In the past decade various phenotypes and functions have been allocated to distinct Treg subsets. These populations include both, CD4⁺ as well as CD8⁺ cells and even CD4⁻CD8⁻ double negative (DN) variants [25]. Based on their ontogeny, Tregs have been dichotomized into two major, from the genomic standpoint distinct, subsets [26]: the natural form (nTregs) arising from the thymus and subsequently populating the periphery and the induced cells (iTregs), which are generated by a conversion of Tconv. This peripheral transformation occurs under various conditions and most times requires a preceding MHC-peptide interaction. Due to the involution of the thymus with age Treg output declines over time but still remains detectable [27].

To date naturally occurring CD4⁺CD25⁺FOXP3⁺ Tregs (nTregs) have been the most extensively studied subset. Characteristically they constitutively express CD25, glucocorticoid induced TNFR family-related protein (GITR) and cytotoxic T lymphocyte antigen (CTLA-4), which are all under the control of FOXP3. Two recent studies demonstrated that the IL-7 receptor α-chain (CD127) could be useful for the discrimination of nTregs from activated Tconvs [28,29]. Since FOXP3 expression correlates inversely with CD127, nTregs consequently depict no or low CD127 levels on their surface. Thus combination of CD127 with CD25 expression (CD3⁺CD4⁺CD25⁺CD127^low/neg) allows a more elaborate identification and even purification of viable Tregs.

The cytokine composition of the milieu, in which Tconvs receive their activating stimuli, plays a decisive role for iTreg generation. A repetitive, antigen-dependent stimulation of naïve or memory CD4⁺ Tconvs in presence of primarily IL-10 (but also vitamin D, dexamethasone, complement factor CD3b and CD4b) leads to the generation of mainly IL-10 producing FOXP3⁺ suppressive cells (T regulatory cells 1; Tr1) [30-32]. Similarly the presence of TGF-β results into the induction of cells that produce high amounts of TGF-β (T helper cells 3; Th3) and can express FOXP3 but not as stably as nTregs [26,33,34]. Once activated by a specific antigen, iTregs do not require additional re-stimulation and suppress in an antigen-independent fashion. This characteristic led to the term “bystander suppression”. The TCR repertoire of iTregs is congruent with T4convs and partly overlaps with nTregs [35]. Other (tumor derived) molecules that have been linked to iTregs include prostaglandin E2 (PGE₂), indoleamine 2,3-dioxygenase (IDO) and Galectin-1 [32]. Triggering the TCR of naïve CD45RA⁺ CD4⁺ Tconvs in presence of TGF-β and/or IL-2 results in cells that are highly suppressive and by expressing FOXP3, CD25, CTLA-4 and GITR resemble the phenotype of nTregs [34].

The quality of T cell stimulation is another important determinant for skewing immune responses towards reactivity versus tolerance. Suboptimal, non-adequate antigen-presentation and/or co-stimulation by defective (tumor-associated) antigen presenting cells (APCs) as it applies for plasmacytoid DCs [36] or so-called myeloid derived suppressor cells found in patients with ovarian and hepatocellular cancer, promotes iTregs [37].

In analogy to their CD4⁺ counterparts, there exist thymic CD25⁺FOXP3⁺CTLA-4⁺ as well as adaptive CD8⁺ Tregs. They similarly exert immune control by cell-to-cell contact, IL-10, TGF-β and CTLA-4 [36,38]. In a large series of patients with various types of malignancies suppressive IL-10-producing CD8+CD28 cells could be isolated [39] and cells from prostate cancer patients expressing CD28, CD25 and FOXP3⁺ suppressed T cells in a contact-dependent fashion [40].

The DN cells compose 0.8%–1% of the total peripheral T cells and have been shown to suppress in an antigen-specific and dose-dependent manner [41]. Uniquely DN cells exhibit a targeted cytotoxicity against syngeneic CD8⁺ T cells of the same TCR-specificity. Furthermore and in addition to cells with an α/β TCR, CD25^-FOXP3^- γ/δ Tregs that inhibit T cells as well as DCs have been reported [42].

