A conditional inducible JAK2V617F transgenic mouse model reveals myeloproliferative disease that is reversible upon switching off transgene expression

Aberrant activation of the JAK/STAT pathway is thought to be the critical event in the pathogenesis of the chronic myeloproliferative neoplasms, polycythemia vera, essential thrombocythemia and primary myelofibrosis. The most frequent genetic alteration in these pathologies is the activating JAK2V617F mutation, and expression of the mutant gene in mouse models was shown to cause a phenotype resembling the human diseases. Given the body of genetic evidence, it has come as a sobering finding that JAK inhibitor therapy only modestly suppresses the JAK2V617F allele burden, despite showing clear benefits in terms of reducing splenomegaly and constitutional symptoms in patients. To gain a better understanding if JAK2V617F is required for maintenance of myeloproliferative disease once it has evolved, we generated a conditional inducible transgenic JAK2V617F mouse model using the SCL-tTA-2S tet-off system. Our model corroborates that expression of JAK2V617F in hematopoietic stem and progenitor cells recapitulates key hallmarks of human myeloproliferative neoplasms, and exhibits gender differences in disease manifestation. The disease was found to be transplantable, and importantly, reversible when transgenic JAK2V617F expression was switched off. Our results indicate that mutant JAK2V617F-specific inhibitors should result in profound disease modification by disabling the myeloproliferative clone bearing mutant JAK2.


Introduction
The discovery of the somatic activating JAK2 V617F mutation in the majority of patients with the chronic myeloproliferative neoplasms (MPNs) polycythemia vera (PV), essential thrombocythemia and primary myelofibrosis marked a breakthrough in the understanding of the underlying molecular mechanism of these diseases [1][2][3][4]. Subsequently, gain-of-function mutations in the thrombopoietin receptor MPL [5] or CBL [6][7][8], loss-of-function mutations in SH2B3 [9,10], and most recently frameshift mutations in CALR [11,12] have been discovered in JAK2 V617F mutation-negative MPN patients. From a mechanistic standpoint, these mutations have a dysregulated, constitutively activated JAK/STAT pathway in common [13]. Accordingly, transplantation of lethally irradiated mice with murine bone marrow cells transduced with a retrovirus expressing either JAK2 V617F or MPL W515L was shown to result in pathological features that closely resemble human PV or myelofibrosis, respectively [14][15][16]. A phenotype mimicking human essential thrombocythaemia and PV is also obtained upon transgenic expression or knock-in of JAK2 V617F in hematopoietic cells of mice [17][18][19][20][21][22], and it was demonstrated that expression of JAK2 V617F in a single hematopoietic stem cell is sufficient to give rise to MPNs [23]. Disease hallmarks observed in the JAK2 V617F mouse models include elevation of hemoglobin and hematocrit, leukocytosis, thrombocytosis, megakaryocyte hyperplasia, extramedullary hematopoiesis resulting in splenomegaly, and increased reticulin fibers in the bone marrow of some of the models.
The identification of the JAK2 V617F mutation has spurred the discovery and development of JAK inhibitors for the treatment of MPNs [24], and the JAK1/JAK2 inhibitor ruxolitinib received regulatory approval for the treatment of myelofibrosis, and for PV patients who are resistant to or intolerant of hydroxyurea [25][26][27]. In the clinical setting, ruxolitinib and other JAK inhibitors have shown remarkable activity in terms of suppressing splenomegaly, constitutional symptoms, and aberrant blood counts in MPN patients [24-26, 28, 29]. Similarly, treatment of mouse MPN models with JAK inhibitors, including ruxolitinib, was shown to strongly decrease splenomegaly, and to rapidly normalize red blood cell and neutrophil parameters, consistent with inhibition of constitutive STAT5 phosphorylation in the bone marrow and spleens of treated animals [19,[30][31][32][33]. However, it was soon realized that in preclinical models treatment with JAK inhibitors did not substantially affect JAK2 V617F mutant allelic burden or eradicated MPN-initiating clones [19,30,32,34]. Similarly, in MPN patients the average reduction of mutant JAK2 allele burden during treatment with JAK inhibitors was modest, although a subset of patients achieved partial or complete molecular responses [26,29,35,36]. The reason for the limited effect of current JAK inhibitors on the mutant allele burden in MPNs has been subject to debate and is not well understood [37].
To gain more insights into the myeloproliferative disease hallmarks that are dependent on JAK2 V617F once the neoplasm has manifested, we generated a conditional inducible transgenic JAK2 V617F mouse model. In our model, expression of JAK2 V617F is under the control of a tetracycline-responsive promoter element (TRE), and transgene expression was directed to hematopoietic stem and progenitor cells using a tet-off system in which the tetracycline transactivator (tTA) is under the control of the stem cell leukemia gene (SCL) regulatory elements (SCL-tTA-2S) [38]. Upon JAK2 V617F expression in double-transgenic mice, MPN-like disease closely resembling key aspects of human PV pathology started emerging after a latency period of several months. While the extent of white blood cell (WBC) count elevation as compared to littermate controls was comparable between female and male double-transgenic mice, the female mice displayed a higher elevation of reticulocyte count and more pronounced splenomegaly, whereas a higher platelet count was preferentially seen in male mice. The disease was found to be transplantable, and importantly, was reversible when transgenic JAK2 V617F expression was switched off. Taken together, our results corroborate that expression of JAK2 V617F in hematopoietic stem and progenitor cells results in a disease phenotype that faithfully mimics the hallmarks of human MPNs and demonstrate that manifested disease is reversible when the mutant JAK2 allele is inactivated.

