Control of Regulatory T Cell Development by the Transcription Factor Foxp3

See allHide authors and affiliations

Science  14 Feb 2003:
Vol. 299, Issue 5609, pp. 1057-1061
DOI: 10.1126/science.1079490


Regulatory T cells engage in the maintenance of immunological self-tolerance by actively suppressing self-reactive lymphocytes. Little is known, however, about the molecular mechanism of their development. Here we show that Foxp3, which encodes a transcription factor that is genetically defective in an autoimmune and inflammatory syndrome in humans and mice, is specifically expressed in naturally arising CD4+ regulatory T cells. Furthermore, retroviral gene transfer of Foxp3 converts naïve T cells toward a regulatory T cell phenotype similar to that of naturally occurring CD4+ regulatory T cells. Thus, Foxp3is a key regulatory gene for the development of regulatory T cells.

To maintain immunological unresponsiveness to self-constituents (i.e., self-tolerance), potentially hazardous self-reactive lymphocytes are eliminated or inactivated during their development (1). Activation and expansion of self-reactive T lymphocytes that have escaped thymic clonal deletion is actively suppressed in the periphery by naturally occurring CD4+ regulatory T cells (TR), the majority of which constitutively express CD25 [interleukin (IL)-2 receptor α-chain] (2–6). CD25+CD4+TR cells are at least in part produced by the thymus as a functionally mature T cell subpopulation (7, 8), and their reduction or functional alteration in rodents leads to the spontaneous development of various organ-specific autoimmune diseases including autoimmune thyroiditis, gastritis, and type 1 diabetes (6–9). TR cells also appear to maintain a balanced response to environmental antigens, preventing inflammatory bowel disease (IBD) and allergy in rodents (10, 11).

Several studies have provided findings that offer clues to the potential pathway by which TR cells develop. Similar multiorgan autoimmune diseases, allergy, and IBD develop in the X-linked recessive disease, IPEX (immune dysregulation, polyendocrinopathy, enteropathy, X-linked syndrome) or XLAAD (X-linked autoimmunity-allergic dysregulation syndrome) (12). A mouse mutant strain, scurfy, also succumbs to similar X-linked recessive autoimmune and inflammatory diseases as a result of uncontrolled activation and expansion of CD4+ T cells (13–16). Recent efforts to identify the genetic defect in IPEX/XLAAD patients or scurfy mice have revealed mutations in a common gene, Foxp3, which encodes a forkhead-winged–helix transcription factor designated Scurfin (17–20). Immunological similarities between the autoimmunity and inflammation produced by manipulating CD25+CD4+ TR cells and those induced by genetic defects in Foxp3 prompted us to investigate the possible contribution of Foxp3 to the development or function of regulatory T cells.

We first examined the expression of Foxp3 mRNA in the thymus and the periphery of normal mice by reverse transcriptase–polymerase chain reaction (RT-PCR) (21). In the thymus, CD25+CD4+CD8thymocytes, which constitute 5% of CD4+CD8thymocytes (7), predominantly transcribedFoxp3, whereas CD4CD8+ and other immature thymocyte populations did not (Fig. 1A). In the periphery, CD4+ T cells specifically transcribed the gene, whereas CD8+ T cells and CD19+ B cells did not (Fig. 1B) (17). Among CD4+ T cells, the CD25+ subset, which constitutes 5 to 10% of CD4+ T cells in normal naïve mice, exhibited predominant transcription. Real-time quantitative PCR analyses revealed that the Foxp3 mRNA level in CD25+CD4+ cells was 100-fold more abundant than in CD25CD4+ cells (Fig. 1C). A low level of expression in CD25CD4+ cells was confined to the CD45RBlow population, which has been previously reported to contain regulatory activity (22, 23). These results indicate that Foxp3 expression is predominantly restricted to the CD25+CD4+population in both the thymus and periphery.

Figure 1

Expression of Foxp3 in a subpopulation of CD4+ T cells in the thymus and periphery. (A) BALB/c thymocytes were sorted into CD4CD8 (DN), CD4+CD8+ (DP), CD48+(CD8) or CD4+8 (CD4) cells (left). CD4+8 thymocytes were further separated into CD25+ or CD25 cells (right). cDNA from each population was subjected to nonsaturating PCR using Foxp3- or HPRT (hypoxanthine-guanine phosphoribosyl-transferase)–specific primers (21). (B) Pooled lymph node and spleen cells from BALB/c mice were sorted into the indicated compartments, and nonsaturating RT-PCR analyses were carried out. CD4+ cells were further separated into CD25+ or CD25cells. (C) Quantification of relative Foxp3 mRNA levels in indicated CD4+ T cell subsets. cDNA samples were subjected to real-time quantitative PCR analyses using primers and an internal fluorescent probe specific for Foxp3 or HPRT. The relative quantity of Foxp3 in each sample was normalized to the relative quantity of HPRT (21). (D) CD25CD4+ (open symbols) or CD25+CD4+ cells (closed symbols) were activated for indicated hours with plate-bound CD3 mAb in the presence of IL-2 (circles) or CD28 mAb (squares) and assessed for the expression ofFoxp3 by real-time quantitative RT-PCR. (A) to (D) each show one representative result of three independent experiments.

