JAK2V617F myeloproliferative neoplasm eradication by a novel interferon/arsenic therapy involves PML

Myeloproliferative neoplasms are partly sensitive to interferon α therapy. Dagher et al. demonstrate that arsenic trioxide sharply potentiates interferon’s ability to eliminate patients’ progenitors or disease-initiating cells from mouse models by targeting PML, an interferon target and key senescence gene.


Introduction
Myeloproliferative neoplasms (MPNs) are chronic stem cell diseases characterized by enhanced production of differentiated blood cells. Most MPN cases are characterized by mutations in JAK2 that result in the expression of the constitutive active mutant protein kinase JAK2 V617F and its downstream cytokine signaling pathways. The JAK2 V617F mutation may yield polycythemia vera (PV), essential thrombocythemia, or primary myelofibrosis (MF), the three diseases that constitute classic BCR-ABL-negative MPN. PV, essential thrombocythemia, and MF share common complications, including thrombosis, hemorrhage, and transformation to acute myeloid leukemia (Pasquier et al., 2014;Vainchenker and Kralovics, 2017). Animal models with hematopoietic cells expressing JAK2 V617F recapitulate many aspects of the human disease, in particular a PV-like disorder evolving into MF. These animal models have allowed exploration of the biology of JAK2 V617F and potential therapeutic approaches (Hasan et al., 2013;Mullally et al., 2013). Apart from phlebotomy in PV, current therapeutic approaches in MPN are based on cytoreductive therapies, JAK2 inhibitors, and IFNα (Spivak, 2019). Although the JAK inhibitor ruxolitinib provides some clinical benefits to MPN patients, it has no significant effect on disease progression and on the JAK2 V617F clone. IFNα can target the JAK2 V617F clone and yield complete molecular response in 8-20% of patients (Kiladjian et al., 2008;Verger et al., 2018). Yet the molecular mechanisms underlying IFNα therapy remain incompletely understood, precluding improvement of its efficacy (Austin et al., 2020;Hasan et al., 2013;Kiladjian et al., 2016;Mullally et al., 2013).
The PML protein is a member of the TRIM family of ubiquitin/SUMO ligases discovered through its implication in acute promyelocytic leukemia (APL). PML is potently transcriptionally activated by IFNs (Stadler et al., 1995). PML has drawn attention from cell biologists due to its ability to nucleate membrane-less organelles, PML nuclear bodies (NBs; Hsu and Kao, 2018;Lallemand-Breitenbach and de Thé, 2018). PML is a sensor for oxidative stress and is essential for induction of senescence (Niwa-Kawakita et al., 2017;Pearson et al., 2000;Vernier et al., 2011). While PML-NBs are disrupted in APL, their reformation drives the therapeutic response (Ablain et al., 2014;Lehmann-Che et al., 2014). Arsenic trioxide (ATO), a curative targeted therapy of APL, directly binds PML to enforce NB formation by promoting PML oxidation and multimerization Zhang et al., 2010;Zhu et al., 1997). PML-dependent growth arrest and senescence require NB formation (Ivanschitz et al., 2015) so that ATO can enhance some PML activities, notably in IFNα-primed cells (Quignon et al., 1998).
Here, we explored the hypothesis that PML may be a downstream effector of IFNα therapy in MPN. We found that ATO greatly enhances the ability of IFNα to eliminate JAK2 V617F MPN stem cells. This elimination is PML dependent and is associated with features of senescence. Our results lay the groundwork for IFNα+ATO-based curative approaches of MPN.

