Novel strategies for the treatment of myelofibrosis driven by recent advances in understanding the role of the microenvironment in its etiology

JAK inhibitors

Following the discovery of JAK2V617, a non-selective JAK1/2 inhibitor, ruxolitinib, was developed and has been shown in two pivotal phase III studies to induce significant spleen volume reduction and improvement in constitutional symptoms, leading to its approval by the USA and European regulatory agencies for intermediate-2 and high-risk MF^29,30. The widespread use of ruxolitinib has changed the therapeutic landscape and significantly impacted the quality of life of many MF patients. Moreover, long-term follow up of the COMFORT studies has suggested a survival advantage with the use of ruxolitinib³¹. Since the biological effects of this drug on mutation allele burden or bone marrow histological findings have been inconsistent and overall modest^32–34, other analyses have attributed the beneficial effect of ruxolitinib on survival to the attenuation of systemic inflammation and reversal of cachexia owing to spleen volume reduction. Furthermore, the use of ruxolitinib is limited in patients with anemia and thrombocytopenia. These considerations inspired the development of second-generation JAK inhibitors that lack myelosuppressive effects and may allow effective treatment for patients with cytopenias. Three second-generation JAK inhibitors have been evaluated in phase III clinical trials, and their efficacy results are summarized in Table 1.

Pacritinib is an oral multi-kinase inhibitor with specificity for JAK2, FLT3, IRAK1, and CSF1R. The PERSIST-1 study randomized ruxolitinib-naïve patients irrespective of platelet count to either pacritinib at a dose of 400 mg once daily or best available therapy (BAT) excluding ruxolitinib. Spleen and symptom responses were significantly superior in the pacritinib arm. Notably, patients with low platelet counts achieved comparable benefits. Toxicities were overall manageable³⁵. PERSIST-2 recruited only patients with thrombocytopenia and randomized them to pacritinib at one of two doses or BAT including ruxolitinib. Prior use of ruxolitinib was allowed. Pacritinib at both doses was superior to BAT and led to significant spleen and symptom responses, with tolerable myelosuppression³⁶. The development of pacritinib was halted by a clinical hold placed by the FDA due to suspicion of excess cardiovascular deaths in patients treated with the drug. This hold was subsequently removed, and the results of a new study that may hopefully lead to pacritinib’s approval are awaited.

Momelotinib is a JAK1/2 inhibitor with activity against activin receptors believed to contribute to a remarkable anemia response, rendering this drug attractive, especially in the setting of transfusion-dependent anemia. The two phase III SIMPLIFY studies that evaluated momelotinib, despite showing a compelling anemia response, each failed to show superiority or non-inferiority in a primary or key-secondary endpoint^37,38. A more pragmatic study design that would take into account the anemia response and superior tolerability of this drug along with spleen and symptom responses could allow for a better evaluation of this potentially beneficial agent⁴¹. Recently, the development of momelotinib was renewed, and a new phase III study is recruiting patients.

Fedratinib is a selective JAK2 inhibitor. The JAKARTA-1 study evaluated two doses of fedratinib versus placebo and showed significantly superior spleen and symptom responses in both treatment arms³⁹. JAKARTA-2 was a single-arm phase II study evaluating fedratinib in patients who were intolerant of or resistant to ruxolitinib⁴⁰. This was the only clinical trial directly evaluating a second-generation JAK inhibitor as second-line treatment, and responses, especially with regards to spleen volume reduction, were compelling for this patient population. The development of fedratinib was also stalled by a clinical hold placed by the FDA due to a suspected association with Wernicke’s encephalopathy, which was eventually lifted following re-evaluation. The phase III FREEDOM study (NCT03755518) is currently recruiting patients and will hopefully contribute to further development of this drug with apparent potential clinical benefit. Based on the results of the JAKARTA-1 trial, on 16 August 2019, the FDA approved fedratinib for the treatment of adult patients with intermediate-2 or high-risk MF.

