Clonal hematopoiesis and its progression to myeloid neoplasms: insights into risk, biology, and therapeutic strategies

Abstract

Clonal hematopoiesis (CH) is defined by the clonal expansion of hematopoietic stem and progenitor cells harboring somatic mutations that confer a fitness advantage. CH is common with advancing age and becomes nearly ubiquitous in middle age. Although typically asymptomatic, CH is associated with an increased risk of hematologic malignancies particularly myeloid neoplasms (MN), diverse non-malignant conditions, and all-cause mortality. Over the past decade, research has provided major insights into the origins of CH. In addition to aging, CH is promoted by environmental exposures, inherited genetic predisposition, and acquired conditions. Large-scale population and longitudinal sequencing studies have identified determinants of clonal behavior. Characterization of the natural history of CH has enabled the development of risk stratification models to identify individuals with CH at high risk of progression to MN, thereby providing a rationale for selecting patient populations best suited for therapeutic intervention trials. Emerging strategies include targeting mutation-specific vulnerabilities, modulating inflammatory pathways, reducing genotoxic therapy-induced clonal selection, and repurposing agents with efficacy in MN. In this review, we summarize current knowledge of the risk factors underlying CH development, highlight recent advances in understanding the determinants of clonal behavior including progression to MN, and discuss emerging therapeutic approaches for preventing malignant transformation and clinical trial design considerations.

Introduction

Clonal hematopoiesis (CH) is characterized by the clonal expansion of hematopoietic stem and progenitor cells (HSPC). These clones arise through the acquisition of somatic DNA alterations, which confer a selective advantage and enable clonal outgrowth. The process of normal stem cells acquiring somatic mutations that modify fitness is pervasive across human tissues and increases with advancing age.^1-3 A variety of DNA alterations can drive CH including point mutations, insertions-deletions (indels), and large-scale chromosomal changes (gains, losses, and copy-neutral loss of heterozygosity); the latter of which are termed mosaic chromosomal alterations.^4,5 Most cases of CH are caused by point mutations in a small number of genes that are recurrently mutated in myeloid neoplasms (MN).^6-8 These include genes involved in epigenetic modification (DNMT3A, TET2, ASXL1), DNA damage response (DDR; TP53 and PPM1D), RNA splicing (SRSF2, SF3B1, U2AF1), and signal transduction (JAK2). CH is ubiquitous with aging and while typically clinically silent, is associated with an increased risk of progression to hematologic malignancies, including MN, as well as a variety of non-malignant adverse health outcomes which collectively result in increased allcause mortality.^6,7 CH is a premalignant state, analogous to monoclonal gammopathy of undetermined significance, which offers a model for understanding the early stages of carcinogenesis in humans. Importantly, CH provides unique opportunities to study clonal dynamics and somatic mutation acquisition in humans since the peripheral blood allows for non-invasive sampling of the entire hematopoietic stem cell pool, which is not possible for other organ systems. This review summarizes current insights into the causes of CH, discusses the factors that drive progression of myeloid CH to MN, and highlights potential therapeutic approaches to prevent progression to MN. It focuses primarily on myeloid CH (commonly referred to as simply CH) driven by somatic mutations, as our understanding of how mosaic chromosomal alterations - which typically involve multiple genes - contribute to progression to MN remains limited. Because the molecular mechanisms underlying many common myeloid CH mutations have been thoroughly reviewed recently,⁹ this review briefly summarizes the most relevant pathways to contextualize the clinically important features of CH.

Key concepts and nomenclature of clonal hematopoiesis

The nosology of CH was comprehensively summarized in a recent excellent review by Weeks and Ebert.¹⁰ Herein, we summarize the key concepts and definitions related to CH nosology that are fundamental to understanding the role of CH in hematologic malignancies.

Humans have an estimated 20,000 to 200,000 HSPC that contribute to hematopoiesis.¹¹ With clonal expansion, the proportion of peripheral blood cells arising from a single HSPC increases and can be quantified by measuring the variant allele fraction (VAF) of the somatic alterations harbored by the clone. CH mutations with a higher VAF are associated with an increased risk of hematologic malignancies and other adverse health outcomes. A minimum peripheral blood VAF threshold of 2% was used to define CH in early studies, a cutoff based on the sequencing error rate of traditional (non-error corrected) Illumina exome sequencing data.^6,7 This definition ultimately led to the term clonal hematopoiesis of indeterminate potential (CHIP), which is defined as CH with a somatic mutation in a MN driver gene with a VAF ≥2% in an individual without hematologic cancer or blood count abnormalities. While CH can now be detected at a VAF of ≥0.001 (0.1%) using error-corrected sequencing methodologies,^12-15 our understanding of the biological relevance of alterations with VAF <2%, termed micro-CH by some,¹⁶ is limited even today.

CH can be driven by diverse alterations with specific genetic drivers influencing both CH biology and outcomes. For example, CH driven by alterations in known myeloid driver genes (which is referred to as myeloid CH) is associated more strongly with an increased risk of MN.^6,7 Conversely, CH driven by alterations in lymphoid drivers (referred to as lymphoid CH) is more strongly associated with risk of lymphoid neoplasms.¹⁷ Myeloid CH is clearly more common than lymphoid CH largely due to highly recurrent mutations in DNMT3A, TET2, and ASXL1.^17-19 The distinction between these classes of CH is complicated by the fact that somatic mutations in many genes can drive both myeloid and lymphoid neoplasms (e.g., TP53, TET2, RUNX1, SF3B1, IDH2).²⁰ We found that 6.2% of individuals in the UK Biobank had somatic mutations in hematologic malignancy driver genes, with 4.6% occurring in myeloid only, 0.4% in lymphoid only, and 1.5% in both myeloid and lymphoid genes.²¹ Like hematologic cancer, multiple classes of somatic events can drive CH. This includes single nucleotide variants, indels, and large-scale copy number events including mosaic chromosomal alterations. Attributing mosaic chromosomal alterations, which typically involve multiple genes, to a binary myeloid versus lymphoid classification is challenging. In aggregate, the prevalence of autosomal mosaic chromosomal alterations was lower than that of somatic CH mutations (3.2 vs. 6.2%) among individuals in the UK Biobank.²¹ Furthermore, estimating the true prevalence of somatic copy number events is challenging, as the sensitivity for detecting even moderately sized copy number alterations is substantially lower than for single nucleotide variants at comparable sequencing depth and remains limited when using single nucleotide polymorphism arrays. The extent to which complex re-arrangements and fusion events (which are common drivers of hematologic cancer) might also drive CH is unclear due to difficulty in detecting these classes of events at a low VAF.

