FP15 Transcriptomic and microenvironment signatures identify five molecular states and predictive signatures along the actinic keratosis to cutaneous squamous cell carcinoma progression
Wendy Tran, Emma Taggart, Ankit Patel, Irene Leigh, Daniel Pennington, Catherine Harwood, Jun WangAbstract
Introduction and aims
Actinic keratoses (AKs) are considered precursor lesions of cutaneous squamous cell carcinoma (cSCC). Although most AKs remain stable/regress, a small proportion will progress into cSCC, the mechanisms by which remain poorly understood. This study aims to identify distinct molecular subtypes that exist across the progression from normal skin to AK to cSCC and to characterize the transcriptomic features that underlie lesion stability, regression or progression.
Methods
Lesional samples from patients with AK, Bowen disease (BD) and cSCC were collected using punch biopsies. For each lesion type, normal skin samples were obtained from proximal and distal sites. Normal skin distant from lesions were also collected. Bulk RNA sequencing was performed on normal, perilesional normal (PN), and lesional AK, BD and cSCC samples (n = 152). Single-cell RNA sequencing was carried out on freshly dissociated, CD45+ fluorescence-activated cell sorting-sorted PN and lesional AK, BD and cSCC samples (n = 16).
Results
Non-negative matrix factorization of bulk gene expression data unveiled five distinct clusters across the normal-to-cSCC continuum. Integration of a differentiated-versus-progenitor signature revealed that these clusters mapped along progressive stages of epithelial dedifferentiation, a hallmark of cSCC development. Two clusters reflected a differentiated state, consisted mostly of normal skin samples, and were enriched for intermediate filament organization and fatty acid metabolism pathways. The remaining three clusters exhibited a more progenitor-like state, consisted mainly of lesional samples, and were enriched in keratinization, mRNA processing, and immune response pathways, respectively. Tumour microenvironment (TME) profiling resolved a depleted TME state associated with the normal clusters and hence regression, while immune-enriched (IE), fibrotic (F), and mixed IE/F states associated with the lesional clusters, with the mixed IE/F state positively correlating with the progenitor state, and most likely disease progression.
Conclusions
Our results provide novel immune/stromal mechanisms and biomarkers predictive of AK regression and progression risks.