Colorectal cancer (CRC) is the fourth most common cause of death from cancer worldwide, accounting for 8% of all deaths from cancer.1 Five-year survival rates range from >90% among patients with stage I disease to <5% among patients with stage IV disease. The vast majority of recurrences occur within 5 years.2

Although survival depends strongly on the stage of the disease at diagnosis, the role of the immune system in patient survival has recently been shown in several studies. Pages et al3 found that CD3⁺CD45RO⁺ lymphocyte infiltration correlated directly with micro-invasive status. The same team then showed that the type, density and location of effector T cells in CRC provided more independent prognostic information than tumour extension.4 While memory T cell infiltration seems to be important for controlling disease progression and dissemination, the regulation of this spontaneous immune response at the tumour site is not fully understood.

During the last 5 years, T4reg cells have been shown to allow tumours to escape immune surveillance. Indeed, in ovarian cancer, T4reg infiltration correlates with poor outcome.5 Elevated proportions of T4reg have since been described in many other cancers,6 including CRC.7^–9 Many tumours have been shown to induce rapid expansion of T4reg in humans and mice, leading to failure to reject immunogenic tumours. Blocking Treg cell migration or function by means of immunotherapeutic approaches might therefore be beneficial.

CD8⁺CD3⁺ regulatory T cells have recently been identified.10 Some regulatory CD8⁺ T cells are reported to suppress immunity non-specifically,11^–13 whereas others are antigen-specific (viral, allogeneic and self antigen have been described).14^–16 CD8⁺CD25⁺Foxp3⁺ regulatory T cells (T8reg) were first described as a subset of single-positive CD8⁺ thymocytes sharing phenotypic, functional and mechanistic features with T4reg.11 T8reg were recently found by confocal microscopy in prostate cancer specimens.17 In this study authors showed that both CD4⁺ and CD8⁺ T cell subpopulations possessed potent suppressive activity. T cell cloning and fluorescence activated cell sorting analyses showed the presence of CD8⁺CD25⁺ Treg cell clones that expressed Foxp3. These CD8⁺ Treg clones (obtained ex vivo) suppressed naive T cell proliferation mainly through a mechanism that is dependent on cell contact. In this study the authors did not demonstrate that T8reg freshly isolated from a tumour have regulatory properties. Thus, their suppressive capacity at the tumour site remains unclear.

Here we extensively studied the number, proportion and phenotype of blood and tissue T4reg and T8reg cells in patients with colorectal adenocarcinomas. The aim of the study was to determine if T8reg could be found in patients with CRC, to assess the suppressive capacities of this new tumour-infiltrating T cell subset, and the cytokines involved in their induction.

MATERIALS AND METHODS

Subjects and samples

Fresh blood, tumour and normal autologous colonic tissues (normal tissue (NT)) were obtained from patients undergoing surgery for colorectal adenocarcinoma. Blood was collected before surgery. Normal autologous colonic tissue was obtained from a macroscopically normal part of the excised colon, at least at 10 cm from the tumour. None of the patients had received radiotherapy or chemotherapy. Blood from 26 normal donors was used for control experiments. Informed consent was obtained from each subject.

Lymphocyte extraction, count and cell sorting

Peripheral blood lymphocytes were isolated by Ficoll density gradient centrifugation. Normal and tumour-infiltrating lymphocytes were isolated from colonic tissue. Freshly isolated tissue was washed four times in wash medium (RPMI 1640 supplemented with penicillin and streptomycin), before being cut into small pieces with a blade. The specimen was then digested in RPMI 1640 medium containing 2% fetal calf serum (FCS) and 1 mg/ml type IV collagenase (Worthington, Freehold, New Jersey, USA) for 2 h at 37°C with 5% CO₂. A single-cell suspension was obtained by passing the digested tissue through a cell strainer.

In some experiments the suspension was used for ex vivo flow cytometric analyses. The single-cell suspension number was assessed using the Trypan blue exclusion test. Size and scatter were used to gate the lymphocyte populations and determine their proportions among all tissue cells. This allowed us to calculate the total number of lymphocytes by using the single-cell suspension number initially defined for each sample. We next normalised this total lymphocytes number on the weight of tissue (per gram). Then CD3⁺ cells were gated to determine the proportion of T cells among all lymphocytes. The absolute number of T lymphocytes could be then determined. Finally, cell proportions and numbers were determined among these CD3⁺ cells, CD8⁺CD25⁺Foxp3⁺ or CD4⁺CD25⁺Foxp3⁺.

