Targeting Abnormal Hematopoietic Stem Cells in Chronic Myeloid Leukemia and Philadelphia Chromosome-Negative Classical Myeloproliferative Neoplasms

Myeloproliferative neoplasms (MPNs) are unique hematopoietic stem cell disorders sharing mutations that constitutively activate the signal-transduction pathways involved in haematopoiesis. They are characterized by stem cell-derived clonal myeloproliferation. The key MPNs comprise chronic myeloid leukemia (CML), polycythemia vera (PV), essential thrombocythemia (ET), and primary myelofibrosis (PMF). CML is defined by the presence of the Philadelphia (Ph) chromosome and BCR-ABL1 fusion gene. Despite effective cytoreductive agents and targeted therapy, complete CML/MPN stem cell eradication is rarely achieved. In this review article, we discuss the novel agents and combination therapy that can potentially abnormal hematopoietic stem cells in CML and MPNs and the CML/MPN stem cell-sustaining bone marrow microenvironment.


Introduction
Myeloproliferative neoplasms (MPN) are a collection of clonal hematopoietic stem cell disorders characterized by the proliferation of one of more of the hematopoietic lineages [1][2][3][4]. The major clinicopathologic entities comprise chronic myeloid leukemia (CML), polycythemia vera (PV), essential thrombocythemia (ET), and primary myelofibrosis (PMF) [5,6]. CML is defined by the presence of the Philadelphia (Ph) chromosome that results from t(9;22) (q34.1;q11.2) and formation of the constitutively expressed oncoprotein BCR-ABL1 [7,8]. Philadelphia chromosome-negative myeloproliferative neoplasms (Ph-negative MPNs) arise from a single clonal hematopoietic stem cell (HSC) leading to proliferation of more than one cell lineage, with transitional forms from one entity to another. Over 95% of PV, ET, and PMF are associated with mutually exclusive somatic driver mutations JAK2V617F, calreticulin (CALR), and myeloproliferative leukemia protein (MPL) [9][10][11][12][13]. Acquisition of somatic driver mutations leads to the development of the MPN stem cells. In addition, disease initiation and progression of Ph-negative MPNs involve the interplay between cell-intrinsic and cell-extrinsic activities. This is characterized by survival advantage of MPN stem cells over normal HSCs that is sustained by a dysregulated bone marrow niche via a positive feedback mechanism [9,14] ( Figure 1). Further acquisition of non-driver mutations then plays a pivotal role in determining disease phenotype and promoting leukemic progression [9,10,13] (Figure 2). At diagnosis, all CML patients harbor abnormal HSCs [8,15,16]. They are characterized by an unlimited potential and unrestricted ability to self-regenerate, remain quiescent, mediate BCR-ABL1-independent tyrosine kinase inhibitor (TKI)-resistance and evade the host immunity, and allowing disease initiation, development, maintenance and progression [8,15,[17][18][19][20][21][22]. The abnormal HSCs in CML are able to survive and thrive through various mechanisms such as modulation of downstream signaling pathways (e.g., JAK/STAT, PI3K/AKT/mTOR, Wnt/β-catenin, Hedgehog signalling), induction of autophagy, selective advantage in homing and engraftment in the bone marrow microenvironment (BMM), and alterations in cellular metabolism [8,15,[23][24][25][26][27][28][29]. While leukemia stem cells (LCSs) are generally CD34 + /CD38 − , the abnormal HSC populations in CML have extremely heterogeneous and unstable cell surface antigens expression, and vary greatly in terms of their leukemogenic capacity [8,17,21,30,31]. Although TKIs show some efficacy in targeting abnormal HSCs in CML, they may not be adequate for disease eradication.  The persistence of abnormal HSCs in CML and MPN has led to the development of novel therapies targeting CML or MPN stem cells. In this review, we discuss the current and emerging therapeutic options that may target CML and MPN stem cells with the aim of disease modification and eradication. The various pathways involved in the biology of CML and Ph-negative MPNs are depicted in Supplementary Materials Files S2-S7, highlighting the rationale of their therapeutic targeting.

Current Therapeutic Options in CML and Their Effects on CML Stem Cells
TKIs competitively bind to the ATP-binding site of the BCR-ABL1 to reduce abnormal phosphorylation of the dysregulated tyrosine kinase and inhibit downstream pathways and leukaemogenesis [7,32,33]. TKIs (Table 1) have modest effects on CML stem cells as single agents may be rendered ineffective in targeting CML stem cells as a result of BCR-ABL1-independent mechanisms. They may arise from mutations in epigenetic regulators (e.g., DNMT3A, EZH2, IDH1/2) [7,34] or the loss of tumour suppressor genes (e.g., TP53, PTEN, TET1/2) [17,29] and genes that code for anti-oxidant systems (e.g., FoxO, EPAS1) [17]. Quiescence of CML stem cells is another major challenge, and may contribute to TKI resistance and relapse in CML. Despite the fact that 50% of CML patients achieve treatmentfree remissions without relapse after achieving deep molecular response, most harbour residual CML stem cells [15,17,35]. Table 1. Tyrosine kinase inhibitors and their effects on CML stem cells.
Reduction in CD26, a specific CML stem cell marker, in circulation after imatinib treatment 2.