Expression of Foxp3 is not a mere consequence of T cell activation, because in vitro stimulation of CD25CD4+ cells for 3 days by monoclonal antibodies (mAbs) to CD3 in the presence of IL-2 or CD28 mAb failed to elicit Foxp3 expression (Fig. 1D). Neither effector Th1 nor effector Th2 cells prepared from naïve T cells expressedFoxp3 (fig. S1). Furthermore, stimulation of CD25+CD4+ cells with CD3 mAb and IL-2 did not alter their expression levels of Foxp3. Thus,Foxp3 expression is stable in CD25+CD4+ TR cells irrespective of the mode or state of activation.

We next determined whether forced expression of Foxp3 in naïve T cells could convert these cells toward a regulatory T cell phenotype. Bicistronic retroviral vectors expressingFoxp3 and green fluorescent protein (GFP) (Foxp3/MIGR1) or GFP alone (MIGR1) were generated (Fig. 2A). Peripheral CD25CD4+ cells from normal naïve mice were stimulated with CD3 mAb and IL-2 and infected with either retrovirus. One week after infection, the proliferative responses of GFP+ cells to T cell receptor (TCR) stimulation, cytokine production, and expression of cell surface molecules were examined. Retroviral transduction led to expression of GFP in 30 to 60% of CD4+ cells. Upon TCR stimulation with CD3 mAb, GFP+ cells from Foxp3/MIGR1-infected cultures proliferated poorly in contrast to the vigorous proliferation of GFP+cells from control MIGR1-infected cells or freshly prepared CD25CD4+ cells (Fig. 2B). In addition, Foxp3/MIGR1-infected GFP+ cells produced very little but detectable IL-2, IFN-γ, IL-4, and IL-10 as compared to GFP+ MIGR1-infected cells, which secreted large amounts of these cytokines (Fig. 2C). Although these Foxp3-expressing cells produced higher amounts of cytokines than freshly isolated CD25+CD4+ cells (fig. S2), cytokine-producing cells were largely confined to cells expressing low levels of GFP (and hence Foxp3).

Figure 2

Retroviral transduction of Foxp3into naïve CD4+ T cells. (A)Foxp3 cDNA was inserted into the MIGR1 retrovirus vector (21). This vector simultaneously expresses two cDNAs,Foxp3 and GFP, with the use of an internal ribosomal entry site (IRES). (B) CD3 mAb–stimulated proliferative response of freshly isolated CD25CD4+ cells and GFP+ Foxp3/MIGR1- or MIGR1-infected CD25CD4+ cells. [3H] thymidine incorporation was measured as an indicator of cell proliferation and expressed as the mean (± SD) of triplicate cultures. cpm, counts per minute. (C) IL-2, IFN-γ, IL-4, or IL-10 concentration in the culture supernatant of CD3 mAb–stimulated culture of GFP+ Foxp3/MIGR1- or MIGR1-infected CD25CD4+ cells (mean ± SD). (D) Foxp3/MIGR1- or MIGR1-infected T cells derived from CD25CD4+ cells were stained with phycoerythrin (PE)–labeled mAb for CD25, GITR, CD103, CD4, CD44, or an irrelevant antigen (control), along with detection of GFP. Intracellular CTLA-4 molecules were detected with PE-labeled CTLA-4 or isotype-matched control mAb (21). (B) to (D) each show one representative result of three independent experiments.

Naturally arising CD25+CD4+ TRcells characteristically express CD25, cytotoxic T lymphocyte-associated antigen–4 (CTLA-4), glucocorticoid-induced tumor necrosis factor receptor family–related gene (GITR), and CD103 (αE integrin) (6–10,24–28). Although the activation of CD25CD4+ cells for retroviral infection led to expression of these molecules in both Foxp3/MIGR1- and MIGR1-infected cells, GFP+ cells in the former expressed CD25, GITR, and CTLA-4 at higher levels than GFP cells or GFP+ MIGR1-infected cells (Fig. 2D, table S1). Among the GFP+ Foxp3-transduced cells, the higher the level of GFP (Foxp3) expression, the higher the expression was of these molecules (Fig. 2D). Furthermore, only GFPhigh (Foxp3 high) cells in the Foxp3/MIGR1-infected cells expressed CD103, with no expression on MIGR1-infected cells. Because Foxp3 transduction did not affect the expression levels of either CD4 or CD44, theFoxp3-induced high expression of these TR-associated molecules is most likely due to specific genetic instruction and not simply excessive cell activation byFoxp3. From these experiments, we conclude that transduction of Foxp3 could render naïve T cells hyporesponsive to TCR stimulation, inhibit cytokine production, and up-regulate expression of cell surface molecules closely associated with the regulatory function of naturally occurring CD25+CD4+ TR cells (29, 30).