Results
Generation of a conditional inducible mouse model that allows control of JAK2 V617F expression in a doxycycline regulatable manner A number of preclinical studies have assessed the pathology associated with the expression of JAK2 V617F in mouse models. These models included bone marrow transplantation experiments where JAK2 V617F expression was driven by retroviral vectors [14,16,39,40], or transgenic mice where JAK2 V617F expression was either under the control of a tissue-specific or the endogenous promoter, respectively [19][20][21]. These studies compellingly corroborated an initiating role for JAK2 V617F in the development of myeloproliferative disorders. However, these models did not enable assessing JAK2 V617F in MPN maintenance, and blocking JAK2 function was mostly interrogated with systemically acting JAK inhibitors. To overcome these shortcomings, we generated a regulatable murine model, where JAK2 V617F transgene expression can be switched on and off in hematopoietic cells (S1A Fig). To this end, a construct was made in which a tetracycline operator drives inducible expression of the human JAK2 V617F cDNA and an IRES-EGFP (S1B Fig). This vector was injected into the pronucleus of C57Bl/6 oocytes, and 7 out of 48 transgenic founder lines were genotyped positively and produced progeny. The 7 founders were bred until the F2 generation.
In MPN patients, JAK2 V617F is expressed under the endogenous promoter, but uniparental disomy in the MPN clone can select for 2 mutant JAK2 V617F genes and loss of the remaining wild type copy [3,41]. In order to avoid artificially high expression of the mutant transgene, we determined the relative human JAK2 copy number in transgenic mice from the 7 founder lines (S1 Table). Founder 29 displayed the lowest copy number of 1.8 on average, suggesting these mice carry two copies of the transgene. We used this mouse line for all future experiments. Primary mouse fibroblasts from this line were cultured in vitro and transfected with a CMV-itTA plasmid to evaluate induction of JAK2 V617F expression (S2A Fig). A strong increase of transgene expression was found after CMV-itTA transfection, which was blunted when transfected fibroblasts were treated with doxycycline (dox) (S2B Fig), indicating that the expression of JAK2 V617F was switchable. By Western blotting, JAK2 protein levels were slightly increased in the itTA transfected samples in absence of dox, and went back to endogenous levels when dox was added (S2C Fig). Although the GFP gene was also expressed (S2B Fig), there was no discernible GFP signal using Western blotting (S2C Fig). Sequencing of the transgene from fibroblast genomic DNA revealed construct integrity. When the JAK2V617F-IRES-EGFP cDNA construct was transiently co-transfected with itTA into NIH3T3 fibroblasts we readily detected high levels of JAK2 and GFP mRNA expression (S2B Fig), indicating that, unfortunately, in the transgenic context of our model IRES-directed GFP expression may be hampered by the relatively large preceding cDNA [42].
Next, the founder 29 responder line was crossed with the SCL-tTA-S2 effector line [38] (S3A and S3B Fig), known to permit reversible on/off switching of transgenes in hematopoietic stem cells and myeloid progenitors [43]. The tTA-2S tetracycline trans-activator under the control of SCL regulatory elements (S3A Fig) allows transgene expression in erythrocytes, megakaryocytes, granulocytes and Kit + Linhematopoietic cells [38]. In a pilot experiment, we assessed transgene expression and phenotype in vivo by following 4 SCL-tTA; JAK2 V617F double-transgenic and 4 littermate control mice (2 wild type, 2 JAK2 V617F littermate controls).
Dox containing food was given during gestation and the first weeks after birth to keep the transgene switched off and avoid potential JAK2 V617F -induced developmental lethality [19]. Mice in the experimental cohort were switched to normal food at 7-9 weeks of age to enable transgene induction. Encouragingly, after an observation period of 138 days off dox, a slight increase in platelet count was discernible in double-transgenic mice as compared with controls (S3C Fig

