The Predictive Value of HLA Class I Tumor Cell Expression and Presence of Intratumoral Tregs for Chemotherapy in Patients with Early Breast Cancer

Breast cancer is the most common cancer in women: it affects one in nine women. Systemic treatment improves disease-free survival and overall survival in patients with early breast cancer (1). Decisions about this systemic therapy are depending on prognostic and predictive factors, which divide patients into different risk groups (2). With the current classifications, however, prediction of outcome is still not optimal and additional prognostic and predictive factors are needed to improve tailored treatment.

It is widely accepted that the adaptive immune system plays an important role in controlling tumor growth and spread (3). CTLs are capable to affect tumor development. However, because of their intrinsic genetic unstable nature, tumor cells may acquire properties to escape from CTL recognition. Among these properties are downregulation or complete loss of human leukocyte antigen (HLA) class I expression. In addition, immunosuppressive regulatory T cells (Treg) may be induced (4).

HLA class I molecules play a pivotal role in CTL-mediated immune responses and have been found to be a prognostic factor in various types of cancer (5–7). Previous studies have shown that HLA class I expression is frequently downregulated in breast cancer (8, 9). However, the reports on prognostic influence of HLA class I expression in breast cancer have contradictory results (10–12). Some found no significant correlation between percentage of tumor cells expressing HLA class I and survival of breast cancer patients (10, 12). In contrast, another study found that total loss of HLA class I was an independent indicator of good prognosis (11).

Tregs act as immunosuppressors and maintain immunologic self-tolerance. Numbers of tumor-infiltrating Treg are known to be increased in several malignancies, and a correlation was found with worse disease stage and prognosis in cancer (13, 14). In breast cancer, the presence of Treg in the tumor environment has been found in several studies (15–17). Moreover, these studies found a higher prevalence of Treg in tumor microenvironment and in peripheral blood of patients suffering breast cancer compared with healthy donors (15–17). One study found higher numbers of Treg to be correlated with worse disease stage and shorter survival (15). Interestingly, chemotherapy has been found to be involved in presence and prognostic influence of Treg (16). The numbers of Treg in tumor tissue decreased after chemotherapy administration, and there was an association between disappearance of Treg and pathologic complete response to preoperative chemotherapy.

The purpose of our study was to analyze the prognostic relevance of HLA class I expression and Treg infiltration in a large cohort of early breast cancer patients. In addition, we explored the predictive value of these markers for chemotherapy response.

The patient population is composed of all nonmetastasized breast cancer patients primarily treated with surgery in the Leiden University Medical Center between 1985 and 1994 (N = 677). Patients with bilateral tumors or a history of cancer, other than basal cell carcinoma or cervical carcinoma in situ, were excluded. The following data were known: age, tumor differentiation grade and morphology, TNM stage, local and systemic therapy, locoregional/distant tumor recurrence, secondary tumor, alive/death, estrogen receptor, progesterone receptor, Ki67, and human epidermal growth factor receptor 2.⁵

The mouse monoclonal antibodies HCA2 and HC10, which recognize HLA class I heavy chains (kindly provided by Prof. Dr. J. Neefjes), were used. The reactivity spectrum of HCA2 is composed of all HLA-A chains (except HLA-A24), as well as some HLA-B, HLA-C, HLA-E, HLA-F, and HLA-G chains (18, 19). HC10 reacts mostly with HLA-B and HLA-C heavy chains and some HLA-A (HLA-A10, HLA-A28, HLA-A29, HLA-A30, HLA-A31, HLA-A32, HLA-A33; refs. 20, 21). Mouse antibodies against human Foxp3 (ab20034 clone 236A/E7; AbCam) were used for Treg identification. The reactivity spectrum of Foxp3 is composed of Treg and may include small numbers of CD8+ cells (22), but thus far, it is the best single marker of Treg (23).

For HLA class I staining, slides of 4 μm were cut from a priory constructed tissue microarray. For staining of Treg sections, 4-μm sections were cut from the original formalin-fixed, paraffin-embedded tumor blocks. Tissue sections were deparaffinized and rehydrated. Endogenous peroxidase was blocked for 20 min in hydrogen peroxide–methanol. For antigen retrieval, 0.01 mol/L citrate buffer (pH 6.0) was used for 10 min at maximum power in a microwave oven. Sections were incubated overnight with HCA2 or HC10 at room temperature using predetermined optimal concentrations. After incubation with secondary antibody envision anti-mouse (Dako Cytomation K4001), sections were visualized using 3,3′-diaminobenzidine solution (25 mL 3,3′-diaminobenzidine in 225 mL 0.05 mol/L Tris-HCl). Tissue sections were counterstained with hematoxylin, dehydrated, and mounted in malinol. All slides were stained simultaneously to avoid interassay variation. For each patient, normal epithelium, stromal cells, or lymphoid cells served as internal positive control for HLA class I antibody reactivity. Slides from human tonsil tissue served as positive control for Treg staining. For each staining, slides that did undergo the whole immunohistochemical staining procedure but without primary antibodies served as negative controls.

