Molecular and clinical correlates of extrachromosomal DNA in oral squamous cell carcinoma

Introduction: Extrachromosomal DNA (ecDNA) consists of circularized chromosomal fragments generated through structural instability. In solid tumors, ecDNA amplifies oncogenes, reshapes cis-regulatory architecture, and alters therapeutic response. Its role in HPV-negative oral cavity squamous cell carcinoma (OSCC) remains poorly defined, with only limited tumor-specific analyses to date. Clarifying its contribution to transcriptional dysregulation and tumor phenotypes may identify drivers of carcinogenesis of OSCC. Methods: We evaluated OSCC cases from Canada's Marathon of Hope Cancer Centres Network (MOHCCN). Whole-genome (~80X depth) and bulk whole-transcriptome sequencing (~80M reads) were performed on 131 primary tumors. ecDNA architecture was resolved using AmpliconArchitect and AmpliconClassifier. Differential gene expression was performed by DESeq2, and immune cell type deconvolution was performed by Cibersort. Matched patient-derived xenografts (PDXs) were generated to assess downstream functional effects. Results: All OSCC cases were surgically treated with 16% stage I-II and 84% stage III-IVB. ecDNA was detected in 33 of 131 (25.2%) tumors. ecDNA positivity showed a significantly different subsite distribution compared with ecDNA-negative cases, with ecDNA positivity enriched in floor-of-mouth and tongue tumors (Fisher's exact p = 1.7×10⁻⁴). Smoking status, drinking history, and stage were not significantly associated with ecDNA burden. Recurrent ecDNA structures harbored key oncogenes and co-amplified partners, including CCND1, FGF19, FGF3, FGF4, MYEOV, PPFIA1, CTTN, and ANO1, with the highest enrichment on chromosomes 11 (n=14) and 7 (n=9). Transcriptomic analysis demonstrated marked up-regulation of ecDNA-enriched genes, with expression levels elevated by approximately six-fold relative to non-ecDNA counterparts (Wilcoxon test, p < 1x10⁻¹⁶). Correlation of ecDNA copy number with RNA expression supported direct dosage effects and regulatory augmentation; promoter-like element density showed a positive association with expression (R=0.36, p=0.046). Immune deconvolution demonstrated a significant reduction in resting natural killer cells in ecDNA-positive tumors (FDR < 0.05), while M1 macrophages (FDR = 0.065) and CD8 T cells (FDR = 0.096) showed similar downward trends. Additionally, matched PDX models were generated and profiled for transcriptome and proteome on 88 of the 131 patients; 26/88 (29.5%) tumors were ecDNA positive. Conclusion: This work delineates the landscape and functional impact of ecDNA in OSCC, linked to transcriptional amplification, regulatory element activity, and shifts in immune contexture. These findings establish ecDNA as a correlate of tumor biology and a potential therapeutic target.