Scurfin (FOXP3) Acts as a Repressor of Transcription and Regulates T Cell Activation*

We have recently identified and cloned Foxp3 , the gene defective in mice with the scurfy mutation. The immune dysregulation documented in these mice and in humans with mutations in the orthologous gene indicates that the foxp3 gene product, scurfin, is involved in the regulation of T cell activation and differentiation. The autoimmune state observed in these patients with the immune dysregulation polyendocrinopathy, enter-opathy, X-linked syndrome, or X-linked autoimmunity-allergic dysregulation syndrome also points to a critical role for scurfin in the regulation of T cell homeostasis. FOXP3 encodes a novel member of the forkhead family of transcription factors. Here we demonstrate that this structural domain is required for nuclear localization and DNA binding. Scurfin, transiently expressed in het-erologous cells, represses transcription of a reporter containing a multimeric forkhead binding site. Upon overexpression in CD4 T cells, scurfin attenuates acti-vation-induced cytokine production and proliferation. We have identified FKH binding sequences adjacent to critical NFAT regulatory sites in the promoters of several cytokine genes whose expression is sensitive to changes in SFN abundance. Our findings indicate that the ability of scurfin to bind DNA, and presumably repress transcription, plays a paramount role in determining the amplitude of the response of CD4 T cells to activation.

scurfy (sf), 1 a naturally occurring, X-linked, recessive mutation, has been described previously (1)(2)(3)(4)(5)(6)(7)(8)(9). The disease observed in hemizygous mutant males features massive lymphoproliferation and subsequent infiltration of several organs, and results in death by ϳ3 weeks of age (1,4). Both depletion and adoptive transfer experiments indicate that CD4 T cells are primarily responsible for the pathology observed in sf mice (2,6). sf CD4 T cells are chronically activated, expressing up-regulated levels of several activation markers and secreting increased levels of several cytokines directly ex vivo (8,9). sf-derived CD4 T cell effector function is also refractory to inhibition with several pharmacological reagents, particularly the immunosuppresants cyclosporin A and rapamycin (9). The phenotype of sf mutant mice is strikingly similar to that observed in mice deficient in expression of CTLA-4, indicating that scurfin may also be an important regulator of the T cell activation program (10 -14).
Recently we have positionally cloned the sf gene (15). The wild type sf gene (Foxp3) encodes a novel 48-kDa protein, scurfin (SFN). The protein is expressed at low levels, primarily in CD4 T cells (15). Interestingly, SFN expression levels do not appear to be modulated following activation. The sf mutation is a 2-bp insertion that results in a premature stop codon. It is not known if the predicted truncated protein is stably expressed in mutants. However, overexpression of a transgene encoding the mutant form of the SFN gene in wild type mice yielded no phenotype (15). This finding suggests that the sf phenotype most likely results from a loss of SFN function. The observations that females heterozygous for the sf mutation are phenotypically normal and display random X-inactivation also support this interpretation.
Concurrent with the cloning of the sf gene, several reports of mutations in the human ortholog of the Foxp3 gene have been made (16 -18). Affected males present symptoms similar to those observed in sf mice and include a predilection to autoimmune diseases and allergy. The majority of the mutations in these patients also result in premature stop codons. Although the phenotype of sf mice and of patients with the immune dysregulation polyendocrinopathy, enteropathy, X-linked syndrome (IPEX), or X-linked autoimmunity-allergic dysregulation syndrome (XLAAD) indicate a critical role for the foxp3 gene product in the negative regulation of T cell activation, it is not immediately evident how the protein exacts this effect.
The most prominent structural feature of SFN is a forkhead/ winged helix domain at the C-terminal end of the protein.
Forkhead/winged helix domain-containing proteins are members of a rapidly growing family of DNA binding factors. Since the identification of the original forkhead (FKH) protein in Drosophila in the early 1990s, over 80 FKH family proteins, classified into 17 different subfamilies, have been described in a variety of species, including nematodes, yeast, and mammals (Ref. 19 and www.biology.pomona.edu/fox.html). Although both transcriptional activators and repressors have been identified in this family, FKH domain-containing proteins typically function in the regulation of lineage commitment and developmental differentiation (reviewed in Ref. 20).
