Background
The immune system has proven to play an important role in tumorigenesis and gained a lot of attention in cancer research [
1‐
4]. Consequently, evasion of immune surveillance has become one of the important hallmarks of cancer [
5]. Tumors are thought to be ‘edited’ through a Darwinian selection process into poorly immunogenic tumor variants, invisible to the immune system and able to grow progressively. Immuno-editing might influence patient’s prognosis substantially [
6].
We have described a few mechanisms responsible for evasion of immune surveillance below.
First, cytotoxic T-cells (CTL) are capable of destroying tumor cells by recognizing tumor-associated antigens (TAA) on the tumor cell surface presented by classical human leukocyte antigen (HLA) class I. Tumor cells can escape this CTL recognition through downregulation or complete loss of HLA class I, resulting in minimization of TAA expression and absence of CTL destruction [
7‐
9]. Second, non-classical HLA-E and HLA-G also play an important role in immune surveillance. Presence of HLA-E and HLA-G causes an inhibitory signal to natural killer (NK) cells, resulting in further immune escape [
7,
10‐
14]. HLA-E is regularly expressed in various healthy tissues and correlates with HLA class I expression [
15]. HLA-G is rarely found in healthy tissues, but is frequently observed in tumors [
16]. Third, immune reactivity can become suppressed by the attraction of immunosuppressive regulatory T cells (Tregs) into the tumor microenvironment [
17,
18].
In colorectal cancer (CRC), the presence of Tregs in the tumor micro-environment has been related to a worse prognosis in some studies, although other studies showed an inverse association [
19‐
22]. Loss of HLA Class I tumor expression was related to a better prognosis in CRC in most studies [
14,
23] and HLA-E and HLA-G tumor expression has been correlated with a poor prognosis and tumor progression [
24,
25].
In rectal cancer specifically, only a few studies reported on the role of the immune system, in which expression of HLA Class I was related to a better prognosis [
26,
27]. Recently, more studies showed differences in biology between colon- and rectal cancer [
28‐
30]. Unfortunately, most studies so far have focused on CRC and did not perform separate analyses. Furthermore, often only one immune marker was investigated in CRC, while recent studies showed the complex interaction between the different mechanisms of immune-escape [
6,
31,
32].
In this study we therefore aimed to investigate the immune-related biomarkers HLA Class I, HLA-E and -G and Tregs, determined with immunohistochemistry, in rectal cancer specifically, and to establish distinct patterns that reflect immune-escape mechanisms of rectal cancer by combining these markers and relate these patterns to clinical outcome.
Methods
Study population
The study cohort consisted of patients obtained from the non-preoperative treated arm of the Dutch TME trial (January 12
th, 1996, DUT-KWF-CKVO-9504, EORTC-40971, EU-96020), a multicenter trial that evaluated total mesorectal excision (TME) surgery with or without preoperative radiotherapy (5 × 5 Gray) from 1996–1999 [
33]. Radiotherapeutical, surgical and pathological procedures were standardized and quality-controlled [
34]. Before the start of the TME trial the Medical Ethical Committee of the Leiden University Medical Center approved the trial and retrospective use of samples. Written informed consent for participation and retrospective use of samples was obtained from all patients enrolled in the TME trial. Previously, a tissue microarray (TMA) including 1208 patients (irradiated and non-irradiated) of the Dutch TME trial was available. Because of insufficient tissue on this TMA a new TMA was constructed for this study. Sufficient formalin-fixed paraffin-embedded tumor material was available for 495 non-preoperative radiotherapy-treated stage I-IV Dutch patients, resulting in a total study cohort of 495 rectal cancer patients who only had surgery.
Antibodies
The mouse monoclonal antibodies HCA2 and HC10 were used, which recognize the heavy chains of HLA Class I, these were kindly provided by Prof. Dr. J. Neefjes (NKI, Amsterdam, The Netherlands). The reactivity spectrum of HCA2 comprises all HLA-A chains (except HLA-A24), as well as some HLA-B, HLA-C, HLA-E, HLA-F, and HLA-G chains. HC10 reacts with HLA-B and HLA-C heavy chains and some HLA-A chains (HLA-A10, HLA-A28, HLA-A29, HLA-A30, HLA-A31, HLA-A32, HLA-A33) [
31]. The mouse antibody against human Foxp3 (ab20034 clone 236A/E7; Abcam) was used for Treg identification. The reactivity spectrum of Foxp3 is composed of regulatory T cells and may include small numbers of CD8+ cells but is generally considered to be the best single marker for Treg identification [
35,
36]. For HLA-E and HLA-G identification mouse monoclonal antibodies against HLA-E (ab2216 clone MEM-E/02: AbCam) and HLA-G (4H84: Exbio, Czech Republic) were used [
32]. MEM-E/02 recognizes denatured HLA-E [
37,
38], while 4H84 recognizes denatured HLA-G molecules and also binds to free heavy chains of classical HLA class I molecules [
38‐
40].
