Active behavior of triple-negative breast cancer with adipose tissue invasion: a single center and retrospective review

Digital image analysis and machine learning-assisted prediction of neoadjuvant chemotherapy response in triple-negative breast cancer

BACKGROUND

Pathological complete response (pCR) is associated with favorable prognosis in patients with triple-negative breast cancer (TNBC). However, only 30-40% of TNBC patients treated with neoadjuvant chemotherapy (NAC) show pCR, while the remaining 60-70% show residual disease (RD). The role of the tumor microenvironment in NAC response in patients with TNBC remains unclear. In this study, we developed a machine learning-based two-step pipeline to distinguish between various histological components in hematoxylin and eosin (H&E)-stained whole slide images (WSIs) of TNBC tissue biopsies and to identify histological features that can predict NAC response.

METHODS

H&E-stained WSIs of treatment-naïve biopsies from 85 patients (51 with pCR and 34 with RD) of the model development cohort and 79 patients (41 with pCR and 38 with RD) of the validation cohort were separated through a stratified eightfold cross-validation strategy for the first step and leave-one-out cross-validation strategy for the second step. A tile-level histology label prediction pipeline and four machine-learning classifiers were used to analyze 468,043 tiles of WSIs. The best-trained classifier used 55 texture features from each tile to produce a probability profile during testing. The predicted histology classes were used to generate a histology classification map of the spatial distributions of different tissue regions. A patient-level NAC response prediction pipeline was trained with features derived from paired histology classification maps. The top graph-based features capturing the relevant spatial information across the different histological classes were provided to the radial basis function kernel support vector machine (rbfSVM) classifier for NAC treatment response prediction.

RESULTS

The tile-level prediction pipeline achieved 86.72% accuracy for histology class classification, while the patient-level pipeline achieved 83.53% NAC response (pCR vs. RD) prediction accuracy of the model development cohort. The model was validated with an independent cohort with tile histology validation accuracy of 83.59% and NAC prediction accuracy of 81.01%. The histological class pairs with the strongest NAC response predictive ability were tumor and tumor tumor-infiltrating lymphocytes for pCR and microvessel density and polyploid giant cancer cells for RD.

CONCLUSION

Our machine learning pipeline can robustly identify clinically relevant histological classes that predict NAC response in TNBC patients and may help guide patient selection for NAC treatment.

Age and obesity-driven changes in the extracellular matrix of the primary tumor and metastatic site influence tumor invasion and metastatic outgrowth

Younger age and obesity increase the incidence and metastasis of triple-negative breast cancer (TNBC), an aggressive subtype of breast cancer. The extracellular matrix (ECM) promotes tumor invasion and metastasis. We characterized the effect of age and obesity on the ECM of mammary fat pads, lungs, and liver using a diet-induced obesity (DIO) model. At 4 week intervals, we either injected the mammary fat pads with allograft tumor cells to characterize tumor growth and metastasis or isolated the mammary fat pads and livers to characterize the ECM. Age had no effect on tumor growth but increased lung and liver metastasis after 16 weeks. Obesity increased tumor growth starting at 12 weeks, increased liver metastasis only at 4 weeks, and weight gain correlated to increased lung but not liver metastasis. Utilizing whole decellularized ECM coupled with proteomics, we found that early stages of obesity were sufficient to induce changes in the ECM composition and invasive potential of mammary fat pads with increased abundance of pro-invasive ECM proteins Collagen IV and Collagen VI. We identified cells of stromal vascular fraction and adipose stem and progenitor cells as primarily responsible for secreting Collagen IV and VI, not adipocytes. We characterized the changes in ECM in the lungs and liver, and determined that older age decreases the metastatic potential of lung and liver ECM while later-stage obesity increases the metastatic potential. These data implicate ECM changes in the primary tumor and metastatic microenvironment as mechanisms by which age and obesity contribute to breast cancer progression.

Digital image analysis and machine learning-assisted prediction of neoadjuvant chemotherapy response in triple-negative breast cancer

Background

Pathological complete response (pCR) is associated with favorable prognosis in patients with triple-negative breast cancer (TNBC). However, only 30-40% of TNBC patients treated with neoadjuvant chemotherapy (NAC) show pCR, while the remaining 60-70% show residual disease (RD). The role of the tumor microenvironment (TME) in NAC response in patients with TNBC remains unclear. In this study, we developed a machine learning-based two-step pipeline to distinguish between various histological components in hematoxylin and eosin (H&E)-stained whole slide images (WSIs) of TNBC tissue biopsies and to identify histological features that can predict NAC response.