Conditional inducible expression of JAK2 V617F in mice recapitulates key hallmarks of human MPNs
To corroborate these results we crossed further animals from the responder line with SCL-tTA-S2 effector mice, and closely followed their development after removing dox in the chow of young adults. After about 6 months post-dox removal body weight gain of SCL-tTA; JAK2 V617F double-transgenic mice, but not of control littermates, plateaued (Fig 1A). Highresolution micro-ultrasound imaging revealed marked splenomegaly in double-transgenic mice (Fig 1B), along with elevated reticulocyte, platelet and WBC counts ( Fig 1C). Interestingly, there was a clear gender difference in the manifestation of MPN phenotypes: While double-transgenic females had more pronounced splenomegaly and higher reticulocyte count, males had a higher platelet count ( Fig 1C). Thus, double-transgenic female mice exhibited disease hallmarks of PV, while male mice recapitulated an ET-like phenotype.
To further characterize the MPN phenotype, mice were sacrificed after 294 days for histopathological assessment of spleen and bone marrow. Spleens of double-transgenic mice displayed marked erythroid expansion (S4A Fig), and the bone marrow was hyperproliferative (Fig 2A). Furthermore, silver impregnation staining of bone marrow sections showed reticulin fiber deposits in double-transgenic mice but not the littermate controls, which may correspond to a pre-fibrotic stage of human MPNs (Fig 2B and S4B Fig). Immunohistochemistry for phosphorylated STAT5 on spleen sections revealed a strong staining in samples from double-transgenic mice ( Fig 2C and S4C Fig), consistent with expression of the JAK2 V617F transgene, as detected by chromogenic in situ hybridization (CISH) (Fig 2D). Levels of phosphorylated STAT5 were also substantially increased in bone marrow sections of double-transgenic mice, as compared to littermate controls (S4D Fig).
Taken together, these results corroborate that transgenic JAK2 V617F expression in the hematopoietic compartment of mice is sufficient to give rise to a phenotype resembling human MPNs.

The MPN-like disease is reversible upon switching off JAK2 V617F expression in vivo
Having established that SCL-tTA-S2; JAK2 V617F double-transgenic mice develop an MPN-like disease several months after induction of JAK2 V617F expression raises the key question whether the disease would be reversible upon switching off transgene expression. On one hand, the cooperation of the JAK2 V617F -bearing MPN clone with its surrounding (JAK2 wild type) stroma sets off a progressive remodeling of the hematopoietic microenvironment in the bone marrow and spleen [44], and it is conceivable that this may lead to gradually reduced dependency on the JAK2 V617F initiating event. This might explain the clinical activity observed with current JAK inhibitors in MPN patients, reducing splenomegaly and aberrant blood counts, while having modest effects on JAK2 V617F allele burden and on reversing bone marrow fibrosis [19,30,32,34]. On the other hand, it may be possible to not only block the vicious cycle, but also to reverse disease hallmarks, including MPN clonal dominance, by specifically thwarting the MPN clone.
To interrogate these two different possibilities, SCL-tTA-S2;JAK2 V617F mice that had been fed normal chow for nearly 8 months, as well as their littermate controls, were administered dox-containing food for a period of 72 days. As additional controls, 2 animals of each genotype continued to receive normal chow. Prior to switching mice to dox-containing food, the double-mutant animals exhibited significantly elevated reticulocyte, WBC and platelet counts, as well as increased spleen volume compared to their littermate controls (S5A Fig  To confirm and expand on these results, mice from an independent cross were fed normal chow until the SCL-tTA-S2; JAK2 V617F mice developed MPN-like disease. Then, the doubletransgenic mice were randomized into 3 groups: in one arm, mice were switched to dox containing chow to turn off JAK2 V617F expression, while in the other 2 study cohorts, mice either received the JAK1/JAK2 inhibitor ruxolitinib at 60 mg/kg twice daily per os, or drug vehicle only. In a fourth additional arm, control littermates were also given vehicle. After 26 days of treatment, mice were sacrificed for analysis. WBC counts ( Fig 3A) and spleen volumes ( Fig  3B) of SCL-tTA-S2;JAK2 V617F mice were comparably reduced in the dox chow and the ruxolitinib arms. However, reduction of reticulocyte ( Fig 3C) and platelet ( Fig 3D) counts was superior in the dox chow arm as compared to the ruxolitinib treated group. In fact, platelet count in the latter group was even higher than in the double-transgenic vehicle control group ( Fig  3D), an effect which has been observed previously in other mouse MPN models treated with JAK inhibitors [32,45]. Levels of STAT5 phosphorylation in bone marrow and spleen of double-transgenic mice were markedly reduced in the dox chow and ruxolitinib arms (Fig 3E and  3F). Histopathological analysis of bone marrow sections showed that either turning off JAK2 V617F expression or ruxolitinib treatment suppressed marrow hypercellularity and reticulin fibers (Fig 3E and S2 Table). In the spleens of double-transgenic mice, turning off JAK2 V617F expression or ruxolitinib treatment both resulted in marked suppression of extramedullary hematopoiesis, along with hemosiderin deposition, which may be the consequence of iron overload [46] once excessive erythroid expansion is blunted (Fig 3F and S2 Table). In situ hybridization on spleen and bone marrow sections confirmed that JAK2 V617F expression was switched off in double-transgenic mice that were fed dox-containing chow (S6A and S6B Although the JAK1/2 inhibitor suppressed several hallmarks of MPN, including splenomegaly, or reduction of WBC counts, JAK2 V617F expressing cells persisted in spleens and bone marrow.
Thus, our data not only show that conditional-inducible JAK2 V617F expression is sufficient to give rise to key MPN disease hallmarks, but also that it is necessary to maintain the disease, evidenced by reversibility of the MPN-like phenotype when transgene expression is switched off. Although JAK inhibitor treatment also impressively impacted a number of MPN disease hallmarks in the model, there were differences in several readouts compared to the cohort in which transgene expression had been switched off. Importantly, JAK2 V617F expressing cells persisted despite JAK inhibitor treatment, corroborating earlier findings [19], and being seemingly consistent with allele burden data in clinical settings [25].