Results and discussion
IFNα or ATO enhances PML-NB formation preferentially in JAK2 V617F cells We first explored the PML-NB status of cells expressing JAK2 V617F . We observed significantly more and larger PML-NBs in CD34 + cells from JAK2 V617F MF patients compared with control CD34 + cells (Fig. 1 a). PML-NB number and staining intensity increased significantly upon IFNα and/or ATO treatment of CD34 + cells. A further significant increase of NB staining intensity was found when these cells were treated with both drugs (Lallemand-Breitenbach et al., 2001;Quignon et al., 1998;Zhu et al., 1997; Fig. 1 a). Using human isogenic UT-7 cell lines with or without exogenous JAK2 V617F , we confirmed that PML-NB formation was consistently more pronounced with JAK2 V617F (Fig.  S1 a). Mouse bone marrow (BM) Lin − c-Kit + (LK) cells harvested from Jak2 V617F knock-in (KI) mice exhibited similar PML-NB numbers as BM LK cells from control littermates. However, after ATO, IFN, or IFN+ATO treatments, Jak2 V617F LK cells consistently contained more PML-NBs than did Jak2 WT LK cells (Fig. 1 b). Western blot analyses performed on these LK cells demonstrated an increased PML level upon ex vivo IFNα treatment (Fig. 1 c). ATO triggered an increase in sumoylated forms of PML accompanied by some decrease in the un-sumoylated PML abundance (Lallemand-Breitenbach et al., 2001 Fig. 1 c). Importantly, both the basal and IFNα-induced PML levels were consistently higher in LK cells derived from Jak2 V617F KI mice compared with their WT counterparts ( Fig. 1 c). Overall, JAK2 V617F expression or IFNα exposure increases PML expression, while ATO enforces its NB formation, suggestive of increased sensitivity of mutated cells for PML-controlled processes.
ATO enhances IFNα-driven growth suppression in JAK2 V617F progenitors We first investigated the proliferation of UT-7 cells expressing or not exogenous JAK2 V617F when treated with increasing concentrations of ATO alone or together with a suboptimal concentration of IFNα (200 IU/ml; Fig. S1 b). Growth inhibition after 3 d was consistently more pronounced in the presence of JAK2 V617F (Fig. 2 a). These data were supported by a decreased proportion of cells in S phase, with an enhanced effect between ATO and IFNα (P < 0.001) in UT-7 JAK2 V617F cells only (Fig. S1 c).
We then assessed the effect of these drugs on the clonogenic potential of primary cells derived from MF or PV patients, exploring the balance between normal and mutant progenitors by genotyping of individual colonies (Fig. 2, b and c). With cells from the 10 MF patients explored, a decrease in erythroid colony number was observed upon treatment with IFNα, but this effect was considerably enhanced by the IFNα+ATO combination. Importantly, while the proportion of JAK2 V617F colonies was not altered by either IFNα or ATO alone, it was significantly decreased (P < 0.01) by the combination (Fig. 2 b). Similarly, in PV patient samples (n = 7), IFNα decreased the number of colonies in a dose-dependent manner, and the addition of ATO significantly enhanced this effect (Fig. 2 c and Fig. S1 d). Importantly, the number of JAK2 mutated colonies decreased after treatment with the combination (P = 0.01), but not with IFNα alone (Fig. 2 c). Apart from clonogenic potential, the growth of burstforming unit-erythroid progenitors was also clearly impacted, as the size of the colonies decreased with IFNα and decreased further with IFNα+ATO ( Fig. 2 d). In contrast, IFNα, ATO, and their combination had only moderate effects on the growth of hematopoietic progenitors from healthy individuals or JAK2 WT progenitors derived from MF and PV patients (Fig. 2 We then explored the anti-clonogenic activity of IFNα and ATO in mouse Jak2 V617F and Jak2 WT hematopoietic progenitors. BM cells from Jak2 V617F KI and Jak2 WT mice were seeded in methylcellulose with cytokines, IFNα, and/or ATO. The Jak2 V617F colony number decreased in response to the IFNα+ATO combination at suboptimal (200 IU) or maximal (1,000 IU) IFNα concentrations, in contrast to Jak2 WT cells (Fig. 2 e). Moreover, the number or proportion of Jak2 mutant colonies was synergistically decreased by IFNα and ATO, while no single agent had a significant effect. Collectively, ATO significantly enhances the growth-suppressive effects of IFNα on JAK2 V617F progenitors ex vivo.

