Abstract
Objectives
To investigate the clinical relevance of somatic mutations in core binding factor (CBF) acute myeloid leukemia (AML) patients at diagnosis and relapse and hierarchy of mutation acquisition using serial sequencing.
Patients and Methods
Eighty seven patients (pts) diagnosed with CBF AML were enrolled in this study (62 pts carrying RUNX1-RUNX1T1 /t(8; 21) and 25 pts carrying CBFB-MYH11 /inv(16). Median follow-up duration among survivors was 56.9 months (range 0-155.3). Using a 84 gene panel, we performed targeted deep sequencing in 357 samples using Illumina Hiseq 2500. Sequenced samples include bone-marrow/peripheral blood taken at initial diagnosis (n=87) and T-cell (CD3+, n=68), CD34+/CD38-(n=66) and CD34+/CD38+ (n=68) fractions. In addition, samples taken at complete remission (n=53), and relapse (n=15) were sequenced. Average of on-target coverage was 1657.8x.
Results
We detected 166 mutations in 79/87 pts at time of diagnosis (90.8%, median 2 mutations/pts, range 0-7). At diagnosis, KIT (39%), NRAS (33%), ASXL2 (14%), KRAS (13%), RAD21 (7%), and FLT3 (6%) were commonly mutated. When grouped by biological pathway, frequencies of mutations in RAS ( KRAS or NRAS ), chromatin modifiers and cohesin complex were significantly different among two sub-populations (p = 0.002, 0.01, and 0.02, respectively, Figure A). Survival analyses show that KIT -D816mut is an adverse prognostic factor for overall survival (hazard ratio (HR) 2.31, 95% CI [1.04-5.14], p=0.04) and relapse (HR 2.76 [1.06-7.16], p =0.04). Mutation in RAS was a favorable factor for relapse (HR 0.13 [0.03-0.56], p=0.001, Figure B). Multivariate analysis confirmed that only RAS mutation was a significant favorable factor for the incidence of relapse (HR 0.20 [0.05-0.86], p=0.03).
Taking advantage of serial sequencing and real-time polymerase chain reaction (PCR) data ( RUNX1-RUNX1T1 ), we inferred mutation dynamics and clonal hierarchies. Mean allelic burden at diagnosis and relapse were comparable (mean variant allele frequency (VAF) = 22.17% and 22.22%, respectively), whereas mutations were nearly cleared at complete remission (mean VAF = 0.25%) (Figure C). With initial genetic rearrangement detected at both diagnosis and relapse, 14/15 pts carried additional 39 mutations throughout the course of the disease. Fifteen mutations were stable, whereas 9 and 15 mutations were cleared/decreased and acquired/selected at relapse. Inference based on mutation dynamics revealed stable mutations are more likely to be earlier events than mutations that are cleared at relapse (p = 0.01, Figure D). Furthermore, integrated analyses of reduction rate of the RUNX1-RUNX1T1 and mutation dynamics showed RUNX1-RUNX1T1 is no later event than all targeted mutations in our cohort, further explaining clonal hierarchies of CBF AML with RUNX1-RUNX1T1(Figure E) .
Conclusion and Summary
Current study provided distinct mutation patterns between two sub-population of CBF AML. Multivariate analyses showed that presence of mutation in KRAS or NRAS was a favorable prognostic factor. In addition, results from longitudinal sequencing combined with RUNX1-RUNX1T1 level provides further insights on the order of genetic mutation acquisition during leukemogenesis.
No relevant conflicts of interest to declare.
