Patient enrollment and sample collection
We retrospectively investigated a total of 8416 breast lesion patients receiving surgeries from January 2008 to December 2012 at the Cancer Hospital, Chinese Academy of Medical Science (CHCAMS), Beijing, China. Patients were eligible for this study if they were: 1) 18 to 55 years old, 2) premenopausal, 3) pathologically diagnosed as invasive lobular carcinoma (ILC) or invasive ductal carcinoma of no special type (IDC, NST) and ER or PR-positive (ER/PR ≥1%), 4) clinical stage Ⅰ to Ⅲ based on the AJCC 8th TNM staging system, and 5) received endocrine therapy for at least five years after surgery with or without systematic adjuvant chemotherapy or radiotherapy. The exclusion criteria included: 1) diagnosis of microinvasive carcinoma, 2) receiving neoadjuvant chemotherapy or radiotherapy, 3) follow-up data unavailable. After applying the inclusion and exclusion criteria, a total of 596 LBC patients were included in this study (Figure 1). We retrospectively reviewed the medical records to collect information on clinicopathologic features (age at diagnosis, tumor size, lymph node status, histological grade, ER, PR and HER-2 status, Ki-67 labeling index, occurrence of peritumoral vascular invasion), type of surgery, and postoperative treatment plan. The follow-up information was obtained through hospital visits or telephone contact with patients or relatives. Until June 11, 2019, 114 patients relapsed and 29 patients died. The median follow-up time was 88.41 months (ranging from 3.27 months to 11.83 years). The study was approved by the CHCAM internal review board on ethical issues.
Breast cancer subtypes
Data on immunohistochemical (IHC) staining of key tumor markers, including ER, PR, HER2, and Ki-67 were collected from patients’ pathological report. IHC was performed for ER (CONFIRMTM, anti-Estrogen Receptor Rabbit Monoclonal Primary Antibody), PR (CONFIRMTM, anti-Progesterone Receptor Rabbit Monoclonal Primary Antibody), Ki-67 (MAIXIN, Ki-67 Rabbit Monoclonal Primary Antibody) and HER2 (VENTANA, anti-HER2/neu Rabbit Monoclonal Primary Antibody) according to the manufacturer’s instructions. All patients with equivocal HER2 status (IHC 2+) were recommended to do fluorescence in situ hybridization (FISH). FISH staining was performed using the PathVysion HER2 DNA Probe Kit (PathVysion, Abbott Molecular, Des Plaines, Illinois, USA) according to the manufacturer’s instructions. All markers were visually assessed by pathologists. For ER and PR, a cut-point of 1% was used to define positive staining, which is consistent with recommendations by international guidelines. For HER2, a score of 3+ on IHC, or amplification on FISH, was considered as positive according to the American Society for Clinical Oncology (ASCO)/College of American Pathologists (CAP) guidelines. Patients with HER2 IHC scores 2+ but for whom information on FISH was not available were classified as HER2-negative. Because Ki-67 threshold of at least 25 % of immunostained cells has been shown to provide the most powerful outcome prognostication, cases with Ki-67 score above this threshold were considered positive(19). Breast cancer molecular subtypes were classified according to the following rules: luminal A (ER positive and/or PR positive, HER2 negative, Ki-67<25%), luminal B/HER2+ (ER positive and/or PR positive, and HER2 positive) and luminal B/HER2- (ER positive and/or PR positive, Ki-67≥25%, and HER2 negative). If the Ki-67 information is missing, tumor grade was used for defining luminal A (grade 1 or 2) and luminal B/HER2- (grade 3) subtypes.
Stromal TILs assessment
Whole sections of hematoxilyn and eosin (H&E) stained slides were used to evaluate stromal TILs (sTILs), strictly following the criteria proposed by the International TIL Working Group (20). Briefly, the percentage of all mononuclear cells (including lymphocytes and plasma cells) in the stromal compartment within the border of the invasive tumor was visually evaluated. We also developed a supervised machine-learning algorithm for the unbiased detection of TILs based on cell contexture, size and shape within well-defined regions of the stroma (12).
Immunohistochemistry (IHC) for CD8+ T-cells
Of these 596 patients, 160 cases were selected to perform CD8 immunohistochemical (IHC) staining on formalin-fixed paraffin-embedded (FFPE) whole tissue sections. These cases were selected based on TIL quantity measured by the algorithm, with 60 and 100 cases from the lowest and highest quartile, respectively. The tumor-infiltrating CD8+ T-cells were evaluated by IHC using a rabbit monoclonal primary antibody against CD8 (ZA-0508, ZSGB-BIO). Automated IHC was performed on 4-μm-thick sections using the Ventana Benchmark ULTRA automated slide processing system according to the manufacturer’s instructions.
All cells stained positive in the stromal compartment and tumor nests within the borders of the invasive tumor were evaluated and reported as a percentage value. CD8+ T-cells were classified as ‘intratumoral CD8+ T-cell’ (iCD8+, the percentage of the area of CD8+ T-cell infiltration within the tumor area if they were in direct contact with tumor cells) and ‘stromal CD8+ T-cell’ (sCD8+, the percentage of the area of CD8+ T-cell infiltration within the tumor stroma if they were in the stromal compartment but not in direct contact with tumor cells).
Stromal TILs and CD8+ T-cells outside of the tumor border, around DCIS and normal breast tissue, or in areas of necrosis were not included in the scoring. The visual evaluation of stromal TILs and CD8+ T-cells was performed by an expert pathologist (JZ) without knowledge of the clinical information. For cases with significant region variations in lymphocyte distribution, we evaluated each region separately and calculated a weighted average of TIL values across different regions according to the size of the region (Figure 2). We used the third quartile as the cutoff point to define high and low TIL levels, specifically, 10% for sTILs, 5% for sCD8 and 1% for iCD8, which were consistent with previous studies (7, 21). We also defined the lymphocyte-predominant breast cancer (LPBC) based on sTILs ≥50%, similar to what was previously reported (18).
Statistical Analysis
Two clinical outcomes were analyzed in this study. Disease-free survival (DFS) was defined as the interval between the first operation time and the date of first relapse (local, regional, contralateral, or metastatic) or death from any reason, and overall survival (OS) was defined as the time from surgery to the date of death from any reason. Patients who were alive (for OS) and disease free (for DFS) were censored at the date of last contact.
Statistical analyses were conducted using SPSS version 25.0 and R version 3.6.1. The differences between categorical variables were evaluated using Chi-square or Fisher's exact test. The differences between two continuous variables were evaluated using Mann-Whitney Test. Univariable and multivariable Cox proportional hazards regression models were used to assess differences in DFS and OS across different groups defined by TILs. The multivariable model contained variables that remained significant in the univariable Cox-regression model after backward elimination (significance level of 0.1). For visualization purposes, Kaplan-Meier plots were used to produce DFS and OS curves stratified by sTILs, sCD8, and iCD8 as binary variables. The log-rank test was used to compare the two groups. All reported p values were two-tailed, and for all analyses, p<0.05 was considered statistically significant.