The evolution of myeloid CH to MN is summarized in Figure 1. Numerous features have been identified that contribute to the risk of CH progression (e.g. high-risk mutations, increased mutation number and VAF, ineffective erythropoiesis, cytopenias) and have been used to develop MN prediction models.^22,23 These are covered in detail later in this review. Myeloid CH with a persistent explained cytopenia and without dysplastic features meeting criteria for myelodysplastic syndrome (MDS) is termed clonal cytopenia of undetermined significance (CCUS).^24,25 Individuals with CCUS have a significantly elevated risk of progression to MN compared to those with CH without cytopenias or idiopathic cytopenia of undetermined significance.²⁶

It is common to detect CH without identifying a somatic mutation or chromosomal alteration in a recognized hematologic malignancy driver gene.^6,11,27 Using distinct analytical approaches and large cohorts, Genovese et al.⁶ and Zink et al.²⁷ found that 39% and 59% of CH cases, respectively, lacked a known driver. Importantly, CH without a known driver carried a risk of hematologic malignancy comparable to CH with an established driver.^6,27 Mitchell et al.¹¹ further confirmed the commonness of CH without a driver using an orthogonal approach, whole-genome sequencing of 3,579 single cell-derived HSPC colonies from ten donors across the human lifespan. Similar observations have been made in MDS, in which ~6% of patients lack genomic alterations in established MDS driver genes.²⁸ Potential explanations for CH with unknown drivers include the presence of novel driver genes (which would likely be relatively uncommon), mutations in non-coding regions (e.g., enhancers), mutations difficult to detect at a low VAF (e.g., complex rearrangements), genetic drift (i.e., random chance), and technical artifacts. Together, these findings highlight the need to better understand the biological underpinnings of CH lacking a known driver and its relationship with the progressive hematopoietic oligoclonality that is pervasive with aging.

Causes of clonal hematopoiesis

The modern concept of carcinogenesis, originally developed through the seminal work of Carl Nordling, Peter Armitage, and Richard Doll in the 1950s,^29,30 posits that cancer arises through the stepwise acquisition of genetic alterations in cells over time. Elucidation of the molecular events underlying carcinogenesis was first demonstrated in colon cancer by Fearon and Vogelstein in 1990 through analysis of biopsies obtained during colonoscopy, with publication of the “Vogelgram” model outlining specific genetic alterations associated with progression of normal colonic epithelium to adenoma and eventual carcinoma.³¹ Subsequent studies using massively parallel sequencing have since characterized the genomic landscape of numerous cancer types, providing insight into the genetic alterations driving specific tumor subtypes.³² The discovery of CH provided critical insights into the earliest stages of hematologic malignancies in humans through analysis of easily accessible peripheral blood samples. Furthermore, the presence of pre-existing large biobanks with linked genetic and longitudinal health data, such as the UK Biobank,³³ has fueled the rapid acquisition of knowledge on factors that influence the initiation and progression of hematologic malignancies, particularly MN.

Intrinsic factors

Age is the strongest risk factor for the development of CH,^6,7 reflecting the fact that age is the principal driver of accumulation of somatic mutations across human tissues.^1,34,35 The rate of somatic mutation acquisition is enhanced in highly proliferative tissues such as the human bone marrow, which produces approximately 200 billion red blood cells and 100 billion neutrophils daily to maintain blood homeostasis.³⁶,³⁷ With each cell division, somatic mutations are inevitably introduced despite the high fidelity of human DNA replication due to the inherent challenge of correctly copying a 3.2 billion base-pair genome. Somatic mutations can serve as a biological clock, marking the passage of time and the accumulation of genetic changes within HSPC. By age 60 years CH is common, with approximately 10% of individuals having CH detectable at a VAF ≥2% and 75% at a VAF ≥0.1%.³⁸ The somatic mutations that drive CH can occur early in life, including in utero, and typically take decades to manifest clinically.^39,40

Figure 1.Evolution of clonal hematopoiesis to myeloid neoplasms. Hematopoietic stem cells acquire somatic mutations throughout life. Infrequently, these mutations confer a selective growth advantage, resulting in clonal expansion, which is termed clonal hematopoiesis (CH). Several risk factors for CH have been identified, many of which preferentially select for specific mutations. Common routes of progression from CH to specific myeloid neoplasms are shown and are driven by the acquisition of additional cooperating mutations. CMUS: clonal monocytosis of undetermined significance; CMML: chronic myelomonocytic leukemia; MPN: myeloproliferative neoplasm; CCUS: clonal cytopenia of undetermined significance; MDS: myelodysplastic syndrome; AML: acute myeloid leukemia.

While the overall prevalence of CH is similar between males and females, sex influences the frequency of CH driven by alterations in specific genes. Males are more likely to have mutations in splicing factor genes (e.g., SRSF2, SF3B1, U2AF1) and ASXL1; and less likely to have DNMT3A mutations.^41-43 Additionally, males are more prone to mosaic loss of the Y chromosome than females are to mosaic loss of the X chromosome (20% vs. 5% in the UK Biobank), and do not develop mosaic loss of the X chromosome.⁴⁴ These sex-based differences persist after controlling for basic potential confounders (e.g., smoking rates). The CH mutational spectrum observed in males is over-represented for genes that confer an elevated risk of MN, consistent with the increased incidence of hematologic malignancies observed in males across ages.^45,46 Our understanding of the mechanisms underlying the sex-based differences in CH and hematologic malignancies are poorly understood. Potential contributors to increased risk of hematologic malignancy in males include differential carcinogen exposure, presence of tumor suppressor genes on the X chromosome that can escape X-inactivation,⁴⁷ and hormonal impacts on HSPC biology and leukemic potential.⁴⁸ These data collectively highlight an important and understudied biological role of sex on MN pathogenesis beyond differential environmental exposures, which warrants further investigation.

Environmental exposures

While smoking is moderately associated with multiple individual CH genes, including SRSF2, SF3B1, and DNMT3A, it is strongly associated with ASXL1 mutations.^41,49 The mechanism underlying the association between ASXL1 mutations and smoking is not understood. Most ASXL1 mutations in CH and MN are caused by heterozygous truncating (frame-shift indels and nonsense) mutations in the C-terminus, which allows for escape from nonsense-mediated decay.⁵⁰ Mutational signature analysis can identify characteristic patterns of somatic mutations induced by distinct mutational processes (e.g., tobacco carcinogens, ultraviolet light) through characterization of base substitution subtypes and their trinucleotide context (bases 5’ and 3’ of the mutated base). This has been used to show that smoking is associated with increased base substitutions and indels in tissues with and without direct exposure to tobacco smoke.⁵² However, the trinucleotide context of CH drivers occurring in smokers and non-smokers is largely similar and predominantly driven by the age-related mutation signatures (e.g., SBS1).⁵³ It is unclear whether carcinogens in tobacco preferentially induce ASXL1 mutations, or whether smoking simply confers a fitness advantage to HSPC with ASXL1 mutations.

