Applicability of DHPLC for Low Sensitive Residual Disease Detection in RUNX1 and CEBPA Mutated AML.

Abstract 3067

Poster Board III-4

During the last years it has been shown that PCR based detection of minimal residual disease (MRD) has high relevance for early detection of relapse and overall prognostication. This has been proven for several fusion transcripts but also for NPM1 as a target in normal karyotype AML (NK-AML). Other mutations frequently found in NK-AML are RUNX1 mutations (8-10%) and CEBPA mutations (10-15%). However, these mutations are distributed throughout the entire coding reading frames of CEBPA and RUNX1 making mutation analyses more laborious compared to analysis of genes with mutational hotspots like NPM1. In addition, it is nearly impossible to establish high sensitive real time PCR assays for every patient specific mutation. In contrast, DHPLC (denaturing high performance liquid chromatography) is a method that effectively can detect unknown mutations and for known mutations has a sensitivity of up to 1%. Therefore we analyzed the impact of DHPLC analysis for the applicability and value as predictive MRD analysis. At diagnosis mutation screening by DHPLC was performed first. Both genes were amplified each by four different PCR reactions that were subsequently analysed on a WAVE system (Transgenomic, Inc., Omaha, USA). Positive reactions were further characterized by sequencing. The respective fragment or fragments containing the mutations was/were subsequently also analyzed in the follow up samples. The sensitivity was dependent on the kind of the mutation and its position within the PCR fragment and was between 1 and 10% as estimated by limited dilution experiments. Paired diagnostic and relapse samples were available in 15 cases (12 RUNX1 and 3 CEPBA). The respective mutations were retained at relapse in all cases indicating the stability of both markers, rendering them eligible for follow up evaluation. Next, we analysed 30 patients with CEBPA mutation detected at diagnosis and further investigated 91 samples during follow up. 12 of these cases had two different mutations that were localized on two different DHPLC fragments and thus could be analysed in parallel. For RUNX1 we analysed 144 follow up samples of 60 patients that revealed one (n=51) or two (n=9) RUNX1 mutations at diagnosis. Six of the CEBPA mutated and 13 of the RUNX1 mutated cases had an FLT3-ITD in addition. The median follow up sample number per patient was 3 (range 2-13) and the median follow up time was 339 days (range: 57-3001 days). In the subsequent analysis both cohorts were combined. The follow up samples were simply rated as negative or positive. According to previous studies in fusion gene positive patients and NPM1 mutated patients the impact of the DHPLC results on survival was analysed for defined time intervals after start of treatment: interval 1 (days 18-60), interval 2 (days 61-120), interval 3 (day 121-365), interval 4 (days >365). DHPLC results within these intervals were as follows: interval 1 (positive: n=16; negative n=17), interval 2 (positive: n=14; negative n=19) interval 3 (positive: n=38; negative n=65), interval 4 (positive: n=22; negative n=38). Whenever in the follow up samples two different mutations were available (n=99), the results were shown to be concordant. The impact of the results was analysed by Kaplan Meier analysis. For overall survival a trend for significance was found for interval 1 (medians not reached; p=0.157) and interval 2 (medians not reached; p=0.090) and a significant impact for interval 3 (median: not reached vs. 981 days; p=0.015) and interval 4 (median not reached: vs. 285 days; p=0.048), demonstrating that negative DHPLC results correlate with longer OS. This effect could even more clearly be shown for event free survival with respective results for interval 1 (median: 463 vs 731; p=0.048), interval 2 (median: 499 vs 731 days; p=0.109) and interval 3 (p<0.001) (too few samples for interval 4). As neither age, WBC or pretreatment FLT3 status were significantly associated with outcome in this cohort a multivariate analysis could not be performed. These data clearly show that in the absence of sensitive markers for RQ-PCR low sensitive PCR can be very useful for follow up controls at least in RUNX1 and CEBPA mutated AML.