HSP27 is a partner of JAK2-STAT5 and a potential therapeutic target in myelofibrosis

Heat shock protein 27 (HSP27/HSPB1) is a stress-inducible chaperone that facilitates cancer development by its proliferative and anti-apoptotic functions. The OGX-427 antisense oligonucleotide against HSP27 has been reported to be beneficial against idiopathic pulmonary fibrosis. Here we show that OGX-427 is effective in two murine models of thrombopoietin- and JAKV617F-induced myelofibrosis. OGX-427 limits disease progression and is associated with a reduction in spleen weight, in megakaryocyte expansion and, for the JAKV617F model, in fibrosis. HSP27 regulates the proliferation of JAK2V617F-positive cells and interacts directly with JAK2/STAT5. We also show that its expression is increased in both CD34+ circulating progenitors and in the serum of patients with JAK2-dependent myeloproliferative neoplasms with fibrosis. Our data suggest that HSP27 plays a key role in the pathophysiology of myelofibrosis and represents a new potential therapeutic target for patients with myeloproliferative neoplasms.

H eat shock proteins (HSPs) are highly conserved molecular chaperones whose synthesis is induced by different stresses in both environmental and pathophysiological conditions 1 . HSP70 (HSPA1A, NP_005336) and HSP27 (HSPB1, NP_001531 (human); Hsp27, Hspb1, NP_038588 (mouse)) are the most strongly induced proteins following aggression due to oxidative stress, anti-cancer agents or ionizing radiation for example 2 . These HSPs were originally characterized as chaperones due to their ability to prevent aggregation and restore proteostasis and to their participation in the transport of proteins. These HSPs also have a strong cytoprotective action through the inhibition of apoptosis and autophagy processes 3,4 . Because cancer cells have to re-wire their metabolism, they need chaperones for their survival. As a consequence, HSPs are abnormally abundant in cancer cells and have become potent targets in cancer therapy 5 .
HSP27 belongs to the family of small heat shock proteins and is overexpressed in solid cancers, including prostate and breast cancers, as well as in haematological malignancies [6][7][8] . It contributes to tumorigenesis and resistance to chemotherapy via its proliferative and anti-apoptotic functions 9,10 . HSP27 is secreted by cancer cells and has been found in the serum of cancer patients, where its presence is often associated with a poor prognosis 11,12 . In addition, our team recently demonstrated the beneficial effect of a new-generation antisense oligonucleotide against HSP27, OGX-427, which targets the human and rodent Hsp27 translation initiation site 13,14 , in a murine model of idiopathic pulmonary fibrosis 13 . This compound is currently under clinical evaluation (phase II clinical trials) as a chemosensitizing agent in solid tumours 15 . Furthermore, we and others have shown that HSP27 was strongly expressed in samples from patients with idiopathic pulmonary and kidney tubulointerstitial fibrosis 13,16 .
Myelofibrosis (MF) is a chronic degenerative disorder associated with megakaryocytic abnormalities and progressive marrow fibrosis, in which fibrous tissues replace red bone marrow. These two distinctive features are used in patients to monitor the progression of the disease 17 . The clinical features of MF generally include constitutional symptoms, splenomegaly and progressive marrow failure, which result in reduced life expectancy. MF is mostly related to myeloproliferative neoplasms (MPN) but can also be induced following treatment with haematopoietic growth factors like thrombopoietin (TPO) 17,18 . One of the most frequent signalling pathways involved in MF pathogenesis is the Janus kinase/signal transducer and activator of transcription (JAK/ STAT) pathway. The aberrant activation of the JAK/STAT pathway may result from somatic mutations directly affecting JAK activity, excessive cytokine stimulation by factors like TPO and/or epigenetic modifications leading to abnormal gene regulation [19][20][21] .
Given the reported role of HSP27 in leukaemia and in fibrotic disorders, we hypothesized that HSP27 might be involved in MF. In this study, we show that specific inhibition of HSP27 using OGX-427 limits myelofibrosis progression in two murine models of MF 15,22,23 and affects the JAK2/STAT signalling pathway. Our data also reveal an increase of HSP27 in samples from patients with MF suggesting that HSP27 represents a new therapeutic target for MF.

