Identification and Validation of Ferroptosis-Related Genes As Novel Prognosis Prediction Panel for Acute Myeloid Leukemia

Acute myeloid leukemia (AML) is a main type of adult acute leukemia, especially in elderly population. The development of AML ascribed to various factors including chromosomal abnormalities, isolated gene mutations, etc. and the 5-year survival rate is around 30%. Currently, the most popular and precise assessment of AML is based on cytogenetic differences, and the patients with higher risk molecular mutations had inferior prognosis. However, with the development and application of new drugs, including bcl-2 inhibitors, many AML with high risks (such as FLT3, P53 mutated) also achieve better remission and longer survival. Up to date, many novel prognosis models are in process, with the better understanding of the initiation and development of leukemia.

Ferroptosis, a widespread and ancient form of cell death driven by iron-dependent phospholipid peroxidation. Its involvement in AML has been reported, as leukemic cells frequently exhibit enhanced iron uptake and reduced iron efflux, resulting in elevated intracellular iron levels. Targeting iron homeostasis and inducing ferroptosis may provide novel insights into AML therapy. Herein, we integrated multi-gene information to identify a novel prediction model for ferroptosis-related genes (FRGs), and validated the accuracy of the model using clinical samples from our center (Figure A).

In this study, we downloaded RNA-seq data and clinicopathological features of 151 AML patients from The Cancer Genome Atlas (TCGA) database (https://www.cancer.gov/ccg/research/genome-sequencing/tcga) and 70 corresponding health samples from the Genotype-Tissue Expression (GTEx) database (https://gtexportal.org/). Retrieve the ferroptosis-related genes (FRGs) set from the FerrDb database (http://www.zhounan.org/ferrdb/current/). The difference analysis, LASSO regression and Cox regression were used to determine the FRGs signature to construct prognostic models (Figure B). We finally identified 10 FRGs signature (ACSF2, SOCS1, CDO1, MYB, LPIN1, DNAJB6, PSAT1, GPX4, MT1G, FH) to establish a risk scoring model, and AML patients were divided into high-risk and low-risk group according to the median risk score. The risk score for each sample was determined using the following formula: . Kaplan-meier analysis showed that the overall survival of patients in the high-risk group was significantly worse( P<0.001) (Figure C). Cox regression analysis, ROC and DCA results found that these features have good predictive ability for prognosis of AML (Figure D-F). We also grouped the TCGA AML patients into poor/adverse and favorable/intermediate group based on NCCN guideline (risk stratification based on genomic mutation background), and found that the survival differences based on classical risk model were not that obvious ( P=0.062) (Figure G) when compared to our model ( P<0.001 in Figure C). Which suggested that our model may have its superior in distinguishing real high risk AML patients.

In order to validate the accuracy of our model, we also conducted a retrospective study on 30 AML patients and 13 healthy controls (bone marrow stem cell donor) from September 2019 to September 2022. Bone marrow mononuclear cells (BMNC) from our biobank were obtained, and the ten FRGs panel was applied. Results indicated that there was a significantly difference between normal control and AMLs in FRGs expression (Figure H), and the survival data of high-risk AML patients with higher FRGs level was significantly inferior than that in low-risk group as expected ( P<0.05) (Figure I). To search for possible therapeutic drugs for high-risk patients, we downloaded AMLs data from the GDSC database (https://www.cancerrxgene.org/) and calculated the IC50 of commonly used chemotherapeutic drugs. We found that bcl-2 inhibitor (Obatoclax Mesylate) and TKI (Gefitinib) may have less sensitivity in our high-risk group ( P<0.05), while some unconventional drugs (paclitaxe, dactinomycin, camptothecin, irinotecan and vinorelbine) may benefit in patients with higher FRGs expression (Figure J).

In summary, we identified and validated 10 FRGs signature to well predict the prognosis and drug sensitivity of AML. Based on our findings, appropriate combination of new drugs at the time of treatment may achieve unexpected therapeutic outcomes, but still need to go through further verification.