More than 70% of breast cancers contain lymphocytic infiltration in the stroma, and preclinical studies suggest that immunoediting and partial control of cancer progression by the local immune microenvironment operate in most breast cancers. Consistent with this hypothesis, a large number of studies demonstrated a favorable prognostic and chemotherapy response predictive role for immune infiltration in breast cancer. The evidence is particularly strong for triple-negative and HER2-positive cancers. The development of clinically effective immune checkpoint inhibitors now provides an opportunity to test the therapeutic potential of augmenting the local antitumor immune response. Several phase I clinical trials using single-agent anti–PD-1 and anti–PD-L1 antibodies demonstrated objective tumor response rates, with remarkably durable responses, in heavily pretreated, metastatic, triple-negative cancers and somewhat lower responses in estrogen receptor–positive cancers. Currently, close to 50 ongoing, or soon to open, clinical trials evaluate the role of this new treatment modality in breast cancer. Clin Cancer Res; 22(9); 2105–10. ©2016 AACR.
The prognostic and predictive roles of the immune microenvironment in breast cancer
The presence of immune cells in the breast cancer microenvironment has long been recognized as a good prognostic indicator (1). More recently, it also became clear that the prevalence of lymphocytic infiltration and its prognostic role varies by molecular subtype. Immune infiltration is most prevalent in triple-negative breast cancers (TNBC) followed by HER2-positive and highly proliferative estrogen receptor (ER)–positive cancers. Immune infiltration is least prominent in low-grade, luminal A type, ER-positive cancers.
In TNBC, high levels of immune infiltration, either measured as tumor-infiltrating lymphocyte (TIL) count or captured by various immune gene signatures, predicts for good survival even in patients not receiving systemic adjuvant therapy, indicating a pure prognostic function (2, 3). Additionally, several neoadjuvant (preoperative) chemotherapy studies demonstrated significantly higher pathologic complete response (pCR) rates among immune-rich compared with immune-poor TNBC, indicating a chemotherapy response predictive role (4–7). Not surprisingly, among TNBC patients who received adjuvant chemotherapy, TIL counts are strongly predictive of cancer-free survival; each 10% increase in TIL count is associated with an 18% reduction of risk of distant recurrence (8, 9). At diagnosis, approximately 5% to 15% of TNBCs are classified as lymphocyte predominant (LPBC), variably defined as either ≥50% or ≥60% lymphocytes in the stroma, another 15% to 20% have no lymphocytic infiltration, whereas the majority (65%–80%) harbor low to moderate level of immune cells (9, 10). Both stromal lymphocytes (residing in the stroma without direct contact with neoplastic cells) and intratumoral lymphocytes (intermingled with and in direct contact with cancer cells) provide prognostic and predictive information, but stromal TILs are more abundant and, therefore, can be quantified more reliably (11). Lymphocyte predominance in residual cancer (≥60% of stromal cells) after neoadjuvant chemotherapy is seen in about 10% of TNBC treated with neoadjuvant chemotherapy and is also associated with excellent survival even in patients who have high-risk pathologic features such as positive nodes or >2-cm residual tumor size (12).
In HER2-positive breast cancer, TIL and immune signatures are also associated with better prognosis with or without systemic adjuvant therapy (13). Similar to TNBC, each 10% increase in TILs is associated with a significantly decreasing risk of distant recurrence in patients receiving adjuvant chemotherapy concomitant with trastuzumab (14). This association with outcome was confirmed in the NeoALTTO, but not in the N9831 trial (15). The expression of CD40 (a costimulatory protein on antigen-presenting cells) related genes was associated with a higher probability of achieving pCR to neoadjuvant trastuzumab containing chemotherapy in HER2-positive cancers (16); however, TILs did not show a linear association with pCR in the NeoALTTO and NeoSphere trials, while it was shown that patients with intermediate TIL infiltration significantly benefited from HER2-targeted therapies (15, 17). NeoSphere also demonstrated a more complex interplay between the immune system and clinical response in the presence of monoclonal antibodies. This trial included a combined treatment arm of trastuzumab and pertuzumab without chemotherapy. Higher expression of several immune genes and metagenes was associated with a higher pCR rate, whereas PDL1 mRNA expression and MHC1 metagenes were associated with resistance (17).
