Characterization of GATA3 Mutations in the Hypoparathyroidism, Deafness, and Renal Dysplasia (HDR) Syndrome*

The hypoparathyroidism, deafness, and renal dysplasia (HDR) syndrome is an autosomal dominant disorder caused by mutations of the dual zinc finger transcription factor, GATA3. The C-terminal zinc finger (ZnF2) binds DNA, whereas the N-terminal finger (ZnF1) stabilizes this DNA binding and interacts with other zinc finger proteins, such as the Friends of GATA (FOG). We have investigated seven HDR probands and their families for GATA3 abnormalities and have identified two nonsense mutations (Glu-228 → Stop and Arg-367 → Stop); two intragenic deletions that result in frameshifts from codons 201 and 355 with premature terminations at codons 205 and 370, respectively; one acceptor splice site mutation that leads to a frameshift from codon 351 and a premature termination at codon 367; and two missense mutations (Cys-318 → Arg and Asn-320 → Lys). The functional effects of these mutations, together with a previously reported GATA3 ZnF1 mutation and seven other engineered ZnF1 mutations, were assessed by electrophoretic mobility shift, dissociation, yeast two-hybrid and glutathione S-transferase pull-down assays. Mutations involving GATA3 ZnF2 or adjacent basic amino acids resulted in a loss of DNA binding, but those of ZnF1 either lead to a loss of interaction with specific FOG2 ZnFs or altered DNA-binding affinity. These findings are consistent with the proposed three-dimensional model of ZnF1, which has separate DNA and protein binding surfaces. Thus, our results, which expand the spectrum of HDR-associated GATA3 mutations and report the first acceptor splice site mutation, help to elucidate the molecular mechanisms that alter the function of this zinc finger transcription factor and its role in causing this developmental anomaly.

The hypoparathyroidism, deafness, and renal dysplasia (HDR) syndrome is an autosomal dominant disorder caused by mutations of the dual zinc finger transcription factor, GATA3. The C-terminal zinc finger (ZnF2) binds DNA, whereas the N-terminal finger (ZnF1) stabilizes this DNA binding and interacts with other zinc finger proteins, such as the Friends of GATA (FOG). We have investigated seven HDR probands and their families for GATA3 abnormalities and have identified two nonsense mutations (Glu-228 3 Stop and Arg-367 3 Stop); two intragenic deletions that result in frameshifts from codons 201 and 355 with premature terminations at codons 205 and 370, respectively; one acceptor splice site mutation that leads to a frameshift from codon 351 and a premature termination at codon 367; and two missense mutations (Cys-318 3 Arg and Asn-320 3 Lys). The functional effects of these mutations, together with a previously reported GATA3 ZnF1 mutation and seven other engineered ZnF1 mutations, were assessed by electrophoretic mobility shift, dissociation, yeast twohybrid and glutathione S-transferase pull-down assays. Mutations involving GATA3 ZnF2 or adjacent basic amino acids resulted in a loss of DNA binding, but those of ZnF1 either lead to a loss of interaction with specific FOG2 ZnFs or altered DNA-binding affinity. These findings are consistent with the proposed three-dimensional model of ZnF1, which has separate DNA and protein binding surfaces. Thus, our results, which expand the spectrum of HDR-associated GATA3 mutations and report the first acceptor splice site mutation, help to elucidate the molecular mechanisms that alter the function of this zinc finger transcription factor and its role in causing this developmental anomaly.
GATA3 belongs to a family of zinc finger transcription factors that are involved in vertebrate embryonic development (1)(2)(3). The six mammalian GATA proteins (GATA-1 to -6) share related Cys-X 2 -Cys-X 17 -Cys-X 2 -Cys (where X represents any amino acid residue) zinc finger DNA-binding domains (see Fig.  1) and bind to the consensus motif 5Ј-(A/T)GATA(A/G)-3Ј (4). The C-terminal finger (ZnF2) 1 is essential for DNA binding, whereas the N-terminal finger (ZnF1) appears to stabilize this binding and to physically interact with other multitype zinc finger proteins, such as the Friends of GATA (FOG) (5-7). Thus, FOG-1 and FOG-2 have been shown, in mammals, to modulate the biological activities of GATA1 and GATA4, respectively (5)(6)(7). Furthermore, the importance of these interactions of GATA and FOG family members are underscored by their evolutionary conservation, because it has been shown that the Drosophila GATA factor, Pannier, interacts with a FOG-like protein referred to as U-shaped (8,9). The mammalian GATA factors can be subdivided into two families based on their structures and patterns of expression (10,11). Thus, the structurally related proteins GATA4, -5, and -6 are expressed in overlapping patterns in the heart, gut, urogenital system, and smooth muscle cell lineages, whereas GATA1, -2, and -3 are expressed in the hematopoietic cell lineages in which they control development of the erythroid, hematopoietic stem cell and T cell lineages, respectively (10,11). In addition, GATA3 is also expressed in the developing parathyroids, inner ear, and kidneys (12,13). These expression patterns are consistent with the disease phenotypes that have been reported in the few patients with genetic abnormalities involving three of the GATA members. Thus, GATA1 mutations lead to dyserythropoietic anemia, thrombocytopenia (14), and the megakaryoblastic leukemia of Down's syndrome (15); GATA3 haploinsuf-ficiency is associated with the hypoparathyroidism, deafness, and renal dysplasia (HDR) syndrome (16); and GATA4 hemizygosity has been observed in some patients with congenital heart disease (17). More than 90% of HDR syndrome patients have hypoparathyroidism and deafness, and more than 80% have renal tract abnormalities (16,18,19). The hypoparathyroidism is characterized by symptomatic or asymptomatic hypocalcemia with undetectable or inappropriately normal serum concentrations of parathyroid hormone (PTH), and normal brisk increases in plasma cAMP in response to PTH infusion, which indicates normal sensitivity of the PTH receptor (18). The sensorineural deafness is usually bilateral, although the hearing loss may vary in its severity (18, 20 -22). The renal tract abnormalities, which may be uni-or bi-lateral, consist of: renal cysts that may cause pelvicalyceal deformities and/or compression of the glomeruli and tubules that may lead to kidney failure; renal aplasia or hypoplasia; and vesicoureteral reflux (16, 18 -22). The precise manner in which GATA3 mutations cause these congenital abnormalities of the parathyroids, inner ear, and kidneys remains to be elucidated. To gain further insights into the structure-function relationships of GATA3, we have studied additional HDR patients for GATA3 abnormalities and have investigated the effects of GATA3 mutations on DNA binding and protein interactions.

