ECP 2023 Abstracts

S164 Virchows Archiv (2023) 483 (Suppl 1):S1–S391 13 *The Fourth Hospital of Hebei Medical University, China Background & objectives: Most of its issues are still undecided on the relationship between tumour mutation burden (TMB) and immune- related genes in the breast cancer. This study explores their relationship based on gene mutation and transcription data in the TCGA database. Methods: Download and extract the somatic mutation data in the TCGA database. Cases are divided into Low-TMB and High-TMB subgroups. Differentially expressed immune-related genes were identified in differ- ent TMB subgroups, and predicted patient prognosis and gene function enrichment analysis, while comparing different TMB subgroups and invasive immune cells, different clinical pathological features. Results: A total of 986 mutation data with breast cancer patients were obtained. Compared with Low-TMB groups, the survival period of High- TMB group was relatively longer, and the high and low expression of TMB was related to some clinical pathological characteristics. There were 337 differential expression genes between TMB subgroups. Of these genes, 7 differentially expressed immune-related genes are associated with progno- sis. Among the High-TMB group, activated CD4 + memory T cells and other cells have high expression, while the expression ratio of memory B cells and other cells in the Low-TMB group is high. But these immunocytes are unrelated to the survival of the patients (P> 0.05). Conclusion: TMB-related immunological infiltration characteristics have good predicted value for prognosis of breast cancer patients. Dif- ferentially expressed immune-related genes between TMB subgroups provide important predictive information for the survival. E-PS-02-045 Histopathological H&E-based evaluation of HER2 status in breast cancer through a novel deep learning approach C. Lupo*, N. Casatta, G. Gerard, G. Cervi, A. Polonia, N. Fusco, G. Curigliano, C. Eloy *Innovation Department, Diapath SpA, Martinengo (BG), Italy Background & objectives: Recent research introduced a new ‘HER2- low’ expression in breast cancer. The accurate definition of the HER2 profile is thus showing the need for more sensitive diagnostic tools to ensure that eligible patients are not deprived of effective therapies. Methods: In computational pathology, self-supervised learning is a new paradigm for learning feature representations without fully labelled data. In particular, Vision Transformers (ViTs), which rely on self-attention mechanisms to process image data, were used as the base for the patch-wise classification of different histopathological datasets. The model aimed at the classification of HER2 biomarker expression on H&E slides of breast cancer. Results: Through the successful project collaboration among the experts of the University of Milan, two Italian innovative companies and an excellence reference centre and through the shared knowledge of different expert teams, a model was developed. It properly defines areas interested by the presence of HER2 in WSI of breast samples on the basis of H&E slides. The ultimate goal of the project is to train the system not only to quantify HER2 expression but also to discriminate HER2 low cases on H&E WSI. The preliminary results of the model were further validated outside the training environment on real cases provided by a reference centre. Conclusion: Preliminary results show that ViTs have great potential in the development of innovative Machine Learning systems for AI appli- cations in Pathology. This works finally aims at providing an H&E- based algorithm that can predict HER-2 status and treatment response in breast cancer at an accuracy that may benefit clinical evaluations, especially for the identification of HER2 low-expression patients. E-PS-02-046 Characterization of EGFR expression in low oestrogen receptor positive breast carcinoma A. Maaoui*, Y. Houcine, S. Elfekih, S. Kamoun, S. Ayadi, M. Driss *Pathology Department, Salah Azaiez Institute, Tunis, Tunisia Background & objectives: EGFR (epidermal growth factor receptor) is expressed in breast cancer, particularly in the triple-negative subtype. Some low positive oestrogen receptor (ER) breast carcinomas may also express EGFR. We aim to characterize the EGFR expression in low positive ER breast carcinomas. Methods: Patients with low ER+ breast carcinoma from 2019-2021 were identified (n=20). Low ER positive if 1 to 10% ER positivity by immunohistochemistry (IHC) was present. An immunohistochemical study using the anti-human EGFR monoclonal antibody was performed. Interpretation was based on a scoring system identical to that of HER2neu. Cases with an overall score ≥ 1 were considered positive. Results: The median age was 57 years (range 35-80). SBR grade III predominated (67%). The presence of tumour emboli was noted in 50% and lymph node metastases in 30% of the cases. Positive EGFR expression was noted in less than half of the cases (6 cases). EGFR seemed to be associated with high grade tumour, tumour emboli and lymph node metastases. EGFR tends to be more expressed in younger patients (≤ 50 years) without statistical significance. The median follow-up was 400 days. Five patients were alive with disease, 4 patients died due to invasive breast carcinoma, 3 developed metastases and 3 recurrences. No evolution data were available for the other patients. Conclusion: EGFR expression in low positive ER breast cancer has been associated with a more aggressive tumour phenotype. However, the clinical significance of EGFR expression in this subset of breast cancer is still unclear and further research is needed to fully understand its role in tumour development and progression. E-PS-02-047 Precursor lesions and biomarker profile of mucinous breast carcinoma F. Mengna*, Y. Zhang, L. Yang, M. Shen, F. Li, B. Wei, H. Bu, Z. Zhang *Department of Pathology, China Background & objectives: Mucinous carcinomas (MCs) of the breast is the most common Invasive breast carcinoma with extracellular mucin, the morphologic evolutionary pathways were not fully studied. We aim to explore the precursor lesions, clinicopathological character- istics and prognosis of MCs with different biomarkers. Methods: A total of 245 MCs diagnosed by surgical specimens among 2014 to 2021 were included, including 163 pure mucinous carcinomas (PMCs) and 82 mixed mucinous carcinomas (MMCs). The morphology of the surrounding breast tissue adjacent to the MCs were evaluated. The follow up data were obtained and compared among PMCs and MMCs with different clinicopathological features and biomarkers expression. Results: The median age at diagnosis was 52 years (range 20-85) and mean tumour size was 2.9 cm (range 0.5-16.0). Of all the 245 cases, DCIS were observed in 87 cases, 15 of them were high grade and coexisted with Luminal B MMCs (12/15). Mucocele-like lesion was observed in 44 cases, 68.2% of them were associated with Luminal A type A PMC. Solid papillary carcinoma was observed in 2 type B MCs and 10 MMC. Besides LN involvement and higher stage, <40 years, more micropapillary architecture and non-mucinous component (>50%), HER2-positive and coexist with high grade DCIS were sig- nificant factors for poorer prognosis. Conclusion: MCs of breast is generally less aggressive carcinoma. Focusing on precursor lesions and high risk clinicopathologic charac- teristics can help provide more effective treatment for such patients. Our findings may support the concept that progression of varies IHC pheno- types of MC with specific precursor lesion is a unique evolution pathway.