Results
OGX-427 limits myelofibrosis progression in mouse models of MF. To assess the role of HSP27 in bone marrow fibrosis, we used two animal models recommended to study the establishment of the myelofibrotic features 24 : the TPO high and the JAK2V617F murine model (Figs. 1 and 2, respectively). These models reproduce some myelofibrotic traits found in human primary MF 22,24 , such as megakaryocyte hyperplasia, anaemia, extramedullary haematopoiesis, splenomegaly and myelofibrosis. Also, in both animal models there is sustained activation of JAK2/ STAT signalling induced by the persistent production of TPO 22,25 or the constitutive active JAK2 mutant (JAK2V617F) 26 (Figs. 1a and 2a).
As shown in Fig. 1b, we clearly observed an increase of HSP27 level in the serum in the TPO high model compared with wild-type mice. One month later, once the myelofibrosis-like features were established, mice were treated with the specific inhibitor of HSP27, OGX-427, using a validated non-toxic dose (10 mg kg −1 ) 14 . These conditions induced a 65% decrease in HSP27 expression in splenocytes from OGX-427-treated mice compared with those from control (CTL) mice (i.e., treated with a non-relevant antisense oligonucleotide, Fig. 1c). Although OGX-427 treatment in this model had no direct effects on blood parameters ( Supplementary Fig. 1a), it resulted in a marked reduction in spleen weight (30%) and size, a major criterion of treatment efficacy in humans (Fig. 1d, e). In agreement with this result, spleen sections revealed hyperplasia of the red pulp in TPO high mice and partial restoration of the white pulp territories in OGX-427-treated mice compared to CTL (Fig. 1f). These results strongly suggest that OGX-427 treatment enable partial restoration of the normal splenic architecture. In addition, the decrease in erythropoiesis in the bone marrow was more pronounced in control mice than in OGX-427-treated mice, whereas megakaryocytic hyperplasia was less marked in OGX-427-treated mice than in controls (Fig. 1g), suggesting a beneficial effect of OGX-427 treatment on erythropoiesis and on the reduction in megakaryocytic expansion. It is worth noting that in our ethically approved protocol to study the effect of OGX-427 on myelofibrosis development, the animals were killed around 5 months after the beginning of the disease. However, even though the protocol was not adequate to study the effect of OGX-427 on animal survival, the few animals that died before the planned killing suggested an improved survival in the mice treated with OGX-427 ( Supplementary Fig. 1b).
We next studied the effect of OGX-427 in the JAK2V617F transgenic mouse model, which has the advantage of being a better preclinical model for fibrosis studies 26 (Fig. 2). Twelve weeks after transplantation, JAK2V617F mice were treated with OGX-427 three times per week for 10 weeks (10 mg kg −1 ) or not (control) (Fig. 2a). In accordance with the results obtained in the TPO high mice model, OGX-427 treatment significantly reduced the weight of the spleen (Fig. 2b), induced a 50% decrease in HSP27 expression in the bone marrow compared with that from control mice (Fig. 2c) and limited megakaryocyte hyperplasia in the spleen and in the bone marrow (Fig. 2d). Of note, in this model, in the OGX-427-treated mice compared with controls, we observed a decrease in reticulin fibrosis in the bone marrow sections (Fig. 2e), and this was associated with a clear fall in platelet and white blood counts (i.e., back to normal levels at the end of the treatment). In contrast, these counts remained very high in the control group (Fig. 2f).
