Always stressed but never exhausted: how stem cells in myeloid neoplasms avoid extinction in inflammatory conditions

Human hematopoiesis generates ∼10⁸ leukocytes, 2 × 10⁸ erythrocytes, and 10¹¹ platelets daily to support oxygen transport, hemostasis, and immune response.¹ Demand for mature hematopoietic cells can increase suddenly because of injury or infection, necessitating a system with large spare capacity and high regenerative potential. To address this challenge, complex regulatory networks maintain balance between quiescence, proliferation, self-renewal, and differentiation. Mouse data suggest that chronic inflammation or recurrent episodes of acute inflammation can exhaust the hematopoietic system’s regenerative capacity prematurely and compromise its function.² In humans, chronic inflammation can aggravate cytopenias in inherited bone marrow failure (BMF) syndromes, suggesting that inflammation may contribute to BMF.³ The opposite seems to apply in myeloid malignancies and their precursor clonal hematopoiesis (CH; the presence of bona fide driver mutations in the blood or BM of people without obvious hematologic abnormalities⁴). Starting with our initial observations, a large body of data has accumulated to implicate inflammation as a critical progression driver in myeloid malignancies, particularly myeloproliferative neoplasms (MPNs), raising the question of how malignant clones avoid attrition in highly inflammatory environments.⁵ Here, we review the mechanisms that protect normal hematopoiesis from exhaustion under inflammatory stress, drawing parallels between chronic inflammation and continuously increased proliferation in MPNs. This is followed by an in-depth discussion of dormancy as a crucial mechanism to preserve hematopoietic stem cells (HSCs) under stress, considering the possible role of stochastic effects. We then cover how myeloid malignancies co-opt physiological mechanisms to avoid extinction, illustrating general principles with selected examples. Lastly, we review emerging clinical implications and potential therapeutic strategies.

At homeostasis, blood cell numbers and lineage composition are maintained within narrow boundaries, with individual variation reflecting genetic and environmental factors.⁶ Increased demand for mature blood cells triggers a coordinated response termed emergency hematopoiesis that involves a complex, context-dependent interplay between pathogen-associated molecular pattern molecules (PAMPs), cytokines, and hematopoietic and stromal cells (reviewed in⁷). For instance, short-term HSCs (ST-HSCs) and multipotent progenitor cells (MPPs) express high levels of toll-like receptors (TLRs) and can be directly stimulated by PAMPs such as lipopolysaccharide (LPS), the ligand for TLR4, whereas nonhematopoietic TLR4-expressing cells are required for LPS-induced, granulocyte colony-stimulating factor (G-CSF)-mediated myelopoietic responses.^8,9 Barcoded HSC transplants in mice and data from human gene therapy trials suggested that steady state hematopoiesis is maintained by small numbers of long-term HSCs (LT-HSCs) that continuously undergo self-renewal divisions, producing a steady stream of maturing blood cells.^10,11 However, barcoding of individual HSCs using Cre recombinase or sleeping beauty transposase subsequently showed that multiple clones contribute to homeostatic hematopoiesis under nontransplant conditions, and identified ST-HSCs and MPPs as the main sources of mature blood cells, suggesting that, in homeostasis, the ST-HSC compartment requires minimal contributions from upstream LT-HSCs to remain balanced.^12,13 In contrast, flux from the LT-HSC to downstream compartments is much greater after treatment with 5-fluorouracil, suggesting that LT-HSCs are recruited to maintain downstream compartments under stress (Figure 1).¹³ 