For functional studies the suspension was sorted into CD4⁺CD25⁺, CD4⁺CD25⁻, CD8⁺CD25⁺ and CD8⁺CD25⁻ populations. The different T cell populations from cancer tissue were sorted on a BD FACSAria cell sorter (BD Biosciences, San Jose, California, USA). The purity of the cell preparations exceeded 97%. Purified CD4⁺CD25⁺, CD4+CD25⁻, CD8⁺CD25⁺ and CD8⁺CD25⁻ populations were then used for functional experiments.

Antibodies and FACS analyses

The following antibodies were used to stain single-cell suspensions ex vivo: CD3–FITC, CD3–PE, CD4–PERCP, CD8–FITC, CD8–PE, CD8–PERCP, CD25–PE, CD45RO–FITC, CTLA4–PC5, GITR–FITC, CCR7–FITC, CD28–PC5 and CD103–FITC, all produced by BD Biosciences (where CCR is CC chemokine receptor, CTLA-4 is cytoxic T lymphocyte-associated antigen-4, FITC is fluorescein isothiocyanate, GITR is glucocorticoid-induced tumour necrosis factor-related protein, PC5 is phycoerythrin-cyanin 5, PE is phycoerythrin, and PerCP is peridinin chlorophyll protein). Cell surface staining of latency-associated peptide (LAP; transforming growth factor (TGF)β1) was done using an anti-human LAP antibody (R&D Systems, Abingdon, UK) followed by FITC–goat–anti-mouse immunoglobulin (Pharmingen, Heidelberg, Germany) and allophycocyanin (APC)–Foxp3 murine monoclonal antibody, using the e-Bioscience kit as recommended by the manufacturer (e-Bioscience, San Diego, California, USA). Cells were analysed by flow cytometry with a FACS Calibur (Becton Dickinson, Oxford, UK). Data were analysed using CellQuest software version 5.1 or FlowJo software version 7.2.

³H-thymidine assay

Suppression experiments

The assay was performed in 96-well plates (Nalge Nunc, Rochester, New York, USA) coated with anti-CD3 (clone UCHT1, at 5 μg/ml; BD Biosciences). Cells were cultured in RPMI 1640 medium supplemented with 5% human AB serum, 2 mmol/l l-glutamine, 100 U/μg/ml penicillin/streptomycin, 0.5 mmol/l sodium pyruvate, 0.05 mmol/l non-essential amino acids (Gibco/Invitrogen, Cergy Pontoise, France) and soluble anti-CD28 (clone 28.2, at 5 μg/ml; BD Biosciences, Erembodegem Belgium). CD4⁺CD25⁻ or CD8⁺CD25⁻ responder cells were cultured at 4×10⁴ cells/well and variable numbers of regulatory cells were added. 3H-thymidine at 18.5 kBq per well was added for the final 16 h of a 5 day assay. In some experiments, supernatants were collected on day 2 for cytokine profiling.

Proliferation experiments

CD4⁺CD25⁻ and T4reg cells were cultured at 1×10⁵ cells/well and CD8⁺CD25⁻ and T8reg were cultured at 1×10⁴ cells/well in 96-well plates coated with anti-CD3. Interleukin 2 (IL2, 1000 IU/ml or 1× phosphate-buffered saline (PBS) was then added.

Protein extraction from normal and cancer tissues

Normal and cancer tissue specimen weights were determined before protein extraction with Tissue Protein Extraction Reagent (T-PER; Pierce, Rockford, USA) as recommended by the manufacturer. Briefly, 20 ml of P-TER was added to 1 g of tissue and homogenised. Samples were centrifuged at 10 000 g for 5 min and the supernatant (protein extract) was stored at−80°C until cytokine/chemokine profiling.

Cytokine/chemokine measurements

The LUMINEX kit was used to determine chemokine and cytokine profiles in tumour protein extracts and culture supernatants, according to the manufacturers’ recommendations (Linco, St. Charles, Missouri, USA/Biosource International, Camarillo, California, USA). The following cytokines and chemokines were determined: IL2, IL4, IL5, IL6, IL8, IL10, IL13, IL17, IFNγ, tumour necrosis factor (TNF)α, TGFβ1, fractalkine (FTK), interferon γ induced protein-10 (IP)-10, monocyte chemoattractant protein (MCP)-1, macrophage inflammatory protein (MIP)-1α, MIP-1β. Concentrations of IL6 and TGFβ1 in tumours were confirmed with ELISA kits according to the manufacturer’s recommendations (R&D Systems).

