Jmjd1c is dispensable for healthy adult hematopoiesis and Jak2V617F-driven myeloproliferative disease initiation in mice

The histone demethylase JMJD1C is overexpressed in patients with myeloproliferative neoplasms (MPNs) and has been implicated in leukemic stem cell function of MLL-AF9 and HOXA9-driven leukemia. In the emerging field of histone demethylase inhibitors, JMJD1C therefore became a potential target. Depletion of Jmjd1c expression significantly reduced cytokine-independent growth in an MPN cell line, indicating a role for JMJD1C in MPN disease maintenance. Here, we investigated a potential role for the demethylase in MPN disease initiation. We introduced a Cre-inducible JAK2V617F mutation into Jmjd1c knockout mice. We show that Jmjd1c is dispensable, both for healthy hematopoiesis as well as for JAK2V617F-driven MPN disease initiation. Jmjd1c knockout mice did not show any significant changes in peripheral blood composition. Likewise, introduction of JAK2V617F into Jmjd1c-/- mice led to a similar MPN phenotype as JAK2V617F in a Jmjd1c wt background. This indicates that there is a difference between the role of JMJD1C in leukemic stem cells and in MPN. In the latter, JMJC domain-containing family members may serve redundant roles, compensating for the loss of individual proteins.


Introduction
The pathophysiology of myeloproliferative neoplasia (MPN) remains incompletely understood, despite the discovery of disease defining mutations such as JAK2 V617F and C-terminal alterations of CALR. In particular, it is not clear which effectors or pathways are required for disease initiation or maintenance in addition to aberrant JAK/STAT signaling. Elucidating additional oncogenic determinants is clinically meaningful in light of the limited therapeutic efficacy of JAK2 inhibition in these disorders. A rational search for drug combinations can be informed and accelerated by pre-clinical investigation of potential targets.
We have previously shown that the histone 3 mono and dimethyl-specific demethylase JMJD1C is overexpressed in MPN patients [1]. JMJD1C participates in an auto-regulatory loop, as it is both a target of the transcription factor NFE2, overexpressed in the large majority PLOS  of MPN patients, and also binds the NFE2 promoter, thereby enhancing NFE2 expression. Moreover, depletion of JMJD1C in JAK2 V617F -expressing BAF/3 cells significantly reduced cell proliferation and this effect was more pronounced during cytokine-independent growth. These data suggest that JAK2 V617F -mediated proliferation and growth-factor independence is at least partially dependent on the presence of JMJD1C. MPN patients are at risk of transformation to acute leukemia, a life-threatening disease exacerbation that is often refractory to treatment. JMJD1C has recently been shown to play a critical role in the survival of acute myeloid leukemia (AML) cells. Depletion of JMJD1C severely impaired proliferation of ten different AML cell lines, carrying various oncogenic fusion genes or mutations, including AML-ETO, PML-RAR, FLT3-ITD, t(3;3) with Evi-1 overexpression and JAK2 V617 among others [2]. Moreover, JMJD1C is required for leukemic stem cell self-renewal in murine models of both MLL-AF9 and HOX-A9-driven AML [3]. In these mice, a genetic knockout of JMJD1C decreased the frequency of leukemic stem cells and caused differentiation. This observation is especially intriguing in light of the finding that in this model, hematopoietic stem cells (HSCs) appeared less effected than leukemic stem cells (LSCs).
We therefore tested the hypothesis that JMJD1C is required for JAK2 V617F -driven MPN disease initiation by engineering mice conditionally expressing the active kinase in absence or presence of the histone demethylase.

