Objective: HER2 expression is a key prognostic and treatment-influencing factor in breast cancer and is assessed for all invasive breast carcinoma (BC). As with all immunohistochemistry (IHC) staining, visual interpretation of HER2 expression is subjective, which leads to intra- and inter-pathologist variability. This study aims to evaluate the clinical utility (concordance, accuracy, and user feedback) of artificial intelligence (AI)-aided HER2 scoring solution on whole slide digital images of HER2 IHCs of breast samples. Methods: The cohort included biopsies and excisions from 1,997 patients from 12 US, EU, and UK clinical laboratories, including academic medical centers and reference/private laboratories. HER2 slides of diverse BC subtypes from primary and metastatic tumors were stained with anti-HER2 antibody (4B5, VENTANA) at each laboratory and scanned with different scanners (Leica GT450DX, Philips UFS, Aperio AT2). This observational two-arm multi-reader study compared the performance of 26 pathologists (“readers”) on HER2 scoring (each reviewed 50-200 slides) unassisted vs. aided by AI HER2 solution (Ibex Breast HER2®), which detects the invasive tumor area and on slide control, classifies tumor cells based on their staining pattern, and derives a slide-level HER2 score by applying 2018 ASCO/CAP guidelines. Both study arms were compared to ground truth (GT), established as majority score of three breast pathologists (“experts”) who reviewed the slides manually. Results: Experts’ overall inter-observer agreement on all HER2 scores was 73.9% (95%CI: 72.6%,75.2%) and for 0/1+/2+/3+ was 80.8%/ 66.2%/ 63.7%/ 94.3%, respectively. Readers’ overall inter-observer agreement was significantly higher when assisted by AI, 87.5% (85.9%,89.0%) vs. 74.3% (72.3%,76.3%) without AI, p <0.05. Moreover, reader's accuracy for all HER2 scores (agreement with GT) was significantly higher with AI 80.9% (79.7%,82.2%) vs without AI, 76.6% (75.2%,77.9%) (p <0.05). For 0/1+ vs 2+/3+ cutoff, readers with AI showed significantly higher inter-observer agreement 93.1% (91.8%,94.2%) vs. without AI 86.8% (85.2%,88.3%), p <0.05, and significantly higher accuracy, 91.9% (91.0%,92.8%) with AI vs 88.8% (87.7%,89.8%) (p <0.05) without AI. For the HER2-low cutoff of 0 vs. 1+/2+/3+, readers with AI showed significantly higher inter-observer agreement of 95.0% (93.9%,95.9%) vs. 88.8% (87.3%,90.2%), p <0.05, with slightly higher accuracy (89.8% (88.8%,90.7%) with AI vs 88.5% (87.4%,89.4%) without AI). The standalone automatic AI solution demonstrated high accuracy for HER2 scoring of 89.4% (88.0%,90.7%), and 91.2% (89.8%,92.3%) for the respective clinical cutoffs of 0 vs. 1+/2+/3+, 0/1+ vs. 2+/3+. Feedback from reader pathologists' user survey indicates an increased confidence in their HER2 scoring accuracy and consistency. Additionally, 77% of the pathologists expressed a preference for HER2 scoring supported by AI over manual scoring. Conclusions: This study reports a large multi-site validation of a fully automated AI solution for HER2 scoring in BC. Pathologists supported by AI showed significant improvements in HER2 scoring consistency, evidenced by inter-reader agreement, and accuracy overall and for other clinical cutoffs. The AI solution demonstrated high accuracy and generalizability to multiple different laboratories (pre-analytics and staining protocols) and scanners. AI solutions, such as the one investigated here, could be used as decision-support tools for pathologists in routine clinical practice, enhancing the reproducibility and consistency of HER2 scoring, thus enabling optimal treatment pathways and improved patient outcomes. Citation Format: Savitri Krishnamurthy, Stuart Schnitt, Anne Vincent-Salomon, Elena Provenzano, Rita Canas-Marques, Laurent Arnould, Gaetan Mac Grogan, Elisabeth Shearon, Derek Welch, Pranil Chandra, Piotr Borkowski, Sabine Declercq, Joseph Loane, Anu Gunavardhan, Luca Di Tommaso, Vitor Krauss, Jeanne Thomassin, Marie Brevet, Maya Grinwald, Dana Mevorach, Sevde Etoz, Raz Ziv, Shai stein, Giuseppe Mallel, Judith Sandbank, Chaim Linhart, Manuela Vecsler. Improved Standardization and Accuracy of HER2 Score with AI support in Breast Cancer: Large Multicenter Study [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P1-07-03.