Altogether, these results suggest a beneficial effect of OGX-427 on the pathogenesis of myelofibrosis, both by limiting splenomegaly and megakaryocyte expansion, and by reducing fibrosis development.  hairpin RNA (shRNA) or OGX-427 in two leukaemic cell lines, HEL92.1.7 (erythroleukaemia cell line) and SET-2 (megakaryoblastic cell line), which bear the JAK2V617F mutation that constitutively activates the JAK2 signalling pathway. HSP27 depletion by the three different approaches affected cell proliferation induced by the constitutively activated JAK2 mutant (Fig. 3a, Supplementary Fig. 2a, b), but not apoptosis (Supplementary Fig. 2c), with a more pronounced effect on the megakaryoblastic SET-2 cell line (Fig. 3a). It is worth noting that HSP27 depletion had no effect on the proliferation of the JAK2V617F-negative cell line K562, suggesting that HSP27 plays a role in JAK2V617F-positive cell proliferation. Our rescue experiments in the shRNA HSP27-depleted HEL92.1.7 cell line confirmed the implication of HSP27 in the cell proliferation ( Supplementary Fig. 3a). An effect of HSP27 on cell proliferation was also observed on primary cells from patients by performing an erythroid colony formation assay. We found that HSP27 depletion by means of a specific shRNA induced a decrease in burst-forming units-erythroid (BFU-E) from MPN patients along with a specific effect on JAK2V617F cells ( Supplementary Fig. 3b-d).
At the molecular level, depletion of HSP27 decreased the amount of phosphorylated STAT5 (Fig. 3b). Consequently, the expression of the STAT5 target c-MYC, a modulator of proliferative signals, was reduced ( Fig. 3b and Supplementary  Fig. 2b). To investigate how HSP27 impacted phosphorylated STAT5 content, we performed both phosphorylation and dephosphorylation in vitro assays. We found that while recombinant HSP27 did not affect the phosphorylation of STAT5 by JAK2 (Fig. 4a), it did affect its de-phosphorylation by tyrosine phosphatase SHP2 (a main phosphatase involved in STAT5 tyrosine de-phosphorylation) [27][28][29] (Fig. 4b). In keeping with these results, AG490, an inhibitor of JAK2 activation 30 , induced a faster de-phosphorylation of STAT5 in HEL92.1.7 cells depleted for HSP27 than in control cells (Fig. 4c). We next performed in vitro, in cellulo and in situ experiments to determine whether HSP27 interacted with JAK2/STAT5. Using biolayer interferometry, we showed that recombinant JAK2 and STAT5A/B interacted directly with immobilized biotinylated HSP27 within the same range as Crystallin alpha B (CRYAB), a well-known partner of HSP27 used here as a positive control 31 (Fig. 5a, Supplementary Fig. 4a-f).
JAK2 or STAT5 association with HSP27 was next determined by co-immunoprecipitation experiments using endogenous (Fig. 5b, c and Supplementary Fig. 5a) or exogenous tagged proteins ( Supplementary Fig. 5b, c). In line with the biolayer interferometry results, HSP27 interacted with JAK2 and with STAT5.
We then confirmed these results by proximity ligation assay (PLA), by analysing the interaction between JAK2 or STAT5 and HSP27 within the HEL92.1.7 cells (Fig. 5d). The number of detected interaction foci per cell was significantly lower in siRNA HSP27 transfected than in control HEL92.1.7 cells (Fig. 5d). Using the same approach, we determined whether HSP27 affected the interaction between JAK2 and STAT5. As shown in Fig. 5d, depletion of HSP27 by siRNA significantly decreased the association between JAK2 and STAT5.
Altogether, our results indicate that the chaperone HSP27 is involved in the proliferative effect of JAK2/STAT5. HSP27, as previously shown with other HSP27 client proteins 32-34 , protects STAT5 from its de-phosphorylation and might also favour the molecular assembly JAK2/STAT5.

Patients with MPN-associated MF display high HSP27 levels.
On the basis of our previous in vivo and in vitro results, we determined the expression levels of HSP27 in patients with an MPN-associated MF. In this type of MF, acquired mutations (JAK2V617F, calreticulin (CALR) or MPL mutations) constitutively induce JAK2 signalling pathways.