Generation of Jmjd1c knockout mice
Knockout first mice (Jmjd1c tm1a(EUCOMM)Wtsi ) were purchased from EUCOMM (ID-No. 71834). These animals referred to as "Jmjd1c-k" mice, contain the tm1a allele which introduces a lacZ exon-trap ( Fig 1A). A gene trap cassette, following exon 8 of the Jmjd1c locus, contains an additional splice acceptor (SA), which should link the spliceosome to an artificial polyadenylation sequence (pA). Translation of lacZ as an independent polypeptide occurs via an internal ribosomal entry site (IRES). Furthermore, the tm1a allele contains a selection cassette allowing the expression of a neomycin resistance gene (neo) under the control of the human ß-actin promoter (Bact). Two Frt sites flank the two cassettes and allow their removal by the application of the FLP recombinase. Similarly, loxP sites surround exons 9 and 10 and represent recognition sequences for the Cre recombinase. Jmjd1c-k mice were bred with FLPe expressing mice (129S4/SvJaeSor-Gt(ROSA)26Sor tm1(FLP1)Dym /J) to remove the lacZ and neo markers, yielding mice with a floxed Jmjd1c locus, termed "Jmjd1c-p". Mating with Cre deleter mice (B6.C-Tg(CMV-cre)1Cgn/J) generated animals deleted for exons 9 and 10 of Jmjd1c, ("Jmjd1c-d"). Both, FLPe and Cre deleter mice, were kind gifts of Prof. R. Schüle, University Medical Center Freiburg.

Animal housing and protection
All experiments conducted on mice were approved by the Environment and Consumer Protection Agency of the State of Baden-Württemberg, Germany (G-17/59). The reviewing animal ethics committee consisted of lay people and animal welfare experts (veterinarians). Mice were maintained under specific pathogen-free conditions at the research mouse facility of the University Medical Center Freiburg. Lighting was adjusted to the circadian rhythm of the animals and temperature was kept between 20 and 23˚C. Mice lived in Type2Long cages, enriched by nesting material such as litter, tunnels and paper towels. Mice had permanent access to water and food (KLIBA NAFAG, Switzerland), which was changed every week or earlier if necessary. Animal health and behavior was monitored once daily by care takers and 5 days per week by research personnel. A special training in animal care and handling (FELASA B certificate) was mandatory for all staff working with mice.
The results of this study are based on 62 mice. Prior to the experiments, humane endpoints were determined to avoid pain and distress of the animals. These include local infections, decrease in body weight, large tumors, bleeding, decrease of activity, paralysis, etc. Once animals reached endpoint criteria, they were sacrificed on the same day. The phenotype of Jmjd1c d/d mice was investigated for 40 weeks. Jmjd1c k/k and Jak2 V617F Jmjd1d d/d mice were Knockout first mice (Jmjd1c-k allele) were crossed with FLPe mice to create conditional knockout mice (Jmjd1c-p allele). Further matings with Cre deleter mice were performed to excise exons 9 and 10 (Jmjd1c-d allele). Numeration of exons is shown in S1 Fig. (B) Beta-galactosidase staining of whole-mount mouse embryos at day E7.5, E9.5 and E11.5 of embryonic development. Scale bars: E7.5 = 250 μm, E9.5 = 500 μm, E11.5 = 1000 μm. (C-G) Sections showing beta-galactosidase signals during hepatogenesis (C), vasculogenesis (D), heart development (E) and myogenesis of skeletal muscles (F+G). E11.5 and E13.5 whole mouse embryos were beta-galactosidase stained with subsequent formalin-fixation, paraffin embedding and sectioning, while E17.5 were first frozen followed by beta-galactosidase staining of the cryosections. All sections were counterstained with Nuclear Fast Red. investigated for 12 weeks. Only one wild type control mouse was found dead without prior symptoms and the cause of death could not be determined. All other animals were sacrificed by carbon dioxide euthanasia followed by cervical dislocation.

Genotyping-PCR, RT-PCR and RT-qPCR
Ear punches were used to extract gDNA for PCR genotyping (S2 and S3 Figs; GeneJET Genomic DNA Purification Kit (Thermo Scientific, No. K0722)). From murine peripheral blood (PB), BM or testicular tissue both gDNA and RNA were extracted (AllPrep DNA/RNA Micro Kit (Qiagen, No. 80204)). Reverse transcription was performed using 400 ng RNA in the Taq-Man Reverse Transcription kit (Applied Biosystems, No. 4368813) and the resulting cDNA assayed for Jmjd1c exon usage by PCR (Figs 2A+2C and 3). Primer sequences are shown in S1 Table. Original, minimally cropped and adjusted gel images are shown in S4 Fig. RT-qPCR was used to determine the efficiency of gene trapping in Jmjd1c k/k mice. The gene trap should cause splicing of exon 8 to an artificial splice acceptor (SA), which is linked to a poly-adenylation site (pA). Jmjd1c mRNA in Jmjd1c k/k mice should therefore lack all exons downstream of exon 8. Quantifying the expression of exons 4/5 and exons 11/12, thus determines the efficiency of gene trapping (Fig 2A). Commercially available reagents were used for exons 4/5 (Applied Biosystems, Mm01150348_g1), exons 11/12 (Applied Biosystems, Mm01150330_m1) and the housekeeping gene murine β-2-microglobulin (Applied Biosystems, Assay on Demand). RT-qPCRs were performed in duplicates using a LightCycler 480 (Roche). Data were analyzed with the -ΔΔCT method.