Ponatinib
Ponatinib, a third generation TKI, is indicated in CML with BCR-ABL1 T315I mutations or refractoriness to ≥2 TKIs. Steric hindrance is produced due to replacement of threonine by isoleucine at the ATP-binding site [47,48]. Presence of a carbon-carbon triple bond in ponatinib allows it to be 500-fold more potent than IM in overcoming TKI-resistance [22, [47][48][49]. Other pathways targeted by ponatinib include VEGFR, KIT, SRC, FGFR, PDGFR, FLT3, and KIT [47]. In-vivo murine models with CML stem cells that are lin − Sca-1 + c-kit + showed that ponatinib was effective in CML stem cell eradication and spleen size reduction. Thirty-percent residual BCR-ABL1 chimerism at 28 days was achieved compared to >50% in dasatinib and IM [22,47,50].

Interferon-α
IFNα was used as first-line treatment before the emergence of TKIs. It induces apoptosis of LSCs via Fas-receptors upregulation, FADD/caspase-8 pathway activation, and cytochrome-c release, leading to mitochondrial disruption and cellular apoptosis independent of anti-apoptotic B-cell lymphoma 2 (Bcl-2), cell-cycle arrest and tumour-suppressor p53 [54][55][56]. IFNα also restores normal function of the dysregulated BMM through β1integrin for cellular differentiation and elimination of the protective barrier established for LSC quiescence [54,57,58]. IFN-α-mediated increase in expression of major histocompatibility complex (MHC) class I molecules and tumour-associated antigens cause reactivation of CTL and prompt CTL-mediated cytotoxicity against LSCs [54,55]. The 5-year survival rate of IFNα was 57% as shown in a meta-analysis of 7 data sets of randomized trials consisting of 1,554 patients [54,59]. In another study using IFNα monotherapy, the 10-year survival rate was 72%, where 46% remained in CCyR [55,60]. These highlight the potential re-emergence of IFNα for LSC elimination, where clinical trials using IFNα alone or in combination with other TKIs showed promising results for TFR (ClinicalTrials.gov identifiers: NCT02001818, NCT01657604, NCT03117816, NCT03831776, NCT04126681, NCT01316250, NCT02381379, and NCT00452023).

IFNα
A major significance of Peg-IFNα-2a is its ability to target MPN stem cells and reduce mutant allele burden in MPN [61][62][63][64][65][66][67][68]. Sustained molecular, haematological response and regression of BM fibrosis were seen in some patients after discontinuation of Peg-IFNα-2a, indicating the eradication of MPN stem cells [65,69] (Table 2). Interestingly, the effect of Peg-IFNα-2a on JAK2V617F+ stem cells was greater than that on CALR-mutated stem cells, with no difference in hematological response [70,71]. This is due to the phosphorylation and activation of JAK1-STAT1 pathway in JAK2V617F cells, but not in CALR-mutated cells, resulting in JAK2V617F-positive cells priming towards Peg-IFNα-2a [70]. There is a paucity of data suggesting that Peg-IFNα-2a targeting MPL-mutated clones could be due to the low frequency of MPL mutations in MPN. Faster response in homozygous JAK2V617F clones than that of heterozygous clones.

3.
Variation in responses in patients with additional molecular mutations [73] Ropeginterferon α-2b PV 1. Preferential inhibition and reduction of JAK2V617F-mutated primary hematopoietic progenitors with sparing of JAK2-wild type cells.

Allogeneic HSCT
Allo-HSCT is able to overcome high molecular risks (HMR) mutations, and its recipients usually harbor additional molecular mutations. These additional molecular mutations, including those conferring a poor prognosis in MPN (ASXL1, EZH2, SRSF2, IDH1/2, TP53) generally do not affect relapse-free survival (RFS) and OS in patients receiving allo-HSCT [87]. However, discrepancies are seen in terms of relapse risks. The ASXL1 mutation, which accounts for >90% in intermediate-2 and high-risk MF patients, is found to be associated with higher relapse risks [88]. In a study assessing the outcome of allo-HSCT in MPL-mutated PMF and secondary myelofibrosis (SMF), the only relapsed patient harbored ASXL1 and EZH2 mutation [89]. Meanwhile, some post-transplant ASXL1-mutated patients may die without relapsing [87].

