It has become evident that self-reactive T cells with the potential to cause autoimmune disease comprise a part of the normal T-cell repertoire, but their activation is prevented by suppressor cells [1–3]. Although originally described in the 1970s [4], significant progress in characterizing suppressor T-cell subsets has been made only recently, where they have been renamed 'regulatory' T cells.

A subset of thymus-derived CD4⁺ cells that constitutively expresses CD25, the α-chain of the IL-2 receptor, protect their host from spontaneous organ-specific autoimmune diseases. These CD4⁺ CD25⁺ cells have been called 'professional' suppressor cells and have a contact-dependent mechanism of action, at least in vitro[5]. Other subsets of CD4⁺ and CD8⁺ cells, natural killer T cells, and cells displaying γδ TCRs also have downregulatory (suppressor) activity. In the periphery, suppressor T cells generated in response to environmental antigens protect their hosts from immune-mediated tissue injury by producing immunosuppressive cytokines.

The mechanisms responsible for the generation of suppressor T cells were poorly understood until recently. Our group has accumulated evidence that the multifunctional cytokine transforming growth factor-β (TGF-β) plays an essential role in the expansion of thymus-derived, professional, CD4⁺ CD25⁺ precursors that circulate in the blood. TGF-β also plays a key role in the generation of peripherally induced CD4⁺ and CD8⁺ cytokine-producing suppressor cell subsets.

This article will briefly review the evidence for contact-mediated and cytokine-producing suppressor cells, especially in humans, and the role of TGF-β in the generation of these cells. This knowledge can be used to generate suppressor T cells ex vivo in large numbers, and raises the possibility that the transfer of these cells back to the donor can serve as a therapy for autoimmune diseases such as systemic lupus erythematosus (SLE). This T-cell-based therapy could also be used to prevent graft rejection.

The existence of thymus-derived suppressor cells was suggested by studies in mice where neonatal thymectomy on day 3 led to the development of a multiorgan autoimmune disease [6]. This disease is due to the loss of CD4⁺ CD25⁺ suppressor cells that do not appear until the first week after birth [7, 8]. Mature CD4⁺ CD25⁺ cells are found in the CD45RB^low activated/memory fraction mouse T cells. Because potentially aggressive, self-reactive T cells are found in the CD45RB^hi naive fraction of mouse T cells, the injection of CD45RB^hi cells from nonautoimmune, normal mice into immunodeficient mice results in generalized, multiorgan inflammatory disease. Similar to neonatal thymectomy, this disease is prevented by supplementing the injected cells with purified CD4⁺ CD25⁺ cells [9, 10]. Because these thymus-derived CD4⁺ CD25⁺ T cells appear to be crucial for the prevention of spontaneous autoimmune diseases, they have been called 'professional' suppressor cells [5, 8].

In general, the properties of rodent and human CD4⁺ CD25⁺ T cells appear to be very similar. In humans, 6–18% of CD4⁺ T cells constitutively express CD25 [11–17]. Purified CD4⁺ CD25⁺ cells do not proliferate in response to cross-linking of their TCRs. They inhibit the activation of other T cells by a contact-dependent mechanism [5–17]. A large percentage constitutively express intracellular cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4 or CD152), the IL-2 receptor β-chain (CD122), transferrin receptors (CD71) and class II MHC markers [17].

Almost all the CD4⁺ CD25⁺ are in the 'activated' state (CD45RB^low in the mouse, CD45RA^- RO⁺ in the human). This suggests they may be continuously stimulated by their internal environment. Although activation of CD4⁺ CD25⁺ cells is antigen specific, once these cells are activated they not only suppress T cells stimulated by the same antigen, but they also inhibit T cells stimulated by other antigens; so-called bystander effects [18]. Although CD4⁺ CD25⁺ cells are nonresponsive to cross-linking their TCRs, they do proliferate when costimulated with IL-2 or anti-CD28.

Cytokine production by CD4⁺ CD25⁺ cells is controversial. While some groups claim that these cells do not produce cytokines [8, 17], other groups have found that they can produce IL-10 [12, 13, 19], TGF-β [15, 19], IL-4 [12] and low amounts of interferon-γ [15]. All groups agree that these cells do not produce IL-2 and that they have a contact-dependent mechanism of action. Their suppressive activities are not abolished by neutralizing antibodies to IL-10, and all groups agree that suppression is not abolished by anti-TGF-β, but for one exception [19]. Nakamura et al. reported that immunosuppression by CD4⁺ CD25⁺ regulatory T cells is mediated by cell surface-bound TGF-β [19]. Many of these differences can possibly be explained by the heterogeneity of CD4⁺ CD25⁺ T cells. One group separated human CD4⁺ CD25⁺ cells into high- and low-intensity fractions by cell sorting, and they found that the suppressive effects were only displayed by the high-intensity fraction [17]. This subset did not produce cytokines.

