Interaction, Cooperative Promoter Modulation, and Renal Colocalization of GCMa and Pitx2*

The transcription factor GCMa is a member of a new small family of transcription factors with a conserved zinc-containing DNA-binding domain. All members of this transcription factor family play crucial roles as master regulators during development. GCMa is restricted to placenta during development and to kidney and thymus at postnatal stages. It is essential for the formation of the placental labyrinth and as a consequence for survival of the embryo from mid-embryogenesis onwards. Here, we identify Pitx transcription factors as GCMa-interacting proteins. We show that Pitx proteins interact via their conserved homeodomain with the DNA-binding domain of GCMa. As a consequence, Pitx proteins and GCMa exhibit cooperative DNA binding. Furthermore, Pitx proteins influence GCMa-dependent promoter activation in a cell-specific manner. One of the three Pitx paralogues in mice, Pitx2, is the predominant Pitx member present in the placenta and colocalizes on the cellular level with GCMa in the kidney. This is the first description of a regulatory cross-talk between a transcription factor of the GCM family and a homeodomain protein.

GCM 1 (glial cells missing) proteins are a small group of transcription factors containing a highly conserved DNA-binding domain with novel mode of conformation and DNA interaction (1)(2)(3)(4)(5)(6). Contrary to their structural conservation, GCM proteins are involved in very different biological events as key regulators (7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19). Whereas the prototype member GCM/glide was identified in the nervous system of Drosophila and shown to play a role in gliogenesis, vertebrate homologs turned out to be located and to function in tissues such as placenta, kidney, thymus, and parathyroid gland (7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19). Knockout mice proved the involvement of GCMa in the formation of the placental labyrinth (11,14). In the absence of placental GCMa expression, the allantois still fuses with the chorion, but subsequent steps of placental labyrinth formation are inhibited. The resulting failure to supply the embryo with nutrients and gases leads to lethality around E9.5 (11,14). Postnatally, GCMa is expressed in kidney and, at low levels, in thymus (18). Renal expression of GCMa is turned on in neonatal mice and is restricted to the S3 segment of proximal tubules (18). GCMb knockout mice lack the parathyroid gland and suffer from partial neonatal lethality caused by defects in calcium and phosphate homeostasis (12). Although the octameric binding motif (ATGCGGGT) of GCM proteins has been identified and its main biochemical properties have been thoroughly studied, up to now only two potential target gene promoters have been described for GCMa, namely aromatase and syncytin (1,20,21). No target gene is known for GCMb.
Transcriptional regulators of the Pitx group also exert crucial roles during mammalian development (22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32). Their common structural characteristic is a bicoid-related homeodomain. There are three Pitx paralogues in mice, which are highly conserved within their amino-terminally located homeodomain (Ͼ97% identity) and still well conserved over the whole remaining carboxyl-terminal length (55-70% identity). A short region near the carboxyl-terminal end has been termed OAR (Otx, Aristaless, Rex) and might modulate transactivation capacities or be involved in protein-protein interactions (28,33). Pitx1 is expressed in many tissues, including the stomodeum and its derivatives such as the pituitary, where a large set of Pitx1 target genes has been identified. Pitx3 is expressed in the midbrain dopaminergic system and in the eyes (34,35). Pitx2 was originally identified by positional cloning in patients with Axenfeld-Rieger syndrome, a rare autosomal dominant hereditary condition characterized by ocular anomalies as well as dental hypoplasia and defects in umbilical development (27,28,30,36,37). The Pitx2 gene encodes three isoform in mice (Pitx2a-Pitx2c), and a fourth isoform (Pitx2d) has been identified in humans (27, 38 -42). The different isoforms arise from alternative splicing and use of different promoters. All isoforms can combine as homodimers, and heterodimers are formed with Pitx2b (43).
We searched for proteins interacting with GCMa and identified Pitx proteins. At least one Pitx protein is localized in the same tissues where GCMa is expressed. Furthermore, the interaction of GCMa and Pitx modulates GCMa-dependent promoter activation in reporter gene assays. We mapped the GCMa DNAbinding domain and the homeodomain of Pitx proteins as the regions that physically associate with each other. Our findings that GCMa interacts and cooperates with a member of the Pitx transcription factor family might open a new avenue for the identification of GCMa target genes and will yield insight into the mechanistic aspects of GCMa function.