Bone marrow transplantation

The discovery by Fialkow in 1976 that MPNs are a clonal disorder of HSCs³, more recently confirmed at the molecular level by the Weissman laboratory⁴, led to the hypothesis that MPNs may be cured by HSCT, a procedure which replaces the malignant HSCs of the recipient with healthy ones provided by the donor. However, since the disease also impairs the supportive bone marrow microenvironment, which remains patient derived after transplantation⁴², this approach was initially received with skepticism. Reports of long-term survival of transplant recipients coupled by histological evidence of resolution of bone marrow fibrosis eventually provided compelling evidence for the curative potential of this approach^43–46. However, HSCT remained limited to a relatively small proportion of patients owing to the older age and frequent comorbidities of most patients and the high non-relapse mortality reported in early studies.

Recently, advances in transplantation medicine and molecular diagnosis are being translated to a consistent improvement in HSCT outcomes in MF^47–49. Several factors contribute to the increased utilization and improved outcomes of HSCT for MF treatment: 1) the use of reduced intensity conditioning (RIC) regimens in older patients, which have demonstrated decreased transplant-related mortality with comparable relapse rates and overall survival^50–52; 2) increased accessibility to HSCT for patients lacking sibling or matched-unrelated donors owing to advances in the use of alternative donors such as haploidentical donors and cord blood^53–55; 3) accumulating data regarding mutational profiles and their prognostic significance allowing earlier and more informed patient selection⁵⁶; and 4) accumulating experience with the use of ruxolitinib before or after transplant.

HSCT is presently considered the standard of care in eligible patients with intermediate-2 and high-risk disease³⁴. There is lack of consensus regarding its use in earlier stages of the disease predicted to have more prolonged survival. A subset of patients classified as having low/intermediate-1 risk disease may eventually experience rapid progression to advanced MF and shortened survival. The recent advances in linking genetic profiles with risk of progression may improve the prediction of disease course and allow us to identify early stage patients who should be considered for HSCT.

Agents that improve anemia

Activins are members of the transforming growth factor (TGF)-β superfamily that inhibit the differentiation of late-stage erythrocyte precursors in a mechanism independent of erythropoietin and are overexpressed in myelodysplastic syndromes (MDS) and in MF. Two agents that antagonize the activity of activins, sotatercept and luspatercept, have demonstrated efficacy in the treatment of anemia in patients with MDS^59–63. These agents are currently being evaluated in patients with MF and anemia with promising results in early phase studies⁶⁴.

Drugs targeting epigenomic alterations found in malignant HSCs

Members of the Bromodomain and Extra-Terminal (BET) family of proteins function as “readers” of histone modification marks, by interacting with acetylated lysine residues on histone tails, and regulate genes that are involved in inflammation and cancer such as MYC, BCL-2, and NF-kB^65,66. The NF-kB pathway downstream to BET has been shown to be activated in MF via JAK–STAT and other inflammatory cytokine signaling⁶⁷. In a preclinical study using two MF mouse models, treatment with a BET inhibitor (BETi) resulted in reduction of cytokine production, spleen volume, and bone marrow fibrosis, and these effects were enhanced in combination with ruxolitinib⁶⁸. These data indicate that drugs that reduce NFκB function and expression of NF-kB target genes might be effective in treating MF⁶⁹. CPI-610, a BET inhibitor, is currently being evaluated for the treatment of MF patients, alone or in combination with ruxolitinib (NCT02158858).

While most of the above investigational agents as well as JAK inhibitors target signaling pathways involved in myeloproliferation and systemic inflammation, agents directly targeting MF-HSCs are lacking. The development of such agents has been a major challenge in the effort to improve the therapeutic armamentarium for myeloid malignancies. Imetelstat is an inhibitor of telomerase enzymatic activity⁷⁰. The rationale for its use in MF stems from the findings of enhanced telomerase activity in MPN granulocytes⁷¹. A pilot study evaluating the use of single-agent telomerase in 33 patients with intermediate-2 and high-risk MF demonstrated a modest overall response rate (ORR) of 21%; however, in all four patients who achieved a complete response, bone marrow fibrosis was reversed, and a molecular response occurred in three of these patients, suggesting that imetelstat targets the malignant HSCs⁷². Preclinical studies confirmed these findings by demonstrating direct targeting of MF-CD34⁺ cells by imetelstat, reflected by decreased formation of megakaryocytic colonies and by reduced human chimerism in an immune-deficient mouse xenotransplantation model. Both of these effects were selective to cells from MF patients, but not normal individuals, and sustained over time⁷³. Recently, a multicenter phase II study evaluating two doses of imetelstat in a cohort of MF patients that progressed after or were refractory to ruxolitinib treatment has demonstrated an estimated twofold prolongation of survival as compared to retrospective reports of comparable patient populations, warranting further clinical testing⁷⁴.

