Using in vitro models to predict imatinib responses in PDGFRA-mutant gastrointestinal stromal tumor

Gastrointestinal stromal tumor (GIST) is the most common sarcoma, and 85% of cases harbor mutations in KIT or PDGFRA receptor tyrosine kinases. Imatinib (IM), a type II TKI, can treat many KIT-mutant GIST, but not GIST with the most common PDGFRA mutation, exon 18 D842V. D842V is resistant to all type II TKIs. Avapritinib (AVA), a type I TKI, was developed and is FDA-approved for all PDGFRA exon 18 mutant GIST but is costly and has severe cognitive side effects in some patients. AVA is also not widely available outside the United States, leaving many exon 18 mutant GIST patients without treatment options. However, limited clinical evidence reports the usage of IM to treat non-D842V exon 18 mutations. As IM is a more tolerable, cost-effective, and accessible drug than AVA, utilizing IM therapy in certain cases would provide treatment for those who cannot access or tolerate AVA. As it is not feasible to model every observed mutation, we utilized in vitro models to predict exon 18 mutations that could be treated with IM. Through collaboration and AACR GENIE, we curated a cohort of 1000+ PDGFRA-mutant GIST. 66% had D842V while the remaining had non-D842V point mutations and complex in/dels. Strikingly, 78% of mutations involved the D842 residue, which plays a key role in the autoinhibition of PDGFRA, and mutations like D842V disrupt this. As nearly every single amino acid substitution at the 842-residue was observed in our cohort, we hypothesized that the characteristics of the 842-position amino acid determine IM sensitivity and will predict treatment responses for any exon 18 mutation. To test our hypothesis, we used Ba/F3 and CHO cells to express every possible D842X and D842_D846delinsX mutation. This 4-residue deletion was the most common in/del in the cohort; therefore, we chose to profile this mutation backbone as well. IM sensitivity was determined by calculating an IC 50 value using immunoblotting for phosphorylated and total PDGFRA. We observed similar trends in IM sensitivity depending on the class of amino acid at the 842-position, with little difference between D842X and D842_D846delinsX mutant kinases. Seven out of eight hydrophobic residues conferred IM resistance while amino acids from other classes (polar, +/- charged, special case) conferred IM sensitivity/intermediate sensitivity. Notably, alanine conferred IM sensitivity, different than the other hydrophobic substitutions and in silico modeling revealed how the side chain structure at the 842-position affects IM binding/activity. Lastly, we determined that our results are concordant with first-line IM response data, as patients with predicted IM-sensitive mutations experienced a longer median progression-free survival than those with predicted or known IM-resistant mutations (30 vs 4 months, p <0.0001). Our work highlights an approach to optimize clinical guidelines for the TKI treatment for PDGFRA-mutant GIST based on specific patient mutations.