Different Genomic Patterns in Patients with Primary Acute Myeloid Leukemia (AML) Compared to Secondary AML in Patients with Normal Karyotype

Background

For decades, cytogenetic analysis has played an essential role in AML risk stratification. Among the 50% of AML patients (pts) with normal karyotype (NK), outcome can vary widely. More recently, whole genome sequencing (WGS) and whole exome sequencing (WES) have identified several recurrent mutations that play an important role in AML pathogenesis and impact outcome. Pts with secondary AML (sAML) have a particularly poor prognosis, are not as responsive to standard induction chemotherapy, and often are referred in first complete remission to hematopoietic stem cell transplantation. We hypothesized that different genomic patterns exist between primary AML (pAML) and sAML that can distinguish the two, and can alter treatment recommendations. To negate the impact of chromosomal abnormalities, we focused our analyses on pts with NK.

Methods

We performed WES and multi-amplicon targeted deep sequencing on samples from bone marrow and peripheral blood of pts diagnosed with sAML at our institution between 1/2003- 1/2013 and who had NK cytogenetics. We compared them to pts with NK primary AML (pAML) whose data were extracted from The Cancer Genome Atlas (TCGA). A panel of 62 gene mutations that has been described as recurrent mutations in myeloid malignancies was included. Mutations were considered individually and grouped based on their functional pathways: RNA splicing (SF3B1, U2AF1/2, SRSF2, ZRSR2), DNA methylation (TET2, DNMT3A, IDH1/2), chromatin modification (ASXL1, EZH2, MLL, SUZ12, KDM6A), transcription (RUNX1, CEBPA, NPM1, BCOR/BCORL1, SETBP1, ETV6), activating signaling (FLT3, JAK2), cohesion (STAG2, SMC3, RAD21), RAS superfamily (K/NRAS, NF1, PTPN11, CBL) and tumor suppressor genes (TP53, APC, WT1, PHF6). Using deep sequencing methodology for resequencing or targeted sequencing, variant allelic frequency (VAF) was measured for each mutation detected. VAF was adjusted by zygosity evaluated by SNP-array karyotyping. For confirmation of clonal architecture, serial sample sequencing and single colony PCR were applied. Differences were compared using Fisher-exact test and Mann-Whitney U test for categorical and continues variables respectively.

Results:

Of 143 pts included, 101 (71%) had pAML and 42 (29%) had sAML. Compared to pAML, sAML pts were older (59 vs 69 years, p <.001), and had lower white blood cell count (28 vs 3.5 X 10⁹/L, p <.001). Median hemoglobin (10 vs 10) g/dl and platelet counts (57 vs 60) k/uL were similar between the two groups. With a median follow up of 26.4 months (mo, range, .93-95.4), median OS was shorter for sAML than for pAML (12.9 vs 16.2 mo, p= .03). Overall, the most common mutations were: NPM1 (35%), DNMT3A (27%), FLT3 (25%), RUNX1 (14%), IDH1 (12%), IDH2 (12%), STAG2 (12%), TET2 (11%), NRAS (8%), ASXL1 (8%), U2AF1 (8%), PTPN11 (7%), WT1 (6%), BCOR (5%), and PHF6 (5%). Mutations in SF3B1, U2AF1/2, BCOR/BCORL1, ETV6, ASXL1, JAK2, STAG2, and APC were more common in sAML compared to pAML, whereas mutations in DNMT3A, NPM1, CEBPA, and FLT3 were more common in pAML. Mutations in activated pathways in splicing machinery, transcription, chromatin modification, cohesion and RAS pathway were more prominent in sAML, while mutations in DNA methylation and signaling pathways occurred more frequently in pAML. Serial sample analyses at multiple time points demonstrated intra-tumor heterogeneity in most cases of sAML, which was supported by additional cross sectional analyses of VAF in multiple gene mutations in each case. These findings prompted us to evaluate secondary events in the cohort of pts whose sAML originated from an initial MDS stage, defined by ancestral mutations. Among genes frequently affected by mutations, TET2 and ASXL1 were identified as founder events, whereas STAG2, NRAS and PTPN11 were observed in subclonal sAML derived from founder MDS clones. In pAML, however, TET2 and ASXL1 mutations were found to be secondary lesions, while IDH1 and DNMT3A were identified as ancestral events.

Conclusion

Clear genomic variations exist between sAML and pAML that suggest differences in the pathophysiology of both diseases. Specific therapies should be directed to the activated pathways according to the unique clonal hierarchy in each AML subtype.