Targeting the inflammatory microenvironment

Several clinical studies have shown that MPN patients express increased levels of inflammatory mediators⁷⁵, including the key inflammatory cytokine interleukin-8 (IL-8), the plasma levels of which predict adverse clinical outcomes⁷⁶. The pathogenetic role of these cytokines is also supported by the observation that JAK inhibition can attenuate features of MPNs in vivo through the inhibition of cytokine production in mutant and non-mutant cells⁷⁷. However, there is limited insight on the pathobiological consequence of aberrant cytokine production in MPNs. It is conceivable that specific targeting of some of the cytokines altered in MPNs may lead to better clinical outcomes than those obtained with JAK inhibitors⁷⁷. An altered cytokine pathway that holds potential clinical interest in MF is the lipocalin-2 (LCN2)/interleukin-8 (IL-8) axis and its down-stream signaling via NF-kB. There is evidence that the sequence of events leading to abnormalities in the microenvironment in MF spleens and marrow includes, at least in part, LCN2⁷⁸. LCN2 increased the proliferation of splenic endothelial cells, and LCN2 treatment of splenic stromal cells led to increased elaboration of IL-8, which contributes to the creation of an endothelial cell niche supporting the proliferation of MF-HSCs^79,80. This endothelial cell niche can be disrupted in vitro by reparixin, an antagonist of the receptors for IL-8 CXCR1/2^81,82 that are highly expressed on MF spleen CD34⁺ cells⁷⁹. These data indicate that reparixin may represent a novel therapy which can antagonize the effects of IL-8 and thereby disrupt the HSC niche function of splenic endothelial cells. A clinical trial to test the effects of reparixin in MF is planned to be opened by the Myeloproliferative Neoplasm Research Consortium.

Boosting the patient’s immune system

An important emerging approach in cancer therapy is activation of endogenous anti-tumor activity by treatment with agents that suppress the immune evasion mechanisms activated by cancer cells⁸³. Checkpoint inhibitors, which block the interaction between programmed cell death (PD)1 and its ligand PD-L1 that is exploited by cancer cells to prevent attack by cytotoxic T-lymphocytes, have shown impressive anti-tumor responses in several solid cancers and B-lymphoid malignancies⁸³. There is a strong pre-clinical rationale for inhibition of PD1/PD-L1 interaction in MF^84–86, and several clinical trials are ongoing (NCT03065400, NCT02421354, and NCT02871323). In addition, recent studies suggest that mutant calreticulin and JAK2 can induce specific anti-tumor T-cell responses, which can be harnessed for the development of mutation-specific peptide vaccines^87,88. These and other immune therapy approaches, such as bispecific antibodies, are still in early stages of development.

Targeting the reduced expression of p53 in MF-HSCs

p53 is a tumor suppressor that regulates cell cycle, apoptosis, DNA repair, and senescence of many cell types^92,93. Hypomorphic/loss-of-function mutations of p53 are associated with tumor progression in most human cancers¹⁸, and inactivating mutations are observed in 20% of MPN-BP while deletion of p53 leads to the development of AML in JAK2V617F-harboring mice⁹⁴. Although in patients with chronic-phase MPN p53 is usually wild-type, as discussed in “Targeting the inflammatory microenvironment”, malignant HSCs are p53 hypomorphic, as they express low levels of the protein^95,96.