Genotoxic therapies, including cytotoxic chemotherapy and radiation therapy, are associated with CH. These therapies induce DNA damage in HSPC and create a selection pressure for clones with genetic alterations in DDR genes including TP53, PPM1D, and CHEK2.⁴⁹ This process is likely mostly driven by selection for pre-existing clones with DDR mutations, rather than generation of de novo mutations by genotoxic therapy, but distinguishing between these two possibilities is challenging.⁵⁴ However, some CH mutations may be induced by genotoxic therapy. A study of CH in long-term survivors of pediatric cancers (median follow up 32 years since diagnosis) identified recurrent STAT 3 mutations only among individuals with a history of Hodgkin lymphoma.⁵⁵ Single-cell whole-genome sequencing on peripheral blood from three individuals with STAT 3 mutations revealed that S TAT 3 -mutant hematopoietic cells (and not wild-type cells) contained an ~3.5 fold increased mutation burden, and were strongly enriched for the mutational signature COSMIC SBS25, which occurs due to exposure to procarbazine, an alkylating drug used in Hodgkin lymphoma.⁵⁵ Cytotoxic chemotherapy regimens containing topoisomerase II inhibitors, platinum agents, alkylating agents, and bleomycin are most strongly associated with DDR CH,^49,55 consistent with the known risk of these agents for secondary MN. Higher radiation therapy doses are associated with increased risk of CH, with a stronger effect observed for DDR CH.⁵⁶ DDR CH was significantly associated with specific anatomic sites of radiation therapy delivery, including the head and neck, pelvis, brain, and thorax.⁵⁶ Interestingly, these sites account for ~65% of active bone marrow tissue in adults,⁵⁷ suggesting that radiation-induced CH expansion may be linked to the total dose applied to the hematopoietic compartment.

Genetic predisposition

Inherited genetic variation has a well-established role in predisposing individuals to CH and modifying clonal behavior. Inherited bone marrow failure syndromes (IBMFS) and familial acute myeloid leukemia and/or myelodysplastic syndromes (AML/MDS) are caused by rare pathogenic germline variants in genes with critical roles in hematopoiesis and have been identified as strong contributors to inherited CH risk at early ages, in addition to their well-established roles in MN predisposition.⁵⁸ While the pathophysiology of IBMFS is variable, these syndromes are united by ineffective hematopoiesis due to impaired HSPC function, which creates a unique selection pressure for somatic clones with a relative fitness advantage to relieve the germline fitness constraints of the specific IBMFS. In these settings, CH arises through two principal mechanisms: (i) somatic normalization – adaptive mutations that compensate for the defect by reversion (the germline defect is corrected directly) or compensation (the germline defect is corrected indirectly), while leaving tumor-suppressor pathways intact; and (ii) somatic transformation – maladaptive mutations that subvert tumor-suppressor pathways, conferring an elevated risk of malignant transformation.⁵⁸ For example, in Shwachman-Diamond syndrome, which involves defects in ribosomal assembly due to loss-of-function mutations in SBDS, somatic normalization results from alterations that partially restore ribosome function through EIF6 disruption (via inactivating mutations or deletion 20q) or partially correct the SBDS mutation (via isochromosome 7q).⁵⁹ In contrast, somatic transformation occurs as a consequence of TP53 inactivation, resulting in uncoupling of ribosomal stress from activation of cellular senescence pathways (without correction of the ribosome function defect) and ultimately increased risk of transformation to MN. In dyskeratosis congenita, which is caused by loss-of-function mutations in genes that regulate telomere maintenance (e.g., TERT), mechanisms of somatic normalization include direct reversion, mutation of the unaffected wild-type TERT promoter (thereby leading to increased telomerase activity), and POT1 loss-of-function mutations (facilitating telomere elongation).⁶⁰ TP53 inactivation and chromosome 7 loss are the major causes of somatic transformation in dyskeratosis congenita.⁶¹

Similar to IBMFS, the unique selection pressures introduced by familial AML/MDS disorder result in the development of specific profiles of CH mutations. Individuals with germline pathogenic variants (without a hematologic malignancy) in RUNX1 and GATA2 had high rates of CH (35% and 22%, respectively) at all ages, whereas individuals with pathogenic DDX41 germline variants had a low CH prevalence (3%).⁶² In addition, RUNX1 germline variant carriers had a unique CH mutational spectrum with a high frequency (42%) of BCOR variants. These findings collectively underscore the importance of integrating germline genetics into our understanding of CH biology. Such integration may improve risk stratification, inform surveillance strategies for individuals with hereditary predisposition to hematologic malignancies, and provide opportunities for mechanistic studies that explore how inherited variation shapes the evolutionary landscape of somatic hematopoietic clones.

Genome-wide association studies using large biobanks with linked phenotypic data have identified several common germline loci that increase susceptibility to CH, both overall and in specific genes. Most of the genes associated with CH globally have known roles in telomere biology, DDR, hematologic malignances (e.g., TET2, SETB1, RUNX1, ETV6, GATA2, PTPN11A, MPL), or HSPC biology (e.g., CD164, LY75, SMC4).^19,21,63,64 Among these, variants at the TERT locus associated with increased telomerase activity and telomere length consistently show the strongest association with overall CH risk.^27,64 Mendelian randomization analysis supports a causal role of increased telomere length on CH risk.⁴¹ In line with this, a small family-based study found that germline heterozygous loss-of-function mutations in POT1, a telomere maintenance gene, confer a high risk of increased telomere length and CH risk, as well as a diverse spectrum of benign and malignant neoplasms.⁶⁵ Rare variants in CTC1, another important regulator of telomeres which has been implicated in dyskeratosis congenita, have been associated with global CH risk.⁶³ These findings collectively demonstrate a critical role for telomere biology in CH pathogenesis, suggesting that reduction of normal telomere shortening with aging promotes the development of CH.

Common germline variants in several DDR genes also predispose individuals to CH overall, including genes associated with well-established cancer predisposition syndromes (e.g., CHEK2, ATM, PARP1, TP53).^41,63 In addition, our group recently identified 22 new CH predisposition genes (20 CH gene-specific),²¹ 14 of which have recognized roles in the DDR (ATR, RAD51D, FANCI, NBN, RTEL1, ERCC1, ERCC2, ERCC3, ERCC4, MUTYH, NTHL1, LIG4, ERCC6L2, PRDM9).