The prognostic and predictive value of immune cells in ER-positive cancers is less extensively studied. However, the available literature suggests that in low-risk ER-positive patients, no such prognostic role is apparent, whereas in highly proliferative ER-positive cancers, immune cells do predict for better prognosis (18).
Overall, a highly consistent body of literature indicates an association between survival and immune cells in the breast cancer microenvironment. These associations are particularly strong in TNBC and HER2-positive cancers, but can also be seen in high-risk ER-positive cancers and raise the possibility that immune cells mediate the observed favorable clinical outcome.
The local immune system in breast cancer
Several outstanding reviews have been published recently on the tumor-promoting and tumor-suppressing role of immune and inflammatory cells (19–21). Cancer tissues are host to multiple different types of immune cells mediating innate and adoptive immunity. However, there are large cancer to cancer differences in the extent and composition of immune cells. In the majority of the literature, T lymphocytes represent the largest proportion of immune cells in breast cancer (70%–80%), followed by B cells (10%–20%), macrophages (5%–10%), natural killer (NK) cells (<5%), and antigen-presenting dendritic cells (4, 10, 22). Each of these main cell types can be subdivided into further functional subtypes [e.g., CD8+ effector T cells, CD4+ T helper cells, and CD4+ regulatory T cells (Th1, Th2, Treg)] and cells can be found in different activity states (e.g., naïve, activated, and memory). The cells form a complex system with dynamic transitions between immune-activating and -suppressing functions. The obligatory, simultaneous presence of multiple different immune cell types in the microenvironment accounts for the highly correlated nature of immune gene expression patterns (5, 17). The strong coexpression of various immune genes explains several seemingly paradoxical associations. For example, high programmed death 1 (PD-1) and programmed death ligand 1 (PD-L1) expression, both negative regulators of the local immune response, is associated with better overall survival and higher pCR rate in TNBC (23, 24). Also, high expression of CTLA-4, a complementary immune checkpoint mechanism to PD-1/scanPD-L1, was associated with benefit from anti–PD-L1 antibody therapy to the same extent as PD-L1 expression in lung cancer (25). The strong correlation between immune marker expression and lymphocyte counts could also limit the independent predictive and prognostic value that immune marker expression adds to TIL counts (5, 7). It is also important to recognize that there are several methodologic concerns regarding detection of PD-L1 (and other immune marker) expression; these include lack of standardized detection methods by IHC, variable cutoffs to determine positivity, and often substantial discordance between mRNA- and IHC-based measurements (7, 24). Furthermore, multiple different cell types (neoplastic and stromal) can express these markers, and expression levels can be upregulated or downregulated in response to an ongoing antitumor immune response, hypoxia, and oncogenic pathway activation (26). These sources of variation contribute to the conflicting results reported in the literature. A central question also remains unanswered: What biologic mechanism underlies the variable levels of immune infiltration and different levels of immune control in different breast cancers?
Results from preclinical experiments and correlative observations in patients suggest that most breast cancers trigger some immune response (22). According the immunoediting hypothesis, the local immune response plays a dual role in cancer progression. On the one hand, it suppresses tumor growth through immune-mediated cell death, which may result in complete elimination of some cancers (before they become detectable) and slow growth or stagnation in others. On the other hand, it also promotes tumor progression by establishing inflammatory conditions that facilitate tumor growth and selecting for tumor cells that survive immunosurveillance (27). An important corollary of this hypothesis is that even during the escape phase, when cancers are clinically apparent, some degree of immune-mediated control is retained, which may account for the better prognosis observed in the immune-rich cancers (Fig. 1).
Another important concept emerging from preclinical models is that tumor response to chemotherapy and trastuzumab is influenced by the host immune system (28). Chemotherapy-induced cellular injury, particularly caused by doxorubicin and cyclophosphamide, can elicit a cytotoxic immune response that partially mediates the clinical response. It has also been suggested that chemotherapy may induce somatic mutations, leading to new antigens that, in turn, elicit immune responses. Chemotherapy, in a drug- and dose-dependent manner, can also stimulate anticancer immune effectors indirectly by inhibiting immunosuppressive regulatory cells (e.g., myeloid-derived suppressor cells and FOXP3+ regulatory T cells; refs. 29, 30). Consistent with these preclinical observations, analysis of TILs in pre- and post-neoadjuvant chemotherapy specimens showed that the development of lymphocytic infiltration during treatment correlates with clinical response (31). The development of clinically effective immune checkpoint inhibitors now provides an opportunity to test the therapeutic potential of augmenting the local antitumor immune response.
