Driver mutation zygosity is a critical factor in predicting clonal hematopoiesis transformation risk

Clonal hematopoiesis (CH) refers to age-related acquisition of mutations that confer selective advantage to hematopoietic stem cells, encompassing both driver single-nucleotide/indel mutations (CHIP) and mosaic chromosomal alterations (mCAs). While most individuals with CHIP do not develop hematologic malignancy, known risk factors for progression include driver gene identity, clonal fraction, clonal complexity, and potentially specific mutated residues. A CH risk score (CHRS) has recently been developed incorporating age, mutation type and number, variant allele fraction, cytopenias, and red cell indices. The authors hypothesized that the zygosity of a driver mutation—heterozygous versus homozygous—materially influences progression risk, with homozygosity potentially arising via co-occurring mCAs (e.g., copy-neutral loss of heterozygosity) at the same locus. They tested this by assessing canonical CHIP mutations with mCAs in the UK Biobank and then focusing on two common driver mutations, DNMT3A R882 and JAK2 V617F, to evaluate how co-occurrence with locus-specific mCAs relates to transformation risk.

Study design and cohorts: The UK Biobank (UKB; n = 451,180 with exome data) served as the discovery cohort, excluding individuals with hematologic cancer prior to or within 6 months of blood draw. BioVU (n = 91,335) was the replication cohort, with linked EHRs. In BioVU, individuals with hematologic malignancy (HM) prior to or within 5 months of blood draw were excluded; incident HM cases were defined by at least three distinct HM ICD-9/10 codes on different days; those with one or two codes were excluded. Controls had no HM codes. CHIP detection (UKB): Exomes were scanned for mutations in 74 canonical CHIP genes using Mutect2. Candidate variants required DP ≥ 20, alternate allele read depth ≥ 5, presence in both read directions (F1R2 ≥ 1 and F2R1 ≥ 1), and presence on a curated CHIP variant list. Population-level filters were then applied to remove false positives per established methods. mCA detection (UKB): mCAs were called per previously published methods from genotyping array data, including CN-LOH, deletions, duplications, and unknown copy changes; sex chromosome losses were excluded. CHIP detection (BioVU): MEGA+ genotyping array intensity probes specifically interrogated JAK2 V617F and DNMT3A R882H (with categorization including DNMT3A R882C). Individuals were designated mutation-positive if normalized alternate allele fraction exceeded 6 standard deviations above the population mean. Validation against targeted NGS (>500× coverage across 24 CHIP genes) in 477 participants showed high concordance for JAK2 V617F and DNMT3A R882C (r > 0.9), lower for R882H; thus, DNMT3A R882H positivity was defined by NGS positivity or R882C positivity. mCA detection (BioVU): Peripheral blood DNA was genotyped on Illumina MEGA+ (>2M SNPs). mCAs were called from raw IDAT files using established pipelines (e.g., MoChA-based methodology), focusing on autosomal CN-LOH, deletions, gains, and unknown copy changes; sex chromosome losses were excluded. Risk factor analyses (BioVU): Associations of CH alterations with baseline demographics and comorbidities (age, sex, race, BMI, smoking, hypertension, diabetes, hyperlipidemia) were tested via logistic regression. A multivariable model included age, sex, race; additional models adjusted for these covariates. Blood count associations (BioVU): Associations between CH alterations and blood counts (RDW, platelets, hemoglobin, WBC, MCV) were assessed using multivariable logistic regressions with baseline characteristics as covariates, comparing against subjects without CH alterations. Outcomes and statistics: Incident hematologic malignancy was the primary outcome. Time-to-event analyses used Cox proportional hazards models. Statistical analyses were performed in R (survival v3.3-1).

