Background
Interactions between cancer cells and the host immune system are important for development, evolution, and progression of cancer [
1]. They influence response to therapy and survival of patients and modulating these effects offer new approaches for cancer therapy.
Immune checkpoint inhibitor (CPI) therapy can enhance therapy response in advanced triple-negative breast cancer (TNBC) [
2], and several clinical trials are ongoing. For a successful implementation of such strategies in clinical practice, a better understanding of tumor-immunological effects is necessary.
Evidence suggests a role of the immune system in breast cancer, as the quantity of tumor-infiltrating lymphocytes (TILs) is associated with better response to neoadjuvant chemotherapy and better patient outcome [
3,
4]. In a recent meta-analysis, TILs were associated with a higher probability of pathologic complete response (pCR) [
5]. Abundant TILs were associated with a longer disease-free survival (DFS) in HER2+ breast cancer and TNBC. In contrast, TILs were associated with a shorter OS in patients with hormone receptor-positive, HER2-negative (HR+/HER2−) disease, pointing at differences according to breast cancer subtype.
Human leucocyte antigen (HLA) class I molecules are expressed on the surface of all nucleated cells and are encoded by the human leukocyte antigens HLA-A, HLA-B, and HLA-C. Their function is to present antigens to CD8+ cytotoxic T lymphocytes to recognize and eliminate infected or tumor cells [
6]. Downregulation or loss of HLA class I expression is a frequent event in pathogen-infected and tumor cells as an effective mechanism to evade immune recognition [
7,
8]. Deficient HLA class I expression can be mediated by promoter methylation [
9], mutations in the HLA class I heavy chains (HC), b
2-microglobulin (b
2-m) or APM components, loss of heterozygosity of HLA gene loci, and transcriptional regulation [
10,
11]. Molecular data suggest an impaired expression of components of the antigen presentation machinery (APM) as a mechanism of resistance to T cell response [
12] and HLA class I abnormalities have been identified in tumors resistant to CPIs or adoptive T cell therapy [
13].
Aim of this study was to evaluate the potential of HLA class I HC expression for prediction of response to neoadjuvant chemotherapy. To this end, we evaluated expression of HLA class I HC in a large cohort of breast cancer patients treated with anthracycline/taxane-based neoadjuvant chemotherapy (within the GeparTrio trial [
14,
15]). We used a clinically annotated microarray cohort from the MD Anderson Cancer Center to validate the findings and investigate correlations with predefined metagenes [
16] of different immune cell populations.
Methods
GeparTrio trial cohort
The neoadjuvant GeparTrio pilot [
17] (NCT00544765) and main (NCT00544765) trials [
14,
15] were prospective, randomized phase II and III trials including 2357 patients with breast cancer (cT2–4 cN0–3 cM0) recruited between 2001 and 2005. Patients received two cycles of docetaxel, adriamycin, and cyclophosphamide (TAC) and response was evaluated by ultrasound. Responders received four more cycles of TAC (pilot study) or were randomly assigned to four or six cycles of TAC (main study). Non-responders were randomized to receive either four cycles of TAC or four cycles of vinorelbine and capecitabine. HR+ was defined as ≥ 10% of tumor cells with estrogen receptor and/or progesterone receptor expression, as defined in the study protocol. HER2 positivity was determined by immunohistochemistry (HER2 score 3+) and in situ hybridization where appropriate (HER2/CEP17 ratio > 2.2). Endocrine treatment for 5 years was planned for patients with HR+ disease but was not part of the protocol. HER2 therapy was not available at that time. We used all samples with available material in the central biobank. Additional file
3: Table S1 lists the pathological and clinical characteristics of the patients. The extend of tumor-infiltrating lymphocytes (TILs) was estimated on H&E-stained whole tissue slides as the area of tumor cells (for intratumoral TILs) or the stromal area (for stromal TILs) that is covered by lymphocytes [
3].
Immunohistochemistry of HLA class I antigens
Immunohistochemical staining was performed on tissue microarrays using the anti-HLA class I antibody EMR-8-5 (dilution 1:600) recognizing the HLA class I HC HLA-A, HLA-B, and HLA-C (MBL, Woburn, MA, USA). The percentage of cells with membranous staining relative to all cancer cells was assigned to five categories: (0) 0%, (1) 1–9%, (2) 10–50%, (3) 51–80%, and (4) 81–100%. The intensity of staining was assigned into four categories: (0) negative, (1) weak, (2) moderate, (3) strong. The groups for percentage and staining intensity were multiplied resulting in the immunoreactive score (IRS) ranging from 0 to 12. Based on the distribution of staining levels and patient outcome, a cut point (IRS > 2) was selected to optimize separation of curves in Kaplan-Meier analysis. In addition, the percentage of positively stained tumor cells was used continuously to evaluate the effect of HLA immunohistochemistry on clinical endpoints without a defined cut-off.
MD Anderson cancer center (MDACC) microarray cohort
Gene expression data and clinical information of the MD Anderson Cancer Center cohort [
18] (Affymetrix U133A microarrays) was downloaded from the GEO repository (GSE25066). A total of 320 tumors were HR+/HER2− and patients received neoadjuvant chemotherapy and adjuvant endocrine treatment if HR+. HR positivity was defined as any number of stained cells. We applied an additional filter based on gene expression of ESR1 (probeset 205225_at) based on its bimodal distribution (> 10.45). In total, 267 patients met these criteria. Additional file
3: Table S2 lists the patient characteristics. HLA-A, HLA-B, and HLA-C metagenes were calculated as mean expression of the respective probesets (HLA-A: 215313_x_at, 213932_x_at; HLA-B: 211911_x_at, 209140_x_at, 208729_x_at, HLA-C: 208812_x_at, 216526_x_at, 214459_x_at, 211799_x_at). Unsupervised cut-offs were chosen by assigning the same fraction of cases to the groups with high (60%) and low (40%) expression as in the GeparTrio dataset (HLA-A > 14.24, HLA-B > 13.63, and HLA-C > 13.69). Probesets 205225_at (estrogen receptor 1) and 208079_s_at (Aurora kinase A) were used to evaluate their association with HLA-A. Immune cell metagenes were calculated as the mean expression of cell-type-specific genes [
16].