Analysis of mismatch repair status by immunohistochemistry

Representative samples from adenocarcinoma and normal mucosa adjacent to the tumour were selected in each case and paraffin sections were sent to an automated immunostainer. Tissue sections were incubated in citrate buffer with the following antibodies: G168–728 antibody (recognises human Mut L homolog 1 (hMLH1) antigen; the antibody was diluted 1:40; Pharmingen, San Diego California, USA), FE11 antibody (recognises human Mut S homolog 1 (hMSH2) antigen; the antibody was diluted 1:25; Calbiochem, Cambridge, Massachusetts, USA), all in a boiling water bath. An avidin–biotin complex (ABC Vectastain Kit; Vector Laboratories, Burlingame, California, USA) was used to reveal the antigen. Positive controls included slides from tumours with normal expression of hMLH1 and hMSH2; negative controls included slides with no primary antibodies. Two observers assessed all cases independently, and cases with discrepancies were further evaluated until agreement was reached between the observers.

Ex vivo expansion of T8reg

CD8⁺ T cells were isolated from PBMCs of normal volunteers by using CD8 microbeads (Miltenyi Biotech, Malmä, Sweden). After counting, CD8⁺ T cells were plated in 96-well plates (100 000 cells/well) in a final volume of 200 μl of AIMV (Gibco). Dynabeads CD3/CD28 T cell expander (1 bead per T cell; Invitrogen, Dynal, Cergy Pontoise, France) and IL2 at 10 IU/ml (proleukin; Novartis, Rueil-Malmaison, France) were added to each well. Recombinant human IL6 and/or TGFβ1 (R&D Systems) was added at final concentrations of 1 ng/ml. On day 5 of culture, T cells were tested for Foxp3 expression by using the APC–Foxp3 murine monoclonal antibody according to the manufacturer’s recommendations (e-Bioscience). Cells were then analysed by flow cytometry.

Statistical analysis

All results are expressed as means with the standard error of the mean (SEM) or as ranges when appropriate. Because the variables studied were not normally distributed, nonparametric statistical methods were used. The Wilcoxon two-sample rank sum test was used to compare continuous variables between groups. Associations between two continuous variables were evaluated with the Spearman rank correlation method. A paired t test was used to compare results for normal and tumour tissue. Statistical analyses were performed using Prism 5 software (GraphPad version 5.0). p Values of <0.05 were considered significant.

RESULTS

Patients’ baseline characteristics

Between November 2005 and November 2007, forty never-treated patients (25 men, 15 women) with colorectal adenocarcinomas were enrolled in this study. Their baseline characteristics are summarised in table 1. Mean age was 65 years. Tumour stages II and III were most frequent, as expected, and at least one item of the VELIPI micro-invasion score (venous embolism, lymphatic invasion, perineural invasion) was found in 53% of cases.

Numbers of CD8⁺CD25⁺Foxp3⁺ T cells with a phenotype similar to CD4⁺CD25⁺Foxp3⁺ T cells are increased in the blood of patients with CRC

As shown in fig 1A, the percentage of T4reg was significantly higher in the blood of patients with CRC than in normal volunteers (5.2%±1.8% vs 3%±0.9%, p<0.05), and so was the percentage of a CD8⁺CD25⁺Foxp3⁺ T cell population (T8reg) (0.45%±0.3% vs 0.22%±0.1%, p<0.05) (fig 1B). The phenotype of these T8reg was similar to that of T4reg, with strong expression of IL2Rβ (CD25) and weak expression of IL7Rα (CD127), a memory phenotype, as shown by CD45RO expression and intracellular staining for regulatory molecules such as glucocorticoid-induced TNF-related protein (GITR) and cytotoxic T lymphocyte-associated antigen-4 (CTLA-4) (fig 1C). Moreover, T8reg were CD28⁺ and CCR4⁺.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

