JAK2V₆₁₇F-positive endothelial cells contribute to clotting abnormalities in myeloproliferative neoplasms

Significance

The myeloproliferative neoplasms (MPNs) are a group of hematological malignancies characterized by increased numbers of myeloid blood cells, such as platelets, erythrocytes, and neutrophils. The main causes of illness and death in patients with MPNs are arterial and venous clotting and also, conversely, bleeding complications. However, the causes of these conditions are poorly understood. In this paper, we use a mouse model of MPNs to determine the cell types responsible for abnormal clotting in MPNs. We demonstrate that endothelial cells, the cell type that lines all blood vessels, have a significant role to play in MPN bleeding complications, potentially identifying a new cellular target for MPN therapies.

Abstract

The Janus kinase 2 (JAK2) V₆₁₇F mutation is the primary pathogenic mutation in patients with Philadelphia chromosome-negative myeloproliferative neoplasms (MPNs). Although thrombohemorrhagic incidents are the most common causes of morbidity and mortality in patients with MPNs, the events causing these clotting abnormalities remain unclear. To identify the cells responsible for the dysfunctional hemostasis, we used transgenic mice expressing JAK2V₆₁₇F in specific lineages involved in thrombosis and hemostasis. When JAK2V₆₁₇F was expressed in both hematopoietic and endothelial cells (ECs), the mice developed a significant MPN, characterized by thrombocytosis, neutrophilia, and splenomegaly. However, despite having significantly higher platelet counts than controls, these mice showed severely attenuated thrombosis following injury. Interestingly, platelet activation and aggregation in response to agonists was unaltered by JAK2V₆₁₇F expression. Subsequent bone marrow transplants revealed the contribution of both endothelial and hematopoietic compartments to the attenuated thrombosis. Furthermore, we identified a potential mechanism for this phenotype through JAK2V₆₁₇F-regulated inhibition of von Willebrand factor (VWF) function and/or secretion. JAK2V₆₁₇F⁺ mice display a condition similar to acquired von Willebrand syndrome, exhibiting significantly less high molecular weight VWF and reduced agglutination to ristocetin. These findings greatly advance our understanding of thrombohemorrhagic events in MPNs and highlight the critical role of ECs in the pathology of hematopoietic malignancies.

Myeloproliferative neoplasms (MPNs) are clonal hematopoietic stem cell disorders, characterized by significant increases in one or more myeloid-cell lineages. Mutations in the Janus kinase 2 (JAK2) and MPL genes are common in the majority of Philadelphia chromosome-negative (Ph⁻) MPNs, which include polycythemia vera (PV), essential thrombocythemia (ET), and primary myelofibrosis (PMF). By far the most frequent mutation in MPNs is JAK2V₆₁₇F (1–4), which occurs in the highly conserved autoinhibitory JAK homology (JH) 2 domain, causing hyperactive kinase activity and hyperproliferation of myeloid progenitor cells, leading to overproduction of red blood cells (RBCs), platelets, and leukocytes. Although ET and PV have a relatively benign clinical course, patients’ life expectancy can be severely reduced by bleeding or thrombosis, the manifestations of which are significantly more common than other MPN-related complications such as myelofibrosis and acute leukemic transformation (5). The frequency and nature of thrombotic and hemorrhagic events vary greatly depending on disease phenotype and patient history. Data taken from a number of previous studies indicate that the probability of major thrombosis ranges between 8–29% (ET) and 11–39% (PV) whereas the incidence of bleeding at initial presentation is less frequent than thrombosis, ranging between 3–18% (ET) and 3–8% (PV) (6–8).

A number of abnormalities that could potentially contribute to this prothrombotic phenotype have been identified in the blood and vascular cells of JAK2V₆₁₇F⁺ MPN patients. Much work has focused on defining platelet abnormalities, including increased expression of membrane proteins such as P-selectin and tissue factor (TF), which would prime platelets for activation and increase levels of platelet-activation markers and platelet factor 4 (PF4) in the plasma (9–12). Interestingly, however, aggregation studies show a decreased response to ADP and epinephrine in platelets isolated from patients with ET and PV compared with controls (10). Furthermore, no correlation has been made between severity of thrombocytosis in ET patients and increased risk of thrombosis (6, 13). In contrast, extreme thrombocytosis (platelets >1,500 × 10⁹/L) is thought to contribute to a hemorrhagic phenotype in ET patients, and is commonly attributed to the development of acquired von Willebrand syndrome (AVWS) (11, 12, 14), where the increased platelets bind to highly prothrombotic, ultralarge von Willebrand factor (VWF) multimers, removing them from the plasma (15).

