Chemotherapy and Radiation Results in Distinct Patterns of Somatic Mutations in Subsequent Myeloid Disease

Approximately 10% of myeloid neoplasia (MN) may be attributed to extrinsic factors. While environmental exposures are difficult to assess in their impact, iatrogenic causes of MDS, such as previous treatment (tx) with chemotherapy (ctx) or radiation (rtx), are easier to obtain from a patient's history. However, the distinction of a therapy-related MN (tMN) from germline-mediated and spontaneous second MN, e.g ., due to aging, may be problematic. Distinct etiologies may imply specific molecular defects; for instance, tMN types include topoisomerase II inhibitor-induced translocations of 11q23 and 21q22. We hypothesize that tMN is driven by a specific subset of somatic hits.

Among 1058 patients (pts) seen at Cleveland Clinic between July 2002 and July 2017 (68% MDS, 24% AML, 8% MDS/MPN); 266 had tMN, with clearly-defined tx history with ctx, rtx, or a combination of ctx and rtx (both) for their primary malignancy (PM) and 109 pts had a history of PM without tx using ctx or rtx (second MN or sMN). Median age at diagnosis (dx) was 64 (9-88), 68 (20-92) and 74 (47-95) years (y) in pMN, tMN and sMN, respectively (p<.0001). Average latency for tMN vs. sMN was 8.6y vs . 10y (p=0.5). Higher-risk MDS/AML was found in 13%, 26% and 22% of pMN, sMN, and tMN, respectively.For tMN and sMN, prostate (20%, 32%), breast (30.5%, 16%), and colorectal cancers (7%, 16%) were the most common PM. Hematologic neoplasia as PM was more common in tMN than sMN (25% vs . 3.7%, p<.0001). Family history of malignancy was enriched in pts with pMN vs. sMN and tMN (OR 1.7, p=0.001, OR 1.6, p=0.057). Del7/7q was most common in tMN vs. pMN (OR 2.8, p<.0001) and sMN (OR 3.6, p=0.0004). Complex karyotype was enriched in tMN vs. sMN and pMN (30%, 13% 16%, p<.0001). NGS for somatic mutations (60-gene panel) was available for 145 tMN, 65 sMN, and the remaining 683 pts with no history of PM prior to MN dx (primary MN or pMN). To determine differences in patterns of somatic driver lesions in each group, we compared frequency of all mutations (^MUT) as well as ancestral hits, as assessed by recapitulation of clonal hierarchy using variant allele frequency and zygosity.

General somatic patterns of acquisition were similar in frequency and ancestral hits in tMN, pMN and sMN and featured TET2, DNMT3A, and SRSF2^MUTS. However, there were significant differences. TP53 (13.1%), EZH2 (7.6%), WT1 (2.8%), and KIT^MUTS (3.5%) were overrepresented in tMN vs. pMN (OR 2.6, p=0.002 ; OR 1.9, p=0.083 ; OR 3.2; p=0.08 ; OR 6.1, p=0.01) and sMN (OR 2.3, p=0.15; OR 1.3, p= 0.78 ; OR 1.8, p=1 ; OR Infinity, p=0.186). TP53 and EZH2^MUTS were also more common ancestral hits in tMN (11%, 4.8%) vs. pMN (3.5%, 2%) (p=0.0004, p=0 .07). TP53^MUTpersisted in being associated with tx as an ancestral hit when sMN (3.1%) was used as a control (p=0.06) . In tMN, IDH1, SRSF2, CUX1, and PHF6^MUTS (0.7%, 10.3%, 3.5%, 4.1%) were underrepresented relative to sMN (7.7%, 20%, 7.7%, 9.2%) (p=0.011, p=0.077, p=0.29, p=0.19). Indeed, these were more frequent in sMN than pMN (OR 3.07, p=0.042 ; OR 1.99, p=0.044 ; OR 3, p=0.042 ; OR 4.85, p=0.005) and were ancestral events in 4.6%, 12.3%, 1.5%, 1.5% sMN cases.

Specific tx modalities have been associated with molecular lesions. We analyzed whether or not rtx and ctx had the same impact on the profile of somatic lesions. TP53^MUT correlated primarily with ctx vs rtx (OR 7.2 , p=0.06). However, it was highest when both were used vs. rtx (OR 11.5, p=0.01). Notably, TP53^MUT tMN cases were associated with complex karyotype relative to TP53^WT tMN cases (OR 20, p<0.0001). ETV6 and EZH2^MUTS were found only in pts with rtx or both, never ctx alone (p=0.05, p=0.004, respectively). In addition to increasing mutational frequency, rtx also increased the propensity for EZH2^MUT to become the ancestral hit relative to pMN (OR 2.4, p=0.07) and sMN (OR 1.5, p=0.7), albeit less significantly. EZH2^MUT had no influence on complex cytogenetics.

In conclusion, our study demonstrates that tMN are molecularly diverse with increased frequency of specific hits including TP53, EZH2, WT1, and KIT^MUT. The tMN entity may be contaminated by cases with sMN with a hereditary component or spontaneous sMN, in whom molecular patterns may be distinct. However, we were unable to establish an entirely distinct molecular signature related to cytotoxic therapy.