Sandrina Martens

Programmed death‐ligand 1 (PD‐L1) expression, commonly quantified as tumour proportion score (TPS), is a key biomarker guiding immunotherapy in non‐small cell lung cancer (NSCLC). Manual scoring of PD‐L1 immunohistochemistry (IHC) is time‐consuming and challenged by preanalytical variability and interpretive subjectivity, which can hinder consistent treatment selection. Artificial intelligence (AI) tools aim to support more standardized PD‐L1 scoring. We compare publicly available evidence on PD‐L1 AI tools and identify barriers to routine implementation. We mapped publicly available evidence for seven commercial and 13 non‐commercial or research‐stage PD‐L1 AI tools across six interconnected levels: data characteristics and validation setting, data preparation and reference standard, supervision strategy, performance evaluation and agreement, generalizability and clinical relevance, and deployment traceability. Commercial sources more often described workflow integration, intended use, reader assistance and certification or product status, whereas non‐commercial studies more often detailed model architecture, supervision strategy, code/data availability and experimental design. Heterogeneity in cohorts, assay–scanner settings, reference standards, agreement metrics and decision‐threshold reporting limited direct comparison. Key gaps included scarce prospective multi‐centre validation, limited reporting at decision‐relevant TPS thresholds, and incomplete separation between peer‐reviewed validation evidence, vendor claims and deployment traceability. The resulting evidence map supports conditional appraisal and local verification rather than a universal ranking. Publicly available evidence for PD‐L1 TPS AI tools remains heterogeneous, and reported performance is not automatically portable between laboratories because it depends on the combined case mix, assay–scanner workflow, reference standard and TPS‐threshold setting. Therefore, current evidence does not support a universal ranking for routine practice. The six‐level comparison structure and end‐user crosswalk support shortlisting intended‐use‐compatible tools and, where relevant, comparing them locally using the same representative cases, assay–scanner workflow, reference standard and prespecified TPS‐threshold endpoints. More transparent reporting, robust external and prospective multi‐centre validation, and ongoing quality monitoring are needed for safe integration into routine human–AI pathology workflows.

Digital pathology has become increasingly popular for research and clinical applications. Using high-quality microscopes to produce Whole Slide Images of tumor tissue enables the discovery of insights into biological aspects invisible to the human eye. These are acquired through downstream analyses using spatial statistics and artificial intelligence. Determination of the quality and consistency of these images is needed to ensure accurate outcomes when identifying clinical and subclinical image features. Additionally, the time-intensive process of generating high-volume images results in a trade-off that needs to be carefully balanced. This study aims to determine optimal instrument settings to generate representative images of pathological tissue using digital microscopy. Using various settings, an H&E stained sample was scanned using the ZEISS Axio Scan.Z1. Next, nucleus segmentation was performed on resulting images using StarDist. Subsequently, detections were compared between scans using a matching algorithm. Finally, nucleus-level information was compared between scans. Results indicated that while general matching percentages were high, similarity between information from replicates was relatively low. Additionally, settings resulting in longer scanning times and increased data volume did not increase similarity between replicates. In conclusion, the scan setting ultimately deemed optimal combined consistent and qualitative performance with low throughput time.

A ‘classical’ and a ‘basal‐like’ subtype of pancreatic cancer have been reported, with differential expression of GATA6 and different dosages of mutant KRAS . We established in situ detection of KRAS point mutations and mRNA panels for the consensus subtypes aiming to project these findings to paraffin‐embedded clinical tumour samples for spatial quantitative analysis. We unveiled that, next to inter‐patient and intra‐patient inter‐ductal heterogeneity, intraductal spatial phenotypes exist with anti‐correlating expression levels of GATA6 and KRAS ^G12D . The basal‐like mRNA panel better captured the basal‐like cell states than widely used protein markers. The panels corroborated the co‐existence of the classical and basal‐like cell states in a single tumour duct with functional diversification, i.e. proliferation and epithelial‐to‐mesenchymal transition respectively. Mutant KRAS ^G12D detection ascertained an epithelial origin of vimentin‐positive cells in the tumour. Uneven spatial distribution of cancer‐associated fibroblasts could recreate similar intra‐organoid diversification. This extensive heterogeneity with functional cooperation of plastic tumour cells poses extra challenges to therapeutic approaches. © 2023 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