Previous studies have also identified common germline variants that drive CH in specific genes.^19,63,64,66 Most notably, a promoter variant at TCL1A associated with decreased TCL1A expression predisposes to DNMT3A CH, but is protective from CH driven by other genes including TET2, ASXL1, SF3B1, and SRSF2.⁴¹ This difference was driven by the differential impact of the allele on CH expansion rate, resulting in slower growth of non-DNMT3A CH.⁶⁷ In addition, functional studies showed that TCL1A drives HSPC clonal expansion and that its expression is induced by introduction of mutations in TET2 or ASXL1, but not DNMT3A. The results implicate TCL1A as a key mediator of the fitness advantage of many commonly mutated CH genes. Similarly, a variant in the CD164 locus, which is involved in HSPC migration, is associated with DNMT3A and ASXL1 CH, but not TET2 CH for reasons that remain unclear.^41,63

Overall, the link between germline variants and gene-specific CH predisposition appears to be driven primarily by the influence of germline variants on clonal fitness rather than direct mutagenesis, as mutational signatures were largely comparable between carriers and non-carriers, and dominated by the age-related SBS1 signature.²¹ However, longitudinal studies comparing CH evolution in germline carriers and non-carriers would be needed to clarify the relative contribution of mutational acquisition versus expansion.

Acquired disorders

Aplastic anemia is a bone marrow failure disorder characterized by pancytopenia and bone marrow aplasia. Acquired forms of the disorder are usually caused by autoimmune destruction of HSPC and associated with CH in approximately 50% of patients.⁶⁸ The genes most frequently involved with somatic mutations in acquired aplastic anemia include DNMT3A, PIGA, ASXL1, BCOR, and BCORL1. The unique enrichment of PIGA, BCOR, and BCORL1 somatic mutations in acquired aplastic anemia may support their role in escape from immune-mediated destruction, although the mechanisms are not well characterized.

Drivers of progression of clonal hematopoiesis to myeloid neoplasms

Clonal evolution

Over the past 5 years our understanding of the determinants of clonal behavior and their influence on progression to MN has advanced substantially. Studies have employed diverse approaches to characterize clonal dynamics including serial measurement of CH through longitudinal aging studies,⁶⁹ phylogenetic analysis through sequencing of individual HSPC clones,^11,69 and clonal growth estimation from single timepoint data using passenger mutation analysis.⁶⁷ These complementary methodologies have yielded several broad insights across CH driver mutations. CH clones generally expand at an accelerated rate earlier in life, followed by a slower, steady exponential rate during older age (at least 55 years old).⁶⁹ Growth rates vary considerably between different CH mutations, with some mutations exhibiting age-dependent variability in growth.⁶⁹ Larger CH clone size, higher clonal growth rates, and a greater number of CH mutations are all associated with an increased risk of progression to MN.^7,69 Finally, CH mutations are typically acquired decades before manifesting clinically.¹¹

Specific CH driver mutation genes play a central role in dictating clonal behavior, including both the risk and type of progression to MN. DNMT3A CH generally has a low clonal growth rate (~5% per year), though there is marked growth variation by age; faster at younger ages and slower at older ones.⁶⁹ The risk of progression to MN with DNMT3A mutations is relatively low and is driven primarily by an increased risk of AML.⁴¹ In contrast, TET2 mutations exhibit an intermediate clonal growth rate (~10% per year) which is stable across all ages, with one study suggesting TET2 overtakes DNMT3A as the most common CH mutation in individuals over 75 years old.⁶⁹ TET2 CH confers an intermediate risk of progression to MN, particularly MDS and chronic myelomonocytic leukemia.⁴¹ Despite having opposite effects on DNA methylation, loss-of-function mutations in TET2 and DNMT3A both drive CH clonal expansion. ASXL1 mutations also demonstrate intermediate growth rates and are strongly associated with progression to AML and MDS risk.^41,69 Mutations in IDH1 and IDH2 are less common, have an intermediate-to-high growth rate (which is stable over time), and are strongly associated with risk of progression to AML.^22,69 In addition, IDH1/2 mutations are largely mutually exclusive with each other and TET2 mutations in MN (reflecting a shared pathogenic mechanism discussed later in this review),⁷⁰ which has been validated using single-cell DNA-sequencing approaches to reconstruct phylogenetic trees of the major driver clones in AML cases.^71,72

Mutations in splicing factor genes (including SRSF2, SF3B1, and U2AF1) are rare before the age of 50 years old, but when present at older ages are associated with rapid clonal growth rates (15-20% per year).⁶⁹ The reasons for their late onset are incompletely understood, but likely reflect age-related changes in selection pressures, rather than a unique restriction of these mutations to older individuals. Recent work has shown that splicing factor mutations promote CH expansion in part by mitigating telomere shortening in HSPC with aging.⁷³ Investigating the molecular basis of this phenomenon has been difficult because current in vitro and in vivo models of splicing factor mutations often show reduced clonal fitness, the opposite of what is observed in humans.⁷⁴ Splicing factor mutations are typically mutually exclusive because their co-expression results in synthetic lethality.⁷⁵ Splicing factor-driven CH has a high risk of progression to all MN subtypes.⁴¹ Interestingly, SF3B1 CH is associated with a favorable prognosis upon progression to MDS,⁷⁶ but an adverse prognosis upon progression to AML,⁷⁷ highlighting the context-dependent effects of specific splicing factor mutations.

While CH clones harboring the JAK2 V617F hotspot mutation generally have intermediate growth rates, they are unique in demonstrating unpredictable growth dynamics at older ages.⁶⁹ In contrast, mutations in other MPN drivers (CALR and MPL) are less frequently observed in CH, consistent with their lower prevalence in MPN.⁴¹ The canonical MPN driver mutations (JAK2, CALR, MPL) are mutually exclusive and exert strong influence on the risk of developing specific MPN subtypes. JAK2 V617F mutations are associated with all MPN types including polycythemia vera, primary myelofibrosis, and essential thrombocythemia, whereas CALR mutations are restricted to essential thrombocythemia and primary myelofibrosis, and MPL mutations to essential thrombocythemia. JAK2 V617F VAF significantly influences MPN subtype distribution: higher VAF (>50%), often reflecting homozygous mutations due to copy-neutral loss of heterozygosity, are strongly associated with polycythemia vera and myelofibrosis, whereas lower VAF, consistent with heterozygous mutations, are more commonly observed in essential thrombocythemia.⁷⁸ The factors that influence the progression of MPN driver mutant CH to specific MPN subtypes remain poorly understood.