JAK2 Inhibition
JAK2 mediates cytokine-mediated signaling in CML cells. It leads to uncontrolled STAT3/5 phosphorylation by directly binding to the SH2 domain of BCR-ABL1, which is stabilized by Abelson helper integration site 1 (AHI-1), an oncogenic adaptor for LSC survival. [18,30,131,132]. In LSCs, induced expression of MPL enhances JAK/STAT signaling to trigger ROS formation and subsequent clonal evolution, contributing to stemness and TKI-resistance [8,18,30,131]. However, LSCs remain sensitive to JAK2 inhibitors such as ruxolitinib. Preclinical studies showed that combining ruxolitinib with the CML-specific TKIs eliminated CD34 + /CD38 − stem cells with no effects on normal HSCs in-vitro, and reduced CD34 + cell engraftment to the BM in-vivo [131]. Sweet et al. showed that 33% of patients had ≥1 log reduction in BCR-ABL1 transcripts and 44% achieved MR4.5 when co-treated with nilotinib in a phase I trial [133]. Another phase I trial with nilotinib demonstrated that 40% of patients had molecularly undetectable BCR-ABL1 transcripts over 6 months [134]. A phase II trial using ruxolitinib alone showed 60% ORR, where 33% observed clinical benefit in one or more categories: platelet count improvement, hemoglobin improvement, ≥50% reduction in spleen size and ≥50% reduction in symptoms [135]. A phase I/II trial in combining ruxolitinib with CML-specific TKIs showed achievement of CCyR in 87.5% and MMR in 37.5% of patients [136].

Targeting PPARγ/STAT5/HIF2α Axis
STAT5 activation leads to the induction of hypoxia inducible factor-2α (HIF-2α)/CITED pathway for adaptation in low oxygen levels of the BMM to maintain LSC dormancy and self-renewal potential [137][138][139][140]. PPARγ, a negative regulator of the STAT5/HIF-2α/CITED pathway inhibits adhesion of LSCs to the extracellular matrix and drives apoptosis [8,137,140,141]. Preclinical studies demonstrated that thiazolidinediones (PPARγ agonists) upregulate matrix metalloproteinase-9 (MMP-9) and MMP-2 to inhibit LSC invasion and adhesion to the BMM. They also activate caspase-3 for LSC apoptosis [8,137,140,141]. Other findings include upregulation of PPARα ligands e.g., clofibrate and enhanced expression of human organic cation transporter 1 (hOCT1) via WY-12643, which increase cellular uptake of TKIs to promote TKI-mediated apoptosis [139,141]. Preliminary clinical studies in 3 CML patients showed that combined use of pioglitazone and IM accomplished sustained complete molecular remission for up to 4.7 years in all patients, even after pioglitazone withdrawal [141]. Phase II ACTIM trial showed that IM and pioglitazone had no drug interactions, yet their combination achieved MR4.5 at 12 months in 56% of patients [138]. Novel STAT3 inhibitor BP-5087, derived from SF-1-066, demonstrated 10-fold greater potency in reducing STAT3 phosphorylation and translocation, inhibiting survival of TKI-resistant CML cells and LSCs in preclinical studies [18,142]. Combination with TKIs showed dramatic increase in effects, whereas monotherapy of either was evidently inferior [142]. However, STAT3/5 inhibition may be less effective than JAK inhibition as other STATs may compensate for STAT3/5 loss [143].

Activation of Promyelocytic Leukaemia-Nuclear Bodies (PML-NB)
Promyelocytic leukaemia (PML) forms PML-NBs to repair DNA double-strand breaks (DSBs), maintain telomere homeostasis and maintain normal HSC asymmetric division through the PML/PPAR/FAO pathway [17,145,146]. Preclinical studies showed that PML upregulation in mesenchymal stromal cells upregulated inflammatory cytokines (IL-6/IL-6R and CXCL1/CXCR2), which are crucial for the maintenance in the BMM and TKIresistance of CML stem cells [146]. Arsenic trioxide (ATO) was used as a first-line treatment for CML before the development of TKIs, but preclinical studies showed limited effectiveness in targeting CML stem cells as a single agent [147][148][149]. However, combination with TKIs showed LSC targeting, downregulation of VEGFR and angiogenesis, upregulation of NKG2D ligands to induce NK-cell mediated cytotoxicity, growth arrest, inhibition of RAS/MAPK and PI3K/AKT pathways, and apoptosis via extrinsic pathways (caspase-8/-10, TNFR1) and intrinsic (BAX) pathways and the induction of ER stress [147][148][149]. ATO/IFNα combination therapy demonstrated superior in-vivo and in vitro results compared to ATO/TKIs, where it induced cell-cycling of dormant LSCs and inhibited the Hh pathway, hence, leading to autophagy-induced cell death [145]. The established ability of ATO/IFNα to overcome TKI-resistance and abolish CML stem cells in preclinical studies [145,150] has led to phase I clinical trials [151]. In a cohort of eight patients, decrease in BCR-ABL1 fusion transcript was seen in 100% and 87.5% patients after trial and 12 months after trial respectively. MR4.5 or above was achieved in 87.5% and 55.6% patients immediately after study and 12 months later, respectively [151].

