Correlation of Mir-155 Expression with Clinical Outcome in Childhood AML: A Report from Children’s Oncology Group

Acute myeloid leukemia (AML) is characterized by genomic abnormalities that impair differentiation and promote proliferation. Karyotypic alterations and somatic mutations are associated with relapse risk but are rarely predictive of response to induction chemotherapy. Aberrant miRNA expression has been implicated in AML pathogenesis. We present association of altered miR-155 expression with specific molecular characteristics and show its association with high risk of induction failure and worse clinical outcome in children with normal karyotype (NK) AML treated on COG AML trial, AAML0531.

Of 1022 patients enrolled, 233 were NK. 198 diagnostic specimens were available for testing. miR-155 expression was evaluated by quantitative TaqMan miRNA assays normalized against normal marrow. Expression varied by 4 log fold. Patients were divided into 4 quartiles (Q1-Q4) based on expression levels. Q1 had patients with the lowest and Q4, with the highest expression. We correlated disease characteristics and clinical outcome across quartiles. All comparisons and data presented is between high (Q4) vs. low (Q1-Q3) expressors.

Males (66% vs. 48%, p=0.027), Asians (15% vs. 5%, p=0.048) and high median diagnostic WBC (59.4 vs. 22.6, p=0.002) were over-represented in high expressors.

Molecular abnormalities were used to stratify risk. Low-risk patients were those with mutations in CEBPA/NPM and high-risk, in FLT3-ITD. Others were designated as standard-risk. High miR-155 expression was associated with a high prevalence of FLT3-ITD mutations (69% vs. 31%, p < 0.001) and high-risk disease (52% vs. 15%, p<0.001), and had an inverse association with low-risk (22% vs. 38%, p=0.041) and standard-risk disease (26% vs. 47%, p=0.008). There was no association with mutations in CEBPA (p=0.710) or NPM (p=0.826).

We evaluated expression levels and response to induction chemotherapy, morphologic complete remission (CR). CR rates for high vs. low expressors was 46% vs. 83% (p<0.001). As high expressors were more likely FLT3-ITD+, we computed CR for those without FLT3-ITD mutations. High vs. low expressors who were FLT3-ITD-, had CR rates of 47% vs. 86% (p=0.002), suggesting that miR-155 expression may independently provide data on response to chemotherapy

We evaluated overall survival (OS) and event-free survival (EFS) at 3 years. OS for high vs. low expressors was 51% vs. 75% (p=0.002) and EFS was 32% vs. 59% (p<0.001). Association of high miR-155 and worse survival was also verified against RNASeq data, in an independent cohort of children with NK-AML (p=0.005). We performed Cox regression analyses to evaluate the impact of miR-155 expression level as a predictor of clinical outcome in the context of prognostic factors (FLT3-ITD, NPM, CEBPA mutations, treatment arm, risk group, age, WBC) and molecular risk groups using them as covariates in both univariate and multivariate models. In the univariate model, high expression was a prognostic factor for inferior OS (HR of 2.19, p= 0.002) and EFS (HR 2.24, p<0.001). Patients with low-risk molecular features also had improved OS (HR 0.29, p<0.001) and EFS (HR 0.22, p<0.001). In a multivariate model that included the aforementioned prognostic features, high miR-155 expression retained prognostic significance for OS (HR 1.92, p=0.022) and EFS (HR 1.75, p= 0.019). High risk molecular features were no longer an independent predictor of outcome. Low-risk molecular features remained an independent prognostic factor for improved OS (HR 0.34, p=0.005) and EFS (HR 0.22, p < 0.0001).

We extended our study to identify target genes whose expression was impacted or regulated by miR-155 expression. Comparison of the mRNA expression profile for the 23 highest vs. 23 lowest expressors identified anti-correlation of miR-155 and 15 target genes. High miR-155 levels were associated with down-regulation of 9 genes, including E2F2 and ETS1. Low miR-155 levels were associated with up-regulation of 6 genes, including DET1, GATM and SMAD1.

We demonstrate association of miR-155 with specific disease characteristics and clinically significant mutations in pediatric AML. We also show that high expression is highly predictive of induction failure and outcome providing potential biomarkers for identifying patients at high risk of poor response prior to therapy. miRNA expression could provide clinically important information for use in therapeutic allocation.