P729: DEREGULATION OF NONCODING RNAS DERIVED FROM KDM GENES IS ASSOCIATED WITH AZACITIDINE RESPONSE IN PATIENTS WITH MYELODYSPLASTIC SYNDROMES AND ACUTE MYELOID LEUKEMIA

Abstract

Background: Prediction of response to azacitidine (AZA) treatment is an important challenge in the management of higher-risk myelodysplastic syndromes (MDS) and acute myeloid leukemia (AML). To date, no clinical, cytogenetic, or molecular markers of AZA treatment outcome have been validated to support clinical decisions. Moreover, little is known about how AZA efficiency can be affected by various noncoding RNAs (ncRNAs), such as long ncRNAs (lncRNAs) and circular RNAs (circRNAs). KDM genes encode demethylases of histone lysine residues (e.g., KDM1 mediates demethylation of H3K4 and H3K9, and KDM4 acts on H3K9, H3K36, and H1K26). KDMs are often differentially expressed in leukemia and they cooperate with transcription factors to activate or repress gene expression. Aims: Our aims were to provide biological insight into the contribution of ncRNAs to AZA mechanisms of action and to propose novel disease biomarkers that would be able to predict future response to AZA treatment. Based on our results, we particularly focused on expression and function of KDM-related ncRNAs. Methods: Whole transcriptome RNA sequencing (RNA-seq) was performed in CD34+ bone marrow cells from 26 patients with MDS or AML with myelodysplasia-related changes (AML-MRC) before AZA treatment (11 responders and 16 nonresponders), and 9 healthy controls. Expression of protein coding genes (PCGs), lncRNAs and circRNAs was bioinformatically processed. Results: Overall, differential expression analysis between AZA responders and nonresponders identified significant deregulation of 202 PCGs, 34 lncRNAs, and 21 circRNAs. Surprisingly, machine learning data showed that predictive potential of ncRNAs was stronger than that of PCGs. Within the data, deregulation of several KDM-related ncRNAs was observed, e.g., CTC-482H14.5 (antisense RNA to KDM4B gene, downregulated in AZA responders), hsa_circ_0003889 (circRNA processed from KDM1A, downregulated in AZA responders), and hsa_circ_0001580 (circRNA processed from KDM1B, upregulated in AZA responders). By contrast with these noncoding transcripts, the levels of corresponding PCGs (namely KDM1A/1B/4B) were not altered between AZA responders and nonresponders, suggesting that the deregulated ncRNAs play specific roles independent of their host genes. Further, the best predictor genes were defined by machine learning using RNA-seq data and pathway analysis linked these markers to cellular processes related to MDS/AML. Interestingly, CTC-482H14.5 was identified as one of the strongest predictors of AZA treatment response. Expression of this KDM-related ncRNA was coregulated with expression of DNMT1, EZH2, and multiple MCM genes, associating this transcript with epigenetic regulation and recombinational repair pathways. Summary/Conclusion: Several ncRNAs processed from KDM genes seem to be closely related to the responsiveness of MDS/AML-MRC patients to AZA treatment. In this study, novel functions were predicted for these ncRNAs, pointing to possible mechanisms by which their deregulation could affect AZA treatment response. Acknowledgements: Supported by GA CR (No. 20-19162S) and MH CZ-DRO (UHKT, 00023736).