Spatial transcriptomics and histopathological analyses reveal potential biomarkers and immune features linked to recurrence in head and neck cancer.

Head and neck cancers (HNCs) are aggressive malignancies, with recurrence contributing substantially to poor patient outcomes. We used spatial transcriptomics to identify tumour and immune features associated with recurrence. Pathologist-annotated tumour and stromal regions from recurrent and non-recurrent HNC samples were analysed using spatially resolved transcriptomics. Exploratory differential expression analysis was used to nominate tumour-associated candidate markers for recurrence. Selected candidates were assessed by immunohistochemistry and further evaluated using independent spatial transcriptomic and bulk RNA-sequencing cohorts. Immune-cell abundance and spatial co-localisation were also examined. PAEP, MARCO, PNPLA3, and FMO2 emerged as leading candidate markers associated with recurrence. Increased expression of PAEP, MARCO, and PNPLA3 was supported by immunohistochemistry. Integrative analysis across independent spatial and bulk RNA cohorts demonstrated consistent PAEP overexpression in recurrent HNC, particularly within tumour regions, with this pattern reproduced across datasets and profiling platforms. Non-recurrent tumours showed a trend towards greater T- and B-cell infiltration, with these populations frequently co-localising within the tumour microenvironment. In recurrent tumour regions, PAEP expression showed an inverse trend with B-cell and plasma-cell abundance, which was higher in non-recurrent samples. Spatial transcriptomics identified tumour-intrinsic and immune microenvironmental features associated with HNC recurrence. PAEP emerged as the most reproducible candidate biomarker across internal and external cohorts and may be associated with reduced B-cell and plasma-cell infiltration. These findings support further validation of PAEP and related immune signatures for recurrence-risk stratification and personalised treatment.