Regulatory T cells (Tregs) are crucial in mediating immune homeostasis and promoting the establishment and maintenance of peripheral tolerance. However, in the context of cancer their role is more complex, and they are thought to contribute to the progress of many tumours. As cancer cells express both self- and tumour-associated antigens, Tregs are key to dampening effector cell responses, and therefore represent one of the main obstacles to effective anti-tumour responses. Suppression mechanisms employed by Tregs are thought to contribute significantly to the failure of current therapies that rely on induction or potentiation of anti-tumour responses. This review will focus on the current evidence supporting the central role of Tregs in establishing tumour-specific tolerance and promoting cancer escape. We outline the mechanisms underlying their suppressive function and discuss the potential routes of Tregs accumulation within the tumour, including enhanced recruitment, in-situ or local proliferation, and de-novo differentiation. In addition, we review some of the cancer treatment strategies that act, at least in part, to eliminate or interfere with the function of Tregs. The role of Tregs is being recognized increasingly in cancer, and controlling the function of these suppressive cells in the tumour microenvironment without compromising peripheral tolerance represents a significant challenge for cancer therapies.
Since Burnet and Thomas proposed the cancer immunosurveillance hypothesis in the late 1950s, the involvement of the immune system in the recognition and eradication of nascent transformed cells and, therefore, in the control of malignancies has been a matter of intensive debate [1, 2]. An improved molecular understanding of tumour-associated antigens (TAAs), and the use of genetically defined animal models to dissect out the role of different components of the immune system in tumour control has led to a general acceptance of the immunosurveillance hypothesis, despite the lack of consensus in results from some experimental mouse models (reviewed in ). It is also clear that the immune system has a role in shaping the immunogenicity of tumours [1, 2, 4], a concept supported increasingly by both epidemiological studies and direct evidence [5, 6]. Known as immuno-editing, and first proposed by Dunn and colleagues, this broader model involves three stages: elimination, equilibrium and escape [1, 2]. The immune system is thus recognized to have dual roles in the control of malignancy; to protect the host from emerging tumours as well as to sculpt them, resulting in variants with reduced immunogenicity or that are better equipped to evade or suppress immunological attack [1, 2, 7]. During the equilibrium phase, the precise nature of the interaction between the immune system and the tumour will determine the eventual outcome, be it elimination or escape [2, 7]. Recently, Hanahan and Weinberg proposed the addition of immune evasion to the original six hallmarks of cancer they outlined in 2000 and tumour-associated inflammation as an enabling characteristic for the acquisition of these hallmarks . Understanding how the balance between the immune system and the tumour is established during the equilibrium phase, and how it can be reorientated towards elimination, will aid in the design and development of improved therapeutic strategies aimed at cancer eradication.
Regulatory T cells
Identification of tumour-specific tolerance without generalized immunodeficiency or immunosuppression has been demonstrated in a number of models, thus defining tumours as sites of immune privilege [8-11]. This is driven by both tumour-intrinsic and -extrinsic mechanisms (summarized in Fig. 1 ). Because most TAAs are modified or aberrantly expressed self-antigens, it is conceivable that regulatory T cells (Tregs) play a critical role in this process . Indeed, accumulating evidence implicates Tregs as one of the principal cell types suppressing TAA-specific lymphocyte activity and tumour eradication, and thus one of the major obstacles to effective anti-tumour immunotherapy [2, 12-17].
Tregs, first identified by Sakaguchi et al. , are the mediators of peripheral tolerance to self- and innocuous environmental antigens [12, 19]. Originally characterized as CD4+CD25+ lymphocytes, they were also found to express the forkhead/winged helix family transcription factor FoxP3. This is critical for both the development of Tregs and for their suppressive capacity, and to an extent has also enabled improved characterization of these cells [12, 20]. Two major subsets of Tregs exist: natural Treg (nT) cells, generated during T cell development in the thymus upon high-avidity interactions with peptide/class II major histocompatibility complex, and adaptive or inducible Treg (iTreg) cells that arise in the periphery upon naive CD4+ T cell interaction with tolerogenic antigen-presenting cells (APCs) . While the nTreg population is considered to have relatively stable FoxP3 expression , plasticity within the iTreg compartment has been identified. In some models, loss of FoxP3 expression and acquisition of effector lineage-defining cytokines occurs , although in other experimental settings such instability has not been observed . As FoxP3 can be expressed transiently in both human activated CD4+ cells [24, 25] and murine T cells  with no demonstrable suppressive capacity, plasticity of FoxP3 expression may be a function of uncommitted cells that express this transcription factor transiently rather than a property of committed Tregs . Therefore, while it appears that Tregs represent a stable subset, there may still be potential for Treg induction from effector T cells (Teff) upon exposure to certain stimuli present such as in the tumour microenvironment. Both the stability of Tregs and potential Teff conversion within the tumour context may have important implications on therapeutic strategies.