Heterotopia of the salivary gland occurs mainly in the head and neck region of the human body, rarely in regions such as the rectum, but has never been demonstrated in the pancreas. Within a screening effort of pancreatic samples for detecting ΔNp63 expression, we discovered two pancreatic samples from a 35-year-old male showing salivary gland heterotopia. Immunohistochemical stainings were done for markers of healthy and neoplastic salivary glands and showed expression of calponin, CD142 and KRT14 but not of S100p, GFAP or CD117. A PAS-staining and Alcian Blue staining showed the presence of acid mucins. These staining patterns were consistent with non-neoplastic submandibular gland tissue comprised of abundant seromucous glands, basal cells and myoepithelial cells, all features typically absent in the pancreas. Also, no pancreatic islets of Langerhans were detected. We show for the first time that salivary gland heterotopia can occur at the location of the pancreas.

Objective: The aggressive basal-like molecular subtype of pancreatic ductal adenocarcinoma (PDAC) harbours a ΔNp63 (p40) gene expression signature reminiscent of a basal cell type. Distinct from other epithelia with basal tumours, ΔNp63⁺ basal cells reportedly do not exist in the normal pancreas. Design: We evaluated ΔNp63 expression in human pancreas, chronic pancreatitis (CP) and PDAC. We further studied in depth the non-cancerous tissue and developed a three-dimensional (3D) imaging protocol (FLIP-IT, Fluorescence Light sheet microscopic Imaging of Paraffin-embedded or Intact Tissue) to study formalin-fixed paraffin-embedded samples at single cell resolution. Pertinent mouse models and HPDE cells were analysed. Results: In normal human pancreas, rare ΔNp63⁺ cells exist in ducts while their prevalence increases in CP and in a subset of PDAC. In non-cancer tissue, ΔNp63⁺ cells are atypical KRT19⁺ duct cells that overall lack SOX9 expression while they do express canonical basal markers and pertain to a niche of cells expressing gastrointestinal stem cell markers. 3D views show that the basal cells anchor on the basal membrane of normal medium to large ducts while in CP they exist in multilayer dome-like structures. In mice, ΔNp63 is not found in adult pancreas nor in selected models of CP or PDAC, but it is induced in organoids from larger Sox9^low ducts. In HPDE, ΔNp63 supports a basal cell phenotype at the expense of a classical duct cell differentiation programme. Conclusion: In larger human pancreatic ducts, basal cells exist. ΔNp63 suppresses duct cell identity. These cells may play an important role in pancreatic disease, including PDAC ontogeny, but are not present in mouse models.