On the Horizon
In most analyses, the prognostic and predictive values of TILs and immune gene signatures are independent of histologic grade, tumor size, or nodal status and, therefore, immune markers hold a potential for increasing the predictive accuracy of existing prognostic models (2–6). Furthermore, the reproducibility of stromal TIL counts among pathologists is high; for LPBC category, inter-pathologist agreement ranged from good to moderate (Cohen's κ = 0.60–0.90), and consistency for semiquantitative TIL scoring was excellent (correlation coefficient, 0.97; ref. 4). These results are similar to other broadly accepted measures such as histologic grading or hormone receptor scoring and better than interobserver agreement for Ki-67. An international guideline was recently published to standardize TIL assessment and reporting that sets the stage for introducing this prognostic variable into routine pathology reporting (11). However, no studies have been performed to date that included immune signatures, or TIL counts, in existing multivariate prognostic models such as Adjuvant Online, Nottingham Prognostic Index, 21-gene Recurrence Score, Risk of Recurrence (ROR) score, or others to demonstrate improved prognostic accuracy and, therefore, lymphocyte markers are not yet recommended for routine clinical use.
Immune parameters are also attractive candidates to be predictors of response to immunotherapy. In the simplest form, one could hypothesize that immunotherapy will be most useful for cancers with intermediate TIL counts because LPBCs already have an excellent prognosis and cancers with no lymphocytes have no local immune surveillance to boost. The validity of this hypothesis is being tested in ongoing immunotherapy trials in breast cancer. The therapeutic anti–PD-1/PD-L1 antibodies represent one of the most exciting novel class of therapies due to the remarkably durable responses in melanoma, lung, head and neck, and bladder cancers (32). PD-1 is broadly expressed on several different cells types, including CD4- and CD8-positive T cells, B lymphocytes, NK cells, and T regulatory cells, and therefore it is considered of limited biomarker value. Most studies focused on PD-L1 expression as a potential response marker for PD-1/PD-L1–targeted therapies. In breast cancer, PD-L1 protein expression (i.e., ≥1% of IHC+ cells) is detected in 20% to 30% of cases, primarily seen in TNBC (7, 26, 33, 34), while PD-L1 mRNA expression is identified in substantially larger subsets of breast tumors (16, 23, 24, 33, 34). The correlation between PD-L1 protein and mRNA levels is modest (Spearman rank correlation coefficient, 0.15–0.17; ref. 34). In other cancer types, there is a statistically significant association between PD-L1 expression and the amount of clinical benefit from immune checkpoint therapy; but in each of these studies, response and clinical benefit are also consistently seen in PD-L1–negative cancers (25). Pembrolizumab, an anti–PD-1 antibody, is currently approved by the FDA for the treatment of metastatic non–small cell lung cancer (NSCLC) that expresses PD-L1 protein detected by a companion diagnostic IHC assay (IHC 22C3 pharmDx test made by Dako North America). Interestingly, another anti–PD-1 antibody, nivolumab, is also approved by the FDA for the treatment of metastatic squamous cell lung cancer, but without the requirement of a companion diagnostic test. However, a recent FDA indication extension of nivolumab to patients with NSCLC (including non-squamous cell) endorses the use of a complementary diagnostic assay (IHC 28-8 pharmDx, also made by Dako North America but distinct from IHC 22C3 and applying a different threshold to define positivity) to help guide patient selection for treatment. The test is considered "complementary," not "companion," diagnostic because its use is not mandated prior to administering nivolumab. It is important to note that most PD-L1 expression is detected on stromal cells and not on cancer cells; hence, the often-cited explanation that PD-L1 expression by tumor cells is a main mechanism of immune escape appears simplistic.
Currently, no published data exist on the predictive value of PD-L1 expression for immune checkpoint inhibitor therapy in breast cancer. However, all phase I trials in breast cancer that reported clinical outcome required PD-L1 expression for eligibility.