Western blotting and antibodies
Murine bone marrow or testicular tissue was homogenized with a 100 μm cell strainer and isolated cells were resuspended in lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA, pH 8.0 and 1% Triton X-100 in ddH 2 O) followed by vigorous vortexing for 10 min at 4˚C. Cell debris was removed by centrifugation. Before proceeding with western blotting, protein concentrations were determined by Bradford assay (Bio-Rad, Hercules, CA, USA, No. 500-0007). Blots were probed with a primary antibody (polyclonal, rabbit) directed against Jmjd1c (Merck Millipore, No. 17-1026), subsequently stripped and re-probed for β-Actin (Sigma-Aldrich, No. A5441) to control for equal loading. Immunocomplexes were detected by enhanced chemiluminescence using an Intas Imager for visualization. Original, minimally cropped and adjusted blot images are shown in S4 Fig.

Blood draws
Blood samples from mice were taken via puncture of retrobulbar veins. A heparin-coated 10 μl capillary was pushed forward into the inner lid angle behind the ocular bulb. With soft rotational movements, some of the small blood vessels in the orbital cavity were disrupted. Depending on the weight of the mouse, 100-150 μl blood was collected in heparin-coated 300 μl Microvette tubes. After removing the capillary, an adequate pressure on the eye was used to stop continued bleeding. Suffering and distress of the animals were minimized by isoflurane narcosis and a heating mat was used to avoid hypothermia. The whole procedure including the narcosis took approximately 2-3 minutes per animal. Analysis of the complete blood count (CBC) was performed on an Advia 120 system (Siemens) and an Animal Blood Counter (Scil Vet).
E17.5 embryos were processed by cryopreservation and cryosectioning. Whole embryos were prepared for cutting by fixation in 4% paraformaldehyde (PBS based) for 30 minutes on ice. Following washes in ice-cold PBS, the samples were incubated in 30% sucrose (PBS based) for 4 hours at 4˚C, transferred to a 1:1 mix of OCT (optimal cutting temperature compound, Sakura Finetek, No. S0378) and 30% sucrose (in PBS) and incubated overnight at 4˚C. Subsequently, embryos were embedded in OCT and frozen on dry ice. 8-10 μm sections were cut, fixed in solution IIa (2% paraformaldehyde and 0.2% glutaraldehyde in washing buffer) for 5 minutes at 4˚C, washed (2mM MgCl 2 and 0.02% NP40 in PBS at RT) and stained in solution IIb (10 mM K 3 Fe(Cn) 6 , 10 mM K 4 Fe(Cn) 6 , 1 mg/ml X-gal in washing buffer) overnight at 37˚C in the dark. Slides were counter stained with Nuclear Fast Red (Sigma-Aldrich, No. N3020).

Hematoxylin and eosin (H&E) staining
For histopathological analysis, murine femur and spleen samples were fixed in 4% formalin overnight at room temperature. Decalcification of femora was performed in 10% buffered ethylene-diamine tetra-acetic acid (EDTA), pH 7.2. Organs were embedded in paraffin as previously described [8] and sections were stained with H&E.

Statistical analysis
GraphPad PRISM version 6 was used to carry out the statistical analysis. Data of two groups were analyzed using the Mann-Whitney U test. Survival analysis was conducted with the Logrank (Mantel-Cox) test. Units, properties of distribution, p-values and n-numbers are specified in the figure legends. In a few cases, data points were lost due to technical issues or due to the premature death of mice (S2 Table).

