[Explainable artificial intelligence in pathology]. / Erklärbare Künstliche Intelligenz in der Pathologie.

With the advancements in precision medicine, the demands on pathological diagnostics have increased, requiring standardized, quantitative, and integrated assessments of histomorphological and molecular pathological data. Great hopes are placed in artificial intelligence (AI) methods, which have demonstrated the ability to analyze complex clinical, histological, and molecular data for disease classification, biomarker quantification, and prognosis estimation. This paper provides an overview of the latest developments in pathology AI, discusses the limitations, particularly concerning the black box character of AI, and describes solutions to make decision processes more transparent using methods of so-called explainable AI (XAI).

High-dimensional in situ proteomics imaging to assess γδ T cells in spatial biology.

This study presents a high-dimensional immunohistochemistry approach to assess human γδ T cell subsets in their native tissue microenvironments at spatial resolution, a hitherto unmet scientific goal due to the lack of established antibodies and required technology. We report an integrated approach based on multiplexed imaging and bioinformatic analysis to identify γδ T cells, characterize their phenotypes, and analyze the composition of their microenvironment. Twenty-eight γδ T cell microenvironments were identified in tissue samples from fresh frozen human colon and colorectal cancer where interaction partners of the immune system, but also cancer cells were discovered in close proximity to γδ T cells, visualizing their potential contributions to cancer immunosurveillance. While this proof-of-principle study demonstrates the potential of this cutting-edge technology to assess γδ T cell heterogeneity and to investigate their microenvironment, future comprehensive studies are warranted to associate phenotypes and microenvironment profiles with features such as relevant clinical characteristics.

Toward Explainable Artificial Intelligence for Precision Pathology.

The rapid development of precision medicine in recent years has started to challenge diagnostic pathology with respect to its ability to analyze histological images and increasingly large molecular profiling data in a quantitative, integrative, and standardized way. Artificial intelligence (AI) and, more precisely, deep learning technologies have recently demonstrated the potential to facilitate complex data analysis tasks, including clinical, histological, and molecular data for disease classification; tissue biomarker quantification; and clinical outcome prediction. This review provides a general introduction to AI and describes recent developments with a focus on applications in diagnostic pathology and beyond. We explain limitations including the black-box character of conventional AI and describe solutions to make machine learning decisions more transparent with so-called explainable AI. The purpose of the review is to foster a mutual understanding of both the biomedical and the AI side. To that end, in addition to providing an overview of the relevant foundations in pathology and machine learning, we present worked-through examples for a better practical understanding of what AI can achieve and how it should be done.

[New information about tumours of the salivary glands : WHO classification 2022]. / Neues zu Tumoren der Speicheldrüsen : WHO-Klassifikation 2022.

The WHO 2022 classification of head and neck tumours contains another slight increase in the number of listed benign and malignant tumour entities of the salivary glands. This includes conceptual changes and alterations in the terminology of some entities. While some new features are regarded as preliminary or provisional, others are strongly disputed (for example the terminology of intraductal carcinoma). The impact of molecular findings, mainly recurrent gene fusions, continues to increase rapidly and some have been included in the definition of certain tumour entities. The significance of molecular findings is, however, still largely restricted to diagnostic aspects. Newly included entities include microsecretory carcinoma (defined by an SS18::MEF2C fusion), sclerosing microcystic adenocarcinoma (similar to skin adnexal tumours of the same name) and mucinous adenocarcinoma (characterized by AKT1 mutations with heterogeneous morphology).

Salivary carcinosarcoma: insight into multistep pathogenesis indicates uniform origin as sarcomatoid variant of carcinoma ex pleomorphic adenoma with frequent heterologous elements.