As a first approach, we used flow cytometry to evaluate the HSP27 levels in CD34 + circulating hematopoietic progenitor cells (HPCs) isolated from the blood of MF patients and healthy donors (HDs). We observed that HSP27 levels in CD34 + HPCs from MF patients were significantly higher than those in HDs (Fig. 6a). In contrast, HSP70 and HSP90 levels within CD34 + HPC in MF patients were not different from those in HDs. Knowing that HSPs can also be secreted, we assessed the extracellular content of HSPs in serum from MF patients by enzyme-linked immunosorbent assay (ELISA). Again, significantly higher levels of HSP27 were detected in the serum of MF patients than in serum from HD as previously observed in our TPO high mouse model of myelofibrosis (Fig. 6b). High levels of extracellular HSP27 have also been reported in diseases involving chronic inflammation 35,36 , an important process in the clonal evolution of MPN 37,38 . Notably, increased HSP27 levels in HPCs and in serum from patients were detected independently of JAK2V617F, CALR or MPL mutations (Supplementary Table 1), all three mutant proteins that are indeed known to constitutively activate the JAK2/STAT pathway. HSP70 levels in patient serum were again not different from those in HDs (Fig. 6a, b). In contrast, HSP90 levels in serum samples were higher in MF patients than in HDs, highlighting the involvement of HSP90 in MPN 39,40 (Fig. 6b).
In line with the results described above, HSP27 was strongly expressed in bone marrow biopsies from MF patients, whereas only a few myeloid cells were positive for HSP90 and none for HSP70 (Fig. 6c). More precisely, HSP27 was detected in megakaryocytes and in endothelial cells (Fig. 6c). Previous studies also reported increased levels of HSP27 in endothelial cells from fibrotic tissues 41 and in MF megakaryocyte progenitors 42 . Altogether, these results suggest that HSP27 is a potential new player in MPN-associated MF.

Discussion
Myelofibrosis is a severe degenerative disorder for which no curative treatment exists, except bone marrow allograft 43 . Although JAK2 inhibitors have given rise to high expectations, clinical studies, which had beneficial effects on the symptoms, demonstrated limited effects on the progression of the disease 44,45 . Because HSP90 is known to stabilize JAK2 and therefore is a potential therapeutic target in JAK2-dependent MPN associated with MF, inhibitors of HSP90 have also been explored (currently in clinical phase II) 39,40 . In the present work, we demonstrated that another essential chaperone, HSP27, is a therapeutic target of remarkable interest.
Inhibitors of HSPs, particularly of HSP90 and HSP27, are currently being tested in clinical trials as chemo-and radiosensitizing agents, in many different cancer therapy protocols (reviewed in refs. 5,46,47 ). The rationale for targeting HSPs in cancer is provided by the fact that cancer cells have a strong need in chaperones (i.e., HSPs) for their survival, because they need to rewire their metabolism. However, HSP90 inhibitors, probably due to their toxicity, induce the compensatory expression of stress-inducible HSPs (i.e., HSP70, HSP27), which, through their cell survival function, may hamper HSP90 inhibitor effectiveness 40,48 . The HSP27 inhibitor used in this work, an antisense oligonucleotide of generation (OGX-427, in phases I and II in cancer), is a very selective inhibitor, with no or negligible OGX-427 limited the progression of bone marrow fibrosis in the JAK2V617F model, but not in the TPO high model (Fig. 2e). In accordance with anterior published work, Fedratinib (a selective JAK2 inhibitor for which clinical studies have been hindered because of its toxicity) also decreased splenomegaly without affecting fibrosis in this TPO high murine model, whereas it impaired fibrosis in the post-polycythemia vera myelofibrosis murine model 51,52 . The differences in results obtained, using different mouse models, might be explained by the fact that the process of fibrosis in the TPO high model is very rapid and  irreversible 22, 25 compared to other models. Despite the limits of this TPO high model of MF, we nonetheless succeeded to demonstrate the effectiveness of OGX-427, not only on splenomegaly, but also on extramedullar hematopoiesis. This suggests a potential implication of HSP27 in the migration of bone marrow stem cells, from the medullary niche to extramedullary foci. It has been shown