Impaired progenitor cell function is reversible upon switching-off JAK2 V617F expression
To characterize more precisely the impact of SCL-tTA;JAK2 V617F phenotypic reversion on hematopoiesis, we analyzed bone marrow cells of mice (S7A Fig) that had been fed normal chow for 275 days. Animals were then distributed in 3 groups: SCL-tTA;JAK2 V617F mice were put either off dox, or on dox for 26 days, or treated with ruxolitinib for 26 days, before bone marrow analysis. FACS quantification of Lin -Kit + Sca1 -(LKS -) progenitors and Lin -Kit + Sca1 + progenitors (enriched in hematopoietic stem cells) revealed an increase in the proportion of both types of progenitors when comparing control animals to diseased animals kept off dox all the time, consistent with clinical reports in which JAK2 V617F was found to increase amounts of HSC and progenitors [47]. Reversion of JAK2 V617F expression for 26 days was not able to revert the increase in these progenitors, whereas ruxolitinib could efficiently revert this effect (S7B  The MPN-like disease of SCL-tTA;JAK2 V617F mice is reversible upon switching-off JAK2 V617F expression or chemically inhibiting JAK2. SCL-tTA/+;JAK2 V617F /+ mice were aged until they developed MPN-like disease and randomized in 4 groups before being subjected to 26 days of treatment. One group was fed with doxycycline (+dox.) containing chow to turn off JAK2 V617F expression, while in the other 2 study cohorts, mice either received ruxolitinib (-dox. +ruxolitinib), or drug vehicle only (-dox. +vehicle). In a fourth arm, wild-type control littermates were also given vehicle (control). Mice were then sacrifized at experimental endpoint and (A) WBC counts and (B) spleen volume, (C) reticulocytes, (D) platelet counts were measured and the % change versus values of control animals was calculated, n = 5-7 mice per group. Mean ± SEM are indicated. Unpaired student t-test #p<0.05 compared to-dox. +vehicle group, or � p<0.05 between groups. NS: not significant, WBC white blood cells, PLT: platelets. (E) The phospho-STAT5 marker was assessed in bone marrow by immunohistochemistry. Bone marrow fibrosis was evaluated on the sternum using a silver impregnation staining. Bars In order to understand the proliferative potential of these progenitors, ex vivo colony-forming assays were performed on methylcellulose with LKS -, CMP, GMP, or MEP multipotent progenitors isolated from aged mice treated as previously (Fig 4A). Without dox, a markedly increased erythropoietic expansion in SCL-tTA; JAK2 V617F was observed in assays performed with various progenitors. This phenomenon was found to be reversible upon switching off JAK2 V617F expression or pharmacological JAK1/2 inhibition, since CFU-counts of multipotent myeloid lineage progenitors approached those of littermate controls.
To further investigate whether JAK2 V617F may alter repopulating properties, we sorted bone marrow LKSpopulations from mice from all groups and performed in vitro CFU serial replating assays (Fig 4B). Compared with controls, JAK2 V617F expressing colonies showed significantly greater colony formation on replating. Inhibition of JAK2 V617F expression or JAK1/ 2 with ruxolitinib reversed this effect completely. Thus, JAK2 V617F reduced clonal exhaustion of multipotent progenitors, but eventually progenitor cell expansion did slow down.