In the past years the modes of Treg-mediated suppression have been diligently studied. A broad repertoire of mechanisms has been discovered, which affect lymphocytes, monocytes and DCs and encompass cell-to-cell contact as well as soluble factors. The importance of cell-to-cell contact is highlighted by the fact that the suppressive activity is abolished in vitro once semi-permeable membranes separating Tregs from effector cells are introduced. This observation is furthermore supported by the critical role of the LFA-1/ICAM-1 interaction [43]. On the other hand, IL-10 and TGF-β have been shown to be of substantial importance for the creation of a tolerogenic cytokine environment [44-46], but also for the full suppressive potency of Tregs. In animal studies Tregs derived from IL10^−/− and TGF-β^−/− mice displayed a substantial lack of suppressiveness. Interestingly Tregs utilize TGF-β not only in its released form to blunt T and NK cell responses, but also bound to their membrane as shown in a study on patients with gastro-intestinal stromal tumors [47]. Of note iTregs isolated from patients with colorectal and head and neck cancer, both inflammation driven malignancies, suppress T cells also in a PGE₂ dependent manner [48,49].

Recent reports state that IL-35 plays a role for the development as well as (maximal) suppressive activity of Tregs [50,51]. IL-35 is a novel member of the IL-12 heterodimeric cytokine family. A constitutive IL-35 expression could not be found in human nTregs and these findings still require further refinements [52].

It is well established that NK as well as T cells lyse infected or transformed (pre-/malignant) cells by the perforin/granzyme pathway. Astonishingly Tregs are capable of killing effector cells in a uniform fashion [53]. In addition, involvement of Fas ligand, TRAIL-DR5 (tumor necrosis factor-related apoptosis inducing ligand/death receptor) and galectins is suggested as part of this “control-by-killing” strategy [54,55].

Apart from their direct suppressive effects on cells, Tregs are able to modulate the physiological activation of T cells. Very sophisticated in vivo models showed that Tregs attenuate the stability of immunological synapses formed between T cells and APCs in the lymph nodes [56]. The affinity of CTLA-4 on Tregs for co-stimulatory CD80/CD86 molecules on APCs exceeds that of CD28 on Tconvs. Binding of CTLA-4 on APCs subsequently leads to a CD80/CD86 downregulation and thereby promotes tolerance instead of reactivity [57]. In addition CTLA-4 mediates an upregulation of indoleamine 2,3-dioxygenase (IDO) in APCs. This molecule metabolizes tryptophan and generates a substantial amount of reactive oxygen species (ROS). Depletion of tryptophan impedes activation and proliferation of T cells while the simultaneously produced ROS exert cytotoxic effects [58]. Recently identified surface molecules important for the Treg-mediated suppression include the lymphocyte activation gene-3 (LAG-3; CD223). It impairs the maturation and immune stimulatory capacity of immature DCs by binding their MHC class II [59]. Neuropilin-1 (Nrp1) on the other hand prolongs the contact (modulation) time of Tregs to DCs. It thereby also restricts the access of effector T cells to the immunological synapse [60].

The suppressive repertoire of Tregs includes a collection of very distinct mechanisms that are summarized in the term metabolic disruption. Extra-cellular adenosine triphosphate (ATP) acts physiologically as a major danger signal. Upon tissue destruction intracellular ATP-reservoirs are released and mediate inflammatory responses through purinergic receptors on immune cells [61]. Tregs express the ectoenzymes CD39 and CD73, which cleave ATP. Thereby they (A) remove the pro-inflammatory stimulus and (B) generate immunosuppressive adenosine. Adenosine suppresses, via A2A-receptors activated lymphocytes, promotes iTregs [49,62] and hampers immunogenicity as well as maturation of DCs [63]. In contrast to regular lymphocytes Tregs have high intracellular levels of the inhibitory second messenger cyclic adenosine monophosphate (cAMP), which they can inject into activated effector T cells leading them to anergy [64]. The role of the high CD25 expression on Tregs has been fiercely discussed. CD25 being a component of the IL-2 high affinity receptor (together with CD122 and CD132) leads to the speculation that Tregs may consummate/deplete the local IL-2. This could lead to IL-2 starvation and finally to Bim-mediated apoptosis of (activated) Tconvs [65]. Nevertheless these hypotheses are challenged by the unaltered suppressive function of Tregs upon antibody-mediated blocking of CD25 [66]. Very recent findings describe that Tregs also interfere at the level of the very sensitive redox-metabolism of immune cells. The consumption of thiols and inhibition of the glutathione synthesis in DCs and T cells result in functional alterations as well as in an increased susceptibility towards ROS mediated cytotoxicity [67,68]. These findings are of special importance taking into consideration that cancer and inflammation are associated with oxidative stress and Tregs on their part depict an increased resilience towards ROS [69,70].

