Somatic Mutations in Therapy-Related Myeloid Neoplasms Are Influenced By Therapeutic Modality and Clonal Hematopoiesis of Indeterminate Potential

Treatment-associated myeloid neoplasias (tMN) are serious iatrogenic complications of cytotoxic therapies applied to primary malignancies (PM). With increased therapeutic success rates and prolonged survival of cancer patients, tMN may become more prevalent. Although tMN diagnosis is trivial to the extent that previous therapies are known, tMN may represent coincidental primary MDS/AML not causally linked to chemotherapy (ctx) or radiation (rtx). Other seemingly tMN cases may carry germ line predisposition responsible for co-occurrence of 2 neoplastic processes, and finally others are truly treatment-related MN. With the recognition of clonal hematopoiesis of indeterminate potential (CHIP), it is also likely the preexisting CHIP would be either selected for, eliminated by cancer therapies, or that ctx or rtx lead to emergence of CHIP. tMN may also manifest without antecedent CHIP, and thus could be either CHIP-derived or de novo.

Some of the problems in assigning somatic mutational pattern to tMN may be alleviated by application of proper control groups which include sMN (MN after PM treated only surgically). In our cohort of 1058 patients, we identified 109 cases of such sMN, 266 tMN with a history of rtx or ctx for PM, and 683 of primary MN (pMN), having no PM. Of these 65 sMN, 145 tMN, and 683 pMN were sequenced by NGS. Using these three patient groups, we sought somatic mutations that distinguish them.

tMN presented as more aggressive disease: diagnosed older vs. pMN/sMN (68 years, p<.001) , shorter latency from PM to MN vs. sMN (8.7 vs. 10.5 years, p=.085), complex cytogenetics vs. pMN, sMN (p=1.4x10^-5, p=2.7x10^-4) including chromosome 7 (p=1.7x10^-7, p=8.4x10^-5) and 5 abnormalities (p=.044, p=.09), 50% tMN were advanced MDS/AML vs. 35% pMN, 42% sMN (p=.00016, p=.25). The most common mutations in all 3 groups were TET2, DNMT3A, ASXL1 and SRSF2. Mutations in SF3B1 and JAK2 were less common in tMN vs. pMN, sMN (p=.058, p=.014; p=.011, p=.327) while those in TP53, KIT, EZH2,WT1 were most frequently mutated in tMN vs. pMN/sMN (OR 2.6, p=.002, OR 6, p=.011, OR 1.9, p=.083, OR 3.2, p=.08).

Mutations in ETV6 and EZH2 were only found in rtx vs. ctx-treated tMN cases (p=.046, p=.004). TP53 mutations were associated with ctx (OR 7.2, p=0.062), and when combined with cases which received both ctx and rtx vs. rtx alone, TP53 mutation was 9.3x was common (p=.014). In tMN TP53 and EZH2-mutated cases, a higher proportion of transversions was observed vs. those found in pMN (p=.055, p= 0.052). The domains mutated in TP53 tMN vs. pMN/sMN cases were similar, while EZH2-mutated tMN were enriched for hits in domain 2 (p=0.0604). This suggests that the type of treatment utilized influences the molecular signature of tMN in terms of frequency of mutations as well as types of mutations found.

A meta-analysis of 9 CHIP studies was performed to pool overall frequencies of MN-related genes. These frequencies were compared to those of ancestral events (determined by recapitulation of clonal hierarchy via variant allele frequency and zygosity) in our cohort. This yielded 3 categories of mutations: those that are CHIP-derived (frequency in CHIP> ancestral frequency in MN), from de novo MN (mutations in gene not seen in CHIP), or those found in both CHIP and MN, termed mix-derived. CHIP-derived hits were TET2, DNMT3A, JAK2 while STAG2, EZH2, APC, MLL, WT1 were de novo hits.

Using this categorization scheme tMN break down as 19% CHIP-derived, 38% de novo, and 24% mix-derived. CHIP-derived tMN were, on average, 6 years older than de novo tMN (p=.019). This also held true for the age of PM diagnosis, where CHIP-derived cases were, on average, 10 years older than de novo tMN (p=.0175), suggesting that age of PM may be correlated with acquisition of CHIP in tMN. CHIP-derived vs. de novo tMN latencies did not differ.

The molecular signatures of tMN are influenced by therapies utilized for PM as well as CHIP. Targeted sequencing for germ line predisposition genes is under way for these patients to further characterize their molecular profiles.