Studies to determine the levels of p53 in MF patients with no obvious p53 mutations were inspired by Takaoka et al., who reported that interferon suppresses tumor progression and activates the immune response by increasing the expression of p53 in mouse models of fibroblast transformation⁹⁷. Since interferon is an effective therapy in patients with MPNs⁹⁸, it was hypothesized that interferon is at least partially effective in these patients by increasing the otherwise low levels of p53 expressed by the malignant HSCs⁹⁹. As in other cancers^95,100, the mechanisms that reduce p53 in MPNs may be represented by overexpression of its major regulators MDM2 (HDM2 in humans) and MDM4 (HDM4/HDMX), which are expressed at levels higher than normal in MF CD34⁺ cells⁹⁹. Overexpression of HDM4 and HDM2 in MF may be genetically determined. The two genes are localized on chromosomes 1 and 12, respectively, and gain of 1q and 12q rearrangements are frequently associated with disease progression in MPNs¹⁰¹. Alternatively, the observation that high TGF-β1 induces the expression of HDM2 in breast cancer driving its late metastatic stage¹⁰² suggests that, even in the absence of chromosomal duplication, the great levels of TGF-β expressed in the microenvironment (see following section) may drive high levels of HDM2 in MF.

The existence of a p53/HDM2 cycle in MF is important because binding of p53 to HDM2 resulting in p53 ubiquitination and degradation is inhibited by a class of compounds called Nutlins. Nutlins (Nutlin-3, RG7112, RG7388, HDM201, and KRT232) bind to the MDM2 p53-binding site, interfering with its interaction with p53, leading to p53 accumulation and activation^{95,100,103–106}. Proof-of-principle for the therapeutic use of Nutlins in MF was provided by the observation that the orally available Nutlin RG7388 (idasanutlin) depletes MPN-HSCs in culture⁹⁹. In addition, low doses of the Nutlin RG7112 induce apoptosis of MF CD34⁺ cells and reduce donor cell chimerism and JAK2V617F allele burden in NSG mice transplanted with MF CD34⁺ cells¹⁰⁷. The promising results of an open label phase I study of idasanutlin as a single agent in patients with PV and ET presented by Dr. Mascarenhas et al. in 2017 at ASH¹⁰⁸ and the observation that in AML the patients with the highest overexpression of HDM2 are most responsive to therapy with idasanutlin¹⁰⁹ suggest that MF patients, who express higher levels of HDM2 than those observed in PV, will respond well to idasanutlin. A clinical trial with idasanutlin in MF will be conducted soon by the Myeloproliferative Neoplasm Research Consortium.

Targeting the TGF-β circuitry in MF

Megakaryocytic hyperplasia was the first histopathological hallmark identified in MF¹¹⁰. These early studies also identified that this abnormality is associated with increased TGF-β content and release due to pathological emperipolesis (the interaction between megakaryocytes and neutrophils)¹¹¹. However, in spite of the overwhelming evidence linking TGF-β with fibrosis in other systems^112,113, the link between megakaryocyte and TGF-β abnormalities and disease progression in MF has been obscured for some time by the fact that they are not direct targets of the driver mutations.

The link between driver mutations, alterations of megakaryocyte maturation, and disease progression in MF has been recently clarified by the observations that mice expressing JAK2V617F only in megakaryocytes develop MF¹¹⁴; the driver mutations induce a ribosomopathy that reduces the content of the transcription factor GATA1 (which is essential for terminal maturation) in megakaryocytes, halting their maturation^115,116; mice lacking the regulatory sequences which specifically drive GATA1 expression in megakaryocytes (Gata1^low mice) develop the same megakaryocyte abnormalities observed in MF and MF with age^117,118; and, finally, treatment with an inhibitor of Aurora kinase A, a protein overexpressed in MF megakaryocytes, rescues GATA1 expression in these cells, curing MF in animal models¹¹⁹, suggesting GATA1 as a druggable target in MF¹²⁰. This clinical hypothesis was tested by demonstrating that an inhibitor of Aurora Kinase A has some efficacy in MF patients¹²¹. Owing to these exciting results, additional megakaryocyte abnormalities are currently being considered as potential therapeutic targets for MF.

