Inflammatory Pathophysiology as a Contributor to Myeloproliferative Neoplasms

Abstract

Myeloid neoplasms, including acute myeloid leukemia (AML), myeloproliferative neoplasms (MPNs), and myelodysplastic syndromes (MDS), feature clonal dominance and remodeling of the bone marrow niche in a manner that promotes malignant over non-malignant hematopoiesis. This take-over of hematopoiesis by the malignant clone is hypothesized to include hyperactivation of inflammatory signaling and overproduction of inflammatory cytokines. In the Ph-negative MPNs, inflammatory cytokines are considered to be responsible for a highly deleterious pathophysiologic process: the phenotypic transformation of polycythemia vera (PV) or essential thrombocythemia (ET) to secondary myelofibrosis (MF), and the equivalent emergence of primary myelofibrosis (PMF). Bone marrow fibrosis itself is thought to be mediated heavily by the cytokine TGF-β, and possibly other cytokines produced as a result of hyperactivated JAK2 kinase in the malignant clone. MF also features extramedullary hematopoiesis and progression to bone marrow failure, both of which may be mediated in part by responses to cytokines. In MF, elevated levels of individual cytokines in plasma are adverse prognostic indicators: elevated IL-8/CXCL8, in particular, predicts risk of transformation of MF to secondary AML (sAML). Tumor necrosis factor (TNF, also known as TNFα), may underlie malignant clonal dominance, based on results from mouse models. Human PV and ET, as well as MF, harbor overproduction of multiple cytokines, above what is observed in normal aging, which can lead to cellular signaling abnormalities separate from those directly mediated by hyperactivated JAK2 or MPL kinases. Evidence that NFκB pathway signaling is frequently hyperactivated in a pan-hematopoietic pattern in MPNs, including in cells outside the malignant clone, emphasizes that MPNs are pan-hematopoietic diseases, which remodel the bone marrow milieu to favor persistence of the malignancy. Clinical evidence that JAK2 inhibition by ruxolitinib in MF neither reliably reduces malignant clonal burden nor eliminates cytokine elevations, suggests targeting cytokine mediated signaling as a therapeutic strategy, which is being pursued in new clinical trials. Greater knowledge of inflammatory pathophysiology in MPNs can therefore contribute to the development of more effective therapy.

Keywords: JAK2; NF kappa B (NFκB); Ruxolitinib; cytokines; intracellular signaling; myelofibrosis; myeloproliferative neoplasms; tumor necrosis factor (TNF).

Figure 1

Mechanisms of JAK2 activation by MPN driver mutations. (A) Normal mechanism for receptor activation of the human thrombopoietin (TPO) receptor MPL (myeloproliferative leukemia proto-oncogene). MPL, expressed in HSPC and megakaryocytic lineage cells, exists in an equilibrium between inactive monomers and active homodimers. Binding of the monomeric ligand TPO stabilizes dimer formation, allowing phosphorylation of the receptor by dimeric JAK2 tyrosine kinase. This step initiates intracellular signaling downstream of active JAK2. The active receptor-kinase holocomplex is subject to inhibition by several inhibitor molecules, notably SOCS (suppressor of cytokine signaling) family proteins, tyrosine phosphatases, and the adaptor protein SH2B3 (also known as LNK, lymphocyte linker protein). Mutations in genes encoding these various inhibitory molecules are rare driver mutations in MPNs, which lead to JAK2 hyperactivation by removal of physiologic inhibition. (B) Activation of JAK2 signaling by mutant CALR. The single-pass transmembrane protein CALR (calreticulin) is a calcium-binding chaperone protein that normally recycles between plasma and intracellular membranes in the secretory pathway. Mutated CALR acquires a neomorphic C-terminus (depicted as striped), which is capable of binding to and dimerizing MPL, consequently producing signaling-active MPL homodimers, which recruit and activate JAK2, in the absence of TPO binding. CALR mutations in MPN are therefore hypothesized to have similar effects to activating MPL W515L/K mutations, which are present in roughly 10% of ET and MF patients. (C) Effect of the JAK2 V617F mutation. JAK2 (Janus kinase), depicted as the ancient Roman god Janus (sculpture in the Vatican Museum, Rome), contains homologous kinase and pseudokinase domains, the pseudokinase being inhibitory to the kinase. V617F mutation in the pseudokinase domain inactivates the inhibition, producing a constitutively active (autophosphorylated) kinase. (D) Direct activation of JAK/STAT signaling by active JAK2. Mutant JAK2 (typically V617F) dimerizes MPL and other cytokine receptors, rendering the receptor active even in absence of a bound ligand. The active receptor-bound JAK2 phosphorylates STAT3 and STAT5 homodimers, which then translocate to the nucleus to activate transcription. Unlike MPL and CALR mutations, JAK2 mutations enable constitutive JAK2 activity even in cells that do not express MPL. Mutant JAK2 has been shown to promote epigenetic changes and increase the potential for unrepaired DNA damage. Active JAK2 collaterally activates MAP kinase (MAPK) and PI3 kinase (PI3K) signaling pathways independently of STAT3 and STAT5.

