(P 182) USING IN VITRO MODELS TO PREDICT IMATINIB RESPONSES OF PDGFRA EXON 18-MUTANT GIST

Objective: Gastrointestinal stromal tumor (GIST) is the most common sarcoma with 85% of cases driven by mutations in KIT or PDGFRA, which are treated with tyrosine kinase inhibitors (TKIs). First-line (1L) imatinib (IM), a type II TKI, is effective for KIT mutations but not for the most common PDGFRA driver mutation, exon 18 D842V. While D842V is resistant to all type II TKIs, avapritinib (AVA), a type I TKI, is FDA-approved for all exon 18 PDGFRA-mutant GIST. Outside the US, AVA is only available for D842V-mutant GIST or not at all, leaving some patients with no treatment. As some clinical evidence suggests the use of IM in some exon 18 cases, its usage is not well defined, therefore the use of predictive in vitro models define which exon 18 mutations could be treated with IM, a more accessible drug than AVA. We curated a cohort of 1019 patients with PDGFRA-exon 18 mutant GIST. ~67% had D842V, and ~33% had non-D842V point mutations or complex in/dels, and ~78% of all cases had mutations involving the 842 residue. We hypothesized that IM sensitivity could be predicted based on the nature of the amino acid in the 842-position.

Methods: We modeled all 20 amino acids at the 842 position in both point mutation and 4-residue deletion contexts, that were expressed in Ba/F3s or CHOs, and assessed TKI sensitivities with immunoblotting. In silico modeling was performed using PyMol.

Results: When in the 842-position, hydrophobic amino acids except for alanine were IM resistant, while all other amino acids in the polar, +/- charged, and special case classes were IM sensitive. 1L IM response data were concordant with these predictions, where the median progression free survival for predicted IM-resistant mutations (n=80) was 4 months compared to 30 months for those with predicted IM-sensitive mutations (n=35) (HR 3.028, CI 2.011-4.561, p< 0.0001). In silico modeling showed that IM binding depends on the structure of the amino acid side chain at the 842 position.

Conclusion: Our models defined cases where 1L IM can be used to treat PDGFRA exon 18 mutant GIST while also defining those that require 1L AVA. We suggest that all patients with predicted IM-sensitive mutations should receive 1L IM, regardless of regulatory approvals in a country, as IM is more widely available and more cost-effective than AVA. Our work highlights a paradigm shifting approach in which clinical guidelines for the treatment of PDGFRA-mutant GIST should be based on specific patient mutations, allowing for the optimization of patient care.