The proportion of regulatory T8 (T8reg) cells is elevated in the blood of patients with colorectal cancer (CRC), and the T8reg phenotype is very similar to that of T4reg. (A) The proportion of CD4⁺CD25⁺Foxp3⁺ cells among CD4⁺ T cells was determined by flow cytometry of peripheral blood mononuclear cells (PBMCs). Means with the SEM are shown. (B) The proportion of CD8⁺CD25⁺Foxp3⁺ cells among CD8⁺T cells was determined by flow cytometry of PBMCs. Means with the SEM are shown. (C) Cells were gated on CD8⁺Foxp3⁺ cells and expression of CD25, CD28, CD45RO, CCR4, CD127, GITR and CTLA-4 was examined. *p<0.05. CCR, CC chemokine receptor; CTLA-4, cytotoxic T lymphocyte-associated antigen-4; GITR, glucocorticoid-induced tumour necrosis factor-related protein.

CD8⁺CD25⁺Foxp3⁺ T with an activated phenotype are increased in colorectal cancer tissue

To examine the presence of these regulatory populations in situ, 32 tumours and distant autologous normal colon specimens were dissociated. The CD3⁺ lymphocyte yield was 0.03×10⁶ to 5.7×10⁶ cells and 0.3×10⁶ to 19×10⁶ cells/g of normal and tumoral colon, respectively. All descriptive statistics are summarised in table 2. As shown in fig 2A,B, the proportion and number of T8reg and T4reg were markedly increased in the tumours as compared to normal tissue. A significant number of T8reg was found in only 34% (12/32) of normal tissue samples but in 96% (31/32) of tumour specimens. We then investigated the expression of regulatory molecules and the secretion of inhibitory cytokines which are known to be involved in the suppressive capacity of regulatory T cells. In the tumours, T8reg had an activated phenotype with significantly increased membrane expression of GITR and CTLA-4 (fig 2C) as compared to normal tissue. Moreover, a significant proportion of intraturmoral T8reg cells could produce TGFβ1 but not IL10, as shown by ex vivo intracellular staining (fig 2D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Regulatory T8 (T8reg) cells are enriched in tumour tissue and have an activated phenotype. (A) Cells from tumour tissue (colorectal cancer; CRC) and normal tissue (NT) were gated on viable cells and double positive CD8⁺CD3⁺ T cells. Then, the proportion of CD25⁺Foxp3⁺ was determined (left panel). The T8reg number (right panel) was determined using mononuclear cell number (obtained after tissue dissociation) per gram of tissue (freshly weighed before dissection). (B) As in (A), but for T4reg. (C) The proportion of GITR⁺ (right panel) and CTLA-4⁺ (left panel) cells among T8reg was determined in CRC and normal tissue. (D) The proportion of TGFβ1⁺ (right panel) and IL10⁺ (left panel) cells among T8reg was determined in CRC and normal tissue. Means with the SEM are shown. *p<0.05, **p<0.01; ***p<0.001. NS, not significant. CTLA-4, cytoxic T lymphocyte-associated antigen-4; GITR, glucocorticoid-induced tumour necrosis factor-related protein; IL, interleukin; TGF, transforming growth factor.

CD8⁺CD25⁺Foxp3⁺ cells correlate with tumour stage and micro-invasive status

We then studied the possible relation of T8reg with the cancer stage and micro-invasive status. As shown in fig 3A, the ratio of T8reg (fig 3A left panel), and T4reg (fig 3A right panel) correlated positively with the tumour stage (AJCC, 6th edition), suggesting that these cells accumulate during tumour progression. However, only T8reg (fig 3B, left panel), but not T4reg (fig 3B, right panel), were significantly increased in tumours with at least one sign of micro-invasion (venous embolism and/or lymphatic invasion and/or peri-neural invasion, ie, VELIPI positive) as compared with tumours with no signs of micro-invasion.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Tumour-infiltrating regulatory T8 (T8reg) cells correlate with tumour stage and VELIPI status. (A) The ratio of T8reg cells in the tumour correlated positively with the tumour stage (left panel, rs = 0.52 (p<0.0022)). As for T8reg cells , the ratio of T4reg cells correlates with the tumour stage (right panel, rs = 0.46 (p<0.01)).The ratio between colorectal cancer tissue (CRC) and autologous normal tissue (NT) was calculated as %Treg CRC/%Treg NT. (B) The mean (with the SEM) of Treg in VELIPI-negative patients was compared to the mean (with the SEM) of Treg in VELIPI-positive patients (T8reg (right panel) and T4reg (left panel)). *p<0.05, **p<0.01. NS, not significant; VELIPI, venous embolism, lymphatic invasion, perineural invasion.