It is likely that tumour-specific tolerance relies on both nTreg and iTreg populations, and independent contributions of these two populations have been shown in a murine tumour-induced tolerance model . However, others  have shown that the Treg population in renal cell carcinoma patients co-expresses FoxP3 and Helios, an Ikaros family transcription factor expressed by nTregs, but not iTregs [28, 29]. Furthermore, it was the FoxP3+Helios+ population that expanded after interleukin (IL)-2 administration . Therefore, it may be that the nTreg population is more important in tumour progression, at least in some tumours. This suggests that previous studies correlating increased Treg presence with disease progression require re-evaluation to determine whether the expanded population represents nTregs. While the implications for expanded nTreg over iTreg populations remain to be elucidated, inhibition of Helios expression has been shown to attenuate the suppressive capacity of nTregs , and as such could be employed therapeutically.
Classical CD4+CD25+FoxP3+ T cells can down-modulate anti-tumour responses by both release of soluble factors and direct cell–cell contact . The suppressive mechanisms employed by these cells, reviewed extensively elsewhere, are outlined in Fig. 2 [12, 19]. The relative contributions of each of these mechanisms remain to be elucidated, particularly in the tumour context . In addition, there are several other subsets of T cells with regulatory capacity, but these are beyond the scope of this review.
Mechanisms driving regulatory T cell accumulation within tumours
It is critical that the processes that contribute to Treg accumulation in tumours are understood, as this underpins successful treatment strategies. Three different, but not mutually exclusive, modes of Treg accumulation within the tumour mass can be envisaged: (i) increased trafficking, (ii) preferential Treg expansion and (iii) de-novo differentiation, where the latter two can occur either locally within the tumour microenvironment or distally in tumour-draining lymph nodes (TDLNs).
It has been suggested that Tregs display an enhanced capacity for infiltration of, and accumulation within, the tumour in comparison to Teffs [2, 12, 13]. In support of this, preferential recruitment has been observed in ovarian  and breast carcinoma , and also Hodgkin's lymphoma . It is reliant on chemokine-driven mechanisms, and several chemokines and their cognate receptors have been implicated. The chemokine receptors CCR4 and CCR8 are expressed by Tregs  and the CCR4 ligand CCL22 has been shown to be produced by both tumour cells and tumour-infiltrating macrophages [13, 17, 34]. Blockade of CCL22 reduced Treg infiltration into ovarian tumours and induced tumour rejection in a murine xenograft model . CCR4 also appears to facilitate Treg tumour infiltration in gastric cancer  and oesophageal squamous cell carcinoma . Another chemokine, CCL28, can be expressed by tumour cells during hypoxia, and it is reported to recruit preferentially Tregs expressing CCR10 , while in a murine model of pancreatic cancer Tregs have been shown to be recruited to the tumour site via a CCL5/CCR5 axis . Targeting these chemokine receptors may inhibit Treg accumulation within tumours, although receptor and ligand redundancy and promiscuity within the chemokine–chemokine receptor system provides a significant hurdle to overcome for therapy.
A second mechanism could be through expansion of Tregs within the tumour mass (in situ) or in the TDLNs. IL-2 is essential for Treg development and homeostasis , and it has been proposed that Teff-derived IL-2 within the tumour mass could promote Treg proliferation, and the presence of Ki-67 (a nuclear protein expressed by proliferating cells) has been observed in tumour-infiltrating Tregs . Furthermore, expansion of Tregs with potent suppressive capacity has been shown upon IL-2 administration in peripheral blood of metastatic melanoma or renal cell carcinoma patients , as well as within the tumour mass of patients with renal cell carcinoma . Dendritic cell-derived transforming growth factor (TGF)-β-dependent proliferation of Tregs within the TDLNs in rodent tumour models has also been reported .