IFNα+ATO clears JAK2 V617F disease-initiating cells in vivo
We next explored the potency of the IFNα+ATO combination in vivo using a Jak2 V617F mouse model (experimental design in Fig. S2), wherein long-term IFNα treatment (14 wk) is required to clear disease-initiating cells in some mice (Hasan et al., 2013). Using an optimized schedule of 5 mg/kg ATO every 2 d with or without IFNα for 8 wk, we found that ATO clearly improved IFNα efficacy to clear MPN (Fig. 3). Although ATO on its own had only modest effects on WBC counts and spleen size, its combination with IFNα significantly accelerated and amplified IFNα-driven hematological response, as assessed by the decrease in WBCs, hematocrit, and splenomegaly ( Fig. 3, a and c). Moreover, the addition of ATO significantly increased the molecular response to IFNα, as assessed on blood GFP + , Jak2 V617F granulocytes (Gr-1/CD11b), platelets (CD42), and RBC (Ter119) counts ( Fig. 3 b). Finally, at the time of sacrifice, combined IFNα+ATO treatment had dramatically decreased the percentage of lineage-negative (Lin − ), LK, Lin − Sca1 + c-Kit + (LSK), and SLAM + (signaling lymphocyte activation molecule) cells from Jak2 V617F KI origin, demonstrating that the combination is considerably more potent than IFNα alone for targeting Jak2 V617F stem cells (Fig. 3 d).
To substantiate any decline in disease-initiating cells, we transplanted whole BM cells (3 × 10 6 ) from treated mice into secondary recipient mice and monitored disease occurrence (Fig. 3 e). Most mice transplanted with BM cells from untreated (n = 15/17) or ATO-treated (n = 12/12) mice developed the disease, while ∼60% (n = 13/22; P = 0.06) of mice transplanted with BM  (upper panel) and quantification of PML-NB numbers and intensities (lower panel) in CD34 + cells from a healthy donor (control) and a JAK2 V617F MF patient, not treated (NT) or treated in vitro with ATO (1 µM) for 1 h, IFNα (1,000 IU/ml) for 24 h, or both (lower panel represents the pooled results of three patients and controls). Two-tailed unpaired from mice treated with IFNα alone became ill (Fig. 3 e). In contrast, none of the mice transplanted with BM from mice treated with the IFNα+ATO combination (n = 0/18; P = 0.0001) developed the disease, strongly suggesting that the combined treatment may have long-term efficacy in primary recipients. We thus monitored disease recurrence in primary animals after treatment discontinuation (Fig. 3 f). All the NT mice remained ill (n = 11/11), while 73% of the mice treated with IFNα alone relapsed (n = 11/15; P = 0.2) at 9 wk after treatment discontinuation. More than half of the mice (53%) treated with the combination (n = 8/15; P = 0.007) remained disease-free 9 wk after treatment discontinuation. Collectively, these results establish that ATO accelerates and amplifies the anti-neoplastic response induced by IFNα, targeting Jak2 V617F stem cells, with long-term clearing of the disease in a substantial number of animals.

PML drives IFNα+ATO responses
The IFNα+ATO combination maximizes PML-NB formation (Quignon et al., 1998). To explore any role of PML in the action of the IFNα+ATO combination, we first examined the role of PML knockdown in CD34 + cells from JAK2 V617F MF patients. After transduction with lentiviruses expressing an shRNA efficiently targeting PML (Ivanschitz et al., 2015;Niwa-Kawakita et al., 2017), CD34 + cells were sorted for GFP expression, seeded in semisolid cultures, and treated or not with IFNα and/or ATO. Remarkably, the decrease in mutant colony formation upon IFNα+ATO treatment was severely blunted by PML downregulation ( Fig. 4 a). These results strongly suggest that the ability of IFNα+ATO to target the clonogenic activity of JAK2 V617F progenitors is PML dependent.
To further explore PML contributions to IFNα or/and ATO selective elimination of JAK2 V617F hematopoietic stem cells (HSCs) in vivo, we generated Jak2 V617F /Pml −/− mice and transplanted syngenic mice with a mixture of Jak2 V617F /Pml −/− and Jak2 V617F /Pml +/+ BM cells (Fig. 4 b). 5 wk after transplantation, mice were treated with IFNα and/or ATO for 12 wk. In this experimental setup, if Pml was required for therapy response, its absence should yield a progressive survival advantage to Jak2 V617F /Pml −/− cells. In NT animals, MPN rapidly developed (Fig. 4 c). The IFNα+ATO combination or IFNα alone elicited a rapid decrease of leukocytosis, thrombocytosis, and erythrocytosis ( Fig. 4 c). A time-dependent enrichment of Jak2 V617F / Pml −/− granulocytes, platelets, and RBCs occurred in response to combined IFNα+ATO therapy ( Fig. 4 d), while this enrichment was observed for IFNα alone only in Jak2 V617F /Pml −/− RBCs. This key observation demonstrates Pml-dependent targeting of HSCs by IFNα+ATO in vivo and a partial Pml dependency for IFN response. Collectively, these results demonstrate that PML is involved in the elimination of JAK2 V617F progenitors or stem cells in response to the IFNα+ATO combination.