CD8⁺CD25⁺Foxp3⁺ and MMR status

Immunohistochemical staining to determine MMR status showed six patients with microsatellite instability (MSI) and 26 were considered to have a stable phenotype (MSS). In our cohort of patients, a trend for a higher proportion of T4reg and T8reg in MSS patients was observed. However, since the population studied was small, with only six MSI positive patients, there were no significant differences between these groups (T4reg ratio CRC/NT was 5.76 (SEM 5.07) vs 3.95 (SEM 1.2) for MSS and MSI status, respectively, p = 0.56; and T8reg ratio CRC/NT was 11 (SEM 12) vs 6.79 (SEM 5) for MSS and MSI status respectively, p = 0.89).

T8reg isolated from the cancer tissue can suppress autologous CD4⁺CD25⁻ T cell proliferation and Th1 cytokine production ex vivo

To determine the suppressive capacity of T8reg infiltrating CRC, we sorted CD8⁺CD25^hi T cells from CRC tissue from eight patients. As shown in fig 4A, CD8⁺CD25^hi T cells were mainly Foxp3⁺. These sorted T8reg cells were able to suppress CD4⁺CD25⁻ T cell proliferation (fig 4B, left panel) as T4reg (fig 4B, right panel). Moreover, T8reg, like T4reg, efficiently suppressed production of the proinflammatory cytokines IL2 and IFNγ by CD4⁺CD25⁻ T cells from tumour tissue (fig 4C). T8reg were more efficient than T4reg on CD4⁺CD25⁻ T cell proliferation and Th1 cytokine production. Finally, like T4reg, these T8reg were anergic at the tumour site and did not respond to IL2, contrary to effector T cells (fig 4D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Tumour regulatory T8 (T8reg) cells have suppressive capacity ex vivo. (A) Foxp3 expression was determined in CD8⁺CD25^hi T cells from CRC tissue. (B) Freshly sorted tumoral CD4⁺CD25⁻, T4reg and T8reg cells were used in suppression assays. Briefly, proliferation of CD4⁺CD25⁻ cells was assessed with and without T8reg (left panel) and T4reg (right panel) at several ratios (see figure). (C) As in (B), but for interleukin (IL)2 (left panel) and interferon (IFN)γ (right panel) secretion. (D) The proliferative capacity of tumoral CD4⁺CD25⁻ (Eff CD4⁺), CD8⁺CD25⁻ (Eff CD8⁺), T4reg and T8reg cells was determined in the presence of 1000 IU/ml recombinant human IL2 (black histograms) or an equivalent volume of phosphate-buffered saline (PBS; white histograms). The experiments were performed twice, with similar results. Means with the SEM are shown. *p<0.05, **p<0.01.

T8reg correlate with IL6 and TGFβ1 and can be induced ex vivo by these cytokines

To determine if T8reg could be recruited from the periphery by local chemokine production, we determined chemokine concentrations in normal and cancer tissues. As shown fig 5A, IL8, MCP-1, MIP-1α and fractalkine levels were significantly higher in cancer tissue than in normal tissue; but none of these chemokines correlated with the number of T8reg. As T8reg express CCR4⁺, a chemokine receptor involved in Treg recruitment to inflammatory tissues, they might be recruited from the periphery by local production of CCR4 ligands such as CC chemokine ligand (CCL)17 and 22.18 Intra-cellular staining showed the presence of these chemokines within tumour tissue, but also in normal tissue (data not shown), where T8reg are relatively scarce.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Regulatory T8 (T8reg) cells are enriched at tumour sites and can be derived from CD8⁺CD25⁻Foxp3⁻ cells ex vivo. (A) Cytokine profiling in colorectal cancer tissue (CRC) and normal tissue (NT). Only significant differences are shown. Values are means with the SEM. (B) The number of T8reg cells in the tumour correlated positively with the interleukin (IL)6 concentration (left panel, rs = 0.47 (p<0.0252)) and the transforming growth factor (TGF)β1 concentration (right panel, rs = 0.52 (p<0.0123)). (C) In vitro differentiation of T8reg from CD8⁺ T cells of normal volunteers; a representative dot plot is shown in the left panel and means with the SEM are shown in the right panel. The experiments were performed three times with similar results. *p<0.05, **p<0.01; ***p<0.001.