Accumulation of Tregs has been described in cancer patients for the tumor microenvironment- as well as systemically [32]. As yet in vitro as well as in vivo observations have revealed a complex underlying system that comprises recruitment, expansion and de novo generation of Tregs (Figure 1).

In cancer patients, CCR4⁺ Tregs show a re-directed trafficking towards malignant tissue following a CCL22 chemokine gradient. This was firstly demonstrated in ovarian cancer patients. Interestingly in these patients CCL22 was not produced by the tumor, but released by the “bystanding”, tumor-associated macrophages (TAMs) [23]. In the follow several chemokine receptors were identified on Tregs, among them CCR2, CCR4, CCR5, CCR7, CCR8 and CXCR4. These receptors enable n/iTregs to migrate towards tumors in response to CCL2, CCL5, CCL17, CCL22 and CXCL12 [32,71-74].

Decreased TCR excision circles and an elevated proportion of Ki67⁺ cells strongly indicate an increased turnover of Tregs in cancer patients [74,75]. Several tumors have been shown to release high amounts of TGF-β and/or to induce bystander cells (e.g., MDSCs and immature DCs). These bystander cells represent additional sources of cytokines (e.g., TGF-β, IL-10, PGE₂ and others) with a fundamental role in Treg expansion as well as maintenance of suppressive features [33,76-78].

Furthermore, and as previously noted, Tregs can be generated from naïve or memory CD4⁺ and CD8⁺ Tconvs. Conditions that favor such a transformation are regularly found in cancer. TGF-β is the key cytokine in such a process while IL-10 holds the secondary role especially for antigen-stimulated cells [78,79]. Other contributing molecules are heme-oxygenase 1 (HO-1), cyclooxygenase 2 (COX-2), H-Ferritin, IDO [58, 81-83] and of course defective APCs [58,80-82].

Altogether it becomes evident that an accumulation of suppressive Tregs results from complex multistep and multifactorial processes, which need to be diligently explored and holistically analyzed in order to develop efficient modes of clinical interventions.

The paradigm of Tregs being unequivocally detrimental in cancer has been mainly challenged by the pivotal work performed in pre-clinical CRC models [93]. At first an increased susceptibility for inflammatory mucinous cancer was seen in IL-10^−/− mice [94]. In APC^Min/+ mice, in which underlying genetic defects of the β-catenin and wnt-signaling pathway lead to cancer formation [95], adoptive transfer of Tregs and IL-10 administration proved to have protective and even therapeutic effects [96,97]. Furthermore, anti-inflammatory treatment with recombinant IL-10 or neutralizing TNF-α antibodies led to a significant numerical reduction and redistribution of Tregs from the periphery into the centre of regressing malignant lesions [98,99]. As previously mentioned TGF-β is crucial for upregulation and stability of FOXP3. At the same time it positively regulates RORγt, the signature TF for pro-inflammatory Th17 cells that have been related to colitis-induced cancer and the promotion of CRC initiating cells [26,100,101]. Generally FOXP3 inhibits RORγt and the thereby promotes the generation of “classical”, suppressive Tregs [102,103]. However in presence of strong inflammatory stimuli, namely IL-6 and TNF-α, FOXP3 is downmodulated and cell differentiation skewed towards the Th17 pathway. It is speculated that persistent inflammation thereby forms a positive, self-amplifying loop (Figure 2). Suppressive Tregs are transformed into Th17 cells overriding any balance between anti- and pro-inflammatory activities and increasing the risk for a malignant transformation [102,103]. Recent studies suggest that mast cells (MC) are an active partner in this diversion of Tregs during polyposis coli and CRC [104,105]. Interestingly, Blatner and colleagues showed that generation of pro-inflammatory Tregs as well as MC-Treg crosstalk in CRC are both independent of IL-6 and IL-17 [105]. These findings suggest the existence of an alternative, to the classical Th17 conversion, transformation of anti- to pro-inflammatory Tregs. For the future it will be very interesting to investigate the therapeutic potential of such MC-Treg interaction in CRC.

