Phenotypic characterization of disease‐initiating stem cells in JAK2‐ or CALR‐mutated myeloproliferative neoplasms

Abstract Myeloproliferative neoplasms (MPN) are characterized by uncontrolled expansion of myeloid cells, disease‐related mutations in certain driver‐genes including JAK2, CALR, and MPL, and a substantial risk to progress to secondary acute myeloid leukemia (sAML). Although behaving as stem cell neoplasms, little is known about disease‐initiating stem cells in MPN. We established the phenotype of putative CD34+/CD38− stem cells and CD34+/CD38+ progenitor cells in MPN. A total of 111 patients with MPN suffering from polycythemia vera, essential thrombocythemia, or primary myelofibrosis (PMF) were examined. In almost all patients tested, CD34+/CD38− stem cells expressed CD33, CD44, CD47, CD52, CD97, CD99, CD105, CD117, CD123, CD133, CD184, CD243, and CD274 (PD‐L1). In patients with PMF, MPN stem cells often expressed CD25 and sometimes also CD26 in an aberrant manner. MPN stem cells did not exhibit substantial amounts of CD90, CD273 (PD‐L2), CD279 (PD‐1), CD366 (TIM‐3), CD371 (CLL‐1), or IL‐1RAP. The phenotype of CD34+/CD38− stem cells did not change profoundly during progression to sAML. The disease‐initiating capacity of putative MPN stem cells was confirmed in NSGS mice. Whereas CD34+/CD38− MPN cells engrafted in NSGS mice, no substantial engraftment was produced by CD34+/CD38+ or CD34− cells. The JAK2‐targeting drug fedratinib and the BRD4 degrader dBET6 induced apoptosis and suppressed proliferation in MPN stem cells. Together, MPN stem cells display a unique phenotype, including cytokine receptors, immune checkpoint molecules, and other clinically relevant target antigens. Phenotypic characterization of neoplastic stem cells in MPN and sAML should facilitate their enrichment and the development of stem cell‐eradicating (curative) therapies.

One approach is to employ immunological targets expressed on the surface of MPN cells, including MPN-initiating cells. [19][20][21] The concept of neoplastic stem cells (NSC) has been developed with the aim to explain cellular hierarchies and related clonal architectures in various cancer types and to improve treatment strategies by eliminating disease-initiating and -propagating cells. [22][23][24][25][26][27][28][29][30] In fact, any form of therapy is only´curative´when eliminating most or all NSC in a given patient.
The disease-propagating capability of NSC in hematopoietic neoplasms is usually demonstrated in highly immuno-deficient mice, such as nonobese diabetic severe combined immunodeficiency (NOD/SCID) mice lacking an IL-2 receptor-gamma chain, also known as NSG.
In most forms of AML, NSG mouse-engrafting leukemic stem cells (LSC) reside in a CD34 + compartment of the malignant clone. [22][23][24]30 Depending on the AML variant and mouse strain employed, AML LSC reside in a CD34 + /CD38 À and often also in a CD34 + /CD38 + fraction. 31 In the blast phase of chronic myeloid leukemia (CML), CD34 + LSC also exhibit CD38, 32 while in chronic phase CML, LSC are primarily detectable in CD34 + /CD38 À subpopulations. 25,26,33,34 Normal hematopoietic stem cells (HSC) also reside in a CD34 + /CD38 À fraction of bone marrow (BM) cells. 22,23 In contrast to normal HSC, CML LSC exhibit CD25, CD26, and IL-1RAP in an aberrant manner. [33][34][35] Cell surface antigens frequently expressed at high levels on CD34 + /CD38 À AML LSC include CD25, CD47, CD96, and CD371 (CLL-1). [36][37][38][39][40] So far, little is known about the phenotype and functional properties of disease-initiating NSC in patients with MPN. A number of previous and more recent data suggest that immature myeloid cells in MPN display CD34. [41][42][43][44][45][46][47] However, only a few attempts have been made to confirm functional stemness of these cells in a xenotransplantation model. A related problem is that it is notoriously difficult to engraft MPN cells in NSG mice. Recent data suggest that MPN cells may engraft better when these mice exhibit one or more human hematopoietic growth factors. 48 Other studies have reported that immature CD34 + MPN cells display disease-related mutations in JAK2 or CALR, thereby confirming that these cells belong to the malignant clone. 44 We have recently shown that putative CD34 + /CD38 À MPN stem cells express pSTAT5. 46 However, little is known about phenotypes and target expression profiles of NSC in ET, PV, and PMF.
The aims of the current study were to establish the phenotype and target expression profile of MPN-propagating NSC and to compare NSC phenotypes in patients with untransformed MPN with target expression profiles of LSC in patients with sAML following MPN. In our xenotransplantation experiments, we used NSGS (NSG-SGM3) mice expressing human interleukin-3 (IL-3), granulocyte-macrophage colony-stimulating factor (GM-CSF), and stem cell factor (SCF).