TMA production and immunohistochemistry
Histo-pathological characteristics of tumor material from all patients were standardized and quality-controlled [
33,
34]. Sections from formalin-fixed paraffin-embedded (FFPE) tumor blocks of the primary tumors were cut for haematoxylin and eosin staining. Based on these slides, histopathologically representative tumor regions were identified and punched for preparation of tumor tissue microarray (TMA) blocks. From each donor block, three 1.0 mm diameter tissue cores were punched from three different identified tumor areas to account for tumor heterogeneity and transferred into a receiver paraffin block using the TMA master (3DHISTECH, Budapest, Hungary). Immunohistochemical staining (IHC) for Foxp3+ cells, non-classical HLA-E and HLA-G, and classical HLA class I tumor expression was performed on 4 μm sections, which were cut from each receiver block and mounted on glass. For each type of primary antibodies, all slides were stained simultaneously to avoid inter-assay variation.
The sections were deparaffinized and rehydrated in accordance with standard protocol. Endogenous peroxidase was blocked for 20 minutes in 0.3% hydrogen peroxide in PBS. For antigen retrieval, slides for staining with HLA-E, HLA-G or Foxp3+ were boiled in a 0.01 M EDTA buffer (pH 8) for 10 minutes at maximum power in a microwave oven. Slides for staining with HCA2 and HC10 were boiled in a 0.1 M citrate buffer (pH 6). Sections were incubated overnight with Foxp3, HLA-E, or HLA-G antibodies at pre-determined optimal dilution. The next day, after 30 minutes of incubation with Envision anti-mouse (K4001; DAKO Cytomation, Glostrup, Denmark), sections were visualized using diaminobenzidine solution (DAB). Tissue sections were counterstained with haematoxylin, dehydrated and finally mounted in pertex.
For the HCA2 and HC10 stainings a double staining was performed to better discriminate between stroma (using a mixture of anti-extracellular matrix antibodies that resulted in brown staining of tumor stroma) and tumor tissue (using a blue staining for the HLA expression to be determined) in the tissue sections. Sections were incubated overnight at room temperature with all primary antibodies simultaneously (anti-collagen I, anti- collagen VI, anti-elastin (all polyclonal rabbit antibodies obtained from AbCam) and HCA2 and HC10). Afterwards, sections were washed three times for 5 minutes in PBS and incubated for 30 minutes with Envision + System HRP anti Rabbit (DAKO, Glostrup, Denmark). After washing the sections three times with PBS, sections were developed using Liquid DAB + Substrate Chromogen System (DAKO, Glostrup, Denmark) following manufacturer’s instructions for visualization of stromal tissue. Then, sections were washed again three times for 5 minutes in PBS followed by 30 minutes incubation with rabbit-anti-mouse antibodies (DAKO, Glostrup, Denmark). Afterwards, the sections were incubated with APAAP (DAKO, Glostrup, Denmark) diluted in PBS/BSA 1% for 30 minutes. And finally, sections were washed three times for 5 minutes in PBS followed by 20 minutes incubation with Vector-Blue following manufacturer’s instructions for visualization of the HCA2 and HC10 antibodies, and mounted in Aquamount (Merck, Darmstadt, Germany).
For each patient, normal epithelium, stromal cells, or lymphoid cells served as internal positive control for HLA class I and HLA-E antibody reactivity [
24]. Tonsil tissue served as external positive control for the HCA2 and HC10 stainings and placenta tissue slides for the HLA-E and HLA-G stainings. Slides from human tonsil tissue served as positive control for Foxp3 staining. Tissue slides that underwent the whole immuno-histochemical staining without primary antibodies served as negative controls (Additional file
1: Figure S1).
