Inclusion of BLIMP-1⁺ effector regulatory T cells improves the Immunoscore in a cohort of New Zealand colorectal cancer patients: a pilot study

Abstract

Analysis of tumour-infiltrating T cells in colorectal cancer can predict disease-free survival. The Immunoscore, obtained by quantifying tumour-infiltrating CD3⁺ and CD8⁺ T cells, may improve current staging. Effector regulatory T cells are a potently suppressive subset in mice and, while present in human colorectal cancer, their role in patient outcome is unknown. Immunofluorescence was used to analyse immune cell infiltrates in patients with early (stage II) colorectal cancer with (n = 13) and without (n = 19) recurrent disease. CD3 and CD8 were used for the Immunoscore; FOXP3, BLIMP-1 and CD3 to identify effector regulatory T cells. Patients with high Immunoscores had increased disease-free survival compared to patients with low Immunoscores (Log-rank test p < 0.01). Prediction of outcome was further improved by stratifying patients with a low Immunoscore according to CD3⁺FOXP3⁺BLIMP-1⁺ cell infiltration at the invasive margin. Patients with a low Immunoscore and high infiltrate of CD3⁺FOXP3⁺BLIMP-1⁺ cells tended to have better disease-free survival than patients with low Immunoscore and low infiltrate of CD3⁺FOXP3⁺BLIMP-1⁺ cells. Patients with a high Immunoscore had better disease-free survival than patients with a low Immunoscore and low infiltrate of CD3+ FOXP3+ BLIMP-1+ cells (Log-rank test p < 0.001). These results indicate that tumour infiltration with effector regulatory T cells improves the prognostic value of the Immunoscore and implies that these cells may play a role in colorectal cancer patient outcome.

Electronic supplementary material

The online version of this article (doi:10.1007/s00262-016-1951-1) contains supplementary material, which is available to authorised users.

Introduction

Colorectal cancer (CRC) incidence is highest in Australia and New Zealand (38.5 per 100 000 people) [1]. The standard treatment for CRC is surgical resection, followed by adjuvant chemotherapy for patients with lymph node metastases (American Joint Committee on Cancer (AJCC) stage III). Patients with early stage disease (AJCC stage I and II) do not usually receive chemotherapy unless other adverse histological (lymphovascular invasion, poor differentiation) or clinical (presentation with perforation or obstruction) features are present because the risks and costs of adjuvant therapy in stage II CRC may outweigh the benefit [2, 3]. Nevertheless, up to 25% of so-called ‘low risk’ patients eventually relapse [4].

The immune response in CRC is one of the most important determinants of outcome [5, 6]. Galon et al., found that quantitative analysis of tumour-infiltrating T cells was superior to TNM (tumour:node:metastasis)/AJCC staging methods at identifying patients who went on to develop recurrent disease [7]. These researchers proposed the Immunoscore, a new tool designed to measure the infiltration of CD3⁺ and CD8⁺ cells in the invasive margin and the centre of the tumour [8]. The Immunoscore is currently undergoing validation in a prospective multi-national study to determine its relevance in effectively staging colorectal cancer patients [9].

Other cell types in addition to CD8⁺ T cells have been implicated in cancer outcome. Regulatory T cells (Tregs), defined by expression of the transcription factor FOXP3, have been associated with both good and poor prognoses in CRC and these inconsistencies may be related to phenotypic plasticity caused by the local immune microenvironment [10]. FOXP3 is often used as a marker of Tregs, despite transient upregulation on activated T cells [11] and lack of expression in some regulatory subsets. Despite this, it remains a consistent marker of regulatory T cells. The transcription factor BLIMP-1 (gene: PRDM1) may be co-expressed with FOXP3 in murine Tregs [12]. BLIMP-1 is necessary for plasma B cell differentiation in both mice and humans, but also has an important role in controlling IL-2 signalling in murine T cells (reviewed in [13]). Furthermore, BLIMP-1 is upregulated in murine Tregs in response to T cell receptor stimulation [14] and has also been shown to be important for IL-10 production in both mice and humans [12, 15]. In mice, BLIMP-1⁺FOXP3⁺ T cells have been termed “effector Tregs” due to their ability to co-express effector-like markers and markers of suppression [16]. To date, very little is known about BLIMP-1⁺FOXP3⁺ T cells in humans. Santner-Nanan et al. showed that BLIMP-1 was expressed in freshly isolated antigen-experienced T cells [17]. There is, therefore, a need to study this cell population with respect to its putative role in immune regulation of disease. While others have defined “effector Tregs” as Tregs with effector functions and have identified these cells in humans both in homeostasis and in the context of CRC [18, 19], the association between FOXP3⁺BLIMP-1⁺ effector Tregs and patient outcome in CRC has not previously been reported.

