Prognostic and Biologic Significance of Long Non-Coding RNA (lncRNA) Profiling in Cytogenetically Abnormal Acute Myeloid Leukemia (CA-AML)

Introduction: Aberrant expression levels of lncRNAs have been shown to independently associate with outcome of younger and older patients (pts) with cytogenetically normal AML. However, the prognostic and biologic significance of lncRNA expression in CA-AML pts have not been extensively studied.

Methods: We performed whole transcriptome profiling (RNA-seq) in 469 pts with de novo CA-AML. Cytogenetic analyses were performed in institutional laboratories and the results were reviewed centrally. All pts were treated on frontline Cancer and Leukemia Group B (CALGB)/Alliance protocols.

Results: To evaluate the prognostic significance of lncRNA expression in CA-AML, we analyzed RNA-seq data of 469 pts by applying a machine learning algorithm-based approach. As CA-AML pts constitute a heterogeneous group, we first determined which other clinical and molecular parameters were prognostic [i.e., associated with event-free survival (EFS)] in our dataset. Among the parameters tested, the European LeukemiaNet (ELN) Risk Group status and age group [i.e., younger than 60 years (y) or aged 60 y and older] significantly associated with clinical outcome of CA-AML pts. Next, we individually identified each lncRNA that associated with EFS while adjusting for ELN Risk Group and age group. We conducted random forest analyses to select the prognostic lncRNAs, whose combined expression levels could generate an effective outcome predictor for CA-AML pts. For each step of the random forest analyses, a bootstrap technique was applied; a simple random sample of pts was drawn which served as the training set and the out-of-sample pts were used as the independent validation set. We identified 55 prognostic lncRNAs and used their expression levels to separate our CA-AML cohort into a lncRNA low-risk (n=161) and a lncRNA high-risk (n=308) group. With regard to clinical characteristics, pts in the lncRNA low-risk group were younger (P<.001) and had lower platelet counts (P<.001) and higher white blood cell counts (P=.01) than pts in the lncRNA high-risk group. Concerning cytogenetic abnormalities, pts in the low-risk group more often had core-binding factor translocations or inversions (P<.001) and less often complex karyotypes (P<.001) than pts in the high-risk group. Pts in the lncRNA low-risk group had higher complete remission (CR) rates than pts in the high-risk group (91% vs 48%, P<.001). LncRNA low-risk group status also associated with longer disease-free survival (DFS; 5-y rates 49% vs 12%, P<.001), overall survival (OS; 5-y rates: 58% vs 14%, P<.001) and EFS (5-y rates: 45% vs 6%, P<.001). With regard to the accuracy of outcome prediction, the lncRNA risk classification had a C-index of 0.73, which compares favorably with other prognostic classifiers of AML pts. In multivariable analyses, lncRNA low-risk status was an independent marker for higher CR rates, as well as for longer DFS, OS and EFS (P<.001 in all comparisons), after adjusting for other covariates. Finally, we examined the prognostic value of the lncRNA risk classification within the Favorable and Intermediate ELN Groups of our dataset, for which lncRNA risk groups had adequate pt numbers. Among pts in ELN Favorable Group, lncRNA low-risk pts (n=128) had higher CR rates (P=.003) and longer DFS (P<.001), OS (P<.001) and EFS (P<.001) than lncRNA high-risk pts (n=32). Similarly, in the ELN Intermediate Group (n=85), lncRNA low-risk group status (n=28) associated with higher CR rates (P=.01), longer OS (P=.01) and EFS (P=.005) and a trend for longer DFS (P=.08).

To gain biological insights, we examined the molecular pathways regulated by the 55 prognostic lncRNAs. To minimize the confounding effects of differences in the concurrent cytogenetic abnormalities, we restricted these analyses to the ELN Favorable Group. We identified approximately 900 transcripts that were differentially expressed between lncRNA low- and high-risk pts. DAVID pathway analyses showed enrichment of genes involved in the processes of phosphorylation, acetylation and RNA-binding. Ingenuity pathway analyses of up-stream regulators identified aberrant activity of homeobox genes such as MEIS1, HOXA9 and HOXA10 in the lncRNA low-risk group and other transcription factors such as MYC, FOSB and JUN in the high-risk group.

Conclusion: We conclude that lncRNA profiling provides meaningful prognostic and biologic information in CA-AML pts.