Results
The emergence of histone demethylases as novel targets in the treatment of myeloid malignancies is highlighted by the fact that several LSD1 inhibitors are already under clinical investigation [9] and JMJC domain-containing demethylase inhibiting substances are in development. While our previous results indicate a possible role for JMJD1C in the pathophysiology of MPN (1), its function in these disorders has not been investigated. Moreover, its role in healthy hematopoiesis remains incompletely understood and developmental expression of JMJD1C has not been examined.
We therefore depicted JMJD1C expression at five stages of embryogenesis, between day E7.5 and E17.5. From the EUCOMM consortium, we obtained Jmjd1c "knockout-first mice" (Fig 1A, top, k allele). This construct is predicted to result in a complete Jmjd1c knockout (ko), as a novel splice site is generated, which causes splicing of exon 8 to the inserted cassette that contains a splice acceptor site and a beta galactosidase (lacZ) gene followed by a poly-adenylation site (pA). In Jmjd1c +/k and Jmjd1c k/k mice, beta-galactosidase staining therefore accurately reflects transcription off the Jmjd1c promoter, visualizing sites of Jmjd1c expression (Fig 1B).
On day E11.5, β-galactosidase signals are present in the urogenital ridge close to the dorsal aorta and the subcardinal vein (S6 Fig) as well as in the hepatic primordium ( Fig 1C). All three represent sites of hematopoiesis during this stage of embryogenesis. Moreover, we detected strong β-galactosidase signals in the vascular system throughout day E11.5 embryos, for instance, around the hindlimb artery, in the cephalic mesenchyme, in the choroid plexus and in the mesentery (Fig 1D). These data suggest that Jmjd1c plays a previously unrecognized role in vasculogenesis.
The strongest Jmjd1c expression was found in myogenous tissues, for instance in the developing heart of E11.5, E13.5 and E17.5 embryos (Fig 1E). Moreover, in day E11.5 embryos βgalactosidase signals indicating Jmjd1c expression are present in the myotome premuscle mass, which forms the epaxial muscles among others (Fig 1F). We also detected strong Jmjd1c signals in the tongue and in the intercostal muscles of E13.5 and E17.5 embryos (Fig 1G and Table 1). Expression of JMJD1C has been shown to repress neural differentiation in human embryonic stem cells (ESCs) [8]. Accordingly, β-galactosidase signals were absent in neuronal tissues, for example in the developing cerebrum, as well as in lung and the developing intestinal tract (S6 Fig). Because we observed Jmjd1c expression at embryonic sites of hematogenesis, we investigated the effect of Jmjd1c deletion on steady-state, adult hematopoiesis. Animals homozygous for the k-allele (Jmjd1c k/k , Fig 1A), which introduces the novel splice acceptor following exon 8, should retain expression of the proximal 8 exons but no longer express exons 9 and beyond which contain the catalytic part of the protein. However, two different quantitative real-time  PCRs showed that mRNA-expression of the proximal exons 4 and 5 and the distal Jmjd1c exons 11 and 12 is equal between Jmjd1c +/+ and Jmjd1c k/k animals (Fig 2A). Further RT-PCR experiments revealed two observations. Firstly, the gene trap did not function efficiently as a substantial amount of normally spliced, wild-type mRNA remained detectable in Jmjd1c k/k animals ( Fig 2B+2C). This was due to inefficiency of the splice acceptor used in the "knockout-first" construct. Secondly, a new mRNA variant was detected that contained 115 bp of the "knockout-first" cassette, yielding a larger mRNA variant (Fig 2B+2C). In addition to the splice acceptor site, the "knockout first" cassette must therefore also contain a cryptic splice donor site, which splices to exon 9, restoring expression of all downstream exons. In effect, Jmjd1c k/k mice, express normal levels of an intact Jmjd1c mRNA. Consequently, blood values of Jmjd1c k/k mice were comparable to wt littermate controls (Fig 2D).