A third mechanism involving de-novo conversion of FoxP3– T cells into Tregs may play an important role in Treg accumulation in tumours . The role of TGF-β in the induction of Tregs is well established , and tumour cell-derived TGF-β can contribute to the induction of Tregs . Similar observations have been made for the effects of indoleamine 2,3-dioxygenase , in addition to its capacity to directly activate Tregs in TDLNs . In support of local differentiation, Kurt and colleagues carried out a comparison of CD3+ T cells in progressing and rejected tumours, and found that tumour-infiltrating lymphocytes (TILs) with suppressive capacity have the same T cell receptor β chain variable region (TCR-Vβ) usage as those that promote tumour rejection . However, more recently it has been reported that local conversion does not contribute significantly to the Treg pool in carcinogen-induced tumours, as the T cell receptor repertoires between Teff and Treg cells are largely distinct within the tumour microenvironment . A high degree of overlap was observed between Tregs in the tumour and TDLNs, suggesting that they are induced in the TDLNs . Indeed, other work has shown that CD4+CD25+ T cells within the tumour microenvironment can be derived from the existing peripheral Treg pool , while Valzasina and colleagues have postulated that Tregde-novo conversion in TDLNs is the principal mechanism of Treg accumulation in tumour-bearing thymectomized and CD25-depleted mice . They also suggested that the expansion of Tregs in the TDLN occurs at the expense of CD4+ Teffs, and thus results in a concomitant reduction in the Teff pool . However, de-novo conversion, even in the TDLNs, may not be the major mechanism in lymphosufficient animals or, indeed, at later time-points during disease. In light of the recently published data on enriched nTreg over iTreg in renal cell carcinoma patients, de-novo induction of Tregs in the periphery does not appear to contribute significantly to the Treg pool, at least in this malignancy . Moreover, data from Curiel and colleagues show decreasing populations of Tregs in TDLNs with cancer progression , and therefore there may be a temporal component that governs which of these modes dominates.
Regulatory T cells and cancer prognosis
Since the first studies suggesting the involvement of suppressive T cells in inhibiting effective anti-tumour responses , Treg participation in anti-tumour responses has been studied extensively in animal models, and many malignancies have now been shown to contain infiltrating Treg populations . Typically constituting 5–10% of murine and human peripheral CD4+ T cell pool, Tregs are enriched significantly within the tumour mass, peripheral blood or ascites, where they may constitute up to 20–30% of total CD4+ cells [12, 17, 48]. This apparent enrichment is of particular importance because Tregs appear to suppress in a dose-dependent manner [2, 12, 30]. While an abundance of TILs was proposed initially to correlate with favourable prognosis , it is now recognized that in many instances it is rather the balance between Teff and Treg populations that determines disease outcome [2, 51, 52]. Furthermore, there exists a strong correlation between tumour progression and Tregs, with increased numbers or frequency of Tregs associated with poor prognosis in a variety of cancers, including ovarian , breast , non-small cell lung , hepatocellular , renal cell [40, 56], pancreatic , gastric  and cervical  carcinomas. This can be demonstrated experimentally, where adoptive transfer of Tregs results in reduced anti-tumour immune responses in mouse models [60, 61]. In contrast, however, increased levels of Tregs in patients with colorectal carcinoma [16, 62, 63], follicular lymphoma, germinal centre-like diffuse large B cell lymphoma and Hodgkin's lymphoma  have been shown to be associated with increased survival, and in head and neck squamous cell carcinoma with improved local tumour control  and survival . This correlation between Treg abundance and favourable prognosis perhaps seems counterintuitive, but might relate to the ability of these cells to suppress inflammation. Inflammation is one of the critical processes promoting tumour growth in many tissues by fostering proliferation, angiogenesis and metastasis [4, 5, 67]. Indeed, in patients with colorectal carcinoma, it has been proposed that Tregs suppress pro-tumourigenic T helper type 17 (Th17) cell-driven responses against the gut microbiota , and a similar down-regulation of pro-tumourigenic inflammatory responses by Tregs has been proposed in head and neck squamous cell carcinoma [16, 65]. Conversely, there is also clear evidence that Tregs in colorectal cancer suppress tumour-specific T cell responses, and increase the chance of recurrence after surgical resection . Thus, the type and location of the tumour, and the inflammatory milieu within the tumour mass, may determine precisely how Tregs impact on tumour progression.
