Introduction
Immune response may play an important role in cancer progression. Tumor-infiltrating lymphocytes (TILs) reflect a local immune response and could be a key mechanism in controlling tumor progression [
1,
2]. A number of studies demonstrate that TILs are associated with clinical outcome in patients with carcinoma and melanoma [
3‐
8]. TILs have been found to be mainly T lymphocytes, and the majority express a cytotoxic effector phenotype (CD8
+) [
9‐
11]. CD8
+ T cell-mediated type 1 immune responses can enhance the accumulation of distinct endogenous CD8
+ and CD4
+ T cells and facilitate their antitumor function within the tumor microenvironment [
12,
13]. Studies in ovarian carcinomas and colon cancer show that high levels of CD8
+ lymphocyte infiltration are associated with better prognosis in these diseases [
3,
14]. In breast cancer, some studies have reported that inflammation and cytotoxic lymphocyte infiltration are associated with better survival [
15‐
17]. In contrast, other groups have reported that high numbers of TILs are related to worse overall survival [
18,
19], whereas still other studies did not find any significant association of TILs with patient outcome [
20,
21]. A recent publication reported that a high ratio of CD8
+ TILs to FOXP3
+ regulatory T cells had a significant relationship to improved patient survival in breast cancer [
22]. Two other studies have tested larger series: one study used a retrospective cohort of 1,334 patients with primary breast cancer diagnosed from 1987 to 1998 in the UK and showed that total CD8
+ TILs were independently associated with better survival in breast cancer [
23], whereas another study with 1,953 breast cancer cases treated in the University Hospital Basel in Switzerland between 1985 and 1996 demonstrated that the independent favorable prognostic effect of total CD8
+ TILs was observed only in those with estrogen receptor-negative (ER
-) tumors (whereas, in univariate analyses, CD8
+ TILs had an unfavorable effect on outcome in ER-positive (ER
+) breast cancers) [
24]. Thus, the extent to which TILs contribute to tumor progression and clinical outcome in breast cancer has remained controversial, possibly because the effect is limited to certain subgroups of patients.
Breast cancer is a heterogeneous disease composed of different intrinsic subtypes, each with distinctive biological and prognostic behaviors and responses to therapy. Although the introduction of adjuvant systemic therapy (AST) has led to a significant reduction in breast cancer mortality, many patients do not benefit. Gene expression studies suggest that predictive indicators should be developed for different breast cancer subtypes [
25,
26]. The interaction between immune response, intrinsic subtype, and treatment strategy all likely contribute to the outcome of the disease. The development of molecular diagnostic techniques has facilitated a better understanding of the heterogeneity of breast cancer and opened up the possibility of more personalized therapy [
27,
28]. Hormone receptor status and human epidermal growth factor receptor-2 (HER2) molecular status are currently used to guide AST strategies for the luminal and HER2
+ intrinsic subgroups, but no targeted therapy for the basal-like subgroup is currently available. Basal-like breast cancer comprises about 15% of all invasive breast cancers and is likely to be high-grade, occur in young women, and have an aggressive clinical course [
29]. Although a majority of basal-like tumors carry a clinical triple-negative phenotype (TNP) (ER
-, progesterone receptor-negative (PR
-), and HER2
-), they are not synonymous [
30], and triple-negative breast cancers include many cases that lack the expression of basal markers - the so-called 'five-marker negative phenotype' (5NP): ER
-, PR
-, HER2
-, epidermal growth factor receptor-negative (EGFR
-), and cytokeratin (CK) 5/6
- - which have been shown to have significantly better outcomes than core basal cases [
31,
32]. Gene expression profiling data suggest that medullary breast tumors (a rare histological subtype with a prominent lymphocytic reaction and a good prognosis) are a specific subgroup within the basal-like class, indicating that the overall poor survival of basal-like breast cancer might be mitigated in cases in which there is a strong immune response [
33‐
35]. On the other hand, a separate body of research has highlighted that recruitment of chronic inflammatory cells, including macrophages, can actually promote cancer progression [
36]. Different types of immune response in different subtypes of breast cancer might explain apparently contradictory results. However, to date, no large immunohistochemistry study has explored the prognostic effect of an immune response in breast cancer stratified by breast cancer intrinsic subtype.
Therefore, there is a clear need for studies with sufficient power for subgroup analysis, employing validated measurements of immune response, to evaluate the significance of TILs in breast tumors. The aim of this study was to examine the prognostic significance of CD8+ TILs in different breast cancer intrinsic subtypes in a large population-based cohort with long-term follow-up. Our hypothesis was that CD8+ lymphocyte infiltration has distinct prognostic effects in different intrinsic molecular subtypes of breast cancer.