Evaluation of immunohistochemistry
Microscopic analysis of HCA2, HC10, HLA-E and HLA-G expression and presence of Foxp3+ cells was performed by two independent observers in a blinded manner (M.S.R.: 100% of the cohort, C.C.E. 30% of the cohort). The kappa values for inter-observer agreement were all between 0.5 and 0.7, indicating substantial agreement between the two observers [
41]. The scores of the three 1.0 mm punches were averaged. For HCA2 and HC10 the percentage of tumor cells with membranous staining was assessed. HLA class I expression status was determined according to the International HLA and Immunogenetics Workshop [
42], with tumor cell HLA class I expression status defined as follows: loss of HLA class I expression: less than 5% of tumor cells expressing both HCA2 and HC10, downregulation of HLA class I; less than 5% of tumor cells expressing either of the markers, and expression of HLA class I: 5% or more expressing both markers. For HLA-E and HLA-G, intensity of tumor staining (absent (undetectable or faint in <20% of the cells), weak (faint to weak in 20% but ≤70% of the cells), moderate (weak to moderate in >70% of the cells) or strong intensity (intense in 20-70% of the cells)) was determined, based on previous studies [
43,
44]. The scores of the three 1.0 mm punches were averaged as well. For analysis these scores were further categorized as weak (absent and weak intensity together) versus strong (moderate and strong intensity together) tumor staining. Quantification of the number of Foxp3+ cells was microscopically assessed in the entire tumor punches of the TMA and the absolute number of positive cells was used for analysis, with the use of the median as cut-off value for categorization in two categories: Foxp3+ below median and Foxp3+ above median.
Statistical analysis
Statistical analyses were performed using the statistical package SPSS (version 17.0 for Windows; SPSS Inc.). The Student’s T-test and the Chi-squared test were used to evaluate associations between tumor expression of classical HLA Class I, non-classical HLA-E and HLA-G and tumor infiltration of Foxp3+ cells and various clinico-pathological variables. Overall Survival (OS) was defined as time of surgery until death; Disease Free Survival (DFS) as time of surgery until death or relapse of disease, whichever came first. The Kaplan-Meier method was used for calculation of survival probabilities and the Log-rank test for comparison of survival curves between expression levels of markers. Cox regression was used for univariate and multivariable analysis for OS and DFS. To preserve statistical power in subgroup analyses, patients with stage IV disease (n = 32) and positive resection margin (n = 98) were included in the final analyses. In multivariable analyses corrections were made for TNM stage, circumferential margin, age, tumor grade and adjuvant therapy.
Discussion
In this study, by combining the immune-related tumor markers HLA class I, HLA-G and Foxp3+, we reported an independent association between tumor immune-phenotype and patient outcome. These phenotypes might represent how the immune system controls tumor growth and metastases in rectal cancer.
Previous studies on HLA class I expression, which focused on a mixed population of colon- and rectal cancer together, have shown inconsistent findings [
13,
14]. Our study showed a survival benefit for patients with tumors expressing HLA class I. These results are partly comparable with results from Watson
et al., who showed that low expression of HLA class I was related to a poor prognosis in a large group of colorectal cancer patients, whereas tumors with loss or expression of HLA class I were associated with a survival benefit [
14]. A substantial part of Watson’s cohort showed HLA class I negative tumors (24.6%). In our cohort, consisting solely of rectal cancer patients, only 15.0% of the patients had tumors with loss of HLA class I expression, which might indicate that colon cancers lose their HLA class I expression more often. Previously, Speetjens
et al. investigated the prognostic value of HLA class I expression in rectal cancer patients from the Dutch TME Trial as well [
27]. In this study, as described in the methods sections, a new TMA was used without complete overlap and thus different patients. Both studies showed a survival benefit for patients with tumors showing expression of HLA class I. Because we have changed the scoring criteria based on recommendation by the International HLA and Immunogenetics Workshop [
42] differences have to be acknowledged. Speetjens
et al. reported that 16% of non-irradiated patients had tumors with loss and downregulation of HLA Class I, whereas our study showed 37% (15% loss and 22% downregulation). Thus, besides a different patient cohort, other possible explanations for inconsistent findings between studies are the use of different definitions of HLA class I expression and differences in staining techniques. Furthermore, tumor microsatellite status might also play an important role. Approximately 50% of all proximal colon tumors show microsatellite instability (MSI), whereas almost all distal colon and rectal cancers are microsatellite stable (MSS) tumors [
46,
47]. Loss of HLA class I has been described more significantly in MSI colorectal tumors compared to MSS right-sided colon tumors [
48,
49]. HLA class I negative tumors are therefore more likely to be MSI tumors with a different clinical behavior than MSS colorectal tumors [
27]. Since MSI tumors have a better prognosis, MSI might influence prognostic results when considering HLA class I expression in colorectal tumors [
46]. In this rectal cancer cohort determination of the microsatellite status would not have been useful. Research has shown that in only 2% of rectal cancers MSI can be found [
50], resulting in insufficient statistical power for separate analyses Finally, colon and rectum are biologically different tissues; the colon epithelium consists of simple columnar epithelium, whereas the rectum is a transition from single columnar epithelium to stratified squamous epithelium, which might result in different outcomes. The Cancer Genome Atlas Network attempted to find biological differences between colon and rectal cancer. However, only differences in anatomical tumor site with more hypermethylation in right-sided tumors were found, which might be explained by different embryonic origins of the right-and left-sided tumors [
28].