As T4reg can be induced from naive T cells by various stimuli,19^–22 we studied whether T8reg were associated with specific cytokines in tumour tissue. As shown in fig 5A, the levels of TNFα, IL6, TGFβ1 and IP-10 were significantly increased in tumour tissue, while IL2, IFNγ and IL17 were lower in cancer tissue than in normal tissue. Only IL6 and TGFβ1 levels correlated positively with T8reg numbers (fig 5B). This correlation, together with the lack of increased CCR4 ligands expression in tumour tissue, is compatible with in situ differentiation rather than recruitment to the tumour site.

We then harvested CD3/CD28 cross-linked CD8⁺ T cells from normal volunteers, in the presence of IL6, TGFβ1 or both, ex vivo. As shown in fig 5C, IL6 significantly increased the proportion of CD8⁺Foxp3⁺ T cells after 5 days of culture. Moreover, IL6 and TGFβ1 acted synergistically on CD8⁺Foxp3⁺ induction in vitro (fig 5C).

DISCUSSION

Immune cells infiltrating colorectal tumours appear to have a major role in tumour control, and their activity correlates with a good prognosis.3 4 23 24 However, tumours can become invisible or inaccessible to cytotoxic effector cells. Indeed, even if tumours are sites of inflammation, the proportion of efficient infiltrating effector cells is low. The three main mechanisms underlying tumour immune escape are (1) tumour expression of molecules such as B7 homologous (B7H)1, B7H4 and the FAS ligand,25 26 which can induce T cell anergy and/or apoptosis; (2) tumour-derived immunosuppressant molecules (TGFβ, IL10, colony stimulating factor (CSF)-1 and vascular endothelial growth factor (VEGF)),27^–29 which can diminish effector cells responses; and(3) tumour-induced suppressive immune cells such as Foxp3⁺ regulatory T cells.30

Here we extensively studied blood and tissue Foxp3⁺ T cells in 32 patients with CRC and compared them with Foxp3⁺ T cells from distant normal colonic tissue and with blood Foxp3⁺ T cells from healthy volunteers. We found, as already described, a large proportion of T4reg with suppressive capacities and an activated phenotype in tumour tissue from patients with CRC. We also detected a new CD8⁺CD25⁺Foxp3⁺ T cell subset (T8reg) that is more abundant in the blood and especially the tumour tissue of patients with CRC. These T8reg were absent from most normal tissue samples but present in more than 90% of CRC specimens. However, their absolute numbers in the tumours was low, and their clinical relevance is not known. These cells were Foxp3⁺ and CD28⁺, thus differing from CD8⁺CD28^–Foxp3^– regulatory T cells recently described in various human cancers by Filaci et al31 and Filaci and Suciu-Foca.32 Similar CD8⁺CD25⁺Foxp3⁺ T cells have been identified in various settings,11 33 34 and notably in a very recent study of prostate cancer specimens.17 However, in this last study, while the suppressive capacity of CD8⁺CD25⁺Foxp3⁺-derived T cell clones was demonstrated, the suppressive capacity of freshly sorted tumour-infiltrating CD8⁺Foxp3⁺ T cells was not shown, and no ex vivo phenotypic or functional characterisation was performed.

CRC T8reg were mostly GITR⁺ and CTLA4⁺ and expressed TGFβ1, suggesting they were activated and had regulatory functions. Suppression experiments performed after cell sorting of CD8⁺CD25^hiFoxp3⁺ T cells from tumour specimens immediately after surgery showed that they could suppress CD4⁺CD25^– T cell proliferation and Th1 cytokine secretion. In addition, like T4reg, T8reg were anergic at the tumour site, while effector T cells remained capable of responding to IL2. Altogether, these data demonstrate for the first time that a new CD8⁺CD25⁺Foxp3⁺ T cell subset with an activated phenotype and suppressive functions is present in malignant colorectal tumours.

Three previous studies have examined T4reg in patients with CRC; these cells were relatively abundant in the cancer tissue, but they did not correlate with the tumour stage or disease outcome.7^–9 In contrast, we found that T8reg correlated with the stage and micro-invasive status of CRC, two major prognostic factors in this disease.