Recent studies suggest that leukocytosis is a potential thrombotic risk factor in young PV and ET patients, possibly through the interactions of leukocytes, especially neutrophils, with platelets and endothelial cells (ECs) (16, 17) or the production of prothrombotic molecules such as TF. Increased basal activation of neutrophils has been shown in PV and ET patients, including elevated expression of CD11b and levels of neutrophil proteases in the plasma, both of which are prothrombotic (9, 18, 19). Studies have also shown increased activation of vascular ECs in JAK2V₆₁₇F⁺ MPN patients. Increased P- and E-selectin levels in the plasma, coupled with decreased levels of nitric oxide (NO), could conceivably contribute to a prothrombotic phenotype. Furthermore, JAK2V₆₁₇F⁺ ECs have recently been reported in a subpopulation of MPN patients, and EC expression was coupled with an increased risk of thrombosis (20, 21). Taken together, previous studies describe physiological abnormalities in a number of cell types in JAK2V₆₁₇F⁺ MPN patients, all of which could contribute toward increased thrombosis and/or bleeding. However, these data are often contradictory and fail to definitively explain the mechanism/s responsible for the development of thrombohemorrhagic disease.

Here, we used FF1 transgenic mice (22) to express human JAK2V₆₁₇F in specific lineages to determine which cells are responsible for the thrombohemorrhagic manifestations seen in patients with MPNs. FF1 mice were crossed with Pf4-Cre or Tie2-Cre mice to express JAK2V₆₁₇F specifically in platelets alone, or in hematopoietic cells and ECs, respectively (23–28). These models have provided us with an unparalleled opportunity to determine the specific role/s of JAK2V₆₁₇F in pathological thrombosis and hemostasis.

Results

Tie2-Cre/FF1 Mice Develop an MPN-Like Phenotype.

FF1 mice were crossed with either Pf4-Cre or Tie2-Cre mice to generate mice with either platelet-specific or hematopoietic and EC-specific JAK2V₆₁₇F expression (Fig. S1A). To confirm lineage-restricted expression of human JAK2V₆₁₇F in these mice, quantitative PCR analysis was performed on isolated megakaryocytes (MKs), marrow neutrophils (BMNs), and ECs (Fig. S1B). As expected, WT mice did not express human JAK2, whereas Pf4-Cre/FF1 mice expressed the human transgene only in MKs. In contrast, Tie2-Cre/FF1 mice expressed human JAK2 in MKs, BMNs, and ECs (Fig. S1B). No significant difference in levels of human JAK2 expression was seen in megakaryocytes isolated from Pf4-Cre/FF1 and Tie2-Cre/FF1 mice, where the ratio of humanJAK2:mouseJak2 expression levels was ∼1:10. Tie2-Cre/FF1 mice additionally showed humanJAK2:mouseJak2 ratios of 1:3 in BMNs and 1:5 in ECs. To determine the percentage of hematopoietic cells that express the JAK2V₆₁₇F mutation, primary marrow was collected from Tie2-Cre/FF1 mice and WT controls for colony assays. After 10 d in culture, individual colony-forming unit-granulocyte (CFU-G) colonies were analyzed for expression of the human JAK2 transgene by RT-PCR (Fig. S1C). One hundred percent of the colonies picked from Tie2-Cre/FF1 mice expressed JAK2V₆₁₇F. Colony assays from isolated primary marrow also showed a significant increase in total colony formation in Tie2-Cre/FF1 mice compared with both WT and Pf4-Cre/FF1, particularly in the number of granulocyte/macrophage (CFU-GM) and megakaryocyte (CFU-MK) progenitor cells (Fig. S1D). Pf4-Cre/FF1 mice also exhibited a modest but significant increase in CFU-MK colonies compared with WT. Thrombocytosis and neutrophilia were observed in Tie2-Cre/FF1 mice at 8 wk of age, both of which were even more significant by 16 wk, although no difference in RBC or total lymphocyte count was observed (Fig. 1A). At 22 wk, Tie2-Cre/FF1 mice exhibited increased red-cell distribution width (RDW) without significant differences in mean cell volume (MCV), compared with WT and Pf4-Cre/FF1 (Fig. S1E), indicative of premyelofibrosis in these mutants.