CD8⁺ and CD4⁺ T cells that produce immunosuppressive cytokines have been described. Those that produce predominantly TGF-β and variable amounts of IL-10 and IL-4 have been called Th3-type cells, and they have been generated in vivo by immunization through an oral or other mucosal route [2, 20]. This route of antigen administration, however, does not only result in Th3 cells. Both Th2 cells and CD4⁺ CD25⁺ cells can also be generated by this procedure [20–22]. The conditions needed for the generation of Th3 cells are poorly understood.

Other workers have produced regulatory CD4⁺ cells by repeatedly stimulating with the antigen in the presence of IL-10 [23–26], or using immature antigen-presenting cells that lack potent costimulatory activity [27]. These regulatory CD4⁺ cells have been called Treg 1 (Tr1) cells and they produce significant quantities of IL-10. They do not proliferate in response to antigen and do not produce IL-2. Therefore, they are anergic.

Th3 and Tr1-like cells have been described in humans. One group has reported the appearance of Th3 cells in patients with multiple sclerosis following oral administration of myelin basic protein [28]. Human Tr1 cells suppressed an alloantigen-induced proliferative response [29]. Th3 or Tr1 cells mediate antigen-specific cellular hyporesponsiveness in patients with chronic helminth infections [30].

The fact that some regulatory T cells produce predominantly TGF-β and others IL-10 is not fortuitous. The combination of TGF-β and IL-10 is more immunosuppressive than either of the cytokines by themselves [31]. Significantly, shortly after antigen activation, T cells downregu-late their signal transducing type II receptor (TGF-βRII) and become refractory to the effects of TGF-β [32]. These cells then become mature effector cells. IL-10 appears as a feedback regulator later in the response and induces the re-expression of TGF-βRII. The synergistic inhibitory effects of TGF-β and IL-10 then terminate the response.

Whether Th3 cells and Tr1 cells come from similar precursors or comprise different subsets of regulatory T cells is not known. Many variables determine the differentiation pathway that a naive T cell will take following activation. These include the antigen concentration and route of administration, the cytokine milieu, and the pattern of costimulatory signals. Self-MHC-reactive T cells in humans can either provide B-cell helper function or suppress antibody production, depending on how they are activated. In each case, regulatory function depends on the cytokines produced [33]. In determining the T-cell response to myelin basic protein, another group found that TCR usage was similar whether the T cells became Th1 encephalitogenic cells or regulatory Th3 cells [34]. These studies suggest common precursors for T cells that take different differentiation pathways.

While TGF-β has well-known inhibitory effects on lymphocyte cytokine production and functional properties [35], our laboratory has accumulated data that these effects can be overcome by IL-2 and can be superceded by costimulatory activities. The net effect is that TGF-β induces IL-2-activated CD8⁺ and CD4⁺ T cells to develop potent suppressive activities. In parallel, other groups have observed that TGF-β inhibits the differentiation of T cells to Th1 or Th2 subsets [36, 37].

The initial observation that TGF-β is an IL-2-dependent differentiation factor for regulatory T cells was made in a study designed to determine the conditions required for human CD8⁺ T cells to become suppressors of antibody production. Using a model where we could induce T-cell-dependent antibody production without accessory cells, we found that CD4⁺ T cells, by themselves, lacked suppressor-inducing activity. The CD4⁺ cells produced IL-2 but, notwithstanding previous reports [38, 39], this cytokine could not induce suppressor cells by itself. We learned that the interaction of IL-2-activated natural killer cells with CD8⁺ cells leads to the production of active TGF-β, and that the presence of this cytokine was critical for CD8⁺ cells to suppress antibody production (Figure 1) [40, 41]. Moreover, the suppression was cytokine dependent and was abolished by a neutralizing anti-TGF-β monoclonal antibody (JD Gray and DA Horwitz, unpublished observation, 2000). Both IL-2 and TGF-β were thus critical for CD8⁺ cells to become Th3-like regulatory cells.