Hematopoiesis handles single acute stress events without significant reduction of LT-HSCs, whereas chronic or repeated acute inflammation can lead to HSC attrition.^2,14-17 In fact, chronic infections are in the differential diagnosis for patients with pancytopenia, and a wealth of data from animal studies support the notion that chronic inflammation can have profound effects on HSC function. LT-HSCs from mice challenged by chronic infection (Mycobacterium avium-intracellulare [MAI]) or repeated exposure to PAMPs (such as poly I:C) exhibit reduced repopulation capacity that resembles aged hematopoiesis, suggesting sustained inflammatory stress reaches the most primitive cells.^2,18 Various cytokines and PAMPs are implicated in mediating inflammatory effects on HSCs, including interleukin-1β (IL-1β), IL-6, and LPS, among others.^17,19 Moreover, there is a widespread role for type I or type II interferon (IFN) signaling, depending on the specific inflammatory stimuli and their corresponding pattern recognition receptors. In MAI infection models, LT-HSC dormancy exit is abrogated by genetic ablation of type II IFN signaling (Ifngr1^−/−or Stat1^−/−), but not type I IFN signaling (Ifnar^−/−). Although LT-HSC numbers in Ifng^−/− and Ifng^+/+ are similar, Ifng^−/− HSCs exhibit improved engraftment, suggesting an increase in dormant cells.¹⁸ Subsequent studies confirmed that IFN-γ signaling causes HSC exhaustion in chronic bacterial and viral infections,^20,21 whereas in poly I:C challenge models, type I IFN signaling promotes recruitment of dormant LT-HSCs.^22,23 In addition, chronic viral infections can profoundly affect the BM stroma. For example, CXCL12 abundant reticular cells are decimated in mice with chronic lymphocytic choriomeningitis as the result of IFN-α and IFN-γ production by virus-specific CD8⁺ T cells.²⁴ Aging and experimentally enforced HSC proliferation induce hypermethylation of polycomb repressive complex (PRC)-regulated hematopoietic lineage genes.^2,25 Altogether, although evolution has devised effective mechanisms to protect the core of hematopoiesis during organismal lifetime, defenses can be breached by persistent stress. As such, CH and MPNs could be regarded as adaptive responses to inflammatory stress, a notion discussed in subsequent sections.^26,27 Because studies in mice suggest that flux from the dormant LT-HSC to downstream compartments is minimal, one would predict that damage to dormant HSCs becomes apparent only when the compensation capacity of downstream compartments is exhausted (Figure 1).¹² This would explain why most cases of aplastic anemia lack an exposure history, but, to our knowledge, has not been tested experimentally. With the limitations of normal HSCs to handle repeated acute or chronic inflammatory stress, how do SCs in myeloid malignancies manage to avoid exhaustion? To approach this question, we will compare inflammatory signaling in MPNs and physiological stress responses.

MPNs and chronic infections share clinical features, including constitutional symptoms, left-shifted granulopoiesis, and elevated inflammation biomarkers. These clinical commonalities are paralleled by similarities in the pathways activated by intrinsic (oncogenic) or extrinsic (inflammatory) signals. However, compared with HSCs, MPN SCs are more resilient or may even exhibit paradoxical responses to cytokines, generating a survival differential between cell populations living in the same ecosystem.

Many MPNs are driven by mutations that activate pathways downstream of cytokine receptors such as JAK-STAT, MAPK, and nuclear factor κB.^28-30,CSF3R (also known as G-CSF receptor) mutations illustrate the complex relationships between physiological and oncogenic signaling. Short term G-CSF induces HSC proliferation; however, this is limited to CD41⁺ cells with transient regenerative potential, whereas dormant HSCs are mobilized but not induced to proliferate.³¹ Accordingly, short term G-CSF–induced transcriptional changes, although indicative of cell cycle activation, are mild.³² In contrast, long-term G-CSF exposure reduces HSC repopulation.³³ Chronic G-CSF treatment of patients with severe congenital neutropenia (SCN) is associated with acquisition of truncating CSF3R mutations that activate Src kinase signaling and predate transformation to myelodysplastic syndrome/acute myeloid leukemia (AML), a frequent complication of SCN.^34,35 Recent work from the Touw laboratory showed that these mutations confer G-CSF hypersensitivity to normal, but not to SCN-derived, induced pluripotent stem cells, explaining why long-tern G-CSF does not lead to HSC exhaustion.³⁶ In contrast, membrane-proximal CSF3R mutations characteristic of chronic neutrophilic leukemia (CNL) activate JAK/STAT, similarly to physiological G-CSF signaling.³⁵ Thus, to avoid exhaustion, CNL leukemic stem cells (LSCs) may depend on additional mutations that activate self-renewal, explaining why CNL with isolated CSF3R mutations is rare.³⁷ 