Statistical methods
Pathologic complete response (pCR) was defined as the absence of invasive cancer in breast and lymph nodes (ypT0/ypTis ypN0). In the MD Anderson cohort, response to chemotherapy was recorded using the residual cancer burden index [
19]
. Disease-free survival (DFS) was defined as time from study entry to local or distant recurrence or death from any cause, distant recurrence-free survival (DRFS) as the interval between diagnosis and distant recurrence or death from any cause, and overall survival (OS) as the time from study entry to death from any cause. To evaluate associations between clinical and pathological variables with HLA I,
χ2 tests were used. Kaplan-Meier estimators and log-rank tests were used for survival analyses (R package
survival). The independent prognostic value was assessed in multivariate Cox regression analyses and likelihood ratio tests. Statistical computations were performed in R version 3.3.1 (R Development Core Team, Vienna, Austria). Significance was based on
P < 0.05 and 95% CI estimates.
Discussion
HLA class I HC expression as evaluated by IHC is predictive for better response and worse DFS and OS in HR+/HER2− breast cancer in a large clinical trial cohort. It was positively associated with TILs and independently prognostic for patient survival adjusted for other clinical and pathological factors. It was neither predictive nor prognostic in TNBC or HER2+ disease. We confirmed the prognostic value of HLA-A expression in an independent microarray-based dataset.
HLA class I HC expression was associated with clinical and pathological tumor characteristics indicative of a more aggressive tumor biology such as the presence of lymph node metastases and high histopathological grade. We considered that this might explain the better response, but worse DFS that was associated with high HLA class I HC expression. However, there was no strong association of HLA-A expression with either ESR1 expression or AURKA expression. HLA was prognostic in a multivariate analysis adjusted for clinical and pathological tumor characteristics including nodal status and tumor grade. Thus, the effects could not be explained by these correlations alone.
The main caveat of the study is that we cannot evaluate the role of HLA class I expression for sensitivity to different treatments, as patients received both cytotoxic and endocrine therapy. Also, we cannot provide data on HLA class I in its most relevant context of therapeutic immunomodulation.
Another major limitation of the study is the lack of an independent validation cohort, especially to validate the cut-point used for immunohistochemistry. Even though the GeparTrio and MD Anderson datasets share clinical and pathological features, a different technology is used to detect biomarkers on different biological levels. This approach comes with limitations, both on the technical and clinical levels. An important clinical difference is that in GeparTrio, hormone receptor positivity was defined as ≥ 10% of tumor cells stained for estrogen or progesterone receptor, in the MD Anderson cohort, as any stained tumor cells. As it is known that low hormone receptor-positive cases may resemble triple-negative disease on the molecular level and clinically [
21,
22], this might influence the comparability of the cohorts. We have therefore applied an additional filter to exclude cases with low ESR1 expression from the MD Anderson dataset.
With these limitations in mind, we believe it is a strength of our study that we are able to show the prognostic relevance of HLA in different cohorts and on the protein as well as mRNA level, which points out the role of HLA as a relevant parameter for the description of the tumor microenvironment.
HLA class I expression was previously evaluated by immunohistochemistry in 212 breast cancer samples not stratified for subtype [
23]. The frequency of strong staining (32.5%) was comparable to our data, but HLA class I was associated with a longer DFS across subtypes. In other cancers, like colorectal [
24] and non-small cell lung cancer [
25], HLA class I expression was positively associated with survival.
For patients with an excellent response to chemotherapy, we observed a trend towards better survival if HLA class I HC expression was high. However, this subgroup was very small, as most patients with HR+/HER2− breast cancer do not show a complete response. In our recent meta-analysis of TILs in breast cancer [
5], TILs were associated with better response to chemotherapy in HR+/HER2− disease, but shorter DFS and OS. The effect was strongest in large tumors and in patients who did not have an excellent response to chemotherapy. Our observations in this study are in line with our current findings, pointing at an interaction between lymphocytes and cytotoxic and endocrine treatment effects.
All of the analyzed immune cell metagenes were positively correlated with HLA-A expression. Only HLA-A, and the monocytic and B cell lineages were statistically significantly associated with higher risk, but as trends and correlations were similar for the different immune cell populations, we found it not possible to pinpoint HLA-A effects to one or more specific immune cell type based on these data.
Patient outcome in HR+/HER2− disease is a function of resistance and sensitivity to neoadjuvant cytotoxic therapy, adjuvant endocrine treatment, and natural biology. Because response to chemotherapy was better in patients with tumors with high HLA class I, we hypothesized that the shorter DFS might be due to an interaction with endocrine treatment. Tumors with a high TIL density might be less responsive to endocrine treatment by unknown mechanisms unrelated to ESR1 expression. Interestingly, immune gene modules can predict poor antiproliferative response to aromatase inhibitors [
26,
27]. In a large study on patients receiving endocrine therapy alone, HLA class I was not prognostic, but the presence of FOXP3-positive cells was associated with better survival [
28].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.