The Vainchenker laboratory hypothesized for the first time that TGF-β represents a target to cure MF by demonstrating that malignant HSCs lacking the TGF-β gene fail to induce MF upon transplantation in healthy recipients¹²². Furthermore, malignant HSCs did not induce MF when transplanted into recipients lacking the TGF-β receptor 1 gene¹²³. These results suggest that TGF-β produced by the progeny of the malignant HSCs exerts its pathobiological effects by activating its receptor on a cell population present in the microenvironment. However, attempts to explore the clinical potential of TGF-β in MF were hampered for some time by the fact that the gene ablation strategy used by Vainchenker is not practical in clinical settings.

TGF-β is known to play a major role in the development of fibrosis in multiple organs¹²⁴ and in the induction of tumor-supporting microenvironments in other cancers^125,126. The experiments by Vainchenker do not clarify whether TGF-β promotes MF by inducing fibrosis and therefore by hampering the ability of the bone marrow microenvironment to support normal HSCs and/or by inducing a microenvironment supporting the malignant HSCs in extramedullary sites. Using the Gata1^low mouse model, our laboratory has identified at least some of the events linking megakaryocyte abnormalities, increased TGF-β, and development of MF. We determined that, although plasma and washes from bone marrow and spleen of Gata1^low mice, as well as those from MF patients, contain levels of total and bioactive TGF-β modestly (twofold) greater compared to normal controls^127,128, the megakaryocytes and their surrounding microenvironment contain levels of TGF-β 1000–2000-fold greater than normal^128,129, suggesting that it is not increased TGF-β content per se but its increased bioavailability that plays a major role in inducing disease progression in MF. By careful electron-microscopy analyses, we also determined that the increased TGF-β bioavailability is established by a sequence of abnormal cellular interactions involving megakaryocytes, neutrophils, and activated fibrocytes that are mediated by adhesion receptors that are druggable (Figure 1). Ablation of one of them, P-selectin, rescues MF in Gata1^low mice, providing clinical rationale for the use of P-selectin inhibitors, alone or in combination with ruxolitinib, in MF.

Figure 1. A cellular model for the establishment of increased transforming growth factor (TGF)-β bioavailability, which leads to fibrosis and disease progression in myelofibrosis (MF).

This model was elaborated thanks to the fact that, in contrast with other animal models that develop a MPN phenotype that rapidly progress into its fatal MF phase (discussed in 114), Gata1^low mice slowly develop MF with age^117,118. From 1–8 months, Gata1^low mice express pre-MF traits such as splenomegaly, increased rates of thrombosis, and osteosclerosis. From 8–12 months, they display MF traits including fibrosis and neovascularization, and from 12 months until their natural death they express a late-MF phenotype which includes increased stem/progenitor cell trafficking and extramedullary hematopoiesis in liver. The various phases are characterized by a sequence of abnormal cellular interactions that finally result in increased TGF-β bioavailability in the microenvironment. First, a pathological neutrophil-megakaryocyte (Mk) emperipolesis leads to death of the megakaryocytes by para-apoptosis, which releases TGF-β into the microenvironment. Second, TGF-β activates fibrocytes to produce collagen and to establish contacts with megakaryocytes, leading to the death of additional megakaryocytes and the release of activated lysyl-oxidase 2 (LOX2) into the microenvironment¹³⁵. LOX2 polymerizes the collagen produced by the activated fibrocytes into collagen fibers, resulting in fibrosis. The collagen fibers are heavy binders of TGF-β, inducing the formation of areas of increased TGF-β bioavailability in the microenvironment (see also 128).