| Reagents and antibodies
Reagents used in this study are described in the supplement.
Fluorochrome-conjugated monoclonal antibodies (mAb) applied to detect NSC and/or molecular targets in these cells are listed in Table S1.

| Statistical analysis
Statistical analyses applied in this study are described in the supplement. Differences were considered significant when p < .05.  Table S3). In limiting dilution experiments, the calculated frequency of NSGS-engrafting MPN NSC within the CD34 + fractions ranged between 0.3% and 1.3% ( Figure S3; Table S4). We also compared engraftment of CD38 + and CD38 À subsets of MPN cells in 3 patients.
Engrafted cells were found to harbor JAK2V617F (0.2%-97%) and consisted of basophils, eosinophils, mast cells, monocytes/macrophages, neutrophils, and blast cells ( Figure 1D). Next, we applied sorted CD34 + /CD38 À and CD34 + /CD38 + MPN cells. In these experiments, only the CD34 + /CD38 À cells produced engraftment ( Figure S4A). By contrast, in sAML, both the CD34 + /CD38 À and the CD34 + /CD38 + cells engrafted in NSGS mice ( Figure S4B,C).  Table S5). In about half of the patients tested, MPN NSC expressed higher levels of CD117 compared to normal HSC ( Figure 2). Furthermore, CXCR4 (CD184) was expressed at higher levels on MPN NSC compared to normal HSC ( Figure 2). In most patients with PV or ET, NSC did not express CD25 (IL-2RA). By contrast, CD25 was expressed on NSC in about half of the patients with MF ( Figure 2; Table S5). CD25 expression levels were slightly higher in patients with post-ET/post-PV MF compared to PMF patients (Table S6). In a subset of patients, NSC expressed the thrombopoietin receptor CD110, insulin-like growth factor-1 receptor (CD221) and/or IL-1RAP ( Figure 2). Except for a slightly elevated median CD25 level on NSC in MPN patients exhibiting CALR mutations and a slightly higher median KIT level on MPN NSC in JAK2V617F+ patients, no difference in the NSC phenotype was found when comparing CALR-  Table S7). In a subset of patients, MPN stem cells displayed higher levels of CD25 and CD221 compared to MPN progenitor cells ( Figure S7). By contrast, the NSC in some patients displayed lower levels of IL-1RAP compared to MPN progenitor cells ( Figure S7).

| Expression of stem cell antigens and molecular targets on MPN NSC
In all types of MPN, CD34 + /CD38 À NSC expressed CD33, CD44, CD97, CD99, CD133, and the multidrug resistance antigen CD243 ( Figure 2; Table S5). We observed a slightly elevated level of CD133 on NSC in MPN patients exhibiting JAK2V617F compared to MPN patients exhibiting CALR mutations ( Figure S5). Unexpectedly, we found a negative correlation between the JAK2V617F allele burden and expression of CD97 and CD243 on NSC in our MPN patients  Table S5). CD34 + /CD38 + MPN progenitors expressed a similar phenotype compared to NSC ( Figure S7; Table S7). In most patients, these cells expressed higher levels of CD33 compared to NSC ( Figure S7). In a subset of patients, MPN progenitors displayed lower levels of CD26, CD52, CD90, and CD133 compared to NSC ( Figure S7).

| Expression of immune checkpoint antigens on MPN NSC
In all MPN donors, NSC displayed the "don't eat me" receptor CD47 and the immune checkpoint antigen PD-L1 (CD274) (Figure 2). In most patients, MPN NSC expressed higher levels of CD274 compared to HSC or LSC of sAML patients. In a few patients, MPN NSC expressed low amounts of CD83 and/or CD96. Other immune checkpoint antigens, including CD28, CD80, CD86, CD273, CD279, CD366 (TIM-3), or CD371 (CLL-1) were not detected on MPN NSC ( Figure S6; Table S5). MPN progenitor cells and NSC displayed a comparable profile of checkpoint antigens ( Figure S7; Table S7). In most patients, these cells expressed higher levels of CD371 (CLL-1) compared to NSC ( Figure S7). The phenotype of the JAK2V617F+ cell lines HEL and SET-2 as well as UT-7 cells (WT for CALR and CALR-mutated) is shown in Table S8. These cells also expressed several targets and immune checkpoint antigens, including PD-L1 (CD274).