To generate a complete ko, we therefore crossed Jmjd1c k/k mice with FLPe mice, thereby excising the lacZ and neomycin cassettes (Jmjd1c-p allele, Fig 1A). The resulting mice were crossed with Cre deleter mice, to excise the floxed Jmjd1c exons 9 and 10. The resulting transcript contains a premature stop codon within the open reading frame and should therefore be subject to non-sense mediated mRNA decay (Jmjd1c-d allele, Fig 1A). However, by RT-PCR we were able to amplify an alternatively spliced, residual mRNA lacking exons 9 and 10 in tissue of Jmjd1c d/d mice (Fig 3A). Nonetheless, neither full length nor truncated or deleted Jmjd1c protein was detectable by western blotting in Jmjd1c d/d mice (Fig 3B), confirming that the animals are Jmjd1c deficient and residual mRNA was subjected to non-sense mediated decay.
We analyzed peripheral blood counts of Jmjd1c d/d mice at 8, 24 and 40 weeks of age (Fig 4A  +4B). While there was a trend towards lower platelet values in Jmjd1c d/d mice compared to wt controls (Fig 4A), no statistically significant differences were detectable. FACS measurements of the bone marrow (BM) showed no differences in the proportions of the myeloid, erythroid or lymphoid compartment, nor changes in the hematopoietic stem and progenitor populations (Fig 4C-4E). BM cellularity was non-significantly increased in Jmjd1c d/d compared to wt littermate controls (S7A Fig). Histopathological analysis of Jmjd1c d/d femora revealed a BM composition comparable to wt littermate controls (Fig 4F). Spleen (SPL) weight of Jmjd1c d/d mice was significantly reduced after 40 weeks (Fig 4G). Overall survival in Jmjd1c d/d mice was similar to wt littermates (Fig 4H). Therefore, absence of Jmjd1c either during fetal development or in adult animals had no detectable effect on the hematopoietic system.
Since Jmjd1c deficiency did not impact healthy hematopoiesis, we wanted to know whether Jmjd1c is required for Jak2 V617F -driven myeloproliferation. JMJD1C expression has been reported in hematopoietic cells [10] and we have demonstrated overexpression of JMJD1C in patients with MPN. (1) We were able to show in vitro that depletion of Jmjd1c by shRNA in Jak2 V617F -transduced Ba/F3 cells led to a significant decrease of cytokine-independent growth [1]. Given the fact that this model is only partly transferrable because of its artificial nature and the lymphoid background of Ba/F3 cells, we wanted to test the dependency of Jak2 V617F -driven MPN on Jmjd1c expression in vivo. Hence, we generated animals carrying a Cre-inducible Jak2 V617F allele either in a wt background or in the context of a Jmjd1c d/d knockout. At the age of seven weeks, Cre expression was provoked by pI:pC injections, initiating Jak2 V617F expression. Five weeks following induction of Jak2 V617F expression blood counts were determined and the animals sacrificed and subjected to histopathological analysis. Complete blood counts of Jak2 V617F and Jak2 V617F Jmjd1c d/d animals revealed the same Jak2 V617F -driven myeloproliferative phenotype in both genotypes (Fig 5A). No substantial differences in leukocyte or platelet counts were seen between Jak2 V617F Jmjd1c d/d mice and the Jak2 V617F control group. Only the red blood cells were significantly increased in Jak2 V617F Jmjd1c d/d animals compared to their controls (Fig 5A). This finding, however, was not supported by the presence of increased  young erythrocytes (reticulocytes, Fig 5B), suggesting that it was not due to an enhanced erythroid drive. Differential analysis of the peripheral blood by FACS measurements revealed that the proportion of mature myeloid and lymphoid cells is unaltered by deletion of Jmjd1c ( Fig  5C). BM cellularity did not differ between Jak2 V617F Jmjd1c d/d mice and Jak2 V617F single mutant mice (S7B Fig). Moreover, spleen weight of Jak2 V617F Jmjd1c d/d mice was comparable to that of Jak2 V617F mice (Fig 5D). Further analysis of the BM (Fig 5E-5G) and the spleen ( Fig  5H-5J) revealed no substantial difference between the anticipated Jak2 mutant phenotype and the Jmjd1c ko, with the exception of a reduced number of mature Mac1 + Gr1 + granulocytes in the BM in Jak2 V617F Jmjd1c d/d mice (Fig 5E). Histopathological slides of femora and spleen of Jak2 V617F Jmjd1c d/d animals showed the same myeloproliferation found in femora and spleens of control mice (Fig 5K+5L). These findings support the conclusion that Jmjd1c is dispensable for a Jak2 V617F -driven myeloproliferative disease.