To identify additional druggable TGF-β targets, we characterized the TGF-β expression profiling of bone marrow and spleen from MF patients and Gata1^low mice^128,130. These experiments identified that the tissues from the patients and the mouse model express the same distinctive abnormalities which include reduced expression of the canonical SMAD 1, 2, and 4 signaling and increased expression of JUNB, EVI1, and STAT1, three genes downstream of the non-canonical MAPK signaling. These abnormalities predict the activation of non-canonical p38/ERK-dependent rather than canonical SMAD-dependent signaling. The knowledge that activation of SMAD signaling induces normal HSCs into quiescence^131,132 while in systemic sclerosis activation of the p38/ERK pathway promotes the development of fibrosis^133,134 suggests a model for progression from the indolent phase of the disease (pre-MF) to MF depicted in Figure 2. In this model, increased levels of TGF-β promote disease progression by reducing the number of normal HSCs (inducing them into quiescence and reducing their supportive microenvironment in the bone marrow) while increasing the number of malignant HSCs (which are insensitive to TGF-β and are sustained by a specific microenvironment induced by TGF-β in the spleen).

Figure 2. Circuitry between p53 abnormalities in the malignant hematopoietic stem cells (HSCs) and transforming growth factor (TGF)-β in the supporting microenvironment leading to disease progression in myelofibrosis (MF).

It is suggested that in MF disease progression is driven by a p53/TGF-β circuitry. In the pre-MF stage, the driver mutations, possibly by inducing an inflammatory milieu (lipocalin-2 [LNC2]/interleukin-8 [IL-8]), reduce p53 in MF-HSCs, making these cells unresponsive to TGF-β and retarding megakaryocyte maturation. Retarded megakaryocyte maturation, associated with high IL-8 expression, induces, in an autocrine fashion, megakaryocytes to increase TGF-β bioavailability, which in turn is responsible for suppressing hematopoiesis from normal HSCs (inducing bone marrow failure) and for promoting an MF-HSC-supporting microenvironment in the spleen, facilitating the transition of pre-MF to MF (modified from 137). MK, megakaryocyte; PMF, primary myelofibrosis

Circumstantial evidence in support of the model for disease progression depicted in Figure 2 exists. First, we observed that the spleen, but not the bone marrow, from MF patients and Gata1^low mice contains greater numbers of activated fibrocytes that establish physical contacts with megakaryocytes, forming “nests”, which lodge hematopoietic cells with the morphology of HSCs^128,136. In the mouse model, the presence of these nests is induced by TGF-β. Second, we demonstrated that in contrast to normal HSCs, MF-HSCs are not induced into cycle by treatment with TGF-β receptor 1 kinase inhibitors¹³⁷.

Based on this overwhelming evidence, a proof-of-principle experiment was conducted that provided findings that one-month treatment with an inhibitor of LK5, the first element of the TGF-β receptor I signaling, completely rescues MF in Gata1^low mice^129,136. Furthermore, a limited clinical trial with a human neutralizing antibody against TGF-β (fresolimumab, Sanofi Aventis) provided evidence for sustained improvements in hemoglobin level^137,138 and of reduced bone marrow fibrosis that lasted for at least one year after the drug was discontinued¹³⁷ in at least two of the three patients treated. A drawback of drugs targeting TGF-β is their lack of specificity that makes them prone to off-target effects. The availability of ligand-traps specific for TGF-β1, such as AVID200 developed by Formation Biologics¹³⁹, makes conducting clinical trials to test the efficacy of TGF-β inhibition in MF and eventually to halt the progression from pre-MF to MF in patients without fear of cardiovascular or osteological complications finally possible. AVID200 will be tested in a clinical trial to be opened soon by the Myeloproliferative Neoplasm consortium.

The existence of a circuitry between p53 and TGF-β in development and cancer progression has been extensively discussed^140–142. In addition to the possibility that high levels of TGF-β may be responsible for reducing p53 in cancer cells by inducing HMD2 discussed in a previous section, “Targeting the reduced expression of p53 in MF-HSCs”¹⁰², other feedbacks between the two pathways have also been described. In some cell models, low levels of p53 are responsible for rendering malignant cells insensitive to the inhibitory effects of TGF-β on proliferation¹⁴³. It is therefore possible that, also in MF, low levels of p53 make the malignant HSCs TGF-β unresponsive. This consideration suggests that combination therapies targeting p53 (Nutlins) and TGF-β (AVID200), by disrupting the cross-talks between the two pathways, have a greater potential than single therapies to cure MF. This hypothesis will be tested by ongoing preclinical studies.