IFNα+ATO drives PML-dependent senescence ex vivo
In the syngenic UT-7 model, IFNα and/or ATO induced SAβ-galactosidase activity and expression of senescence-associated genes, which were significantly enhanced in JAK2 V617F -expressing cells (Fig. S3, a and b). We then compared the behavior of CD34 + hematopoietic progenitors from healthy individuals and MPN patients in a condition of megakaryocytic differentiation. Normal cells exhibited massive spontaneous senescence, while those derived from MF patients did not (Besancenot et al., 2010). Yet IFNα+ATO treatment restored senescence in MF patient cells, as assessed by SA-β-galactosidase activity and expression of senescence-associated genes (Fig. 5,a and b;and Fig. S3 c). Induction of senescence by the IFNα+ATO combination in primary cultures of progenitors from JAK2 V617F MF patients was abrogated upon lentiviral-introduced shRNA targeting PML (Fig. 5, c-e).
We similarly explored ex vivo IFNα and ATO treatments of primary BM progenitor cells derived from Jak2 V617F KI or littermate Jak2 WT mice, deficient or not for Pml. The expression of senescence-associated genes (Cdkn2a, Pai1, and Gdf15) was significantly more induced by the IFNα+ATO combination in Jak2 V617F Lin − cells than in Jak2 WT Lin − cells, and the absence of Pml blunted their induction (Fig. 5 f). Collectively, the IFNα+ATO combination preferentially induces senescence markers in JAK2 V617F primary cells, at least in part through PML.

Immunofluorescence microscopy
Human cell lines, patient-derived cells, or mouse Lin − or LK cells were treated with ATO alone, IFN alone, or both and then washed with PBS, cytospined, fixed using 4% paraformaldehyde (Thermo Fisher Scientific) for 10 min, and permeabilized with 0.5% (human) or 0.2% (mouse) Triton X-100. Human cells were labeled with a murine monoclonal Ab detecting human PML and a secondary anti-mouse Ab conjugated to Alexa Fluor 488 (Thermo Fisher Scientific;Lallemand-Breitenbach et al., 2001). Mouse monoclonal anti-mouse PML, clone 36.1-104 (MAB3738; Merck Millipore) was directly coupled to FITC fluorochrome or Alexa Fluor A594 for immunostaining. Slides were washed and then incubated in DAPI 1 µg/ml. After several washes, the slides were mounted in Vectashield H-1000 Mounting medium (Vector Laboratories, Inc.). For human samples, slides were kept at +4°C and analyzed under confocal microscope (LSM 800; Zeiss) under a 63× objective oil immersion. Z-stack images were processed with the image software ZEN (Zen 2 blue edition, 2011; Carl Zeiss Microscopy GmbH), and dots were counted and intensity measured on at least 20 cells spread over five to six fields with the image analysis software ImageJ (ImageJ2, Dec 2009). Mouse samples were imaged with a reversed confocal point-scanning microscope (SP8-Leica Microsystem) equipped with four lasers, 405, 488, 561, and 638 nm. An oil immersion, high numerical aperture (1.40) 63× objective was used for imaging. The scan protocol was preserved throughout all experiments. Two sequences were used to spectrally separate the two wavelengths, one for DAPI and the other for Alexa Fluor 594. The emission signal was collected on two HyD detectors for higher sensitivity. For the 594-nm signal, the "photon-counting" mode was used to remove any false numerical gain of signal. At least 20 cells per condition were analyzed using an ImageJ plugin developed by Collège de France microscopy core facilities.