Targeting the CML Stem Cell Microenviroment, Survival and Self-Renewal
Normal HSCs interact with endothelial cells, neural cells, osteoclasts, mesenchymal stromal cells and osteoblasts in the BMM [17,25,152,153]. Selectins, integrins, and CD44 expressions are required for HSC engraftment and adhesion between fibronectin on the extracellular matrix and CD106 (VCAM-I) on the BM endothelium [23,25, [152][153][154]. HSC rolling and homing is mediated by interaction between constitutively expressed E-and Pselectins and VLA4, where SDF1 and its receptor CXCR4 acts as a chemo-attractant through β1/2− integrins and SDF1 for stable engraftment [23,25, [152][153][154][155][156]. BCR-ABL1 impairs the SDF1/CXCR4 axis in normal HSCs but upregulates it in CML stem cells, conferring selective homing and survival in the BM niche [17,23,152,153,157,158]. In addition, CML stem cells have defective β1-integrin levels (VLA4 or VLA5), allowing redistribution and mobilization into the PB and other organs, e.g., spleen with the potential of uncontrolled extramedullary myeloproliferation and local LSC reservoirs [152,154,157]. CML stem cells alter extrinsic factors and upregulate expression of CD44 + and E-selectin to promote prominent BMM changes such as marrow fibrosis for exclusive stem cell engraftment and dormancy, offering protection from drug-targeting [25,152,154,158-162].

Dipeptidyl-Peptidase (DPP-4) Inhibition
DPP-4 (CD26) is an overtly expressed protease on LSC surface, where it cleaves the SDF1/CXCR4 axis to facilitate LSC mobilization into the PB independent of niche regulations [16,17]. TKIs decrease CD26 + LSCs but levels dramatically increase following resistance or relapse [16,17,137]. CD26 is not expressed on normal HSCs, suggesting that it may be a marker for LSCs as concentrations correlate with white blood cell (WBC) count [16]. DPP-4 inhibitors (gliptins) normalize the dysregulated SDF1/CXCR4 axis to restore and promote homing of LSCs [16,36]. Interestingly, Willmann et al. demonstrated that single agent nilotinib could inhibit engraftment and induce apoptosis of LSCs, while neither vildagliptin nor imatinib addition exhibited these effects [36]. Combination of nilotinib with vildagliptin also did not produce cooperative results, suggesting insignificant effects of co-administration [36]. However, vildagliptin alone reduced disease expansion through limiting LSC mobilization [16,36]. In samples of two nilotinib-pretreated CML patients with diabetes mellitus using gliptins for diabetic control, BCR-ABL1 transcripts were near undetectable or undetectable [16].

E-Selectin Antagonist
Uproleselan (GMI-1271) is an E-selectin inhibitor which dislocates homed LSC from the BM niche into PB for cellular differentiation [154,163]. Promising phase III study results in acute myeloid leukaemia (AML) for LSC eradication [163] has led to preclinical studies in CML. In vitro studies demonstrated cell cycle progression via upregulated CDK6 (cell cycle promotor) and downregulated p16 (cell cycle inhibitor), leading to an increase in cells in G-phase and increase G 2 /S/M phase when used as monotherapy or in combination with IM [152,154]. Reduced CD44 + expression via the Scl/Tal1 pathway, increased CML stem cell cycling, and restoration of TKI-sensitivity were also noted [152,154]. Murine models showed depletion of LSC and BCR-ABL1 + cells, spleen size reduction, impaired LSC engraftment to the BM and spleen, and improved OS [152,154].