that HSP27 is overexpressed in fibrotic conditions such as pulmonary or renal fibrosis 16,53,54 . For instance, we previously unravelled the overexpression of HSP27 in the pleura of patients with pulmonary fibrosis, as well as in two rat models of idiopathic pulmonary fibrosis 13 . The identification, in the present work, of HSP27 overexpression in MF patients (Fig. 6a-c), as well as in mouse models of MF (Figs. 1 and 2), again supports association between HSP27 and fibrotic pathologies. How HSP27 impacts myelofibrosis development is not yet understood, but it may involve the JAK2/STAT5 signalling pathway. Indeed, our results demonstrate that HSP27 interacts directly with JAK2 and STAT5 (Fig. 5a-d), probably favouring their assembly into a stable complex, and protecting STAT5 from its de-phosphorylation (Fig. 4b, c). Additionally, we show that HSP27 is involved in the proliferation of JAK2V617F cells and the expression of c-MYC (Fig. 3b), a downstream target of STAT5 and a known actor in fibrosis development 55,56 . Interestingly, the HSP70/90 co-chaperon STIP1 (stress-induced phosphoprotein 1) has been reported to be necessary for the control of the stability of the JAK2/HSP90/STAT3 complex 57 , suggesting that JAK2/ STAT5 activity is tightly regulated by different chaperones. Other signalling pathways might also be involved in the myelofibrotic process. In particular, it has been reported that stimulation of platelets by collagen leads to a HSP27-dependent release of platelet-derived growth factor (PDGF), a growth factor that participates in the pathogenesis of MF 58 . In addition, aberrant expression of PDGF and its receptor (PDGF receptor alpha) has been identified in bone marrow cells of patients with idiopathic MF at the advanced stages of the disease 59 .
Here, we also identify for the first time an extracellular increase of both HSP90 and HSP27 in the sera of MF patients (Fig. 6b), as well as of mice developing the disease (Fig. 1b). Although HSP90 extracellular function in MF remains to be studied, in cancer it has been reported to participate in carcinogenesis 60,61 . In the context of pathological remodelling heart, García et al. 62 Supplementary Fig. 8d. c Immunoprecipitation of endogenous STAT5 was followed by immunodetection of endogenous HSP27 and JAK2. Inputs: proteins in total cell lysates. IP IgG: immunoprecipitation with a non-relevant antibody (IgG mouse) (n=2 independent experiments). Uncropped blots presented in Supplementary Fig. 8e. d Immunofluorescence analysis of the endogenous interaction (red foci) of JAK2 (upper panel) or STAT5 (middle panel) with HSP27 visualized in situ by PLA in HEL92.1.7 cells transfected or not (NT) with a HSP27 siRNA. Nuclei are stained with DAPI. Images were taken randomly and obtained using an Axio Imager 2 at ×40 magnification and analysed using ICY software. Cells were segmented manually, and the number of interaction foci in each cell was counted using the spot detector plugin. Right panel, graphs represent quantification of the interaction of JAK2 or STAT5 with HSP27 visualized in situ by PLA. Each data point corresponds to an analysed cell, placed according to the number of detected interaction foci (lower panel). Immunofluorescence analysis of the endogenous interaction (red foci) of JAK2 with STAT5 visualized in situ by PLA in HEL92.1.7 cells transfected with a scramble (CTL) or a HSP27 siRNA (siHSP27). Nuclei are stained with DAPI. Images were obtained using an Axio Imager 2 at ×63 magnification and analysed using ICY software as in upper panel. Right panel, graphs represent quantification of the interaction of endogenous JAK2 with STAT5 visualized by PLA. P values were calculated using the Mann-Whitney test. ****P < .0001. Scale bar 10 µm collagen production as well as the canonical transforming growth factor β (TGF-β) signalling cascade. It has been demonstrated that extracellular HSP90 binds to the extracellular part of TGFβ receptor I (TGFβR1) in cardiac fibroblasts, thereby inducing the canonical pathway of collagen production. Accordingly, Hsp90 knockout mice showed a dramatic decrease in the production of collagen by fibroblasts in response to TGF-β stimulation 62 . These results highlight the existence of a functional cooperative partnership between HSP90 and TGFβRI in the fibrotic process. Since we also observed a marked increase in serum HSP90 levels in MF patients, this suggests that extracellular