AIMS: The formal pathogenesis of salivary carcinosarcoma (SCS) remained unclear, both with respect to the hypothetical development from either preexisting pleomorphic adenoma (PA) or de novo and the clonal relationship between highly heterogeneous carcinomatous and sarcomatous components. METHODS AND RESULTS: We performed clinicopathological and molecular (targeted RNA sequencing) analyses on a large series of 16 cases and combined this with a comprehensive literature search (111 cases). Extensive sampling (average 11.6 blocks), combined with immunohistochemistry and molecular studies (PA-specific translocations including PLAG1 or HMGA2 proven in 6/16 cases), enabled the morphogenetic identification of PA in 15/16 cases (93.8%), by far surpassing a reported rate of 49.6%. Furthermore, we demonstrated a multistep (intraductal/intracapsular/extracapsular) adenoma-carcinoma-sarcoma-progression, based on two alternative histogenetic pathways (intraductal, 56.3%, versus myoepithelial pathway, 37.5%). Thereby, early intracapsular stages are identical to conventional carcinoma ex PA, while later extracapsular stages are dominated by secondary, frequently heterologous sarcomatous transformation with often large tumour size (>60 mm). CONCLUSION: Our findings strongly indicate that SCS (almost) always develops from PA, with a complex multistep adenoma-carcinoma-sarcoma-sequence, based on two alternative histogenetic pathways. The findings from this novel approach strongly suggest that SCS pathogenetically is a rare (3-6%), unique, and aggressive variant of carcinoma ex PA with secondary sarcomatous overgrowth. In analogy to changes of terminology in other organs, the term "sarcomatoid carcinoma ex PA with/without heterologous elements" might be more appropriate.

DNA methylation-based classification of sinonasal tumors.

The diagnosis of sinonasal tumors is challenging due to a heterogeneous spectrum of various differential diagnoses as well as poorly defined, disputed entities such as sinonasal undifferentiated carcinomas (SNUCs). In this study, we apply a machine learning algorithm based on DNA methylation patterns to classify sinonasal tumors with clinical-grade reliability. We further show that sinonasal tumors with SNUC morphology are not as undifferentiated as their current terminology suggests but rather reassigned to four distinct molecular classes defined by epigenetic, mutational and proteomic profiles. This includes two classes with neuroendocrine differentiation, characterized by IDH2 or SMARCA4/ARID1A mutations with an overall favorable clinical course, one class composed of highly aggressive SMARCB1-deficient carcinomas and another class with tumors that represent potentially previously misclassified adenoid cystic carcinomas. Our findings can aid in improving the diagnostic classification of sinonasal tumors and could help to change the current perception of SNUCs.

Subgroups in the treatment of nasal polyposis with dupilumab: A retrospective study.

Dupilumab has been shown to be safe and effective in treating chronic rhinosinusitis with polyposis (CRSwNP). There is to this date no published data whether subgroups like patients with aspirin exacerbated respiratory disease (AERD), increased histologic eosinophilia or elevated blood eosinophil or IgE-levels benefit greater from dupilumab therapy. Moreover, there is no data comparing the efficacy of functional endoscopic sinus surgery (FESS) with dupilumab therapy. We conducted a retrospective chart review of all patients that were treated at a tertiary referral center for CRswNP with dupilumab. We also contacted the patients with a questionnaire to evaluate the efficacy of previous surgeries and dupilumab therapy by visual analogue scale (VAS) and the glasgow benefit inventory (GBI) as well as report on side effects. Overall, 75 patients were included in the study at hand that reported back 138 times. While dupilumab treatment was efficient, we found no systematic evidence of greater efficacy of dupilumab in patients with AERD, histologic eosinophilia or increased blood eosinophil or IgE-levels. All patients showed a considerable decrease in subjective burden of disease, objective smell tests and endoscopic findings. From the patients point of view, dupilumab therapy showed greater efficacy both in the VAS and the GBI overall and all subcategories but "social support." Dupilumab is efficient in treating CRSwNP; this effect is independent from disease characteristics like AERD, histologic eosinophilia, serum IgE-levels or eosinophil counts. There seems to be a group of patients that benefit greater from dupilumab therapy compared to FESS.

New entity of microsecretory adenocarcinoma of salivary glands: first case with recurrence and metastases - proof of malignancy.

Microsecretory adenocarcinoma (MSA) of the salivary glands is a recently described entity. Due to lack of reported metastases, in 30 cases described until now, the designation as low-grade cancer was so far solely based on demonstration of local tumor invasion and in a single case with perineural invasion. We herein describe the first documented case with local recurrence and hematogenous metastases.

Multiparametric Phenotyping of Circulating Tumor Cells for Analysis of Therapeutic Targets, Oncogenic Signaling Pathways and DNA Repair Markers.