MPN and normal HSCs inhabit an ecosystem containing their own progeny, cells of the innate and acquired immune system, nonhematopoietic cells, and noncellular components.^38-40 Elevated inflammatory cytokines, are characteristic of MPNs,^41-43 associated with various mutations, and contributed by bidirectional interactions between MPN and normal cells.^5,44-46 Differential sensitivity to specific cytokines is implicated in CH progression to clonal dominance. For instance, tumor necrosis factor α (TNF-α) signaling through TNF receptor 1 (TNF-R1) promotes expansion of mutant HSCs in Dnmt3a^R882H/+ transgenic mice.⁴⁷ TNF-α signaling through TNF-R2 promotes expansion of JAK2^V617F mutant human and mouse HSPCs.^5,48 To which degree normal HSCs survive in inflammatory MPN ecosystems differs between subtypes. In chronic myeloid leukemia (CML), BCR::ABL1 inhibitors can induce cytogenetic responses, even in long-standing disease, and chemotherapy is followed by mobilization of BCR::ABL1⁻ cells, but this is not seen in other MPNs.^49,50 Whether residual normal HSCs are irreversibly exhausted or can be resuscitated will define the limits of nontransplant therapies.

Registry studies found that prior infection is associated with a higher risk of myelodysplastic syndromes/AML⁵¹ and MPN.^51,52Tet2^+/+ mice that received transplantation with a mix of Tet2^−/− and Tet2^+/+ BM develop an incompletely penetrant MPN that is limited to those animals with demonstrable 16S bacterial rRNA gene in hematopoietic organs, suggesting host and environmental factors influence clonal expansion.⁵³ Autoimmune disorders also increase the risk for myeloid neoplasms, with the excess risk emphasizing entities without detectable antibodies, such as Crohn disease, hinting at cellular immunity as the driver.⁵² Given the wealth of genome wide association studies on autoimmune diseases, it may be of interest to mine this data for potential associations with myeloid malignancies.⁵⁴ 

MPN SCs exhibit impaired cell cycle control. For instance, transcriptional profiling revealed that, unlike their normal counterparts, quiescent CML LSCs are set to cycle.⁵⁵ The activity of TP53, a potent inducer of quiescence, is often reduced in the chronic phase of MPNs, with inactivating mutations limited to blast phase. Accordingly, chronic phase-CML LSCs are sensitive to combined disruption of p53 and Myc.^56,57 As forced cell-cycle entry due to knockout of cell cycle regulators such as Cdkn2 can cause HSC exhaustion, MPN SCs rely on mechanisms to fine-tune cell cycle control (Figure 2).⁵⁸ During the transition from homeostatic to emergency hematopoiesis and vice versa, the fundamental cell fate decisions between quiescence vs cell cycle entry and between self-renewal vs differentiation are recalibrated. To ensure long-term maintenance of a fully functional hematopoietic system, robust mechanisms preserve LT-HSC numbers and functionality.⁵⁹ Dormant HSCs are at the core of this highly resilient system. In homeostasis, most LT-HSCs are quiescent. Pulse-labeling studies using 5-bromo-2′-deoxyuridine or histone 2B (H2B)–green fluorescent protein revealed that the quiescent LT-HSC compartment comprises a large pool of non–label-retaining cells (NLRC) and a much smaller pool of label-retaining cells (LRCs) (Figure 3A).^60-62 Quiescent NLRCs are set to cycle, evidenced by upregulation of positive (eg, Cdk6) and downregulation of negative cell cycle regulators (eg, p57), whereas opposing patterns characterize LRCs. Upon cytokine stimulation, LRCs enter cell cycle with longer latency compared with NLRCs, indicating greater distance from cell cycle entry, a state referred to as dormancy. Dormant LT-LSCs undergo only 4 to 5 divisions over the life of a mouse, remember each division, and contain the entire long-term repopulation activity. Which mechanisms control this striking cell division limit is unknown.^61-63 Single-cell RNA sequencing demonstrated a continuum of transition states between NLRC and LRC, allowing response in proportion to signal strength (Figure 3B).⁶⁴ Numerous transcription factors are implicated in the regulation of quiescence and dormancy, of which p53, Foxo family, and Myc seem the most important (Table 1).