HSP90 might likewise play a role in the pathogenesis and/or could be used as a biomarker in MF 63 . Extracellular increase of HSP27 has already been observed in pathological conditions, such as chronic pancreatitis, pancreatic cancer or breast cancer [64][65][66] . However, in pulmonary and renal fibrosis, in which a significant increase of intracellular HSP27 was previously reported, its levels in the patient's sera have not yet been investigated 16,53,54 . The level of HSP27 in the extracellular medium is generally correlated with an increase in its intracellular concentration 67 , which is observed under extreme conditions such as very intense physical effort 68 or pathological secretion by cancer cells 11,69,70 . Accordingly, we show, here, that the high level of HSP27 in the serum of MF patients, compared to healthy donors, correlated with an increase in HSP27 intracellular levels in CD34+, megakaryocytes and endothelial cells (Fig. 6a, c). We believe that HSP27 secretion by these cells contributes to the high levels of HSP27 detected in the blood of MF patients, although we do not exclude that other types of cells may also participate to this process. In particular, it has already been shown that HSP27 can be released by human platelets under normal conditions 71 . Platelets might therefore also be involved in HSP27 secretion in the context of MF, especially as they are deregulated in this disease 42 . Altogether, the increase in the expression of HSP27 observed in CD34+ HPC, as well as in the serum from MF-associated MPN patients, suggests that HSP27 is a potential therapeutic target in this disease. To summarize, using two murine models of myelofibrosis, we show that a specific HSP27 inhibitor, OGX-427, limits the progression of myelofibrosis by (i) reducing both spleen weight and size, (ii) decreasing myeloid proliferation in the spleen and bone marrow, and by reducing megakaryocytic expansion, (iii) decreasing reticulin fibrosis and (iv) normalizing the platelet and white blood counts. We also show, for the first time, that HSP27 may chaperone JAK2/STAT5, a central signalling pathway activated in MF-associated MPN, and also in other kinds of diseases 72,73 . Altogether, our results support a key role for HSP27 in the pathophysiology of MF and highlight a potential interest of HSP27 inhibitors as a complementary approach for MFassociated MPN treatment. Western blot analysis. Protein extracts from cells and spleen tissues were prepared using a modified Laemmli buffer (5% sodium dodecyl sulphate, 10% glycerol, 32.9 mM Tris-HCl pH 6.8) supplemented with protease and phosphatase inhibitors (Roche). Then, 30 μg of proteins from each lysate were subjected to migration on 8-12% acrylamide gels and transferred on to polyvinylidene difluoride membranes (GE Healthcare Europe GmbH) in borate buffer (50 mM Tris-HCl and 50 mM borate) for 1 h and 45 min at constant voltage (48 V). The membranes were incubated with primary antibodies overnight at 4°C, then washed in Tris-buffered saline-Tween 0.1% and incubated for 1 h with horseradish peroxidase (HRP)coupled secondary antibody or beta-actin peroxidase antibody (Sigma-Aldrich). The signal was revealed using a chemiluminescent reagent (Luminata Crescendo Western HRP substrate reagent, Merck Millipore) and was detected using a Che-miDoc XRS System (Bio-Rad). Uncropped blots are presented in Supplementary  Fig. 7, 8.
For in vitro immunoprecipitation, recombinant proteins were produced using TNT Quick Coupled Transcription/Transcription System (L1170, Promega) as follows: 1 mg of template plasmid DNA was added to the reaction mixture, which was afterwards incubated at 30°C for 90 min. The in vitro translated proteins were incubated for 15 min at 37°C in buffer (Hepes pH 7.4 10 mM, NaCl 25 mM, MgCl 2 5 mM, MnCl 2 5 mM, dithiothreitol (DTT) 1 mM, protease and phosphatase inhibitors) and then subjected to immunoprecipitation for 1 h and 30 min at RT using an anti-mouse HSP27 antibody (2 µg, clone G3.1, SPA-800, Enzo Life Sciences) or a mouse IgG (serum, I5381, Sigma-Aldrich) as a negative control that were pre-incubated 1h at RT with Protein G UltraLink Resin (53132, Pierce). Protein complexes were then washed 4 times in lysis buffer and suspended in 2X laemmli buffer. After boiling, the immunoprecipitates were resolved in 8-12% SDS-PAGE and immunoblots were performed using an anti-rabbit STAT5 (#9363, Cell Signalling technologies).