Therapeutic targeting of regulatory Tregs
Strategies for tumour eradication have focused mainly on either enhancing tumour immunogenicity or promoting anti-tumour effector responses, although both have been largely unsuccessful [12, 13]. This is thought be due primarily to the highly immunosuppressive environment found within the tumour. As Tregs are involved critically in shaping the tumour microenvironment to prevent optimal function of effector cells, Treg depletion or manipulation of their suppressive functions have been explored increasingly as therapeutic strategies.
Augmenting anti-tumour immune responses by anti-CD25 antibody-mediated Treg depletion has been explored extensively, but has generated variable outcomes. Not surprisingly, depletion of Tregs has been shown to increase anti-tumour immunity and vaccine efficacy, and to result in overall improved responses to immunotherapy in murine studies [14, 34, 61, 69, 70]. Two studies using depletion of CD4+CD25+ cells with an anti-CD25 antibody (PC61) in mice showed the efficient eradication of syngeneic tumours in vivo [69, 71]. Depletion of CD4+CD25+ alone or in combination with vaccination has since been shown to contribute to improved anti-tumour responses and/or tumour regression in various tumour models [70, 72, 73], although some studies have failed to show any effect on tumour growth.
Clinically, non-specific depletion of CD25+ cells using daclizumab, a monoclonal antibody (mAb) against CD25, has been effective in the induction of anti-tumour responses when administered together with peptide vaccines in metastatic breast cancer patients . Denileukin diftitox [DAB389IL-2 (ONTAK)], a recombinant protein composed of IL-2 and the active domain of diphtheria toxin, has been shown useful in the treatment of cutaneous T cell lymphoma . It has also been shown to enhance anti-tumour effects of dendritic cell vaccines in renal cell carcinoma patients . However, it had no effect on survival in patients treated previously for non-small cell lung carcinoma , and had no impact on Treg numbers, Treg function or disease progression in patients with metastatic melanoma or renal cell carcinoma . Furthermore, Powell and colleagues reported that, while administration of LMB-2 (a CD25-directed immunotoxin) to metastatic melanoma patients resulted in preferential reduction of circulating Treg numbers, it did not have any significant effects on the anti-tumour immune responses induced by vaccination .
Interestingly, the efficacy of several non-specific chemotherapeutic agents may be due in part to the depleting effects on Treg populations . Among the most studied is cyclophosphamide, an alkylating agent which has been shown to induce protective anti-tumour responses in both animal models [61, 80, 81] and in combination with vaccination in patients with metastatic melanoma  or breast cancer . Peripheral blood samples from patients with metastatic cancers treated with metronomic doses of cyclophosphamide had a selective reduction in circulating Tregs and improved anti-tumour immunity . Cyclophosphamide also potentiated anti-tumour responses in hormone-refractory prostate cancer patients , and there is some evidence that cyclophosphamide inhibits the optimal function of Tregs . However, no effect on the total Treg pool, or Treg suppressive capacity, was observed with a single administration of cyclophosphamide in combination with bacilli Calmette–Guérin (BCG) immunotherapy in a Phase I trial of metastatic carcinoma . In tumour-bearing mice, paclitaxel (an inhibitor of mitosis) given in metronomic doses with antigen-specific immunotherapy was able to reduce the number of Tregs independent of its known role in angiogenesis inhibition . Previous studies in a murine mammary carcinoma model showed a small but significant immunomodulatory effect of doxorubicin or paclitaxel administration with anti-tumour vaccination . Reduction in Treg numbers and suppressive capacity upon paclitaxel administration has also been confirmed in patients with non-small cell lung carcinoma .
An important consideration for therapies relying on Treg depletion is the rapid recovery. It has been described previously that the Treg compartment is repopulated in 48 days following depletion in mice , and similar effects have been reported in humans [76, 92]. Furthermore, depletion of Tregs should be carried out with caution because the agents described may target both Teff and Treg populations, as CD25 may be expressed on more than 25% of human T cells . Clinically, treatment of renal cell carcinoma with autologous T cells depleted of Tregs produced a transient reduction in peripheral Treg numbers, which coincided with increased antigen-independent and tumour-specific T cell responses. This suggests that although Treg depletion is generally transient, it can result in a potential therapeutic window for generating effective anti-tumour responses .