Late relapses in patients with AML or the fact that only ∼30% of patients with CML achieve treatment-free remission speak to the recalcitrance of dormant LSCs.⁹⁵ Understanding dormancy regulation may be central to understanding how LSCs survive in inflammatory environments. Although both dormant and quiescent HSCs are in G₀, metabolic properties distinguish dormancy from mere quiescence.

In contrast to NLRC, LRCs exhibit low oxidative phosphorylation (OXPHOS) and low glycolysis,⁹⁶ challenging the view that glucose usage within the LT-HSC compartment is universal.^97-99 Rather than glucose, dormant HSCs may use fatty acids, amino acids, and nucleic acids to meet energy needs.^100-103 HSCs with low mitochondrial membrane potential, consistent with low OXPHOS, have the highest repopulating potential.^96,104 

Protein synthesis in dormant HSCs is low and tightly regulated by intrinsic and extrinsic factors; increasing or decreasing protein synthesis compromises HSC repopulation capacity.¹⁰⁵ Several mechanisms are implicated in fine-tuning HSC proteostasis that converge on the translation machinery, with a central role for the PI3K/AKT/mTOR axis. HSCs null for the PI3K antagonist Pten exhibit increased protein synthesis but reduced long-term repopulation potential. Deletion of the mTOR activator Rictor in Pten^−/− HSCs normalizes protein synthesis, and improves repopulation, suggesting Pten deletion depletes HSCs by increasing protein synthesis.¹⁰⁶ Similarly, protein synthesis is reduced in mice with a hypomorphic mutation in the Rpl24 ribosome subunit (Rpl24^Bst/+). Rpl24^Bst/+ HSCs exhibit reduced engraftment that is restored by combining Rpl24^Bst/+ with Pten^−/−107. The transcription factor LAG1 inhibits translation initiation by 4E-BP through complex mechanisms, promoting HSC expansion.^105,108 Interestingly, some extrinsic signals elicit differential responses in HSCs and hematopoietic progenitor cells (HPCs), optimizing protection of HSCs, while maximizing differentiated cell output. Angiogenin, an RNase secreted by niche cells, is endocytosed by HSPCs and promotes generation of transfer RNA–derived stress-induced small RNA (tiRNA) in HSPCs but of ribosomal RNA in differentiated progeny. Consequently, protein synthesis is reduced in HSPCs but increased in differentiating cells, promoting HSC quiescence but myeloid progenitor proliferation.¹⁰⁹ Stressors, such as changes in redox balance or energy status, induce endoplasmic reticulum stress and the unfolded protein response.¹¹⁰ In response to endoplasmic reticulum stress, HSCs undergo apoptosis, whereas HPCs adapt and survive.¹¹¹ Together, these data support a critical function of proteostasis for HSC maintenance, and, by proxy, for regulation of dormancy.