EXPERIMENTAL PROCEDURES
Patients-Ten patients with HDR from seven unrelated families were ascertained (Table I) (16/1998) was from Samoa. All 10 patients had hypoparathyroidism with serum calcium ranging from 1.01 to 2.00 mM, and this was associated with tetany or seizures in four patients, but was asymptomatic in six patients (Table I). Bilateral sensorineural deafness was found in all 10 patients with the age at diagnosis ranging from Ͻ1 to Ͻ30 years. Renal abnormalities were found in seven patients, of which two patients had developed end-stage renal failure, two had hypoplastic kidneys, and another two had agenesis of the right kidney.
DNA Sequence Analysis of the GATA3 Gene-Venous blood was ob-tained after informed consent, as approved by the local ethical committee, and used to extract leukocyte DNA (23). Nine pairs of GATA3specific primers were used for the PCR amplification of the six exons and ten intron-exon boundaries ( Fig. 1) utilizing 150 ng of genomic DNA as described (24). The DNA sequences of both strands were determined by Taq polymerase cycle sequencing (24) and resolved on a semi-automated detection system (373 sequencer, Applied Biosystems, Foster City, CA). DNA sequence abnormalities in the probands, which were confirmed either by restriction endonuclease analysis (24), by allelespecific oligonucleotide hybridization (25), or by a modified version of the amplification refraction mutation system (26), were demonstrated to co-segregate with the disorder and to be absent in the DNA obtained from 55 unrelated individuals. Electrophoretic Mobility Shift Assays-COS-1 cells, which do not endogenously express GATA3, were transfected using LipofectAMINE Plus (Invitrogen, Carlsbad, CA) with either a wild type GATA3 construct prepared in pcDNA 3.1 (GATA3-pcDNA3) (Invitrogen) or a construct harboring the mutation that was introduced by the use of site-directed mutagenesis (QuikChange, Stratagene, La Jolla, CA) (16). Forty-eight hours post-transfection, the cells were harvested, and nuclear extracts were prepared for use in binding reactions that utilized a 32 P-labeled double-stranded oligonucleotide containing the GATA3 consensus as described (16). The binding reactions were resolved by non-denaturing 6% PAGE. Western blot analysis using HG3-31 monoclonal antibody against GATA3 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) was used to detect the presence of GATA3 protein in the nuclear extracts (16). For dissociation shift assays (14,27), unlabeled competitor DNA was added to an 100-fold excess to the binding reactions, and aliquots were removed after 0, 10, 30, and 60 min for non-denaturing PAGE.
Nuclear Localization Studies Using GATA3-Green Fluorescent Protein Fusion Constructs-The wild type and mutant GATA3 constructs were subcloned in-frame into the mammalian expression vector pEGFP-C1 (BD Biosciences Clontech, Palo Alto, CA) as previously described (28). COS-1 cells were transfected with the GATA3-GFP constructs, using LipofectAMINE Plus (Invitrogen), and after 24 h the cells were replated at lower density onto 70% ethanol-treated coverslips and cultured for a further 24 h. The cells were then washed with phosphate-buffered saline (PBS), fixed with freshly prepared 4% paraformaldehyde/PBS for 30 min, washed with PBS, and mounted with 4Ј,6-diamidino-2-phenylindole (DAPI)-containing Vectashield (Vector Laboratories, Burlingame, CA), as described (28). The DAPI/ GFP images were visualized using a Nikon Eclipse E400 microscope with a Y-FL Epi-fluorescence attachment and a triband DAPI-fluorescein isothiocyanate-Rhodamine filter (28). GATA3 Minigene Construct for mRNA Splicing Studies-A minigene containing GATA3 exons 4, 5, and 6 was constructed. Each exon was PCR-amplified from genomic DNA using exon-specific primers and conditions that were utilized for DNA sequence analysis (16). The PCR products were cloned directly into pGEM-T (Promega, Madison, WI) and sequenced to determine orientation and absence of Taq-introduced secondary mutations. Each exon was excised from pGEM-T using appropriate restriction endonucleases and directionally subcloned in a four-way ligation reaction into pcDNA 3.1. COS-1 cells were transfected with the plasmid, as described above, and after 48 h the cells were harvested and RNA prepared (24) for use in reverse-transcription PCR (RT-PCR) that utilized avian myeloblastosis virus reverse-transcriptase (Life Sciences Inc., St. Petersburg, FL) and a reverse pcDNA3.1 primer to synthesize the first-strand cDNA. A control reaction without reverse transcriptase was also performed. The PCR reaction contained 1.5 mM MgCl 2 , 250 M dNTP (Invitrogen), 0.3 M of each primer (Forward exon 5 primer, 5Ј-TCTGCAATGCCTGTGGGCTC-TAC-3Ј, and reverse exon 6 primer, 5Ј-CTAACCCATGGCGGTGAC-CATGC-3Ј), and 1 unit of Taq DNA polymerase (Invitrogen) in 50 l of standard PCR buffer (16). Amplification conditions were, denaturation at 95°C for 5 min, followed by 30 cycles of 94°C for 15 s, 65°C for 15 s, and 72°C for 1 min, followed by final extension at 72°C for 5 min and rapid cooling to 20°C.