The growth of DDR-mutated CH is generally modest in the absence of genotoxic therapy-induced selective pressure. TP53 mutations are typically associated with a low clonal growth rate, which further decelerates with aging, whereas PPM1D-mutant CH demonstrates an intermediate growth rate.⁶⁹ Both TP53- and PPM1D-mutant clones are linked with a relatively average risk of progression to MN. This may help to explain why MN uncommonly occur in patients with Li Fraumeni syndrome (pathogenic germline TP53 variants) in the absence of prior genotoxic therapy.^22,23,79,80 In contrast, exposure to cytotoxic chemotherapy or radiation therapy drives expansion of TP53- and PPM1D-mutant CH clones and increases the risk of MN.^49,54,81 The selection advantage provided by somatic TP53 mutations is not restricted to genotoxic therapies, but is also observed with lenalidomide⁸² and MDM2 inhibitors.⁸³

Although NPM1 and FLT3 are common driver mutations in de novo AML, they are rarely detected in CH. In the UK Biobank cohort, only two out of ~200,000 individuals were found to have a NPM1 driver mutation with normal blood counts.⁸⁴ Both individuals developed AML within 6 months of the sample collection. The near lack of FLT3 and NPM1 mutations in CH studies may reflect a rapid progression to de novo AML.

Factors other than CH mutations play an important role in governing clone behavior and include both cell-intrinsic and extrinsic factors. The relevance of non-mutational factors is highlighted by several observations. Low frequency CH mutations (micro-CH) are ubiquitous with aging (>95% after 50 years old), however only a small minority of these clones expand to meet criteria for CHIP (VAF of 2%).¹³ Parallel observations have been made in other somatic tissues including the skin and esophagus, which have frequent oncogenic mutations, but low overall rates of progression to carcinoma.^2,3 The same driver mutations have variable growth rates between different clones.^67,69 Expanded clones lacking known drivers are not uncommon, particularly at advanced ages.²⁷ While acquisition of somatic alterations is typically involved in driving progression to MN, a large proportion of JAK2 CH that progresses to MPN⁸⁵ and a subset of SF3B1 CH to MDS⁸⁶ without the acquisition of new somatic alterations. Cell-intrinsic factors that govern clonal behavior beyond individual CH mutations are diverse, and several examples are highlighted below. Clinical markers of ineffective erythropoiesis, such as increased mean corpuscular volume and red cell distribution width, as well as the presence of cytopenias, are associated with an elevated risk of MN even after accounting for specific CH mutations.⁴¹ Germline genetic variation also plays a substantial role. In the UK Biobank cohort, 14 out of 98 CH predisposition genes were associated with increased risk of hematologic malignancy, eight of which were linked with CH overall.²¹ Moreover, CH carriers with germline variants in these 14 CH predisposition genes had a higher risk of developing a hematologic malignancy compared with CH carriers lacking the variants.²¹ The previously discussed TCL1A promoter variant, which confers gene-specific effects on CH predisposition, is also associated with slower expansion of non-DNMT3A CH.⁶⁷ In addition, germline pathogenic variants in DDX41 exhibit a markedly higher penetrance of MN in men compared to women for unknown reasons.⁸⁷ Down syndrome (trisomy 21) is frequently associated with hematologic abnormalities, including neonatal transient abnormal myelopoiesis (TAM). TAM is a preleukemic syndrome unique to Down syndrome, which typically presents with circulating blasts that originate from the fetal liver and harbor somatic G ATA1 mutations, and resolves spontaneously within a few months.⁸⁸ Notably, approximately 20% of neonates with Down syndrome have detectable somatic G ATA1 mutations in the absence of hematologic features of TAM.⁸⁹

Beyond selection of DDR mutations with genotoxic therapy, the extent to which cell-extrinsic factors influence CH progression to MN is less well established. Multiple preclinical models have demonstrated that systemic inflammation can promote the expansion of specific CH clones, particularly those with TET2- and DNMT3A-mutations.⁹⁰ However, there are limited data supporting this in humans. We recently identified a number of plasma proteins associated with CH (N=34) and subsequent risk of progression to MN (N=115) in the UK Biobank cohort.⁹¹ The proteins associated with MN risk were enriched for involvement in regulation of the innate and adaptive immune system, and improved risk prediction beyond clinical and CH-related features.⁹¹ Longitudinal studies in humans are needed to further characterize the extent to which inflammatory stress might influence CH evolution.

Risk stratification of clonal hematopoiesis

CH is ubiquitous with aging, but progression to MN only occurs in a small subset of patients (<1% per year among those with CHIP). As knowledge of the factors driving progression of CH to MN increases, there has been growing interest in developing models to predict risk of progression. Such models are essential for consideration of interventional trials to prevent malignant transformation, given that the overall risk of MN in unselected CH populations is low, and therefore the potential harms of preventative approaches could outweigh their benefits. Recently, two groups developed models for predicting MN risk that have gained prominence in the CH field.^22,23 Both efforts used the UK Biobank, which is uniquely suited for this purpose due to its in-depth genetic (including CH mutation profiling) and longitudinal (>10 years) health information on over 450,000 individuals.

Weeks et al.²³ employed a decision tree-based machine-learning model to identify demographic factors, CH clone characteristics, and laboratory values predictive of incident MN. From this, they created the Clonal Hematopoiesis Risk Score (CHRS), which stratifies individuals with CHIP and CCUS into three risk groups. The estimated 10-year probability of incident MN was 52% in the high-risk group (1% of individuals), 8% in intermediate-risk group (10%), and <1% in low-risk group (89%). The CHRS model incorporates the following binary features, listed in order of decreasing risk contribution: presence of high-risk mutation, mean corpuscular volume ≥100 f L, red cell distribution width ≥15, VAF ≥0.2, ≥2 CH mutations, presence of cytopenia, age ≥65 years, and presence of a single DNMT3A mutation. Gu et al.²² used Cox regression to develop MN-predict, a set of models that estimate the risk of progression to MPN, MDS, or AML separately These models rely on features similar to those in CHRS, but also allow for utilization of optional parameters such as body mass index and clinical laboratory values, which can improve prediction accuracy.

However, it is important to note that both CHRS and MN-predict were developed using data from the UK Biobank, a cohort enriched for healthy volunteers predominantly of European ancestry, which limits their generalizability to other patient populations.³³ This cohort also lacks longitudinal blood count data to assess persistence of cytopenias and bone marrow biopsies were not performed to exclude MN including MDS, both of which are required to meet diagnostic criteria for CCUS. To address these limitations, Xie et al.⁹² developed the Clonal Cytopenia Risk Score (CCRS) using real-world data from patients evaluated at tertiary referral centers who met diagnostic criteria for CCUS (and underwent bone marrow biopsy). The CCRS model stratifies individuals with CCUS into three risk groups based on three binary variables, listed in order of decreasing risk contribution: presence of ≥2 mutations, platelet count <100x10⁹/L, and presence of a splicing factor mutation. The estimated 2-year probability of incident MN was 37% in the high-risk group (10% of individuals), 14% in the intermediate-risk group (39%), and 6% in the low-risk group (51%). Notably, the incidence of MN observed in this real-world cohort was substantially higher than that observed among healthy participants in the UK Biobank cohort (~1%).²³ Further refinement of CH/CCUS risk prediction models in real-world patient populations will be critical to guide future therapeutic intervention trials.

