Session: 618. Acute Myeloid Leukemias: Biomarkers and Molecular Markers in Diagnosis and Prognosis: Poster III
Hematology Disease Topics & Pathways:
Research, Acute Myeloid Malignancies, AML, Adult, Translational Research, Genomics, Diseases, Myeloid Malignancies, Biological Processes, Molecular biology, Study Population, Human
Methods: A total of 81 non-acute promyelocytic leukemia AML patients were enrolled in the study. Leukemic blasts (CD34+/CD117+) were sorted from peripheral blood (PB) at diagnosis using autoMACS (Miltenyi Biotec). High-quality, DNAse-treated RNA (RIN>8) was ribodepleted using the RiboCop rRNA Depletion kit (Lexogen) and used for library preparation with the NEBNext Ultra II Directional RNA Library Prep Kit (NEB). Transcriptome of 35 patients was sequenced on NovaSeq (Illumina). For statistical analysis, patients were divided into responding (n=16) and refractory (n=19) groups based on a complete remission (CR) achievement after the first cycle of a standard 3+7 chemotherapy. Gene expression quantification was performed using StringTie2 software (version 1.3.6) and the resulting data were analyzed and visualized in R software 4.0.0. The most significant results were validated in a cohort of 46 PB samples from AML patients (responders=18; refractory=28) using TaqMan Gene Expression assays and a StepOne Plus instrument (Applied Biosystems). To further analyze our dataset, a logistic regression model was constructed based on the expression values of IFIT5, IFI44L, and IFI44 genes. Model coefficients were used to calculate log odds (IRDS-score, combining the expression of all three genes) and probability of being refractory. Log odds were dichotomized according to the mean value and further used to evaluate the model´s accuracy. Nomogram prediction models including standard prognostic variables such as age, white blood cell count (WBC), and ELN risk, were also created.
Results: Among genes differentially expressed between responding and refractory AML subgroups, we identified a noticeable cluster of interferon-related DNA damage resistance signature (IRDS) genes, including IFIT5, IFI44L, IFIT1, IFI44, IFIT3, IFI6, IFI16, IFITM1, IFIT2, and IFIH1. These IRDS genes were significantly (p<0.05) overexpressed in refractory patients. Given that IRDS genes are linked to chemoresistance and poor prognosis in multiple cancers, we focused on them in a subsequent, more detailed analysis. When applying FDR correction (<0.05), four genes stood out – IFIT5 (p<0.0001), IFI44L (p<0.0001), IFI44 (p=0.0002), and IFIT1 (p=0.0001). Due to its very low expression level, IFIT1 was excluded from further analyses. The overexpression of IFIT5 (p=0.0426), IFI44L (p=0.0284), and IFI44 (p=0.0042) was validated in an independent cohort of PB AML samples. Next, we calculated the IRDS-score. The concordance index of a CR achievement prediction model containing age, WBC, and ELN risk was 0.782. The accuracy of our model, derived only from the dichotomized IRDS-score, was 0.713. Finally, when we combined age, WBC, ELN, and IRDS-score into one nomogram model, the concordance index increased to 0.913.
Conclusions: Among the upregulated genes in refractory AML, we discovered a distinct cluster of IRDS genes. These genes are known to drive cell resistance to DNA damage, leading to resistance against various DNA-damaging drugs, including anthracyclines. We present the first comprehensive transcriptome profile of IRDS genes in AML blasts and introduce a model for predicting AML chemoresistance using the IRDS-score. This model demonstrates significantly higher accuracy compared to existing models. The IRDS-score relies on a manageable number of genes, making it accessible for measurement in most laboratories. While further validation with larger sample sets is necessary, these initial results are highly promising.
Funding: This project was supported by MH CZ-DRO (UHKT 00023736) and by AZV CR (NU20-03-00412).
Disclosures: No relevant conflicts of interest to declare.