The autoimmune regulator (AIRE) is a DNA-binding protein.

The autoimmune regulator (AIRE) protein is a putative transcription regulator with two plant homeodomain-type zinc fingers, a putative DNA-binding domain (SAND), and four nuclear receptor binding LXXLL motifs. We have shown here that in vitro, recombinant AIRE can form homodimers and homotetramers that were also detected in thymic protein extracts. Recombinant AIRE also oligomerizes spontaneously upon phosphorylation by cAMP dependent protein kinase A or protein kinase C. Similarly, thymic AIRE protein is phosphorylated at the tyrosine and serine/threonine residues. AIRE dimers and tetramers, but not the monomers, can bind to G-doublets with the ATTGGTTA motif and the TTATTA-box. Competition assays revealed that sequences with one TTATTA motif and two tandem repeats of ATTGGTTA had the highest binding affinity. These findings demonstrate that AIRE is an important DNA binding molecule involved in immune regulation.

The autoimmune regulator (AIRE) protein is a putative transcription regulator with two plant homeodomain-type zinc fingers, a putative DNA-binding domain (SAND), and four nuclear receptor binding LXXLL motifs. We have shown here that in vitro, recombinant AIRE can form homodimers and homotetramers that were also detected in thymic protein extracts. Recombinant AIRE also oligomerizes spontaneously upon phosphorylation by cAMP dependent protein kinase A or protein kinase C. Similarly, thymic AIRE protein is phosphorylated at the tyrosine and serine/threonine residues. AIRE dimers and tetramers, but not the monomers, can bind to G-doublets with the ATTGGTTA motif and the TTATTA-box. Competition assays revealed that sequences with one TTATTA motif and two tandem repeats of ATTGGTTA had the highest binding affinity. These findings demonstrate that AIRE is an important DNA binding molecule involved in immune regulation.
Autoimmune polyendocrinopathy candidiasis ectodermal dystrophy (APECED), 1 also known as autoimmune polyglandular syndrome type 1 (APS1), is a rare autosomal recessive disorder common in isolated populations such as Finns, Sardinians, and Iranian Jews (1). This syndrome is characterized by destructive autoimmune diseases of the endocrine organs, chronic candidiasis of mucous membranes, and ectodermal disorders. APECED is caused by mutations in the autoimmune regulator (AIRE) gene on chromosome 21q22.3 (2)(3)(4). The AIRE gene has recently been cloned by two independent groups of investigators (5,6). The AIRE gene consists of 14 exons coding for a 2445-base pair mRNA transcript, and the translated product is expected to have 545 amino acids with a predicted molecular mass of 57.5 kDa. The predicted AIRE protein has several domains indicative of a transcriptional regulator protein (6). AIRE harbors two zinc fingers of plant homeodomain (PHD) type. A putative DNA binding domain named SAND as well as four nuclear receptor binding LXXLL motifs, an inverted LXXLL domain, and a variant of the latter (FXXLL) hint that this protein functions as a transcription coactivator (5)(6)(7). Furthermore, a highly conserved N-terminal 100-amino acid domain in AIRE has a significant homology to the homogenously staining (HSR) domain of Sp100 and Sp140 proteins (7). This domain has been shown to function as a dimerization domain in several Sp-100 related proteins (8). At the subcellular level, AIRE can be found in the cell nucleus in a speckled pattern in domains resembling promyelocytic leukemia nuclear bodies, also known as ND10, nuclear dots, or potential oncogenic domains, associated with the AIRE homologous nuclear proteins Sp100, Sp140, and Lysp100 (9).
Interestingly, it has recently been shown that AIRE can activate transcription from a reporter gene when fused to a heterologous DNA binding domain. This activation required the full-length protein or the presence of more than one activation domain. A glutathione S-transferase pull-down assay showed that AIRE formed homodimers in vitro, probably through the N-terminal domain (amino acids  or through the minimal 1-100 amino acid domain resembling Sp-100. It has also been shown that AIRE interacts in vitro with CREBbinding protein (CBP) through the CH1 and CH3 conserved domains, which has led the investigators to suggest that the transcriptional activities of AIRE might be mediated through its physical interaction with the common coactivator CBP (10). CBP is a key coactivator that modulates the transcriptional regulation dependent on adenylate cyclase-signaling pathway in eukaryotes. The signal processing is mediated by a family of cyclic AMP-responsive nuclear factors, including CREB, cAMP response element modulator (CREM), and activating transcription factor 1 (ATF-1). These factors contain the basic domain/leucine zipper motifs and bind as dimers to cAMP-responsive elements (CREs). The activation function of CRE-binding proteins in turn is modulated by several kinase-dependent phosphorylations and is mediated by coactivators such as CBP and p300 (11).
Although it is important to note that the interaction of AIRE with CBP might influence signal transduction mediated by cAMP-responsive nuclear factors, some of the structural properties of AIRE are suggestive of its function beyond the suspected transactivation. The presence of two PHD zinc fingers and a leucine zipper was the focus of our attention. The objectives of the present study were 1) to verify whether recombinant human AIRE oligomerizes in vitro and 2) to examine the possibility of AIRE possessing DNA binding properties. In this paper, we report that AIRE forms homodimers and homotetramers in vitro and that both the dimer and tetramer forms of AIRE possess DNA binding activity. We further demonstrate that AIRE exists in oligomerized forms in vivo and that phosphorylation of AIRE by cAMP-dependent PKA and/or PKC could trigger its dimerization. * This work was supported by National Institutes of Health Grants 1R24DK58778 (NIDDK), 1RO1HD37800 (NICHD), and 1P01AI-42288 (NIAID) and by a grant from the Juvenile Diabetes Research Foundation. 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 U.S.C. Section 1734 solely to indicate this fact.