The aims of the current study were first, to validate the Immunoscore in a New Zealand cohort of patients with AJCC stage II disease, and second, to determine whether quantification of the FOXP3⁺BLIMP-1⁺ effector Treg infiltrate can improve prediction of disease recurrence, compared to the Immunoscore alone.

Materials and methods

Patient selection and clinical data collection

Patients undergoing elective surgery for colorectal cancer at Dunedin Hospital were recruited from 1995 to 2006. The study was approved by the Lower Regional Ethics Committee (LRS/11/04/017/AM02) and all patients gave written informed consent prior to inclusion in the study in accordance with the Treaty of Helsinki. Resected specimens were fixed in 4% formaldehyde and tumour samples embedded in paraffin wax. Formalin–fixed, paraffin-embedded (FFPE) tumour blocks were used for staging and for immunohistochemical analysis. Patients were staged according to the AJCC or TNM classification. Patients were followed up prospectively for a minimum of 5 years. Recurrent disease was confirmed by radiological and clinical criteria with tissue confirmation in most cases.

In this pilot study, the original cohort of 60 patients was chosen primarily on the basis of AJCC stage II disease with minimum 5 year follow-up. Patients were divided into those with and without recurrence disease, but laboratory investigators were blinded to the clinical data including patient outcome. Fifteen of the 60 original patients were excluded due to lack of distinct invasive margin and the centre of the tumour on the histology slides available. A further 13 patients who received pre-operative radiation therapy were excluded because of evidence that radiotherapy alters the immune cell infiltrate and destroys the invasive margin and centre of the tumour [20]. The characteristics of the remaining 32 patients that comprise the final cohort are shown in Table 1.

Table 1.

Patient characteristics

	Recurrent 13 (%)	Non-recurrent 19 (%)
Age (mean ± SD)	72.86 ± 10.5	72.72 ± 10.37
Gender
Female	7 (54)	8 (42)
Male	6 (46)	11 (58)
Tumour side
Righta	9 (70)	10 (53)
Leftb	4 (30)	9 (47)
Tumour stage
II A	9 (70)	17 (89)
II B	4 (30)	2 (11)

^aTumour proximal to splenic flexture

^bSplenic flexture and beyond

Immunohistochemical analysis

FFPE tumour blocks were cut into 4 µm sections and processed onto glass slides by the Histology Unit at the University of Otago. Haemotoxylin and Eosin stains were performed on one slide from each block to allow identification of the invasive margin and the centre of the tumour.