Yeast Two-hybrid Assays-In vivo interactions between the GATA3 N-terminal zinc finger (ZnF1) (Fig. 1) and FOG2 ZnF1, -5, -6, and -8 were studied using a yeast two-hybrid system (BD Biosciences Clontech) (29). GATA3 ZnF1 (amino acids 261-293) was generated by cloning a PCR product, amplified from the wild type GATA3 expression construct (16), in-frame, into the Gal4 DNA-binding domain (BD)-encoding plasmid, pGBKT7 (30). Mutations were introduced into this construct by site-directed mutagenesis (QuikChange, Stratagene). FOG2 ZnFs were generated by RT-PCR using human embryonic kidney (HEK) 293 cell RNA as template. Each FOG2 ZnF (ZnF1 amino acids 236 -290; ZnF5 amino acids 531-617; ZnF6 amino acids 661-747; and ZnF8 amino acids 1100 -1151) was cloned in-frame into the Gal4 activation domain (AD)-encoding plasmid, pGADT7. The p53-pGBKT7 and Large T antigen-pGADT7 plasmids (BD Biosciences Clontech) were used as controls (31,32). Competent AH109 yeast cells were transformed sequentially with the appropriate GATA3 and FOG2 ZnF plasmid constructs using the LiAc/single-stranded DNA/polyethylene glycol procedure (33). The transformants were selected on Leu Ϫ Trp Ϫ (double drop-out, DDO) minimal media plates by growth at 30°C for 3 days. Transformants were then patched onto His Ϫ Ade Ϫ Leu Ϫ Trp Ϫ (quaternary drop-out, QDO) media plates and monitored for growth for up to 3 days. Expression of GATA3 and FOG2 Gal4 fusion proteins was confirmed by preparing protein extracts from each clone according to the manufacturer's instructions (BD Biosciences Clontech) and analyzing them by SDS-PAGE in Tris-glycine-SDS buffer (Bio-Rad, Hercules, CA) and electro-transference onto PolyScreen polyvinylidene difluoride transfer membrane (PerkinElmer Life Sciences, Boston, MA) in CAPS buffer (10 mM, pH 11; Sigma Chemical Co., St. Louis, MO). Western blot analysis was performed with antibodies to either the Gal4-AD (FOG2-pGADT7 constructs) or the Gal4 DNA-BD (GATA3-pGBKT7 constructs), according to the manufacturer's instructions (BD Biosciences Clontech) except that Gal4 DNA-BD antibody was used at 50 ng/ml and Gal4-AD antibody at 100 ng/ml (29). A secondary antibody, goat anti-  Table II) together with the six previously reported mutations (labeled a-f: a, R277X; b, R367X; c, deletion frameshift (del, fs) from codon 156; d, insertion frameshift (ins, fs) from codon 301; e, in-frame deletion (del, inf) 316 -319; f, W275R (16,22). In addition, six whole gene deletions (del) have been previously reported (16,22), yielding a total of 19 GATA3 abnormalities identified in HDR patients. Nine of the 10 HDR mutations, which affect the region encompassing the two zinc fingers and the adjacent C-terminal region, are further detailed above in the amino acid sequence, in which every tenth amino acid is numbered. The amino acids altered by the nine HDR mutations are highlighted in black, and the seven mutations (E263V, C264R, GA268/269QT, P273T, R276Q, D278G, and D278Y) of ZnF1 generated for additional functional studies (Figs. 4 and 5) are highlighted in gray.
GST Fusion Proteins and Pull-down Assays-The glutathione Stransferase (GST) fusion proteins contained FOG2 ZnF1, ZnF5, ZnF6, and ZnF8 fused downstream of the GST protein in the vector pGEX-4T-1 (Amersham Biosciences). The expression of GST fusion proteins was carried out in Escherichia coli BL21 (34). The 35 S-labeled wild-type or mutant GATA3 proteins were prepared by in vitro transcription/ translation (TNT system, Promega, Madison WI) using GATA3-pcDNA3 or constructs harboring selected mutations, and aliquots were utilized to monitor [ 35 S]methionine incorporation by SDS-PAGE (16). In vitro binding assays using 1 g of the fusion protein attached to glutathione-Sepharose 4B (Amersham Biosciences), and 1 l of the radiolabeled GATA3 protein were performed in 300 l of binding buffer (150 mM NaCl, 20 mM Tris-HCl, pH 7.5, 0.1% Igepal CA-630, 20 M ZnSO 4 , 0.25% bovine serum albumin, 1 mM ␤-mercaptoethanol, 1.5 mM phenylmethylsulfonyl fluoride) and incubated with mixing for 1 h at 4°C (35). The glutathione-Sepharose 4B-FOG2 fusion protein-GATA3 complexes were recovered by centrifugation (20,000 ϫ g, 2 min) and washed four times with 450 l of cold binding buffer. The proteins were released by boiling in 15 l of Laemmli sample buffer (Bio-Rad) and analyzed by SDS-PAGE (12% polyacrylamide resolving gel in Tris/glycine/SDS running buffer (Bio-Rad). The gel was fixed and then soaked in Amplify (Amersham Biosciences) prior to autoradiography (36).