Methods

H&E-stained WSIs of treatment-naïve biopsies from 85 patients (51 with pCR and 34 with RD) were separated through a stratified 8-fold cross validation strategy for the first step and leave one out cross validation strategy for the second step. A tile-level histology label prediction pipeline and four machine learning classifiers were used to analyze 468,043 tiles of WSIs. The best-trained classifier used 55 texture features from each tile to produce a probability profile during testing. The predicted histology classes were used to generate a histology classification map of the spatial distributions of different tissue regions. A patient-level NAC response prediction pipeline was trained with features derived from paired histology classification maps. The top graph-based features capturing the relevant spatial information across the different histological classes were provided to the radial basis function kernel support vector machine (rbfSVM) classifier for NAC treatment response prediction.

Results

The tile-level prediction pipeline achieved 86.72% accuracy for histology class classification, while the patient-level pipeline achieved 83.53% NAC response (pCR vs. RD) prediction accuracy. The histological class pairs with the strongest NAC response predictive ability were tumor and tumor tumor-infiltrating lymphocytes for pCR and microvessel density and polyploid giant cancer cells for RD.

Conclusion

Our machine learning pipeline can robustly identify clinically relevant histological classes that predict NAC response in TNBC patients and may help guide patient selection for NAC treatment.

Association of quantitative analysis of intratumoral reduced E-cadherin expression with lymph node metastasis and prognosis in patients with breast cancer

Loss of E-cadherin expression is a poor prognostic factor in patients with breast cancer. Breast cancer cells co-cultured with adipocytes reportedly promote E-cadherin attenuation and tumor progression. The current study aimed to investigate the association of reduced E-cadherin expression with adipose tissue invasion (ATI) and prognosis in breast cancer. Surgical specimens were collected from 188 women with invasive ductal carcinoma of the breast who had undergone surgery without neoadjuvant treatment. We compared E-cadherin expression in ATI and invasive front (IF) using immunohistochemistry with ImageJ. Reduced E-cadherin expression was detected not only in the ATI area but also in the IF, and the degree of reduced E-cadherin expression was positively correlated with both areas. In patients with lymph node metastasis compared to those without, E-cadherin expression was reduced and this reduction was associated with poor recurrence-free survival. We concluded that E-cadherin expression is reduced not only at the ATI area but also at the IF of the tumor. Reduced E-cadherin expression is a clear prognostic factor for breast cancer. Hence, future research is warranted for establishing an objective and quantitative E-cadherin staining assay that will allow clinical use of E-cadherin as a prognostic factor.

Clinical Significance of Peritumoral Adipose Tissue PET/CT Imaging Features for Predicting Axillary Lymph Node Metastasis in Patients with Breast Cancer

We investigated whether textural parameters of peritumoral breast adipose tissue (AT) based on F-18 fluorodeoxyglucose (FDG) PET/CT could predict axillary lymph node metastasis in patients with breast cancer. A total of 326 breast cancer patients with preoperative FDG PET/CT were retrospectively enrolled. PET/CT images were visually assessed and the maximum FDG uptake of axillary lymph nodes (LN SUVmax) was measured. From peritumoral breast AT, 38 textural features of PET imaging were extracted. The diagnostic ability of PET based on visual analysis, LN SUVmax, and textural features of peritumoral breast AT for predicting axillary lymph node metastasis were assessed using the area under the receiver operating characteristic curve (AUC) values. Among the 38 peritumoral breast AT textural features, grey-level co-occurrence matrix (GLCM) entropy showed the highest AUC value (0.830) for predicting axillary lymph node metastasis. The value of GLCM entropy was higher than that of visual analysis (0.739; p < 0.05) and the AUC value was comparable to that of LN SUVmax (0.793; p > 0.05). In the subgroup analysis of patients with negative findings on visual analysis, GLCM entropy still showed a high diagnostic ability (AUC: 0.759) in predicting lymph node metastasis. The findings suggest a potential diagnostic role of PET/CT imaging features of peritumoral breast AT in predicting axillary lymph node metastasis in patients with breast cancer.