SFN, with an FKH domain of only 84 amino acids, is an atypical winged helix family member. QRF-1, a previously de-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: sf, scurfy; SFN, scurfin; IPEX, immune dysregulation polyendocrinopathy, enteropathy, X-linked syndrome; XLAAD, X-linked autoimmunity-allergic dysregulation syndrome; FKH, forkhead; GST, glutathione S-transferase; V1P, V1 immunoglobulin heavy chain variable-region (VH) promoter; FCS, fetal calf serum; Tet i , tetracycline-inducible; GFP, green fluorescence protein; ␣, anti; EMSA, electromobility shift assay; TTR-S, transthyretin site; NFAT, nuclear factor of activated T cells; GM-CSF, granulocyte macrophagecolony stimulating factor; IL, interleukin; TNF, tumor necrosis factor ␣; PMA, phorbol myristate acetate; STAT, signal transducer and activator of transcription; bp, base pair(s). scribed partial protein sequence (21), contains a similarly truncated FKH domain. Li and Tucker (21) have demonstrated that the FKH domain of QRF-1 is necessary and sufficient for binding to oligonucleotides containing DNA sequences described to bind FKH domain-containing proteins. The sequence identity between the FKH domain of QRF-1 and scurfin suggested that the latter would also function as a DNA binding protein. Another distinct feature of SFN is the location of the FKH domain at its C terminus. Generally, the FKH domain is located near the amino terminus of proteins in this family. This unique structural feature may be significant in the protein's function and help distinguish SFN from other FKH family proteins active in CD4 T cells.
With the ultimate goal of identifying how SFN acts to regulate T cell function, we have carried out fundamental analyses of this protein. Here, we report that scurfin is a DNA binding protein that can repress transcription. The FKH domain is required for this activity. Complimentary to the observation of hyper-responsive T cells in mice carrying the inactivating sf mutation, and the description of T cell-mediated autoimmunity in humans with the IPEX or XLAAD syndrome, overexpression of scurfin attenuates the T cell activation response, as evidenced by reduced cytokine production in response to activating stimuli. The ability to bind DNA, i.e. inclusion of the FKH domain, is required for this effect. Although the exact mechanisms by which SFN regulates T cell activation remain to be elucidated, the findings presented here suggest that the ability of scurfin to act as a negative regulator of cytokine production by CD4 T cells may involve direct repression of NFAT-mediated transcription.

EXPERIMENTAL PROCEDURES
Constructs cDNA encoding full-length human scurfin (amino acids 1-431) or a fragment lacking the FKH domain (amino acids 1-327) was inserted into the expression vector pIRES2-EGFP (CLONTECH, Palo Alto, CA), the GST fusion vector pGEX-4T-1 (Amersham Pharmacia Biotech, Piscataway, NJ), and the tetracycline-responsive vector pREV-TRE (CLONTECH). A multimeric FKH binding site construct was created by annealing complementary oligonucleotides containing three tandem repeats of the FKH binding site V1P (sequence given below). Restriction sites were included at the ends of the oligonucleotides to facilitate subcloning of the multimer into the SV40 promoter-driven luciferase reporter, pGL-3 Promoter (Promega, Madison, WI) to create 3xFKHluc. Sequencing confirmed all constructs to be correct and free of mutation.

Cell Lines
HEK 293T cells and COS-7 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% FCS, penicillin, and streptomycin. The Jurkat subline D4 was maintained in RPMI 1640 supplemented with 10% FCS, penicillin, and streptomycin. Tetracycline-inducible (Tet i ) Jurkat clones were generated following neomycin selection of D4 cells transfected with pREV-Tet-On (CLONTECH). Coexpression of a Tet-responsive GFP construct allowed selection of a clone yielding minimal promoter activity in the absence of treatment with the tetracycline analog, doxycycline (CLONTECH). This clone, hereafter referred to as TO, was used to create Tet i expressers of empty vector (TO.TRE), full-length scurfin (TO.SFN), or scurfin lacking the FKH domain (TO.⌬FKH).

Localization Studies
Confocal Microscopy-The full-length (SFN), deletion mutant (⌬FKH), or empty pIRES2-EGFP construct was transiently expressed in HEK 293T cells using a calcium phosphate transfection kit (5 Prime 3 3 Prime, Inc., Boulder, CO). After 24 h of culture, ϳ5000 cells/well were plated on poly-lysine-coated microscope slides and allowed to adhere overnight at 37°C. 48 h post-transfection, cells were fixed in 4% paraformaldehyde and permeabilized with 0.1% saponin. Cells were stained with pre-immune sera (Pre) or polyclonal antisera raised against full-length human scurfin (␣SFN), followed by goat ␣-rabbit IgG Alexa-568, the actin stain phalloidin-Alexa 488, and the DNA intercalating dye Toto-3 (all from Molecular Probes, Eugene, OR). After several phosphate-buffered saline washes, mount medium (50% glycerol in phosphate-buffered saline) was added to each well. A coverslip was then applied and sealed. The Laser Scanning Confocal Imaging System, MRC 1024 (Bio-Rad, Hercules, CA), was used to visualize subcellular localization of full-length and ⌬FKH scurfin.