Fig. 1.

Tie2-Cre/FF1 mice, which express JAK2V₆₁₇F specifically in hematopoietic and endothelial cells, develop a severe MPN. FF1 mice were crossed with Pf4-Cre and Tie2-Cre-expressing mice to activate JAK2V₆₁₇F expression specifically in platelets (Pf4-Cre/FF1), or hematopoietic and ECs (Tie2-Cre/FF1), respectively. Periodic complete blood counts (CBCs) were performed at 4, 8, and 16 wk of age (A). Sections of bone marrow stained with hematoxylin and eosin revealed severe osteopetrosis in the Tie2-Cre/FF1 mutants at 32 wk, and spleens collected at 16 wk show splenomegaly (B).

Similar to previously characterized Vav-Cre/FF1 mice (22), bones removed from Tie2-Cre/FF1 mice appeared pale on gross examination (Fig. S1F). Upon isolation of marrow cells, significantly reduced levels of TER-119 and CD71 expression, measured by flow cytometry, confirmed decreased marrow erythropoiesis in Tie2-Cre/FF1 mutants, which appears to be compensated by increased splenic erythropoiesis (Fig. S2). The predominant cell type in the marrow of these mutants was GR1/Mac-1–positive myeloid cells. CD41/CD61-positive megakaryocytes were also increased in both the marrow and spleen of Tie2-Cre/FF1 mice although a significant reduction in the number of CD3-positive T cells was observed in both tissues, compared with controls. Marrow histology at 16 wk further confirmed greatly increased megakaryopoiesis in Tie2-Cre/FF1 mice compared with controls (Fig. S1G), and, at 32 wk, Tie2-Cre/FF1 mice had developed extensive marrow osteopetrosis (Fig. 1B). Moreover, by 16 wk, Tie2-Cre/FF1 mice developed significant splenomegaly (Fig. 1B and Fig. S1H), and normal splenic architecture was destroyed (Fig. S1I).

Tie2-Cre/FF1 Mice Develop Significant Abnormalities in Hemostasis.

To determine the effects of JAK2V₆₁₇F expression on hemostasis and thrombosis in vivo, we performed carotid artery occlusion assays on WT, Pf4-Cre/FF1, and Tie2-Cre/FF1 mice. Five days before ferric chloride (FeCl₃) injury, complete blood counts (CBCs) were performed to confirm the same Tie2-Cre/FF1 thrombocytosis and neutrophilia phenotypes observed previously (Fig. 1A), in the mice to be used for occlusion assays (Fig. S3). The FeCl₃ occlusion assays (Fig. 2 A and B) revealed no significant differences in occlusion time between WT (7.4 min ± 0.4) and Pf4-Cre/FF1 (8.2 min ± 1.0) mice, in which JAK2V₆₁₇F expression is restricted to platelets (Fig. 2B). Surprisingly, however, Tie2-Cre/FF1 mice failed to occlude in response to FeCl₃ injury over a 30-min period (Fig. 2 A and B). Increasing the FeCl₃ concentration to 10% or 12% also failed to induce arterial occlusion. To determine the composition of the FeCl₃-induced thrombus, carotid arteries were removed following the assays and histologically stained using the Carstair’s method (Fig. S4). Injured WT carotid arteries clearly exhibit platelet and fibrin-rich occlusive thrombi at the point of injury and extending across the lumen (Fig. S4B). However, Tie2-Cre/FF1 mice show only mild focal platelet plug formation and no deposition of fibrin, the remaining lumen filled with postmortem coagulated blood (Fig. S4C), as seen in uninjured mice (Fig. S4A). Next, we determined whether this antithrombosis phenotype was also present in a different vascular environment, by assessing thrombus formation following injury to the tail (Fig. 2C). Again, we found no significant difference in bleeding times between WT and Pf4-Cre/FF1 mice, but Tie2-Cre/FF1 exhibited a significantly increased bleeding time (Fig. 2C).