Immunofluorescence analysis

Preparation for immunofluoroescence was carried out as published [9]. Briefly, the paraffin wax from FFPE slides was removed from each slide using three washes with xylene (Thermo Fisher, MA, USA) then rehydrated with two washes of 100% ethanol (Thermo Fisher), one wash of 95% ethanol and one wash of 70% ethanol. The slides were then washed with phosphate-buffered saline (PBS; Sigma–Aldrich, MS, USA) before heat-induced antigen retrieval in which slides were boiled for 20 min in a conventional microwave in 10X pH6 Citrate antigen retrieval buffer (Sigma–Aldrich). After the slides had cooled to room temperature they were washed with PBS and incubated with protein block solution (3% Bovine Serum Albumin) (BSA, Life Technologies, CA, USA) for one hour at 37 °C. Slides were incubated with primary antibodies (Supplementary Table 1) overnight in the dark at 4 °C then washed three times with PBS (5 min/wash). Optimial concentrations of antibodies were determined by titration. Slides were then incubated with secondary antibodies (Supplementary Table 2) and 4′,6-diamidino-2-Phenylindole, Dilactate (DAPI; Biolegend, CA, USA), in the dark at room temperature for 60 min. Slides were washed three times in PBS, dried and a cover slip was applied using a drop of DAPI Prolong gold anti-fade (Life Technologies). Immunofluorescence, rather than immunohistochemistry, was used to quantify T cell populations, to allow for the inclusion of multiple T cell markers in analysis.

Quantification of the Immunoscore

Briefly, the invasive margin and centre of the tumour regions of tumour FFPE sections were identified. The invasive margin was defined as the 100 μm region on each side of the border between tumour cells and non-tumour stroma. The selection was determined by a histopathologist, blinded to clinical outcome. Immunofluorescence microscopy was used to measure the CD3⁺ and CD8⁺ T cell infiltrate. The T cell infiltrate for each patient was measured by counting 10 high-power fields; T cell frequencies were validated by an independent investigator with good inter-observer agreement [intraclass correlation coefficient (ICC) of 0.80–0.86] [21]. Manual counting, rather than automated counting, was chosen due to the ability to avoid necrotic areas and identify the tumour locations more effectively [21, 22]. The Immunoscore was calculated as follows: an infiltrate above the median at the invasive margin (where tumour tissue and normal bowel tissue intersect) or the centre of the tumour for CD3 or CD8 resulted in a score of 1 and below the median was a score of 0, these scores added together gave a total Immunoscore [9]. The minimum score was Immunoscore (IS)-0; the maximum was IS-4.

Quantification of Treg infiltrate

CD3⁺ and FOXP3⁺ double positive cells were defined in this study as “Tregs” and were counted in 10 high-power fields and validated by an independent investigator. Data from the median CD3⁺FOXP3⁺ cell percentages at the invasive margin was analysed, as this has been reported to be the tumour location where regulatory T cells had the most impact on patient outcome [22]. FOXP3⁺ Tregs were scored as a percentage of CD3⁺ cells to exclude the bias of a high T cell infiltrate. Patient samples were scored ‘high’ if the infiltrate was above the median percentage of Tregs, and ‘low’ if below the median.

Quantification of effector Treg infiltrate

In this study, “effector Tregs” were defined as CD3⁺FOXP3⁺BLIMP-1⁺ T cells, in accordance with the literature describing murine effector Tregs [16]. Effector Tregs were identified and the frequency of infiltrating cells was quantified by counting CD3⁺, BLIMP-1⁺ and FOXP3⁺ triple positive cells in 10 high power fields, validated by an independent investigator. Effector Tregs were scored as a percentage of CD3⁺FOXP3⁺ Tregs. The frequency of infiltrating effector Tregs was compared between patients with recurrent and non-recurrent disease using the cohort median as the cut off for ‘high’ and ‘low’.

Statistical analysis

Dichotomous data were compared using the Fisher exact test. Kaplan–Meier survival curves were used to estimate disease-free survival, and comparisons between groups were performed with the log-rank test. Disease-free survival was measured in months between the date of primary tumour removal and diagnosis of recurrent disease. p < 0.05 was considered statistically significant. Multivariable analysis was not performed due to the limited sample size.

Results

Clinicopathological characteristics

The cohort of stage II patients included 13 patients with and 19 without recurrence (Table 1). The mean age of the recurrent and non-recurrent groups was similar (71 versus 73 years). The recurrent group had a higher proportion of right-sided tumour location (70 versus 50%) and a higher proportion of females (54 versus 40%) than the non-recurrent group.

