Interplay between Chromosomal Alterations and Gene Mutations Shapes the Evolutionary Trajectory of Clonal Hematopoiesis

Background

Stably acquired mutations in hematopoietic cells represent substrates of selection that may lead to clonal hematopoiesis (CH), a common state in cancer patients that is associated with a heightened risk of leukemia development. Owing to technical and sample size limitations most CH studies have characterized gene mutations or mosaic chromosomal alterations (mCAs) individually. The relationship between acquired gene mutations and mCAs in CH and their joint roles in leukemia development have not been systematically investigated.

Methods

We developed a method to reliably map mCAs at low cell fractions from deep targeted sequencing data. We applied this method in a cohort of 32,442 solid tumor patients who have undergone prospective clinical sequencing (MSK-IMPACT). We characterized gene mutations in our patient cohort using an established variant calling procedure from our previous studies.

Results

We jointly characterized 383 mCA events (median aberrant cell fraction 32%, range 10%-90%) and 14,789 mutations across 457 genes. mCA was significantly associated with age (OR=1.8, P<0.001), male gender (OR=1.4, P=0.012), white race (OR=1.5, P=0.033) and prior receipt of external beam radiation therapy (OR=1.7, P=0.022). 217 (63%) mCAs co-occurred with at least one gene mutation, while 129 (37%) did not (OR=3.9, P<0.001). mCA was especially enriched in CH cases with high mutation number and VAF, detectable in 5.8% of subjects with ≥3 gene mutations and 4.8% of those with mutations at >20% VAF, compared to 1% of the general cohort.

We identify co-mutational patterns characteristic of diverse mechanisms of clonal selection. We observe that mutations in DNMT3A, TET2, JAK2, MPL, EZH2, TP53 and ATM form recurrent double-hits with deletions or CNLOHs, resulting in either oncogene mutant dosage adjustment or inactivation of tumor suppressors. Notably, certain mCA events were highly directed events acting on previously acquired gene mutations in the corresponding loci. Of six events of 7qCNLOH, all six co-localized with an EZH2 (7q36.1) mutation (q<0.001). Of 12 cases with 9pCNLOH, 11 (92%, q <0.001) co-localized with a JAK2 V617F mutation. 4 out of 9 (44%, q <0.001) 1pCNLOH events co-localized with a MPL (1p34.2) mutation. In addition, we observe recurrent composite genotypes (4q24-/SRSF2, 7qCNLOH/ASXL1, 20q-/U2AF1) indicative of co-operating or epistatic interactions as well as loss of gatekeeper function (i.e. TP53) presenting with multiple chromosomal aneuploidies (5-, 7-, 3+). In total, these recurrent composite genotypes resembling known genetic interactions in leukemia genomes underlie 23% of all detected autosomal mCAs.

During patient follow-up, the 3-year cumulative incidence of leukemias was significantly higher in patients with composite CH genotypes (14.6%, CI: 7-22%) as compared to patients with either mCA, gene mutation alone or no CH, of which all had a 3-year cumulative incidence of <1% (Figure 1). We performed a multivariable cause-specific Cox regression model and showed that mCA was independently predictive of subsequent leukemia diagnosis (HR=14, 95% CI: 6-33, P=1.2e-09) after adjusting for number of gene mutations and VAF in putative drivers.

Conclusions

Our joint characterization of gene mutations and mCAs in a large prospective sequencing cohort reveals a previously unrecognized layer of complexity in the evolutionary dynamics of clonal hematopoiesis that converges towards characteristic genotypes associated with distinct leukemia subtypes. This puts mCAs in the context of the continuous evolutionary process of oncogenesis that can often span years and sheds new lights on its patterns of acquisition and progression. We demonstrate that the integration of chromosomal aberrations provides additional resolution to risk stratification as well as interpretation of clinical phenotypes and that mCAs should be screened in conjunction with gene mutations to improve existing CH surveillance programs in cancer patients.