MATERIALS AND METHODS
Plasmid Constructs Protein Expression-The full-length AIRE (amino acids 1-545) was amplified by polymerase chain reaction. The fragment was cloned into pET32 containing a thioredoxin tag (Novagen) for 6ϫHis fusion protein expression. Escherichia coli BL21 (DE3) were transformed and selected on ampicillin. Starting with a single colony, 500 ml of an overnight culture of a positive clone was set up. The expression of the AIRE gene was induced with 10 M isopropyl-1-thio-␤-D-galactopyranoside, and the culture was continued for 3 h. The cells were harvested, and the recombinant protein was purified using the B-PER 6ϫ histidine fusion protein purification system (Pierce) following the manufacturer's instructions with slight modifications. Briefly, the cells harvested from fresh induction were resuspended in 20 ml of B-PER reagent by pipetting up and down until the cell suspension was homogenous. The homogenous suspension was shaken gently at room temperature for 10 min. The soluble proteins were separated from the insoluble fraction by centrifugation at 27,000 ϫ g for 15 min at 4°C. The supernatant was separated into two aliquots of 10 ml each and loaded onto two nickel-chelated columns equilibrated with B-PER reagent. The columns were washed three times with the wash buffers following the instructions of the manufacturer. The elution was performed on an imidazole gradient (0.1-1ϫ), and 20 fractions of 0.5 ml each were collected. The fractions were analyzed on SDS-PAGE, the fractions containing the recombinant AIRE were pooled, and the protein concentration was adjusted to 10 g/l. Protein concentrations were determined by BCA protein assay reagent (Pierce) using the standard protocol. This preparation is designated as AIRE monomer.
Oxidation and Refolding of AIRE-A 20-ml fraction of the eluate pool containing AIRE was resuspended in 5 volumes of ice-cold water containing 10 mM benzamidine hydrochloride. Guanidine hydrochloride and sodium phosphate (pH 7.0) were added such that the final concentrations of 6 M guanidine and 10 mM phosphate were obtained, and the mixture was allowed to stir at room temperature for 2 h. The denatured soluble fraction was then diluted 10-fold over the course of 2 h into 1 liter of ice-cold oxidation, refold buffer containing Tris-HCl (pH 9.0), 1 M urea, and a redox couple (3 mM L-cysteine, 0.3 mM cystamine dihydrochloride). The solution was kept at 4°C for 20 h. The solution was then concentrated, and buffer was exchanged against 10 mM sodium phosphate (pH 7.0) containing 2 M urea using Centricon centrifugal filtration device (Millipore) with a molecular weight cut-off of 10,000. This preparation was then adjusted to pH 5.2 with acetic acid and allowed to stir at 4°C overnight. The solution was clarified by centrifugation, adjusted to pH 8.0, and concentrated, and the buffer was exchanged against Tris buffer (pH 8.5) using Millipore centrifugal filtration device. As a final refolding step, the pH of the preparation was adjusted to pH 5.0 with acetic acid, and buffer exchange was performed with 10 mM sodium phosphate buffer containing 150 mM NaCl (pH 7.2). The final volume of the preparation was adjusted so as to have a protein concentration of 10 g/l. This preparation is designated as refolded AIRE.
Immobilization of Anti-AIRE Antibody on Carbolink TM Gel-Total IgG from the sera of rabbits immunized with recombinant human AIRE (rhAIRE) was purified using IgG purification kit (Pierce) with manufacturer's instructions. The IgG was oriented on a Carbolink TM gel employing an immobilization chemistry that used sodium periodate to oxidize carbohydrates located on the Fc portion of antibody molecules, converting vicinal hydroxyl groups to reactive aldehyde groups. These aldehyde groups were allowed to react with the hydrazide groups of the resin to form hydrazone bonds. The resin is designated as immobilized anti-AIRE hereafter. IgG recovered from non-immunized animals were also immobilized in a similar fashion, which is designated as immobilized normal rabbit IgG.
Immunoprecipitation of AIRE from Mouse Thymus-Thymus tissues from four immature B6 mice were excised and homogenized in radioimmune precipitation buffer (50 mM Tris, pH 8.0, 1 mM EDTA, 150 mM NaCl, 0.1% SDS, 1% Nonidet P-40, and 0.5% sodium deoxycholate) containing 0.1% phenylmethylsulfonyl fluoride (1 ml of buffer/mg of fresh tissue weight). The tissue homogenate was cleared by centrifugation at 20,000 ϫ g for 20 min at 4°C. The supernatant was recovered and pre-cleared with immobilized normal rabbit IgG (60 l of hydrated beads/1 ml of lysate). The mixture was incubated for 1 h at 4°C on a rocker. The supernatant was recovered after centrifugation at 16,000 ϫ g (5 min) and mixed with immobilized anti-AIRE antibody. The mixture was incubated overnight at 4°C on a rocker. After incubation, the preparation was centrifuged at 16,000 ϫ g for 5 min to recover the beads. The beads were washed five times with 1 ml of fresh radioimmune precipitation buffer each time. Bound AIRE recovered from the immunocomplex was batch-eluted using a 100 -500 mM glycine gradient and buffer-exchanged with phosphate-buffered saline.