Targeting SDF1/CXCR4/CXCR7 Axis
Preclinical studies showed that disruption of the SDF1/CXCR4/CXCR7 axis of mesenchymal stromal cells reduced EZH2 expression [164], increased self-renewal capacity in LSCs and the ability to override TKI-resistance with no effect on osteoprogenitor cells, mesenchymal stromal cells and normal HSCs [164][165][166]. NOX-A12, a pegylated Spiegelmer, inhibits SDF1 and antagonizes the SDF1-CXCR4 or -CXCR7 interactions to inhibit LSC homing and causes TKI-sensitization [167]. In-vitro studies showed enhanced abolishment of SDF1-mediated migration in BCR-ABL1 + cells and induction of apoptosis when combination with imatinib was used (p < 0.00005) [168]. In-vivo studies showed eradication of FLT3-ITD + cells and inhibition of SDF1-mediated migration of FLT3-ITD + cells [168]. Plerixafor, an allosteric CXCR7 agonist and CXCR4 antagonist/partial agonist, is clinically used for stem cell mobilization in HSCT in multiple myeloma and non-Hodgkin lymphoma [165,168,169]. Its use in in vitro studies with K562 and KU812 cell lines showed reduction of drug-resistance, cellular migration and adhesion to BMM and sensitization to TKI [165]. Plerixafor in in-vivo murine models mobilized LSCs to the PB, potentiating TKIinduced tumor bulk eradication [165]. However, Agarwal et al. presented contradicting in-vivo results, which demonstrated that TKI plus plerixafor led to stem cell infiltration of the central nervous system (CNS) and subsequent development of neurological deficits [170].

Targeting Wnt/β-Catenin Signalling
Porcupine (PORCN)-dependent acetylation of Wnt ligands is essential in Wnt/βcatenin signalling for maintenance of cellular functions [17,178,179]. BCR-ABL1 drives constitutive secretion of Wnt-ligands and overexpression of frizzled-4 (FZD4) receptors to promote nuclear transduction and stabilization of β-catenin, mediating TKI-resistance [178][179][180]. Riether et al. proposed that it might be induced by prolonged TKI exposure as TKI-therapy depleted miR29 and amplified CD70 expression, leading to CD27-mediated Wnt activation for LSC quiescence and therapy resistance [181]. In an in-vivo study using transgenic murine models with CD34 + and c-kit cells, potent PORCN inhibitor WNT974 in combination with nilotinib was efficacious in reducing neutrophils, white blood cells and myeloid cells in PB, with eradication of CML stem cells and other progenitors in the BM and spleen [18,178,179]. Mice treated with nilotinib monotherapy died after 30 days while nilotinib plus WNT974-treated mice had prolonged survival with prominent suppression of c-Myc, cyclin-D1 and Axin-2 expression [18,178]. C82, a β-catenin inhibitor, downregulated CD44, c-Myc, STAT5, survivin, and CRKL in T315I and E255V mutant cell lines, eliminating LSCs in-vitro and in-vivo [27].

Sirtuin 1 (SIRT1) Inhibition
SIRT1, a NAD + -dependent deacetylase, is a potent suppressor of tumour suppressor p53 found to be overly expressed in CML stem cells [18,221,226]. It activates PGC-1α to promote mitochondrial DNA replication to maintain the bioenergetic demands of LSCs [226]. SIRT1 deletion in vitro and in vivo demonstrated downregulation of mitochondrial genes and upregulation of p53 acetylation in LSCs and progenitor cells [221,226]. While TKI treatment did not affect mitochondrial respiration [226], combination with SIRT inhibitors restored sensitivity to TKIs and subsequent TKI-mediated apoptosis [221,226].

Human Double Minute 2 Protein (HDM2) Inhibition
HDM2, another p53 negative regulator, inhibits TP53 transcription via binding to its transactivation domain [18,227,228]. Hyperactivity of CML LSCs leads to p53 proteasomal degradation and evasion of apoptosis. DS-5272, an HDM2 antagonist, restored TKI sensitivity via p53 reactivation and induction of NOXA, leading to silencing of antiapoptotic Mcl-1 [229]. Combination with TKIs or BET inhibitors suppressed downstream Myc-related pathways and upregulated p53, NOXA and BAXA, reducing the threshold for TKI-mediated apoptosis [229]. In vitro and in vivo results demonstrated high selectivity and near complete eradication of LSCs [229]. MI-219 directly stabilized and reactivated p53, reduced CD44 + for LSC homing and engraftment, and depleted important genes for LSC self-renewal (e.g., JARID2, PRDM16) both in vitro and in vivo [18,227]. MI-219 had limited effects on normal HSCs, and it upregulated IFNAR1 to drive LSCs into the cell cycle and exhaust them [227].

Targeting Autophagy in CML Stem Cells
Autophagy is the stress-induced formation of autophagosome for recycling and degradation of damaged and/or aged cytoplasmic components to sustain bioenergetic and nutritional demands [24,29, [230][231][232]. A metabolic shift in LSCs results in increased glucose influx, pyruvate shuttling, glycolysis, anaplerosis, oxidative phosphorylation, and ROS overload (Warburg effect) for survival and maintenance of stemness [8,29,171,232]. Moreover, the upregulation of Beclin-1 is essential in autophagic flux [29, 230,231], acting as a protective mechanism to avoid oxidative stress and apoptosis for the maintenance of stemness [24,29,230-232].