A recent study suggested that higher tumoral expression of Foxp3, IL17, IL1β, IL6 and TGFβ1 was associated with the mismatch repair (MMR)-proficient (MSS) phenotype. In our cohort of patients, we found only six with an MSI phenotype tumour and 26 with an MSS phenotype tumour. Even if a trend for a higher proportion of T4reg and T8reg in MSS patients than in MSI patients was observed, since the studied population was small, no significant differences were observed between these groups. Moreover, in the study by Le Gouvelo and colleagues,35 the authors showed an increase of IL17 mRNA in MSS tumours. In our work, IL17 protein was not increased in tumour tissue as compared with normal homologous tissue considering the whole population (fig 5A) as already described by others.36 However, when comparing IL17 regarding MMR status of the tumour, none of the MSI patients had detectable IL17 levels (normalised from the number of CD3⁺ T cells per gram of tissue) in their tumours. The ratio of IL17 (CRC/NT) was superior to 1 in 8 out of 26 MSS patients and was equal to 1 or lower in all MSI patients (data not shown). Altogether these data may be in accordance with the findings by Legouvelo and colleagues35 but the small sample size does not allow us to reach definitive conclusions.

The mechanisms involved in the accumulation of T4reg in cancer tissue are poorly understood. Both recruitment5 and in situ differentiation29 30 37 38 may be responsible. The conversion of naive CD4⁺ T cells into T4reg cells can be mediated by TGFβ139 and regulatory dendritic cells.29 30 40 41 Although T8reg cells express CCR4, no increase in CCR4 ligand expression – which could support local recruitment of these cells – was found in CRC specimens as compared to normal colon specimen (data not shown). Finally, the numbers of T8reg cells were very low in the blood of patients with CRC. These data argue for in situ differentiation rather than recruitment from blood, but this remains to be demonstrated unequivocally. Multiplex analysis after protein extraction showed that several cytokines were increased in CRC specimens compared to normal tissue, but that only IL6 and TGFβ1 levels correlated with T8reg infiltration. In vitro experiments showed that IL6 and TGFβ1 added to the culture medium synergistically induced T8reg differentiation from blood CD3⁺CD8⁺ T cells. In our hands, CD3/CD28 cross-linking and IL6 alone induced a level of conversion similar to that found in cancer tissues.

IL6 has been shown to induce Th17 cells from TGFβ1-treated CD4⁺CD25⁻ T cells.42 We observed no IL17 production by IL6/TGFβ1-treated CD8⁺ T cells, as described by Liu et al,43 and IL17 expression was not increased in CRC tumour specimens (data not shown). However, we used an IL6 concentration of 1 ng/ml, whereas Liu and colleagues used 100 ng/ml. Moreover, we were unable to induce T8reg differentiation with higher concentrations of IL6 (10 and 100 ng/ml; data not shown). Because we had no specific marker allowing us to isolate this IL6/TGFβ1-induced CD8⁺CD25⁺Foxp3⁺ population, we could not demonstrate their suppressive capacity and must therefore remain cautious regarding their functionality, as far as Foxp3 upregulation on cultured lymphocytes is not necessarily linked to suppressive functions.44 The correlations observed here between the number of T8reg and IL6 and TGFβ1 expression could be due to production of these cytokines by T8reg themselves. However, contrary to TGFβ1, we observed no production of IL6 by T8reg ex vivo (data not shown).

IL6 has been implicated in cancer genesis and invasiveness and is frequently associated with a poor prognosis in cancer patients.45^–49 If IL6 participates to the induction of T8reg with suppressive capacities in CRC tissue, this would represent another mechanism implicated in the deleterious role of IL6 in cancer patients. However, other experiments are needed to fully evaluate this hypothesis.

In conclusion, we show the involvement of a new regulatory CD8⁺CD25⁺Foxp3⁺ T cell subset in human cancer. These T8reg cells have an activated phenotype and suppressive functions and accumulate within colorectal cancer tissue. They can be induced by IL6, the action of which is potentiated by TGFβ1. However, T8reg cells represent only a small fraction of CD8 T cells in vivo, and their origin and clinical relevance remain to be determined. The different factors found here to be involved in Treg conversion, ie, IL6 and TGFβ1, could synergistically worsen the prognosis of CRC. Strategies targeting both IL6 and TGFβ1 might help to counteract immune lethargy in patients with colorectal cancer in the future.