PURPOSE The proven efficacy of human epidermal growth factor receptor 2 (HER2) antibody-drug conjugate therapy for treating HER2-low breast cancers necessitates more accurate and reproducible HER2 immunohistochemistry (IHC) scoring. We aimed to validate performance and utility of a fully automated artificial intelligence (AI) solution for interpreting HER2 IHC in breast carcinoma. MATERIALS AND METHODS A two-arm multireader study of 120 HER2 IHC whole-slide images from four sites assessed HER2 scoring by four surgical pathologists without and with the aid of an AI HER2 solution. Both arms were compared with high-confidence ground truth (GT) established by agreement of at least four of five breast pathology subspecialists according to ASCO/College of American Pathologists (CAP) 2018/2023 guidelines. RESULTS The mean interobserver agreement among GT pathologists across all HER2 scores was 72.4% (N = 120). The AI solution demonstrated high accuracy for HER2 scoring, with 92.1% agreement on slides with high confidence GT (n = 92). The use of the AI tool led to improved performance by readers, interobserver agreement increased from 75.0% for digital manual read to 83.7% for AI-assisted review, and scoring accuracy improved from 85.3% to 88.0%. For the distinction of HER2 0 from 1+ cases (n = 58), pathologists supported by AI showed significantly higher interobserver agreement (69.8% without AI v 87.4% with AI) and accuracy (81.9% without AI v 88.8% with AI). CONCLUSION This study demonstrated utility of a fully automated AI solution to aid in scoring HER2 IHC accurately according to ASCO/CAP 2018/2023 guidelines. Pathologists supported by AI showed improvements in HER2 IHC scoring consistency and accuracy, especially for distinguishing HER2 0 from 1+ cases. This AI solution could be used by pathologists as a decision support tool for enhancing reproducibility and consistency of HER2 scoring and particularly for identifying HER2-low breast cancers.

The core Hippo pathway module consists of a tumour-suppressive kinase cascade that inhibits the transcriptional coactivators Yes-associated protein (YAP) and WW domain-containing transcription regulator protein 1 (WWTR1; also known as TAZ). When the Hippo pathway is downregulated, as often occurs in breast cancer, YAP/TAZ activity is induced. To elaborate the roles of TAZ in triple-negative breast cancer (TNBC), we depleted Taz in murine TNBC 4T1 cells, using either CRISPR/Cas9 or small hairpin RNA (shRNA). TAZ-depleted cells and their controls, harbouring wild-type levels of TAZ, were orthotopically injected into the mammary fat pads of syngeneic BALB/c female mice, and mice were monitored for tumour growth. TAZ depletion resulted in smaller tumours compared to the tumours generated by control cells, in line with the notion that TAZ functions as an oncogene in breast cancer. Tumours, as well as their corresponding in vitro cultured cells, were then subjected to gene expression profiling by RNA sequencing (RNA-seq). Interestingly, pathway analysis of the RNA-seq data indicated a TAZ-dependent enrichment of 'Inflammatory Response', a pathway correlated with TAZ expression levels also in human breast cancer tumours. Specifically, the RNA-seq analysis predicted a significant depletion of regulatory T cells (Tregs) in TAZ-deficient tumours, which was experimentally validated by the staining of tumour sections and by quantitative cytometry by time of flight (CyTOF). Strikingly, the differences in tumour size were completely abolished in immune-deficient mice, demonstrating that the immune-modulatory capacity of TAZ is critical for its oncogenic activity in this setting. Cytokine array analysis of conditioned medium from cultured cells revealed that TAZ increased the abundance of a small group of cytokines, including plasminogen activator inhibitor 1 (Serpin E1; also known as PAI-1), CCN family member 4 (CCN4; also known as WISP-1) and interleukin-23 (IL-23), suggesting a potential mechanistic explanation for its in vivo immunomodulatory effect. Together, our results imply that TAZ functions in a non-cell-autonomous manner to modify the tumour immune microenvironment and dampen the anti-tumour immune response, thereby facilitating tumour growth.

