Genomic characterization of FGFR-altered solid tumors using a clinically annotated pan-cancer repository.

3074 Background: Somatic mutations, gene fusions, and amplifications in FGFR1/2/3 are found across solid tumors. FGFR2/3 fusions and select mutations have matched FDA-approved therapies in intrahepatic cholangiocarcinoma (iCCA) and urothelial carcinoma (UC). Identification of FGFR-altered tumors may inform patient selection for future drug development. Methods: All solid tumors that underwent MSK-IMPACT, a DNA-based NGS panel with coverage of FGFR1-4, were queried for FGFR1-4 alterations. FGFR structural variants (SV) were manually reviewed and compared to RNA sequencing data (RNAseq, Archer FusionPlex) when available. Alteration (alt) frequencies of iCCA and UC tumors with target FGFR alt (level 1 OncoKB evidence), non-target FGFR alt (lacking level 1 evidence) and no FGFR alt were compared using chi-squared tests and odds ratios. Euclidean distance was calculated between cohorts based on prevalence of frequently altered genes (>5%). Consensus clustering was used to identify optimal number of clusters and compare genomic profiles between clusters. Results: At data lock of June 1, 2022, 5695 tumors among 48,416 patients harbored an FGFR alteration (11.8%). FGFR SV were identified in 347 tumors on MSK-IMPACT. Of these 111/140 (79.2%) with RNAseq were positive. 98% (203/207) of SV without RNAseq were considered likely activating (canonical SV, n=125; in-frame inclusive of kinase domain, n=78). The most frequent fusion was FGFR3-TACC3 (n=124), followed by fusions unique to a single patient (n=100). Fusions were discovered in >15 tumor types, most frequently in iCCA (n=93), UC (n=50) and high-grade glioma (n=44). Among 1176 iCCA tumors, 9.9% (n = 117) had a target FGFR2 fusion. Target FGFR tumors were enriched for concurrent BAP1 alt (18%) compared to tumors without FGFR alt (58/969, 5.9%) and with non-target FGFR alt (7/90, 7.8%). Both groups without target alt were enriched in TP53, ARID1A and IDH1 alts compared to target group. Four optimal genomic clusters were identified, with FGFR2 fusions enriching in cluster 4 (29% vs 3%, 7% and ≤1%); 21 non-target FGFR alt were also in cluster 4. Among 2720 UC tumors, 24.7% (n=672) had target FGFR alt (select FGFR3 mutations and FGFR2/3 fusions). Three genomic clusters were identified, with target FGFR enriched in cluster 3 (42% target FGFR vs cluster 1: 11%, cluster 2: 18%) and non-target FGFR alterations enriched in cluster 1 (13% vs. cluster 2: 9.4 %, cluster 3: 7.0%); 68 non-target FGFR alt were in cluster 3. Tumors with non-target FGFR alterations and no FGFR alterations had close genomic association (Euclidean distance: 0.18). Conclusions: Activating FGFR fusions, which occur across tumor types with marked diversity in fusion partners, present opportunities for future drug development. Non-target FGFR alterations in iCCA and UC, particularly those that cluster with target FGFR alterations, may be candidates for future FGFR-targeted strategies.