Figure 2.Summary of the intervention approaches to mitigate malignant transformation in high-risk clonal hematopoiesis and clonal cytopenia of undetermined significance. CH: clonal hematopoiesis; CCUS: clonal cytopenia of undetermined significance; IDH1/2: isocitrate dehydrogenase 1 and 2; TET2: tet methylcytosine dioxygenase 2; IL-1β: interleukin-1 beta; NLRP3: NLR family pyrin domain containing 3; TGF-β: transforming growth factor beta; SMAD: SMAD family member 2; CDK4/6: cyclin-dependent kinase 4 and 6.

Interventional studies in high-risk clonal hematopoiesis

Advances in CH risk stratification have identified populations at significant risk of progression to MN based on clinical and molecular features,^22,23 thereby creating opportunities for therapeutic intervention trials. The increasing use of next-generation sequencing in oncology, both for detecting circulating tumor DNA and assessing for hereditary cancer syndromes, has also led to more frequent detection of CH as part of routine clinical care. Together, these developments have provided strong impetus to investigate targeted approaches to mitigate malignant transformation, particularly in individuals with high-risk clonal profiles. Several such approaches are now being investigated in clinical trials, which are summarized in Figure 2 and Table 1.

Ascorbic acid (vitamin C)

TET family methylcytosine dioxygenases, including TET2, catalyze the oxidation of methylated DNA, thereby promoting DNA demethylation, which has key roles in the regulation of gene expression.⁹³ Loss-of-function mutations in TET2 are very common in CH (second after only DNMT3A CH) and confer a fitness advantage to HSPC by impairing DNA demethylation, ultimately resulting in enhanced self-renewal and impaired differentiation. Preclinical studies demonstrated that treatment with ascorbate (vitamin C), which is a cofactor for TET2, can restore TET2 activity in the setting of TET2 haploinsufficiency, resulting in increased DNA demethylation, normalization of HSPC function, and impaired leukemogenesis.^94,95 This compelling preclinical work led to the initiation of two clinical trials studying vitamin C in CCUS. The phase II trial of single-agent high-dose intravenous ascorbic acid in TET2-mutant CCUS (NCT03418038) has readout and did not identify any clinical responses at 20 weeks by International Working Group MDS criteria (out of 8 patients eligible for response assessment) or significant changes in TET2 V A F . ⁹⁶ Results from a phase II study of oral vitamin C in low-risk MN including CCUS (NCT03682029) are pending.

IDH1/2 inhibition

Mutations in IDH1 (R132) and IDH2 (R140 or R172) are associated with high-risk CH and represent common early clonal leukemogenic events in MN. Isocitrate dehydrogenase (IDH) enzymes normally catalyze the conversion of isocitrate to alpha-ketoglutarate (αKG). These missense IDH mutations result in production of the oncometabolite 2-hydroxyglutarate, which competitively inhibits αKG-dependent enzymes (including TET2), resulting in broad effects including altered metabolism, aberrant DNA and histone methylation, and differentiation blocks.⁷⁰ Mutant-specific small molecule inhibitors of these enzymes are already approved for IDH-mutated AML, including ivosidenib and olutasidenib (IDH1), as well as enasidenib (IDH2). Given that IDH mutations are early clonal events in the natural history of MN and that IDH inhibitors have a reasonably good safety profile, IDH inhibition is an ideal target for initial CH intervention studies. Our group is leading two early phase clinical trials studying enasidenib in IDH2-mutant (NCT06240754) and ivosidenib in IDH1-mutant CCUS (NCT05030441). There is also an active trial (NCT06566742) assessing olutasidenib in IDH1-mutant CCUS and low-risk MN. Initial results from our IDH1-mutant CCUS trial (median follow-up of 16 months) demonstrated that ivosidenib was well tolerated and induced high rates of hematologic response (>80%) and IDH1 mutation clearance (>40%).⁹⁷

Table 1.Overview of the interventional clinical trials in clonal hematopoiesis and clonal cytopenia of undetermined significance.

Interleukin-1β inhibition

The relationship between CH, aging, and inflammation is complex and likely involves bidirectional causality between the three. This has been thoroughly reviewed in detail elsewhere.^98,99 In brief, both CH (particularly TET2 mutant) and aging are associated with elevated levels of proinflammatory cytokines, which can provide a selection pressure for expansion of specific CH clones with a relative fitness advantage in the context of systemic inflammation (TET2 and DNMT3A CH).⁹⁰ Preclinical studies in mice have shown that TET2 loss in hematopoietic cells results in accelerated atherosclerosis and increased proinflammatory cytokines (including interleukin [IL]-1β and IL-6) secretion by macrophages.^100,101 Furthermore, IL-1 (which increases with aging) induces expansion of Te t 2 ^+/- HSPC in mouse models of Te t 2 ^+/-driven CH, whereas genetic and pharmacological inhibition of IL-1 signaling impairs expansion of Tet 2 ^+/- clones.¹⁰² These findings collectively provide a strong biological rationale for the therapeutic targeting of IL-1β in CH and CCUS, which has fueled subsequent clinical investigation in this space. Canakinumab is an anti-IL-1β monoclonal antibody that inhibits IL-1β from binding to and activating the IL-1 receptor. The large phase III CANTOS trial assessed whether treatment with canakinumab reduces major adverse cardiovascular events in patients with prior myocardial infarction and an elevated baseline C-reactive protein level.¹⁰³ The trial found that treatment with canakinumab (compared to placebo) resulted in an ~15% relative risk reduction in major adverse cardiovascular events, increased incidence of fatal infections, and no difference in overall survival. Subsequent exploratory analyses of CANTOS showed that: CH mutations in TET2 were more common than in DNMT3A (which could be related to the inclusion criteria requirement of an elevated C-reactive protein level); patients with TET2 CH who were treated with canakinumab had a decreased risk of major adverse cardiovascular events (compared to those given a placebo); and canakinumab treatment was associated with decreased risk of incident anemia and improved hemoglobin response (particularly among patients with concurrent CH mutations and anemia).^104,105 A randomized placebo-controlled clinical trial (NCT05641831) of canakinumab in patients with CCUS is currently underway and aims to determine whether canakinumab IL-1β inhibition can prevent or delay MN development (primary endpoint of time to MN). A recent small single-arm trial of canakinumab in lower-risk MDS demonstrated a modest 17% overall response rate, with all responders (N=4) having an IPSS-M score <0.¹⁰⁶

The NLRP3 inflammasome pathway is aberrantly activated in many MN, leading to caspase-1-dependent secretion of the pro-inflammatory cytokines IL-1β and IL-18, which drive inflammation and ineffective hematopoiesis.¹⁰⁷ The small molecule NLRP3 inhibitor DFV890 is currently under investigation in low-risk MN including high-risk CCUS (NCT05552469) as well as TET2- and DNMT3A-mutant CHIP with coronary heart disease (NCT06097663).