SUMMARY: Objective: An aggressive basal-like molecular subtype of pancreatic ductal adenocarcinoma (PDAC) exists, driven by ΔNp63. In other epithelia, ΔNp63 ⁺ basal cells have stem cell capacity and can be at the origin of tumors. In the pancreas, basal cells have not been identified. Design: We assessed basal cell markers in human and mouse pancreas, chronic pancreatitis and PDAC, and developed a 3D imaging protocol (FLIP-IT) to study sizeable samples at single cell resolution. We generated organoid cultures of ducts from Sox9-eGFP reporter mice. Results: In normal human pancreas, rare ΔNp63 ⁺ cells exist in ducts that expand in chronic pancreatitis. ΔNp63 ⁺ cells express KRT19 and canonical basal markers (KRT5, KRT14 and S100A2) but lack markers of duct cells such as CA19.9 and SOX9. In addition, ΔNp63 ⁺ cells pertain to a niche of cells expressing gastrointestinal stem cell markers. 3D views of the ductal tree in formalin fixed paraffin embedded samples show that basal cells are localized on the basal membrane of medium to large ducts and expand as multilayer dome-like structures in chronic pancreatitis. In mice, ΔNp63 expression is induced when culturing organoids from Sox9-low ductal cells but could not be found in normal pancreas nor in models of pancreatitis or pancreatic cancer. Conclusion: We discovered a novel ductal cell population in normal human pancreas similar to basal cells in other tissues. Using FLIP-IT, we provide unprecedented 3D visualization of these cells in archival clinical specimens. ΔNp63 ⁺ cells may play an important role in pancreatic tissue regeneration and cancer. SUMMARY BOX: What is already known about this subject? ΔNp63 has a central role in determining the basal-like subtype of pancreatic ductal adenocarcinoma (PDAC). Different to other tissues with basal cancers, the normal pancreas reportedly does not contain (ΔNp63-expressing) basal cells. Current protocols face severe limitations for marker-based identification and 3D imaging of individual (rare) cells in archival pancreatic samples. What are the new findings? We report a rare and atypical pancreatic duct cell that expresses ΔNp63, other basal cell markers and g.i. stem cell markers. The number of these basal cells increases in diseases such as chronic pancreatitis and pancreatic cancer. We provide an easy to implement protocol for 3D clearing and high-resolution imaging of sizeable samples of (fresh or FFPE) human pancreas or an entire mouse pancreas. Except after culturing medium to large ducts as organoids, we fail to detect basal cells in mouse experimental pancreatic models. How might it impact on clinical practice in the foreseeable future? Extrapolating from knowledge in other organs, basal cells in the pancreas may have a stem cell/progenitor role, including in diseases such as (basal) pancreatic cancer. Use of the 3D imaging protocol in archival clinical specimens will allow unprecedented insights in pancreatic histopathology. For above mentioned diseases, we caution for findings in experimental mouse models that may not (fully) recapitulate the etiopathogenesis.

Immune checkpoint inhibition (ICI) is a promising cancer therapy, which has progressed rapidly from a preclinical concept to clinical implementation. Commonly considered targets in ICI are CTLA-4, PD-1/PD-L1, and LAG-3, and the list grows. As ICI is generally only beneficial for a subset of patients, there is a need to select patients that are eligible for therapy as well as to monitor therapy response. There is growing interest to do this noninvasively, by molecular imaging with target-specific tracers. To this day, noninvasive imaging has focused on CTLA-4 and PD-1/PD-L1, while there is no noninvasive tool available to accurately assess LAG-3 expression in vivo. In this proof-of-concept study, we developed nanobodies, the smallest functional fragments from camelid heavy chain-only antibodies, to noninvasively evaluate mouse LAG-3 expression using single positron emission computed tomography (SPECT)/CT imaging. The in vitro characterization of 114 nanobodies led to the selection of nine nanobodies binding to mouse LAG-3. The injection of ^99mTechnetium-labeled nanobodies in healthy mice showed specific uptake in immune peripheral organs like the spleen and lymph nodes, which was not observed in LAG-3 gene knock-out mice. Moreover, nanobody uptake could be visualized using SPECT/CT and correlated to the presence of LAG-3 as assessed in flow cytometry and immunohistochemistry. SPECT/CT scans of tumor bearing mice further confirmed the diagnostic potential of the nanobodies. These findings substantiate the approach to use nanobodies as a tool to image inhibitory immune checkpoints in the tumor environment.

Different histological and molecular subtypes of pancreatic ductal adenocarcinoma (PDAC), with different molecular composition and survival statistics, have recently been recognised. This review describes the currently available studies regarding molecular and histological subtypes in PDAC. Studies from major cohorts such as International Cancer Genome Consortium as well as smaller cohorts are reviewed. We discuss where the described subtypes overlap, where the discrepancies are and which paths forward could be taken regarding diagnosis, ontogeny and therapy. Four molecular subtypes with strong overlap among the different studies can be found, next to a list of mixed findings. Two of the four subtypes (epithelial classical and mesenchymal basal-like) were represented in every study and were often discriminated in other solid tumours as well. These two subtypes differ substantially in prognosis. One biomarker has been discovered, only discriminating these two subtypes, and insights into subtype-specific therapeutic vulnerabilities are scarce. Subtypes can be reproducibly detected in cohorts of PDAC patients and two of them directly relate with prognosis. A consensus on the subtypes is warranted. Further discovery and validation studies are needed to identify strong biomarkers, to comprehend subtype ontogeny and to define strategies for precision medicine.