Computer Modeling of GATA3 ZnF1 Structure-The three-dimensional structure of the murine GATA1 N-terminal zinc finger has been reported (37), and because the N-terminal zinc fingers of GATA1 and GATA3 are over 90% identical, we modeled the position of the GATA3 mutants on this framework. The GATA1 ZnF1 three-dimensional structure is archived in the Protein Data Bank (PDB) at the European Bioinformatics Institute (EBI) with the accession number 1GNF (available at oca.ebi.ac.uk/oca-bin/ccpeek?id ϭ 1GNF) and was visualized using the MDL Chime program (MDL Information Systems, Inc., San Leandro, CA)

RESULTS
Mutations in HDR Families-DNA sequence analysis of the entire 1332-bp coding region together with the associated splice sites and 5Ј and 3Ј untranslated regions of the GATA3 gene from each of the seven probands with HDR revealed the pres-ence of seven heterozygous mutations ( Fig. 1 and Table II), six of which were novel and one of which had previously been reported in an unrelated Japanese family (22). Thus, two of the mutations were nonsense mutations (Fig. 2), two were frameshifting deletions, two were missense mutations, and one was an acceptor splice site mutation (Fig. 3). The occurrence of the nonsense, frameshifting deletions and missense mutations in the probands was confirmed either by restriction enzyme analysis (Fig. 2), or by allele-specific oligonucleotide hybridization analysis, or by amplification refraction mutation system (Table  II). The acceptor splice site mutation, which involved a g to t transversion of the invariant ag ( Fig. 3) was confirmed by repeat DNA sequence analysis on independently obtained PCR products. This predicted a loss of this acceptor splice site and the possible use of another naturally occurring, but normally unused acceptor splice site at codons 351-353 (Fig. 3). These predicted effects on mRNA splicing were assessed by expressing wild-type and mutant GATA3 minigene constructs that encompassed exons 4 -6 in COS-1 cells. This revealed utilization of the alternative acceptor splice site that would lead to a loss of 8 nucleotides from the mRNA. This resulted in a frameshift that, if translated, would produce a missense peptide with a premature termination at codon 367. Co-segregation of the GATA3 mutations and HDR was demonstrated in the available members from families 8/2000 (Fig. 2), 9/2000, and 2/2001, whereas in the probands from families 13/2001 and 16/2001, the mutations were demonstrated to be absent in the parents and hence were arising de novo (Table II). In addition, the absence of these DNA sequence abnormalities in 110 alleles from 55 unrelated normal individuals indicated that these abnormalities were mutations and not functionally neutral polymorphisms that would be expected to occur in Ͼ1% of the population. All of the seven mutations, which occurred in exons 3-6 ( Fig. 1), predict structurally significant changes (Table II). Thus, the E228X (Glu-228 3 Stop) and frameshift deletion  Fig. 1. b Family identification refers to clinical details shown in Table I. c Analysis by restriction enzymes (RE), or allele specific oligonucleotide (ASO) hybridisation, or amplification refractory mutation system (ARMS)-PCR analysis, or sequence analysis (SA). occurring in codon 201 are predicted, if translated, to lead to truncated GATA3 proteins that lack both ZnFs; the R367X (Arg-367 3 Stop), the frameshift deletions occurring in codon 355, and the acceptor splice site mutation at the intron 5/exon 6 boundary are predicted to lead to truncated GATA3 proteins that lack the C-terminal region adjacent to ZnF2, and the missense mutations C318R (Cys-318 3 Arg) and N320K (Asn-320 3 Lys) are predicted to disrupt ZnF2 of GATA3 (Fig. 1). The effects of these mutations together with the W275R (Trp-275 3 Arg) that was reported in a Japanese HDR patient (22) were further assessed in DNA binding studies. The effects of the acceptor splice site mutation found in family 16/2001 (Fig.  3) were not assessed separately, because the predicted protein is almost identical to that resulting from the frameshift deletion found in family 16/1998 ( Fig. 1 and Table II). DNA Binding and Subcellular Localization Studies-All of the HDR-associated GATA3 mutations, with the exception of one, W275R, are predicted to disrupt ZnF2 or its adjacent C-terminal region (Fig. 1), and the results of Western blot analysis are consistent with this (Fig. 4). ZnF2, which is the C-terminal zinc finger, is essential for DNA binding, and thus all of these HDRassociated GATA3 mutations would predict a disruption of DNA binding (Table II). However, the W275R mutation lies within ZnF1, and its effects are more difficult to predict, although some naturally occurring and some engineered GATA1 mutants of the N-terminal zinc finger, ZnF1, have been shown to destabilize DNA binding or protein-protein interactions (14, 35, 38 -40). We therefore engineered the equivalent seven GATA3 mutants, E263V (Glu-263 3 Val), C264R (Cys-264 3 Arg), GA268/269QT (GlyAla-268/269 3 GlnThr), P273T (Pro-273 3 Thr), R276Q (Arg-276 3 Gln), D278G (Asp-278 3 Gly), and D278Y (Asp-278 3 Tyr), so as to facilitate a more comprehensive study of the 25 residues forming the GATA3 ZnF1 (Fig. 1). These residues were selected for engineering mutants, because they either are the non-conserved residues of ZnF1 when compared with their respective ZnF2 counterpart, or they are the equivalent counterparts to GATA1 disease-causing mutations (38 -40). We assessed these GATA3 mutants (i.e. the ones associated with HDR ( Fig. 1) and the seven engineered GATA3 ZnF1 mutations) initially for altered DNA binding by EMSAs (Fig. 4), using nuclear extracts from COS-1 cells transfected with either the wild-type or mutant GATA3 constructs. In addition, an assessment of the subcellular localization of the GATA3 mutants, using GATA3-GFP constructs, was also undertaken, and this revealed that 12 of the mutants, which retained ZnF1 (Fig.1), accumulated in the nucleus and were indistinguishable from the WT-GATA3 (Fig. 4). However, the two mutants (deletion of C in codon 201 and E228X) that lacked ZnF1 did not accumulate in the nucleus.