Western Blot Analysis-Cytoplasmic and nuclear protein was isolated from HEK 293T transient transfectants as previously described (22). Protein concentration was determined by Bradford assay. Equal quantities of cytoplasmic and nuclear protein were resolved by polyacrylamide gel electrophoresis and transferred to nitrocellulose. The subcellular localization of scurfin was assessed by immunoblotting with ␣SFN, followed by detection with a chemiluminescent reagent. GST fusion proteins were affinity purified from isopropyl-1-thio-␤-D-galactopyranoside-induced bacterial lysates in the presence of protease inhibitors using pre-swelled glutathione-agarose beads (Sigma Chemical Co., St. Louis, MO). Nuclear extracts were prepared from D4 cells as previously described (22). Nuclear protein (4 g/lane) or a comparable amount of purified GST fusion protein was preincubated at room temperature with binding buffer supplemented with 500 ng of poly(dI-dC) (Roche Molecular Biochemicals, Indianapolis, IN) and a 50-to 100-fold molar excess of cold oligonucleotide competitor for 10 min (23). Probe (2 ϫ 10 4 cpm, endlabeled with [␥-32 P]ATP) was then added to a final volume of 25 l, and the incubation continued for an additional 15 min at room temperature. In some reactions, rabbit pre-immune sera or antisera raised against full-length SFN was incubated with nuclear protein for 30 min on ice prior to addition of the probe. All reactions were analyzed by polyacrylamide gel electrophoresis using 4% gels in 0.027 M Tris borate, 0.6 mM EDTA buffer.

Evaluation of the Effect of Scurfin on Transcriptional Activation
The activity of the 3xFKH-luc or the pGL-3 Promoter luciferase reporter construct was assessed in COS-7 cells. Calcium phosphate transfection was used to transiently express the ␤-galactosidase control vector, pSV-␤-gal (Promega), the luciferase reporter pGL-3 Promoter or 3xFKH-luc, and the SFN, ⌬FKH, or empty pIRES2-EGFP expression construct in COS-7 cells. 48 h post-transfection, cells were harvested and lysates were prepared in 1ϫ Reporter lysis buffer according to the manufacturer's specifications (Promega). Luciferase activity was measured in cellular lysates using the Luciferase assay system (Promega) and the EG & G Berthold Lumat 9507 luminometer (PerkinElmer Life Sciences, Boston, MA) according to the manufacturers' specifications. The ␤-galactosidase activity in duplicates of each lysate was determined using the ␤-galactosidase enzyme assay system (Promega) and used to normalize measured luciferase activity.

Analysis of Cytokine Production
Detection of Secreted IL-2-The Jurkat subline D4 was transiently transfected with the expression construct encoding full-length human SFN or the parent vector as described previously (23). After 24 h of culture in the absence or presence of 1 or 10 g/ml CD3 antibody (OKT3), culture supernatants were collected. IL-2 secretion was evaluated by bioassay using the IL-2-dependent murine cell line HT-2 (9).
Evaluation of Cytokine Promoter Activity-The activity of a luciferase reporter containing a multimer of a regulatory NFAT site (Ϫ280) from the murine IL-2 gene (NFAT-luc) (24) or the SV40 promoterdriven luciferase reporter, pGL-3 Promoter, was assessed in the TO.TRE, TO.SFN, and TO.⌬FKH cell lines. Cells were cultured in 1 g/ml doxycycline for 24 h prior to transfection. Equal concentrations (10 7 cells/0.4 ml) of each clone were electroporated with 10 g of the control vector pSV-␤-gal and either NFAT-luc or pGL-3 Promoter as previously described (23). Cells were rested for 2 h at 37°C. Cell viability was determined by trypan blue exclusion, and 10 6 cells per well were cultured in the absence or presence of ionomycin (1.5 M) and PMA (25 ng/ml) for 6 h. Cells were harvested, and lysates were prepared in 1ϫ reporter lysis buffer. Luciferase activity was measured and normalized based on the ␤-galactosidase activity detected in duplicate lysates.

