Tumor-Infiltrating Lymphocytes: A Predictive and Prognostic Biomarker in Neoadjuvant-Treated HER2-Positive Breast Cancer Free

Tumor-infiltrating lymphocytes (TIL) are a specific histologic feature of various cancer types and vary clearly in number between individual cases. In breast cancer, higher levels of TILs are found in highly proliferative tumors like triple-negative breast carcinomas (TNBC) and HER2-positive carcinomas (9, 10). There are several recent studies, showing a prognostic and predictive impact of TILs. In the adjuvant therapy setting, Loi and colleagues demonstrated that in TNBC for every 10% increase in TILs, there is a 15% to 20% increase in disease-free survival (DFS) and overall survival (OS; ref. 11). These data were validated in an independent data set by Adams and colleagues corroborating the strong prognostic value of TILs in the adjuvant therapy of TNBC (12). In HER2-positive breast cancer, higher TIL levels at time of diagnosis resulted in higher response to adjuvant trastuzumab treatment (11). In the neoadjuvant therapy setting, TILs are a predictor for an increased pathologic complete response (pCR) after anthracycline- and taxane-based chemotherapy in HER2-positive and TNBC (13).

HER2 is a tyrosine kinase receptor, activated by homo- or heterodimerization, therefore promoting proliferation and cell survival by prosurvival pathways. HER2 protein overexpression and gene amplification are detectable in about 15% of breast carcinomas and are associated with a more aggressive phenotype and poor prognosis (14). The addition of the monoclonal antibody trastuzumab to the adjuvant treatment of HER2-positive breast cancer has shown a survival advantage in early and metastatic disease (15). Trastuzumab has several mechanisms of action. First, it binds to the extracellular domain of the HER2 receptor and prevents receptor dimerization, inhibiting signaling through the downstream cascade. It increases endocytotic deconstruction of the receptor and enhances tumor cell destruction by activating the immune system (15). Preclinical models suggest a strong effect via antibody-dependent cellular cytotoxicity (ADCC), by which binding of an antibody to a cell induces immune effector cells to kill the antigen-expressing cells (16). This is supported by the fact that animals deficient for Fcγ receptors do not show any response to trastuzumab (17). Fcγ receptors are crucial for ADCC because the interaction of antibodies with Fcγ receptor on innate effector cells like natural killer cells, neutrophils, and γδT cells (18) regulates the immune response. However, besides the innate immune system, adaptive immune response seems also to be needed for a sufficient therapeutic result. The effect of trastuzumab was significantly reduced, followed by a rapid tumor relapse, when treating B- and T-cell–deficient mice or wild-type mice depleted of CD8-positive T cells (17). Furthermore, stimulation of the immune system by anti–PD-1 or anti–CD137 antibodies strengthens the immune effect of trastuzumab and enhances the therapeutic effect (19). Overall, these data corroborate that the immune system plays an essential role for the trastuzumab effect.

Lapatinib is a small-molecule, dual tyrosine kinase inhibitor of HER2 and EGFR (also known as HER1). It binds to the ATP-binding site of the protein kinase domain, and therefore prevents self-phosphorylation and subsequent activation of the signaling cascade (20, 21). Preclinical studies in the MMTV-neu mouse model (mouse mammary tumor virus) also suggest an antitumoral immunomodulation of lapatinib by attracting CD4- and CD8-positive IFNу-producing T cells (22). In vivo depletion of CD8-positive cells in these mice significantly reduced the antitumoral effectiveness of lapatinib.

Pretherapeutic core biopsies of untreated breast cancer samples are a perfect tool to evaluate prognostic and predictive markers of neoadjuvant therapy regimen. In this study, we retrospectively validated TILs in 498 centrally confirmed HER2-positive breast cancer samples of the neoadjuvant GeparQuattro and GeparQuinto trial. In the GeparQuattro study, all patients with HER2-positive breast cancer received trastuzumab (3), whereas in the GeparQuinto study, patients were randomly assigned to receive either trastuzumab or lapatinib as part of the neoadjuvant therapy (7).

We re-evaluated TILs as an independent predictor for response to neoadjuvant chemotherapy. We investigated a potential predictive impact of TILs on different anti-HER2 blocking strategies to evaluate if TILs could be of more importance in antibody-based anti-HER2 therapy (trastuzumab) than in inhibition of HER2 by small molecules (lapatinib). In addition, we tested whether higher levels of TILs are also associated with better outcome.