Detection of circulating tumor cells (CTCs) has been established as an independent prognostic marker in solid cancer. Multiparametric phenotyping of CTCs could expand the area of application for this liquid biomarker. We evaluated the Amnis® brand ImageStream®X MkII (ISX) (Luminex, Austin, TX, USA) imaging flow cytometer for its suitability for protein expression analysis and monitoring of treatment effects in CTCs. This was carried out using blood samples from patients with head and neck squamous cell carcinoma (n = 16) and breast cancer (n = 8). A protocol for negative enrichment and staining of CTCs was established, allowing quantitative analysis of the therapeutic targets PD-L1 and phosphorylated EGFR (phospho-EGFR), and the treatment response marker γH2AX as an indicator of radiation-induced DNA damage. Spiking experiments revealed a sensitivity of 73% and a specificity of 100% at a cut-off value of ≥3 CTCs, and thus confirmed the suitability of the ISX-based protocol to detect phospho-EGFR and γH2AX foci in CTCs. Analysis of PD-L1/-L2 in both spiked and patient blood samples further showed that assessment of heterogeneity in protein expression within the CTC population was possible. Further validation of the diagnostic potential of this ISX protocol for multiparametric CTC analysis in larger clinical cohorts is warranted.

DNA methylation-based machine learning classification distinguishes pleural mesothelioma from chronic pleuritis, pleural carcinosis, and pleomorphic lung carcinomas.

OBJECTIVES: Our goal was to evaluate the diagnostic value of DNA methylation analysis in combination with machine learning to differentiate pleural mesothelioma (PM) from important histopathological mimics. MATERIAL AND METHODS: DNA methylation data of PM, lung adenocarcinomas, lung squamous cell carcinomas and chronic pleuritis was used to train a random forest as well as a support vector machine. These classifiers were validated using an independent validation cohort including pleural carcinosis and pleomorphic variants of lung adeno- and squamous cell carcinomas. Furthermore, we performed differential methylation analysis and used a deconvolution method to estimate the composition of the tumor microenvironment. RESULTS: T-distributed stochastic neighbor embedding clearly separated PM from lung adenocarcinomas and squamous cell carcinomas, but there was a considerable overlap between chronic pleuritis specimens and PM with low tumor cell content. In a nested cross validation on the training cohort, both machine learning algorithms achieved the same accuracies (94.8%). On the validation cohort, we observed high accuracies for the support vector machine (97.8%) while the random forest performed considerably worse (89.5%), especially in distinguishing PM from chronic pleuritis. Differential methylation analysis revealed promoter hypermethylation in PM specimens, including the tumor suppressor genes BCL11B, EBF1, FOXA1, and WNK2. Deconvolution of the stromal and immune cell composition revealed higher rates of regulatory T-cells and endothelial cells in tumor specimens and a heterogenous inflammation including macrophages, B-cells and natural killer cells in chronic pleuritis. CONCLUSION: DNA methylation in combination with machine learning classifiers is a promising tool to reliably differentiate PM from chronic pleuritis and lung cancer, including pleomorphic carcinomas. Furthermore, our study highlights new candidate genes for PM carcinogenesis and shows that deconvolution of DNA methylation data can provide reasonable insights into the composition of the tumor microenvironment.

[Artificial intelligence: a solution for the lack of pathologists?] / Künstliche Intelligenz als Lösung des PathologInnenmangels?

Given the rapid developments, there is no doubt that artificial intelligence (AI) will substantially impact pathological diagnostics. However, it remains an open question if AI will primarily be another diagnostic tool, such as immunohistochemistry, or if AI will also be able to replace human expertise. Most current studies on AI in histopathology deal with relatively simple diagnostic problems and are not yet capable of coping with the complexity of routine diagnostics. While some methods in molecular pathology would already be unthinkable without AI, it remains to be shown how AI will also be able to help with difficult histomorphological differential diagnoses in the future.

The single-cell transcriptional landscape of lung carcinoid tumors.