RESULTS
Distinct sequences present in the FKH domain have been described to mediate nuclear localization and DNA binding by proteins containing this structural feature (25). To analyze scurfin localization, we first assessed the ability of transiently expressed scurfin to localize to the nucleus. A construct directing expression of scurfin (SFN), a fragment lacking the FKH domain (⌬FKH), or the parent vector pIRES2-EGFP (293) was transiently transfected into HEK 293T cells as described under "Experimental Procedures." Scurfin localization was assessed at both the single cell level by confocal microscopy (Fig. 1, A-D) and at the level of the transfected population by fractionation of total cellular protein into nuclear and cytoplasmic components and subsequent Western blotting (Fig. 1E). Scurfin (in red) predominantly localizes to the nucleus of HEK 293T cells (Fig.  1C). As predicted, removal of the FKH domain results in the exclusion of the majority of scurfin from the nucleus (Fig. 1D). The residual red staining observed in the nucleus of ⌬FKHtransfected cells could reflect staining of scurfin endogenously expressed in HEK 293T cells or, more likely, cross-reactivity of the antisera with other FKH domain-containing proteins expressed in these cells. A similar, low level of reactivity with the SFN antisera is observed in the nucleus of cells transfected with vector alone (Fig. 1B). Finally, staining with the scurfin antisera is specific, because no detectable signal is observed when pre-immune sera is used for staining (Fig. 1A).
To demonstrate that the results observed by confocal micros-copy were representative of the entire transfected population, cytoplasmic and nuclear protein were concomitantly isolated from transfectants. Equal quantities of cytoplasmic (c) and nuclear (n) protein were immunoblotted with scurfin antisera (Fig. 1E). As observed with confocal microscopy, the majority of scurfin (ϳ48 kDa) localizes to the nuclear protein fraction, and removal of the FKH domain (ϳ34-kDa band) results in its retention in the cytoplasm. These findings indicate that scurfin can localize to the nucleus and that the FKH domain is required for this to occur. Confocal microscopy and Western blotting of nuclear protein isolated from Jurkat cells both indicate that endogenously expressed SFN also localizes to the nucleus (data not shown). Based on the sequence identity between the FKH domain of the previously described partial protein QRF-1 and scurfin, we predicted that scurfin would form a complex with the same DNA sequences (21). We tested the ability of both recombinant GST fusions with scurfin and endogenously expressed scurfin to bind to oligonucleotides containing these sequences. As shown in Fig. 2A, a fusion of GST with full-length scurfin binds to a probe containing a dimer of the FKH consensus binding site from the TTR-S gene ( Fig. 2A, lane 3). Addition of a 50-fold molar excess of cold competitor of the same sequence eliminates formation of the lowest band ( Fig. 2A, lane 4), but inclusion of the same amount of an irrelevant DNA sequence does not ( Fig. 2A, lane 5). Addition of polyclonal antisera raised against SFN in the gel shift reaction results in retardation of the lowest band ( Fig. 2A, lane 6). This indicates that SFN is a primary component of the lowest DNA-protein complex and that the two higher bands may represent nonspecific binding activity. As expected, a fusion of GST with a deletion mutant of scurfin, which lacks the FKH domain, fails to form a complex with the probe (Fig. 2A, lane 2). Finally a fusion between GST and c-Jun, a protein known to bind a distinct DNA sequence, fails to form a complex with the TTR-S probe (Fig. 2A, lane 1). These results suggest that scurfin is a DNA binding protein that specifically complexes with a consensus FKH binding site and that the FKH domain is required for this activity.
The FKH domain-dependent ability of recombinant scurfin to bind to DNA was not surprising. However, this finding did not prove that scurfin binds DNA under more physiologically relevant circumstances. Because CD4 T cells express scurfin and are most affected by the inactivating sf mutation, we next studied the DNA binding ability of endogenously expressed scurfin. Nuclear protein was extracted from Jurkat cells activated with ␣CD3 for 2 h and used in gel shift assays. As observed with recombinant full-length protein, endogenously expressed scurfin forms a complex with a probe that contains a FKH protein consensus DNA binding site from the V1P promoter (Fig. 2B, lanes 1 and 5). A 50-fold molar excess of cold oligonucleotide competitor of the same DNA sequence as the probe (Fig. 2B, lane 2) or the FKH binding site from the TTR-S promoter (Fig. 2B, lane 3) competes away the protein/probe complex, indicating that a specific complex is formed. Furthermore, addition of an irrelevant DNA sequence does not disrupt complex formation (Fig. 2B, lane 4). The same findings were attained when the FKH binding site from the TTR-S promoter was used as the probe (data not shown). Inclusion of antisera raised against scurfin (Fig. 2B, lane 6), but not pre-immune sera (Fig. 2B, lane 7), results in retardation of the migration of the complex, demonstrating that scurfin is a major component of the protein-DNA complex. These findings confirm that scurfin is present in the nucleus of CD4 T cells and is capable of binding DNA. The same DNA binding activity is observed in nuclear protein isolated from unstimulated Jurkat cells (data not shown).
