RNA Expression Profile Using Targeted NGS As a Potential Predictor of Early Molecular Response and Relapse in Core-Binding Factor Acute Myeloid Leukemia

Background: Core-binding factor (CBF) acute myeloid leukemias (AML) are classified as favorable prognosis subgroup but outcomes do not approach that of acute promyelocytic leukemia. We and others have shown that that 3-log reduction in fusion transcript by quantitative PCR (qPCR) at the end of induction portends for better outcome. In this study, we focused on finding gene expressions by RNA sequencing from samples at first diagnosis that could predict early molecular remission. In addition, RNA seq allowed us to identify biological difference distinguishing between the two cytogenetic subgroups [Inversion (Inv) 16 vs t (8;21)] that potentially can be used therapeutic approaches and to detect low frequency mutations.

Methods: Samples collected at diagnosis from forty-five patients with CBF-AML, treated with fludarabine, cytarabine and granulocyte colony stimulating factor (FLAG) based regimen were analyzed. RNA sequencing was performed by hybridization based targeted panel that detects fusion, expression, and mutations in the 1385 genes (Illumina, San Diego, CA). Each sample was deeply sequenced (>55M reads) using Illumina paired-end 75 bp sequencing and duplicate reads were removed before fragment per kilobase of transcript per million mapped reads (FPKMs) were calculated for each gene. The expression data was mean centered around expression of ABL1. Cytogenetic subgroups were classified by a Bayes multi-variate logistic regression model. The models were evaluated by the Leave One Out Cross-validation technique.

Results: The median age of patients was 36 years (range, 22-78); 26 patients (58%) had Inv16 and 19 (42%) had t(8;21). All patients achieved remission and 7 patients relapsed; 4 had t(8;21) and 3 had Inv16. The expression of RUNX1T1, NR6A1, VEGFA, PLCG1, and CD58 mRNA was significantly (P<0.00001) higher and MYH11, IRF8, SPECC1, LPAR1, and TNFRSF11A were significantly (P<0.00001) lower in t(8;21) than in Inv16. High expressions of GSN (p = 0.0005), TACC3 (p = 0.0012) were significantly associated with 3-log reduction in qPCR post-induction in patients with Inv16, as compared to CCDC6, SDHA, NPM1 (all p < 0.005) in patients with t(8;21). In addition high expressions of CD44, FUS, MKL1, TCF7L2 and ELK4 were associated with t(8;21) with a relatively lower degree of significance (all p<.015). Expression of ENPP2, SOCS3, and CIC correlated negatively (P<0.001) with relapse. In multivariate analysis, expression of CIC, ENPP2, and USP6 genes were strong predictor of relapse for all patients (F-score=92.6%), while the expression of USP6 and FASLG genes were predictor of relapse in Inv16 patients (F-score= 97.9) and the expression of CSBPD and PRICKLE1 were predictor of relapse in t(8;21) with F-score=100%.

Our analysis also identified several mutations not commonly described in CBF AML; RICTOR (n=5), ACACA (n=4), FIP1L1, KDM5A, MSH6 (all n=3), KMT2D, ATM, PRKCA (all n= 2). Mutations in 43 other genes were seen in only one patient each. Presence of KIT (n = 16, 36%) or FLT3 (n = 10, 22%) mutations did not predict for relapse and this was consistent our prior analysis.

Conclusion: Our data suggests that baseline targeted RNA NGS profiling, not only will detect the Inv16 and t(8;21) fusion RNA, but might help to predict early molecular response and relapse in patients with CBF-AML. Pathway based analysis to understand underlying biology of cytogenetic subgroups and molecular response prediction is ongoing. Our next steps will include accessing publicly available databases for further validation of data.