Distinction of FLT3-ITD transcriptomic signature from FLT3-TKD and the frequency of this signature in acute myeloid leukemia without <i>FLT3</i> mutation.

Abstract

6538 Background: The FLT3 gene plays a major role in acute myeloid leukemia and frequently overexpressed irrespective of its mutation status. FLT3 internal tandem duplication (ITD) and tyrosine kinase domain (TKD) mutation are both associated with adverse prognosis, but the ITD is associated with more aggressive disease and higher relapses than the TKD. Therapeutically, FLT3-TKD has different resistance implications from FLT3-ITD. We used transcriptomic data to compare AML FLT3-ITD-positive (ITD+) with FLT3-TKD-positive (TKD+), then tested if AML FLT3-negative (FLT3-) may have RNA signature similar to those seen in ITD+ or TKD+ AML. Methods: Between 2023 and 2025, 906 AML samples were RNA profiled using next generation sequencing. A targeted NGS panel of approximately 1,600 genes was used. More than 100 million reads were required to accept RNA results. The required percentage of spliced RNA reads was above 20%. Of these samples, 171 had FLT3-ITD and 115 had FLT3-TKD mutations and 619 were FLT3-. Results: We first explored if expression profiling can define a subgroup of TKD+ cases that are similar to those with ITD+ AML. Using 22 genes in random forest modeling, expression profiling segregated ITD+ AML from TKD+ AML with AUC of 0.734 (95% CI: 0.638-0.831, Precision: 0.940). The top genes relevant for this segregation were GRB10, PTK7, WT1, TEC, CASP3, NR6A1, and CD47. We then build a random-forest-based model to distinguish ITD+ from FLT3- cases. We used the 171 ITD+ AML and a set of 220 FLT3- AML cases to distinguish between the two groups. Using 45 genes in this model we were able to segregate with the ITD+ cases based on expression with AUC of 0.889 (95% CI: 0.833-0.945, precision : 0.935). Top genes were LUC7L2, NFYC, PRPF40B, PRPF8, SF3A1, SMAD5, THRA, ACVR1, EVI2A, MCM3AP. This highly predictive model was then used to test independent set of 399 FLT3- AML cases. Of these cases 319 (80%) had a signature similar to ITD+ AML. Similarly, we used the 115 TKD+ with the 220 FLT3- AML cases and built a random forest-based model to distinguish between the two groups. This model required 70 genes and was less robust but segregated the TKD+ AML with AUC of 0.784 (95% CI: 0.701-0.868, precision: 0.842). Testing an independent set of 399 FLT3- cases using this model showed TKD signature in 236 (59%) of FLT- cases. All these positive TKD signature cases were also positive for the ITD signature. Conclusions: This data suggests that significant number of AML cases are biologically driven by changes similar to those seen in FLT3-mutant cases. While clinical trial is needed for confirmation, these cases with FLT3-ITD signature may respond efficiently if treated with therapy including FLT3 inhibitors in fashion similar to that seen ITD+ cases. Furthermore, this data suggests that ITD signature might be a better biomarker than TKD mutation for selecting patients to be treated with FLT3 inhibitors.