Next we investigated the functional significance of the ability of scurfin to bind DNA. The FKH family of DNA binding factors contains transcriptional activators as well as repressors of transcription. The documented transactivation domains of other FKH family proteins are located C-terminal of the FKH domain (26 -28). As previously mentioned, the scurfin FKH domain is unique in that it ends only 11 amino acids from the C terminus of the protein. We hypothesized that SFN might lack a transactivation domain and possibly inhibit transcrip-tion. To address this question, we examined the effect of SFN on the expression of the SV40 minimal promoter reporter flanked by multiple FKH binding sites. COS-7 cells were transiently transfected with an expression construct directing the production of scurfin (SFN), a mutant lacking the FKH domain (⌬FKH), or the parent vector pIRES2-EGFP (Control). The control vector pSV-␤-gal and either 3xFKH-luc (Fig. 3, speckled bars) or the parent luciferase reporter pGL-3 Promoter (Fig. 3, solid bars) were co-transfected with each expression construct. Lysates were harvested and luciferase and ␤-galactosidase activity measured 48 h later. Comparable levels of luciferase activity are detected in cells transfected with the pGL-3 Promoter, irrespective of the expression construct co-transfected (Fig. 3, solid bars). Activity of the 3xFKH-luc construct in COS-7 cells transfected with the parent vector (Fig. 3, Control, speckled bar) and SFN lacking the DNA binding domain (Fig. 3, ⌬FKH, speckled bar) is comparable. However, overexpression of full-length scurfin consistently results in more than a 50% reduction of transcription of the reporter gene (Fig. 3, SFN, speckled bar), suggesting that SFN can repress transcription. The observation that SFN overexpression does not affect promoter activity in the absence of FKH binding sites (Fig. 3, SFN, solid bar) and that removal of the DNA binding domain eliminates the reduction in expression of the 3xFKH-luc reporter gene (Fig. 3, ⌬FKH, speckled bar) both indicate that this repression is specific and most likely mediated by the FKH domain. Although Fig. 3 shows an apparent reduction in expression of the 3xFKH-luc in ⌬FKHtransfected cells, relative to that observed in control transfected cells, this trend has not been consistently observed in duplicate experiments.
Based on the results of the multimer study in COS-7 cells, scurfin appears to repress transcription. How this activity, or other as yet unidentified scurfin functions, is directly involved in the regulation of T cell activation is not immediately obvious. Nonetheless, based on the phenotype of CD4 T cells derived from what are functionally scurfin knockout mice, i.e. mice hemizygous for the sf mutation, we predicted that overexpres- sion of scurfin would have a complementary effect on CD4 T cell effector function. Although sf CD4 T cells are hyper-responsive to activation, secreting augmented levels of cytokines and displaying enhanced proliferative capacity (8,9), we hypothesized that CD4 T cells overexpressing scurfin would display attenuated cytokine production in response to activating stimuli. We first tested this assertion directly, by determining the effect of SFN overexpression on activation-induced IL-2 expression in CD4 lineage T cells.
An expression construct encoding scurfin or the parent vector was transiently expressed in Jurkat cells as described under "Experimental Procedures." After transfection, cells were cultured in increasing concentrations of ␣CD3 (OKT3) for 24 h. Culture supernatants were collected and tested for the presence of IL-2 by bioassay using HT-2 cells. In the absence of the CD3 cross-linking, negligible amounts of IL-2 were produced in control transfected or SFN-overexpressing cells (Fig. 4, 0 g/ml ␣CD3, filled versus empty bars). As predicted, SFN overexpression (Fig. 4, empty bars, 1 and 10 g/ml ␣CD3) yielded a dramatic reduction in the amount of IL-2 produced by activated Jurkat cells, as evidenced by the decreased proliferative response of HT-2 cells. The reduction in IL-2 production was observed at both antibody concentrations. A comparable reduction in activation-induced IL-2 production was observed when Jurkat cells stably transfected with the SFN construct or CD4 T cells derived from mice that overexpress scurfin as a transgene were similarly analyzed (data not shown). 2 In the absence of functional SFN, the expression of IL-2 and other cytokines genes dependent on NFAT for transcription is heightened, whereas abnormally high levels of scurfin result in attenuated cytokine production. Given the observation that IL-2 production by CD4 T cells derived from sf mutant mice is refractory to inhibition with cyclosporin A (9), we surmised that NFAT-mediated transcription might be specifically affected by alterations in SFN expression. To address this possibility, we examined the effect of SFN overexpression on NFAT-mediated transcription using reporter gene analysis.