Cell lines and primary cells
The human megakaryoblastic UT-7 11oc1 cell line was a kind gift of Dr. Komatsu (Juntendo University Graduate School of Medicine, Tokyo, Japan). These cells were lentivirally transduced either by JAK2 WT -or JAK2 V617F -expressing vectors as previously described (Besancenot et al., 2010). They were grown in DMEM-10% FBS-1% penicillin-streptomycin in the presence of erythropoietin (1 IU/ml). Human erythroleukemia cells (obtained from the American Type Culture Collection) were grown in RPMI 1640 medium supplemented with 10% FBS and antibiotics (100 mg/ml penicillin-streptomycin mix; Thermo Fisher Scientific). Human CD34 + cells were purified from peripheral blood mononuclear cells of PV/MF patients using the human CD34 MicroBead Kit from Miltenyi Biotec. Peripheral blood mononuclear cells from healthy donors were obtained at the French National Blood Service (French transfusion institution).
Mouse BM cells from Jak2 V617F KI or Jak2 WT littermate mice were used pure or mixed at 50% ratio and treated or not with IFNα 200 and 1,000 IU/ml (Miltenyi Biotech) with or without ATO 0.2 µM (As 2 O 3 , A1010). Cells were seeded at a density of 5 × 10 4 /ml in semi-solid cultures (Methocult, M3434; Stem Cell Technologies) supplemented with thrombopoietin (10 ng/ml). Triplicate experiments were performed, twice with BM cells from Jak2 WT and Jak2 V617F GFP + and once with Jak2 WT GFP + and Jak2 V617F .
Viral transductions CD34 + cells derived from MPN patients were transduced by a pLVTHM-eGFP-shPML lentiviral vector (Ivanschitz et al., 2015;Niwa-Kawakita et al., 2017). Briefly, lentiviral vectors were packaged in 293T cells by transfecting pCMV-8.91 and pMD.G (Addgene) for 2 d. The supernatants containing the lentiviruses were concentrated with RetroX solution (Takara Bio Europe). Hematopoietic cells were then spinoculated for 90 min with the lentiviral supernatant and then cultured for 2 d. On the third day, the lentiviral supernatant was removed, and the percentage of transduced cells was evaluated by the percentage of GFPpositive cells in a FACS Canto II and then sorted in a FACS Aria III (Becton Dickinson).

Western blot
Cell pellets were lysed in 2× Laemmli buffer, sonicated on ice for 15 min, and denatured for 7 min at 95°C. Samples were separated by a 4-12% polyacrylamide gel. Proteins were transferred on a nitrocellulose membrane for 2 h. The membrane was blocked with 10% milk-Tris-buffered saline-Tween 0.1% and then incubated overnight at 4°C with mouse monoclonal anti-mouse PML, clone 36.1-104 (MAB3738). After several washes, the membrane was incubated with anti-mouse secondary Ab (Jackson ImmunoResearch) for 1 h at room temperature and washed, and protein expression was revealed using a luminescent Image Analyzer (ImageQuant LAS 4000; GE Healthcare Bio-Sciences).