A third metabolic pillar of HSC dormancy is autophagy, the transport of cytoplasmic materials to lysosomes for degradation and/or reuse. Deletion of autophagy genes such Atg7 promotes mitochondrial metabolism in quiescent HSCs, leading to cell cycle entry, differentiation, and exhaustion.^112-115 Cytosolic proteins that contain a KFERQ motif form complexes with HSP70, which interact with lysosomal-associated membrane protein 2 for translocation into lysosomes, a process referred to as chaperone-mediated autophagy.¹¹⁶ Chaperone-mediated autophagy promotes protein quality control and balanced energy metabolism, and its age-related decrease contributes to functional HSC decline.¹¹⁷ However, decline is not universal, as one-third of aged HSCs resemble young HSCs, with high autophagy, low metabolism, and robust regeneration potential.

Pathways governing metabolic responses to environmental cues, such as mTOR are often evolutionarily conserved, suggesting that important insights may be derived from other species responding to metabolic stress.¹¹⁸ Many aspects of dormancy resemble diapause, a program of suspended embryonic development used by selected species during unfavorable environmental conditions.^119,120 Diapause arrests the embryo at the blastocyst or even late embryonic stage, delaying development.¹²¹ In contrast to senescence in aging SCs, diapause is reversible. There are striking similarities between the factors regulating entry to, and exit from, dormancy and diapause, respectively. Deletion of both c-Myc and n-Myc from embryonic SCs reduces transcription, splicing, and protein synthesis, leading to a reversible growth arrest with retained pluripotency that transcriptionally resembles that of diapaused epiblasts.¹²² Similarities extend to PRC1 and PRC2, epigenetic master regulators implicated in hematopoietic development and LT-HSC maintenance. The PRC1 component BMI1 promotes HSC maintenance in postnatal tissues, partly by inhibiting the expression of p16^Ink4a and p19^Arf, tumor suppressors associated with senescence.¹²³ Hematopoietic-specific deletion of Bmi1 slowly depletes HSCs, possibly reflecting erosion of a quiescent HSC pool. Surprisingly, deletion of p16^Ink4a and p19^Arf failed to correct the defect, suggesting involvement of mechanisms other than senescence.¹²⁴ Indeed, Bmi1 deficiency increased protein synthesis, protein aggregation, and protein ubiquitylation independently of its effects on senescence regulators.¹²⁴ Mice with deletion of Eed, a core subunit of PRC2, fail to maintain LT-HSC, and develop BMF soon after birth.¹²⁵ These data parallel studies in African turquoise killifish (Figure 4), which have adapted to intermittent drying of their habitat by evolving desiccation-resistant eggs that remain in diapause in the dry mud for months but complete development within weeks once rains resume.¹²⁶ Diapause in killifish is associated with profound upregulation of CBX7 and EZH1, members of PRC1 and PRC2. The similarities between dormancy and diapause are striking but questions remain: is dormancy simply a low-energy state or an active metabolic program? What triggers the return from cycling to dormancy? Of particular interest in the context of this review, some cancer cells, including AML cells, survive chemotherapy by activating diapause-like programs.^127-129 In a critical experiment, mice were injected with cells from a barcoded colorectal cancer cell line. Tumors recurring after a temporary response to chemotherapy retained barcode complexity, which suggests that clonal survival is not genetically determined but either random or because of a specific functional state.¹²⁸ Given the widespread dysregulation of PRC in myeloid malignancies, similarities between dormancy and diapause may point to fruitful future studies to identify strategies to eliminate dormant LSCs.

