Basal Cell Carcinoma

Our understanding of how human skin cells differ according to anatomical site and tumour formation is limited. To address this, we have created a multiscale spatial atlas of healthy skin and basal cell carcinoma (BCC), incorporating in vivo optical coherence tomography, single-cell RNA sequencing, spatial global transcriptional profiling, and in situ sequencing. Computational spatial deconvolution and projection revealed the localisation of distinct cell populations to specific tissue contexts. Although cell populations were conserved between healthy anatomical sites and in BCC, mesenchymal cell populations including fibroblasts and pericytes retained signatures of developmental origin. Spatial profiling and in silico lineage tracing support a hair follicle origin for BCC and demonstrate that cancer-associated fibroblasts are an expansion of a POSTN+ subpopulation associated with hair follicles in healthy skin. RGS5+ pericytes are also expanded in BCC suggesting a role in vascular remodelling. We propose that the identity of mesenchymal cell populations is regulated by signals emanating from adjacent structures and that these signals are repurposed to promote the expansion of skin cancer stroma. The resource we have created is publicly available in an interactive format for the research community.

Cutaneous squamous cell carcinoma (cSCC) is a serious public health problem due to its high incidence and metastatic potential. It may progress from actinic keratosis (AK), a precancerous lesion, or the in situ carcinoma, Bowen's disease (BD). During this progression, malignant keratinocytes activate dermal fibroblasts into tumor promoting cancer-associated fibroblasts (CAFs), whose origin and emergence remain largely unknown. Here, we generate and analyze >115,000 single-cell transcriptomes from healthy skin, BD and cSCC of male donors. Our results reveal immunoregulatory and matrix-remodeling CAF subtypes that may derive from pro-inflammatory and mesenchymal fibroblasts, respectively. These CAF subtypes are largely absent in AK and interact with different cell types to establish a pro-tumorigenic microenvironment. These findings are cSCC-specific and could not be recapitulated in basal cell carcinomas. Our study provides important insights into the potential origin and functionalities of dermal CAFs that will be highly beneficial for the specific targeting of the cSCC microenvironment.

Hedgehog-GLI (HH) signaling plays an essential role in embryogenesis and tissue homeostasis. Aberrant activation of the pathway through mutations or other mechanisms is involved in the development and progression of numerous types of cancer, including basal cell carcinoma, medulloblastoma, melanoma, breast, prostate, hepatocellular and pancreatic carcinomas. Activation of HH signaling sustains proliferation, suppresses cell death signals, enhances invasion and metastasis, deregulates cellular metabolism and promotes angiogenesis and tumor inflammation. Targeted inhibition of the HH pathway has therefore emerged as an attractive therapeutic strategy for the treatment of a wide range of cancers. Currently, the Smoothened (SMO) receptor and the downstream GLI transcriptional factors have been investigated for the development of targeted drugs. Recent studies have revealed that the HH signaling is also involved in tumor immune evasion and poor responses to cancer immunotherapy. Here we focus on the effects of HH signaling on the major cellular components of the adaptive and innate immune systems, and we present recent discoveries elucidating how the immunosuppressive function of the HH pathway is engaged by cancer cells to prevent immune surveillance. In addition, we discuss the future prospect of therapeutic options combining the HH pathway and immune checkpoint inhibitors.

Distinguishing infiltrative basal cell carcinoma (BCC) from poorly differentiated cutaneous squamous cell carcinoma (cSCC) remains a significant histopathological challenge. Automated deep learning approaches hold promise for improving diagnostic reliability, yet robust external validation is essential. In this study, we developed a weakly supervised deep learning model to classify these diagnostically challenging subtypes and evaluated its generalizability across internal and external cohorts, as well as in comparison to a dermatopathology foundation model (HistoGPT). The model employed a multiple-instance learning framework (CLAM) using the histopathology-specific transformer Phikon for feature extraction from whole-slide images. Slide-level ground-truth diagnoses from the collected images (n = 335, University Hospital Erlangen) were derived from routine clinical practice and re-evaluated by two board-certified dermatopathologists. Performance was assessed on an internal test set of 84 whole-slide images (27 cSCC and 57 BCC) and two external datasets: Queensland cohort (n = 10, curated in-distribution cases) and the COBRA cohort (n = 200, broad, partly out-of-distribution cases). Model discrimination was quantified using ROC curves, while accuracy, sensitivity, and specificity were reported alongside 95% Wilson confidence intervals (CIs). On the internal test set, the model achieved perfect classification [area under the receiver operating characteristic (AUC) = 1.0; 100% accuracy, sensitivity, and specificity]. Similarly, strong performance was observed in the Queensland cohort (AUC = 1.0), although limited by sample size. In the more heterogeneous COBRA cohort, discrimination remained high (AUC = 0.923, 95% CI 0.885-0.961), requiring threshold adjustment to correct for marked calibration shift (balanced accuracy 86.5% at Youden's J). Attention heatmaps highlighted histologically meaningful regions. In zero-shot evaluation on the internal test set, HistoGPT achieved an overall accuracy of 77%, with high class-wise sensitivity for BCC (98%, 95% CI 91-100) but markedly reduced sensitivity for cSCC (33%, 95% CI 19-52). Fine-tuning a task-specific classifier on the HistoGPT backbone substantially improved performance, achieving near-perfect discrimination and 98% balanced accuracy. These findings demonstrate that weakly supervised deep learning enables highly accurate classification of diagnostically challenging BCC and cutaneous squamous cell carcinoma subtypes. However, reliable deployment across institutions necessitates careful calibration and domain adaptation, and even powerful foundation models such as HistoGPT benefit from targeted fine-tuning to ensure robust performance in dermatopathology.