Objective: Tumor HER2 expression is a key prognostic and treatment influencing factor in breast cancer. As with all immunohistochemistry (IHC) staining, visual interpretation of HER2 expression is subjective, which leads to intra- and inter-pathologist variability. Recent findings on the efficacy of HER2-targeted therapy on HER2-low patients raise the need for accurate and reproducible scoring. We developed a fully automated, artificial intelligence (AI) -based algorithm for HER2 scoring. The algorithm was based on ASCO/CAP 2018 guidelines and validated against rigorous ground truth (GT) established by multiple blinded expert pathologists. Methods: Algorithm development: We developed a solution that employs two steps: The first step consists of an ensemble of Deep Learning networks that process tissue regions and classify them as various tissue classes: Invasive cancer, Ductal Carcinoma In Situ (DCIS) and other morphologies. These networks were trained on slides that were automatically labeled by a separate AI system that analyzed the corresponding H&E slides and projected its findings to the HER2 IHC slides using a registration algorithm. To further enrich the training set, especially with rare and difficult cases, a team of 8 expert pathologists manually marked tissue areas and assigned them to one of the tissue classes. In total, the training set consisted of 6,400 manual annotations and 1,300 automatically-annotated slides, both collected from 9 laboratories and scanned using 3 different scanners. The second step is an ensemble of Object Detection networks that process only the regions classified as invasive cancer, detect the tumor cells within them, and classify their staining pattern (e.g., Not stained, Moderate incomplete, etc.). Finally, the detected cells are counted, and the ASCO/CAP guidelines are applied to derive the slide-level HER2 score. Validation: The validation set was comprised of 453 HER2 slides stained using the VENTANA anti-HER2/neu (4B5) Rabbit Monoclonal Primary Antibody as per manufacturer’s instructions. HER2 slides included biopsies and excisions with different breast cancer diagnoses (e.g., Infiltrating Ductal Carcinoma (IDC), Infiltrating Lobular Carcinoma (ILC), rare invasive subtypes, with and without DCIS) from 3 different laboratories. Ground truth was established by the consensus scores of a panel of 3 pathologists, who scored HER2 according to the guidelines without additional clinical considerations, such as scoring borderline 1+/2+ cases as 2+ to have additional tests performed. Results: The algorithm showed very high performance for detecting invasive cancer in HER2 tissue sections, with AUC of 0.967 (measured on 4-fold Cross-Validation classifying invasive vs. other regional classes). The algorithm demonstrated an overall accuracy of 80.3% for the HER2 scores when compared to the GT. When using different cutoffs for binary classification the resulting performance was: for 0 vs 1+/2+/3+ Kappa was 0.800; 0/1+ vs 2+/3+ Kappa was 0.728; for 0/1+/2+ vs 3+ Kappa was 0.954. The Quadratic Kappa between the AI score and the GT was 0.898, which is considered almost perfect. The performance of the AI was similar across the different laboratories and diagnoses(e.g. IDC, ILC). Conclusion: This study reports the successful development and independent validation of a fully automatic AI-based solution for accurate HER2 scoring in breast cancer. AI solutions, such as the one reported here, could be used as decision-support tools for pathologists in routine clinical practice, enhancing the reproducibility and consistency of HER2 scoring, thus enabling optimal treatment pathways and better patient outcomes. Accurate and automatic IHC scoring solutions can also contribute to the development of new prognostic, predictive and companion diagnostic tools. Citation Format: Yuval Globerson, Lilach Bien, Jonathan Harel, Giuseppe Mallel, Geraldine Sebag, Michel Vandenberghe, Craig Barker, Tsuyoshi Matsuo, Charo Garrido, Judith Sandbank, Chaim Linhart. A fully automatic artificial intelligence system for accurate and reproducible HER2 IHC scoring in breast cancer [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P6-04-05.

Introduction: Ex vivo organ cultures (EVOC) were recently optimized to sustain cancer tissue for 5 days with its complete microenvironment. We examined the ability of an EVOC platform to predict patient response to cancer therapy.Methods : A multicenter, prospective, single-arm observational trial. Samples were obtained from patients with newly diagnosed bladder cancer who underwent transurethral resection of bladder tumor and from core needle biopsies of patients with metastatic cancer. The tumors were cut into 250 mu M slices and cultured within 24 h, then incubated for 96 h with vehicle or intended to treat drug. The cultures were then fixed and stained to analyze their morphology and cell viability. Each EVOC was given a score based on cell viability, level of damage, and Ki67 proliferation, and the scores were correlated with the patients' clinical response assessed by pathology or Response Evaluation Criteria in Solid Tumors (RECIST).Results: The cancer tissue and microenvironment, including endothelial and immune cells, were preserved at high viability with continued cell division for 5 days, demonstrating active cell signaling dynamics. A total of 34 cancer samples were tested by the platform and were correlated with clinical results. A higher EVOC score was correlated with better clinical response. The EVOC system showed a predictive specificity of 77.7% (7/9, 95% CI 0.4-0.97) and a sensitivity of 96% (24/25, 95% CI 0.80-0.99).Conclusion: EVOC cultured for 5 days showed high sensitivity and specificity for predicting clinical response to therapy among patients with muscle-invasive bladder cancer and other solid tumors.

