P751: CHARACTERIZING CIRCULAR RNA EXPRESSION IN MYELODYSPLASTIC SYNDROME

Background: Circular RNA (circRNA) research is an expanding field, reflecting a recognition that circRNAs may have various roles in cancer pathology. CircRNAs are formed when back-splicing events create covalently closed loop structures from pre-mRNA molecules (Kristensen LS, Nat Rev Genet, 2019). CircRNA expression is cell-type specific, also during hematopoietic differentiation (Nicolet BP, Nucleic Acids Res, 2018), but their biological significance in myeloid cancers is still largely unknown. Aims: We aimed to profile circRNA expression in myelodysplastic syndrome (MDS), along with the related diseases chronic myelomonocytic leukemia (CMML) and clonal cytopenia of undetermined significance (CCUS), in order to identify disease-specific expression patterns and relevant candidates for biological or biomarker exploration. Methods: Seventy-one patients and eight healthy age-matched controls were included. Bone marrow aspirate underwent FACS sorting to obtain CD34+ stem and progenitor cells. RNA was extracted and total RNA libraries were created using enzymatic ribosomal RNA removal. The RNA was sequenced on a NovaSeq 6000 (mean 270 million paired-end reads per sample, read length 150 bp). Bioinformatic pipelines find_circ and CIRI2 were used to identify and quantify reads originating from the back-splicing junctions of circRNAs. Detection by both pipelines was required, plus presence in at least 10% of the samples. DESeq2 was used for normalization and comparison of groups. CircRNAs with a mean expression of >10 normalized reads were considered highly abundant and this subset was the focus of analysis. Results: In total, 8,651 unique circRNAs passed the filtering criteria, and the top 826 were highly abundant. Global circRNA expression showed upregulation in CCUS relative to healthy controls (p<0.001) and MDS relative to CCUS (p<0.001) (Figure A). We found that 110 unique circRNAs were significantly upregulated in MDS relative to healthy controls, whilst none were significantly downregulated (Figure B). In addition, high or very high IPSS-R score was associated with significantly higher circRNA expression than lower scores (p<0.001). We used LASSO regression to identify 14 circRNAs which may be related to clinical outcomes, and used these to calculate a Myeloid Circ Score (MCS). Patients were designated ‘high MCS’ or ‘low MCS’ with a cut-off at the median. A high MCS was associated with poorer progression-free survival (PFS) with a hazard ratio (HR) of 33.5 (95%CI 7.8-143.5, p<0.001) and poorer overall survival (OS) with a HR of 16.5 (95%CI 3.8-71.5, p<0.001) in the whole cohort. The same could be seen in analysis of MDS patients only (PFS HR 14.2 (95%CI 3.2-62.7, p<0.001), OS HR 8.5 (95%CI 1.9-38.3, p=0.005)). This a greater hazard ratio than that of the IPSS-R in this cohort (Very high/high/Intermediate IPSS-R relative to low/very low IPSS-R; HR for PFS of 4.27 (95%CI 1.7-11.1, p=0.003), and HR for OS of 2.5 (95%CI 0.9-7.3, p=0.09)). Image:Summary/Conclusion: We have observed associations between overall circRNA abundance and disease severity in myeloid cancer, including a global upregulation along the spectrum of disease from healthy to CCUS to MDS, and upregulation in high risk MDS compared to lower risk MDS. CircRNAs may be directly implicated in disease biology or be reflective of other changes at the cellular level. The Myeloid Circ Score has potential in risk stratification but requires validation in independent cohorts and further investigation in more readily available tissues such as bulk bone marrow or peripheral blood.