In vitro transcription and translation reactions were performed using TNT Quick Coupled Transcription/Transcription System (L1170, Promega) as recommended by the supplier. Next, 1 µg of the plasmid DNA template was transcribed and the protein was translated at 30°C for 90 min.
For the de-phosphorylation assay, a phosphorylation kinase assay JAK2/STAT5 was first done in the presence of ATP as described above and the reaction was stopped by adding 20 µM of Ruxolitinib (sc-396768A, Santa-Cruz) for 5 min. Then, HSP27 (produced in reticulocyte lysate) was added or not to the samples and incubated for 10 min at 30°C. The recombinant SHP2 (590 ng, SRP0217, Sigma-Aldrich) was added or not to the samples for 12 min at 30°C and the reaction was stopped by adding 2× Laemmli buffer.
Murine models and analysis of mice. For the TPO high murine model, specific pathogen-free C57BL/6J female mice between 2 and 4 months of age were purchased from Janvier (Lyon, France) and maintained in sterile housing in accordance with the guidelines of the Ministère de la Recherche et de la Technologie (Paris, France). Rodent laboratory food and water were provided ad libitum. All experiments were granted by the ethics committee CE26 (Paris, license D94-076-11SCEA ; IGR 23-11-12 2012-061). All animal procedures were performed with inhalation anaesthesia with isoflurane (TEM, Lormont, France). To generate the TPO high model, mice were lethally irradiated (9.5 Gy, 60 Co gamma rays) and animals were injected intravenously (IV) with 1 to 5 × 10 6 transduced bone marrow cells (retrovirus MPZenTPO) 22,25 . The infection of bone marrow cells was done as follow: 4 days after 5-fluorouracil injection in mice (150 mg kg −1 administered IV), bone marrow cells (from two femurs) were co-cultivated with 1 × 10 5 MPZenTPOproducing GP/E-86 cells in media containing DMEM (20 ml), 20% foetal calf serum (FCS), 10% pokeweed mitogen-stimulated spleen cell conditioned medium (PWM-SCM), in a 100 mm tissue culture petri dish. After 4-5 days, nonadherent cells were harvested and injected into lethally irradiated mice. The mice were randomized based on blood count 30 days after transplantation and injected intraperitoneally, three times per week, with 10mg kg −1 OGX-427, or a control oligonucleotide. TPO high mice were treated for 5 weeks and euthanized 4 weeks after the last injection (Fig. 1a).
For the JAK2V617F murine model, bone marrow cells (2 × 10 6 per recipient) from 2-month-old female C57BL/6J mice (control mice, Harlan Laboratories) and SclCre;FF1;GFP mice were transplanted into 2-month-old female C57/BL6J lethally irradiated recipients (5 mice per group) in a 1:1 ratio 26 . At 12 weeks post transplantation, a first blood cell count was done to check and confirm the MPN phenotype. The animals were then divided into two groups of five mice: one group was treated with the vehicle alone (3 times per week) and the other mice were treated with the HSP27 inhibitor OGX-427 (10 mg kg −1 ; 3 times per week). The mice were treated for 10 weeks and cell blood count was performed every 3 weeks. At the end of the experiment, the mice were killed, and spleens were weighed and stored at −20°C. The female mice had free access to food and water and were kept under specific pathogen-free conditions (laboratory animal husbandry license CH-1007H). Animal experiments were done in strict compliance with Swiss laws for animal welfare and were authorized by the Swiss Cantonal Veterinary Office of Basel-Stadt, license 2581.