One of the first studies in CRC patients showed that peripheral blood derived CD4⁺CD25⁺ T cell lines inhibited via TGF-β cytotoxicity and proliferation of autologous HLA-A1 restricted CD4⁺ CTLs [106]. An early quantitative analysis of circulating CD4⁺CD25⁺ Tregs in patients with epithelial cancers, including nine CRC cases, revealed increased frequencies. These cells expressed CTLA-4, produced TGF-β and impaired both, T and NK cell functions [107]. In the follow several studies confirmed the findings on the elevated proportions of CD4⁺ Tregs in peripheral blood [108-111].

Initial analyses of TILs evidenced increased levels of CD4⁺ Tregs compared to autologous healthy tissue without any clear correlation to disease stage [108,111,112]. Tregs infiltrated predominantly the tumor stroma resulting in a reversal of the stroma/epithelium ratio seen in healthy tissue [112]. The number of Tregs was two-fold higher in limited as compared to metastatic disease. This observation led to the not yet verified speculations that Tregs may migrate from the primary lesion towards the metastases during disease spreading. The positive correlation of stromal DCs with infiltrating Tregs implies a potential link between both populations [113]. In line with the other examined compartments CD4⁺CD25^highFOXP3⁺CTLA-4⁺ Tregs were also increased in the tumor draining mesenteric lymph nodes (TDLNs) [110]. Noticeably Tregs from TDLNs can express COX-2 by which they suppress IFN-γ as well TNF-α production of peripheral T cells; thus providing an additional argument in favor of COX-2 inhibition in CRC patients [48].

Several groups have investigated the impact of Tregs on the specific immune responses against tumor-associated antigens (TAAs) in CRC. The grade of local infiltration did not correlate with responses against well-defined TAAs as EpCAM, Her-2/neu and CEA [113]. Depleting Tregs in PBMCs from CRC patients dramatically boosted the IFN-γ and TNF-α production in T cells, which were stimulated with a CEA peptide [48]. In spite the unmasking of responses against several other TAAs, recall antigens like PPD and HA were not affected suggesting a TAA-specific rather than a systemic immune suppression [110]. In a very comprehensive analysis various TAA-specific Tregs were exclusively identified in CRC patients. Peptides for CEA, telomerase, Her-2/neu and MUC-1 all led to an activation of Tregs [114]. TAA-specific Tr1 cells were successfully identified using a p53 peptide [115]. In addition to CD4⁺ Tregs also CD8⁺CD28^- Tregs could be isolated from peripheral blood, tumor tissue and metastatic lymph nodes of CRC patients [39]. These cells suppressed T cells in an IL-10 dependent fashion and were mainly CCR4⁺, which may have contributed to their accumulation via recruitment. A recent study identified circulating and tumor infiltrating CD28⁺ CD8⁺ Tregs with a CD25⁺, FOXP3⁺, CTLA-4⁺, GITR⁺, CCR4⁺, TGF-β⁺ and CD127^low/neg phenotype [111]. Remarkably this type of Tregs was found in 90% of the CRC specimens but was totally absent in normal colonic tissue suggesting a cancer-specific presence without contribution to the physiologic epithelial homeostasis [90]. Ligands for CCR4 (e.g., CCL17 and/or CCL22) were in contrast to IL-6 and TGF-β not highly expressed in the tumor tissue, altogether indicating a conversion from CD8⁺ Tconvs rather than a tumor directed migration as the cause for the observed infiltration. In another recent study CXCL11 produced by CRC-derived CD68⁺ myeloid cells is suggested to be a promising chemoattractant for Tregs [116].