For electroblotting, the gels were recovered and introduced into the transfer module. The proteins were transferred onto a PVDF membrane pre-wet in methanol and equilibrated in the transfer buffer using a Mini Transblot cell (Bio-Rad). The transfer buffer had the following composition: 40% methanol (v/v), 25 mM Tris, and 190 mM glycine (pH 8.2). The transfer was performed at 30 mA of constant current for 12-16 h, after which the PVDF membrane was recovered and air-dried.
For the development of blots, the membranes were pre-wet in methanol and then incubated in the Opti-4CN blocking reagent (Bio-Rad) for 1 h at 37°C followed by extensive washing in phosphate-buffered saline with 0.1% Tween 20 at room temperature. The membranes were incubated with appropriate primary antibodies for 1 h at room temperature. After incubation, the membranes were washed with three changes of phosphate-buffered saline, 0.1% Tween 20 and then incubated with goat anti-rabbit IgG-horseradish peroxidase at a dilution of 1:5000. The membranes were washed extensively after incubation and then developed with amplified Opti-4CN reagents (Bio-Rad) following the recommendations of the manufacturer.
N-terminal Microsequencing-N-terminal microsequencing of the 54-, 110-, and 220-kDa bands affinity-purified from mouse thymus on anti-AIRE columns was performed as follows. 100 -300 mM glycine (pH 2.5) eluates from immobilized anti-AIRE immunocomplexes were pooled and separated on native-PAGE polymerized for 18 h after the protocol described in the earlier section. The separated proteins were transferred onto a PVDF membrane (Millipore pSQ) pre-wet in methanol and equilibrated in the transfer buffer (0.1 M MES buffer (pH 6.0) containing 40% v/v methanol). The transfer was performed at 30 mA of constant current for 12-16 h, after which the PVDF membrane was recovered and washed extensively in deionized water. The membranes were briefly stained in Coomassie Brilliant Blue R-250 for 1 min and de-stained in two changes of de-staining solution. The membranes were dried, and bands were excised. The excised bands were subjected to Edman degradation and the PTH derivatives were detected on a microbore liquid chromatography system.
Labeling of Oligonucleotides-The oligonucleotides used in this study are listed in Table I. The oligonucleotides were end-labeled with [␥-32 P]ATP after a standard T4 polynucleotide kinase reaction. The phosphorylation reaction was continued for 30 min at 37°C, after which the reaction was stopped by the addition of 1 l of 0.5 M EDTA/10 l of total reaction volume. The unincorporated label was removed using QIAQUICK nucleotide removal kit (Qiagen) following the instructions of the manufacturer. The labeled oligonucleotides were eluted in 100 l of TE buffer. The specific activities of the labeled oligonucleotides were measured and were confirmed to be close to 20,000 cpm/100 fmol before they were used in gel-shift assays.
Gel Shift Assays-Gel shift assays were performed with purified AIRE (100 pM/reaction), and radiolabeled double-stranded oligonucleotides were prepared as stated above. The oligonucleotides used in this study are listed in Table I. The protein preparations were preincubated in binding buffer (10 mM Tris-HCl (pH 7.5), 50 mM NaCl, 0.5 mM dithiothreitol, 0.5 mM EDTA, 1 mM MgCl 2 , 4% glycerol) in the presence of 0.05 mg/ml poly(dI-dC)⅐poly(dI-dC). Competition studies were performed by the addition of 1.75 pmol of unlabeled double-stranded oligonucleotide into the preincubation step. DNA-protein complexes were formed by the addition of 0.175 pmol (40 -100,000 cpm/liter) of probe and resolved by separation from unbound radiolabeled oligonucleotides through 6% non-denaturing polyacrylamide gels (acrylamide:bisacrylamide ratio, 80:1) containing 2.5% glycerol in 0.5ϫ TBE (44.5 mM Tris, 44.5 mM boric acid, 1 mM EDTA) using a Protean II electrophoresis system (Bio-Rad). The gels were pre-run at 100 V for 30 min before loading of the binding reaction. After loading the samples, the DNAprotein complexes were separated by electrophoresis at 150 V for 3.5 h. All procedures were performed at room temperature. After electrophoresis, the gels were exposed to x-ray film (Fuji Film).
Gel shift competition assays were performed by the introduction of a synthetic zinc finger peptide (ZIF 268) in the reaction. The basic reaction was set up as stated above, with a modification involving the addition of ZIF 268 at a concentration of 100 pM. This reaction was allowed to incubate for 15 min at room temperature, after which AIRE was introduced at concentrations ranging from 100 pM through 800 pM. Gel shift assays of these reactions were performed in the same fashion as discussed above.
