Abstract
Clonal hematopoiesis (CH) occurs when somatic mutations confer a fitness advantage in blood cells enabling them to outcompete, increases with age, and has been linked to greater risk of hematologic cancer and cardiovascular disease. Studies report increased prevalence of certain CH mutations with radio/chemotherapy and exposures such as smoking, thereby potentiating the risks of CH. Few efforts have aimed to broadly define associations between therapeutic exposures and CH.
To comprehensively map these associations, we connected real-world therapeutic exposure data to CH mutation data from >30,000 peripheral blood samples of patients diagnosed with 30 solid tumor subtypes, mainly arising from breast, pancreatic, prostate, ovarian, bladder, colon, and non-small cell lung cancer. CH was detected via targeted next-generation sequencing. Analysis was constrained to reported CH and myelodysplastic syndrome/acute myeloid leukemia-associated genes, including DNMT3A, TET2, ASXL1, TP53, PPM1D, and spliceosome genes among others.
To assess correlations between CH mutations and therapeutic exposures, >10,000 therapeutic exposures were recorded and categorized into 84 treatment classes such as analgesics, platinum compounds, hypomethylating agents, cell-based therapies, and anthracyclines. We then functionally classified CH mutations into distinct categories including age-related (i.e., DNMT3A, TET2, ASXL1 [DTA]), DNA damage response (DDR), and spliceosome mutations. Feature selection was performed to identify therapeutic exposures that were informative of each CH category. Penalized likelihood logistic regression was used to quantify and test for significant associations between CH mutations and therapeutic exposures adjusting for covariates including age, sex, and known co-occurring exposures.
CH mutations were detected in 16% of this cohort and were strongly age-associated (P < 0.001), with 13% of patients demonstrating a single CH mutation and 3% demonstrating multiple. When adjusting for age and sex, the frequencies of CH were highest in meningioma, endocrine tumors, lung cancer, and ovarian cancer. The most frequently mutated genes were DNMT3A, TET2, ASXL1, TP53, PPM1D,CHEK2, and ATM, with the DDR genes being overrepresented in this cohort as the probable result of prior therapy-related exposures.
From the correlation map of CH mutations and therapeutic exposures, we found that highly clonal DTA mutations (variant allele fraction > 10%) were significantly increased with thyroid hormone replacement (odds ratio [OR]: 1.94, 95% confidence interval [CI]: 1.20 - 3.02; P < 0.01) and aromatase inhibitor therapy (OR: 1.72, CI: 1.30 - 2.25; P < 0.001) accounting for menopause status. Significant, albeit lower magnitude, upticks in the odds of DTA mutations were noted for pyrimidine analogs, tubulin binding agents, and immunotherapy. DTA mutations were not significantly associated with anthracycline or platinum-based therapy. In contrast, DDR mutations were most strongly associated with PARPi therapy (OR: 3.47, CI: 2.2 - 5.31; P < 0.001), platinum agents (OR: 2.40, CI: 1.97 - 2.92; P < 0.001), and anthracyclines (OR: 1.89, CI: 1.32 - 2.68; P < 0.001), consistent with previous reports. Lower magnitude associations with DDR mutations were noted for alkylating agents, pyrimidine analogs, and tubulin binding agents. Spliceosome and IDH1/2 mutations were most strongly associated with treatments for myeloid/lymphoid neoplasms, such as hypomethylating agents, a probable result of these drugs being used in treatment of underlying hematologic disease rather than being causal for CH. Intriguingly, a novel association was revealed between IDH1/2 mutations and progestin (OR: 22.47, CI: 2.26 - 109.74; P = 0.013).
Through a systematic evaluation of real-world evidence, we revealed several novel associations between functional CH categories and therapeutic exposures, especially among highly clonal DTA mutations. As these data are strictly correlative, further prospective studies formally assessing causality and potential confounding variables such as therapy-associated comorbidities are needed. Nonetheless, these findings provide a basis for using multimodal data to develop predictive models of CH modulation and progression to hematologic disease, which can be crucial in weighing the relative risks of therapeutic interventions in some patients with solid tumors.
Disclosures
Sonnenschein:Tempus Labs, Inc.: Current Employment. Moore:Tempus Labs, Inc.: Current Employment. Lo:Tempus Labs, Inc.: Current Employment. Kang:Tempus Labs, Inc.: Current Employment. Taxter:Tempus Labs, Inc.: Current Employment. Mahon:Tempus Labs, Inc.: Current Employment. Hassane:Tempus Labs, Inc.: Current Employment.
Author notes
Asterisk with author names denotes non-ASH members.