Figure 2

Distinct coexpressed groups of cytokines overproduced in MPNs. Data specific to MF, also utilized in Fisher et al. (48). (A, B) Presence or absence of tandem regulation of cytokine induction by TPO. Biaxial plots show in rows (from top to bottom) ex vivo cell samples from healthy control blood and blood from the JAK2 V617F mutant MF patient MF20. Columns show cytokines as identified in cells by mass cytometry (CyTOF) after 4-hour incubation either without stimulation (Basal) or stimulated by TPO. (A) In monocytes, TNF (Y axis) showed coexpression with IL-8/CXCL8 (X axis) when induced by TPO. Combined induction is illustrated by schematic below lower right plot. (B) TGFβ (X axis) showed minimal basal coexpression with TNF (Y axis) in monocytes. TPO stimulation, however, induced TNF but not TGFβ, as illustrated by schematic under lower right plot. (C) Schematic depicting “axis” groups of cytokines overproduced in MF myeloid cells ex vivo. The majority of cytokines could be separated into TNF or TGFβ “axis” groups (related to biaxial plots in A, B), based on co-expression with either TNF or TGFβ after stimulation with TPO, TNF, or a TLR receptor ligand (R848 or PAM3CSK4), with Pearson R>0.25 (48). Of 15 cytokines assayed by CyTOF in Fisher et al. (48), only VEGF and IFNγ failed to demonstrate coexpression with either TNF or TGFβ, while IL-8/CXCL8 showed some evidence of coregulation with both (R>0.25).

Figure 3

Signaling pathways activated by TNF receptors. (A) Bivalent cell-death regulating Complex I formed at plasma membrane by TNFR1 but not TNFR2. TNF receptors TNFR1 (encoded by TNFSF1A) and TNFR2 (encoded by TNFSF1B) are homotrimeric receptors binding homotrimeric TNF ligand. TNFR1, unlike TNFR2, contains an intracellular “death domain” (DD), which binds homologous DDs on intracellular RIP1, FADD, and TRADD, to compose the core of Complex (I) RIP1 recruits the pro-apoptotic RIPK1 kinase, while TRADD recruits the anti-apoptotic TRAF adaptor protein family members, which are necessary for activation of NFκB and MAP kinase signaling downstream of TNFRs. TNFR2, which lacks a “death domain”, can recruit TRAFs directly, but with a lower binding affinity than Complex (I) TRAFs can recruit CIAP1 and CIAP2, anti-apoptotic proteins which form heterodimers (encoded by BIRC2 and BIRC3). (B) Mechanism of cell death inhibition by CIAPs at TNFRs. CIAPs are ubiquitin ligases, which polyubiquinate RIPK1, targeting it for degradation. CIAPs can be recruited intracellularly at TNF-bound TNFR1 or TNFR2. SMAC/DIABLO is an inhibitor of CIAPs: hence, SMAC/DIABLO mimetics (such as LCL-161 and birinapant) are pro-apoptotic. (C) Cell death pathways downstream of TNFR1. The uninhibited Complex I (with active RIPK1) promotes formation of the cytoplasmic death promoting Complexes IIa and IIb. Both complexes can activate cytoplasmic CASP8, the cleavage of which, to an active protease (cCASP8), initiates the apoptotic cascade. CASP8 activation can be inhibited by the long form of FLIP, encoded by the NFκB target gene CFLAR: a mechanism by which NFκB activation can promote cell survival. Complex IIb (containing RIPK3 as well as RIPK1) activates the kinase MLKL, which activates the necroptotic signaling cascade. (D) Activation of NFκB by TNFRs. TRAFs bound to TNFR1 or TNFR2 recruit TAK1, an essential activator of NFκB signaling. TAK1 activates the canonical NFκB pathway by recruiting the IKK complex, consisting of NEMO/IKKγ, IKK1, and IKK2. The IKK complex phosphorylates IκB family members, dissociating them from NFκB subunits and targeting them for degradation. NFκB, freed from IκB, is phosphorylated by both the IKK complex and casein kinase II (CK2), which in tandem activate NFκB subunits, which translocate to the nucleus to act as transcription factors. Bortezomib (directly) and pevonedistat (indirectly) inhibit IκB degradation. Among NFκB target genes are those encoding several cytokines and their receptors, as well as both antiapoptotic and pro-apoptotic components of the TNFR-NFκB signaling cascades.