Oligo Library Screening-We determined the ability of purified recombinant AIRE to bind DNA employing two oligo libraries. The first library (N 26 oligo library) had randomized a 26-mer stretch flanked by arms for PCR amplification. The second library (G 2 N 7 G 2 library) had N 9 GGN 7 GGN 6 sequence flanked by adapter sequence for PCR amplification. The libraries were PCR-amplified under standard conditions (melting at 94°C for 2 min followed by cycling at 94°C for 30 s, annealing at 63°C for 30 s, and extension at 72°C for 1 min; a final hold at 72°C for 10 min was given for completion of chain elongation). After amplification, the PCR product was suspended in 1ϫ binding buffer and incubated with purified and refolded rhAIRE for 1 h at room temperature with constant shaking. After this, anti-AIRE antibody (1:100 dilution) was added to the mixture, and the incubation was continued for an additional 1 h. Protein A-agarose with preblocked nonspecific binding sites (1:10 dilution) was added to the mixture, and the incubation was allowed to continue for 1 h. At the end of this final incubation, the suspension was centrifuged at 16,000 rpm on an Eppendorf centrifuge. The supernatant was carefully removed, and the pellet was washed 5 times with 1.5 ml of 1ϫ binding buffer each time. The final pellet was resuspended in 1ϫ TE. A 2-l aliquot of the resuspended pellet was PCR-amplified under the settings mentioned above. The whole procedure was repeated 5 times as discussed earlier. After the last amplification, the PCR product was separated on a 3% agarose gel, and the band was eluted and cloned using pCR4-TOPO cloning kit for sequencing (Invitrogen). TOPO10 cells were transformed adopting a one-shot chemical transformation protocol.
Thirty-two clones each were selected and sequenced from the AIREselected N 26 library and the G 2 N 7 G 2 library. Plasmid minipreps were made from the selected colonies and sequenced using M13 forward primer. The insert sequences were aligned. Two different controls were included in this experimentation. The first control involved the PCR amplification of G 2 N 7 G 2 library (before selection by AIRE) and its cloning into pCR4-TOPO vector. Forty-four clones were sequenced, and the sequences were analyzed to evaluate the bias in the random areas of the oligonucleotides in this library (control 1). The second control involved the cloning of oligonucleotides selected by reagents excluding AIRE from the experimental reaction set-up (control 2).
Relative Binding Affinity Assays-Relative binding affinity assays were performed in a manner similar to that outlined for gel-shift assays. In this assay, we constructed four independent sets of four reactions, each using hot oligonucleotides oligo-TGG, oligo-TG, oligo-GG, and oligo-G, respectively. Each set had a gel shift with the given oligonucleotide and competitions with the other three oligonucleotides. The concentrations of the protein, hot oligonucleotide, and competitors were the same as those in the gel-shift assays detailed earlier in this section.
In Vitro Phosphorylation Assay-rhAIRE monomer was resuspended in kinase buffer containing 40 mM HEPES (pH 7.4), 10 mM MgCl 2 , and 3 mM MnCl 2 . 100 g/ml ATP was added into the suspension. The kinase reaction was initiated by the introduction of the corresponding protein kinases. We tested the phosphorylation of AIRE by Abl protein kinase (New England Biolabs), casein kinase I, casein kinase II (Promega Corp.), cAMP-dependent protein kinase A and PKC (Cell Signaling technology). The reactions were incubated for 30 min at 27°C, after which AIRE was immunoprecipitated with anti-AIRE-agarose. The immunocomplexes were dissociated with 100 mM glycine buffer (pH 2.6), and the anti-AIRE-agarose conjugate was separated by centrifugation. The supernatant was neutralized with 100 mM Tris buffer (pH 10.0). The eluates were separated on native PAGE and transferred onto PVDF membranes. The membrane strips were probed with the corresponding anti-Tyr(P) or anti-Ser(P)/Thr(P) antibodies. One of the strips was probed with anti-AIRE antibody to verify the position of AIRE.
Homology Modeling-The full-length sequence of human AIRE (Gen-Bank TM accession number NM 000383.1) was subjected to a global threading, and significant threading matches were filtered out considering high global and local Z scores. This process yielded a few templates for further use in homology modeling. The templates used include 11GSA, 1PYDA, 1OTCA, 1PRCH, 2RAMB, and 2YHX. Alignments were performed using the program package GCG, and the model was built using Insight II and Discover (Molecular Simulations Inc., San Diego, CA) and evaluated with Swissmodel program (12).

RESULTS
AIRE Purification, Antibody Production, and Refolding Assays-The full-length construct of human AIRE (hAIRE) in pet32 showed a very low basal level of expression of AIRE, which was augmented severalfold after induction with isopropyl-1-thio-␤-D-galactopyranoside (Fig. 1A, lanes 2 and 3). Because B-Per reagent could extract AIRE in the soluble fraction, as shown in lane 3, we discarded the pellet after confirming the efficiency of the recovery. The chelated nickel affinity chromatography followed by imidazole gradient elution produced a protein fraction at an estimated molecular size of 56 kDa, as represented in lane 4 (Fig. 1A), which shows a purified fraction of AIRE. This fraction is termed AIRE monomer for descriptive purposes.