Discussion
The role of epigenetic "writers", "readers" and "erasers" in the pathophysiology of human diseases, especially in neoplasias, has gained increased attention in recent years. Because of their enzymatic activity, these proteins constitute potential drug targets. Moreover, contrary to what could have been expected given the ubiquitous importance of epigenetic modifications, the first clinical trials using epigenetic drugs have revealed only moderate clinical side effects [11]. Therefore, understanding the contribution of individual epigenetic enzymes to molecular disease etiology is important to identify novel druggable pathways.
The molecular pathophysiology of myeloproliferative neoplasms is incompletely understood. Most notably, inhibition of JAK2 V617F , one of three identified driver mutations in MPN patients fails to induce molecular remissions in the majority of patients. We therefore searched for additional drug targets concentrating on epigenetic modifiers as several of the co-occurring mutations in MPN patients affect epigenetic enzymes, for example TET2, DNMT3A and ASXL1.
We and others have shown that the JAK2 V617F MPN driver mutation exerts its effect at least in part via up-regulation of the transcription factor NFE2. We have recently shown that among 60 epigenetic modifiers identified as possible downstream NFE2 effectors, the histone demethylase JMJD1C constitutes a novel NFE2 target gene. JMJD1C participates in a positive feedback loop, as it binds the NFE2 locus, thereby increasing NFE2 expression and, in turn, its own transcription. We therefore hypothesized that JMJD1C is required for disease initiation in a Jak2 V617F -driven murine model of MPN. However, our results clearly demonstrate that in the chosen model, JMJD1C is dispensable for the MPN phenotype.
Two specific limitations of our model must be considered. Firstly, in the constitutive Jmjd1c knock-out strain used, the protein is absent through embryonic development, during which, as we show here (Fig 1B), it is normally robustly expressed. As the JMJD-family of proteins contains 17 highly homologous proteins with demethylase activity, absence of a single member may be compensated by the function of related proteins. In this case, only the deletion of multiple enzymes would visibly alter the Jak2 V617F -driven murine phenotype.
Secondly, we interrogated whether Jmjd1c is necessary for Jak2 V617F -driven MPN disease initiation by introducing Jak2 V617F into a Jmjd1c-deficient background, perhaps already adapted to this deficiency. However, it is still possible that in an established MPN, de novo inhibition of JMJD1C interferes with disease maintenance. Our data showing reduced proliferation of Jak2 V617F -positive Ba/F3 cells upon JMJD1C inhibition, support this hypothesis. Given the possible redundancy of JMJC domain-containing family proteins, this hypothesis is better addressed by pharmacological inhibition of closely related enzymes than by the genetic approach targeting a specific protein that we employed. As pan-JMJC domain containing family inhibitors with acceptable toxicity profiles are not yet available, experimental evidence in murine models must await further developments.
We were encouraged to investigate our hypothesis by the published role of JMJD1C in MLL-driven AML stem cells. However, our data may also reflect a fundamental biological difference between murine models of MLL-driven AML and JAK2 V617F -driven MPN. JMJD1C and KDM3B inhibitors exert their effect mainly through induction of differentiation [12]. By nature of the diseases modeled, differentiation is heavily disturbed in MLL rearranged leukemias but it is not affected in MPN. Similarly, the therapeutic use of retinoids in APL, induction of differentiation by JMJD-inhibition may therefore hold therapeutic potential in AML but not in MPN.
There is growing evidence that JMJD1C plays an important role in myogenesis ( [13] and Fig  1), highlighted by the fact that expression of MyoD, a key regulator of myogenesis, is dependent on Jmjd1c. The demethylase decreases H3K9me 2 marks at the MyoD locus, thereby increasing its transcription [13]. Similarly, JMJD1C is a target of DPF3b, a transcriptional activator important for heart and muscle development, that associates with the BAF chromatin remodeling complex [14]. Our observation of strong embryonic Jmjd1c expression in developing muscle tissues suggests additional roles for JMJD1C in organs besides the hematopoietic system.  Fig (F) and S3 Fig (G).