Microscopic analysis of HCA2 and HC10 was assessed by two independent observers (E.M. de Kruijf and Q.R.J.G. Tummers) in a blinded manner. Percentage of tumor cells that showed membranous staining was assessed. HCA2 and HC10 staining were scored in five categories according to the defined standard method of the International HLA and Immunogenetics Workshop (score 1, 0-5% of tumor cells positively stained; score 2, 5-25%; score 3, 25-50%; score 4, 50-75%; score 5, 75-100%; ref. 24). Quantification of Treg within the tumor was microscopically assessed in 10 high-power fields by two observers (E.M. de Kruijf, 100%; A. Sajet, 30%) in a blinded manner. Treg was scored into two categories: absence and presence of Treg infiltration.

Statistical analyses were done using the statistical package SPSS (version 15.0 for Windows; SPSS, Inc.). Cohen's κ coefficient revealed a satisfactory agreement in classification (κ = 0.73). The χ² test was used to evaluate associations between various clinicopathologic variables and HLA class I expression and infiltration of Treg. Relapse-free period was the time from date of surgery until a locoregional recurrence and/or distance recurrence, whichever came first. Clinical follow-up policy was equal for all patients in the study. Overall survival was defined from date of surgery until death. The Kaplan-Meier method was used for calculation of survival probabilities and the log-rank test for comparison of survival curves. Relapse-free period is reported as cumulative incidence function after accounting for death as competing risk (25). Cox regression was used for univariate and multivariate analyses for relapse-free period and overall survival. Significant or close-to-significant variables (P < 0.1) in univariate analysis were included in multivariate analysis. To analyze the predictive effect of HLA class I and Treg, analyses were done and stratified for adjuvant chemotherapy administration.

Tumor material was available and incorporated in the tissue microarray of 86% (574 of 677) of the patients. Clinicopathologic and treatment characteristics are shown in Table 1. Median age of patients was 57 years (range, 23-96 years). Median follow-up of patients alive was 19 years (range, 0-23 years). Chemotherapy treatment is consisted of a combination of cytostatic drugs always containing cyclophosphamide.

Microscopic quantification was successful in 94% (538 of 574) of tumors for HC10 and in 96% (548 of 574) for HCA2 (Fig. 1A and B). A total of 523 of 574 tumors (91%) could be quantified for both and were therefore available for total HLA class I expression evaluation. Three groups were defined for HLA class I expression: (a) HLA class I loss (both HCA2 and HC10 scored 0-5%), 23% of tumors; (b) HLA class I downregulation (either HCA2 or HC10 scored 0-5%), 38% of tumors; and (c) HLA class I expression (both HCA2 and HC10 scored 5-100%), 39% of tumors (Table 1).

A total of 556 of 574 (97%) tumors could be evaluated for Treg infiltration (Fig. 1C). Tumors with absence of Treg (0 Treg per 10 high-power fields) and presence of Treg (≥1 Treg per 10 high-power fields) were seen in 57% and 43% of patients respectively (Table 1).

To analyze the prognostic effect of HLA class I and Treg, all patients who did not receive any systemic therapy were analyzed. There were no statistically significant differences in outcome for relapse-free period (log-rank P = 0.27) or overall survival (log-rank P = 0.55) between different HLA class I expression levels (Fig. 2A and B). Treg showed no differences, neither in outcome for relapse-free period (log-rank P = 0.93) nor overall survival (log-rank P = 0.14) between intratumoral absence and presence (Fig. 3A and B). In contrast with the data in patients who did not receive systemic treatment, analysis of HLA class I expression in chemotherapy-treated patients showed statistically significant differences for relapse-free period between groups (log-rank P = 0.003; Fig. 2C). Of patients with expression of HLA class I, 68% were relapse free after 15 years compared with 51% and 30% for downregulation and loss of HLA class I expression, respectively. Infiltration of Treg showed, similarly to HLA class I, moderate differences in outcome between groups among chemotherapy-treated patients (log-rank P = 0.06; Fig. 3C). Patients with intratumoral infiltration of Treg had less relapses compared with patients with no infiltration of Treg. Cox proportional multivariate analysis was done with data from chemotherapy-treated patients, including the variables that showed a trend of influence on outcome (P < 0.1) in Cox proportional univariate analysis: lymph node status, HLA class I, and Treg (Table 2). This analysis revealed that lymph node status, HLA class I [P = 0.002; hazard ratio for downregulation, 2.11; 95% confidence interval (95% CI) for downregulation, 1.13-3.95; hazard ratio for loss, 3.34; 95% CI for loss, 1.67-6.67], and Treg (P = 0.01; hazard ratio, 2.04; 95% CI, 1.16-3.57) were all independent prognostic factors for relapse-free period among chemotherapy-treated patients.