Figure 4

Cell-autonomous and non-cell-autonomous paths to NFκB pathway activation in MPNs. (A) Hypothesized pathways for cell-autonomous activation of NFκB downstream of active JAK2. JAK2 is activated by cytokine receptors such as MPL, or constitutively resulting from the V617F mutation. Active JAK2 activates RAS, in turn activating the MAP kinase and PI3 kinase signaling cascades, which are active in MPN HSPCs. Reported inputs of these pathways to NFκB include phosphorylation of IKK1 by AKT and phosphorylation of p65/RELA by S6 kinase (111, 112). The cell cycle kinase CDK6, a transcriptional target of both JAK/STAT and NFκB signaling, also can phosphorylate RELA (113). Active phosphorylated RELA translocates to the nucleus to mediate transcription. (B) Several non-cell-autonomous inputs overproduced in MPNs can activate NFκB. MPN driver mutations produce JAK2 hyperactivity, which leads to pathophysiologic production of TNF, IL-1, and TLR ligands S100A8/A9 and related family members. NFκB is activated either by TNFRs via TRAFs and TAK1, or by TLRs or IL-1 receptor, via the adaptor protein MYD88, which binds to these receptors, and recruits IRAK1/IRAK4 kinase heterodimers, which in turn activate the IKK complex.

Figure 5

Transmission of JAK/STAT and NFκB pathway activation from malignant to non-malignant cells. (A) Malignant hematopoietic MPN cells transmit NFκB signaling activation to non-malignant cells, according to the hypothesized mechanism described by Kleppe et al. (86) MPN driver mutations produce hyperactive JAK2, leading to overproduction of cytokines, including IL-6. IL-6 receptor activation (in non-malignant cells) activates JAK2 and STAT3, which are required for maximal non-cell-autonomous NFκB activation in non-malignant cells. STAT3 shares multiple target genes with NFκB. NFκB mediated transcription requires BET bromeodomain proteins (BRDs) as cofactors, and therefore is subject to inhibition by the BRD inhibitor JQ1. (B) Active cytokine receptor signaling (such as from MPL or IL-6R) in malignant cells activates JAK2 and phospho-STATs 3 and 5, which co-activate multiple target genes along with NFκB. Among JAK2/STAT3,5 and NFκB co-induced targets are genes encoding several cytokines overproduced in MPNs: TNF, CCL3 (MIP-1α), CCL4 (MIP-1β), IL-6, IL-8, IL-1α, and IL-1β (41, 48, 53, 84, 86). IL-6 and IL-8 (bold) can activate JAK2/STAT3,5 signaling in non-mutant cells. These cytokines in turn act non-cell-autonomously on the endothelial cells of blood vessels (BV), mesenchymal stromal cells (MSC), and endosteal cells in the bone marrow. MSC are sources of SCF and IL-6 and endosteal cells are sources of TPO, which can activate JAK2/STAT3,5 signaling non-cell-autonomously (also see Figure 6 ). (C) Malignant neutrophils (PMN) in MPNs can produce IL-1α, and IL-1β, while malignant monocytes produce IL-6, IL-8, and TNF. TNF, IL-1α, and IL-1β, can activate NFκB signaling in endothelial cells of blood vessels (BV), leading them to release SCF and IL-6. These cytokines can combine with IL-6 and IL-8 secreted by malignant monocytes to produce activated JAK2/STAT3,5 signaling in both malignant and non-malignant cells.