MPN-like disease is transplantable, and disease manifestation in the host animals can shift compared to the donor phenotype
Next, we assessed if the JAK2 V617F expression-driven MPN-like disease in the transgenic model would be transplantable. To this end, bone marrow cells were harvested from 4 male and 3 female SCL-tTA;JAK2 V617F mice that had been fed normal chow for nearly 13 months and developed MPN-like disease (S3 Table). Bone marrow cells from each of the respective double transgenic animals were transplanted into six lethally irradiated C57Bl/6J recipient mice (3 female and 3 male recipients). Blood counts were then analyzed 28, 68 and 161 days after transplantation. At the earliest time of analysis platelet count was elevated in most of the transplanted animals ( Fig 5A). However, this may be interpreted as a platelet rebound subsequent to irradiation-induced thrombocythemia. Notwithstanding, 40 days later the blood count analysis revealed increases in WBC in the majority of animals. Interestingly, the recipient group without elevated platelet and WBC counts displayed an elevated reticulocyte count (S3 Table). Approximately 9 weeks post-transplantation micro-ultrasound imaging on 3 randomly picked animals showed that spleen volumes were roughly 4 times larger than volumes previously determined in control mice (Fig 5B, S8 Fig, and S3 Table).
By day 161 after transplantation, platelet count had returned to normal levels in the transplanted groups, while WBC count increased in three of four groups when compared to prior readings ( Fig 5A). The exception was the recipient group without elevated platelet and WBC counts, and animals in this group continued to have a higher reticulocyte count as compared with the other transplanted groups (Fig 5A). This finding was unexpected, as the animals in this group had been transplanted with bone marrow from the donor mouse that had the highest WBC count, but an unremarkable reticulocyte count (S3 Table). It was also evident that myeloproliferation mainly seemed to involve WBCs, as groups transplanted with bone marrow from donor mice with either increased platelet or reticulocyte counts did not develop thrombocythemia or erythrocytosis. Taken together, these results demonstrate that the JAK2 V617F expression-driven MPN-like disease is transplantable, with a bias towards development of leukocytosis, along with splenomegaly.

Transplanted MPN-like disease is reversible upon switching off JAK2 V617F expression
To determine whether the MPN-like disease was dependent on JAK2 V617F expression in the transplant setting, transplanted mice were randomized based on hematocrit values into two cohorts of 12-13 mice. They were either continued on normal chow or switched to dox containing chow for one month prior to analyses. The remaining transplanted mice were divided into four cohorts of 4 mice to explore the short-term kinetics of histological and pharmacodynamic marker changes subsequent to modulation of JAK2 V617F expression by either diet switching or acute JAK inhibitor treatment. Animals were either administered ruxolitinib at 60 mg/kg (3 consecutive doses, 12 hours apart) or switched to dox containing chow for 1, 2, or 3 days, and given drug vehicle every 12 hours, prior to analyses. In agreement with the notion that JAK2 V617F expression is required to initiate and maintain MPN disease, the cohort of transplanted mice switched to dox containing food for one month exhibited reduced spleen weight when compared to the group that continued on normal chow (Fig 6A). Leukocytosis was also significantly suppressed by switching mice to dox containing chow (Fig 6B), while reticulocyte and platelet counts were within normal range (Fig S9A,B). A switch to dox containing chow for 3 days exhibited a trend for spleen weight reduction, demonstrating that the MPN-like disease is impacted rapidly by modulating JAK2 V617F expression (Fig 6A).
Histopathological analysis of spleen sections showed reduced extramedullary hematopoiesis in the group switched to dox containing chow for one month (Fig 6C), consistent with decreased spleen weight in this cohort. Analyses of bone marrow histopathology also demonstrated normalization of myeloproliferation, accompanied by expansion of erythroid and megakaryocytic lineages, in this cohort (Fig 6D), which also displayed markedly lower numbers of reticulin fibers in the bone marrow when compared with the cohort on normal chow (S9C Fig). Immunohistochemistry detection of STAT5 phosphorylation was markedly reduced in the bone marrow and spleen of animals in the cohorts switched to dox containing chow (Fig 6C and 6D and S9C Fig), with reduced levels already evident already in the group that was put on dox containing diet for one day. Levels of STAT5 phosphorylation were also substantially lower in the bone marrow and spleen of the cohort treated with ruxolitinib (Fig 6C and  6D and S9C Fig). Cleaved-caspase 3, a marker of apoptotic cell death, exhibited a modest temporal trend with higher levels in the cohort on dox containing chow for less than 3 days and lower levels at later time points. In contrast, acute ruxolitinib treatment increased the cleavedcaspase 3 histology score in the bone marrow by approximately 2-fold, suggesting that the reduction of the phospho-STAT5 signal in the ruxolitinib cohort may be attributable to both JAK2-STAT5 pathway inhibition and induction of cell death, whereas in the cohorts switched to dox containing chow it likely predominantly reflects normalization of JAK2 signaling. Thus, in the transplant setting continuous expression of JAK2 V617F is required to sustain MPN hallmarks and the remodeling of the bone marrow, including prevalence of reticulin fibers, is impacted swiftly upon turning off JAK2 V617F expression.