The GeparQuattro trial (G4; NCT00288002) was a prospective, randomized, multicenter phase III study. In total, the trial recruited 1,509 patients between August 2005 and November 2006, and 1,495 patients were assigned for initial treatment. Patients received four cycles of epirubicin/cyclophosphamide (EC) followed by four cycles of docetaxel with or without capecitabine, whereas capecitabine was administered concomitantly or in sequence. In the HER2-positive setting (n = 445, based on local pathology report), patients were treated with trastuzumab (3).

The GeparQuinto trial (G5; NCT00567554) was a prospective, randomized, open-label, multicenter phase III clinical trial to evaluate the integration of targeted therapies to neoadjuvant breast cancer chemotherapy approaches. Patients were recruited from November 2007 to July 2011. Chemotherapy comprised four cycles of EC followed by four cycles of docetaxel. In the HER2-positive setting, 620 patients were enrolled and randomized to receive either trastuzumab or lapatinib. Finally, 307 patients were randomly assigned to chemotherapy with trastuzumab, and 308 patients to chemotherapy with lapatinib (7). Trial inclusion was based on local pathology results, and participation in biomaterial collection was not mandatory for the patients.

Survival data were available from all patients. DFS (invasive) was calculated as the time from study registration to any invasive recurrence (local, contralateral, distant), any second invasive cancer or death of any cause.

Written informed consent for clinical trial participation and use of biomaterials was obtained from all patients. Ethic committee approval was obtained from all centers participating in the clinical study and from the Institutional Review Board of the Charité hospital. This study is reported according to the Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK) criteria (23).

All clinicopathologic data were extracted from the clinical study database and refer to the local results, except HER2. HER2 status was reviewed centrally. pCR was defined as the absence of invasive tumor cells in breast and lymph nodes (ypT0is/ypN0).

In summary, following inclusion criteria were used: (i) Patients treated with anti-HER2 therapy on the HER2-positive arms of G4 and G5; (ii) Available pretherapeutic formalin-fixed paraffin-embedded core biopsy for reviewing HER2-status; (iii) Centrally confirmed HER2 positivity; (iv) Available tumor tissue for evaluating TILs; and (v) Available data on pCR (Fig. 1).

Tumor sections were hematoxylin and eosin stained for histologic evaluation. HER2 expression was determined by immunohistochemistry using a monoclonal anti-HER2 antibody (clone 4B5; Ventana Medical Systems). Analysis of HER2-gene amplification was performed by silver-enhanced in situ hybridization (SISH; ultra View SISH Detection kit; Ventana Medical Systems). Both analyses were run on the Ventana BenchMark ultra-automated staining system.

HER2 status was determined according to the consensus panel recommendation on HER2 testing in breast cancer (24): Scores 3+ were reported as positive, scores 0/1+ as negative. Tumors with scores 2+ were reported as equivocal and further tested by SISH. An average number of 20 tumor cells per sample was evaluated. A HER2/CEN17 ratio ≥2 was considered positive, and a ratio <2 negative (24).

Stromal lymphocytes were evaluated according to the current recommendations (25) by one experienced pathologist (B. Ingold Heppner) who has successfully participated in the quality assurance TILs ring trial. Critical cases were evaluated by a second pathologist (C. Denkert). Breast cancer samples with ≥ 60% stromal TILs were defined as lymphocyte-predominant breast cancer (LPBC).

Statistical analysis was performed using SPSS version 22 (IBM Corp.). pCR rates were reported in subgroups defined by binary parameters (LPBC and clinical/pathologic features) and compared between subgroups using the Fisher exact test. The probability of pCR as a function of TILs was determined by univariate logistic regression analysis. The multivariate logistic regression was used to adjust analysis for known clinical parameters influencing pCR, such as local tumor extent (T-Stage), nodal status (N-stage), tumor grading, hormone receptor status, and therapy regimen. For survival analyses, the Kaplan–Meier method was applied to estimate time to event outcome parameters, and different groups were compared using the log-rank test. To calculate hazard ratios (HR), univariate Cox proportional hazard models were used. Multivariate Cox models were applied to adjust analysis for the relevant baseline characteristics as predefined. Due to the small number of events (n = 48), data on OS are not presented. P values less or equal to 0.05 were considered to be statistically significant.