Objective This study aimed to clinically validate the use of an AI-based solution by pathologists for the primary diagnosis of breast core needle biopsies as compared with the gold standard practice (review on the microscope). Methods A two-arm prospective reader study comparing the performance of pathologists using an AI-based solution with pathologists using a microscope was performed at two sites (different staining and digital scanners). Both arms were compared to ground truth (GT) established by the consensus of two breast pathologists. Rates of major discrepancies between each arm and GT, as determined by an adjudicating pathologist, were compared. Results Eight pathologists participated in the study and reported on 385 cases (442 HES and 330 H&E slides), each case being reported twice, once in each study arm. Pathologists first reviewed only H&E/HES slides, if requested and available, they were provided with IHCs, while the AI results were on H&E/HES only. The major discrepancy rates of the microscope arm and of the AI arm against GT were 4.42% and 3.12%, respectively, demonstrating a 29.4% reduction in major discrepancies. Pathologists with AI demonstrated very high accuracy for the detection of invasive carcinoma with sensitivity and specificity of 100% for both, as well as for DCIS/ADH with sensitivity of 92.4% and specificity of 97.8%. Conclusions This multi-site reader study reports diagnostic accuracy improvements by pathologists performing diagnosis and reporting with the support of a first read AI solution for breast biopsies. The AI solution performed accurately and generalized well for different staining platforms and different scanners. Thus, AI solutions could be used as significant aiding tools for pathologists in clinical decision-making in routine pathology practice, enhancing the quality and reproducibility of diagnosis. Citation Format: Anne Salomon, Alona Nudelman, Joanna Cyrta, Marina Maklakovski, Anat Albrecht Shach, Geraldine Sebag, Giuseppe Mallel, Ira Krasnitsky, Tali Feinberg, Manuela Vecsler, Judith Sandbank. Primary Diagnosis of Breast Biopsies supported by AI versus Microscope: Multi-Site Clinical Reader Study [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P6-04-07.

Purpose of study: Understanding the epigenetic mechanism underlying luminal-to-basal plasticity of breast cancer subtypes Experimental procedures: Epigenetic focused CyTOF, ATAC-seq, RNA-seq, FACS-sorting, Proximity Ligation Assay, orthotopic syngeneic injections of tumor cells to mouse mammary gland, and others. Summary: Breast cancer is the most frequent cancer in women. It is defined by specific types of tumors that are classified into distinct histological and molecular subtypes. However, lately, with the advent of single-cell technologies, the extent of cellular and functional intratumoral heterogeneity within breast cancer subtypes is being appreciated. Much of this heterogeneity, including transitions of luminal-to-basal cell identity, can be attributed to alterations in the activity of chromatin remodelers. The transcriptional corepressor, NCOR1, and the tumor suppressor, LATS1, are downregulated in luminal breast cancer tumors with poor prognosis. Here, we show that LATS1 impedes the emergence of an epigenetically distinct basal-like population in luminal B tumors by augmenting NCOR1 repressive activity and facilitating deacetylation of H3K27ac. Conclusion: Rigorous repression of ERα-repressed genes facilitated by LATS1-NCOR1 maintains luminal cell identity and restricts progression of luminal breast cancer. Citation Format: Yael Aylon, Noa Furth, Giuseppe Mallel, Nishanth Nataraj, Gilgi Friedlander, Ori Hassan, Rawan Zoabi, Benjamin Cohen, Tomer Salame, Saptaparna Mukherjee, Randy Johnson, Efrat Shema, Moshe Oren. Breast cancer plasticity is restricted by a LATS1-NCOR1 repressive function [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 832.

The TP53 gene is mutated in approximately 60% of all colorectal cancer (CRC) cases. Over 20% of all TP53-mutated CRC tumors carry missense mutations at position R175 or R273. Here we report that CRC tumors harboring R273 mutations are more prone to progress to metastatic disease, with decreased survival, than those with R175 mutations. We identify a distinct transcriptional signature orchestrated by p53R273H, implicating activation of oncogenic signaling pathways and predicting worse outcome. These features are shared also with the hotspot mutants p53R248Q and p53R248W. p53R273H selectively promotes rapid CRC cell spreading, migration, invasion and metastasis. The transcriptional output of p53R273H is associated with preferential binding to regulatory elements of R273 signature genes. Thus, different TP53 missense mutations contribute differently to cancer progression. Elucidation of the differential impact of distinct TP53 mutations on disease features may make TP53 mutational information more actionable, holding potential for better precision-based medicine.