We next assessed the potential of Foxp3-transduced T cells to show suppressive activity. Foxp3/MIGR1-infected CD25CD4+ T cells could specifically reduce proliferation of freshly prepared CD25CD4+responder T cells when stimulated with CD3 mAb (Fig. 3A). This suppression was associated with the inhibition of IL-2 production of the responder population in a fashion similar to that of natural CD25+CD4+TR cells (Fig. 3B) and correlated with the level of GFP (Foxp3) expression in the infected population [Supporting Online Material (SOM) Text, fig. S3A]. These GFPhigh cells also suppressed IL-4 secretion by CD25CD4+ responder T cells (fig. S3B). CD25+CD4+ cells, whether infected with Foxp3/MIGR1 or MIGR1, showed equally potent suppressive activity (Fig. 3A).

Figure 3

Suppressive activity ofFoxp3-transduced naïve CD4+ T cells. (A) Graded doses of GFP+ cells infected with either Foxp3/MIGR1 (closed circles) or MIGR1 (open circles) derived from BALB/c CD25CD4+ (left) or CD25+CD4+ cells (right) were cultured with 2.5 × 104 freshly prepared CD25CD4+ cells for 72 hours with CD3 mAb and 5.0 × 104 antigen-presenting cells (APCs). Proliferation of cells was assessed as in Fig. 2B. Freshly isolated CD25+CD4+ cells were also included in the assay (open squares). (B) Freshly prepared CD25CD4+ cells alone or mixed with the same number of GFP+ Foxp3/MIGR1-(+Foxp3/MIGR1) or MIGR1-infected CD25CD4+ cells (+MIGR1) or freshly isolated CD25+CD4+cells (+CD25+) were stimulated with CD3 mAb, and IL-2 concentration in the culture supernatant was measured as in Fig. 2C. (C) GFP+ cells from DO11.10/RAG-2–/– CD4+ cells infected with Foxp3/MIGR1 (closed circles) or MIGR1 (open circles) were cultured with 2.0 × 104 freshly prepared DO11.10 CD4+cells in the presence of OVA peptide and 4.0 × 104APCs. (D) BALB/c CD25CD4+ cells were transduced with Foxp3 (closed circles) or GFP alone (open circles) and sorted for GFP. Indicated numbers of GFP+ cells were mixed with 2.5 × 104DO11.10 TCR transgenic CD4+ cells and stimulated with either specific OVA peptide (left) or CD3 mAb (right). (E) GFP+ Foxp3/MIGR1-infected T cells derived from BALB/c CD25CD4+ cells and an equal number of freshly prepared CD25CD4+ cells were separated or nonseparated by a semipermeable membrane and stimulated with CD3 mAb and APCs on each side (left). IL-10 receptor (IL-10R) mAb, TGF-β mAb, or the mixture of these two was added to the culture of freshly prepared CD25CD4+ cells either alone (white bars) or in the presence of GFP+ Foxp3/MIGR1-infected CD25CD4+ cells (gray bars) or freshly prepared CD25+CD4+ cells (black bars) (right) (21). (A) to (E) each show one representative result of three independent experiments.

To examine whether antigen-specific naïve T cells could be converted to regulatory T cells by ectopic Foxp3 expression, we used recombination-activating gene (RAG)–deficient DO11.10 CD4+ T cells, which express ovalbumin (OVA) peptide–specific transgenic TCRs, are deficient in natural CD25+CD4+ TR cells (7), and scarcely express Foxp3 (SOM Text, fig. S4). In the absence of specific antigen, almost all T cells from these animals remain in a naïve state. Transgenic CD4+ T cells infected with Foxp3/MIGR1 suppressed the proliferation of freshly prepared noninfected transgenic T cells upon stimulation with specific OVA peptides, whereas those infected with MIGR1 did not (Fig. 3C). These results collectively indicate that ectopic expression of Foxp3 was sufficient to convert otherwise nonregulatory naïve T cells toward a regulatory T cell phenotype capable of suppressing proliferation of other T cells, presumably through inhibition of IL-2 production (SOM Text, fig. S5).

