ECP 2022 Abstract Book

Virchows Archiv (2022) 481 (Suppl 1):S1–S364 13 by MANTIS. Their MSI status was confirmed by standard PCR analyses, while none of the other MEITLs or EATLs showed insta- bility. Altogether, our analyses estimated a prevalence of MSI of 11% (3/28) in the MEITL subtype. Notably, none of the MSI samples presented somatic mutations within MMR pathway genes (MLH1, PMS2, MSH2, MSH6), suggesting alternative mutagenic mechanisms. Conclusion: Taken together, these data reveal a relatively high TMB in ITCLs when compared to other peripheral T-cell lym- phomas, and an MSI status in a subset of MEITLs. This suggests a role of MMR deficiency in the oncogenesis of a proportion of ITCLs, with potential clinical implications. Alternative muta- genic mechanisms, possibly involving the intestinal environ- ment, seem to play a role in the tumorigenesis of the majority of ITCLs. MD-04-005 Extensive spatial characterisation of the tumour-microenviron- ment in a large clinical non-small cell lung cancer cohort and correlation with clinical and pathological parameters S. Ruane*, S. Schallenberg, C. Böhm, G. Dernbach, M. Gottschalk, N. Aldoj, X. Liang, R. Büttner, A. Quaas, S. Merkelbach-Bruse, K. Hekmat, M. Heldwein, G. Schlachtenberger, D. Horst, M. Drago- mir, J. Rückert, N. Frost, F. Leiss, M. Alber, F. Klauschen *Aignostics, Germany Background & objectives: We present an AI-driven image analysis workflow combining histomorphological and multiplex immunofluorescence data for automated, cell-level characterization of the tumour microenvironment (TME) that can facilitate biomarker identification. The workflow is utilized for the exploration of a clinical NSCLC cohort. Methods: 340 clinical NSCLC cases from an FFPE tissue micro- array were stained with a 12-plex immunofluorescence panel (IF) and subsequently with hematoxylin and eosin (H&E). All stains were scanned and co-registered with single cell accuracy. Deep learning models were developed to detect tumour regions from H&E and cell subtypes from IF and H&E. The resulting readouts were correlated with clinical parameters. Results: Deep learning models were trained to quantify the presence of tumour cells expressing PD-L1 and lymphocyte cell subtypes. For the task of separating PD-L1+ carcinoma, PD-L1- carcinoma and other cells a balanced accuracy (BA) of 92% was reached. The task of separating FoxP3+ T-cells, CD8+ T-cells and B-cells respectively from other cells was performed with a BA of > 90%. The models were trained and evaluated using over 13,000 IF-informed manual pathologist annotations. The evaluation was performed on a held out TMA section with separate cases. The correlation of PD-L1 expression and vari- ous tumour infiltrating lymphocyte subpopulation levels with patient prognosis and other clinical parameters was computed. Conclusion: The presented workflow allows for scalable char- acterization of the NSCLC TME by (1) staining the same sec- tion with H&E and multiplex IF images and registering result- ing scans and (2) applying deep learning for robust cell subtype detection from both modalities at once. First results show that cells involved in the immune response within the TME can be precisely quantified and correlated with clinical parameters. Posters PS-01 | Poster Session Breast Pathology PS-01-001 Insulinoma-associated protein 1 (INSM1) expression in breast carcinomas with neuroendocrine morphologies: application and future perspectives T. Kawasaki*, T. Tashima, Y. Nakamura, A. Fujimoto, Y. Usami, A. Fujita, H. Imai, Y. Baba, M. Nakahira, S. Ryozawa, K. Oka- moto, H. Kagamu, T. Inozume, C. Muramatsu, M. Saitoh, K. Taniyama, H. Nagai, T. Kondo, A. Enomoto, K. Kaira *Department of Pathology, Saitama Medical University Interna- tional Medical Center, Japan Background & objectives: Insulinoma-associated protein 1 (INSM1) is a zinc-finger transcription factor initially isolated from a human insulinoma subtraction library. INSM1 was recently demonstrated to be a better diagnostic and prognostic indicator for small cell lung carcinoma than the traditional neuroendocrine (NE) markers. Methods: Herein, for the first time, we present eight cases with NE phenotype mammary neoplasms in which the NE nature of the tumours was confirmed solely by INSM1. Patients were 35–64 (mean: 48.9) year-old women with breast tumours showing char- acteristic NE morphologies, i.e. solid growth of polygonal, short- spindle or plasmacytoid cells with fine-granular cytoplasm and nuclei, and a well-developed vascular network. Results: On immunohistochemical examinations, these malignancies showed diffuse nuclear expressions of INSM1 (mouse monoclonal, clone A-8: sc-271408, dilution 1:100; Santa Cruz Biotechnology, Inc., Dallas, TX), whereas chromogranin A [three sources: 1) mouse monoclonal, clone LK2H10; Roche Diagnostics, Mannheim, Germany, 2) rabbit polyclonal, dilution 1:500; Dako, Copenhagen, Denmark, and 3) rabbit polyclonal, 412751; Nichirei Bioscience Inc., Tokyo, Japan] and synaptophysin [two sources: 1) rabbit polyclonal, dilution 1:50; Dako, and 2) mouse monoclonal, clone 27G12: 413831; Nichirei] staining did not correspond to distinct NE features in the neoplastic cytoplasm. Finally, we diagnosed these cancers of luminal-like immuno-subtype as 4 neuroendocrine neoplasms (NENs), three hypercellular mucinous carcinomas, and one neuroendocrine ductal carcinoma in situ. Conclusion: Based on the establishment of INSM1, a promising NE marker with high sensitivity and specificity, accompanied by our current immunohistochemical results, the frequency of detect- ing NE differentiation in systemic neoplasms, including mammary NENs as well as carcinomas with NE differentiation such as type B mucoid carcinoma and solid papillary carcinoma, is anticipated to increase. Our observations might contribute to the development of novel treatments including molecular-targeted therapies for these tumour entities. Funding: Tomonori Kawasaki is supported by Grants-in-Aid for Scientific Research (No. 21K06910 and No. 20K08131) from the Japanese Ministry of Education, Culture, Sports, Science and Technology and the National Hospital Organization (NHO) Grant (H29-NHO-01). PS-01-002 PRAME expression in invasive breast carcinoma L. Korša*, M. Abramovic, T. Pikivaca, L. Kovačević, M. Milošević, M. Prutki, Z. Marušić *Clinical Department of Pathology and Cytology, University Hos- pital Centre Zagreb, Croatia Background & objectives: PRAME (PReferentially expressed Antigen in MElanoma) is a carcinoma testis antigen expressed in numerous tumour types. The aim of this study was to assess PRAME expression in different surrogate subtypes of breast car- cinoma and its correlation with other prognostic factors. S60