Fig. 2.

Tie2-Cre/FF1 mice fail to clot in response to injury. Differences in in vivo thrombosis in response to injury were studied in age-matched WT (n = 10), Pf4-Cre/FF1 (n = 7), and Tie2-Cre/FF1 (n = 11) mice (A). Occlusion times in WT, Pf4-Cre/FF1, and Tie2-Cre/FF1 mice (B) (***P < 0.001). Tail-bleeding assays were also performed on the three mouse models (C) (***P < 0.001).

JAK2V₆₁₇F Expression Has No Effect on Platelet Function.

We next established whether the abnormal thrombotic response to injury observed in Tie2-Cre/FF1 mice was due to platelet dysfunction. Initial whole-blood aggregometry studies showed that blood from Tie2-Cre/FF1 mice aggregated quicker, and to a greater extent, in response to a combination of epinephrine (10 μM) and ADP (20 μM), or collagen (10 μg/mL), than whole blood from both WT and Pf4-Cre/FF1 mice (Fig. 3A and Fig. S5A). Due to the significant thrombocytosis in Tie2-Cre/FF1 mice, compared with the other mouse models, it was possible that the increased aggregation was merely due to increased platelet numbers, rather than increased activity. To test this, aggregation was measured using equal numbers of washed platelets isolated from each of the genotypes, in response to the same platelet agonists (Fig. 3B and Fig. S5B). No significant differences in aggregation were seen between each of the three mouse lines, suggesting that the differences observed in whole-blood aggregation were indeed due to increased platelet number in Tie2-Cre/FF1 mice. To determine whether platelet integrin expression differed between WT and Tie2-Cre/FF1 mice, platelet levels of integrin β3, αIIb, β1, and α2 were measured by flow cytometry in resting platelets; no significant differences in membrane expression were observed (Fig. S5C). Moreover, no significant differences were seen in the activity of the platelets, determined by P-selectin expression and activated αIIbβ3 (clone JON/A) levels, both in the absence and presence of thrombin (0.1 U/mL) (Fig. S5C). However, Western-blot analysis suggested moderately increased levels of phospho-tyrosine in both Tie2-Cre/FF1 and Pf4-Cre/FF1 platelets compared with WT (Fig. S5D).

Fig. 3.

Comparison of platelet-aggregation phenotypes in WT, Pf4-Cre/FF1, and Tie2-Cre/FF1 mice. Differences in platelet function were analyzed by whole blood (A) and washed platelet (B) aggregometry. Aggregation was monitored for up to 10 min following agonist [epinephrine (10 μM) plus ADP (20 μM) or collagen (10 μg/mL)] stimulation. In all experiments, data represent mean ± SEM (**P < 0.01; n = 4).

ECs Are Critical in the Development of Dysfunctional Hemostasis.

As platelet function is not responsible for the abnormal hemostasis observed in Tie2-Cre/FF1 mice, we next aimed to determine precisely which cell type was responsible for this phenotype. We performed two types of marrow transplantation experiments, to more specifically restrict the cell types in which JAK2V₆₁₇F is expressed. First, marrow was collected from either Tie2-Cre/FF1 mice or age-matched WT controls and transplanted into irradiated WT (B6.SJL-Ptprca Pep3b/BoyJ, CD45.1) mice (Fig. S6A). These transplants generated mice in which JAK2V₆₁₇F expression was limited to hematopoietic cells whereas ECs were normal. Chimerism was confirmed by peripheral-blood analysis at 4 wk (93% ± 2.3 CD45.2⁺). By 8 wk posttransplant, recipient mice that had received Tie2-Cre/FF1 marrow had developed blood counts characteristic of an MPN, exhibiting significant thrombocytosis and neutrophilia (Fig. 4A), whereas RBCs were normal (Fig. S6B). However, carotid artery occlusion assays (Fig. 4B) demonstrated no significant difference in occlusion time between the Tie2-Cre/FF1 donor;WT recipient mice and the WT donor;WT recipient controls (Fig. 4C), and Carstair’s staining showed platelet and fibrin-rich thrombi in both (Fig. S4 D and E). Tail bleeding time assays were performed in the same mice ∼1 wk before carotid-occlusion assays and showed no difference in bleeding time between these two transplant models (Fig. 4D).