Normal hematopoiesis can attenuate stress at multiple levels, including cytokine secretion, cytokine-receptor engagement, receptor endocytosis, signal amplification, and signal duration, allowing for a fine-tuned emergency response that protects HSCs. MPN SCs co-opt these mechanisms to avoid exhaustion, win clonal competition, and establish disease (Table 2). How can some HSCs and LSCs remain dormant in the presence of activating cytokines? Certain adhesion molecules and cytokines, such as N-cadherin, jagged-1, and angiopoietin-1 contribute to quiescence maintenance, but the most extensively studied quiescence enhancer is transforming growth factor β (TGF-β).^130-132 Canonical signaling downstream of TGF-β involves activation of SMADs, but additional components contribute to a system of bewildering complexity.¹³³ For instance, TGF-β signals to JNK through TAK1, a MAP kinase with a role in stress and inflammation signaling, and Tak1^−/− mice develop BMF.^134,135 Mice with hematopoietic-specific deletion of Tif1g, another noncanonical TGF-β target, develop a chronic myelomonocytic leukemia (CMML)-like MPN upon aging. This MPN, characterized by reduced LT-HSCs but increased ST-HSCs, MPPs, and granulocyte-macrophage progenitors, is transplantable, suggesting that maintained canonical TGF-β signaling is sufficient to avoid exhaustion. TIF1G is epigenetically silenced in one-third of human CMML, and its expression restored with hypomethylating agents.¹³⁶ Cytokine signaling can also be attenuated at the level of receptor endocytosis. The tetraspanin MS4A3 marks myeloid lineage cells and is induced upon HSC differentiation.¹³⁷ We discovered that MS4A3 promotes common β chain (βc) cytokine (IL-3/granulocyte-macrophage colony-stimulating factor) receptor signaling by escorting receptor endocytosis to enhance signaling.¹³⁸ Low MS4A3 in CML LSCs reduces response to βc cytokines, preventing exhaustion.⁵⁷ Similarly, lysosomal degradation of membrane receptors promotes LT-HSC quiescence, consistent with a critical role for the endo-lysosomal pathway.¹³⁹ In contrast, Myc-driven repression of lysosomal catabolism activates LT-HSCs, validating Myc’s role for switching between quiescence/dormancy and proliferation.¹³⁹ 

Downstream of stress signaling, transcriptional output orchestrated by chromatin modifiers, transcription factors, and RNA processors determines HSC fate between quiescence and cycling. Because each signaling pathway mobilizes multiple regulatory factors and vice versa, the decision between quiescence or self-renewal must reflect the sum of multiple signaling inputs.¹⁹² Somatic mutations in epigenetic modifiers such as TET2, ASXL1, IDH1/IDH2, EZH2, and DNMT3A are common in myeloid malignancies and have a profound impact on cell fate decisions and dormancy. As a comprehensive review of this subject is beyond the scope of this manuscript, we have focused on PRC1/2 to illustrate the importance of balancing effects on SC maintenance vs proliferation. Clearly, hematopoiesis has evolved multiple mechanisms to insulate dormant HSCs from extrinsic stimulation, ensuring a high threshold for cell cycle entry to protect the most valuable cells from unnecessary proliferation.

Differences in methylation levels and patterns cause stochastic transcriptional heterogeneity among genetically identical cells.^193,194 It is conceivable, but not experimentally proven, that this translates into differences in quiescence and repopulating potential that may insulate a portion of HSCs from stress.^195-197 HSC heterogeneity likely involves additional mechanisms of expression control (eg, RNA splicing), posttranslational modifications, and signaling oscillation. Heterogeneous states may explain divergent outcomes following HSC transformation by the same oncogene. For instance, NRAS^G12D is a CH mutation, a founder in CMML and AML, and a single driver in juvenile myelomonocytic leukemia, validating its ability to drive clonal expansion.^198,199 However, NRAS^G12D alone causes senescence, raising the question of how NRAS^G12D-expressing HSCs circumvent extinction.²⁰⁰ Label retention studies showed that Nras^G12D expression can either increase proliferation and reduce self-renewal or the exact opposite,²⁰¹ suggesting that expansion of an NRAS^G12D mutant clone may depend on a specific cellular state. Can LSCs transition between states that do or do not permit access to dormancy? If this were the case, it would explain why barcode heterogeneity is preserved in some cancer cells persisting during chemotherapy.^127,129 Characterizing the state from which leukemic cells access dormancy may inform new therapeutic strategies.