The processing of the rhAIRE under oxidation and refolding conditions brought about a major shift in the pattern of migration of this molecule (Fig. 1A, lane 5). Although a minor fraction of AIRE still appears at the 56-kDa position, the major bands appear at estimated molecular sizes of 110 and 220 kDa when the refolded rhAIRE was analyzed on native PAGE. Both these bands migrated to the 56-kDa position on SDS-PAGE in the presence of 8 M urea (Fig. 1A, lane 6), suggesting the formation of dimers and tetramers of AIRE during the refolding process.
AIRE Forms Dimers and Tetramers in Vivo-Polyclonal antibodies raised in rabbit in our laboratory against the full- length rhAIRE recognized the AIRE in mouse thymus extracts on Western blots (Fig. 1B, lane 1). Along with the expected 56-kDa band, we also observed the presence of anti-rhAIREpositive bands at molecular weight positions of 110 and 220 kDa, suggesting the possible oligomerization of AIRE in vivo (Fig. 1B, lane 1). Both these high molecular weight bands were microsequenced, confirming that they are derivatives of AIRE (data not shown).
AIRE Is Phosphorylated in Vivo-A computer-based prediction (www.cbs.dtu.dk/services/NetPhos) of the properties of AIRE had indicated that this molecule is quite possibly phosphorylated at serine and/or at threonine at several locations, with one possible tyrosine phosphorylation site at Tyr-394. To verify this prediction, we immunoprecipitated AIRE from total protein extract of mouse thymus using anti-AIRE antibody immobilized on Carbolink TM resin using a Carbolink TM IgG Immobilization Kit (Pierce). Immunoprecipitated AIRE was recovered using a glycine gradient. The fractions were neutralized, separated on native PAGE, and analyzed on Western blots. Native immunoblots of the immunoprecipitated fractions probed with anti-AIRE antibody detected bands at 56, 110, and 220 kDa, indicating the presence of monomers, dimers, and tetramers of AIRE in tissue extracts (Fig. 1B, lane 1). Identical lanes were probed with anti-Ser(P)/Thr(P) (Fig. 1B, lane 2) and anti-Tyr(P) (Fig. 1B, lane 3) antibodies, which indicated that AIRE existed in a phosphorylated condition in mammalian tissues.
AIRE Can Be Phosphorylated in Vitro-Five different kinases (Abl protein kinase, CK1, CK2, cAMP-dependent PKA and PKC) were selected considering the predicted ability (www.cbs.dtu.dk/services/NetPhos) of one or all of them to phosphorylate recombinant human AIRE produced by prokaryotic expression. Abl protein kinase and the casein kinases used in this experiment could not bring about phosphorylation of rhAIRE in vitro (Fig. 1C, lanes 2-4). Interestingly, cAMP-dependent PKA brought about phosphorylation of rhAIRE (Fig.  1C, lane 5). To our surprise, we observed that the rhAIRE band shifted toward the 100-kDa position, indicating a possibility of dimerization of this molecule in response to phosphorylation. Protein kinase C also brought about phosphorylation of rhAIRE. Again, rhAIRE appeared to dimerize in response to phosphorylation although the monomer form of rhAIRE was also visible after the termination of the phosphorylation reaction (Fig. 1, lane 6).
AIRE Can Bind to a Zinc Finger Binding Consensus-We screened seven consensus sequences known to be recognized by zinc finger/leucine zipper proteins (Table I) to examine the possibility of AIRE interacting with DNA through its zinc finger domains. In our initial attempts, we conducted gel shift assays with monomeric AIRE. None of the oligonucleotides brought about a gel shift when we used AIRE monomer in the assays (data not shown). Interestingly, the refolded recombinant protein showed a strong gel shift with EGR consensus, producing two bands corresponding to the respective dimer and tetramer positions of AIRE (Fig. 2A, lane 2). The addition of a 10 M excess of cold consensus oligonucleotide brought about significant reduction in the intensity of the gel-shifted bands, signifying the specificity of the interaction (Fig. 2A, lane 3). Furthermore, the use of a mutant EGR consensus with base replacements at the underlined positions (GAATTCAGCGGG-GGCGAGCGGGGGCGA, GG 3 TA in mutant) did not bring about competitive removal of the gel-shifted bands ( Fig. 2A,  lane 4). Moreover, hot mutant EGR consensus also failed to show a gel-shifted band, confirming this observation ( Fig. 2A,  lane 5). This indicated the possible requirement of those underlined guanine residues in this oligonucleotide for its binding to AIRE.
To further confirm that the zinc finger domain of AIRE is involved in this interaction, we performed a competition assay where AIRE was made to interact with EGR consensus in the presence of a known synthetic zinc finger peptide ZIF 268 showing strong affinity toward the EGR consensus. In the absence of the competitor peptide, AIRE brought about a gel shift as shown in Fig. 2A, lane 2. In the presence of 100 pM ZIF 268, AIRE could not bring about visible mobility shifts over a concentration range of 100 -300 pM (Fig. 2B, lanes 3-5), but a 4-fold excess of AIRE (400 pM) restored the mobility shift (Fig.  2B, lane 6). A further increase in the concentration of AIRE resulted in the appearance of the lower gel-shifted band (corresponding to AIRE dimer) as well (Fig. 2B, lanes 7-10).
