Assessment of tumor-infiltrating lymphocytes (TILs) is increasingly recognized as an integral part of the prognostic workflow in triple-negative (TNBC) and HER2-positive breast cancer, as well as many other solid tumors. This recognition has come about thanks to standardized visual reporting guidelines, which helped to reduce inter-reader variability. Now, there are ripe opportunities to employ computational methods that extract spatio-morphologic predictive features, enabling computer-aided diagnostics. We detail the benefits of computational TILs assessment, the readiness of TILs scoring for computational assessment, and outline considerations for overcoming key barriers to clinical translation in this arena. Specifically, we discuss: 1. ensuring computational workflows closely capture visual guidelines and standards; 2. challenges and thoughts standards for assessment of algorithms including training, preanalytical, analytical, and clinical validation; 3. perspectives on how to realize the potential of machine learning models and to overcome the perceptual and practical limits of visual scoring.

Gene expression profiles provide important information about the biology of breast tumors and can be used to develop prognostic tests. However, the implementation of quantitative RNA-based testing in routine molecular pathology has not been accomplished, so far. The EndoPredict assay has recently been described as a quantitative RT-PCR-based multigene expression test to identify a subgroup of hormone–receptor-positive tumors that have an excellent prognosis with endocrine therapy only. To transfer this test from bench to bedside, it is essential to evaluate the test–performance in a multicenter setting in different molecular pathology laboratories. In this study, we have evaluated the EndoPredict (EP) assay in seven different molecular pathology laboratories in Germany, Austria, and Switzerland. A set of ten formalin-fixed paraffin-embedded tumors was tested in the different labs, and the variance and accuracy of the EndoPredict assays were determined using predefined reference values. Extraction of a sufficient amount of RNA and generation of a valid EP score was possible for all 70 study samples (100%). The EP scores measured by the individual participants showed an excellent correlation with the reference values, respectively, as reflected by Pearson correlation coefficients ranging from 0.987 to 0.999. The Pearson correlation coefficient of all values compared to the reference value was 0.994. All laboratories determined EP scores for all samples differing not more than 1.0 score units from the pre-defined references. All samples were assigned to the correct EP risk group, resulting in a sensitivity and specificity of 100%, a concordance of 100%, and a kappa of 1.0. Taken together, the EndoPredict test could be successfully implemented in all seven participating laboratories and is feasible for reliable decentralized assessment of gene expression in luminal breast cancer.

Purpose: Accurate response assessment during neoadjuvant systemic treatment (NST) poses a clinical challenge. Therefore, a minimally invasive assessment of tumor response based on cell-free circulating tumor DNA (ctDNA) may be beneficial to guide treatment decisions. Experimental Design: We profiled 93 genes in tissue from 193 patients with early breast cancer. Patient-specific assays were designed for 145 patients to track ctDNA during NST in plasma. ctDNA presence and levels were correlated with complete pathological response (pCR) and residual cancer burden (RCB) as well as clinicopathologic characteristics of the tumor to identify potential proxies for ctDNA release. Results: At baseline, ctDNA could be detected in 63/145 (43.4%) patients and persisted in 25/63 (39.7%) patients at mid-therapy (MT) and 15/63 (23.8%) patients at the end of treatment. ctDNA detection at MT was significantly associated with higher RCB (OR = 0.062; 95% CI, 0.01–0.48; P = 0.0077). Of 31 patients with detectable ctDNA at MT, 30 patients (96.8%) were nonresponders (RCB II, n = 8; RCB III, n = 22) and only one patient responded to the treatment (RCB I). Considering all 145 patients with baseline (BL) plasma, none of the patients with RCB 0 and only 6.7% of patients with RCB I had ctDNA detectable at MT, whereas 30.6% and 29.6% of patients with RCB II/III, respectively, had a positive ctDNA result. Conclusions: Overall, our results demonstrate that the detection and persistence of ctDNA at MT may have the potential to negatively predict response to neoadjuvant treatment and identify patients who will not achieve pCR or be classified with RCB II/III.