Among CRC 10%–15% of the cases are characterized by high levels of micro-satellite instability (MSI-H) that results from defective DNA mismatch repair (MMR)-systems. These types of CRC depict an increased infiltration of activated immune cells and have a relatively better prognosis [117]. The augmented immune response is ascribed to the highly immunogenic frame-shift derived TAAs. A stratification based on MSI status revealed that patients with MMR deficiency had a significantly higher infiltration with FOXP3⁺ CD4⁺ Tregs potentially as result of the pronounced immune response and its accompanying controlling components [118]. These findings were similar to observations by Sinicrope and Nosho [119,120], but in opposition to a study, which assessed FOXP3 mRNA levels instead of staining for cells [121]. Differences in the patients groups and more importantly in the applied methodologies may account for the varying observations and interpretations. It should be taken into consideration that FOXP3 mRNA can be present without translation and a potential FOXP3 expression by tumor cells can severely influence the results.

In a clinical trial for CRC patients combining chemotherapy (gemcitabine plus FOLFOX) with immune stimulants (granulocyte macrophage colony-stimulating factor and interleukin-2) increased levels of circulating Tregs almost normalized after two treatment cycles. The drop in Tregs was accompanied by an increase in T cell responses against tumor cell lysates, a patients' response-rate of almost 70%, and restoration of the CD4⁺/CD8⁺ T cell ratio [109]. This data could represent a strong indication for the clinical relevance of Tregs for anti-tumor responses and the potential benefit of their removal. However, it cannot be precluded that the beneficial effects were confounded phenomena and it still remains unsettled whether Tregs exert a substantial inhibitory in vivo effect in CRC patients.

In analogy to the numerous studies on the prognostic impact of Tregs in various types of solid tumors [32], for which the presence of Tregs is mostly associated with a worse prognosis, several groups are interested in their predictive value for CRC (selected studies are summarized in Table 1). At first, research focused on the proportion of Tregs among the TILs. Most studies were carried out by immunohistochemistry (IHC) using the FOXP3 antibody clone 236A/E7 that has been extensively tested and shown to detect mostly CD4⁺CD25⁺ Tregs [11]. However, it has to be emphasized that enumeration and visualization techniques (FOXP3 single- versus co-staining with other surface molecules, e.g., CD4) as well as the chosen FOXP3 clone have been shown to substantially influence the results making further standardization obligatory in order to achieve direct inter-study comparability [122,123]. In a French study on a large data set (n = 967) high density of Tregs within the tumor and weak infiltration of the surrounding healthy tissue was an independent positive prognostic marker [124]. The observed beneficial effect of intratumoral Tregs is in contrast to most findings in other solid cancers [32]. It reflects however very well the conclusions drawn from several pre-clinical models [93,96,97,99,125]. Similar observations were made in a cohort of 1420 patients, in which strong infiltration with Tregs was linked to a better survival, especially in MMR-proficient CRC [126]. In a side study of a phase-3 trial for relapsed CRC patients comparing standard FOLFOX-4 treatment with GOLFIG-2, high up-front infiltration with Tregs was of positive prognostic value for treatment response as well as for survival [127].

As described previously Tregs infiltrate more preferably the tumor stroma [108,112]. A higher stromal infiltration has been associated with limited disease and less metastatic lymph nodes [112,119]. Despite the rather low infiltration, the epithelial CD3/FOXP3 ratio had a significant impact for the five-year disease free survival (DFS). It even exceeded the prognostic potency of metastatic lymph nodes and TNM [119]. Similarly an increased CD8/Treg ratio (without discriminating between epithelial and stromal infiltrates) was an independent positive predictor for a better overall survival (OS) in patients undergoing curative surgery [128].

In TDLNs tumor antigens are firstly presented to cells of the adaptive immune system initiating a specific anti-tumor response. Since they simultaneously also represent the preferential site for CRC-metastases, TDLNs are an immunological compartment of great interest. In a study evaluating TDLNs from patients that underwent a radical resection, CD4⁺CD25⁺FOXP3⁺ Tregs correlated with disease stage and functional alterations of the adjacent CD8⁺ T cells. Suppression was reverted in vitro by depleting the Tregs [129]. However high Treg infiltration in sentinel lymph nodes appears to be associated with a better survival as indicated in a study that includes a rather limited number of cases (n = 30) [130] and further larger scale investigations in TDLNs remain mandatory.