Owing to difficulties attaining consistent transient expres-sion of SFN in Jurkat cells, and to avoid the possibility that its constitutive overexpression might select for the outgrowth of cells with compensatory mutations, we utilized the tetracycline-inducible system offered by CLONTECH. D4 clones directing expression of empty vector (TO.TRE), scurfin (TO.SFN), or a deletion mutant lacking the FKH domain (TO.⌬FKH) were created (see the "Experimental Procedures" for details). These cell lines were transiently transfected with a luciferase reporter gene whose expression is driven by the SV40 promoter (pGL-3 Promoter) or a multimer of an essential regulatory NFAT site from the murine IL-2 promoter. The pSV-␤-gal reporter was co-transfected to allow correction for transfection efficiency. The control cell line, TO.TRE regulates NFAT-mediated transcription as expected (Fig. 5A, TO.TRE). There is low basal expression of the reporter gene (black bar) and activating stimuli (gray bar) results in a dramatic increase in luciferase production. As predicted, there is a significant (67%) reduction in PMA and ionomycininduced expression of NFAT-luc in the TO.SFN cell line (Fig.  5A, TO.SFN, gray bar). The FKH domain is required for this effect on cytokine gene transcription as NFAT-luc-mediated luciferase expression in the TO.TRE or TO.⌬FKH cell lines is indistinguishable. There is no significant difference between luciferase expression in TO.TRE, TO.⌬FKH, and TO.SFN in the absence of activating stimuli (Fig. 5A, black bars), suggesting that the difference between reporter gene expression in PMA-and ionomycin-treated TO.SFN or TO.TRE and TO.⌬FKH cells is not the result of global reduction of transcription in SFN overexpressing cells. This finding is also consistent with the observation that T cell activation is required to elicit disease in scurfy mutant mice (29). Importantly, expression of the SV40 promoter-driven luciferase reporter pGL-3 Promoter is the same in all cell types and for each condition (Fig. 5B), indicating that the reduction of NFAT-mediated transcription in TO.SFN cells is specific and not simply the result of reduced cell viability. Similar results have been observed when cross-linking of CD3 and CD28 is used as activating stimuli or when the proximal IL-2 promoter is used to drive luciferase expression (data not shown). We would predict that the activation-dependent transcription of other cytokines genes, documented to be dysregulated in sf mutant mice, would similarly be repressed in response to SFN overexpression. Given the ability of SFN to function as a transcriptional repressor and the observation that its overexpression affects IL-2 production and NFAT-mediated transcription, we wished to identify promoters targeted by SFN. Because IL-2, IL-4, GM-CSF, and TNF expression is dysregulated in mice with the sf mutation, and because NFAT is known to play a primary role in the regulation of these promoters, we examined the wellcharacterized regulatory sequences of these genes for putative FKH binding sites (8,9,(31)(32)(33)(34)(35). Unlike other DNA binding factors, FKH domain-containing proteins have been described to bind to a variety of sequences. Using several reported FKH binding sites to derive a "consensus" sequence ( Fig. 6A) to guide our search (20,36), we identified at least one putative FKH binding site in the minimal promoter of each of these genes. Interestingly, several of these were in close proximity to known NFAT binding sites.
The ability of SFN to bind to a putative FKH/NFAT composite site in the human GM-CSF enhancer (NFAT site Ϫ330), human IL-2 gene (NFAT site Ϫ125), and the murine IL-2 gene NFAT site used in the multimeric luciferase reporter (NFAT site Ϫ280) was assessed by EMSA. The sequences of these FKH/NFAT sites are shown in Fig. 6A. In the case of the GM-CSF enhancer, the potential FKH binding sequence is promoter-driven luciferase reporter construct pGL-3 Promoter (solid bars) or the 3xFKH-luc reporter (speckled bars) was co-transfected with each expression vector. The control plasmid pSV-␤-gal was also included in each reaction to allow correction for transfection efficiency. Cell lysates were prepared after 48 h of culture, and luciferase activity was determined using a luminometer. Luciferase activity, normalized to ␤-gal activity detected in duplicates of each lysate, is presented in arbitrary light units on the ordinate. The expression vector used for each data set is indicated on the abscissa. Error bars represent the S.E. of duplicate determinations on each lysate. The data are representative of three independent transfection experiments. located in the non-coding strand. Nuclear protein isolated from Jurkat cells activated with ␣CD3 treatment for 2 h was incubated with probe containing the V1P FKH binding site (Fig.  6B). As shown in Fig. 2B, a specific complex is formed between endogenously expressed SFN and the V1P probe (Fig. 6B, lane  1). Inclusion of a 50-fold molar excess of cold oligonucleotide competitor of the same DNA sequence as the probe (Fig. 6B,  lane 2) but not an irrelevant sequence (Fig. 6B, lane 3) competes for SFN binding. Putative composite FKH/NFAT sites from the GM-CSF enhancer (Fig. 6B, lane 4), IL-2 promoter (Fig. 6B, lane 5), and the NFAT reporter construct (Fig. 6B,  lane 6) all compete with the probe for SFN binding, confirming the proposed FKH binding sites in these cytokine gene promoters. The NFAT multimer-derived DNA sequence does not compete for complex formation as efficiently as the GM-CSF and IL-2 (Ϫ125 NFAT site) competitors. However, the corresponding Ϫ280 NFAT site in the human IL-2 promoter diverges from the murine sequence at three base pairs. Two of these changes are adjacent and fall in the putative FKH binding region, possibly explaining the less robust competition between the murine DNA sequence and human-derived nuclear protein.