Immunotherapeutic Targeting of CML Stem Cell Targeting PD-1/PD-L1 Axis
The PD-1/PD-L1 axis is responsible for self-tolerance [44,239]. CML induces IFNγmediated PD-L1 expression for LSCs to aid evasion of CTL-cytotoxicity and recruitment of MDSCs and regulatory T cells for immune-evasion. Preclinical studies showed that T-cell immunotherapy with PD-1 inhibition eliminated LSCs [239]. Nivolumab, a monoclonal IgG4 antibody (Ab) against PD-1, was used on an 82-year-old man, with the ability to achieve undetectable BCR-ABL1 transcripts as a single agent [240]. Results of phase I clinical trials with dasatinib are pending (ClinicalTrials.gov identifier: NCT02011945). Avelumab [241], a monoclonal IgG1 Ab against PD-L1, is currently in clinical trials with various TKIs (ClinicalTrials.gov identifier: NCT02767063).
Twenty-five percent reduction of JAK2V617F mutant allele burden is seen in 1 patient 3.
Reduction of reticulin fibre content from grade 1 to 0.5 at week 24; the effect is restricted in patients not treated with hydroxyurea. The mechanism is unknown [256] ActRIIA (luspatercept) DIPSS int-1, int-2, high risk MF  Decrease in symptom burden: 32% 3.
Inhibition of marrow fibrosis 4.
Primary outcome: efficacy and clinical activity in MF 2.
Terminated due to serious adverse effects in 75% patients [261] DIPSS

Heat Shock Protein (HSP) Inhibition
HSP is a family of ATP-dependent, cytoprotective, stress-response protein chaperones that bind and stabilize client proteins in their functional active form, thus maintaining survival advantage of MPN cells. [248,[272][273][274][275]. HSP90 chaperones JAK2 while HSP27 modulates STAT5 phosphorylation, making them therapeutic targets in MPN [273]. HSP90 inhibitor disrupts association between HSP90 and JAK2, leading to JAK2 misfolding and degradation by 26S proteasome [272,275]. In vitro and in vivo studies have demonstrated that HSP90 inhibitor (PU-H71) inhibited, and even degraded JAK2, resulting in abrogation of downstream signaling pathway (e.g., STAT3, STAT5, MAPK). A reduction of MPLW515L allele burden, EMH, and normalization of blood counts were seen in PU-H71-treated murine models [276]. However, preclinical study using another HSP90 inhibitor AUY922 showed a rapid elevation of JAK2V617F level upon AUY922 termination [275]. A phase 2 study also showed that AUY922 failed to demonstrate consistent reduction of JAK2V617F mutant allele burden. This suggests that HSP90-mediated JAK/STAT pathway inhibition may be short-lived [248] (Table 3). An important finding is that the combination of JAK inhibitor (TG101209) with AUY922 acted synergistically to induce apoptosis in primary MF CD34 + cells [273][274][275]. The pro-apoptotic effect was also seen in JAK1/2 inhibitor-resistant cells [275], suggesting a possible solution to overcome ruxolitinib resistance. An emerging role of HSP27 inhibitor (KNK437) is also displayed by its synergistic action with ruxolitinib in JAK2V617F cell lines and PV patient cells. In murine models, HSP27 inhibitor (OGX-427, Apartosen) reduced splenomegaly, BM fibrosis and normalized counts, reflecting the therapeutic potential of HSP27 inhibitors [273].

Poly-ADP-Ribose Polymerase (PARP) Inhibition
PARP is a protein involved in DNA repair. It maintains MPN cell survival alongside two other DNA repair mechanisms: BRCA1/2-mediated homologous recombination repair (HRR) and DNA-dependent protein kinase, catalytic subunit-mediated non-homologous end-joining (D-NHEJ) [277]. ROS levels are found to be elevated in MPN LSCs, predisposing cells to toxic DNA DSBs [277,278], which activate DNA repair by PARP via the recruitment of repair proteins [279]. The significance of PARP inhibitors is revealed when used in combination with ruxolitinib [277]. Ruxolitinib downregulates important molecules in HRR and D-NHEJ in cell lines of all three driver mutations and sensitizes both proliferating and quiescent MPN stem cells to PARP inhibitors. Preclinical studies demonstrated promising synergistic effects of PARP inhibitors (Olaparib and BMN673) with ruxolitinib in eradicating MPN stem cells by apoptosis [277].