Fig. 4.

JAK2V₆₁₇F expression only in hematopoietic cells does not cause hemostatic abnormalities whereas expression only in endothelial cells causes attenuated arterial thrombosis following injury. Bone marrow from Tie2-Cre/FF1 mice was isolated and transplanted into lethally irradiated WT (CD45.1) mice to generate a model of hematopoietic-restricted JAK2V₆₁₇F expression. Conversely, bone marrow was isolated from WT (CD45.1) mice and transplanted into lethally irradiated Tie2-Cre/FF1 mice to generate a model of endothelial-restricted JAK2V₆₁₇F expression. Blood samples were taken 8 wk after transplant from (1) WT donor;WT recipients (n = 9), (2) Tie2-Cre/FF1 donor;WT recipients (n = 10), (3) WT donor;Tie2-Cre/FF1 recipients (n = 9), and (4) Tie2-Cre/FF1 donor;Tie2-Cre/FF1 recipients (n = 4) for CBC analysis (A) (***P < 0.001). Following injury to the carotid artery, mice with endothelial only JAK2V₆₁₇F exhibited significantly longer occlusion times and unstable thrombus formation (B and C). However, there was no significant difference in tail-bleed assay times between the restricted expression transplant models and the WT donor;WT recipient controls (D).

We then performed the reverse transplant, in which WT (CD45.1⁺) marrow was transplanted into either Tie2-Cre/FF1 or WT control mice (Fig. S6C). These experiments generated a mouse model in which only the ECs express JAK2V₆₁₇F whereas the transplanted hematopoietic compartment was normal. Peripheral blood chimerism was again confirmed at 4 wk (94% ± 3.1% CD45.1⁺). CBCs at 8 wk confirmed no significant differences between the WT donor;Tie2-Cre/FF1 recipient mice and the WT donor;WT recipient controls (Fig. 4A and Fig. S6D). However, carotid artery occlusion assays (Fig. 4B) showed a significant increase in occlusion time in Tie2-Cre/FF1 recipients (18.7 ± 3.3 min), compared with WT recipients (8.7 ± 0.3 min) (Fig. 4C). Furthermore, the thrombus that did form was unstable in Tie2-Cre/FF1 recipient mice, as shown by subsequent reacquisition of flow, not seen in WT recipients (Fig. 4B). Additionally, Carstair’s staining confirmed a stable platelet and fibrin-rich occlusive thrombus in WT donor;WT recipients (Fig. S4D) whereas WT donor;Tie2-Cre/FF1 recipients formed only a small platelet plug at the site of injury (Fig. S4F). Despite exhibiting abnormal arterial thrombosis following injury, Tie2-Cre/FF1 recipients did not display a prolonged tail bleeding time compared with WT recipients (Fig. 4D). Tie2-Cre/FF1 donor;Tie2-Cre/FF1 recipients were phenotypically identical to congenital Tie2-Cre/FF1 mice (Fig. 4 A–D).

Tie2-Cre/FF1 Mice Exhibit von Willebrand Disease.