The potential composite sites in the TNF and IL-4 gene promoters are similarly predicted to allow binding of FKH domaincontaining proteins. Although this finding does not demonstrate any functional significance of SFN binding to these sites, it is consistent with a model in which SFN attenuates the T cell activation response, at least in part, through repression of NFAT-mediated cytokine gene transcription. DISCUSSION The investigation of mice or human patients with immune dysfunction of unknown origin has led to the identification of a multitude of immunologically important molecules and has aided the elucidation of their roles in the immune response. The recent positional cloning of the gene defective in scurfy mutant mice and the discovery of phenotypically similar humans with mutations in the orthologous gene have provided us with the opportunity to gain further insight into regulation of T cell-mediated immunity (15)(16)(17)(18). The massive lymphoproliferation and autoimmune state documented in sf mutant mice and patients with the IPEX or XLAAD syndrome implicate the foxp3 gene product as a critical regulator of T cell homeostasis. In this paper we report the first direct characterization of the FOXP3 gene product, SFN.
Not surprisingly, the FKH domain contained in the protein mediates its nuclear localization and ability to bind DNA. Predominantly expressed in CD4 T cells, SFN present in nuclear extracts from cells of this type form a complex with the canonical FKH binding sites present in the TTR-S and V1P genes. The FKH binding site is rather degenerate, but reports have allowed the identification of a "consensus" sequence VAWTRT-TKRYTY (where V ϭ A/C/G; W ϭ A/T; R ϭ A/G; K ϭ G/T; Y ϭ C/T (20,36)). As observed with other families of DNA binding proteins with multiple members, e.g. the STAT family, the activity of each FKH domain-containing transcription factor may be regulated, at least in part, by its preference for variations of the canonical recognition sequence (37). Some reports indicate that amino acid residues adjacent to the DNA binding portion of FKH family members, such as HNF-3 and members of this subgroup, influence their binding specificities (36,38,39). Extrapolation of these findings to the FKH domain in SFN reveals that the A384T substitution mutation observed in some IPEX patients is adjacent to a residue predicted to contact DNA (17,18). Interestingly, the alanine present in wild type SFN diverges from a serine residue conserved at this position in most FKH family members, possibly affecting the DNA binding specificity of scurfin. Site selection studies will be required to definitively identify the preferred binding sequence for SFN.
Experiments with the 3xFKH-luc reporter construct indicate that SFN can function as a repressor of transcription. This finding is strengthened by the recent description of two other members of the Foxp subfamily (Foxp1 and Foxp2) that function as transcriptional repressors in the lung (40). Expression of the 3xFKH-luc reporter gene was similarly examined in the Tet i clones TO.TRE, TO.SFN, and TO.⌬FKH. The expression of 3xFKH-luc was reduced in TO.SFN, relative to the levels detected in TO.TRE and TO.⌬FKH. However, the results of these experiments cannot be definitively interpreted because several FKH family proteins are expressed in CD4 T cells and presumably capable of influencing expression of the FKH multimer (41)(42)(43). Verification of SFN repressor activity in T cells will require a different experimental approach. The use of a fusion protein between the DNA binding domain of Gal4 and SFN should allow the examination of the effect of SFN on transcription independent of the contribution of other FKH domaincontaining proteins (40). These experiments are currently being pursued.
The sf mutation results in a premature stop codon. Missense mutations described in the FOXP3 gene of males with the XLAAD or IPEX syndrome are expected to similarly disrupt SFN expression (16 -18). The dysregulation of T cell-mediated immunity resulting from the loss of protein function demonstrates the critical role SFN plays in the control of activationinduced T cell expansion and differentiation. Interestingly, the phenotype of patients with point mutations in the FOXP3 gene is just as severe, suggesting that the amino acids affected by these mutations may be critical for normal protein function (17). Most of these mutations lie in the FKH domain, possibly disrupting the DNA binding activity of the protein. Reproduction and study of these point mutants may prove invaluable for gaining further insight into the mechanism and regulation of SFN function.