Discussion
Our present report directly addresses i) disease initiation and ii) disease maintenance using an inducible conditional JAK2 V617F allele. By utilizing the SCL-tTA-2S tet-off system, we have been able to successfully generate the first conditional inducible mouse model of JAK2 V617Fdriven MPN-like disease. Our SCL-tTA;JAK2V617F mouse model meets the criteria of the Bethesda proposed classification of murine nonlymphoid hematopoietic neoplasms [48] to be defined as a myeloproliferative disease. Indeed, elevated reticulocytes and platelets were observed in peripheral blood (Fig 1, S3 Fig, and S5 Fig). Moreover, histopathology assessment showed increased hematopoiesis and myelopoiesis in spleen and bone marrow, and increased erythrocytic and megakaryocytic expansion in spleens (Fig 2, S4 Fig, and S2 Table). Finally, we demonstrate that this disease is transplantable (Fig 5, S9 Fig, and S3 Table), thus excluding a reactive lesion. Compared to other JAK2 V617F mouse models described so far, our model is unique in that it enables assessing reversibility of disease phenotypes upon switching off transgene expression. Double transgenic mice, in which JAK2 V617F expression was induced gradually, developed hallmarks of MPN, including elevated blood counts, splenomegaly, bone marrow hyperproliferation and reticulin fiber deposition. Interestingly, female mice exhibited disease hallmarks of PV, characterized by more pronounced splenomegaly and higher reticulocyte count, while male mice displayed a phenotype resembling ET, characterized by an elevated platelet count. To our best knowledge, such gender differences in MPN-like disease manifestation have not been described with other JAK2 V617F mouse models. In humans, PV appears to be diagnosed somewhat more frequently in women, manifesting at younger age and exhibiting a higher likelihood to evolve from essential thrombocythaemia as compared to men [49]. The gender differences in MPN-like disease manifestation in our model may be attributable to sex hormones. For instance, estrogen was shown to be responsible for increased hematopoietic stem-cell self-renewal and splenic erythropoiesis in female mice [50], and in a mouse model of JAK2 V617F -driven MPN-like disease, tamoxifen treatment counteracted the survival advantage of mutant hematopoietic stem and progenitor cells, thereby suppressing disease hallmarks [51]. Our model should allow to genetically test modification of the JAK2 V617F -induced MPN phenotype by hormone receptors and other genes.
Consistent with previous reports [52], the JAK2 V617F expression-driven MPN-like disease is transplantable, manifesting itself quite rapidly in younger recipient mice, and arguing that the disease-causing clone is endowed with multi-potency and self-renewal properties. However, our results also show that a shift in MPN phenotype can occur in recipient mice compared to the MPN-like disease that was originally observed in donor mice. MPN-like disease in transplanted recipient mice predominantly manifested as leukocytosis. The reasons for skewing towards this particular MPN phenotype remain to be determined, but may involve factors such as the initial hematopoietic stress related to the transplant procedure, and/or the interaction that transplanted JAK2 V617F expressing hematopoietic stem-and progenitor cells establish with their new stroma and hematopoiesis-inducing microenvironment. In the transplant setting, switching off JAK2 V617F expression also led to a reversal of the MPN-like phenotype. While acute JAK inhibitor treatment increased apoptosis in the bone marrow, there was a tendency for a gradual decrease in the number of cleaved-caspase 3-positive cells as compared with baseline upon switching off JAK2 V617F expression. The latter finding is seemingly consistent with normalization of both JAK/STAT signaling and hematopoiesis, as aberrant JAK2 V617F signaling and myeloproliferation have been attributed to increased reactive oxygen species and DNA damage [23,53], with ensuing damage and death of cells in the marrow microenvironment [54].
Our transgenic mouse model corroborates that JAK2 V617F expression is sufficient to initiate MPN-like disease [17][18][19][20][21][22][23]. The fact that disease manifests with high penetrance and largely contemporaneously further supports JAK2 V617F as the critical driver. It is conceivable that JAK2 V617F triggers epigenetic changes in the MPN clone, and additional mutations may be required for, and acquired during pathogenesis. Seemingly consistent with this notion, additional mutations have been found to co-occur with JAK2 V617F in the MPN clone, including mutations in TET2, ASXL1, DNMT3A, EZH2 or IDH1/2, which are thought to impact epigenetic regulation and contribute to clonal skewing [55,56]. Furthermore, the MPN clone might induce a self-perpetuating remodeling of the hematopoiesis inducing micro-environment, and these stromal changes cooperate with the mutant clone in MPN pathogenesis [44,57]. Therefore, at least some of the initial dependency on JAK2 V617F may be lost during MPN pathogenesis, and JAK inhibition may only be able to tackle those aspects of the diseases that continue to rely on constitutive activation of the JAK/STAT pathway. On the other hand, conditional deletion of JAK2 in the MPN clone of a mouse model of MPL W515L -driven myelofibrosis-like disease was shown to lead to substantially improved disease modification as compared with JAK inhibitor therapy, providing compelling genetic evidence for continued dependency of key pathological features on constitutive activation of the JAK/STAT pathway in the MPN clone [30]. Similarly, in our model of JAK2 V617F -driven MPN we observe full reversibility of MPN-like disease upon turning off transgene expression. This observation demonstrates that continuous expression of JAK2 V617F is required to sustain disease.
Collectively, our findings also have important therapeutic implications, arguing for the discovery and development of pharmacological agents that specifically target V617F-mutant JAK2, while sparing the wild type protein. Such agents would be expected to profoundly modify disease in mutation-positive MPN patients. In pharmacological terms, our transgenic mouse model of MPN-like disease, when switched to dox containing diet, may mimic the actions of a V617F mutation-specific JAK2 inhibitor in the context of the heterozygous mutant clone, i.e. with one wild type JAK2 allele still being present: In the transgenic mouse model JAK2 V617F expression is switchable in the background of wild type murine Jak2. Thus, when transgene expression is turned off, the cell's Stat signaling relies again entirely on normal Jak regulation. Our model does not enable us to study the consequences of blunting aberrant Stat signaling in a setting of homozygous JAK2 V617F mutation. However, compelling transgenic mouse model studies have established that Jak2 is essential for hematopoietic stem and progenitor cell functions [58,59], and that conditionally deleting Jak2 in the MPL W515L -mutant MPN clone profoundly modifies myelofibrosis-like disease, including allelic burden, which has not been achieved to this extent with first-generation JAK inhibitors [30]. Therefore, specifically targeting JAK2V617F in the homozygous mutant MPN clone during the early course of disease might not only have the potential to eradicate clonal dominance, but also the disease, and to enable wild type hematopoietic stem cells to resume normal hematopoiesis.