CD123 Targeting
Interleukin-3 receptor (IL-3R) consists of an alpha chain (CD123) and a common beta chain (CD131) [86,280]. IL-3, a cytokine released by activated T-lymphocytes, binds to CD123 which dimerizes with CD131 to trigger downstream JAK2 signalling pathway [280,281]. CD123 is overexpressed in CD34 + /CD38 − LSCs in AML, but not in normal HSCs [282]. Overexpression of CD123 is also found in some MF patients, especially in patients with monocytosis, which confers a poor prognosis [249,281]. Tagraxofusp is a recombinant protein genetically engineered from the fusion of IL-3 to the catalytic and translocation domains of diphtheria toxin [86,281]. Upon binding to CD123 and internalization into MPN LSCs, the catalytic domain of diphtheria toxin is cleaved. This inactivates elongation factor 2 (EF2), which is responsible for protein synthesis. Thus, the apoptosis of LSC is driven [280]. Early phase clinical trials performed to evaluate the efficacy of tagraxofusp in advanced MF patients showed promising results [249,283] (Table 3).

Proviral Integration Site for Moloney Murine Leukemia Virus (PIM) Kinase Inhibition
PIM kinase is a family of anti-apoptotic serine/threonine proto-oncogenes that are transcriptionally activated by JAK/STAT signalling [272,[284][285][286]. In MPN, continuous activation of JAK/STAT pathway is observed, making PIM kinase a potential therapeutic target [272,286]. Although the efficacy of PIM kinase inhibitor monotherapy is limited [285], it acts synergistically with ruxolitinib to suppress JAK2 signalling effectively [285,286]. In MPN, pro-apoptotic BAD protein is phosphorylated and inhibited by PIM kinase, prolonging MPN cell survival. Therefore, PIM kinase inhibitor (AZD1208) liberates BAD protein to induce JAK2V617F cellular apoptosis with ruxolitinib [285]. Enhanced cleavage of DNA repair enzyme PARP was also demonstrated to promote apoptosis in a preclinical study [286]. Importantly, preclinical studies showed that PIM kinase inhibitor resensitized ruxolitinib-resistant cell lines to apoptosis [285], while combination treatment prevented disease progression of myeloproliferation and splenomegaly in murine models [286].

Bcl-xL Inhibition
Bcl-2 family includes anti-apoptotic proteins (e.g., Bcl-xL, Bcl-2) and pro-apoptotic proteins (e.g., BAX, Bcl-2 homologous antagonist killer (BAK)). Bcl-xL heterodimerizes with BAX and BAK to exert anti-apoptotic effect [290]. In MPN, Bcl-xL is overexpressed with highest level displayed in MF, followed by PV then ET regardless of JAK2V617F mutation status [290]. Significance of Bcl-xL inhibitor is seen in the cotreatment with ruxolitinib. Invitro studies showed ABT-737, a BH3 mimetic inhibitor that inhibits both Bcl-xL and Bcl2, acted synergistically with ruxolitinib in driving cellular apoptosis [290]. It is also important to note that Bcl-xL inhibitor is a possible treatment to overcome ruxolitinib resistance. The activation of RAS and its downstream pathways inhibit pro-apoptotic Bcl-2-antagonist of cell death (BAD). Hence, BAD cannot bind and inhibit Bcl-xL, contributing to JAK2 inhibitor resistance [291]. Resensitization in JAK2 inhibitor-resistant cells was manifested by co-treatment of JAK2 inhibitor and ABT-737 in preclinical study [292]. Phase 2 clinical study has demonstrated promising results with combination treatment [293].

Lysine Specific Demethylase-1 (LSD-1) Inhibition
LSD-1 is an epigenetic enzyme which demethylates lysine residue on histone H3 to sustain MPN LSC self-renewal [272,294]. In MPN, LSD1 is overexpressed and accounts for 58% of MF patients [272]. An LSD-1 inhibitor, bomedemstat (IMG-7289), is shown to markedly decrease mutant allele burden by TP53 activation and cell cycle arrest [294]. IMG-7289 restores TP53 methylation, leading to an increase in PUMA to induce apoptosis. Meanwhile, anti-apoptotic Bcl-xL is suppressed by TP53, further facilitating apoptosis and reducing mutant allele burden in murine models. Other promising effects include the reduction of BM fibrosis, EMH-mediated splenomegaly and inflammation, as well as the normalization of blood counts. Furthermore, ruxolitinib plus IMG-7289 were shown to eliminate MPN stem cells in murine models, encouraging further investigations [294]. Clinical trials of IMG-7289 are currently underway (Table 3).