Consensus Sequence Recognized by AIRE Is Partly Different from That of EGR-We screened an oligo library with preselected guanine doublets with a random seven-nucleotide spacer (G 2 N 7 G 2 library) and a random oligo library (N 26 library) to identify the AIRE binding consensus sequence. The libraries were subjected to five generations of pull-down assays, and the products after the fifth amplification were cloned and sequenced.
We analyzed 59 clones selected out by AIRE from the G 2 N 7 G 2 library by direct sequencing using M13 forward primer. The sequences were aligned and shaded using Genedoc software (Fig. 3A). Alignment of the sequences from 32 clones is presented because of software limitations. A quantity-based generation of consensus from the aligned sequences was performed using Genedoc software, selecting the base that shows the highest frequency of representation at every given position.
The consensus line appears at the bottom of Fig. 3A. The pre-selected G-doublets are shaded gray in the consensus line and shaded black in the alignment. It is clear that AIRE preferred an A/T-rich neighborhood around GG for its binding. A careful examination of the consensus line showed that this sequence contained two ATTGGTTA motifs (underlined in double lines) and a weak TATA-box (TTATTA) (underlined in a single line).
To verify whether this observation was because of a bias in the library used, we cloned the unselected library into TA-  TGG  TTATTAATTGGTTATATTGGTTA  TG  TTATTAATTGGTTA  GG  ATTGGTTATATTGGTTA  G  ATTGGTTA cloning vector, and 44 clones were sequenced (control 1). Fig.  3B represents an alignment of the sequences from the unselected clones analyzed. The consensus sequence was generated as described above. The library had a G/C-bias, as could be interpreted from Fig. 3B. This observation strengthens the conclusions derived from Fig. 3A that rhAIRE selected A/T-rich motifs around the G doublets. In the screening experiment in the absence of AIRE (control 2), no PCR amplification occurred after the second generation of immunoprecipitation, indicating the specificity of the complex formation and the screening experiments.
The sequences of the clones derived from the N 26 library were aligned as shown in Fig. 3C. AIRE showed clear selection of two different motifs from the oligo library pool, the first one being a T-box (TTATTA) (underlined in single line in the consensus line) and the second a G-box (ATTGGTTA) (underlined in double lines). Three clones (clones 6, 32, and 25 in Fig. 3C) possessed only a T-box, and three (clones 8, 9, and 10) did not show any clear consensus motif. The majority of the selected clones had either a TG-box (12 of 32) or a GG-box (11 of 32) combination. A minor population (3 of 32) had TGG (clones 15, 18, and 22 of Fig. 3C). This observation is in perfect agreement with the interpretations that we deduced from our G 2 N 7 G 2 library-screening experiments.
Relative Binding Affinity of AIRE to TTATTA and to ATTG-GTTA Motifs-To address the relative binding affinities of AIRE to the TTATTA (T-box) motif and to the ATTGGTTA motif (G-box), we performed gel shift assays and competitions with various combinations of four oligonucleotides: oligo-TGG, oligo-TG, oligo-GG, and oligo-G (Table I). Two major gel shift patterns emerged from these analyses. When the labeled oligo-GG and oligo-G are used, two shifted bands were observed (Fig. 4 lanes 9 -16). The upper band corresponds to the AIREtetramer position, and the lower band corresponds to the dimer position. Interestingly, a higher band was observed when both labeled oligos with a T-box (oligo-TGG and oligo-TG) were analyzed (Fig. 4, lanes 1-8).
AIRE showed strong binding with hot oligo-TGG (Fig. 4, lane  4), which could be competed out substantially with cold oligo-TG (Fig. 4, lane 1). Cold oligo-GG brought about partial inhibition in gel shift (Fig. 4, lane 2), but cold oligo-G did not reduce the shifted band of hot oligo-TGG (Fig. 4, lane 3). Oligo-TG also showed good binding with AIRE protein (Fig. 4,  lane 8). This binding could be totally abolished in competitions with oligo-TGG (lane 5), oligo-GG (lane 6), or oligo-G (lane 7). Furthermore, the binding of AIRE to oligo-GG (lane 12) could be blocked significantly by oligo-TGG (lane 9) and to a lesser extent by oligo-TG (lane 10) and oligo-G (lane 11). The binding of AIRE to oligo-G showed a pattern similar to that of oligo-GG (Fig. 4, lanes 13-16). It is interesting to note that the shifted AIRE dimer is not detectable without competitor oligos (Fig. 4, lanes 12 and 16) but appears in the presence of competitors. This is because the labeled oligos only become available for binding with the AIRE dimers when the competitor oligos occupied the binding sites of the AIRE tetramers. These results further suggest that the AIRE tetramers have higher binding affinity than the AIRE dimers.
Homology Modeling-A threading of the sequence of the full-length human AIRE protein returned several templates that were used to generate an energy-minimized model for the full-length protein (Fig. 5). The model depicts an important conformational ingenuity in that both the PHD domains and the SAND domain are arranged in a linear fashion, providing maximal DNA binding surface. The positioning of the leucine zipper well within the middle DNA binding domain (PHD1) would permit the zip-locking of the protein dimer onto the DNA. The leucine-zipper skeleton is shown in red in Fig. 5.

