Mutational Signatures Associated with Intensity and Duration of Smoking in Myelodysplastic Syndromes (MDS)

Background:MDS pathogenesis is a multi-step molecular process resulting from somatic mutations acquired over time. Smoking is a known risk factor for MDS which may accelerate this process. To further investigate the contribution of tobacco use in MDS development, we studied the association between smoking and specific genomic alterations in MDS patients (pts).

Methods: MDS and CMML pts diagnosed between 1996 and 2016 with both smoking histories and sequencing data were included. The Center for Disease Control definition defined current, ex- and never smokers, the latter defined as pts with zero smoking duration/packs per day (PPD). Next generation targeted sequencing was performed for a panel of 60 genes known to be commonly mutated in myeloid malignancies. Fisher's exact test and Wilcoxon's test compared proportions and medians, respectively. Poisson regression modeled total numbers of mutations per pt. The probability of a mutation in a particular gene was modeled using conditional logistic regression. Adjustment was done for both sex and age, fits were stratified using age cut points of 0, 40, 50, 60, 65, 70, 75, 80, 85 and 90 years (y). Odds ratios (OR) were found as exponentials of gene status coefficients of such fits.

Results: In 672 pts, the median age at diagnosis was 68y (range, 20-103), for pt characteristics, see Table 1. Median OS 34 mo (95% CI 30-39) and median follow-up time was 21.5 mo (IQR, 10-45). Median tobacco exposures were 30 pack-years (py) in current smokers (range, 6-80) and 25 in ex-smokers (range, 0.3-175), P=.2. More men than women ever smoked (74% v 45%; P <.0001). Longer duration of smoking associated with higher IPSS-R categories (P=.012) and cytogenetic risk scores, P=.037. Most pts (502, 75%) had at least one mutation; 23% had a single gene mutation, 19% had 2, 15% had 3, and 18% had >3. The most frequently mutated genes were TET2 17%, ASXL1 16%, SRSF2 14%, SF3B1 12%, DNMT3A 10%, RUNX1 9%, STAG2 9%, and TP53 7%.

Poisson regression analysis was used to assess the impact of smoking duration (y) or PPD on mutation acquisition. A positive correlation of greater smoking intensity (> 2 PPD v <0.5, relative increase (RI) =1.46, P=.011) and longer duration of smoking (>40y v <10, RI = 1.31, P=.011) was seen with increased number of mutations [Figure 1].

Impact of duration of smoking was evaluated using logistic regression. Duration of smoking >40y v <1 was significantly associated with the following molecular abnormalities: FLT3 (P= .036), EZH2/ del7 (P=.038) and NRAS (P=.046). Comparing 20-40y v <1 yielded SETBP1 (P=.005), DNMT3A (P=.010) and EZH2 /del7 (P=.036). Comparing 10-20y v <1 yielded STAG2 (P=.039) and NOTCH1 (P=.046). Finally, comparing durations of 1-10y v <1 yielded PHF6 (P=.016). OR per mutation, see Table 2.

Focusing on intensity of smoking, >2 PPD v <0.5 was significantly associated with higher rates of mutations in FLT3 (P=.013), EZH2 /del7 (P=.024), IDH1 (P=.041), and NPM1 (P=.044). Even low intensity smoking, comparing 0.5-1 PPD v <0.5, was significantly associated with mutations in SETBP1 (P=.002), DDX54 (P=.042) and DNMT3A (P=.046). For the combination of smoking duration and intensity, >40py v <10 was associated with IDH1 mutation (P=.046), while 20-30 py v <10 significantly associated with RAD21 (P=. 016), DDX54 (P=.018), SRSF2 (P=.041), del7/ EZH2 (P=.042), and NF1 (P=.045) mutations. Comparing 1-10py v <1, significant associations were found for IDH1 (P=.004), PRPF8 (P=.011), SETBP1 (P=.022), ETV6 (P= .040) and TET2 (P=. 047)mutations [Table 2]. Analyses were repeated after excluding CMML and MDS/MPN pts to control for the common genomic phenotype seen in these disorders, with identical results.

Conclusions: Tobacco smoking predisposes to either acquisition or selection of distinct somatic mutational patterns from those seen in "spontaneous/non-tobacco exposed" MDS. EZH2/ del7 abnormalitiesassociated with longer durations and heavier smoking patterns, while SETBP1 mutation associated with lighter smoking for prolonged durations. Lesions associated with light smoking for years may be related to chronic inflammatory changes, whereas those associated with higher intensity may reflect direct DNA damage. Validation of these findings warrants earlier sequencing in pts with pre-MDS conditions such as clonal hematopoiesis of indeterminate potential to better identify the onset of somatic mutations and their role in MDS pathogenesis.

View largeDownload slide

View largeDownload slide

Close modal