Dysregulation of Developmental Transcription Factors in Acute Myeloid Leukemia Predicts Poor Outcome and Impacts Adaptive Immune Pathway Genes

Introduction

Pediatric acute myeloid leukemia (pAML) involves uncontrolled proliferation of myeloid progenitor cells. Recent studies, including the TARGET initiative, show a majority of pAML to be driven by molecular abnormalities, frequently in genes regulating transcription. The homeobox domain, a highly conserved 60 amino acid DNA binding motif, present in over 300 human transcription factor genes is key in regulating early development. It is observed in all 39 HOX genes and TALE family co-factors. Understanding the global patterns of dysregulation in these master developmental genes is key to a comprehensive understanding of leukemogenesis and better outcome prognostication for patients.

Identifying Dysregulation

To identify which developmental transcription factors are dysregulated in pAML we employed RNA sequencing data from 1,534 diagnostic samples compared to 68 non-diseased bone marrow and 16 CD34+ peripheral blood controls. Unsupervised hierarchical clustering methods utilizing the landscape of homeobox domain containing genes revealed distinct inter and intra subtype segregation as well as separation from normal bone marrow controls. Subtypes including RUNX1-RUNX1T1, CBF-MYH11, CBFA2T3-GLIS2, FLT3+ and KMT2A-rearranged pAML differentiated themselves on expression of these early developmental transcription factors and formed homogeneous clusters of up and down-regulated homeobox containing genes. Importantly, of the 346 human homeobox genes we identified 17 which are not expressed in normal bone marrow, yet differentiated pAML. This cluster included nine members of the HOX A cluster and 4 TALE family genes (MEIS1, PBX3, MKX, IRX3) which we call our DDTF (Dysregulated Developmental Transcription Factor) cluster. Kmeans clustering of pAML using DDTF genes resulted in three groups with mean expressions of 21, 55 and 103 transcripts per million.

Survival Analysis

We evaluated the prognostic value of DDTF gene expression by comparing expressors vs non-expressors using the Kaplan-Meier method. Expression threshold was determined dynamically using the log-rank statistic (Survminer package, R). Cluster expression was defined as the mean value of gene expression across the DDTF cluster in transcripts per million. Outcomes were markedly different between expressors and non-expressors. 5 year overall survival (5-yr OS) for expressors of the DDTF cluster was 53% vs. 79% for non-expressors (p < 0.0001). In KMT2A rearranged pAML clusters had 5-yr OS of 31, 59 and 70% respectively. High DDTF expression consistently indicated poor outcome. To further validate that DDTF expression was associated with poor outcome and not simply identifying poor risk pAML subtypes we separated high risk from low and standard risk into two different cohorts and re-applied our survival analysis to very similar results. In 38 samples with diagnostic, remission and relapse samples we tracked DDTF expression and found a decrease in DDTF at the remission timepoint followed by a concordant increase at relapse.

Transcriptomic Impact

To identify the global transcriptomic impact the dysregulation of DDTF genes has in the context of KMT2A rearranged pAML we performed a differential expression analysis comparing above median expressors to below median expressors. Greater than 1,000 genes were significantly (p. adj < 0.05) differentially expressed. Gene set enrichment analysis identified significant suppression of pathways related to adaptive immune response including ‘T cell activation’ and ‘immune response mediation’ in the over DDTF expressing group. Activated pathways were enriched for early developmental pathways and included ‘embryonic morphogenesis’ and ‘pattern specific processing’.

Conclusions

Our data indicate that dysregulation of developmental transcription factors is both widespread in pAML and occurs in coordinated subtype specific patterns. We further find that increased expression of specific HOX and TALE family factors we call DDTFD, are prognostic of patient OS, especially in KMT2A-rearranged pAML. Dysregulation of these genes leads to global changes in gene expression impacting functional pathways including crucial immune response pathways.