Interfering with some of the principal functions of Tregs is also being investigated. Cytotoxic T lymphocyte-associated antigen (CTLA)-4, expressed constitutively by Tregs, appears to be implicated in tumour control [2, 12]. Studies targeting this molecule have been studied extensively in murine tumour models [94-96]. Clinically, ipilimumab (human anti-CTLA-4 mAb; MDX-010) has been effective in Phase III trials, increasing overall survival in patients with advanced metastatic melanoma . However, patients with stage IV melanoma or renal cell carcinoma showed no reduction in Treg numbers or inhibited suppressive capacity, and it has been postulated that the effect of anti-CTLA-4 treatment is due to increased activation of Teffs . Others have suggested that there is a Treg-depleting effect early after initial administration which permits the optimal induction of effector cell responses . There is additional evidence that the suppressive mechanisms of CTLA-4 and Tregs are, at least in part, independent because even in the absence of CD25+ cells CTLA-4 blockade has been shown to induce anti-tumour responses . Targeting glucocorticoid-induced tumour necrosis factor receptor-related protein (GITR), a receptor expressed on Tregs, CD4+CD25– T cells and APCs, has also been explored . With murine cells in vitro it was observed that administration of the cognate ligand (GITRL), as well as an agonistic antibody (DTA-1), resulted in the suppression of regulatory T cells [100, 101]. Anti-GITR antibody treatment has been effective in promoting anti-tumour responses in the murine B16 tumour model  as well as in carcinogen-induced sarcoma in mice , but its effect in humans remains to be investigated.
Certain therapies have been shown to modulate the Teff/Treg ratio. A combination of Gvax [granulocyte–macrophage colony-stimulating factor (GM-CSF)-secreting B16 tumour cell-based vaccine] or Fl3vax (Flt3 ligand-secreting B16 tumour cell-based vaccine) with CTLA-4 blockade has been shown to be effective in increasing the ratio of Teff to Treg, resulting in rapid tumour rejection in murine tumour models [103, 104]. In combination with anti-OX40, cyclophosphamide has been shown to have similar effects . Interfering with the programmed cell death (PD)-1/PDL-1 axis has also been explored and has been shown to be effective in tumour rejection through a similar increase in the Teff/Treg ratio . This suggests that approaches aimed at both interference with Treg function and enhancement of Teff activation and function may be equally important, and through synergy yield effective anti-tumour responses.
An important implication to therapeutic targeting of Tregs is their involvement in maintaining peripheral tolerance. A number of studies have shown that depleting Tregs or interfering with their function may lead to the development of mild to severe autoimmune reactions in humans. Similarly, anti-CD25 antibody treatment in mice resulted in increased anti-tumour immunity but also a concomitant increase in autoimmunity . These potential side effects need to considered and monitored in future clinical studies.
The immune system shapes established tumours, resulting in an increased ability to withstand immunological attack. When a dynamic state of equilibrium is established upon interaction of the immune system with the developing tumour, strong T cell responses are key to acquiring long-lasting tumour-specific immunity. The mechanisms of establishing tumour-specific tolerance, with a principal contribution from Tregs, are becoming increasingly a focus of research, yet the field is replete with unanswered questions. Development of new therapeutic strategies that interfere with the generation and function of Tregs, or result in their specific depletion, will require improved targeting of Tregs, a better understanding of the role these cells play in different tumour types and clearer insight into the molecular mechanisms that are responsible for their accumulation and survival within tumours. Based on the data currently available, it is clear that the timing and dose of certain therapeutics is of key importance. One of the major caveats for cancer immunotherapy is the incidence of adverse reactions, such as hypersensitivity and autoimmunity, which arise as a consequence of targeting Tregs. Development of effective therapeutic strategies to target cancer will rely on combining control of Treg function and Teff/Treg ratios, while minimizing the loss of peripheral tolerance. This remains a significant challenge for the future of cancer treatment.
ARF is supported by a Medical Research Council Program grant held by RJN and GJG.
The authors have no financial conflict of interest.