A total of 498 patients fulfilled the inclusion criteria, 178 patients of the G4 trial and 320 of the G5 trial. A total of 104 cases were classified as LPBC (20.9%; G4: 47 patients; G5: 57 patients). A total of 219 cases revealed pCR (44.0%), 85 patients in G4 (47.8%) and 134 patients in G5 (41.9%). For detailed patient and tumor characteristics, see Supplementary Table S1.

TILs were reported either in a binary fashion (non-LPBC- vs. LPBC phenotype) or as a continuous parameter per 10% increase.

LPBC phenotype was strongly associated with pCR in the complete cohort as well as in the G4 subgroup. Although this was not statistically significant, LPBC tumors in G5 had a tendency for increased pCR rate (Fig. 2).

In univariate analysis, TILs are an independent predictor for pCR in the complete cohort. These statistically significant effects were also seen in the subgroup analysis of G4 but not in the G5 cohort (Table 1).

In multivariate analysis adjusted for age, tumor stage, nodal stage, hormone receptor status, grading, and anti-HER2 therapy, TILs 10% and LPBC phenotype were independent predictors for pCR in the complete cohort. However, in the subgroup of G5, this effect was not detectable (Table 2).

An additional independent predictor for pCR in multivariate analysis was a negative hormone receptor status, in the complete cohort and both subgroups (Table 2).

Furthermore, TILs were significantly associated with high-grade tumors, hormone receptor negativity, and lymph node metastases (Supplementary Table S2).

In our cohort, 340 (68.3%) patients were treated with trastuzumab (G4: 178; G5: 162) and 158 (31.7%) with lapatinib. Overall, anti-HER2 therapy with trastuzumab resulted in higher pCR rates than with lapatinib (Fig. 3). In both groups, the proportion of LPBC tumors was similar (trastuzumab group: n = 73, 21.5%; lapatinib group: n = 31, 19.6%). In the trastuzumab group, LPBC phenotype was significantly associated with pCR [61.6% vs. 43.8%, P = 0.008; OR: 2.06; 95% confidence interval (CI), 1.21–3.5]. In multivariate analysis adjusted for hormone receptor status, age, histologic subtype, grade, tumor- and nodal stage as well as chemotherapy regimen, TILs were an independent predictor for pCR (TILs 10%: OR 1.12, 95% CI, 1.02–1.24, P = 0.018; LPBC: OR 2.08, 95% CI, 1.17–3.687, P = 0.013).

Although statistically not significant, in G5, tumors with higher TIL levels displayed a tendency to higher pCR rates (Fig. 3). In the lapatinib subgroup, there was no significant association between TILs and pCR detectable in both univariate and multivariate analyses (univariate: TILs 10%: OR 1.12, 95% CI, 0.98–1.27, P = 0.091; LPBC: OR 1.9, 95% CI, 0.86–4.2, P = 0.115; multivariate: TILs 10%: OR 1.1, 95% CI, 0.95–1.27, P = 0.199; LPBC: OR 1.85, 95% CI, 0.76–4.46, P = 0.173).

Survival data were available for all patients. Median follow-up was 60.39 months (59.485–61.295). As expected, DFS was significantly associated with pCR rate (P = 0.018; Fig. 4A). LPBC cases tended to have a better DFS compared with no LPBC cases, although this was statistically not significant (P = 0.231; Fig. 4B). When combining both parameters, pCR and LPBC, two distinct subgroups were detectable: A low-risk group consisting of pCR/LPBC cases and the subgroup of no pCR/no LPBC tumors, representing a high-risk group (P = 0.039; Fig. 4G). However, TILs as a continuous variable were not prognostic in general, either in univariate or multivariate analysis.

We then evaluated the prognostic impact of TILs depending on the hormone receptor status. Interestingly, in hormone receptor-negative cases (Fig. 4C and D), pCR appears to be a significantly stronger prognostic variable for DFS (multivariate: HR 2.49, 95% CI, 1.422–4.363, P = 0.001), whereas TILs are of more relevance in triple-positive carcinomas (Fig. 4E and F; multivariate LPBC: HR 2.8, 95% CI, 0.987–7.909, P = 0.053).

This study showed that the level of TILs is an independent positive predictive marker for response to neoadjuvant therapy in HER2-positive breast cancer. It confirms the predictive impact of TILs on response to neoadjuvant chemotherapy in high-risk breast cancer subgroups as demonstrated in previous studies (8, 9, 13).