DISCUSSION
Because of the presence of very characteristic domains found in classical DNA-binding proteins in AIRE, it was suspected from the very day of its discovery that AIRE may function as a DNA-binding protein. Though the DNA-binding activity of this protein was not established, recent investigations by a few groups were successful in unfolding the molecular basis of AIRE action. Recently, Bjorses et al. (9,13) demonstrate that wild type AIRE was translocated into the nucleus where it forms distinct speckled domains throughout the nucleoplasm, excluding the nucleoli. These authors further demonstrate the transcriptional transactivation potential of this protein by means of cotransfection experiments with Gal4 DBD-AIRE fusion protein and the luciferase reporter genes under either thymidine kinase promoter or the adenovirus E1b TATA sequence with upstream Gal4 response elements (13). Later, Pitkanen et al. (10) performed a series of Gal4 system reporter assays in HUH-7 and COS-1 cells, which showed that the expression of full-length AIRE cDNA as the fusion protein with GAL4 DNA binding domain caused ϳ30 -40-fold activation of the chloramphenicol transferase (CAT) reporter from the pG5-CAT plasmid containing three GAL4 response elements upstream of the E1b minimal promoter (10).
The ability of AIRE to bind itself was recently assessed using a glutathione S-transferase pull-down assay. It was found that in vitro translated, labeled full-length AIRE bound specifically to glutathione S-transferase-AIRE fusion but not to glutathione S-transferase alone, suggesting that AIRE homodimerizes  (Table I)  in vitro very much like Sp100 (10). Recombinant full-length hAIRE from the E. coli expression system did not dimerize, probably because post-translational processing of AIRE may be prerequisite for dimerization. Because it is known that proteins could be unfolded and refolded successfully in vitro, our initial attempts were to examine whether we could develop a strategy to refold rhAIRE in vitro. Under the conditions stated in this paper, AIRE refolded and dimerized as expected. At the same time, we observed the formation of tetramers as well (Fig. 1). If tetramer formation were not an artifact, it would imply that AIRE has dimerization and tetramerization domains in its structure. This appears to be the case if we combine the observation of Pitkanen et al. (10) that AIRE could dimerize utilizing its N-terminal HSR domain and the fact that AIRE does have a leucine zipper extending over amino acids 319 -341. At this moment, the possibility of oligomerization of AIRE through its leucine zipper is not experimentally proven. But it is established that leucine zipper proteins dimerize, engaging the zipper domains of the interacting partners. In that event, it is quite likely that AIRE might dimerize using the leucine zipper motif and would tetramerize using the HSR domain.
The functional significance of the formation of dimers and tetramers of AIRE would depend on whether such molecular interactions take place in vivo or not. To address this issue, we performed a native PAGE followed by Western blot analysis of human thymus extracts. Anti-rhAIRE antibody recognized three bands at molecular weight positions of 56, 110, and 220 kDa. These are the expected positions of monomer, dimer, and tetramer of AIRE in the respective order (Fig. 1). The identity of the high molecular weight bands as derivatives of AIRE was established by microsequencing analysis. Thus, the potential presence of dimer and tetramer forms of AIRE in human thymus is established.
Homo-and/or hetero-oligomerization of proteins is important in the modulation of their function. In vivo, proteins oligomerize in response to stimuli of intrinsic or extrinsic origin. Phosphorylation or dephosphorylation of proteins at specific targets by respective kinases or phosphatases is a common pathway through which biological signal transduction propagates. Because recombinant AIRE did not possess the capability to autodimerize, we suspected a possible activation of this molecule by phosphorylation that might be required for its dimerization. A computer-based prediction of the possible phosphorylation sites in AIRE pointed toward a high chance of phosphorylation of serine/threonine at multiple locations. In the light of this prediction, we tried a series of kinases for their ability to phosphorylate AIRE in vitro. Interestingly, we found that cAMPdependent protein kinase A and PKC could heavily phosphorylate AIRE (Fig. 1C). It is also important to state at this point that the phosphorylation of AIRE by cAMP-dependent PKA and PKC also led to the dimerization of this molecule. Although the relevance of this in vitro demonstration cannot be applied to the in vivo scenario, we feel that there is a strong possibility that AIRE is a substrate for cAMP-dependent protein kinase A and/or PKC. Furthermore, phosphorylation of AIRE appears to be the trigger for its dimerization in vitro and possibly in vivo. This observation, in conjunction with the report that AIRE interacts with CREB-binding protein (10) suggest that AIRE could act as a downstream modulator of cAMP-dependent signal transduction and transcription regulation. Because the human CREB protein contains the leucine zipper motif (14), it is also possible that there is a direct interaction between CREB and AIRE.