Lung carcinoid tumors, also referred to as pulmonary neuroendocrine tumors or lung carcinoids, are rare neoplasms of the lung with a more favorable prognosis than other subtypes of lung cancer. Still, some patients suffer from relapsed disease and metastatic spread. Several recent single-cell studies have provided detailed insights into the cellular heterogeneity of more common lung cancers, such as adeno- and squamous cell carcinoma. However, the characteristics of lung carcinoids on the single-cell level are yet completely unknown. To study the cellular composition and single-cell gene expression profiles in lung carcinoids, we applied single-cell RNA sequencing to three lung carcinoid tumor samples and normal lung tissue. The single-cell transcriptomes of carcinoid tumor cells reflected intertumoral heterogeneity associated with clinicopathological features, such as tumor necrosis and proliferation index. The immune microenvironment was specifically enriched in noninflammatory monocyte-derived myeloid cells. Tumor-associated endothelial cells were characterized by distinct gene expression profiles. A spectrum of vascular smooth muscle cells and pericytes predominated the stromal microenvironment. We found a small proportion of myofibroblasts exhibiting features reminiscent of cancer-associated fibroblasts. Stromal and immune cells exhibited potential paracrine interactions which may shape the microenvironment via NOTCH, VEGF, TGFß and JAK/STAT signaling. Moreover, single-cell gene signatures of pericytes and myofibroblasts demonstrated prognostic value in bulk gene expression data. Here, we provide first comprehensive insights into the cellular composition and single-cell gene expression profiles in lung carcinoids, demonstrating the noninflammatory and vessel-rich nature of their tumor microenvironment, and outlining relevant intercellular interactions which could serve as future therapeutic targets.

Aberrant Expression and Subcellular Localization of ECT2 Drives Colorectal Cancer Progression and Growth.

ECT2 is an activator of RHO GTPases that is essential for cytokinesis. In addition, ECT2 was identified as an oncoprotein when expressed ectopically in NIH/3T3 fibroblasts. However, oncogenic activation of ECT2 resulted from N-terminal truncation, and such truncated ECT2 proteins have not been found in patients with cancer. In this study, we observed elevated expression of full-length ECT2 protein in preneoplastic colon adenomas, driven by increased ECT2 mRNA abundance and associated with APC tumor-suppressor loss. Elevated ECT2 levels were detected in the cytoplasm and nucleus of colorectal cancer tissue, suggesting cytoplasmic mislocalization as one mechanism of early oncogenic ECT2 activation. Importantly, elevated nuclear ECT2 correlated with poorly differentiated tumors, and a low cytoplasmic:nuclear ratio of ECT2 protein correlated with poor patient survival, suggesting that nuclear and cytoplasmic ECT2 play distinct roles in colorectal cancer. Depletion of ECT2 reduced anchorage-independent cancer cell growth and invasion independent of its function in cytokinesis, and loss of Ect2 extended survival in a Kras G12D Apc-null colon cancer mouse model. Expression of ECT2 variants with impaired nuclear localization or guanine nucleotide exchange catalytic activity failed to restore cancer cell growth or invasion, indicating that active, nuclear ECT2 is required to support tumor progression. Nuclear ECT2 promoted ribosomal DNA transcription and ribosome biogenesis in colorectal cancer. These results support a driver role for both cytoplasmic and nuclear ECT2 overexpression in colorectal cancer and emphasize the critical role of precise subcellular localization in dictating ECT2 function in neoplastic cells. SIGNIFICANCE: ECT2 overexpression and mislocalization support its role as a driver in colon cancer that is independent from its function in normal cell cytokinesis.

Mucosal melanomas of different anatomic sites share a common global DNA methylation profile with cutaneous melanoma but show location-dependent patterns of genetic and epigenetic alterations.