The Immunoscore predicts disease-free survival

Of the 32 cases, 10 (31%) had Immunoscores scored as 0 (IS-0), three (8%) were IS-1, five (16%) were IS-2, five (16%) were IS-3 and nine (28%) were IS-4. Immunofluorescence images of a tumour with IS-4 and IS-0 are shown (Fig. 1a, b). Patients with a high Immunoscore (IS-3 or IS-4) had significantly better disease-free survival than those with a low Immunoscore (IS-0, IS-1 or IS-2) (Fig. 1c, p < 0.05). Although a low Immunoscore was associated with worse survival, only 50% of patients with IS-2 had disease recurrence. These results demonstrate that the Immunoscore is sensitive but cannot identify all patients at high risk of recurrent disease.

Fig. 1.

Patients with a high Immunoscore are less likely to have recurrent disease than patients with a low Immunoscore. a, b Representative images of the invasive margin of formalin–fixed,paraffin-embedded (FFPE) colorectal tumour with a high (a) or low (b) Immunoscore. DAPI (blue), CD3 (green) and CD8 (red). c Kaplan–Meier disease-free survival curve of colorectal cancer patients with Immunoscore (0–4). Statistical analysis was performed using a Log-rank test. *p < 0.05. n = 32

The frequency of Tregs infiltrating the tumour was not associated with disease-free survival

CD3⁺FOXP3⁺ cells were scored as a percentage of CD3⁺ T cells to ensure that FOXP3 was associated with T cells, not tumour cells, and to eliminate bias associated with high levels of T cell infiltrate (high Immunoscore). In our cohort, 71% of patients with a high absolute number of CD3⁺FOXP3⁺ cells also had a high Immunoscore. The proportion of patients with a high or low percentage of FOXP3⁺ cells was similar between those with and without recurrent disease. No association of the FOXP3⁺ Treg infiltrate and disease-free survival in colorectal cancer was seen in this cohort (Fig. 2). There was no significant difference in patient outcome between groups with high or low absolute numbers of CD3⁺FOXP3⁺ Tregs (data not shown).

Fig. 2.

The infiltrate of regulatory T cells is not associated with disease-free survival. Representative images of FFPE colorectal tumour with a high (a) or low (b) frequency of infiltrating regulatory T cells. DAPI (blue), CD3 (green) and FOXP3 (red); Tregs shown with yellow arrows. c Kaplan–Meier disease-free survival curve of colorectal cancer patients with frequency of regulatory T cells. High and low frequencies were determined by the cohort median. Statistical analysis was performed using a Log-rank test. n = 32

The frequency of FOXP3⁺BLIMP-1⁺ effector Tregs infiltrating the tumour was associated with disease-free survival

CD3⁺FOXP3⁺BLIMP-1⁺ cells were scored as a percentage of CD3⁺FOXP3⁺ Tregs to eliminate high T cell infiltrate bias, similar to the scoring for Tregs (CD3⁺FOXP3⁺). As described earlier, FOXP3⁺BLIMP-1⁺ were labelled in this study as “effector Tregs”. These cells also produced more IL-17 than FOXP3⁺BLIMP-1⁻ T cells (Supplementary Fig. 1). Patients with a low frequency (below median) of BLIMP-1⁺ effector Tregs made up 46% of the total cohort, but 69% of patients with recurrent disease. A high percentage of BLIMP-1⁺ effector Tregs in the invasive margin, but not the centre of the tumour, was associated with recurrent disease (Fisher’s Exact test, P < 0.0001). Images of high and low percentages of BLIMP-1⁺ effector Tregs are shown in Fig. 3a, b. The frequency of BLIMP-1⁺ effector Tregs at the invasive margin was associated with disease-free survival (Log-rank P < 0.06), independent of the T cell infiltrate and the FOXP3⁺ Treg infiltrate of the patient (Fig. 3c, data not shown). These data indicate that, independent of the Immunoscore, patients with high frequencies of BLIMP-1⁺ effector Tregs in the invasive margin are less likely to experience recurrent disease than those with low frequencies of BLIMP-1⁺ effector Tregs.