Figure 6

Hematopoietic bone marrow niches disrupted in MPNs. (A, B) Bone marrow niches in healthy hematopoiesis. (A) The endosteal niche: HSPC reside in contact with the endosteum, composed of osteoblasts that release TPO, promoting HSC quiescence. CXCL12 secreted by CAR cells promotes HSPC stasis in the bone marrow, while mesenchymal stromal cells (MSC) secrete SCF and IL-6. (B) The perivascular niche: HSPC reside in contact with blood vessels (BV), which are also contacted by the CXCL12-secreting CAR cells. This niche, however, is more prone to HSPC circulation than the endosteal niche. Monocytes (Mono) and megakaryocytes (Mega) secrete cytokines active on HSPC. (C, D) Disruption of hematopoietic niches in MF. (C) In MF, MSCs are abundant but CAR cells are reduced. The endosteum can be disrupted, and MSCs can differentiate into fibrocytes and deposit collagen, disrupting blood vessels in the hematopoietic space. Consequently, HSPC become mobilized. (D) Monocytes and megakaryocytes become abundant in the MPN bone marrow, releasing cytokines including TNF and the fibrogenic TGFβ. Monocytes, as well as MSCs, can differentiate into fibrocytes (71).

Figure 7

Hypothesized mechanisms of bone marrow niche remodeling in MPNs based on studies in mouse models. This figure illustrates mechanisms hypothesized from mouse model studies by Ramalingam et al. (127) Nature Communications (127) (A) and by Arranz et al. (128) Nature (128) (B). (A) (from top left, following arrows indicating course of pathogenesis): Expression of MAPKK1 S218D, S22D mutant in endothelial cells, resulting in constitutive activation of MAP kinase signaling, also produced constitutive NFκB activation, possibly by a cell-autonomous mechanism, as described in Figure 4A . This led to HSPC proliferation and losses of stemness and regeneration potential in mouse HSPC: phenotypes derived non-cell-autonomously, since mutant MAPKK1 expression was confined to endothelial cells. HSPC phenotypes were dependent on NFκB hyperactivation in HSPC, as they could be entirely rescued by hematopoietic expression of the non-degradable IκBα S32A, S36A “super repressor” mutant. NFκB hyperactivation in HSPC promoted myeloid differentiation with loss of HSPCs and lymphocytes, resulting in pancytopenia and bone marrow failure (a phenotype also observed to result from pan-hematopoietic NFκB hyperactivation in mice) (–133). (B) Bone marrow remodeling by hematopoietic Jak2 V617F, analogous to human MPNs. 1. In healthy mouse bone marrow, Cxcl12 secreted by CAR cells and cytokines secreted by MSCs maintain HSPCs in the perivascular niche (analogous to Figure 6B ). Bone marrow also contains sympathetic nerve fibers, which secrete norepinephrine (NE). Schwann cells (SC) are associated with the sympathetic neuronal fibers, and essential for their survival. 2. Jak2 V617F, expressed in hematopoietic cells, causes secretion of Il-1β (Il-1α was not assayed). Il-1β caused apoptosis of CAR cells, other MSCs, and Schwann cells, leading to sympathetic denervation of bone marrow. The exact downstream signaling pathways to apoptosis and marrow fibrosis were not defined in this study. These features were, however, substantially rescued by either a catecholaminergic agonist or the natural Il-1 receptor antagonist Il-1ra, establishing essential roles of both Il-1 and sympathetic denervation in bone marrow pathophysiology caused by Jak2 V617F.