Introduction: Internal tandem duplications (ITDs) in the FLT3 gene occur in ~25% of adult AML patients and are associated with relapse and poor prognosis. Although hundreds of ITD variants have been detected in patients, ITDs are traditionally classified by insertion site—juxtamembrane domain (JMD-ITD) or tyrosine kinase domain (TKD-ITD) —with several studies suggesting greater resistance against chemotherapy and FLT3 inhibitors (TKIs) in TKD1-ITDs. However, due to the lack of systematic and functional analysis of these highly diverse variants, the clinical relevance of ITD heterogeneity remains largely unknown.

Methods: We collected 1,334 FLT3-ITD variants from patient cohorts and public databases and analyzed their sequence diversity. For functional studies, we generated a Ba/F3 cell line panel of 47 representative variants reflecting clinical diversity. Drug sensitivity to six TKIs and two chemotherapy agents was evaluated using CCK-8–based IC50 assays. Selected variants were validated in a murine bone marrow transplantation model. RNA-seq was performed on engineered cell lines, and AlphaFold2 was used to predict variant-specific structures.

Results: We observed extensive heterogeneity in insertion site, length, and presence of exogenous sequences among the 1,334 ITDs, resulting in 952 unique non-redundant variants. To investigate functional differences, we assessed drug responses in a panel of 47 FLT3-ITD–expressing Ba/F3 cell lines, collectively spanning all known FLT3 subdomains and capturing key clinical features. Drug profiling revealed four distinct therapeutic patterns, each corresponding to a subset of ITD variants: (i) pan-resistance to TKIs and chemotherapy; (ii) pan-sensitivity; (iii) selective resistance to chemotherapy; and (iv) selective resistance to type II TKIs. Representative variants from each group showed consistent in vivo responses in murine transplantation models, supporting the robustness of this pattern-based framework. These response patterns emerged independently of insertion site and reflected variant-specific mechanisms.

To explore the basis of therapeutic diversity, we first examined ITD sequence features across patterns. Strikingly, we found all pan-resistant variants (pattern i) harbored duplications spanning the JM-S subdomain (aa579–592), essential for FLT3 autoinhibition and STAT5 regulation, while no shared sequence features were identified in patterns ii–iv. Transcriptomic profiling revealed STAT5 and DNA repair pathway activation in pattern i, and upregulation of cell cycle programs in pattern iii, consistent with known chemoresistance mechanisms. Given that type II TKIs target the DFG-out conformation, AlphaFold2 modeling showed that pattern iv variants favored a DFG-in state, potentially limiting drug access to the kinase pocket.

As pattern i variants exhibited pan-resistance, we further investigated the biological and mechanistic basis of this subtype. Mice transplanted with JM-S–ITD cells showed accelerated disease onset, higher residual burden, and shorter survival. Transcriptomic profiling revealed significantly elevated STAT5 activity and homology-directed repair (HDR) pathway expression in JM-S–ITDs compared to pattern ii variants lacking JM-S coverage. While previous studies linked FLT3-ITD to DNA repair upregulation such as alt-NHEJ, our data reveal enhanced HDR signaling—a upstream repair mechanism—specifically associated with JM-S–ITDs. Elevated STAT5 phosphorylation and target gene expression, such as RAD51, suggest that STAT5 may drive HDR activation. Combining TKIs with the clinically used alt-NHEJ inhibitor Olaparib showed limited efficacy, whereas MRN inhibition via Mirin blocked HDR and markedly enhanced TKI sensitivity—raising cytotoxicity from ~20% to ~80%. This highlights HDR dependency as a tractable vulnerability, offering a potential precision strategy for treating this resistant subtype.

Conclusion: This study presents a comprehensive functional dissection of FLT3-ITD mutational diversity. We identify four distinct drug response patterns that transcend traditional insertion-site–based classifications and pinpoint JM-S as a determinant subdomain of pan-resistance, linked to STAT5 and HDR activation. Targeting the HDR machinery re-sensitizes this subtype and offers a precision therapeutic strategy. These findings establish the first functional classification of highly diverse FLT3-ITDs and reveal actionable vulnerabilities to guide future treatments.

This content is only available as a PDF.
Sign in via your Institution