Discussion
The prognostic significance of TILs in breast cancer has been debated, but no consistent conclusion has yet been drawn. We implemented this study, using a particularly large, well-annotated cohort comprising nearly 4,000 patients, in an attempt to definitively assess the clinical implication of TILs in breast cancer. In addition to addressing the question of whether immune response (as measured by CD8+ TILs) has a prognostic role in breast cancer in general, we examined the effect of TILs in the major breast cancer intrinsic biological subtypes. To our knowledge, this is the first study sufficiently powered for multivariate analysis to investigate the association of CD8+ TILs with patient survival within the breast cancer intrinsic subtypes. Our results demonstrate that the presence of iTILs is independently associated with a significantly superior outcome in women with diagnosed core basal tumors. Although the presence of CD8+ iTILs is also an independent prognostic indicator for improved patient survival in triple-negative breast cancers, this favorable prognostic effect cannot be detected among those lacking expression of basal biomarkers (5NP). In the core basal subgroup, patients having tumors with CD8+ iTILs survived an average of 3.5 years longer than did patients with basal tumors lacking evidence of a CD8+ iTIL immune response.
Breast cancer is both clinically and molecularly heterogeneous and is, in practice, stratified by hormonal receptors (ER and PR), by HER2 status, and, increasingly, by expression of other biomarkers such as Ki67 or by gene expression profiling methodologies. Dissecting the heterogeneity of breast cancer is critically important for understanding the underlying mechanisms of the disease and for identifying subpopulations that are most likely to respond to particular therapies [
43]. In general, ER
- breast cancers have a worse prognosis than those that are ER
+, but not all patients with ER
- breast cancer have poor survival. Teschendorf and colleagues [
44] applied an integrative analysis of three gene expression datasets to assess the prognostic value of molecular signatures and found that most prognostic markers of better prognosis in ER
- breast cancer are associated with the activation of immune response pathways. Furthermore, a seven-gene immune response classifier was constructed and showed significant good prognostic value in patients with ER
- breast cancer [
45]. Meta-analytic studies of clinical and gene expression data have demonstrated that immune response is significantly associated with prognosis in breast cancer [
46], primarily in rapidly proliferating [
47] and ER
- [
48,
49] subgroups. Results from some studies indicate that TILs could be a protective factor reducing the likelihood of distant metastasis in patients with triple-negative breast tumors [
50] and among those with medullary carcinoma [
17]. Moreover, two recently published gene expression profiling studies demonstrated that effective immune (particularly cytotoxic T-cell) response plays a favorable prognostic role in basal breast cancer subgroups [
51,
52]. In our study, the multivariate analysis clearly demonstrates that the presence of CD8
+ iTILs has a different prognostic value in breast cancer with different intrinsic biological subtypes. Even among the triple-negative cases, immune response has different meanings in core basal versus 'five negative' phenotypes. Evidence from previous studies has shown that core basal-like tumors are associated with a poorer prognosis and appear biologically different from 5NP tumors [
31,
32]. Our results suggest that local immune response characterized by CD8
+ lymphocyte infiltration might be considered an important factor differentiating the core basal from 5NP breast tumors within the class of triple-negative breast cancers.
Tumor-infiltrating lymphocytes and macrophages are thought to be molecular determinants of clinical outcome in breast cancer. Although cytotoxic T lymphocytes and natural killer cells have been found to have antitumor activity, some lymphocytes such as B cells exhibit bipolar roles in breast cancer development. Distinct cell-mediated immune responses also play antagonistic roles in disease prognosis. T helper cell 1 (Th1)-mediated immune response pathways are considered to have an inhibitory effect, whereas Th2 immune response pathways may promote development and metastasis of breast cancer. It has been found that CD4
+ T lymphocytes can promote metastasis by activating the EGFR signaling pathway in a Th2-type tumor microenvironment [
53]. Identification of interactions between immune response and other molecular pathways may define novel prognostic subtypes. In ER
- breast cancer, those characterized with high expression of EGFR and low expression of Th1-mediated pathway-related markers such as interleukin-12 and interferon-gamma were found to have a poor prognosis [
54]. TILs in the tumor microenvironment are predominantly CD8
+ T cells [
55,
56], which are considered to be the effector cells in Th1 antitumor immune responses. CD8
+ T cells produce interferon-gamma through interaction with tumor-related antigens, potential leading to tumoricidal activity by induction of apoptosis or macrophage tumor killing activity or both [
57]. Studies indicate that tumor-specific or even non-cancer-specific antigens such as p53 and β-actin are common targets of cytotoxic T lymphocytes and can induce immunological and clinical effects in patients with breast cancer [
58‐
60]. Findings from our study suggest that core basal-like breast cancer is more immunogenic than other intrinsic subgroups, as measured by CD8
+ T-cell infiltration. Tumors of this subtype have a high expression of basal markers, some of which (such as EGFR) may interact with T cell-mediated immune response to affect clinical outcome in breast cancer. We would suggest a hypothesis that certain 'basal proteins' expressed on the cell surface can be recognized as tumor antigens and that the consequent induction of adaptive basal marker-specific immunity can enhance the local Th1-mediated antitumor immune response in these breast cancers. The absence of these surface markers in 5NP breast cancers could underpin the observed difference in prognostic significance of TILs in core basal compared with 5NP breast cancers.