FIG. 2. Detection of GATA3 mutation in exon 3 in family 8/2000 with HDR by restriction enzyme analysis.
DNA sequence analysis of individual III, 1 revealed a G to T transversion at codon 228, thus altering the wild-type (WT) sequence GAG, encoding a glutamine to the mutant (m) sequence TAG, which is a termination (Stop) codon. This nonsense mutation also resulted in the loss of the wild-type BsoBI restriction enzyme (C/ CCGAG), and this facilitated the confirmation of the mutation (b). PCR amplification and BsoBI digestion would result in two products of 167 and 100 bp from the normal, i.e. wild type (WT) sequence, but an additional band of 267 bp would be expected from the mutant (m) sequence as is illustrated in the restriction map in c. Co-segregation of this E228X mutation and its heterozygosity in the affected members (II, 2 and III, 1) was demonstrated (b), and the absence of this E228X mutation in 110 alleles from 55 unrelated normal individuals (N 1 and N 2 shown) indicates that it is not a common DNA sequence polymorphism. Similar restriction enzyme analysis was used to confirm and demonstrate co-segregation of the codon 201 deletion, and the C318R and the R367X mutations (Table II) These findings are consistent with the nuclear localization signal for GATA3 being contained within residues 249 -311 that encompass ZnF1 (41). The EMSA studies revealed that the GATA3 mutants, which disrupted or lead to a loss of ZnF2 ( Fig. 1 and Table II), all resulted in a loss of DNA binding. Furthermore, addition of a 2-fold excess of these mutant GATA3 nuclear extracts to the wild type, did not significantly alter binding by WT-GATA3 (data not shown), thereby suggesting an absence of a dominant-negative effect due to heteroduplex formation. This is consistent with the development of an HDR phenotype in patients who have haploinsufficiency due to a deletion of the GATA3 gene (16). The GATA3 mutants involving ZnF1, all retained DNA binding (Fig. 4). However, these ZnF1 mutants differed in the stability of their binding to DNA, which resulted in altered rates of dissociation. Thus, the HDR-associated mutant W275R and the engineered mutants GA268/269QT, D278G, and D278Y had dissociation rates similar to that of the wild-type GATA3 (Fig. 4), whereas the engineered mutants E263V, C264R, P273T, and R276Q had a more rapid rate of dissociation (Fig. 4).
These results indicate that the ZnF1 GATA3 residues Glu-263, Cys-264, Pro-273, and Arg-276 are critical for stabilizing the DNA binding by ZnF2 and that this is likely to involve interactions with other multitype zinc finger proteins, in a manner similar to that reported for GATA1 ZnF1 (14, 35, 38 -40). For example, the engineered GATA1 mutant C204R, which is equivalent to the GATA3 C264R, has been reported to destabilize DNA binding (14) and to abolish the interaction with FOG ZnF6 (35). However, the HDR GATA3 mutant W275R and the engineered mutants GA268/269QT, D278G, and D278Y did not alter the stability of the DNA binding (Fig. 4), and, to further elucidate the role of these residues and their mutations, we utilized a yeast two-hybrid assay.
Yeast Two-hybrid Assay-GATA1 ZnF1 and GATA4 ZnF1 interact with the zinc finger proteins FOG1 and FOG2, respectively (5-7). We investigated FOG2 for interactions with GATA3, because of their similar temporo-spatial expression patterns (6,12). Thus, in mouse embryos older than 11.5 days, both GATA3 and FOG2 are expressed in the same tissues that include the otic vesicle and the developing kidney (6,12). In addition FOG2 has been shown to interact with GATA3 in mouse embryos (7). These interactions between GATA factors and the FOG proteins involve the GATA ZnF1 and several of the zinc fingers of the FOG protein. For example, the GATA1 ZnF1 interacts with four of the nine zinc fingers (ZnF1, -5, -6, and -9) of FOG, and four of the eight zinc fingers (ZnF1, -5, -6, and -8) of FOG2 (42). We selected to investigate the four involved zinc fingers (ZnF1, -5, -6, and -8) of FOG2 for interactions with wild-type and mutant GATA3 ZnF1 in a yeast twohybrid assay. One GATA3 construct and one FOG2 construct were sequentially transformed into the yeast reporter strain AH109, and yeast containing both plasmids were selected on minimal DDO medium that lacked leucine and tryptophan (Fig. 5a). Co-expression of the GATA3 and FOG2 Gal4 fusion proteins was confirmed by Western blotting of yeast protein extracts, prepared from each clone, and detected using antibodies against either the Gal4 DNA-BD or the Gal4-AD (data not  (Table II). DNA sequence analysis of the affected proband (Table I) revealed a g to t transversion at the Ϫ1 position, which resulted in an alteration of the invariant ag acceptor splice site (a). Analysis of 110 alleles from 55 unrelated normals revealed the presence of the normal ag acceptor splice site and an absence of the at sequence, thereby indicating that the g to t transversion at position Ϫ1 was not a common sequence polymorphism but a likely mutation that would alter mRNA splicing (data not shown). In addition, an examination of the DNA sequences of codons 351-353 revealed another naturally