Immunotherapy of breast cancer
Preliminary results from five phase I clinical trials testing the activity of immune checkpoint inhibitors in metastatic breast cancer are currently available in abstract form. There is also one published phase I trial (n = 26) that reported results for tremelimumab (anti–CTLA-4 antibody) in combination with exemestane in ER-positive metastatic breast cancer and demonstrated stable disease for ≥12 weeks in 11 patients (42%) as the best overall response (35). The KEYNOTE-012 trial assessed the safety and efficacy of single pembrolizumab (10 mg/kg every 2 weeks) in metastatic TNBC that showed ≥1% PD-L1 positivity by IHC. One hundred and eleven patients were screened for PD-L1 expression using the 22C3 antibody and 59% were positive. In the 27 patients who were evaluable for efficacy assessment, the overall response rate was 18.5%, and the median duration of response was not reached at the time of the presentation at the 2014 San Antonio Breast Cancer Symposium (36). The KEYNOTE-028 trial assessed the same drug in metastatic ER-positive breast cancer and also required ≥1% PD-L1 positivity by IHC; the PD-L1 positivity rate was 19%. In the 25 patients who were evaluable for efficacy, the overall response rate was 12% and all 3 responders remained on study treatment for ≥26 weeks at the time of presentation at the 2015 San Antonio Breast Cancer Symposium (37). Adverse events were mostly grade 1–2 and included arthralgia, fatigue, myalgia, and nausea in both studies. Another phase I trial tested the efficacy and safety of the anti–PD-L1 antibody atezolizumab (15 mg/kg, 20 mg/kg, and 1,200 mg fixed dose) in metastatic TNBC and also required ≥5% PD-L1 positivity by IHC using the SP142 antibody (38). Sixty-nine percent of patients tested positive for PD-L1 expression, 21 were evaluable for efficacy, and 19% objective response rate was observed, the 24-week progression-free survival rate was 27%. Adverse events were mostly ≤ grade 2, but 11% of pastients had treatment-related ≥ grade 3 adverse events. The JAVELIN study tested the anti–PD-L1 antibody avelumab (10 mg/kg every 2 weeks) and included all breast cancer subtypes regardless of PD-L1 status (39). In the TNBC cohort (n = 58), the response rate was 8.6%. In the ER-positive/HER2-negative (n = 72) and HER2-positive (n = 26) cohorts, the response rate was 2.8% and 3.8%, respectively. The preliminary results suggested a higher response rate in tumors with PD-L1–positive immune cells [33.3% (4/12) vs. 2.4% (3/124)]. Preliminary results were also reported from a study that combined atezolizumab (anti–PD-L1 antibody) with nab-paclitaxel in metastatic TNBC (n = 24; ref. 40). The combination was well tolerated, and 42% of patients had objective response. Due to these promising early results, there are currently around 50 clinical trials that evaluate this class of drugs in breast cancer in the metastatic, neoadjuvant, and adjuvant treatment settings (Table 1).
Immune checkpoint inhibitors emerged as a new and effective treatment modality for melanoma, NSCLC, and renal cell carcinoma, where these drugs are now approved by the FDA. Clinical trials also show activity in a broad range of solid tumors, including TNBC and, to a lesser extent, ER-positive breast cancer. A large number of clinical trials in the neoadjuvant and metastatic setting are now under way to determine the clinical role of immunotherapies and their combinations in breast cancer.
Disclosure of Potential Conflicts of Interest
L. Pusztai reports receiving research support from AstraZeneca, Genentech, and Merck and speakers bureau honoraria from Merck. G. Bianchini reports receiving speakers bureau honoraria from Roche. No potential conflicts of interest were disclosed by the other authors.
Conception and design: L. Pusztai, T. Karn, G. Bianchini
Development of methodology: L. Pusztai
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Pusztai
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Pusztai, G. Bianchini
Writing, review, and/or revision of the manuscript: L. Pusztai, T. Karn, A. Safonov, M.M. Abu-Khalaf, G. Bianchini
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Pusztai
Study supervision: L. Pusztai
This work was supported in part by grants from the Breast Cancer Research Foundation and Yale Cancer Center Core Grant NIH/NCI P30CA16359 (to L. Pusztai), the H.W. & J. Hector-Stiftung, Mannheim (M67; to T. Karn), and the Associazione Italiana per la Ricerca sul Cancro (AIRC) (MFGA 13428; to G. Bianchini).
- Received November 2, 2015.
- Revision received January 25, 2016.
- Accepted January 26, 2016.
- ©2016 American Association for Cancer Research.