Next, we aimed to identify the potential mechanism/s responsible for the abnormal hemostasis phenotype in Tie2-Cre/FF1 mice. Bleeding diatheses in patients with MPNs has commonly been attributed to AVWS (8). ECs play an essential role in regulating VWF levels, and our data suggest that JAK2V₆₁₇F⁺ ECs contribute greatly to dysfunctional hemostasis. Therefore, we determined whether the Tie2-Cre/FF1 mice had a condition similar to AVWS. Levels of plasma VWF were not significantly different between WT and Tie2-Cre/FF1 mice, and both were similar to levels of VWF in pooled plasma taken from six C57BL/6 mice (Fig. S7A). However, significant differences were observed in the distribution of VWF multimers (Fig. 5A). Tie2-Cre/FF1 mice exhibit a significant reduction in ultralarge multimers and a subsequent compensatory increase in the levels of smaller VWF multimers (Fig. 5A). Densitometry revealed that the low molecular weight multimers (bottom four bands) contribute ∼60% of the total plasma VWF in WT mice whereas the same multimers make up 82% of plasma VWF in Tie2-Cre/FF1 mice (Fig. S7B). To characterize the nature of the AVWS in Tie2-Cre/FF1 mice, we performed ristocetin-induced platelet agglutination (RIPA) assays on whole-blood samples. We found that 2 mg/mL ristocetin led to significantly greater agglutination in WT blood compared with Tie2-Cre/FF1 (Fig. 5 B and C). The failure to agglutinate in the presence of ristocetin implies a condition similar to AVWS in Tie2-Cre/FF1 mice. Because ADAMTS13 is responsible for cleaving VWF, we examined whether increased levels of the protease in Tie2-Cre/FF1 ECs (Fig. S8) may account for the reduction of high molecular weight VWF multimers. However, ADAMTS13 levels were not increased in Tie2-Cre/FF1 ECs (Fig. S8), compared with WT ECs, and may even have been slightly decreased (Fig. S9). Finally, we investigated whether either of the transplant mouse models also displayed a comparable AVWS phenotype. The shift in VWF multimer distribution seen in Tie2-Cre/FF1 mice was absent in either of the transplant models, and instead each showed a similar VWF multimer pattern to the WT donor;WT recipient controls (Fig. 5D). Although the Tie2-Cre/FF1 donors;WT recipients also showed normal agglutination in RIPA assays (Fig. 5 E and F), the WT donor;Tie2-Cre/FF1 recipient mice showed a defective response to ristocetin (Fig. 5 E and F). Furthermore, endothelial cells isolated from the WT donor;Tie2-Cre/FF1 recipient mice showed increased levels of VWF compared with controls although the multimer pattern appeared similar (Fig. 5G).

Fig. 5.

Tie2-Cre/FF1 mice exhibit an acquired von Willebrand syndrome phenotype. VWF multimer distribution was determined by SDS agarose electrophoresis and Western blot analysis (A). Plasma from two representative WT controls and two representative Tie2-Cre/FF1 mice are shown. Ristocetin-induced platelet agglutination (RIPA) was carried out on whole blood isolated from WT controls (n = 4) and Tie2-Cre/FF1 mice (n = 3) (B). Agglutination was averaged in each group (±SEM) (C) (**P < 0.01). Plasma VWF multimer distribution (D), whole blood RIPA (E), and average agglutination (F) were also determined with the transplant mice. 1, WT donor;WT recipient; 2, Tie2-Cre/FF1 donor;WT recipient; 3, WT donor;Tie2-Cre/FF1 recipient. Additionally, the VWF multimer pattern was determined in endothelial cells isolated from the WT donor;Tie2-Cre/FF1 recipient transplant mice and WT donor;WT recipient controls (G). Each lane represents 20 µg of protein from lysed endothelial cells pooled from two transplant mice.

Discussion

Despite significant advances in our understanding of the development of MPNs, the mechanisms that lead to MPN-mediated hemostatic abnormalities, a primary cause of JAK2V₆₁₇F⁺ MPN morbidity and mortality, remain elusive. By crossing previously described JAK2V₆₁₇F knock-in mice (FF1) (22) with platelet-specific Pf4-Cre and hematopoietic/endothelial-specific Tie2-Cre mice, we have been able to highlight the importance of JAK2V₆₁₇F-positive endothelial cells in the dysfunctional thrombosis associated with MPNs.

The development of transgenic models of JAK2V₆₁₇F⁺ MPNs has allowed a physiologically relevant evaluation of the effects of the mutant protein, at differing levels of expression and in various cell types. The JAK2V₆₁₇F “Flip-Flop” model (22) has enabled us to express JAK2V₆₁₇F in platelets alone, or in the entire hematopoietic compartment and, importantly, in ECs. Pf4-Cre/FF1 mice, in which JAK2V₆₁₇F is expressed in platelets alone, did not develop an MPN phenotype, presumably because the Pf4 promoter is activated during early megakaryopoiesis, and will therefore not affect cell proliferation. Conversely, Tie2-Cre/FF1 mice exhibit a robust ET-like phenotype, similar to the previously described Vav-Cre/FF1 mouse (22). However, several significant differences remain between the Tie2-Cre/FF1 and Vav-Cre/FF1 mouse models, including the development of more significant neutrophilia, pronounced osteopetrosis, and, notably, endothelial expression of JAK2V₆₁₇F in Tie2-Cre/FF1 mice. Importantly, the MPN phenotype seen in Tie2-Cre/FF1 mice is transplantable into WT mice, demonstrating that the mutation is present in primitive hematopoietic stem cells.