Cutaneous, ocular, and mucosal melanomas are histologically indistinguishable tumors that are driven by a different spectrum of genetic alterations. With current methods, identification of the site of origin of a melanoma metastasis is challenging. DNA methylation profiling has shown promise for the identification of the site of tumor origin in various settings. Here we explore the DNA methylation landscape of melanomas from different sites and analyze if different melanoma origins can be distinguished by their epigenetic profile. We performed DNA methylation analysis, next generation DNA panel sequencing, and copy number analysis of 82 non-cutaneous and 25 cutaneous melanoma samples. We further analyzed eight normal melanocyte cell culture preparations. DNA methylation analysis separated uveal melanomas from melanomas of other primary sites. Mucosal, conjunctival, and cutaneous melanomas shared a common global DNA methylation profile. Still, we observed location-dependent DNA methylation differences in cancer-related genes, such as low frequencies of RARB (7/63) and CDKN2A promoter methylation (6/63) in mucosal melanomas, or a high frequency of APC promoter methylation in conjunctival melanomas (6/9). Furthermore, all investigated melanomas of the paranasal sinus showed loss of PTEN expression (9/9), mainly caused by promoter methylation. This was less frequently seen in melanomas of other sites (24/98). Copy number analysis revealed recurrent amplifications in mucosal melanomas, including chromosomes 4q, 5p, 11q and 12q. Most melanomas of the oral cavity showed gains of chromosome 5p with TERT amplification (8/10), while 11q amplifications were enriched in melanomas of the nasal cavity (7/16). In summary, mucosal, conjunctival, and cutaneous melanomas show a surprisingly similar global DNA methylation profile and identification of the site of origin by DNA methylation testing is likely not feasible. Still, our study demonstrates tumor location-dependent differences of promoter methylation frequencies in specific cancer-related genes together with tumor site-specific enrichment for specific chromosomal changes and genetic mutations. © 2021 The Authors. The Journal of Pathology published by John Wiley & Sons, Ltd. on behalf of The Pathological Society of Great Britain and Ireland.

Artificial intelligence and pathology: From principles to practice and future applications in histomorphology and molecular profiling.

The complexity of diagnostic (surgical) pathology has increased substantially over the last decades with respect to histomorphological and molecular profiling. Pathology has steadily expanded its role in tumor diagnostics and beyond from disease entity identification via prognosis estimation to precision therapy prediction. It is therefore not surprising that pathology is among the disciplines in medicine with high expectations in the application of artificial intelligence (AI) or machine learning approaches given their capabilities to analyze complex data in a quantitative and standardized manner to further enhance scope and precision of diagnostics. While an obvious application is the analysis of histological images, recent applications for the analysis of molecular profiling data from different sources and clinical data support the notion that AI will enhance both histopathology and molecular pathology in the future. At the same time, current literature should not be misunderstood in a way that pathologists will likely be replaced by AI applications in the foreseeable future. Although AI will transform pathology in the coming years, recent studies reporting AI algorithms to diagnose cancer or predict certain molecular properties deal with relatively simple diagnostic problems that fall short of the diagnostic complexity pathologists face in clinical routine. Here, we review the pertinent literature of AI methods and their applications to pathology, and put the current achievements and what can be expected in the future in the context of the requirements for research and routine diagnostics.

Comparative investigation of cell cycle and immunomodulatory genes in mucosal and cutaneous melanomas: Preliminary data suggest a potential promising clinical role for p16 and the PD-1/PD-L1 axis.

Mucosal melanomas arise from the mucosal lining of various organs. Their etiology is currently unknown and there are no tissue-based methods to differentiate it from cutaneous melanomas. Furthermore, prognostic and predictive markers (e.g. for immune checkpoint inhibition) are lacking. In this study, we aimed to assess the protein expression levels of cell cycle-associated proteins and immune checkpoint markers in a cohort of mucosal melanomas in comparison to cutaneous melanomas and evaluated the effect of potential regulatory mechanisms. We performed immunohistochemistry, DNA methylation analysis and copy number profiling of 47 mucosal and 28 cutaneous melanoma samples. Protein expression of CD117, Ki67 and p16 was higher in mucosal melanomas, while BCL2, Cyclin D1, PD-1 and PD-L1 were overexpressed in cutaneous melanomas. CDKN2A deletions were the most prevalent numeric chromosomal alterations in both mucosal and cutaneous melanoma and were associated with decreased p16 expression. KIT was frequently amplified in mucosal melanomas, but not associated with CD117 expression. On the other hand, amplification of CCND1 lead to Cyclin D1 overexpression. In mucosal melanoma patients high PD-1 expression and high PD-L1 promoter methylation levels were associated with improved survival. PD-L1 expression correlated with response to immune checkpoint inhibitor therapy in the combined group of melanoma patients. Mucosal and cutaneous melanomas show different expression levels of cell cycle-associated and immunomodulatory proteins that are partially regulated by DNA methylation and copy number alterations. PD-1 expression and PD-L1 promoter methylation levels might be a prognostic marker for mucosal melanomas.

Machine learning models predict the primary sites of head and neck squamous cell carcinoma metastases based on DNA methylation.