Generation of the construct and reversible JAK2 V617F transgenic mice
The human JAK2 V617F cDNA was inserted into pCe-lox511-CMVmin and an IRES-EGFP was added. The microinjection was done in male C57Bl/6 pronuclei. Injected embryos were implanted into pseudopregnant females. 48 transgenic founders were obtained and presence of transgene was assessed by PCR. The founder line 29, B6-Tg(TetOCMV-JAK2V617F-IRE-S-EGFP)29Npa, was used for further experiments.

Experimental animal studies
All animal studies were conducted in accordance with Kantonales Veterinäramt Basel-Stadt and strictly adhered to the Eidgenössisches Tierschutzgesetz and the Eidgenössische Tierschutzverordnung. The procedures were approved by permit number BS-2139 and BS-2287 and issued by the Kantonales Veterinäramt Basel-Stadt. All animals had access to food and water ad libitum and were identified with transponders. Mice were housed in a specific pathogen-free facility with a 12-h light/12-h dark cycle. Animals were monitored carefully and regularly in all experiments. Mice were euthanized by carbon dioxide overdose.

Genotyping of animals
DNA was extracted from tail clips and was digested with proteinase K. SCL-tTA-S2 was detected by PCR (annealing temperature of 58C) with the following primers 5'-ccctgctc gatgccctggc-3' and 5'-aggaaggcaggttcggctcc-3'. An amplicon of 420bp was generated in presence of the knock-in allele. Human JAK2 was genotyped by PCR (annealing temperature of 58C) with the following primers 5'-ACTAAATGCTGTCCCCCAAAG-3' and 5'-CCCATGCCAACTGTTTAGC-3'. An amplicon of 480bp was generated in presence of the transgene.

Ultrasound imaging of the spleen in vivo and measurement of spleen dimensions
Mice spleens were imaged by ultrasound using a Vevo2100 1 High-Resolution Imaging System (VisualSonics, Toronto, Canada). Briefly, mice were starved 3 hours before imaging and fur was removed from the abdomen of mice using depilatory cream (Nair), then the skin was washed with water to remove excess lotion. Mice were maintained anesthetized for the duration of the imaging (1.5% isoflurane; 0.8L O2) on the heated imaging platform to preserve body temperature. With a generous amount of warmed ultrasound gel applied to the skin, the mice were restrained with surgical tape and ultrasound images were taken of its abdomen using the MS500D transducer equipped with 3D Motor. The sagittal and transverse planes are defined. More than 300 individual transverse image slices are obtained and then assembled together on the VevoLAB software for a 3D reconstruction.

Animal treatments
The expression of the transgene was switched off by doxycycline treatment. Mice were given doxycycline food pellet from Provimi Kliba. The food contained doxycycline at a concentration of 200mg/kg. Ruxolitinib was administrated at 60mg/kg in 0.5% hydroxypropyl methyl cellulose (HPMC) orally twice a day in a solution of 10ml/kg. On the bone marrow transplantation models, dosing of transplanted mice started 21 to 28 days following transplant when disease was established by blood count.