Recent studies have suggested that a pre-existing immune response can strengthen the effect of conventional chemotherapy [
61,
62], enhancing destruction of tumor cells [
63], and this favorable effect could become stronger in patients with highly immunogenic tumors, perhaps including the core basal group. Basal-like breast cancers have distinctive survival patterns, many relapses and deaths during the first 5 years after diagnosis, but fewer events after this period [
32], indicating that basal-like breast cancers encompass both poor and good prognostic subgroups responding variably to conventional therapies. In our cohort, systemic treatment decisions were not randomized, making outcomes stratified by treatment difficult to interpret; nevertheless, an exploratory analysis suggests that pre-treatment CD8
+ lymphocyte infiltration is an independent favorable predictive indicator of good outcomes in basal-like cases treated with chemotherapy (HR = 0.29, 95% CI = 0.16 to 0.55,
P < 0.001,
n = 107) (Table S4 of Additional file
4). Our results indicate that efforts toward developing immuno-stimulative therapies might be best directed to the core basal group. The recognition of tumor-associated antigens by CD8
+ cells is a significant contributor to the detection and ultimate destruction of tumor cells [
64]. Basal-like breast cancer could be particularly suitable for targeted immunotherapy. The lack of success of prior attempts at immunotherapy for breast cancer may be attributable, in part, to the lack of focus on appropriate breast cancer subtypes. A better understanding of the interaction between immune response, intrinsic subtype, AST, and patient outcome is critical to more effective and targeted clinical management for patients with breast cancer, especially those with basal-like breast tumors.
Studies on TILs in breast cancer have come to inconsistent conclusions. We believe that one of the underlying reasons could be inconsistency in defining and measuring TILs. Some research considered only the presence of peritumoral stromal lymphocytes [
65,
66], and many considered all T lymphocytes (which might include larger numbers of regulatory T cells that could in some cases reflect immune suppression instead of activation). In our study, specific immunohistochemistry was used with a mouse monoclonal anti-human CD8 antibody to detect cytotoxic effector CD8
+ TILs in intratumoral and stromal locations for each tumor tissue core. We evaluated the reliability of repeated scoring by the same scorer and between different scorers, and it was demonstrated that our visual CD8
+ TILs scoring was highly reliable (Figure S4 of Additional file
6). Analyses with intratumoral, stromal, and total CD8
+ TILs were conducted, and consistent results were obtained. We also did analyses using relapse-free survival as an outcome and obtained results similar to those using BCSS as the outcome (Figures S5 and S6 of Additional files
7 and
8 and Tables S5 to S7 of Additional file
4). Thus, we are confident that the identification and quantification of TILs and the assessment of the association of TILs with clinical outcome in breast cancer are reliable and valid in this study. One potential limitation of our methods is that TMAs may not adequately represent breast tumor heterogeneity. Several studies nevertheless have shown that findings from TMAs are consistent with those from full-face tissue sections [
67,
68]. Although we observed a trend to a favorable prognostic effect of CD8 TILs in the HER2
+/ER
- subgroup (and this trend is consistent with a gene expression study [
69]), the effect was not statistically significant in our univariate or multivariate analyses. Research with more power particularly for this subgroup needs to be done to draw a more definitive conclusion among HER2
+ cases. We were not able to measure changes in immune response induced by chemotherapy, as all of the tissue samples were collected before patients received systemic therapy. Further studies would need to be conducted to assess the interaction of TILs with chemotherapy, ideally in randomized trials.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
SLi coordinated the study, analyzed data, and drafted the manuscript. JL advised on scoring and edited the manuscript. SLe assisted with statistical analyses. DG generated primary data. WDF provided the idea for the study, helped with data analysis, and edited the manuscript. TON organized the study, directed data generation and analysis, and edited the manuscript. All authors read and approved the final manuscript.