Transforming growth factor-β/SMAD Inhibition

Luspatercept is a recombinant fusion protein composed of the extracellular domain of the activin receptor type IIB and the Fc domain of immunoglobulin G1 that is approved by the United States Food and Drug Administration for treating anemia in patients with low-risk MDS and β-thalassemia. It binds to multiple transforming growth factor-beta (TGF-β) superfamily ligands, resulting in inhibition of SMAD2 and SMAD3 signaling, which promotes effective maturation of erythroid progenitor cells.¹⁰⁸ Luspatercept is particularly efficacious for treating anemia due to MDS with SF3B1 mutations and/or ring sideroblasts,¹⁰⁹ which is characterized by pronounced ineffective erythropoiesis and an indolent clinical course.¹¹⁰ Interestingly, no change in the VAF of SF3B1 or other somatic mutations has been observed in luspatercept responders compared with non-respond-ers.¹¹¹ Additional clinical trials will be required to determine whether luspatercept has disease-modifying activity in MDS or CCUS. A clinical trial (NCT06788691) of luspatercept in patients with CCUS is ongoing.

CDK4/6 inhibition

Cyclin-dependent kinases 4 and 6 (CDK4/6) play a key role in cell cycle regulation, promoting progression through the G1/S checkpoint by phosphorylating retinoblastoma protein. CDK4/6 inhibitors are used widely in combination with endocrine therapy to treat hormone receptor-positive breast cancers, which are sensitive to the inhibitors due to intact retinoblastoma protein function and reliance on CDK4/6 signaling for cell cycle progression.¹¹² HSPC are also reliant on CDK4/6 activity for cell cycle progression and undergo transient G1 arrest after CDK4/6 inhibition.¹¹³ Trilaciclib is a CDK4/6 inhibitor that was developed to protect HSPC from DNA damaging agent-induced myelosuppression by inducing G1 arrest prior to administration of DNA damaging agents.¹¹⁴ It is now FDA-approved for the prevention of chemotherapy-induced myelosuppression in patients with extensive-stage small cell lung cancer, a malignancy intrinsically resistant to CDK4/6 inhibition due to loss of retinoblastoma protein function, thereby explaining why trilaciclib does not compromise the antitumor efficacy of chemotherapy. Given the myeloprotective effects of trilaciclib, our group recently investigated the effect of trilaciclib on chemotherapy-related expansion of CH clones with DDR mutations (e.g., TP53, PPM1D, CHEK2). We found that trilaciclib reduced expansion of DDR CH by 32% when given prior to cytotoxic chemotherapy in four randomized clinical trials and in a TP53-mutant CH mouse model.¹¹⁵

Given the poor outcomes associated with therapy-related MN, strategies to mitigate cytotoxic chemotherapy-driven expansion of DDR CH, particularly TP53-mutant clones, are warranted and may reduce progression to therapy-related MN. A pragmatic approach is to quantify MN risk using validated clinical prediction models that integrate CH mutational data with routine laboratory parameters,^22,23,92 and to incorporate these individualized estimates when weighing the absolute benefit of adjuvant chemotherapy, whose survival advantage in some early-stage settings is modest, against long-term risk of therapy-related MN. CH status could also be determined from cell-free DNA assays already obtained as part of standard oncological care. Modeling analyses suggest that a subset of patients with early-stage breast cancer have a predicted MN risk that exceeds the expected absolute survival benefit from adjuvant chemotherapy (which frequently includes doxorubicin, a therapy associated with a high risk of MN), highlighting the potential utility of CH-based risk stratification to inform adjuvant treatment decisions.⁴⁹

Hypomethylating agents

Hypomethylating agents such as decitabine have demonstrated efficacy in most MN, while maintaining a favorable safety profile. A clinical trial (NCT06802146) evaluating decitabine/cedazuridine (an oral formulation of decitabine) in patients with high-risk CCUS is ongoing.

Metformin

Metformin is widely used due to it being the first-line treatment for type 2 diabetes mellitus. It has a complex and incompletely understood mechanism of action that involves inhibition of hepatic gluconeogenesis and mitochondrial respiration.⁸⁹ A pair of preclinical studies from earlier this year demonstrated that the competitive advantage of murine Dnmt3a R878H HSPC is dependent on increased mitochondrial respiration and can be reduced by treatment with metformin.^116,117 Functional studies revealed that metformin treatment normalized the aberrant DNA methylation and gene expression profiles driven by Dnmt3a R878H. In addition, analysis of the UK Biobank showed that metformin use was associated with a lower prevalence of DNMT3A R882 CH. A small phase II clinical trial (NCT04741945) investigating metformin in patients with CCUS and low-risk MDS is ongoing.

Miscellaneous

Preclinical studies have demonstrated that statins can exert anti-cancer effects in vitro.¹¹⁸ Statins also possess anti-inflammatory effects in addition to their lipid-lower-ing properties.¹¹⁹ Observational data in MDS indicate that statin use is associated with improved overall survival and a reduced risk of progression to acute leukemia, as shown in a large retrospective database analysis employing propensity score matching.¹²⁰ These findings provided the impetus for a prospective clinical trial (NCT05483010), which is currently evaluating high-intensity statin therapy in patients with CCUS and low-risk MDS.

Curcumin, the active ingredient in turmeric, is a polyphenol with antioxidant and anti-inflammatory properties. It is currently being evaluated in a placebo-controlled clinical trial (NCT06063486) of patients with low-risk MN including CCUS. The primary endpoint of the study is the change in inflammatory cytokine levels and inflammation-related symptomatology.

