Lars Andreas Akslen

Young age is associated with poorer outcomes and more aggressive clinicopathological features in breast cancer, but the biological underpinnings are incompletely understood. With the emergence of infiltrating immune cell tissue‐based biomarkers, we aimed to investigate a relation to age by mapping of CD45, CD3, CD8, FOXP3, and CD163 positive cells. We examined a population‐based cohort of women under the age of 50 years (BCY‐B cohort, n = 311) in comparison to a previously reported series of patients aged 50–69 (TBC cohort, n = 282) years. The BCY‐B cohort was additionally stained for two PD‐L1 markers. Young age (under 50 years) was not associated with higher levels of tumor‐infiltrating lymphocytes (TILs), represented by markers such as CD45 and CD3. However, younger patients presented with lower levels of FOXP3‐positive TILs, both in luminal and non‐luminal subgroups, and higher levels of CD163‐positive tumor‐associated macrophages (TAMs) in combined analyses only. CD163 showed prognostic value when analyzing all patients together and within the luminal subgroups. For non‐luminal subgroups, CD3 expression was prognostic, while CD8 expression showed a trend. Notably, no age‐associated difference in prognostic value was observed for any marker. Among patients within the BCY‐B cohort (aged under 50 years), those below 40 years demonstrated higher PD‐L1 tumor cell expression, also within the triple‐negative subgroup, when compared to those aged 40–49 years. In conclusion, our study found only limited differences in tumor‐infiltrating immune cells with age, with higher density of FOXP3‐positive TILs in patients aged 50 years or above. Within the BCY‐B cohort (age under 50), patients aged under 40 years showed higher tumor expression of PD‐L1 when compared to those aged 40–49 years, suggesting age‐associated differences in immune checkpoint regulation.

Purpose: Mitotic count (MC) and Ki67 expression are well-known prognostic factors in breast cancer. Whereas MC is part of histologic grade but not always reported separately, Ki67 is hampered by lack of methodological standardization. Our aim was to assess how proliferation markers MC, Ki67, and PHH3 perform in prognostication based on different tissue categories (WS, whole sections; CNB, core needle biopsies; TMA, tissue microarrays), and how these markers can stratify hormone receptor positive/HER2 negative breast cancer. Methods: We examined a prospective population-based breast cancer series including 534 women (50–69 years) diagnosed during 1996–2003, with a median follow-up time of 13 years. MC, Ki67, and mitotic count by PHH3 were assessed in paired samples of WS ( N = 534), CNB ( N = 154) and TMA ( N = 101). The level of MC in four additional and independent cohorts are presented for comparison (1598 cases). Results: By univariate survival analysis, all proliferation markers showed significant prognostic impact. By multivariate models, MC and Ki67 demonstrated independent prognostic significance. In the luminal/HER2 negative subgroup, which is currently the primary target group for proliferation assessment, MC retained prognostic significance in multivariate analysis, whereas Ki67 and PHH3 did not. In whole sections, a mitotic count of 2.5 per mm ² corresponded to Ki67 of 20% as cutoff value for separating luminal B from luminal A tumors (HER2 negative cases). Conclusion: Manual proliferation markers should preferably be assessed on WS for prognostication. Our findings indicate that mitotic count might support or replace Ki67 as a marker for stratification of luminal breast cancer, being particularly useful as an accessible biomarker in settings with limited resources.

Breast cancers are biologically and clinically diverse. While large-scale gene expression analyses have enabled epithelial-centered molecular classifications, studies of the tumor microenvironment (TME) remain limited, especially at the proteome level using tissue-specific resolution. By laser capture microdissection and mass spectrometry-based proteomics followed by unsupervised clustering of the stromal proteome, we discovered three patient subgroups. The largest cluster revealed the most discrete representation of stromal proteins, including a 35-protein (35P) panel linked to extracellular matrix biology, tumor progression programs, and increased abundance of tumor-associated macrophages (TAM) by single-cell profiling. Clinical validation of 35P, using whole tissue protein and mRNA values from different cohorts of ER+/HER2− breast cancer, including a large randomized controlled trial (STO), identified that more aggressive (or ‘high-grade’) stromal features were independent of current molecular subtypes. The 35P stromal panel may reflect important clinical information by improving patient stratification beyond current epithelial-based classification of breast cancer.

Whole slide imaging has transformed the field of pathology by enabling high-resolution digitization of histopathological slides. However, the large image size and variability in morphology, tissue processing, and imaging can pose challenges for robust computational analysis. When working with specific tasks in digital pathology, conventional feature extractors pretrained on general images may not provide features as relevant as those trained on histopathological images. To address this, foundation models pretrained on histopathological images have been developed. Yet, their large size and computational demands might limit widespread adoptions to specific tasks. To facilitate the low-cost adoption of these models, we utilized low-rank adaptation for finetuning the model and developed evolving prototype-based multiple instance learning (EP-MIL). Our method’s capabilities were demonstrated by applying it to the classification of two histological subtypes of lung cancer. The results show that our approach achieves competitive performance when benchmarked against a state-of-the-art technique (CLAM), while offering improvements in efficiency. Specifically, our proposed method requires 8.3 times less training runtime compared with CLAM, uses less than 200.0 MB of memory during training, and enables 73.8 times faster inference runtime. These efficiency gains, combined with competitive performance, suggest that utilizing evolving prototypes with LoRA-tuned foundation models offers a more efficient and practical approach for broader use of foundation models in resource-constrained clinical settings.