| Effects of targeted drugs on expression of cytokine receptors and other target antigens on MPN NSC
To evaluate whether surface antigens that were up-or down-  Figure S17). Similarly, dBET6 was found to exert most potent apoptosis-inducing effects in MPN cell lines and in primary CD34 + /CD38 À MPN stem cells ( Figure S18).
Next, the effect of the CD33-directed drug GO and CD52-directed drug alemtuzumab were tested. In primary MPN MNC, GO suppressed proliferation in a dose-dependent manner ( Figure S19A).
Moreover, GO was found to induce apoptosis in primary MPN NSC in our in vitro experiments, whereas GO failed to induce apoptosis in CD34 + /CD38 + MPN progenitors ( Figure S19B). Next, we tested the  Table S9 and a summary of apoptosis-inducing drug effects in Table S10.

| DISCUSSION
Although MPN are considered to be stem cell-derived neoplasms, little is known about phenotypic and functional properties of MPN NSC. 46,51,52 In this study, we show that MPN-initiating NSC reside in a CD34 + /CD38 À fraction of neoplastic cells and that these MPN NSC A number of different mouse models have been examined for their ability to serve as a tool for studying engraftment of MPN NSC. [41][42][43]47,48,53 In earlier studies, NSG mice were employed, but in these mice, only little if any engraftment was found. More recently, newborn MISTRG or NSG mice were found to provide engraftment in some MPN patients. 45,47,48 However, in these models, the xenograft technique is difficult to standardize. NSGS mice have been used successfully to engraft stem cells in chronic myelomonocytic leukemia (CMML) with reproducible engraftment results. [54][55][56] Since CMML may sometimes present as a myeloproliferative disorder (proliferative type of CMML), we employed NSGS mice in our studies. We found that these mice provide suitable engraftment results for MPN NSC although engraftment levels varied among donors, ranging between 0.1% and 40%.
To define the phenotype of MPN NSC, we examined the engraftment potential of purified (sorted) subfractions. First, we depleted the CD34 + cell fraction in MPN samples and found that these stem celldepleted fractions are unable to engraft NSGS mice whereas the bulk MNC (control cells) engrafted with clonal MPN cells. Next, we confirmed that the purified, sorted, CD34 + MPN cells can engraft in NSGS mice and determined their frequency in cell dilution experiments. The calculated frequency of MPN NSC among all injected CD34 + cells was found to be 0.3%-1.3%. Compared to other myeloid neoplasms examined by us and by others, this frequency of MPN NSC is rather high. For example, the frequency of NSGSengrafting LSC among CD34 + CMML cells ranged between 0.009% and 0.04%. 56 52,59,61 Since PD-L1 expression is considered to mediate stem cell resistance, knowledge about strategies to block PD-L1 expression in MPN NSC may provide a suitable basis for the design of improved NSC-targeting treatment strategies. We also asked whether the phenotype of MPN NSC changes during progression to sAML. Indeed, pro-oncogenic alterations in the NSC compartment may play a role in transformation to sAML. 62 However, in our study, most markers exhibited by MPN NSC were also found on sAML LSC, with comparable expression levels.
Because of resistance of MPN NSC, new drug therapies are currently being developed, including novel targeted drugs and immunotherapies. In the current study, we were able to show that MPN NSC express several clinically relevant surface targets, including CD33, CD44, CD47, CD52, CD117, CD123, CD184, and CD274. Since normal HSC also display these antigens, long-term immunotherapy may still be too toxic because of HSC exhaustion. On the other hand, antibody-based therapies, like treatment with GO may be a less toxic and thus a feasible approach. In order to test the hypothesis that CD33 may be a clinically relevant stem cell target, we exposed MPN NSC to GO. In these experiments, we found that GO inhibits proliferation and induces apoptosis in MPN NSC. Moreover, we were able to show that pre-incubation of MPN NSC with GO completely disrupts their ability to engraft NSGS mice. We can of course not exclude that GO also inhibits engraftment of normal HSC in these experiments. An interesting observation was that the effects of GO in primary MPN cells appeared to be specific for NSC whereas no apoptosis-inducing effect of this drug on MPN progenitor cells were found. Another interesting observation was that fedratinib, although effective in vitro in inducing apoptosis in MPN NSC, was not able to suppress the engraftment potential of MPN NSC in our NSGS mice. This unexpected observation is best explained by the fact that fedratinib is primarily acting on proliferating cells and is washed away after short term incubation whereas GO, once bound to NSC cannot be washed away and will enter NSC to induce cell death even after cells were injected into NSGS mice. It is also worth noting that the CD52-targeting antibody alemtuzumab did not induce cell death in MPN cells in our experiments which may be due to the lower expression of CD52 on these cells. Indeed, expression levels of CD52 on LSC may correlate with responsiveness to alemtuzumab. 63 Together, our data show that MPN-initiating NSC can engraft NSGS mice and reside in a CD34 + /CD38 À cell fraction. We also show that MPN NSC display a number of surface antigens, including cyto-