In our cohort, we found a significant association of TILs and pCR in the trastuzumab subgroup when assessing both cohorts, GeparQuattro and GeparQuinto. However, this association is no longer significant, although it shows a trend when only evaluating the trastuzumab and lapatinib subgroups of GeparQuinto. It is known that at least part of the trastuzumab effect is immune-mediated. Therefore, high levels of TILs may further enhance the efficacy of an anti-HER2 therapy by trastuzumab when compared with lapatinib monotherapy. However, to test this more rigorously, larger case numbers are needed. Furthermore, the anti-HER2 blocking effectiveness of lapatinib monotherapy in general is discussed controversial. When comparing trastuzumab and lapatinib therapy in the neoadjuvant setting, some studies showed significantly lower pCR rates in the lapatinib arm (7, 26, 27), whereas other trials could not confirm this (28).

We were not able to confirm TILs as an independent prognostic marker in general as showed by Salgado and colleagues in the phase III NeoALLTO trial (26). Moreover, it became apparent that the prognostic impact of TILs depends on the hormone receptor status. In triple-positive breast cancer, TILs are of more relevance than pCR, whereas the opposite is observed in hormone receptor-negative HER2-positive cases. It should be noted that most of the neoadjuvant trials investigating anti-HER2 blocking strategies showed significant lower pCR rates for triple-positive tumors (27, 29). This suggests that hormone receptor-negative HER2-positive may be a different subtype from triple-positive tumors. This difference might also be relevant for the interpretation of TILs as a biomarker in general, when discordant observations have been made in different clinical trials (10, 11, 30).

By combining both parameters—TILs and pCR—we were able to identify a low-risk group (pCR/LPBC) with a good DFS as well as a high-risk group (no pCR/no LPBC), which has a high risk of recurrence. As already suggested (31), the combination of both parameters may help to identify more defined prognostic subgroups. High TILs/pCR tumors may have an excellent prognosis that may allow to de-escalate therapy, whereas low TILs/no pCR tumors may need additional therapy strategies, for example, promoting immune response (32).

In conclusion, we evaluated TILs in a large cohort of 498 HER2-positive breast carcinomas. Higher levels of TILs are associated with higher therapeutic efficacy, which confirms recent studies on TILs in high-risk breast cancer subgroups. TILs and pCR may contribute independent prognostic information. By combining both parameters, TILs might become a promising biomarker that may also guide further therapy decisions. This should be confirmed in other HER2-positive neoadjuvant and adjuvant cohorts.

C. Denkert has ownership interest (including patents) in Sividon Diagnostics, and is a consultant/advisory board member for AstraZeneca. P.A. Fasching reports receiving speakers bureau honoraria from Genomic Health, GlaxoSmithKline, Novartis, Pfizer, and Roche, is a consultant/advisory board member for Novartis, Roche, and Teva, and reports receiving commercial research grants from Amgen and Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: B. Ingold Heppner, M. Untch, C. Denkert, H. Eidtmann, F. Holms, S. Loibl

Development of methodology: C. Denkert, F. Holms, S. Loibl

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Ingold Heppner, M. Untch, C. Denkert, B.M. Pfitzner, W. Schmitt, H. Eidtmann, P.A. Fasching, H. Tesch, C. Solbach, D.M. Zahm, F. Holms, M. Glados, P. Krabisch, A. Ober, P. Lorenz, J.-O. Habeck

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B. Ingold Heppner, M. Untch, C. Denkert, B. Lederer, F. Holms, P. Lorenz, S. Loibl

Writing, review, and/or revision of the manuscript: B. Ingold Heppner, M. Untch, C. Denkert, B.M. Pfitzner, B. Lederer, W. Schmitt, H. Eidtmann, P.A. Fasching, C. Solbach, F. Holms, P. Lorenz, S. Loibl

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B. Ingold Heppner, C. Denkert, B. Lederer, H. Eidtmann, M. Rezai, F. Holms, P. Lorenz, K. Diebold, S. Loibl

Study supervision: M. Untch, C. Denkert, E. Heck

The authors thank all patients, clinicians, and pathologists for their remarkable support by participating in clinical studies and collecting biomaterial. They also thank Christiane Rothhaar, Britta Beyer, Sylwia Handzik, Peggy Wolkenstein, Ines Koch, Petra Wachs, and Christoph Weber for their excellent technical assistance.

This study was supported by European Commission, grant 278659, RESPONSIFY.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.