Complementary to the phenotype observed in sf mice, SFN overexpression dampens activation-induced cytokine production (results shown here). 2 Augmented SFN expression also reduces the proliferative capacity of activated CD4 T cells (data not shown). 2 Taken as a whole, these findings implicate SFN as a potent regulator of T cell function, with the outcome of its activity being exquisitely sensitive to changes in its abundance. Although the results presented here indicate that SFN functions as a repressor of transcription, it is unclear whether the protein affects cytokine secretion and T cell proliferation directly, i.e. repressing the transcription of growth factors or genes promoting mitogenesis, or via an indirect mechanism, perhaps down-regulating the expression of genes that perpetuate an activated state in T cells. Our identification of FKH binding sites adjacent to critical NFAT elements in several cytokine promoters, favors the former hypothesis. Nonetheless, these possibilities are not mutually exclusive. Identification and more definitive study of promoters targeted by SFN will be required to elucidate the mechanisms of SFN action.
A final question is elicited by these initial studies. Assuming that SFN also functions as a repressor in T cells, what is the mechanism of SFN-mediated repression? Several models could account for its repressor activity. First, SFN could indirectly repress transcription by competing for binding of positive regulatory elements in genes induced following T cell activation. Our documentation of FKH binding sites flanking critical NFAT sites in several cytokine promoters is consistent with this possibility. Of note, a similar mechanism for repression of NFAT-mediated cytokine gene transcription in thymocytes has been described (44,45).
Another possibility is that SFN actively represses target promoters. Certainly other FKH proteins with direct repressor FIG. 6. FKH binding sites are adjacent to NFAT sites critical for activation-induced cytokine gene transcription. A, the consensus FKH binding sequence shown was used to search for putative FKH binding sites in the regulatory sequences of cytokine genes dysregulated by the sf mutation (V ϭ A/C/G; W ϭ A/T; R ϭ A/G; K ϭ G/T; Y ϭ C/T). Putative FKH binding sites (in boldface) were identified adjacent to NFAT sites (underlined) in the human GM-CSF enhancer (GM, non-coding strand), the human IL-2 promoter (IL2), and the NFAT site used in the multimeric luciferase construct (NF). The FKH site present in the V1 promoter is also included (V1P). B, EMSAs were performed with nuclear protein isolated from activated CD4 lineage T cells. The V1P FKH binding site was radiolabeled and incubated with nuclear protein isolated from Jurkat cells cross-linked with activity, e.g. Genesis and TTF-2, have been described (30,46). However these proteins, like most members of this family of transcription factors, have the FKH domain at the N terminus of the protein. Also, the TTF-2 repression domain has recently been mapped to a 21-amino acid region (30), and although there is a similar sequence present in Genesis, there is no homologous sequence readily identifiable in SFN.
A final possibility is that SFN could simply function as a scaffold, recruiting other proteins with repressor activity to target genes. The single zinc finger and the leucine zipper also present in SFN are obvious structural components that could mediate the protein-protein interactions required for such a mechanism. Indeed, the phenotype of an XLAAD patient with a single amino acid deletion in the putative leucine zipper was just as severe as that observed in patients harboring missense mutations in the gene, indicating a role for structural components outside of the FKH domain in protein function (16). Moreover, these structural features are conserved in the other members of this Fox subfamily (Foxp1 and Foxp2), and their suppressor activity has been mapped to a region containing both the zinc finger and leucine zipper (40). If SFN functions as a scaffolding protein required for the recruitment of other factors that negatively regulate transcription, one might expect to see a phenocopy of the sf mutation, IPEX or XLAAD syndrome with no apparent mutation in the FOXP3 gene of the affected individual. Of interest, one such patient with the IPEX syndrome has been described (17).
The severity of the immune dysregulation observed in mice with the sf mutation has for a long time been interpreted as an indication of the importance of the affected gene product in the regulation of T cell homeostasis. Identification of the foxp3 gene marked a significant advancement toward developing an understanding of how the encoded protein functions in this capacity. That this novel protein is a member of the widely studied FKH family of transcription factors should also help expedite the elucidation of its role in the immune response. The observation that changes in SFN expression levels drastically alter the response of CD4 T cells to activating stimuli indicates that identification of the components of this pathway may ultimately aid in the development of therapeutics to treat immunodeficient patients or those afflicted with autoimmune disease.