In head and neck squamous cell cancers (HNSCs) that present as metastases with an unknown primary (HNSC-CUPs), the identification of a primary tumor improves therapy options and increases patient survival. However, the currently available diagnostic methods are laborious and do not offer a sufficient detection rate. Predictive machine learning models based on DNA methylation profiles have recently emerged as a promising technique for tumor classification. We applied this technique to HNSC to develop a tool that can improve the diagnostic work-up for HNSC-CUPs. On a reference cohort of 405 primary HNSC samples, we developed four classifiers based on different machine learning models [random forest (RF), neural network (NN), elastic net penalized logistic regression (LOGREG), and support vector machine (SVM)] that predict the primary site of HNSC tumors from their DNA methylation profile. The classifiers achieved high classification accuracies (RF = 83%, NN = 88%, LOGREG = SVM = 89%) on an independent cohort of 64 HNSC metastases. Further, the NN, LOGREG, and SVM models significantly outperformed p16 status as a marker for an origin in the oropharynx. In conclusion, the DNA methylation profiles of HNSC metastases are characteristic for their primary sites, and the classifiers developed in this study, which are made available to the scientific community, can provide valuable information to guide the diagnostic work-up of HNSC-CUP. © 2021 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

Single-cell RNA sequencing reveals distinct tumor microenvironmental patterns in lung adenocarcinoma.

Recent developments in immuno-oncology demonstrate that not only cancer cells, but also the tumor microenvironment can guide precision medicine. A comprehensive and in-depth characterization of the tumor microenvironment is challenging since its cell populations are diverse and can be important even if scarce. To identify clinically relevant microenvironmental and cancer features, we applied single-cell RNA sequencing to ten human lung adenocarcinomas and ten normal control tissues. Our analyses revealed heterogeneous carcinoma cell transcriptomes reflecting histological grade and oncogenic pathway activities, and two distinct microenvironmental patterns. The immune-activated CP²E microenvironment was composed of cancer-associated myofibroblasts, proinflammatory monocyte-derived macrophages, plasmacytoid dendritic cells and exhausted CD8+ T cells, and was prognostically unfavorable. In contrast, the inert N³MC microenvironment was characterized by normal-like myofibroblasts, non-inflammatory monocyte-derived macrophages, NK cells, myeloid dendritic cells and conventional T cells, and was associated with a favorable prognosis. Microenvironmental marker genes and signatures identified in single-cell profiles had progonostic value in bulk tumor profiles. In summary, single-cell RNA profiling of lung adenocarcinoma provides additional prognostic information based on the microenvironment, and may help to predict therapy response and to reveal possible target cell populations for future therapeutic approaches.

DNA methylation profiling identifies two distinct subgroups in breast cancers with low hormone receptor expression, mainly associated with HER2 amplification status.

BACKGROUND: Current clinical guidelines suggest that breast cancers with low hormone receptor expression (LowHR) in 1-10% of tumor cells should be regarded as hormone receptor positive. However, clinical data show that these patients have worse outcome compared to patients with hormone receptor expression above 10%. We performed DNA methylation profiling on 23 LowHR breast cancer specimens, including 13 samples with HER2 amplification and compared our results with a reference breast cancer cohort from The Cancer Genome Atlas to clarify the status for this infrequent but important patient subgroup. RESULTS: In unsupervised clustering and dimensionality reduction, breast cancers with low hormone receptor expression that lacked HER2 amplification usually clustered with triple negative breast cancer (TNBC) reference samples (8/10; "LowHR TNBC-like"). In contrast, most specimens with low hormone receptor expression and HER2 amplification grouped with hormone receptor positive cancers (11/13; "LowHR HRpos-like"). We observed highly similar DNA methylation patterns of LowHR TNBC-like samples and true TNBCs. Furthermore, the Ki67 proliferation index of LowHR TNBC-like samples and clinical outcome parameters were more similar to TNBCs and differed from LowHR HRpos-like cases. CONCLUSIONS: We here demonstrate that LowHR breast cancer comprises two epigenetically distinct groups. Our data strongly suggest that LowHR TNBC-like samples are molecularly, histologically and clinically closely related to TNBC, while LowHR HRpos-like specimens are closely related to hormone receptor positive tumors.