occurring, but normally unused, acceptor splice site sequence (ncag) (61,62). The effects of the likely mutation were therefore investigated by using wild-type (WT) and mutant (m) minigene constructs containing exons 4, 5, and 6 in the mammalian expression vector pcDNA 3.1, and transfecting these into COS-1 cells. Total RNA was extracted from the cells and utilized with exon 5-and exon 6-specific primers in RT-PCRs. The mutant RT-PCR products are smaller and DNA sequence analysis of these revealed splicing of exon 5 to an internal site in exon 6 that resulted in a new sequence, which encoded a missense peptide with a premature termination at codon 367. Thus, the mutation had resulted in utilization of an alternative, naturally occurring, but normally non-utilized, acceptor splice sequence. Exon sequence (uppercase), intron sequence (lowercase); ϩ, with reverse transcriptase; Ϫ, without reverse transcriptase; size markers are in base pairs. FIG. 4. Analysis of DNA-binding properties and subcellular localization of GATA3 mutants associated with HDR. a, Western blot analysis of in vitro translated wild-type (WT) and GATA3 mutants revealed the expected 50-kDa WT product. The missense mutations C318R and N320K also yield a 50-kDa product, whereas the nonsense (E228X and R367X) and frameshift deletions (201⌬C and 355⌬CT) yield the predicted shown). These yeast colonies were then patched onto minimal QDO medium that lacked leucine, tryptophan, histidine, and adenine to select for those yeast in which a protein-protein interaction had occurred (Fig. 5b). Interaction between the GATA3 and FOG2 zinc fingers would bring the Gal4 DNA-BD into close juxtaposition with the AD at the reporter gene promoter, thereby enabling transcription of the reporter gene. Disruption of this interaction by the GATA3 mutant would lead to a loss of expression of the reporter genes. The results revealed interactions between the wild-type GATA3 ZnF1 and each of the four FOG2 zinc fingers (ZnF1, -5, -6 and -8) (Fig. 5). However, the GATA3 mutants C264R, E263V, and GA268/ 269QT did not interact with any of the FOG2 zinc fingers as evidenced by an absence of yeast growth. The GATA3 mutant W275R similarly abolished interaction with FOG2 zinc fingers 1, 5, and 8, but retained interaction with ZnF6, whereas the D278G and D278Y mutants retained interaction with all FOG2 zinc fingers with the exception of ZnF8. However, the P273T and R276Q mutants retained interaction with all four FOG2 zinc fingers, thereby suggesting that they exert their effect solely by loss of DNA-binding stabilization. These results of the yeast two-hybrid assay were confirmed by GST pull-down assays.
GST Pull-down Assays-GST pull-down assays were performed using full-length GATA3 expressed in a rabbit reticulocyte system and FOG2 ZnF-GST fusion proteins. The wildtype GATA3, and the P273T and R276Q mutants, were retained by FOG2 ZnF1, -5, -6, and -8, whereas the W275R mutant was retained only with FOG2 ZnF6 (Fig. 5c, data shown for wild-type and W275R). In contrast, the E263V, C264R, and GA268/269QT mutants were not retained by any of the four FOG2 ZnFs (Fig. 5c, data shown for C264R), whereas the D278G and D278Y mutants were retained by FOG2 ZnF1, -5, and -6 but not FOG2 ZnF8. These GST pull-down results, which confirm the results of the yeast two-hybrid assay, are in agreement with those previously reported for interactions between GATA1 and FOG2 ZnFs (42). DISCUSSION Our results, which have identified seven mutations of the GATA3 gene in seven HDR probands and their families (Table  II), expand the spectrum of mutations, report the first acceptor splice site mutation (Fig. 3), and further establish the role of GATA3 haploinsufficiency in the etiology of this developmental disorder. In addition, our studies of these GATA3 mutations help to increase our understanding of the underlying DNA binding and protein interactions that are involved for the function of this zinc finger transcription factor. Thus, all the mutations that disrupt either ZnF2 or the basic amino acids located C-terminal to it, lead to a loss of DNA binding (Fig. 4), whereas those that disrupt ZnF1 do not lead to a loss of DNA binding but instead alter interactions with FOG2 (Fig. 5) and/or change DNA binding affinity (Fig. 4). For example, the two missense mutations, C318R and N320K ( Fig. 1 and Table  II), which result in alterations of evolutionarily conserved residues in ZnF2 of the GATA family members, are predicted to disrupt the tertiary structure either directly or via a loss of co-ordination of the zinc ion. This in turn results in a loss of DNA binding and hence a likely alteration in the transcription of target genes. Similarly, the three mutations (two frameshifts starting at codons 351 and 355 and the nonsense mutation R367X) involving the residues on the C-terminal side of ZnF2 ( Fig. 1 and Table II) also result in a loss of DNA binding (Fig.  4). These three mutations involve codons 364 -369, whose equivalents in GATA1 have been shown to be essential for DNA binding, either by direct contact with DNA or by stabilization of nearby residues that contact DNA (43).