Background: This study investigated the performance of ensemble learning holomic models for the detection of breast cancer, receptor status, proliferation rate, and molecular subtypes from [¹⁸F]FDG-PET/CT images with and without incorporating data pre-processing algorithms. Additionally, machine learning (ML) models were compared with conventional data analysis using standard uptake value lesion classification. Methods: A cohort of 170 patients with 173 breast cancer tumors (132 malignant, 38 benign) was examined with [¹⁸F]FDG-PET/CT. Breast tumors were segmented and radiomic features were extracted following the imaging biomarker standardization initiative (IBSI) guidelines combined with optimized feature extraction. Ensemble learning including five supervised ML algorithms was utilized in a 100-fold Monte Carlo (MC) cross-validation scheme. Data pre-processing methods were incorporated prior to machine learning, including outlier and borderline noisy sample detection, feature selection, and class imbalance correction. Feature importance in each model was assessed by calculating feature occurrence by the R-squared method across MC folds. Results: Cross validation demonstrated high performance of the cancer detection model (80% sensitivity, 78% specificity, 80% accuracy, 0.81 area under the curve (AUC)), and of the triple negative tumor identification model (85% sensitivity, 78% specificity, 82% accuracy, 0.82 AUC). The individual receptor status and luminal A/B subtype models yielded low performance (0.46–0.68 AUC). SUV_max model yielded 0.76 AUC in cancer detection and 0.70 AUC in predicting triple negative subtype. Conclusions: Predictive models based on [¹⁸F]FDG-PET/CT images in combination with advanced data pre-processing steps aid in breast cancer diagnosis and in ML-based prediction of the aggressive triple negative breast cancer subtype.

Specific biological properties of those circulating cancer cells that are the origin of brain metastases (BM) are not well understood. Here, single circulating breast cancer cells were fate-tracked during all steps of the brain metastatic cascade in mice after intracardial injection over weeks. A novel in vivo two-photon microscopy methodology was developed that allowed to determine the specific cellular and molecular features of breast cancer cells that homed in the brain, extravasated, and successfully established a brain macrometastasis. Those BM-initiating breast cancer cells (BMIC) were mainly originating from a slow-cycling subpopulation that included only 16% to 20% of all circulating cancer cells. BMICs showed enrichment of various markers of cellular stemness. As a proof of principle for the principal usefulness of this approach, expression profiling of BMICs versus non-BMICs was performed, which revealed upregulation of NDRG1 in the slow-cycling BMIC subpopulation in one BM model. Here, BM development was completely suppressed when NDRG1 expression was downregulated. In accordance, in primary human breast cancer, NDRG1 expression was heterogeneous, and high NDRG1 expression was associated with shorter metastasis-free survival. In conclusion, our data identify temporary slow-cycling breast cancer cells as the dominant source of brain and other metastases and demonstrates that this can lead to better understanding of BMIC-relevant pathways, including potential new approaches to prevent BM in patients. Implications: Cancer cells responsible for successful brain metastasis outgrowth are slow cycling and harbor stemness features. The molecular characteristics of these metastasis-initiating cells can be studied using intravital microscopy technology.

The role of anthracyclines in the treatment of early breast cancer (EBC) is increasingly being challenged, especially in de-escalation strategies. However, owing to their immunogenic effects, anthracyclines are promising combination partners with immunotherapies. In the randomized phase 2 trial ABCSG-52 (EudraCT no. 2019-002364-27), we investigated epirubicin plus immunotherapy in women with human epidermal growth factor receptor 2 (HER2)-positive EBC. A total of 58 patients were randomized 1:1 to two cycles of a chemotherapy-free induction phase (part 1) of dual HER2 blockade with trastuzumab and pertuzumab (TP) plus the anti-programmed death ligand 1 antibody atezolizumab (TP-A) or TP alone. Thereafter, all patients received four cycles of TP-A in combination with epirubicin (part 2). The primary endpoint, pathological complete response (pCR), was met in 35 patients (60.3%; 95% confidence interval (CI) 47.5% to 71.9%), 19 patients (65.5%) in the TP-A group and 16 patients (55.2%) in the TP group. The residual cancer burden 0/I rate and objective response rate (secondary endpoints) in all patients with evaluable data were 80.0% (n = 44/55; 95% CI 67.6% to 88.4%) and 89.3% (n = 50/56; 95% CI 78.5% to 95.0%), respectively. Grade ≥3 adverse events were reported in 17 patients (29.3%). Based on our findings, we conclude that a neoadjuvant chemotherapy de-escalation immunotherapy regimen with trastuzumab, pertuzumab, atezolizumab and epirubicin is effective and safe in patients with HER2-positive EBC.