Fig. 3.

The infiltrate of BLIMP-1⁺ effector regulatory T cells may be associated with disease-free survival. Representative image of FFPE colorectal tumour with a high (a) or low (b) frequency of BLIMP-1⁺ effector regulatory T cells. DAPI (blue), CD3 (green), FOXP3 (red) and BLIMP-1 (white); BLIMP-1⁺ effector Tregs indicated by white arrows, BLIMP-1⁻ Tregs indicated by yellow arrows. c Kaplan–Meier disease-free survival curve of colorectal cancer patients with BLIMP-1⁺ regulatory T cells. High and low was determined by the cohort median. Statistical analysis was performed using a Log-rank test. p < 0.06. n = 32

FOXP3⁺BLIMP-1⁺ effector Tregs can improve sensitivity of the Immunoscore

A proportion of patients with a low Immunoscore did not have recurrent disease (Fig. 1c). Because of the association between BLIMP-1⁺ effector Tregs and the absence of recurrence, as shown in Fig. 3, we wanted to determine if stratifying patients with a low Immunoscore according to BLIMP-1⁺ effector Treg infiltration could improve the sensitivity (that is, by identifying patients with a low Immunoscore who did not experience recurrence). Patients were therefore separated into two groups: high Immunoscore (IS-3 or IS-4) or low Immunoscore (IS-0, IS-1 or IS-2; Fig. 4a). Those with a low Immunoscore were stratified according to the frequency of FOXP3⁺BLIMP-1⁺ effector Tregs infiltrating the invasive margin of the tumour. Patients with a low Immunoscore (IS-0, IS-1 or IS-2) but high effector Treg infiltrate at the invasive margin tended to have better disease-free survival than those with a low effector Treg infiltrate. Patients with a high Immunoscore had better disease-free survival than those with either a high or low infiltrate of effector Tregs (Log-rank p < 0.05 and p < 0.001, respectively; Fig. 4b). These data indicate that the quantification of effector Treg populations alongside the Immunoscore may significantly improve identification of patients at high risk of developing recurrent disease.

Fig. 4.

The addition of BLIMP-1⁺ effector regulatory T cells into the Immunoscore may improve estimation of disease-free survival of colorectal cancer patients. a Kaplan–Meier disease-free survival curve of patients with high Immunoscore (3–4) and low Immunoscore (0–2). b Kaplan–Meier disease-free survival of patients with a high Immunoscore, low Immunoscore with low frequency of BLIMP-1⁺ effector Tregs at the invasive margin and low Immunoscore with high frequency of BLIMP-1⁺ effector Tregs at the invasive margin. Statistical analysis was performed using a Log-rank test. *p < 0.05, **p < 0.01, ***p < 0.001. n = 32

Discussion

We have shown that the Immunoscore was able to predict disease-free survival in a cohort of New Zealand stage II colorectal cancer patients. Tregs were not associated with disease-free survival when identified using FOXP3 alone, however the frequency of effector Tregs (that simultaneously expressed BLIMP-1 and FOXP3) was associated with disease-free survival in our cohort. Furthermore, stratification according to tumour infiltration with effector Tregs at the invasive margin improved the prediction of disease-free survival in patients with a low Immunoscore (IS-0, IS-1 or IS-2). This study indicates that this novel T cell subset may be used to complement the Immunoscore to improve the ability to identify stage II patients with the highest risk of recurrent disease (low Immunoscore, low effector Treg infiltrate).

Previous research has associated tumour-infiltrating Tregs with both good and poor outcomes in colorectal cancer (reviewed in [23]). These studies often identified Tregs using FOXP3 alone. FOXP3 can be upregulated transiently on activated T cells and these activated FOXP3⁺ cells are unable to suppress proliferation or cytokine production [24]. The present study confirms that identification of Tregs using FOXP3 alone has limited predictive power and justifies the use of additional markers to further characterise and quantify Tregs.