In contrast to these GATA3 ZnF2 mutants, the 8 ZnF1 mutants (the HDR-associated W275R and the seven engineered mutants) all retained DNA-binding activity (Fig. 4). These findings for human GATA3 ZnF1 are consistent with those reported for the chicken GATA3 ZnF1 (44), which has been shown to bind GATA or GATC motifs even in the absence of ZnF2. Such studies (27,44,45) have indicated that GATA ZnF1 may serve to stabilize the binding of ZnF2 to gene promoters or enhancers that contain double or palindromic GATA sites, and thereby help in distinguishing between genes that are regulated by different GATA members. However, the GATA3 ZnF1 mutants in our study did show differences in both their DNA binding affinities (Fig. 4) and in interactions with the four of the eight FOG2 ZnFs that were studied (Fig. 5). Thus, the mutants E263V and C264R had low DNA binding affinities and a lack of interactions with FOG2 ZnF1, -5, -6, and -8; the P273T and R276Q mutants had low DNA binding affinities but retained interactions with the four FOG2 ZnFs; the W275R, D278G, and D278Y had a normal DNA binding affinity and interacted with some FOG2 ZnFs, e.g. W275R interacted with FOG2 ZnF6, and D278G and D278Y interacted with FOG2 ZnF1, -5, and -6; whereas GA268/269QT had a normal DNA binding affinity but a lack of interactions with any of the four FOG2 ZnFs. The altered DNA binding affinities observed with E263V and C264R, but not GA268/269QT, may be attributed to the disruption of the ZnF1 structure and a lack of zinc ion coordination that is likely to result with the E263V and C264R mutants, but not the GA268/269QT mutant that involves substitutions for residues that are present in equivalent positions in ZnF2 (Fig. 1). However, any further explanation for these results is difficult to provide on the basis of the primary structure of GATA3 ZnF1, but an analysis of the predicted threedimensional structure of ZnF1 (Fig. 6) may be useful, because it indicates that there may be specific DNA and protein binding surfaces. Thus, Glu-263, Cys-264, Gly-268, and Ala-269 are clustered to form a surface that is important for protein bindtruncated products (Table II). b, electrophoretic mobility shift assays, EMSAs. COS-1 cells were transfected with either the WT or mutant GATA3 constructs, and nuclear extracts were prepared for binding reactions, which used a radiolabeled ( 32 P) double-stranded oligonucleotide containing the GATA consensus DNA sequence (16). Control binding reactions using untransfected (UT) cells and the oligonucleotide alone (OA), i.e. without nuclear extract, were performed. The WT GATA3 bound to double-stranded (ds) DNA, and the method was sensitive enough to detect 10% of the WT GATA3 binding reaction. GATA3 mutants, which disrupted or lead to a loss of ZnF2 (Fig. 1), all resulted in a loss of DNA binding. However, EMSAs revealed normal DNA binding by all the GATA3 ZnF1 missense mutants, whether they were associated with HDR (W275R) or had been engineered (E263V, C264R, GA268/269QT, P273T, R276Q, D278G, and D278Y), (panel c, 0 min). The stability of the DNA binding of all these eight GATA3 mutants that occur in ZnF1 were further studied using dissociation gel shift assays (c) in which unlabeled dsDNA was added, and the effects on the binding of GATA3 to the radiolabeled dsDNA measured over a time course of 60 min by autoradiography. The wild-type (WT) GATA3 and mutants GA268/269QT, W275R, D278G, and D278Y dissociated from the radiolabeled DNA at similar rates, whereas the E263V, C264R, P273T, and R276Q mutations dissociated more rapidly such that the 100-fold excess of unlabeled DNA had replaced all, or a substantial amount, of the radiolabeled DNA by 30 min. Subcellular localization studies (d) revealed that WT-GATA3 and the 12 mutants (E263V, C264R, GA268/269QT, P273T, W275R, R276Q, D278G, D278Y, C318R, N320K, 355⌬CT, and R367X) that contained ZnF1 (Fig. 1) accumulated in the nucleus, whereas the two mutants (⌬201C and E228X) that lacked ZnF1 did not accumulate in the nucleus but instead had a pattern similar to that observed in the cells transfected with GFP alone (GFP). Green and blue labeling represents GFP and nuclear DAPI staining, respectively. Nuclear GFP staining masks DAPI staining, and hence the presence of blue nuclei represents untransfected cells. The scale bar represents 10 m.

FIG. 5.
Interactions between GATA3 ZnF1 and FOG2 ZnFs using a yeast two-hybrid assay. The interaction between wild-type (WT) or mutant GATA3 N-terminal ZnF1, and FOG2 ZnF1, -5, -6, and -8 was studied in the yeast reporter strain AH109 following transformation with the vectors containing GATA3 ZnF1 (pGBKT7) and each FOG2 ZnF (pGADT7) in turn. Yeast growth was monitored 48 h after streaking and incubation at 30°C using either double dropout, DDO (Leu Ϫ Trp Ϫ ), media (a) as a control, or quaternary drop out, QDO (Leu Ϫ Trp Ϫ Ade Ϫ His Ϫ ) media (b) in which growth is dependent on the physical interaction between the GATA3-Gal4 DNA-BD and FOG2-Gal4-AD fusion proteins (29,63). The SV40 large T antigen and p53 proteins, which are known to interact (32), were used as positive controls. Co-expression of the GATA3 and FOG2 Gal4 fusion proteins in the yeast colonies was confirmed in each case by Western blot analysis. The WT GATA3 fusion protein interacted with FOG2 ZnF1, -5, -6, and -8 fusion proteins, whereas the engineered mutant E263V, C264R, and GA268/269QT GATA3 proteins showed an absence of interaction with these FOG2 ZnFs. However, the W275R mutant, which was reported in an HDR patient (22), significantly interacted with FOG2 ing, e.g. with FOG2 ZnFs; whereas Trp-275 and Asp-278 reside on another surface that may be important for interactions with ZnF1, -5, and -8, and ZnF8, respectively, and whereas Pro-273 and Arg-276 reside on a different surface that is involved in binding DNA but not FOG2 ZnFs.