Bone marrow transplantation
Bone marrow cells were isolated from tibias and femurs of 15-17-month-old SCL-tTA / JAK2 V617F mice under sterile conditions. The bones were crushed in a mortar with RPMI + 10% FCS. The medium containing bone marrow cells was filtered through a cell strainer (BD Falcon, # 352340). Cells were centrifuged, the supernatant was removed, and red blood cell lysis buffer was added for 10 minutes at room temperature (Stemcells technologies #07800). Then, RPMI + 10% FCS was added and cells were washed and centrifuged again. Cells were resuspend in pure sterile PBS, counted using Sysmex Nucleocounter, and aliquoted for bone marrow transplantation. 8-9 weeks old C57Bl/6 recipient mice were subjected to total body irradiation (Biobeam, GSM Gamma-Service Medical GmbH) on the day of cells transplantation (2 doses of 4,5Gy at 4hrs interval). Immediately following the second irradiation, mice received a transplant of 5x10 6 bone marrow cells (from aged SCL-tTA / JAK2 V617F double mutant donor animals), injected in tail vein i.v. in a total volume of 100ul in HBSS (Gibco 1 , Cat#14175) at pH 7.4. Body weights were monitored daily for 2 weeks post-transplant. Mice were given antibiotics (Bactrim (Roche Pharma AG) 5ml in 250mL drinking water) to prevent infections for 14 days post bone marrow transplant, and DietGel Boost.

Blood counts
Blood was collected from the tail vein using Microvette 100K3E tubes with EDTA (SAR-STEDT, cat#. 20.1278) for blood analysis. Blood parameters (red blood cells, white blood cells, platelets, and reticuocytes counts) were determined on Sysmex blood analyzer.

Histopathological assessment
Sterna, femurs and spleens were collected, fixed in formalin (10% neutral buffered formalin) for 48 hours at room temperature, EDTA decalcified for 3x24 hours at 37˚C for bones, rinsed in PBS, processed for dehydration, cleared and paraffinized. After embedding in paraffin, 3μm sections were prepared. H&E staining was performed on all organs. Bone marrow fibrosis was evaluated on the sternum using a silver impregnation kit for reticulin fibers (Bio-Optica, #04-04080). Immunohistochemistry 3um thickness tissue section were stained on a BondRX platform (Leica Biosystems) using Epitope Retrieval 2 (ER2) conditions for 20mn at 100˚C for all tested markers. The detection system used was either Refine DAB kit (Leica Biosystems) for p-STAT5 and Cleaved Caspase 3 immunohistochemistry or Refine Red kit (Leica Biosystems) for GFP immunohistochemistry. Slides were hematoxylin counterstained, dehydrated and cover slipped. Primary antibodies used were anti-phospho-STAT5 (clone C11C5, CST, #9359) or cleaved Caspase3 (CST, #9661) or GFP (MBL, #598).

mRNA in situ hybridization
In situ hybridization was performed using either the 1-plex Quantigene View RNA ISH Tissue assay kit (Affymetrix, #QVT0051) or the 2-plex Quantigene View RNA ISH Tissue Assay kit (Affymetrix, #QVT0012). Assays were performed as per the manufacturer's instructions with stock solutions. Before beginning the hybridization protocol, 5um thickness tissue section were baked at 60˚C for 30 min to increase tissue adhesion. The paraffin was removed from the slides with xylene before being boiled in a pretreatment solution (Affymetrix) for 10 min and incubated with Protein Kinase K (Affymetrix) at 40˚C for 10 min. Custom probes for Human JAK2 and Mouse Jak2 (Affymetrix) were then hybridized to the tissue. Signal amplification was accomplished by hybridizing Type1 and Type2 specific pre-amp oligonucleotides, amp oligonucleotides and labeled oligonucleotides sequentially. Slides were then mounted with an aqueous mounting medium (Dako, Ultra mount). Digital slide image data was generated from the glass slides using the Aperio AT2 slide scanner (Leica Biosystems) for bright field, and the P250 Scanner (3D Histech) for fluorescent acquisition. ARIA and data analyzed using FlowJo software (Tree star, Ashland, OR). Briefly, bone marrow (depleted of red blood cells) was labeled with APC-conjugated mature blood cell lineage-specific antigens CD3, CD4, CD8, Gr-1, Mac-1, B220, CD19 and TER119. Doublets, DAPI positive dead cells, and cells expressing lineage markers were excluded from the analysis. More details in Supporting Materials and methods.

Hematopoietic cell colony-forming assays
Mouse tibiae were used to extract bone marrow cells and phenotypically defined hematopoietic progenitor-cell populations were obtained by FACS. Myeloid lineage colony-forming assays were performed in SCF, IL-3, IL-6, EPO supplemented methylcellulose-based medium (Methocult M3434, Stem Cell Technologies). Doxycycline (1ug/ml) and ruxolitinib (1μmol/l) were applied to the medium where indicated and replaced every three days. For LKS-, CMP, GMP and MEP methylcellulose cultures, 2000 cells were seeded in duplicate and scored for colony formation at 7-10 days. For serial replating experiments, LKS+ cells were plated at 2,000 cells per well and scored for colony formation after 10-12 days of primary culture. Cells were subsequently counted and resuspended in fresh medium with 2 x 10 5 cells replated for each of three rounds of serial replating. Serial plating efficiency was scored as percentage of initial CFU per well obtained at 10-12 days culture.  Table. SCL-tTA/+;JAK2 V617F /+ induced MPN-like disease is transplantable and disease manifestation in the host animals can shift compared to the donor phenotype. S1 Methods (JPG) S1 Methods. (DOC)