For the two models, haematocrit, platelet and white blood cell (WBC) counts were determined using an automated counter (Scil Vet abc Plus+, Horiba) on blood collected from the tails in EDTA tubes throughout the study. At the end of the study, spleens were weighed, measured and stored in 4% paraformaldehyde for histopathology studies or frozen at −20°C for western blot analysis. The concentration of HSP27 in serum from control mice was measured by ELISA at 3 months after the transplantation. For bone marrow analysis, femurs were collected, fixed in 4% phosphate-buffer formalin, embedded in paraffin and sectioned. Tissue sections were stained with haematoxylin and eosin for morphology analysis, which was performed in a blinded fashion. Images were organized in folders identified by letters by one person and quantified by another. A minimum of n =5-9 mice were used for each experiment, which was sufficient to provide adequate power for these experiments (inVivoStat analysis).
Patient's samples. Patient samples (serum, blood and histological section) were obtained from the University Hospital of Dijon and Gustave Roussy (France). The study was conducted in accordance with the Declaration of Helsinki with an approved written consent form for each patient (CPP ESTI: 2014/39 ; N°ID: 2014-A00968-39), and approval was obtained from the local ethics committee ESTI (license: NCT02873832).
Peripheral blood mononuclear cells (PBMCs) from peripheral blood or cord blood were collected in EDTA tubes and isolated by Ficoll-Paque density gradient centrifugation. PBMCs were immuno-stained with anti-CD34-PE-Cy7 and anti-CD45-APC antibodies in stain Buffer (BD Biosciences) for 45 min and washed. The cells were fixed and permeabilized in commercial solution (BD Biosciences) and then stained using anti-HSP FITC/Alexa 488 or anti-mouse IgG FITC control antibody for 30 min. After washing, the cells were analysed by flow cytometry. The CD45 low /CD34 high population was gated and the median fluorescence intensities of intracellular HSPs were evaluated (FACs gating strategy, Supplementary Fig. 9b).
Purification of CD34+ cells and erythroid progenitor assays. Mononuclear cells from MPN patients were obtained over a Ficoll density gradient and CD34+ cells were purified by a double-positive magnetic cell sorting system (AutoMACS, Miltenyi Biotec). CD34+ cells were cultured 2 days in a liquid culture system stimulating erythropoiesis, in the presence of 50 ng ml −1 recombinant human stem cell factor (Amgen, Thousand Oaks, CA, USA), 50 U ml −1 interleukin-3 (Novartis, Basel, Switzerland) and 3 U ml −1 erythropoietin (EPO; Orthobiotech, Paris, France). Cells were then transduced with lentiviral vector encoding shRNA-HSP27 or control vector (CTL). The CD34+/GFP+ were sorted 24 h later and 750-1000 cells were plated in duplicate in semi-solid conditions (methylcellulose) (H4230, Stem Cell Technologies) supplemented with interleukin-3, stem cell factor and EPO. BFU-E colonies were counted 14 days later and were plucked from methylcellulose for genotyping. The JAK2V617F mutational status was analysed by quantitative real-time PCR (qRT-PCR), using fluorescent competitive probes.
Immunoassays. Blood samples were centrifuged at 20°C for 15 min at 2500 × g, and serum aliquots were stored at −80°C for subsequent analysis. Since some HSPs are known to be unstable in plasma, we used serum for the analysis 75 . Serum levels of HSP70, HSP90α and HSP27 in patients and serum level of HSP27 in mice were quantified using specific enzyme-linked (ELISA) kits according to the manufacturer's protocol (human HSPs: ADI-EKS-715, ADI-EKS-895 ADI-EKS-500, (Enzo Life Sciences); mouse HSP27: SEA693Mu, (Cloud Clone Corp)). Before quantification, sera were diluted at 1:50 for HSP90, 1:4 for HSP70 and 1:10 dilution for HSP27 or 1:2 to detect mouse HSP27.
Statistics. Data are displayed as means ± s.e.m. or +s.d. Statistically significant differences between two groups was assessed using the unpaired t test with Welch's correction (for CD34 + HPCs from patients), the Mann-Whitney test or Student's t test depending on the distribution of the tested population. Similarity of variance was tested using GraphPad PRISM (San Diego, CA, USA) before the application of any statistical test. P values less than 0.05 were considered significant.
Data availability. The authors declare that the other data supporting the findings of this study are available within the paper (and its supplementary information files).