Most reported spontaneous remissions of solid tumors were associated with infection and attributed to pathogen-stimulated antitumor immunity.²⁰² The rare spontaneous remissions of established MPN involve various genotypes, including BCR::ABL1, FLT3^ITD, and JAK2^V617F, and likely reflect involvement of host immunity.^203-207 However, the detection of BCR::ABL1 transcripts in the blood of healthy volunteers and the subsequent discovery of CH are evidence that mutant clones frequently arise.^208,209 It is intriguing to speculate that many more mutant clones become exhausted before they reach detectability. Advances in simultaneous single-cell genotyping and phenotyping may make this question accessible to experimental interrogation.

Dormant MPN SCs are intrinsically resistant to therapies aimed at proliferating cells. Some effective drugs may induce quiescence, the best example being tyrosine kinase inhibitors in CML, which not only fail to eliminate, but increase, quiescent primitive CML cells.^210-212 Can we use our understanding of quiescence to develop new therapies? As a starting point we propose to consider MPN hematopoiesis as an ecosystem thrown out of balance by an invasive species. Although eradicating the invaders remains the most desirable outcome, in complex myeloid neoplasms, a more realistic goal may be to rebalance the system to maintain function by weakening invasive species, strengthening resident species, or both.²¹³ In the last section we will review several possibilities that might accomplish this.

Because elevated inflammatory cytokines are an MPN hallmark, numerous cytokine-mediated interactions exist in the MPN ecosystem (Figure 5), but it is largely unknown which of these are critical for MPN progression. The most compelling evidence may exist for TNF-α. MPN cells expressing JAK2^V617F or BCR::ABL1 generate TNF-α, which inhibits HSCs but is neutral to, or even stimulates, MPN SCs.^5,178 Although in JAK2^V617F-positive MPN, the signal is transduced through TNF-R2,⁴⁸ TNF-R1 is implicated in Dnmt3a^R882H/+-driven CH,⁴⁷ suggesting that the effect of blocking either TNF-R1 or TNF-R2 may depend on the genetic context. Context dependency extends to the biological setting. In mice with chronic MAI infection, Dnmt3a^−/− HSCs outcompete Dnmt3a^+/+ cells in an IFN-γ–dependent manner, suggesting that blocking IFN-γ signaling may suppress mutant clones in the setting of chronic infection.^27,47 As an emergency signal, IL-1β enhances HSC division and differentiation but chronic IL-1β exposure reversibly reduces HSCs.²¹⁴ A retrospective analysis of a cardiovascular prevention trial of the IL-1β–blocking antibody canakinumab showed the greatest benefit in patients with TET2-mutated CH.²¹⁵ Because IL-1β secreted by Tet2^−/− monocytes promotes AML growth, canakinumab may restrain MPN cells.^216,217 Clearly, at this point we are only scratching the surface of the complex cytokine networks in myeloid malignancies.

Another angle may be to disrupt MPN SCs’ insulation from cell cycle–inducing signals to drive them out of dormancy. This is nothing new, because cytokine priming as a strategy to mobilize quiescent LSCs into cycle has been evaluated in clinical trials of AML and CML, with limited success.^218,219 If studies on normal hematopoiesis in mice are predictive, these interventions may not have reached dormant LSCs that have adapted to an inflammatory, cytokine-rich environment. Exploiting the specific mechanisms used by LSCs may be more promising. For instance, small-molecule inhibitors of transcription factor TFEB may disrupt self-insulation by restoring surface expression of cytokine receptors on LSCs.¹³⁹ TGF-β antagonists such as luspatercept may restore responsiveness to cytokines.^139,220 We have shown that delivery of MS4A3 protein to CML CD34⁺ cells promotes differentiation and reduces colony formation.¹³⁸ If induction and maintenance of dormancy are active processes, this will offer therapeutic opportunities. For instance, dormancy in AML cells is vitamin A dependent, and antagonizing vitamin A may drive dormant AML LSCs into cycle.⁶⁴ Dormant LSCs may rely neither on glycolysis nor on OXPHOS, but may be dependent on amino acids.²²¹ Understanding how this metabolic program is initiated could identify vulnerabilities. If dormancy in hematopoiesis is induced by master switches, in analogy to PRC1 regulation of diapause in killifish, then identifying stress sensors and pathways that transmit the dormancy command may inform strategies to eliminate dormant LSCs.