It is well known that zinc finger proteins represent an important class of regulatory DNA-binding proteins. Different varieties of zinc finger proteins are all linked by the utilization of zinc ions to add a structural element to the binding component. One characteristic of this motif is that it recognizes Grich DNA (15)(16)(17)(18)(19). Because AIRE contained both leucine zipper and zinc finger domains, we decided to characterize the potential DNA binding activity of AIRE using oligonucleotides known to be recognized by proteins having these distinct domains. Thus, we performed gel shift assays with AIRE and consensus sequences recognized by a number of leucine zipper/ zinc finger proteins (Table I). The results showed that AIRE introduced band shift with the EGR consensus but not with the EGR nonconsensus oligonucleotide. The band shift could be competed out with an excess of cold EGR oligonucleotide and also with ZIF 268 synthetic peptide, indicating that the band shift is introduced by the interaction of zinc finger motifs in AIRE and EGR consensus (Fig. 2). Although AIRE monomer failed to bring about band shift, the AIRE dimer and tetramer forms brought about strong band shifts under our assay conditions, implicating the requirement of oligomerization of this molecule for it to gain the DNA binding activity. Pull down assays with an oligonucleotide library with sequence of N 9 GGN 7 GGN 6 indicated the selection of two motifs in the oligonucleotides that would bind AIRE. The first motif is a weak TATA-box (TTATTA), and the second motif is a tandem repeat of ATTGGTTA (G-box). Gel shift competition assays with oligonucleotides oligo-TGG, oligo-TG, oligo-GG, and oligo-G showed that oligo-TGG exhibited the highest AIRE binding capability that could not be abolished by oligo-GG or oligo-G (Fig. 4, lanes 1-4). The binding of AIRE to oligo-TG appeared to be highly unstable, as all the three competitors used in this study could abolish the gel shift (Fig. 4, lanes 5-8).
The binding of AIRE to oligo-GG (Fig. 4, lanes 9 -12) or oligo-G (Fig. 4, lanes 13-16) appeared to be comparable. Based on the binding affinity of AIRE with these oligonucleotides under the given conditions of competition, we conclude that we could arrange them in the following descending order: oligo-TGG Ͼ oligo-GG Ͼ oligo-G Ͼ oligo-TG. It is noteworthy to state that the complexes between AIRE and oligo-TGG and oligo-TG migrated to a larger molecular mass position compared with the expected positions of AIRE dimer and AIRE tetramer. It is likely that high affinity binding between AIRE and the TGG or TG oligonucleotides produce complexes with more than one tetramer unit of AIRE per oligonucleotide molecule in our assay system.
The construction of a three-dimensional model for the fulllength human AIRE protein with the available templates gives us an insight into the structural architecture of this molecule in space. As could be interpreted from this model, this protein has the following visible domains: SAND (197-262), PHD (298 -341 and 433-476), the N-terminal ␤-helix-loop-helix (HSR), and the leucine zipper (319 -341). In the folded form, both the HSR and the leucine zipper domains are available for molecular interaction. A recent determination of the three-dimensional structure of the SAND domain from Sp100b represented a novel ␣/␤-fold, in which a conserved KDWK sequence motif is found within an ␣-helical, positively charged surface patch (20). In our model for AIRE, the proposed DNA binding SAND and the known DNA binding zinc finger domains are also exposed for DNA-protein interactions. These interpretations match well with our experimental observations.
The importance of the leucine zipper domain and the zinc finger domains in deciding the functional competence of AIRE FIG. 5. Molecular modeling of human AIRE. Templates 11GSA, 1PYDA, 1OCTA, 1PRCH, 2RAMB, and 2YHX were the result of a threading of AIRE sequence against the available Protein Data Bank entries followed by a filtering considering the local and global Z scores. All the three putative DNA binding domains (PHD1, PHD2, and SAND) are surface-positioned in this molecule, making them available for DNA-protein interactions. The leucine zipper (drawn in a red skeleton) is within the PHD1. The SP-100 HSR is free on the surface. From this model, we interpret that AIRE could dimerize using the leucine zipper motif and could tetramerize using the HSR domain. becomes evident if we correlate the known AIRE mutations with the APECED disease. So far more than 20 different mutations of the AIRE gene have been identified in APECED patients (21). The PHD domains in AIRE are located at amino acids 299 -340 (PHD1) and 434 -475 (PHD2). The leucine zipper is located at position 319 -341. The major mutations identified so far include the Finnish mutation R257X and the 13-bp deletion in exon 8 (nucleotides 1094 -1106), producing truncated proteins of 28 and 47 kDa, both lacking either both or the second PHD zinc finger domains (10). Two other potential pathogenic mutations detected are a 2-base pair insertion in exon 11, resulting in a premature termination codon at amino acid 479 and a protein that is 66 amino acids shorter and a T 3 C transition at ϩ2 of intron 3 (22). A recent report revealed novel frameshift mutations of the AIRE gene. These include an insertion of a cytosine at nucleotide 29,635 at exon 10 (29635insC), which should lead to a premature termination at the codon 371, producing a truncated protein missing the second plant homeodomain-type zinc finger motif and the third LXXLL motif, and a deletion of a guanine at nucleotide 33,031 at exon 13 (33031delG), which should result in a premature termination at the codon 520, yielding a truncated protein missing the third LXXLL motif (21). The majority of the mutations listed above does affect the PHD domain(s) and/or the leucine zipper domain, although two of the mutations affect only the last LXXLL motif. In addition to what was thought about the implications of these mutations to date, our data strongly indicate that mutations affecting the leucine zipper domain and/or the zinc finger domain would make AIRE incompetent to bind DNA. This would affect the expression of gene(s) under the regulatory influence of AIRE.
With the identification of AIRE as a DNA-binding protein, two relevant points to be addressed are the factors influencing AIRE expression and the gene(s) under the regulatory influence of AIRE. It appears that the first question has already been addressed, since a recent report states that AIRE was absent in RelB-deficient mouse (23). The target genes of AIRE are yet to be identified.