Genes of the BAF/PBAF chromatin-remodeling complexes are among the most frequently mutated in human cancer, but whether those mutations shape response to immune-checkpoint inhibitors was unresolved. Mining genomic and clinical data from 43,728 tumors on cBioPortal, the team found that patients with BAF/PBAF-mutated tumors treated with checkpoint inhibitors lived nearly twice as long — a median 28 versus 15 months. The signal held after adjusting for tumor mutational burden and was reversed in untreated patients, marking it as genuinely predictive rather than merely prognostic.
Published as a Letter to the Editor in Biomarker Research, this analysis was led by Prof. Antoine Italiano — of INSERM U1218, the University of Bordeaux, and the Early Phase Trials and Sarcoma Units at Institut Bergonié — with Kevin Courtet as first author. Explicyte contributed to the genomic data analysis
The team screened 43,728 tumors across cancer types in cBioPortal for nonsynonymous somatic mutations in six BAF/PBAF subunit genes. Mutation frequencies were 6.6% for ARID1A, 4.1% for SMARCA4, 3.4% for both ARID1B and ARID2, 3.2% for PBRM1, and 1.2% for SMARCB1. Uterine carcinoma, melanoma, and bladder cancer carried the highest proportion of BAF/PBAF mutations.
In a previously described cohort of 1,661 patients treated with an immune-checkpoint inhibitor, BAF/PBAF-mutated tumors were associated with markedly longer overall survival — a median of 28 months (95% CI 21.6–34.3) versus 15 months (95% CI 12.9–17.0) for wild-type, p < 0.0001. The magnitude of the difference nominated BAF/PBAF status as a candidate predictive biomarker.
To rule out an intrinsic prognostic effect, the team analyzed 27,870 metastatic-cancer patients who did not receive checkpoint inhibitors. Here BAF/PBAF-mutated tumors did worse, not better — a median 61.8 versus 109.2 months for wild-type (p < 0.001). The reversal indicates the survival advantage under immunotherapy reflects a real predictive value, not baseline biology.
Stratifying by TMB, the benefit concentrated in low-TMB tumors (< 10 mutations/megabase) — 21 versus 14 months, p = 0.024 — which made up 70.6% of the cohort (n = 1,173). In a multivariable Cox model, BAF/PBAF status remained an independent predictor of overall survival (wild-type HR 1.2, 95% CI 1.05–1.4, p = 0.017), holding across most carcinoma types except NSCLC, unknown-primary, and renal cancer.
The work reframes BAF/PBAF mutations as a candidate predictive biomarker for checkpoint-inhibitor benefit — one that operates independently of tumor mutational burden and is most informative in the low-TMB tumors where existing markers underperform.
For drug developers and trial designers, BAF/PBAF mutation status offers a genomically defined way to enrich checkpoint-inhibitor trials — with particular value in low-TMB tumors that current burden-based selection leaves behind. Because the mutations reverse direction without immunotherapy, they behave as a predictive rather than purely prognostic signal, strengthening the case for prospective validation. The authors also flag a combination rationale: pairing checkpoint inhibitors with small-molecule inhibitors of chromatin-remodeling pathways such as EZH2.