BLIMP-1⁺ effector Tregs are a population of FOXP3⁺ Tregs characterised by the simultaneous expression of regulatory and effector-like molecules in mice. Cytokine-producing FOXP3⁺ Treg populations have been identified in human colorectal cancer and inflammatory bowel disease, but the role of BLIMP-1⁺ effector Tregs in patient outcome has not been investigated [18, 25]. This study has identified that colorectal cancer patients with a high frequency of BLIMP-1⁺ effector Tregs at the invasive margin are less likely to get recurrent disease than those with a low frequency of BLIMP-1⁺ effector Tregs.

A higher frequency of BLIMP-1⁺ effector Tregs in the tumour may be reflective of an environment rich in T cell activation. BLIMP-1 is upregulated in Tregs that have experienced T cell receptor stimulation in cytokine environments (reviewed in [14]). Therefore, a higher frequency of BLIMP-1⁺ effector Tregs may be a result of a highly activated T cell response, which is indicative of a good anti-tumour immune response [26]. This does not, however, exclude the possibility that BLIMP-1⁺ effector Tregs may have a direct role in the anti-tumour immune response by suppression of inflammatory responses.

The transcription factor, BLIMP-1, has been shown to be essential for IL-10 production by effector Tregs and other subsets of regulatory T cells in mice and humans [12, 15]. The role of IL-10 in colorectal cancer is currently poorly understood, but recent evidence has indicated that IL-10 may be associated with good anti-tumour immune responses in cancer (reviewed in [27]). Inflammatory responses, particularly local production of IL-17, in colorectal cancer have been associated with poor patient outcome (reviewed in [28]); however, it has been proposed that IL-10 may have a role in inhibition of the IL-17-mediated inflammatory responses [29]. In colorectal cancer, IL-10 production by T cells was inversely correlated with IL-17 production [30], although this was not directly analysed in the current study. It is possible that effector Tregs in CRC suppress inflammatory T cell responses via IL-10 production. It can be hypothesised, therefore, that higher frequencies of BLIMP-1⁺ effector Tregs in CRC infiltrates may suppress the inflammatory responses in the tumour and thereby improve outcome.

This study has several limitations. The pilot study used highly labour intensive methods and the cohort was small, and by design, was confined to stage II disease. A quarter of the original cohort could not be included because the available archival material did not include the invasive margin, and further patients with rectal cancer were excluded because they received pre-operative radiotherapy [20]. On the other hand clinical data collection and follow up were performed prospectively and were complete for all patients with a minimum of 5 years follow up. Only stage II CRC patients were chosen for this study because this is the group most likely to benefit from further prognostic partition [2, 3]. However, future validation studies should include patients with a range of disease stages. Finally, microsatellite-instability status was not available for these patients [6]. Despite the relatively small sample size, we were able to demonstrate highly significant recurrence risk stratification, according to the immune infiltrate. Thirteen cases with recurrent disease is sufficient for the accurate comparison of at least one variable, in this case immune infiltrate [31], nevertheless these data need to be validated in a larger cohort to enable multivariable analysis and inclusion of a broader range of disease stages.

Conclusion

This study validates the Immunoscore in a New Zealand cohort of stage II colorectal cancer patients and found that stratification according to FOXP3⁺BLIMP-1⁺ effector Treg infiltrate in patients with a low Immunoscore further improved the identification of patients at high risk of disease recurrence. This study raised the question of the functional role of BLIMP-1⁺ effector Tregs in CRC and whether these cells are merely prognostic markers or play a direct and influential role in disease outcome.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

AJCC

American Joint Committee on Cancer

BLIMP-1

B lymphocyte-induced maturation protein-1

CRC

Colorectal cancer

FFPE

Formalin-fixed, paraffin-embedded

HPF

High power field

ICC

Intraclass correlation coefficient

IRIISS

Independent Research Institute Infrastructure Support Scheme

IS

Immunoscore

NHMRC

National Health and Medical Research Council

Treg

Regulatory T cell

Compliance with ethical standards

Conflict of interest

The authors declare that no conflict of interest exists.