- UK Biobank discovery cohort (n = 451,180): - mCAs detected in 5,745 (1.3%); most frequent on chromosomes 1 (n = 809), 11 (n = 742), 22 (n = 734), and another report of chromosome 1 (n = 628). - CHIP mutations in 15,304 (3.3%); most common genes: DNMT3A (n = 8,988), TET2 (n = 1,829), ASXL1 (n = 1,545). DNMT3A R882 in 1,152 (0.3%); JAK2 V617F in 121 (0.02%). - Associations with higher incidence of hematologic malignancy (HM): mCAs HR = 9.22 (95% CI 8.24–10.33, p < 0.001); DNMT3A R882 HR = 46.69 (95% CI 3.46–635.9, p < 0.001); JAK2 V617F HR = 39.54 (95% CI 23.39–66.74, p < 0.001); TET2 HR = 2.61 (95% CI 1.87–3.65, p < 0.001); TP53 HR = 4.51 (95% CI 2.03–10.05). - Co-occurrence patterns transforming heterozygous to homozygous increased risk: JAK2 V617F with 9p CN-LOH; TET2 with chromosome 4 mCA; TP53 with chromosome 17 mCA. Among 26 individuals with JAK2 V617F plus concurrent mCA who developed incident HMs: 10 polycythemia vera, 4 myeloproliferative neoplasms NOS, 4 myelofibrosis, 3 essential thrombocythemia, 2 AML, and 1 each of APML, CML, and MDS. - BioVU replication cohort (n = 91,335): - mCAs in 765 (0.8%); recurrent events included 9p CN-LOH (n = 105, 6.9% of mCAs), del(20q) (n = 98, 6.4%), del(13q) (n = 67, 4.4%), among others. - CHIP calls (limited to DNMT3A R882 and JAK2 V617F) in 503 (0.6%): DNMT3A R882 in 161 (0.2%); JAK2 V617F in 345 (0.4%). Combined mCA + DNMT3A/JAK2 in 130 (0.1%). Strong co-occurrence of JAK2 V617F with 9p CN-LOH (n = 100; OR ~5693, q < 0.001). - Age-dependence: Prevalence of mCAs and DNMT3A/JAK2 mutations increased with age. - Risk of HM: mCAs HR = 23.41 (95% CI 19.26–28.46, P < 0.001); JAK2 V617F HR = 48.05 (95% CI 35.50–55.35, P < 0.001); DNMT3A R882 not significantly associated with increased HM risk in this cohort. - Co-occurrence effects: JAK2 V617F plus 9p CN-LOH, which converts heterozygous to homozygous JAK2, conferred markedly increased HM risk (HR = 54.76, 95% CI 33.92–88.41, P < 0.001). DNMT3A R882 plus mCA did not increase risk beyond mCA alone. - Incident HM subtypes among 17 individuals with JAK2 V617F plus 9p CN-LOH: 10 myeloproliferative neoplasms, 4 MDS, 2 CML, 1 AML. - Additional observations: - Co-occurrence patterns suggest selective advantages via biallelic alterations are gene- and locus-specific; strong enrichment of JAK2 V617F with 14q CN-LOH observed, whereas similar enrichment was not seen with some other loci. - CH alterations were associated with differences in blood counts (e.g., lower platelets and abnormalities) and varied by specific combinations of mCAs and SNVs/indels.

The study demonstrates that driver mutation zygosity critically influences the risk of progression from clonal hematopoiesis to hematologic malignancy. Specifically, co-occurrence of a driver mutation with a locus-specific mCA that induces copy-neutral loss of heterozygosity can convert a heterozygous mutation into a homozygous state, markedly elevating transformation risk. This effect is robustly shown for JAK2 V617F with 9p CN-LOH across two large cohorts, while DNMT3A R882 co-occurrence with mCAs does not uniformly increase risk beyond mCAs alone. These findings refine risk stratification beyond mutation presence and clone size by incorporating zygosity status, suggesting that comprehensive profiling of both SNVs/indels and mCAs provides a more accurate prognostic picture. The results align with emerging understanding that specific mCA-SNV pairings confer selective advantages via biallelic inactivation/activation and may involve pathways such as TCL1A on 14q. Given the age-related rise in both CHIP and mCAs, integrating zygosity into clinical reporting could help identify individuals at highest risk for malignant transformation and guide monitoring strategies.

Across the UK Biobank and BioVU cohorts, co-occurrence of driver mutations with locus-specific mCAs that convert heterozygous to homozygous status, particularly JAK2 V617F with 9p CN-LOH, substantially increases the risk of hematologic malignancy. Mutation zygosity has important prognostic implications and should be incorporated into risk scores and clinical reports alongside other factors such as clone size and cytopenias. Future work should expand analyses beyond DNMT3A R882 and JAK2 V617F to additional recurrent CH drivers to generalize these findings and refine risk stratification tools.

- Co-occurrence of CHIP and mCAs was not evaluated using multivariable models alongside other predictors (e.g., CHRS components) due to limited numbers of co-occurrence events. - The BioVU replication focused on two common driver mutations (DNMT3A R882 and JAK2 V617F), enabling single-variant resolution but limiting generalizability to other mutations and genes. - Potential measurement limitations include array-based detection of specific CHIP variants (with lower concordance for DNMT3A R882H) and reliance on genotyping-based mCA calls. - Differences in cohort characteristics and case definitions based on EHR coding may introduce ascertainment or selection biases.