Introduction: Head and neck squamous cell carcinoma (HNSCC) tumors demonstrate significant inter- and intra-tumoral genetic and epigenetic heterogeneity, and this heterogeneity has been linked with poor clinical outcomes. Traditional two-dimensional (2D) cell lines grown in monolayer culture fail to capture this heterogeneity and thus incompletely model the biology of the tumors from which they are derived.

Methods: Three-dimensional (3D) patient derived organoids (PDO) were derived from 17 HNSCC, six oral dysplasia, and 11 normal oral mucosa samples according to our published protocols. HNSCC samples included five HPV-positive and 12 HPV-negative samples. PDOs were characterized by whole exome sequencing, RNA-sequencing, and methylome profiling. Sequencing data were compared with genomic, expression, and methylation data from HNSCC patients in The Cancer Genome Atlas to establish the degree to which canonical features of HNSCC tumors, including genetic and expression heterogeneity, were captured by our bank of organoids.

Results: Consistent with genomic features of their tissues of origin, normal mucosal organoids demonstrated no copy number variations (CNV), while HNSCC organoids demonstrated many CNVs; dysplasia organoids were present on a spectrum between normal and HNSCC organoids. HNSCC organoids were enriched for deletions of 3p, 8p, and 9p, consistent with well-described CNVs in HNSCC. Known cancer-associated mutations, including TP53, MAP3K13, TP63, CCR4, MLH1, and ETV5 were captured at high frequencies in HNSCC organoid samples and were not detected in normal mucosal organoids. Moreover, mutant allele frequencies varied widely within samples, indicating significant genetic heterogeneity within samples. Clustering of organoids by expression profile demonstrated distinct clusters of normal mucosal and HNSCC organoids, with dysplasia organoids spanning the space between these two clusters. Normal and dysplasia organoids demonstrated normal methylation profiles, while HNSCC organoids demonstrated significant and variable aberrations in methylation across samples, as well as a general hypermethylated phenotype, consistent with TCGA analysis of HNSCC. Integrated multiomic analysis revealed alterations in several cancer-associated pathways in HNSCC and dysplasia organoids, including regulation of TP53 activity, DNA repair, DNA replication, cell cycle checkpoints, MHC class II antigen presentation, and stem cell proliferation.

Conclusions: Multiomic profiling of our large and deeply characterized bank of PDO derived from normal oral mucosa, oral dysplasia, and HNSCC revealed that HNSCC organoids capture genomic and epigenomic features of HNSCC tumors. Critically, these models also capture inter- and intratumoral genetic heterogeneity, underscoring the fidelity of 3D organoids as a model for in vivo tumor biology and enabling modeling of patient-specific tumor biology and treatment response.