IFN-α is active in polycythemia vera (PV), essential thrombocythemia, and early-stage myelofibrosis, and in the pretyrosine kinase inhibitor era, was the only nontransplant therapy known to induce cytogenetic responses in CML.^222,223 IFN-α drives dormant HSCs into cycle, suggesting that cytogenetic responses may reflect mobilization of normal hematopoiesis.^22,224 Indeed, this is supported by genetic mouse models, as well as studies in primary MPN cells.^23,225,226 Whether clinical responses to IFN-α indeed reflect mobilization of normal HSCs or specific suppression of MPN cells remains debatable. Recent work identified a pathway involving PKCδ-dependent phosphorylation of ULK1 with subsequent activation of p38 MAPK as essential for IFN-α–suppressive effects on malignant erythroid precursors from patients with MPN. Supporting the clinical relevance of this observation, ULK1 and p38 MAPK levels positively correlate with response to IFN-α therapy in humans.²²⁷ 

There are more opportunities for interventions aimed at rebalancing the MPN ecosystem, such as strategies targeting interactions between MPN SCs and the BM stroma. Success may depend on the fitness of remaining normal HSCs but we know little about these cells. For instance, are they epigenetically reprogrammed after spending years in an inflammatory environment? If so, is this program reversible? Directing more research at the state of residual HSCs may be fruitful.

Few would disagree that eradicating the malignant clone is the most desirable outcome of MPN therapy. However, the realization that CH is widespread and universal in older individuals has blurred the boundary between physiological age–associated processes and malignancy. Oligoclonality is prevalent in many aging tissues and could be seen as a mechanism that mitigates aging-related functional decline of tissue SCs. In the case of CH, the situation is further complicated by the discovery of a seemingly limitless number of medical conditions whose incidence, clinical course, or responsiveness to therapy are affected by CH, although studies with limited size must be interpreted with caution.²²⁸ More than any other organ, hematopoiesis is repeatedly subjected to extreme proliferation stress, forcing evolution to develop effective strategies to protect HSCs from damage and exhaustion. As MPNs must solve the same problem, it does not come as a surprise that MPN SCs use mechanisms used by normal HSCs to avoid exhaustion. Eliminating dormant MPN SC while sparing their normal counterparts will be challenging and may require targeting state transitions rather than the state of dormancy itself. For instance, preventing LSCs stressed by therapy from transitioning into dormancy may eliminate residual leukemia. In fact, maintaining MYC expression in AML cells treated with chemotherapy may accomplish this.¹²⁷ A deeper understanding of the mechanisms triggering state transitions should reveal additional options. In the meantime, rebalancing the perturbed ecosystem of MPN hematopoiesis may be a more achievable goal, and this may include avoiding interventions that promote the outgrowth of clinically aggressive clones by suppressing targetable but nonlethal clones.²²⁹ Clinical management of myeloid malignancies is likely to become more complex.

Given the limited space, many relevant publications could not be included, and the authors apologize to those authors whose important work is not cited. Figures were generated using BioRender.

This work was supported in part by National Institutes of Health grants R01CA268496, R01CA257602, and R01CA254354 (M.D.).

Contribution: H.G.Z. and M.D. reviewed literature, designed the structure of the review, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Michael Deininger, Versiti Blood Research Institute, 8727 W Watertown Plank Rd, Milwaukee, WI 53226; e-mail: mdeininger@versiti.org.