Cell analysis
For proliferation assays, cells were seeded in triplicates at 3 × 10 5 /ml in a 96-well plate in the presence of IFN, ATO, or both. Cells were counted at indicated time points with a TC20 counting apparatus (Bio-Rad), and cell viability was measured using trypan blue. Senescence-associated β-galactosidase (SAβ-Gal) staining was performed in vitro on cells at 10 d of treatment with a Senescence β-Galactosidase Staining Kit (9860; Cell Signaling) following the manufacturer's recommendations. The percentages of stained cells were counted in 20 fields in each condition using an optical microscope in >4,000 cells.
Quantitative PCR of gene transcripts Total RNA was isolated from cell lines or cells derived from human primary CD34 + cells after 10 d of culture. RNA was extracted using Trizol (Thermo Fisher Scientific). First-strand cDNA was synthesized using SuperScript III reverse transcription (Thermo Fisher Scientific). Quantitative RT-PCR was performed using TaqMan Fast Universal PCR Master Mix on a 7500 Fast Real-Time PCR System (Applied Biosystems). Predesigned TaqMan assays for PML, CDKN1A, PAI1, and GDF15 genes were purchased from Applied Biosystems. TBP was used as an endogenous control to calibrate the amount of target mRNA. Mouse Lin − cells were cultured during 24 h, and then total RNA was extracted using Direct-zol RNA MicroPrep (Zymo Research) and cDNA was synthesized using SuperScript IV VILO Master Mix with the ezDNase Enzyme kit (Thermo Fisher Scientific). Quantitative RT-PCR was performed in an Applied Biosystems 7500 Real-Time PCR System thermocycler using the TB Green Premix Ex Taq (Takara Bio Europe). mRNA was normalized by measuring HPRT mRNA levels. The following primers were used: Cdkn2a forward 59-GCTCTGGCTTTCGTGAACAT-39 and Cdkn2a reverse 59-TACGTGAACGTTGCCCATCA-39; Pai1 forward 59-TGGGTGGAAAGGCATACCAAA-39 and Pai1 reverse 59-AAG TAGAGGGCATTCACCAGC-39; Gdf15 forward 59-GAGCTACGG GGTCGCTTC-39 and Gdf15 reverse 59-GGGACCCCAATCTCACCT-39; and Hprt forward 59-AAGACTTGCTCGAGATGTCATGAA-39 and Hprt reverse 59-ATCCAGCAGGTCCCTTGAACCTT-39.

Study approval
Blood samples were obtained from Gustave Roussy (Villejuif, France) and Saint Louis Hospital (Paris, France), with agreement from the Comité de Protection des Personnes Ile de France IV institutional review board (agreement from US Department of Health and Human Services: n°IRB 00003835-Protocol 2015/59-NICB) and Commission Nationale de l'Informatique et des Libertés (authorization #915663). Written informed consent forms were obtained in accordance with the Declaration of Helsinki.
Animal experiments were conducted in the Gustave Roussy animal facility and approved by the Gustave Roussy review board, authorization for the use of animals for scientific purposes protocol #2016-104 and project titled "Treatment and evolution of MPN."

Statistical analysis
In vivo data are presented as mean ± SEM and were analyzed with the Dunnett method as part of one-way ANOVA with multiple comparison tests (GraphPad Prism) or with the twotailed Student's t test. In vitro data were analyzed using the Student's t test and two-way ANOVA for clonogenic assays. Error bars represent the mean ± SEM. In all figures, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Fig. S1 compares the effects of IFN and ATO on cells expressing-or not expressing-JAK2 V617F . Fig. S2 depicts the MPN mouse model used in Fig. 3. Fig. S3 compares senescence induction in control and JAK2 V617F -expressing MPN cells. Figure S3. IFN/ATO drives senescence more efficiently in the JAK2 V617F context. (a) Percentages of JAK2 WT (WT) or JAK2 V617F (VF) UT-7 cells untreated (NT) or treated with ATO (0.25 µM), IFNα (1,000 IU/ml), or both drugs for 7 d stained positively for SA-β-Gal. (b) Relative expression to untreated control of the senescence-associated genes PML, PAI1, and CDKN1A in UT-7 cells. Color code as in panel a. Error bars represent mean ± SEM of three independent experiments; unpaired two-sided Student's t test. (c) SA-β-Gal staining assay of CD34 + -derived cells from three MF patients or three healthy controls after 10 d of megakaryocytic differentiation and untreated or treated with the combination of IFNα (1,000 IU/ml) and ATO (0.25 µM). The values shown are mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001.