malignant HSCs leading to the progressive disappearance of normal HSCs. We therefore examined whether JAK2V617F HSCs had increased proliferative or self-renewal properties compared with JAK2WT HSCs and whether differences between PV and PMF could be observed.

Because an acquired proliferative advantage in malignant HSCs would result in a progressive increase of human CD45+ cells in the mouse BM, we analyzed the kinetics of human cell engraftment by sequential BM aspiration of the femur (RF) 6, 12, and 15 weeks after transplantation (Table 1). In the majority of the cases (13 of 19 samples), the percentage of human cells progressively decreased between 6 and 15 weeks, irrespective of the pathology (Figure 4). Interestingly, a group of 5 mice receiving transplants of 4 different samples (PV8, 9, PMF1, 2) displayed a progressive increase in the percentage of human cells, suggesting that a small fraction of patients had SRCs with increased proliferative capacity. Nevertheless, the human cells present in these mice were not JAK2V617F-positive in PV8 and PV9, ruling out the hypothesis that JAK2V617F SRCs would be responsible for the observed proliferation. To further determine whether JAK2V617F SRCs from PV and PMF patients had acquired a proliferative advantage over JAK2WT SRCs, we analyzed the proportion of JAK2WT, JAK2V617F heterozygous and homozygous human progenitors that were present in NOD/SCID mice engrafted with PV or PMF CD34+ cells at 6, 12, and 15 weeks after transplantation. In the BM of mice receiving transplants of PV CD34+ cells, the proportion of JAK2V617F did not increase over time (Table 2), demonstrating that PV JAK2V617F SRCs do not proliferate more than JAK2WT SRCs. In PMF, the proportion of JAK2V617F colonies was already very high before transplantation, rendering it difficult to discern any posttransplantation increase. However, in all cases, JAK2V617F allelic frequency remained at least stable. Interestingly, we could show the persistence of a few JAK2WT colonies (Table 2). This result shows that JAK2V617F SRCs are predominant in PMF and have no proliferative advantage over JAK2WT SRCs. Collectively, these data argue against the hypothesis that the clonal process in MPD is the result of an acquired proliferative advantage of HSC. In addition, they did not reveal any difference in terms of proliferation properties between PV and MF SRCs.

Figure 4

Kinetics of engraftment suggests that leukemic SRCs have not acquired proliferative advantage, except for a subset of patients. Analysis of the kinetics of human cell engraftment by sequential BM aspiration of the right femur (RF) 6, 12, and 15 weeks after transplantation.

To test whether the clonal process was related to an increase in self-renewal capability of JAK2V617F HSCs, we studied the ability of CD34+ cells of PV, PMF, and post–PV-PMF patients to reconstitute NOD/SCID mice in a secondary transplantation assay. In all but one mouse, no human cells could be detected in secondary transplantations (data not shown). This result was not surprising because in most cases the degree of human cell engraftment was very low in the primary mice. Nevertheless, the percentage of human cells was 43.2% (± 13.2%) in 4 primary mice at 15 weeks after transplantation, and it was expected that the human cells would have implanted into the secondary hosts. The only secondary mouse that was positive for the presence of human cells was engrafted with BM cells from a primary mouse receiving a transplant of CD34+ cells from patient PMF2. Despite the presence of 78% human cells in the primary mouse, only 1% of human CD45+ cells were detected 6 weeks after transplantation in the secondary host. The high engraftment level in the primary mouse and the very low level seen in the secondary host suggest that the human cells present in the primary mouse were mostly differentiated cells. Genotyping of human cells at week 6 in the secondary host showed the presence of JAK2V617F-positive cells (60% of JAK2V617F/total JAK2). Together, the data show that JAK2V617F SRC, either in PV or PMF patients, do not have a dramatic self-renewal advantage over JAK2WT SRCs in a xenogenic assay.


In this study, we have used an in vivo NOD/SCID xenotransplantation assay to investigate the properties of primary human HSCs from JAK2V617F-positive PV, PMF, and post-PV MF patients. We found that JAK2V617F long-term and multipotent SRCs were present in these 3 diseases but that the proportion of JAK2V617F compared with JAK2WT SRCs was markedly different between PV and PMF. Furthermore, we did not find major difference in the proliferation and self-renewal properties of JAK2V617F SRCs between PV, PMF, and post-PV MF. Our data thus show that disease heterogeneity is clearly correlated with quantitative differences in the burden of JAK2V617F SRCs compared with JAK2WT SRCs, and the reasons for these differences still need to be investigated.