Considering that MPN patients are commonly prothrombotic, we were surprised that Tie2-Cre/FF1 mice failed to form an effective thrombus following injury. However, although bleeding is less common than thrombosis, a significant proportion of MPN patients (3–18% in ET and 3–8% in PV) (6–8) have bleeding complications. Histological analysis of the Tie2-Cre/FF1 thrombus strongly suggests that the platelets adhere properly to the injury site, but that the platelet plug then fails to propagate into an occlusive thrombus, suggesting that the defect may not be in the platelets themselves, but rather in other cell types or factors involved in thrombus propagation. In addition, results from both Tie2-Cre/FF1 and Pf4-Cre/FF1 mouse models showed that the presence of the JAK2V₆₁₇F mutation in platelets modestly increased tyrosine phosphorylation but had no significant effect on platelet function or membrane adhesive glycoprotein expression. These findings contradict some previous reports, which suggest that platelets isolated from MPN patients have increased levels of P-selectin (17), decreased platelet functionality characterized by attenuated aggregation, and reduced levels of membrane adhesion molecules (10). However, our data suggest that ECs in JAK2V₆₁₇F⁺ MPN patients also warrant further investigation.

Although the bleeding diathesis in ET patients with extreme thrombocytosis has often been attributed to AVWS, a detailed understanding of how this condition develops remains elusive. The most accepted theory is that the increased platelets bind to highly prothrombotic ultralarge VWF multimers and remove them from the plasma (15). Additionally, enhanced VWF cleavage has also been associated with the increased number of platelets in ET patients with bleeding complications (29). Interestingly, in Tie2-Cre/FF1 mice, we also observe a significant reduction in ultralarge VWF multimers and a defective response to ristocetin, both of which are common in AVWS. However, when JAK2V₆₁₇F expression was restricted to hematopoietic cells, normal thrombosis was observed following injury, even though these mice showed increased platelet counts similar to Tie2-Cre/FF1. Moreover, when JAK2V₆₁₇F was only expressed in ECs, these animals failed to occlude normally and showed a defective response to ristocetin, despite a normal platelet count. These data suggest that JAK2V₆₁₇F-mediated changes in EC-derived VWF processing and/or release may be involved in the AVWS phenotype. However, it is intriguing that, despite an overall increase in EC VWF levels, the multimer pattern appeared normal in these animals, perhaps suggesting a dysfunction, rather than absence, of the high molecular weight VWF multimers. Investigating ECs in patients with ET may provide useful insights into the mechanisms that contribute to AVWS.

Although JAK2V₆₁₇F expression in ECs alone was sufficient to cause delayed occlusion after injury, these animals showed a less profound clotting disorder than the Tie2-Cre/FF1 mice, raising the possibility that an interaction between, or factors expressed in, both JAK2V₆₁₇F⁺ hematopoietic cells and ECs may be important in pathological hemostasis. Additionally, mice deficient for VWF showed impaired, but not completely deficient, thrombus formation after FeCl₃-induced vascular injury (30), similar to our EC-only JAK2V₆₁₇F mouse model, but in contrast to the complete failure to occlude seen in the Tie2-Cre/FF1 mouse. This intermediate phenotype again demonstrates the contribution of both the endothelial and hematopoietic compartments to thrombosis and further implicates the involvement of VWF. Furthermore, the reduction in ultralarge VWF multimers, characteristic of AVWS, was observed only when JAK2V₆₁₇F was expressed in both hematopoietic cells and ECs. Its absence from the EC-only JAK2V₆₁₇F mice agrees with current opinions that increased platelet numbers are responsible for clearing of the high molecular weight VWF multimers. However, the shift was also absent in the hematopoietic-only JAK2V₆₁₇F mice, which show high platelet counts, again suggesting that expression of JAK2V₆₁₇F in both hematopoietic and endothelial compartments contributes to the phenotype.