Considerations on clinical trial design

Designing interventional trials for CH presents significant challenges. Because CH is a premalignant condition, potential interventions must be risk-informed and have a favorable toxicity profile. Trials should have clinically meaningful end-points and avoid the overmedicalization of otherwise healthy individuals with low-risk CH features. To date, CH trials have largely focused on patients with CCUS; accordingly, the following discussion of clinical trial design considerations centers on this population. CCUS is particularly appropriate for interventional trials, as the risk of progression to MN among individuals with CH appears to reside almost entirely in those who develop a preceding cytopenia.¹²¹

Most CH trials have modeled their key endpoints from those used in low-risk MDS, such as hematologic improvement (Table 1). However, reductions in transfusion burden or infections are the only cytopenia-related outcomes recognized as clinically meaningful by the FDA in hematologic malignancies. Thus, studies that show improvement in neutropenia, for example, may not be considered “clinically meaningful”. The clinical relevance of currently accepted endpoints in studies of low-risk MDS remains under debate. For example, while there are multiple FDA-approved therapies for low-risk MDS that result in reduced transfusion burden, none has been prospectively shown to alter the natural history of the disease, such as improvement of overall survival.¹²² Thus, it would be prudent for the CH field to learn from the persistent challenges faced in drug development for low-risk MDS¹²³ and design feasible studies that can demonstrate improvement in quality of life and the potential for disease modification.

Randomized, placebo-controlled trials with disease-free survival as the primary endpoint are the gold-standard for chemoprevention trials. However, this endpoint is limited by the natural history of CH, in which progression typically occurs over many years. Among healthy individuals in the UK Biobank, the 10-year risk of MN is approximately 52% for those with high-risk CH, a group that represents only 1% of individuals with CHIP/CCUS.²³ Intermediate-risk CH is more common (~10% of CHIP/CCUS), but only 8% of carriers progress to MN over a 10-year period.²³ Progression rates are higher among patients with CCUS evaluated at tertiary referral centers, but vary widely between studies ranging from ~20% at 4 years⁹² to 95% at 10 years.²⁶ In addition, the absence of standard-of-care interventions may make participation in long-term placebo-controlled studies undesirable for some patients.¹²⁴

Surrogate endpoints may enable more feasible trial designs but must be chosen carefully to ensure clinical relevance. Given that CH expansion is well-recognized to be part of the natural history of CH progression to MN, interventions that reduce the burden of high-risk CH clones (as reflected by VAF) should reduce MN development. However, the direct impact of a reduction in CH VAF on MN risk has not been measured in prospective trials. Furthermore, reductions in CH mutant VAF can be confounded by the variable distribution of CH-mutant cells across hematopoietic lineages, which in turn can fluctuate with clinical context (e.g., active infection).¹²⁵ However, the achievement of measurable residual disease negativity in hematologic malignancies is clearly associated with improved clinical outcomes across a variety of malignancies including chronic phase chronic myeloid leukemia, AML, acute lymphocytic leukemia, and multiple myeloma.¹²⁶ Thus, we believe that achievement of measurable residual disease negativity for high-risk CH clones represents a valid potential surrogate endpoint for MN prevention. This outcome is feasible for potent targeted therapies directed against high-risk CH mutations.

Finally, the inherent rarity of high-risk CHIP/CCUS, especially specific biological/molecular subtypes, poses significant challenges in recruiting study participants. Currently, screening for CH is not part of standard clinical practice. While targeted sequencing to identify somatic mutations in myeloid driver genes is increasingly common in the work-up of unexplained cytopenias, it is not uniformly applied. Thus, identification of individuals with high-risk forms of CH is challenging. Disease rarity combined with inefficiencies within the current clinical trial infrastructure and difficulties in securing interest from pharmaceutical industry partners for rare diseases make high-risk, gene-specific CH trials challenging to conduct. Decentralized clinical trials may represent an attractive approach for uncommon diseases. These trials are clinical studies in which some or all trial-related activities are conducted remotely using digital health technologies (e.g., telemedicine visits), instead of at a traditional centralized clinical research site (e.g., office visit at an academic medical center).¹²⁷ This model can reduce barriers to clinical trial participation including geographic distance, time toxicity, and financial burden. Hybrid versions can also be used which incorporate traditional design features (e.g. sample collection at local laboratories). Decentralized clinical trials are best suited for well-characterized therapies with good safety profiles used in relatively healthy patient populations with a low rate of serious adverse events.

In summary, to fully realize the potential of CH interventional studies to prevent MN, innovative and flexible clinical trial designs are needed. Rare diseases are subject to similar unique investigational and regulatory approaches.¹²⁸ High-risk CH should be included under this umbrella. Adoption of the rare disease regulatory paradigm including flexible evidence standards, and consideration of alternative data sources (including historic controls) will be needed to facilitate development and approval of therapies for individuals with high-risk CH. Encouragingly, the FDA has recognized the need to evolve its regulatory strategies for personalized therapies in rare diseases and recently announced plans for new drug development processes in this space, including the Plausible Mechanism Pathway and the Rare Disease Evidence Principles program.¹²⁹

Conclusions

The CH field has made remarkable progress in the short time since the seminal studies by Jaiswal et al.⁷ and Genovese et al.⁶ were published in 2014. These investigations provided an unprecedented window into the earliest stages of carcinogenesis, elucidating how intrinsic, environmental, genetic, and acquired factors – often in a mutation-specific manner – govern CH initiation, clonal dynamics, and risk of progression to MN. Much of this progress has been driven by large-scale population (e.g., UK Biobank) and longitudinal sequencing studies, which are only feasible due to the unique ability to sample the hematopoietic system through peripheral blood draws. In turn, these discoveries have enabled the development of risk stratification models to identify individuals with CH at high risk of progression to MN, thereby opening the door for therapeutic intervention trials, which are now actively enrolling patients.

Despite these advances, significant unanswered questions remain. Key priorities for the field include defining the cell-intrinsic and -extrinsic mechanisms underlying the distinct clonal behaviors of specific driver mutations over time (particularly splicing factor mutations); elucidating the basis of hematopoietic oligoclonality that emerges with aging and its frequent association with CH lacking a known driver; and determining whether therapeutic intervention can alter the natural history of CH.

Moving forward, translating insights from CH biology into effective invention approaches will require careful balancing of the risk and benefits of intervention, as the absolute risk of progression to MN is modest even among individuals with high-risk CH. Ultimately, these efforts will be essential for determining optimal clinical management of CH, which is increasingly being identified in clinical practice, and will deepen our understanding of early carcinogenesis in humans, regardless of the trial outcomes.

Footnotes

• Received September 29, 2025
• Accepted January 23, 2026

Correspondence

K.L. Bolton bolton@wustl.edu

Disclosures

JSB and MJW have no conflicts of interest to disclose. KLB receives research funding from and has served on a medical advisory board for Servier.

Contributions

JSB, MJW and KLB conceived and wrote the manuscript.

Funding

This work was supported by NIH grant T32HL007088 (to JSB).