Our data demonstrate that JAK2V617F SRCs are multipotent, a finding consistent with a previous work showing that JAK2V617F cells are detected in all blood lineages and common lympho-myeloid progenitors8 in PV and PMF patients. Because the JAK2V617F mutation was detected in B and NK cells in half the patients with PMF but in a minority of those with PV, it was suggested that the burden of mutated HSCs was different.8 The present findings demonstrate that the HSC compartment is indeed different in PV and MF, as suggested in a recent report with LTC-IC experiments.10 In this work, we studied the PB CD34+ cells for PMF and post-PV MF patients because their BM is fibrotic and devoid of CD34+ cells. For PV patients, CD34+ cells were collected from the BM. It is thus possible that the observations reported herein are related to the source of CD34+ cells but are anyway reflecting the state of the HSC compartment in the patients.

Our data highlight similarities between PV and chronic phase-CML (CP-CML). In CP-CML, the majority of SRCs are BCR-ABL–negative.1618 BCR-ABL–positive HSCs are highly quiescent,19 with no self-renewal advantage20 and skewed toward myeloid differentiation.21 Besides, xenograft assays have always failed to demonstrate any proliferative advantage of BCR-ABL–positive SRCs over normal SRCs, consistent with our observations with JAK2V617F SRCs. We cannot definitively conclude that JAK2V617F SRCs do not have any proliferating advantage because the model of NOD/SCID mice is perhaps not appropriate to test this property in JAK2V617F SRCs. Indeed, it has been demonstrated that JAK2V617F cells are still responsive to extrinsic regulators, and it is possible that the murine microenvironment does not produce certain cytokine or chemokine that would be necessary for the amplification of JAK2V617F HSCs. Difference between human and mouse microenvironmental factors could also explain why we did not observe any erythroid skewing of PV SRCs, in contrast to the in vitro observations of Jamieson et al.9 It is also possible that JAK2V617F SRCs have such a weak proliferative advantage over normal SRC that it could not be assessed in our model because of the low efficiency of human cell engraftment. The use of the recently developed NOD/SCID/gamma-c deficient mice, which are certainly more tolerant for human cells than NOD/SCID mice treated with an anti-CD122 antibody, would be of interest to address this question in the future.

PV is a well-characterized MPD where polycythemia is associated with the JAK2V617F mutation found in almost 95% of the patients. Nevertheless, this group of patients is not homogeneous, as some patients will need more or less hydroxyurea and some will evolve to myelofibrosis or transform into acute leukemia. This work shows some marked differences within the HSC compartment between PV patients. The mice engrafted with CD34+ cells from patients PV8 and PV9 showed an increase with time in human chimerism with a predominance of myeloid cells, an observation opposite to what is seen in the majority of mice engrafted with PV cells. Very surprisingly, we found that the human cells that proliferated in the mice were mostly JAK2WT, suggesting that the HSCs from these 2 patients may have another genetic event that would modify the biology of primitive progenitors and that would precede the occurrence of the JAK2V617F mutation. This is in accordance with multiple studies showing that the JAK2V617F mutation is clearly not the sole event responsible for PV.22,23 Future challenges will be to identify this molecular event and to determine whether a subtype of patients is predisposed to acute leukemia progression, especially to JAK2V617F negative leukemia.

Our data showing that only a minority of JAK2V617F SRCs is detected in PV, whereas all SRCs are mutated in post-PV, MF, and PMF, despite any detectable increase in their self-renewal properties, suggest that the development of myelofibrosis is not only associated with a constitutive mobilization of CD34+ cells but also with a selective disappearance of normal SRCs. Selective apoptosis of normal HSCs was reported in a murine model of CML.24 BCR-ABL–positive cells secrete a glycoprotein, the lipocalin 24p3, that induces apoptosis of normal hematopoietic cells, including HSCs. BCR-ABL–positive cells down-regulate the lipocalin cell surface receptor and become resistant to lipocalin. By analogy, the fibrogenic cytokine TGF-β1 is likely to play a key role in JAK2V617F diseases. Indeed, it is overexpressed in PMF patients,25 it plays a major role in the development of myelofibrosis,26 and it is known to inhibit normal HSCs,27 whereas PMF CD34+ cells are relatively insensitive to its negative regulation through the decreased expression of type II TGF-β1 receptor.28 Investigation of the molecules secreted by cells expressing JAK2V617F may provide further insights in our understanding of the pathogenesis of primitive and secondary myelofibrosis.