Targeting the MPN Stem Cell Niche and Marrow Microenvironment
The BM niche is essential for sustaining self-renewal of MPN stem cells, hence, providing them with survival advantages over normal HSCs via generation of ROS proinflammatory cytokines [14,302]. In view of the positive feedback loop between BM niche and MPN stem cells, various novel agents are developed.

Aurora Kinase A (AURKA) Inhibition
Malignant, atypical megakaryocytes in MPN suppress the expression of GATA binding protein 1 (GATA1), which is responsible for megakaryocyte differentiation and maturation [259,260,310]. MLN8237, an AURKA inhibitor, showed efficacy in enhancing GATA1 expression, promoting megakaryocyte differentiation and polyploidization, as well as reducing BM fibrosis in animal models [310]. Additionally, MLN8237 and ruxolitinib produced synergistic effects to eliminate BM fibrosis and reduced burden of immature megakaryocytes [310]. Clinical investigations of alisertib, another AURKA inhibitor, demonstrated promising results [259,260] (Table 3).

Antifibrotic Therapy
The pentraxin family consists of C-reactive protein (CRP/PTX1), pentraxin-2 (serum amyloid P, SAP), and pentraxin-3. SAP is a 125-kD protein synthesized by hepatocytes to inhibit differentiation from monocytes to fibrocytes and further fibrocyte proliferation [311]. Besides the release of growth factors from atypical megakaryocytes, neoplastic fibrocytes also play a vital role in inducing BM fibrosis. Analysis via quantitative allele-specific PCR showed the presence of JAK2V617F and CALR mutations in fibrocytes, but not MSCs, suggesting that these fibrocytes were derived from a malignant clone [312]. PRM-151 is a recombinant SAP which inhibits PMF fibrocyte differentiation in BM and spleen to slow down development of fibrosis [312]. Due to the unique activity of PRM-151, it is being tested in clinical trials in combination with ruxolitinib [313,314].

Targeting PD-1/PD-L1 Pathway
An elevated expression of PD-1/PD-L1 has been observed in all 3 types of classical Ph-negative MPNs [273,315,316]. In MPN, JAK2V617F increases phosphorylation of STAT3 and STAT5 to promote PD-L1 expression mainly on the surfaces of monocytes, MDSCs, megakaryocytes and platelets [317]. In preclinical studies, a positive feedback was shown in MDSCs, in which it interacted with T cells and resulted in IL-10 secretion by activated T cells, hence resulting in the phosphorylation of STAT3 and induction of PD-L1 expression in MDSCs [273,315,318]. Furthermore, PV and ET patients have increased toll-like receptor 2 (TLR2) levels. This activates the MEK/ERK and STAT pathway, enhancing expression of PD-L1 [273,315]. All these contribute to the oncogene-mediated immune escape via JAK2/STAT3/STAT5/PD-L1 axis. The enhanced PD-L1 expressed in MPN binds to PD-1 on T cells to suppress their cysteine metabolism. This led to anergy, decreased cell cycle activity and exhaustion of T cells [317]. Several anti-PD-1 and anti-PD-L1 antibodies are developed with ongoing clinical trials. Their results, however, were not very encouraging [319][320][321] ( Table 3).

Peptide Vaccination in CALR Exon 9
Mutant CALR possesses a new and large C-terminal peptide sequence which is completely distinct from wild-type CALR. This warrants interests in using CALR as an immunotherapeutic target [322,323]. In MPN, mutant CALR is overexpressed. This reduces MHC-I assembly and loading on cell surface, and impairs CD8 + T-lymphocyte targeting [323]. Interestingly, it was observed that some healthy donors harboured CD4 + memory T-cells towards CALR epitopes [323]. Based on the above observation, a CALR-mutated CD4 + T-lymphocyte clone has been designed to induce cytotoxic effects against CALR-mutated cells [323]. A phase I clinical trial is underway, but no result has been released yet [324] (Table 3).

Other Potential Therapeutic Strategies Targeting MPN Stem Cells
Several new approaches have been observed to potentially target MPN stem cells, either as single agents or in combination with Peg-IFNα-2a or ruxolitinib (Tables 4 and 5). Table 4.
Novel therapies in combination with standard treatment (interferon-alpha/ruxolitinb) in targeting Ph-negative MPNs.

Conclusions
Understanding the biology of CML and Ph-negative MPNs is an important area of scientific research. With considerable data providing insights into the biology and therapeutic targeting abnormal HSCs in CML, efforts towards identifying and quantifying abnormal HSCs in CML may facilitate efforts in achieving treatment-free remissions. In Phnegative MPNs, this understanding may shift the treatment paradigm from cytoreduction and symptom control to effective disease modification.