To prove that HLA class I and Treg were statistically significant cooperating with chemotherapy, an interaction term was introduced in Cox regression analysis. This analysis showed that HLA class I (P < 0.001; hazard ratio for downregulation, 2.15; 95% CI for downregulation, 1.17-3.96; hazard ratio for loss, 3.15; 95% CI for loss, 1.92-5.15) and Treg (P < 0.001; hazard ratio, 2.47; 95% CI, 1.54-3.95) significantly interacted with chemotherapy administration. These data indicated that HLA class I expression and Treg tumor infiltration possess prognostic value, specifically in breast cancer patients who are treated with chemotherapy.

Our study showed that HLA class I and Treg are independent prognostic markers for chemotherapy-treated patients with substantial discriminative power. Variables that are able to determine which breast cancer patients may benefit from adjuvant chemotherapy are few (26). Known factors that tend to indicate better chemotherapy response are negative estrogen receptor, high tumor grade, and high proliferative activity, but their predictive value is marginal (27). In our study, independently of those factors, high levels of HLA class I expression and presence of Treg resulted in statistically significant less relapses over time. Most importantly, these results can be explained by underlying biology and correspond with results of previous studies.

In concordance with previous studies, downregulation and loss of HLA class I was frequently seen in our study (9–11). Previous studies indicate that breast cancer is immunogenic and induces tumor-associated antigen–specific CTL (28). These findings may imply that breast cancer cells with downregulation or loss of HLA class I expression escaped from immune destruction and therefore selectively grew out (29). This seems quite a common phenomenon in breast cancer, considering the fact that we and others found HLA class I downregulation or loss in more than half of the tumors. In addition, Tregs were found in a significant number of tumors. Tumors may either attract these immune-suppressing cells to evade attack from effector T cells, or Treg may consider tumor cells as normal cells and thus prevent immune attack. Our data indicate that the immune system is closely involved in the development of breast cancer. At the time of a clinically manifest tumor, the balance between immune attack and tumor growth obviously is at the site of the tumor (29).

We showed that expression levels of HLA class I had a specific prognostic effect but only in chemotherapy-treated patients. Previous studies on HLA class I expression in breast cancer did not stratify for systemic therapy. Three studies have evaluated HLA class I expression and its effect on prognosis in breast cancer (10–12). Two studies found that HLA class I expression levels had no influence on the prognosis of patients, which is in concordance with our findings in patients that were not systemically treated (10, 12). Our study also showed that infiltration of Treg was a predictive marker for chemotherapy response in breast cancer patients. These finding are supported by Ladoire et al. (16), who found that the number of Treg declined due to chemotherapy, showing that chemotherapy affects Treg and thus may counteract by restrained CTL. More importantly, complete absence of Treg after chemotherapy administration resulted in a better response with higher rates of pathologic complete response, further supporting our findings of a predictive role of Treg. Other studies have shown that infiltration of Treg in breast tumors resulted in a worse prognosis in terms of relapses and survival (15, 30). Our study could not statistically prove such a relation in patients who did not receive systemic treatment. To unravel the complex tumor–immune system interactions during tumor development, further studies are needed.

The specific prognostic effects found for HLA class I and Treg among chemotherapy-treated patients can be explained by the following biological explanation. In our population, choice of chemotherapy is composed of cyclophosphamide, which positively influences host immune responses against cancer (31–33). It is hypothesized that several mechanism are the basis for this phenomenon: enhanced homeostatic expansion of antigen-specific T cells by creation of a niche in the immune system, stimulation of dendritic cells, induction of T-cell growth factors such as type I IFNs, and selective elimination of Treg (31–33). Ceasing of Treg through cyclophosphamide effects results, among other things, in enhanced expansion and function of responding of CTL (32, 33). Ladoire et al. (16) found that, after preoperative chemotherapy, absolute numbers of tumor-infiltrating Treg significantly declined and numbers of effector T cells and CTL remained stable. In addition, a pathologic complete response was associated with a combination of absence of infiltration of Treg and presence of CTL after chemotherapy. In our study, this phenomenon was associated only in tumors that retained HLA expression, suggesting that, upon counteraction of Treg, CTLs are able to affect tumor metastases development.

In summary, HLA class I and Treg have an independent prognostic effect for chemotherapy-treated patients, which can be explained by underlying biology. Both factors resulted in a very high differentiation in sensitivity to chemotherapy. Predictive factors for chemotherapy response in breast cancer are highly necessitated. Therefore, we conclude that HLA class I and Treg are candidate markers for further investigation in randomized studies and may be applied for chemotherapy response prediction.

No potential conflicts of interest were disclosed.

We thank J. Molenaar for his help with the database, K. van de Ham for pictures of representative immunohistochemical stainings, and colleagues at the research laboratory of the surgery department at the Leiden University Medical Center for their help and advice.

Grant Support: Financial support from the Dutch Cancer Society (UL 2007-3968).

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.