Abstract Background: Hypoxia contributes to tumor therapy resistance and it can be evaluated through imaging, such as [18F]fluoromisonidazole (FMISO)-PET, or hypoxia-associated gene expression profiling, among other modalities. Integrating imaging with tumor microenvironment features may provide a more comprehensive prognostic assessment. This study evaluates the prognostic value of combining FMISO-PET hypoxia imaging with baseline gene expression profiles in head and neck squamous cell carcinoma (HNSCC) patients undergoing radiochemotherapy (RCTx). Methods FMISO-PET images were obtained at baseline and at weeks 1, 2, and 5 during RCTx to assess tumor-to-background ratios (TBR) and hypoxic volumes (HV). Gene expression from pre-treatment biopsies was analyzed using Affymetrix microarrays. Gene set enrichment analysis of cancer hallmarks was conducted, and cell type fractions were imputed using CIBERSORTx. Correlation and LASSO analyses were conducted to identify baseline gene expression and cell type fractions associated with hypoxia imaging at baseline and during RCTx. Cox regression models were used to integrate imaging and gene expression data for predicting loco-regional tumor control (LRC). Results Baseline Epithelial to Mesenchymal Transition and Hypoxia gene sets were strongly correlated with hypoxia imaging parameters at week 2. A LASSO model identified baseline gene expression associated with FMISO-PET parameters (TBRpeak, HV16, and residual hypoxia) at week 2, indicating a predisposition for hypoxia persistence. AKR1A1, PITPNB, CTSK, and PEF1 gene expression levels were associated with LRC. In the tumor microenvironment, CD4 memory resting T cells and fibroblasts at baseline correlated with hypoxia imaging at weeks 1 and 2, with high levels of these cell fractions linked to decreased LRC. The fibroblast cell fraction positively correlated with CTSK transcript levels (Spearman ρ = 0.59, p < 0.05). K-means clustering based on CTSK and fibroblast levels identified patient subgroups with distinct LRC outcomes (HR [95% CI]: 3.491 [1.165-10.46], p = 0.0256). Importantly, clustering remained an independent predictor of LRC when analyzed with baseline FMISO TBRpeak and HV16 in multivariate Cox regression (HR [95% CI]: 3.883 [1.224-12.321], p = 0.0213). However, when TBRpeak and HV16 at week 2 were included in the model, the impact of the fibroblast cell fraction/CTSK clustering was attenuated (HR [95% CI]: 2.899 [0.918-9.156], p = 0.0697). Conclusion This study highlights the potential for integrating baseline tumor microenvironment markers, such as fibroblast fraction and CTSK expression, with dynamic hypoxia imaging to improve HNSCC prognostic accuracy. High fibroblast cell fraction and CTSK levels at baseline may predispose tumors to treatment-resistant hypoxia, confirmed by persistent hypoxia measured by FMISO-PET at week 2 . By capturing both predisposition to and hypoxia imaging, this approach may enhance the personalization of radiochemotherapy for high-risk HNSCC patients. Citation Format: María J. Besso, Verena Bitto, Cristina Conde-Lopez, Mareike Roscher, Steffen Löck, Annett Linge, Mechthild Krause, Michael Baumann, Ina Kurth. Combining baseline microenvironment biomarkers with hypoxia imaging to predict radiochemotherapy outcomes in head and neck cancer. [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Translating Targeted Therapies in Combination with Radiotherapy; 2025 Jan 26-29; San Diego, CA. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(2_Suppl):Abstract nr B017.
Read full abstract