In conclusion, we propose the following model (Figure 5). In PV, JAK2V617F targets a multipotent LT-HSC without increasing their self-renewal or proliferation properties, explaining why the majority of SRCs are JAK2WT. JAK2V617F LT-HSCs then differentiate and, when committed progenitors express homodimeric cytokine receptors such as the erythropoietin, thrombopoietin, and granulocyte colony-stimulating factor receptors, JAK2V617F cells proliferate more than normal cells.29 Indeed, it was shown that JAK2V617F-positive cells are hypersensitive to cytokine signaling through homodimeric receptors.30 The hematopoietic amplification characteristic of MPD would only affect the compartments of mature progenitors to mature differentiated cells, as already demonstrated in PV.10 We hypothesize that the malignant JAK2V617F cells (committed cells) would secrete some cytokines or chemokines, such as TGF-β,26 that would induce marrow fibrosis and be toxic for normal HSCs. As JAK2V617F HSCs would be more resistant to apoptosis because of the hyperexpression of bcl-xl, for example, JAK2WT HSCs would become scarce and JAK2V617F HSCs more numerous after many years of disease evolution. In PMF, we hypothesize that the JAK2V617F mutation would also target a multipotent HSC without modifying its self-renewal and proliferation properties, but in this case, another molecular event would be responsible for a massive production of the abnormal cytokine or chemokine, leading to a rapid onset of myelofibrosis and destruction of the normal HSC compartment.

Figure 5

Putative model of evolution from PV toward post-PV myelofibrosis, through modification of the HSC compartment. The left diagram shows a model of hematopoiesis from the HSCs to mature cells in PV. JAK2V617F cells are in gray, and JAK2WT cells in white. The properties of JAK2V617F HSCs are not modified, compared with JAK2 WT HSC, either in term of self-renewal or proliferation. However, a major amplification of terminal myeloid differentiation is observed, from the stage of differentiation when progenitors express homodimeric type I cytokine receptors. We hypothesize that JAK2V617F cells would produce TGF-β1, responsible for the progressive development of myelofibrosis, and another cytokine or chemokine (X-factor) responsible for the progressive destruction of normal hematopoiesis, that could also be TGF-β1. As JAK2WT HSCs would express the receptors of X-factor, they would be sensitive to X-factor–negative regulatory signals and therefore progressively destructed. On the contrary, JAK2V617F HSCs would have a decreased expression of X-factor receptor or signaling and would express antiapoptotic proteins, such as bcl-xl, being therefore insensitive to apoptosis. The right panel shows the stage of post-PV myelofibrosis, when the increasing and persistent production of TGF-β1 would induce myelofibrosis and when JAK2WT HSCs have been destructed following X-factor stimulation. The only remaining HSCs are thus JAK2V617F, without any increase in self-renewing properties.

Having established that the JAK2V617F mutation occurs in an HSC, it is now clear that a JAK2 inhibitor must target HSC if it aims to be curative. The problem is similar to that seen in CP-CML, where most tyrosine kinase inhibitors, despite being very effective on committed cells, are almost ineffective on BCR-ABL HSCs.31 Our results demonstrate that normal SRCs are very scarce, if present at all, in PMF and post-PV MF patients. If a JAK2 inhibitor suppresses JAK2V617F HSCs, it remains to be seen whether rare normal HSCs will be able to reconstitute normal hematopoiesis in these patients.


Contribution: C.J., W.V., and H.d.V. designed research, analyzed data, and wrote the paper; C.J., I.L.-G., S.D., R.C., J.-M.P., and M.T. performed cellular experiments; C.J., S.D., F.D., E.L., and F.-X.M. performed genotyping experiments; C.J. and F.M. performed mice experiments; and G.E., F.D., and S.G. contributed to the recruitment of patients.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Chloe James, Inserm U876, Université Bordeaux 2, 146 rue Léo Saignat, 33076 Bordeaux Cedex, France; e-mail: chloe.james{at}


The authors thank N. Casadevall, J. F. Viallard, F. E. Nicolini, O. Fitoussi, P. Duffau, and A. M. Ferrer for the recruitment of patients, J. Wang and F. Wendling for critical reading of the manuscript, A. Masse, A. C. Pons, and C. Pain for excellent technical assistance, and B. Trotat for the irradiation of mice.

This work was supported by grants from Inserm, La Ligue Nationale contre le Cancer (équipe labellisée 2007), and the INCa (projets libres 2006). Genotyping experiments were performed at the Genotyping and Sequencing facility of Bordeaux (grants from the Conseil Régional d'Aquitaine no. 20030304002FA and 20040305003FA and from the European Union, FEDER no. 2003227).


  • *W.V. and H.d.V. contributed equally to this work.

  • The publication costs of this article were defrayed in part by page charge payment. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

  • Submitted February 5, 2008.
  • Accepted July 2, 2008.


View Abstract