We have also induced CD4⁺ T cells to become Th3 cells. We used the superantigen, staphylococcal enterotoxin B, as the T-cell activating agent. Low-dose staphylococcal enterotoxin B can induce T-cell-dependent antibody production without additional accessory cells [42]. Briefly exposing CD4⁺ cells to TGF-β downregulated B-cell helper activity and induced certain CD4⁺ cells to develop suppressive activity that was neutralized by anti-TGF-β. Activating both CD4⁺ and CD8⁺ cells in the presence of TGF-β thus induced them to develop cytokine-dependent suppressive activity [43]. Other workers have also reported similar effects of TGF-β on CD8⁺ T cells [44]. One group found that IL-4 and TGF-β are involved in the differentiation of naive CD4⁺ cells to cytokine-producing Th3-type cells [45]. Another group reported that in vitro differentiation of Th3-type cells from Th0 precursors from TCR transgenic mice is enhanced by culture with TGF-β [20].

We next focused our attention on the induction of naive (CD45RA⁺RO^-) CD4⁺ T cells to become suppressor cells. Using the alloantigens as the T-cell activating agent, we found that TGF-β induced naive CD4⁺ T cells to develop extremely potent suppressive activity. These CD4⁺ cells had the phenotype and functional characteristics of 'professional' regulatory T cells. Using the generation cytotoxic T-cell activity and T-cell proliferation to assess suppressive activity, we learned that the suppressor cells were CD25⁺, and that a large percentage expressed CTLA-4. Their suppressive effects were contact dependent and were not neutralized by anti-TGF-β or IL-10. Adding less than 1% of these cells to T cells strongly inhibited the generation of cytotoxic T-lymphocyte activity by preventing the activation of CD8⁺ cells [14]. Other workers have also reported that CD4⁺ CD25⁺ cells have potent suppressive effects on CD8⁺ cells. Rodent CD4⁺ CD25⁺ regulatory cells cause CD8⁺ cells to enter cycle arrest [46].

The precursors of the human CD4⁺ CD25⁺ T cells induced by IL-2 and TGF-β appear to be the small number of CD25⁺ cells in the naive fraction. Although <1% of these cells express CD25, depletion of these cells abrogated the generation of suppressive activity in some experiments [14]. The principal difference between the cytokine-induced CD4⁺ CD25⁺ cells and the murine and human positively selected CD4⁺ CD25⁺ cells that are predominantly found in the CD45RO⁺ 'memory' fraction is their capacity for expansion. The positively selected cells are anergic while the CD4⁺ CD25⁺ cells generated from naive cells can be expanded in IL-2 and retain their suppressive activity [14].

Studies on the mechanism of action of TGF-β have revealed that this cytokine has potent costimulatory effects on IL-2-activated T cells. These effects include upregulation of CD25, CTLA-4 and CD40 ligand expression on CD4⁺ cells [14, 47], and increased tumor necrosis factor-α production by both CD4⁺ and CD8⁺ cells [47]. The TGF-β costimulated human CD4⁺ T cells are resistant to activation-induced apoptosis. They took up less annexin and expanded fivefold greater in primary cultures than control, alloactivated CD4⁺ T cells [14] (SG Zheng and DA Horwitz, unpublished observations, 2001). Some workers have reported that TGF-β can accelerate activation-induced cell death of some T cells [48, 49], while others observed that this cytokine protected T cells from apoptosis [50, 51]. We favor the hypothesis that TGF-β promotes the death of mature Th1 and Th2 cells while protecting newly generated regulatory T cells from undergoing apoptosis. This view is consistent with a report indicating positive effects of TGF-β on naive T cells [52].

In summary, using several different stimuli to activate T cells, we have found that the combination of IL-2 and TGF-β can induce CD4⁺and CD8⁺ T cells to become either cytokine-producing Th3-like or contact-dependent professional suppressor cells. In our studies with CD8⁺ cells, the cultures were always supplemented with IL-2. When human CD4⁺ cells are activated in the presence of TGF-β by irradiated allogeneic stimulator cells or with superantigens, however, sufficient IL-2 is produced for the costimulatory effects of TGF-β and suppressor cell differentiation. By contrast, cultures with mouse lymphocytes must generally be supplemented with IL-2.

As shown in Figure 2, we propose that TGF-β induces thymic-derived CD25 precursors in the naive fraction of CD4⁺ cells to expand and to become contact-dependent 'professional' regulatory T cells. TGF-β also induces CD4⁺ and CD8⁺ cells that are CD25^- to become Th3-like cells. Although almost all naive CD4⁺ cells are CD25^-, why the predominant TGF-β effect on T cells in this fraction is the generation of 'professional' regulatory T cells remains to be determined. Our finding that both IL-2 and TGF-β are critical in the generation of regulatory T cells is of particular importance in patients with SLE since production of IL-2 and the active form of TGF-β is decreased [53].