Signaling through nitric oxide may potentially provide a mechanism for the AVWS-like phenotype in Tie2-Cre/FF1. Nitric oxide synthesized by endothelial (e) NOS plays numerous critical roles in EC regulation and hemostasis, including the inhibition of Weibel–Palade body (WPB) exocytosis (31), which attenuates the release of ultralarge VWF multimers (32). Furthermore, JAK2 has been shown to mediate eNOS activity via AKT activation (33). Therefore, the role of eNOS in JAK2V₆₁₇F⁺ ECs and its potential effect on VWF release may be an important focus for future research. It is probable that JAK2V₆₁₇F also affects other, as yet unidentified, signaling pathways in ECs, which would also contribute to the dysfunctional thrombosis.

In support of our results highlighting the importance of ECs, recent reports have confirmed JAK2V₆₁₇F expression in ECs of patients with MPNs (21, 34) and even suggest that expression of the mutant protein may be responsible for thrombosis in PV patients with Budd–Chiari syndrome (20). In our model, ECs are, at least in part, responsible for dysfunctional hemostasis in response to injury, opposing some previous reports suggesting that JAK2V₆₁₇F⁺ ECs may be prothrombotic (35). One potential explanation for these differences is the remarkable heterogeneity of ECs. Indeed, we found a discrepancy in bleeding phenotypes when comparing different sites of injury; the EC-only JAK2V₆₁₇F mouse model showed a prolonged bleeding time after carotid artery injury, but abnormal thrombosis was not observed in the tail-bleeding assays. The two assays examine very different vascular beds. In the first, the injury is specifically localized to the carotid artery whereas, in the latter, amputation of the distal tip of the tail injures a number of vessel types, including arteriole, venule, and capillary. There is much heterogeneity in structure and function between the ECs that line different vessels. Specific locations also expose the cells to particular growth factors, cytokines, and shear stresses, which in turn regulate mRNA and ultimately protein expression (reviewed in ref. 36). Indeed, ECs throughout the vascular tree express different anticoagulant proteins; endothelial protein C receptor is predominantly expressed in large vessels (37), tissue type plasminogen activator expression is particularly high in cerebral and pulmonary arteries (38), and microvessels exhibit high levels of tissue factor pathway inhibitor (TFPI) expression (39). Site-specific differences in ECs may potentially underlie the variation in thrombohemorrhagic complications associated with MPNs.

In summary, our data show that the bleeding diathesis seen in a number of MPN patients may be due to the expression of JAK2V₆₁₇F in ECs, in addition to hematopoietic cells. We have identified a potential mechanism whereby mutant JAK2 may contribute to the bleeding phenotype via regulation of VWF. However, it is highly likely that a number of other, as yet unidentified, mechanisms contribute to EC-mediated thrombohemorrhagic events in MPNs. Further work is required to fully characterize expression of hemostatic proteins in JAK2V₆₁₇F⁺ ECs. It will also be particularly important to compare expression of hemostatic factors in ECs isolated from various points of the vascular tree and specific vascular beds. It is possible that JAK2V₆₁₇F⁺ ECs may be antithrombotic in some tissues and prothrombotic in others, providing a novel explanation for the variability in thrombohemorrhagic events in MPN patients.

Materials and Methods

JAK2V₆₁₇F Flip-Flop (FF1) (22) and Pf4-Cre (26) mice were provided by Radek Skoda (University Hospital, Basal, Switzerland), and Tie2-Cre mice (40) were provided by Mark Ginsberg (University of California, San Diego). FF1 mice were crossed with either Pf4-Cre or Tie2-Cre to generate cell lineage-specific JAK2V₆₁₇F knock-in mouse lines. All mice used were on a C57BL/6 background and were bred in a pathogen-free mouse facility at Stony Brook University. CD45.1⁺ congenic mice (SJL) were purchased from The Jackson Laboratory. Animal experiments were performed in accordance with the guidelines provided by the Institutional Animal Care and Use Committee at Stony Brook University.

SI Materials and Methods describes real-time PCR, CBCs, colony assays, histology, thrombosis and hemostasis assays, platelet assays, hematopoietic stem cell transplants, endothelial cell and neutrophil isolation, plasma VWF analysis, and Western blot analysis.