Results in our study regarding non-classical HLA-G are remarkable. HLA-G expression can inhibit NK-cells from lysing tumor cells that have lost or downregulated classical HLA class I expression as a secondary immune escape [
51,
52]. However, in this study, positive HLA-G expression was correlated with a longer disease free survival.
The antibody used to stain HLA-G can also bind to free heavy chains of classical HLA class I molecules as well, possibly explaining the remarkable results. We therefore performed an interaction analysis between these antibodies. However, no interaction between HLA-G and HLA class I expression was found. Furthermore, HLA-G is found to be highly immunosuppressive by directly inhibiting NK cells, but also by recruitment of Tregs and induction of Treg differentiation [
53]. Our study showed that strong HLA-G expression was significantly related to presence of more Foxp3+ cells, possibly explaining the favourable prognosis of tumors with strong HLA-G expression, since tumors that attracted more Foxp3+ cells had a better outcome in our cohort. Immune regulation in cancer still remains complex and multifaceted, and not all immune related mechanisms are completely clear. To our knowledge, no other studies on HLA-E and HLA-G are performed on rectal cancer specifically and therefore no other comparisons could be made.
The presence of Foxp3+ cells in the tumor microenvironment is thought to inhibit host-protective antitumor responses and especially CTL activity [
6]. A high density of tumor infiltrating Foxp3+ cells has shown to be associated with an unfavorable prognosis in a wide range of human carcinomas [
54,
55]. However, in accordance with our results, opposite results are described in CRC [
20,
21]. A possible explanation could be a significant association between HLA class I tumor expression and Foxp3+ tumor infiltration in our cohort. Foxp3+ infiltrating cells might be necessary to counteract CTL activity in tumors expressing HLA class I to prevent an auto-immune response on other bodily cells as well. Another explanation might be a different micro-environment of rectal cancer, which is colonized with many gastro-intestinal bacteria, triggering the production of pro-inflammatory cytokines causing tumor-enhancing effects. Instead of the specificity of infiltrating T-cells for tumor-antigens, T-cells in rectal cancer could be more specific for the microflora and suppress inflammation and immune responses from bacterial invasion, resulting in an anti-tumorigenic effect [
56].
As shown in our results and results from our previous studies in breast cancer, immune markers are related to each other [
31,
32]. Studying solely one marker might not be enough to truly understand cancer immune surveillance. When we combined our markers, patients showing the worst prognosis were patients with tumors bearing 2 or 3 negative prognostic markers; patients with loss of HLA class I tumor expression, weak HLA-G tumor expression and low tumor infiltration with Foxp3+ cells. These patients therefore qualify as very low immune susceptible. They probably were able to elicit only a minimal CTL attack and subsequently attracted little to no Foxp3+ cells in their tumor micro-environment, possibly explaining their worse prognosis. Furthermore, patients with tumors showing loss of HLA class I expression, low Foxp3+ cell infiltration and strong HLA-G expression showed the worst outcome perspectives. These patients probably had tumors which were highly ‘edited’ as well, causing a minimal CTL attack and subsequently attracted little to no Foxp3+ cells, and because of strong HLA-G expression were able to escape further immune recognition through inhibition of NK cell recognition and subsequently no elimination [
51,
52].
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
MSR participated in the design of the study, carried out the immunohistochemistry and scoring, performed statistical analysis and drafted the manuscript. CCE performed scoring of the immunohistochemical stainings, interpreted the data and aided in drafting the manuscript. HP participated in the design of the study and performed statistical analysis. HM participated in the design of the study, interpreted data and provided pathological support. GJL interpreted data and aided in drafting the manuscript. CJHC participated in the design of the study (principle investigator Dutch TME trial), and interpreted data. PJKK participated in the design of the study, interpreted data and aided in drafting the manuscript. All authors read and approved the final manuscript.