Foxp3-transduced CD4+ T cells appeared to exert the suppressive activity in a similar manner to that of naturally occurring CD25+CD4+ TR cells (SOM Text, fig. S6). First, Foxp3-transduced cells derived from BALB/c CD25CD4+ cells failed to suppress the proliferative response of DO11.10 TCR transgenic CD4+ T cells when stimulated with OVA peptide, whereas polyclonal stimulation with CD3 mAb induced suppression (Fig. 3D). This indicates thatFoxp3-transduced T cells require stimulation through TCRs to exert suppression (29, 30). Second,Foxp3-infected T cells failed to suppress other T cells when separated by a semipermeable membrane, in contrast to effective suppression that was observed when cell-cell contact was allowed (Fig. 3E). These separated Foxp3-infected T cells even enhanced T cell responses across the membrane, presumably via cytokines they produced (fig. S2). In addition, neutralization of transforming growth factor–β (TGF-β) or blocking of IL-10 receptor, either alone or in combination, failed to abrogate suppression, as was the case with natural CD25+CD4+ TRcells. These results indicate that in vitro suppression byFoxp3-transduced T cells may not be mediated by humoral factors but requires cell contact (29, 30).

Finally, we examined whether Foxp3/MIGR1-infected T cells could suppress in vivo the inflammation and the autoimmune disease that have been shown to develop in the absence of T cell regulation. For this, we used a model of IBD and autoimmune gastritis that can be induced in severe combined immunodeficiency (SCID) mice by the transfer of CD25CD45RBhighCD4+ T cells from normal BALB/c mice and prevented by cotransfer of CD25+CD4+ TR cells (10, 31) (Fig. 4). CD25CD45RBhighCD4+ cells from BALB/c mice were infected with Foxp3/MIGR1 or MIGR1, and GFP+ cells were cotransferred with freshly prepared CD25CD45RBhighCD4+ cells. The GFP+ Foxp3-transduced cells inhibited weight loss, diarrhea, and histological development of colitis and gastritis induced by the transfer of CD25CD45RBhighCD4+ cells as effectively as naturally occurring CD25+CD4+cells. By contrast, the cells transduced with GFP alone failed to prevent disease, and rather enhanced weight loss and the development of colitis. Thus, transduction of Foxp3 can render naïve T cells capable of preventing autoimmune gastritis and IBD caused by dysregulated immune responses toward gastric self-antigens and commensal bacteria, respectively (SOM Text).

Figure 4

Prevention of IBD and autoimmune gastritis byFoxp3-transduced T cells. (A) C.B-17scid mice received 4 × 105 fresh CD25CD45RBhighCD4+ cells either alone (n = 6, where n is the number of mice) (open squares) or together with 1.2 × 106GFP+ sorted cells derived from CD25CD45RBhighCD4+ cells infected with Foxp3/MIGR1 (n = 7) (closed circles) or MIGR1 (n = 5) (open circles). Body weight is represented as the percentage of initial weight (mean ± SD). Astericks indicate significant difference, P < 0.01, Foxp3/MIGR1 versus other two groups by Mann-Whitney test. (B) Histopathology of the colon and stomach in each group and in an unreconstituted SCID mouse (None). (C) Colitis (left) and gastritis (right) were histologically scored. Two mice in the group cotransferred with MIGR1-infected cells and one transferred with CD25CD45RBhighCD4+ cells alone died of debilitation before histological examination. Results shown in (A) to (C) are from a total of three independent experiments.

The present study shows that Foxp3 is predominantly expressed in the CD25+CD4+ TRpopulation naturally arising in the thymus and periphery and thatFoxp3 expression in naïve T cells can convert these cells to a regulatory T cell phenotype functionally similar to naturally occurring CD25+CD4+ TRcells. This result suggests that Foxp3 may be a master regulatory gene for cell-lineage commitment or developmental differentiation of regulatory T cells in the thymus and the periphery. Our results also indicate that, in defining the naturally occurring CD4+ regulatory T cells engaged in preventing autoimmune disease and immunopathology, Foxp3 represents a more specific marker than currently used cell-surface molecules (such as CD25, CD45RB, CTLA-4, and GITR), which are unable to completely discriminate between regulatory T cells and activated, effector, or memory T cells.

Mutations in the Foxp3 gene culminate in the development of a fatal lymphoproliferative disorder associated with multiorgan pathology both in mice and humans (12–20).FOXP3 is predominantly expressed in human CD25+CD4+ T cells as well (32). Furthermore, transduction of a mutant Foxp3 lacking the forkhead domain, similar to the mutated Foxp3 in scurfy mice (17), failed to confer suppressive activity to CD25CD4+ T cells (fig. S7). The present results therefore suggest that mutations of the Foxp3 gene may cause these disorders through developmental or functional abnormality of the CD25+CD4+ TRpopulation.

Potentially, generation of TR cells byFoxp3 transduction of naïve T cells may provide a previously unstudied therapeutic mode for treatment of autoimmune and inflammatory diseases and in transplantation tolerance.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S7

Table S1


  • * To whom correspondence should be addressed. E-mail: shimon{at}


View Abstract

Navigate This Article