In breast cancer (BC), the transcription factor GATA3 is linked to estrogen receptor (ER) alpha biology, and its loss is associated with aggressive tumor features. Little is reported about potential roles and implications of GATA3 independent of ER, and possible relationships to the BC tumor microenvironment (TME) have not been much explored. Thus, the discovery of novel biomarkers potentially linked to ER and GATA3 functions and predicting aspects of the TME could significantly improve precision in the management of patient subgroups. We examined GATA3 protein and mRNA expression in a large in‐house population‐based BC series ( n = 837), and in the METABRIC datasets (METABRIC Discovery, n = 997 and METABRIC Validation, n = 995). Associations with primary BC phenotypes, transcriptional programs, TME features, clinical outcomes, and potentially independent roles of GATA3 are reported. We find that low GATA3 expression associates with aggressive features like increased tumor diameter, higher histological grade, triple negative BC, and a basal‐like (CK5/6 positive) phenotype. Low GATA3 mRNA expression associated with downregulation of ER‐related genes, upregulation of transcriptional signatures reflecting hypoxia, and enrichment of gene sets reflecting tumor cell proliferation, epithelial‐mesenchymal transition, and stemness. Low GATA3 protein and mRNA expression both associated with overall reduced BC‐specific survival. Notably, low GATA3 expression strongly associated with upregulation of immune checkpoint markers, T‐cell activation, and metabolic alterations not previously described in BC. Gene expression patterns underlying GATA3 ‐low tumors, independent of ER status, reflected activation of immunological and metabolic processes. This study suggests that GATA3 might influence the TME independent of ER status. Our results point to metabolic and immunophenotypic alterations in GATA3 ‐low BCs, in particular with T‐cell activation and increased expression of immune checkpoints. These findings could be relevant for patient selection in the context of immunotherapies and potential targeting of metabolic pathways.

Introduction: Breast cancer remains a major health challenge due to its molecular heterogeneity and complex interactions with the tumor microenvironment. Adrenergic signaling, mediated by stress hormones such as noradrenaline, has emerged as a potential regulator of cancer progression, influencing cell proliferation, cell adhesion, migration, and invasion.Methods: This study investigates the effects of adrenergic modulation on breast cancer spheroids from basal-like (MDA-MB-231, BT549) and luminal-like (T47D, MCF7) cell lines, using 3D culture systems as a more physiologically relevant model compared to traditional 2D monolayer cultures. The 3D spheroid model better recapitulates the structural complexity of tumors, providing insights into cell–cell and cell-matrix interactions that influence signaling pathways and drug responses.Results: Noradrenaline treatment significantly reduced spheroid size, invasion capacity, and the expression of EMT-related markers and integrins in MDA-MB-231 cells. These effects were partially reversed by propranolol, a non-selective beta-adrenergic receptor antagonist. Luminal-like spheroids, characterized by low ADRB2 abundance, displayed limited responsiveness to adrenergic modulation. Proteomic analysis revealed distinct subtype-specific responses, with basal-like spheroids showing pronounced alterations in pathways related to proliferation, cytoskeletal dynamics, epithelial-mesenchymal transition, and metabolism, whereas luminal-like spheroids exhibited minimal changes.Discussion: Our findings reveal heterogeneity in adrenergic receptor signaling across basal-like and luminal-like breast cancer cell lines, and also within the basal-like subgroup. This diversity underscores the complexity of adrenergic signaling in breast cancer and highlights the advantages of 3D culture systems. These results provide valuable insights into the subtype-specific patterns of response to adrenergic signaling that contribute to tumor progression and may inform future studies including evaluation of therapeutic strategies.

Summary: The limited efficacy of immunotherapy in clinical trials in high-grade serous ovarian cancer (HGSOC) may improve by implementing models more reflective of human biology into preclinical studies. To address this, we developed and validated a humanized patient-derived xenograft mouse model of HGSOC. Human hematopoietic stem cells and patient-derived HGSOC were engrafted into immunodeficient mice. The mice were administered durvalumab and/or oleclumab intraperitoneally semi-weekly for five weeks. The immunotherapy was well-tolerated, though no responses occurred. Leukocytes in primary tumors were analyzed immunohistochemically, and circulating T cells were characterized using spectral flow cytometry. All tumors exhibited an immune-excluded immunophenotype. No significant inter-group differences in disease burden, intratumoral leukocyte density, or circulating T-cells were observed. In the durvalumab-only group, tumor burden significantly positively correlated with intratumoral cytotoxic and regulatory T-cell densities. This model reflects human disease biology and clinical findings, providing a robust platform for studying tumor-immune interactions and immunosuppressive mechanisms in HGSOC.

Digital pathology enables automatic analysis of histopathological sections using artificial intelligence. Automatic evaluation could improve diagnostic efficiency and find associations between morphological features and clinical outcome. For development of such prediction models in breast cancer, identifying invasive epithelial cells, and separating these from benign epithelial cells and in situ lesions would be important. In this study, we trained an attention gated U-Net for segmentation of epithelial cells in hematoxylin and eosin stained breast cancer sections. We generated epithelial ground truths by immunohistochemistry, restaining hematoxylin and eosin sections with cytokeratin AE1/AE3, combined with pathologists’ annotations. Tissue microarrays from 839 patients, and whole slide images from two patients, were used for training and evaluation of the models. The sections were derived from four breast cancer cohorts. Tissue microarray cores from a fifth cohort of 21 patients was used as a second test set. In quantitative evaluation, mean Dice scores of 0.70, 0.79, and 0.75 were achieved for invasive epithelial cells, benign epithelial cells, and in situ lesions, respectively. In qualitative scoring (0-5) by pathologists, the best results were reached for all epithelium and invasive epithelium, with scores of 4.7 and 4.4, respectively. Scores for benign epithelium and in situ lesions were 3.7 and 2.0, respectively. The proposed model segmented epithelial cells well, but further work is needed for accurate subclassification into benign, in situ, and invasive cells.