Abstract

Even though BRAF fusions are increasingly detected in standard multigene next-generation sequencing panels, few reports have explored their structure and impact on clinical course. We collected data from patients with BRAF fusion-positive cancers identified through a genotyping protocol of 97,024 samples. Fusions were characterized and reviewed for oncogenic potential (in-frame status, non-BRAF partner gene, intact BRAF kinase domain). We found 241 BRAF fusion-positive tumors from 212 patients with 82 unique 5' fusion partners spanning 52 histologies. 39 fusion partners were not previously reported, and 61 were identified once. BRAF fusion incidence was enriched in pilocytic astrocytomas, gangliomas, low-grade neuroepithelial tumors, and acinar cell carcinoma of the pancreas. 24 patients spanning multiple histologies were treated with MAPK-directed therapies of which 20 were evaluable for RECIST. Best response was partial response (N=2), stable disease (N=11), and progressive disease (N=7). The median time on therapy was 1 month with MEK plus BRAF inhibitors ([N=11], range 0-18 months) and 8 months for MEK inhibitors ([N=14], range 1-26 months). 9 patients remained on treatment for longer than 6 months [pilocytic astrocytomas (N=6), Erdheim-Chester disease (N=1), extraventricular neurocytoma (N=1), melanoma (N=1)]. Fifteen patients had acquired BRAF fusions. BRAF fusions are found across histologies and represent an emerging actionable target. BRAF fusions have a diverse set of fusion partners. Durable responses to MAPK therapies were seen, particularly in pilocytic astrocytomas. Acquired BRAF fusions were identified after targeted therapy underscoring the importance of post-progression biopsies to optimize treatment at relapse in these patients.