The role of the HDR-associated W275R mutation is of further interest in this model. The W275R mutation is located among residues (Pro-273 and Arg-276) that form a DNA binding surface (Fig. 6), and yet it leads to a loss of protein interactions with FOG2 ZnF1, -5, and -8 and not an alteration in DNA binding affinity. This suggests a dual role for the WRR peptide (codons 275-277), which is conserved in both ZnF1 and ZnF2 (Fig. 1), in binding to DNA as well as FOG2. These results are consistent with those reported from studies of GATA4 ZnF2, in which the equivalent conserved residues were mutated and shown to be critical for DNA binding and for interactions between GATA4 ZnF2 and the protein p300/CBP (46,47). Furthermore, a GATA1 ZnF1 mutant, which involved the equivalent GATA3 residue Arg-276, failed to bind GATA motifs but interacted normally with FOG (40). All these observations indicate that the WRR peptide is involved in separate FOG-GATA-interacting and DNA-binding functions, and a threedimensional model of ZnF1 (Fig. 6) is consistent with this if the aromatic side chain of the Trp-275 residue projects away from the DNA binding surface formed by Pro-273, Arg-276, and Arg-277, and is thereby available to interact with FOG2 ZnFs. We have concentrated on studying the effects of GATA3 mutants on the interactions with FOG2 because of their similar temporo-spatial expression patterns (6,7). However, GATA3 also interacts with other transcription factors that include GATA1, GATA2 (48,49), Sma-and Mad-related protein 3 (smad3) (50), specificity protein 1 (51), erythroid Krü ppel-like factor (51), and rhombotin 2 (52). GATA2 (53), smad3 (54), and rhombotin 2 (55) are expressed in kidney, whereas specificity protein 1 is expressed in both kidney (56) and parathyroids (57), and thus, it may be possible for HDR-associated GATA3 mutations to disrupt interactions with these proteins, provided that they were expressed contemporaneously.
An examination of the HDR-associated GATA3 mutations together with the observed phenotypes does not establish a correlation (Tables I and II), and this is well illustrated by the two unrelated families from Britain and Japan (22) who had an identical R367X mutation but different phenotypes. Thus, the British patient 13/2001 (Tables I and Table I) had hypoparathyroidism and deafness but no renal abnormalities, whereas both Japanese patients had hypoparathyroidism and renal abnormalities but no deafness (22). Furthermore, even within families with patients harboring identical GATA3 mutations, there appears to be a variable expression of renal abnormalities as illustrated by family 8/2000 (Fig. 2 and Table I). The basis of ZnF6 but not with FOG2 ZnF1, -5, and -8. The engineered mutants P273T and R276Q interacted with FOG2 ZnF1, -5, -6, and -8, whereas D278G and D278Y interacted with FOG2 ZnF1, -5, and -6. These results were confirmed by GST pull-down assays (c) that utilized in vitro translated 35 S-labeled GATA3 and GST-FOG2 ZnF fusion proteins (data shown for wild-type, C264R, and W275R). The input row demonstrates that equal amounts of the wild-type and mutant GATA3 protein were loaded, and the Coomassie-stained gel (d) shows that approximately equal amounts of GST-FOG2 fusion proteins were used in the GST pull-down assay. These results of GATA3-ZnF1 interactions with FOG2, are consistent with the findings of the proposed GATA1-ZnF1 three-dimensional model (Fig. 6).
FIG. 6. Three-dimensional structure of the human GATA3-ZnF1 based on the model of murine GATA1 ZnF1 (37). Human GATA1, which consists of 413 amino acids, and human GATA3, which consists of 444 amino acids, belong to the same subfamily (10) and share structural similarities that include two ZnFs (Fig. 1). The three-dimensional structure of the murine GATA1-ZnF1(residues 201-243) has been characterized, and this has 91% identity to the human GATA3-ZnF1 (a), thereby enabling us to use this to construct a three-dimensional model of human GATA3-ZnF1 (residues 261-303). The residues shown in the ribbon (b) and space-filing (c) models refer to those of the equivalent human (h) GATA3 ZnF1, and the corresponding murine (m) GATA1 ZnF1 residues are as follows: hE263 ϭ mE203, hC264 ϭ mC204, hG268 ϭ mG208, hA269 ϭ mA209, hW275 ϭ mW215, hR276 ϭ mR216, and hD278 ϭ mD218. Residues participating in the interaction between mGATA1 and FOG ZnFs, which include the human equivalents of Glu-263, Cys-264, Gly-268, and Ala-269, are shown. They are seen to form a binding surface distinct from that containing Trp-275 and Asp-278, which have been shown to interact with different FOG zinc fingers, whereas Arg-276 lies at the DNA binding surface and does not participate in binding to FOG2 ZnFs. Pro-273, which also resides at a DNA binding surface, is not visible in the projection shown. The backbone is shown as dark magenta; hydrophobic side chains as gray; polar side chains as magenta; acidic side chains as red; and basic side chains as blue. This color scheme derives from the Corey, Pauling, Koltun (CPK) color scheme as follows. hydrophobic ϭ carbon; acidic ϭ oxygen; basic ϭ nitrogen; polar but uncharged ϭ a mixture of oxygen (red) and nitrogen (blue), namely magenta. The backbone is polar but less likely (dark magenta) than side chains to hydrogen bond to non-backbone moieties, because most backbone hydrogen bonding occurs within the backbone. these phenotypic differences in patients with the same mutation remains to be elucidated. One possibility is that there may be different levels of compensation by other GATA family members in different patients. This hypothesis seems attractive, particularly because GATA2 and GATA3 have been shown to be able to partially compensate for the loss of GATA1 in the differentiation of hematopoietic lineages, when placed under the control of the GATA1 locus in transgenic mice (58,59). However, it is important to note that studies of mice lacking GATA1 have demonstrated that GATA2 does not compensate for the loss of GATA1 function in vivo (60), thereby indicating that extrapolation of compensatory mechanisms to the native situation requires cautious interpretation. Additional studies investigating for genotype-phenotype correlations in HDR patients and for alterations in the expression of GATA family members that may compensate for reduced GATA3 expression are required. In summary, our studies have shown that HDRassociated GATA3 mutations may either disrupt DNA binding or protein interactions with FOG2 and that these are consistent with the roles of the zinc finger domains and with the proposed three-dimensional model. However, the manner in which these GATA3 mutations lead to parathyroid, otic vesicle, and renal anomalies remains to be elucidated.