For expression and immunoprecipitation, GCMa constructs were myctagged and transferred to pCMV5 by PCR amplification using the 5Ј-primer: 5Ј-tat aga tct gcc acc atg gag gag cag aa-3Ј annealing 5Ј to the myc-epitope of pGBKT7 and the above mentioned 3Ј-primers with one of the different bait plasmids as template. Additionally a myc-tagged GCMa167, containing only the DNA-binding domain of GCMa was created by using the same 5Ј-primer and 5Ј-cag tcg act cac atg tgc act ttc ttc atg-3Ј as 3Ј-primer.
The insert of the prey plasmid containing of human Pitx1 (Acc. No. XM_047746.1, pos. 560 -1273) was transferred to pCMV5 and at the same time fused at the 5Ј-end with a T7 tag by use of BamHI and XhoI. For cloning of full-length mouse Pitx2c, BamHI and NotI were used as restriction sites and RT-PCR was carried out using adult mouse kidney as RNA source, and primers 5Ј-cgg gat ccg atg aac tgc atg aaa ggc-3Ј and 5Ј-gaa tgc ggc cgc tca cac cgg ccg gtc gac-3Ј (see Acc. No. NM_011098.1).
For identification of the interacting epitope of Pitx2c a number of deletion mutants have been generated (Fig. 1c). Primers 5Ј-gat cga att cca acg ccg gca gag gac tca-3Ј and 5Ј-gat cgt cga ctt agc gtt ccc gct ttc tcc att-3Ј were used to amplify the homeodomain via PCR. The homeodomain containing PCR fragment was fused to a 5Ј-located T7 tag by subcloning it into pCMV5 using restriction sites BamHI and SalI.
N-terminal deletions of Pitx2c were produced by using different primer pairs and restriction sites BamHI and SalI (5Ј-primers: 1, gatc ggatcc aac cag cag gcc gag ctg; 2, gat cgg atc cct gtc ctc tca gag tat gtt t; 3, gat cgg atc ctg tcc tta cgc gcc gcc gac t and 3Ј-primer: gat cgtc gac tca cac cgg ccg gtc gac). The same primers were used to subclone parts of Pitx2 into pGEX-KG to create GST fusions.
Yeast two-hybrid analysis was performed mainly according to the manufacturer's instructions (Clontech). In brief, after yeast transformation, transformants were selected on agar synthetic complete medium lacking leucine, tryptophan, or both. Screening of a human HeLa cDNA Matchmaker library (Clontech) was performed with GCMa220 fused to the Gal4-DNA-binding domain as bait. Potential candidates were selected on synthetic complete medium lacking leucine and tryptophan (positive control) or leucine, tryptophan, histidine, and adenine. Growth was evaluated after 5 days of incubation at 30°C. To ensure that positive clones do interact with the bait and not with the Gal4-DNA-binding domain instead, positive clones were separately checked for interaction with the isolated Gal4-DNA-binding domain. Selected positive clones were further confirmed by colony-lift filter assays for ␤-galactosidase activity as recommended (Clontech).
A reporter plasmid containing five modules of adjacent GCM and Pitx-binding sites was constructed. For that purpose, a doublestranded oligonucleotide with the sequence 5Ј-cgg gat cct gcc atg cgg gtg gat gct aag cct ctg tcc agt aga tct tc-3Ј was cloned five times in tandem in front of the ␤-globin minimal promoter into the BglII site of pGL2-Basic (Promega).
Amplification products were separated on 2% (w/v) agarose gels after 20 cycles (for glyceraldehyde-3-phosphate dehydrogenase) or 40 cycles (in mismatch PCRs and for mouse Pitx). In PCR experiments with mismatch primers, products were subcloned into pGEM-Teasy (Promega) and sequenced to determine their identity.
First strand cDNA from dissected rat kidney cortex and outer and inner medulla were kindly provided by Drs. K. Höcherl and A. Kurtz (University of Regensburg). Analytical PCR studies with these first strand cDNAs were performed using rat-specific primers for GCMa (5Ј-ctc cag ctc ctt acg gat ga-3Ј and 5Ј-agg agg ctg gaa ctg gtc tt-3Ј), Pitx2 (5Ј-gac tgg agg tgc ata caa tc-3Ј and 5Ј-ata ctg gca agc act cag gt-3Ј), and ␤-actin (5Ј-tct aca atg agc tgc gtg tg-3Ј and 5Ј-cca tct ctt gct cga agt ct-3Ј). Quantitative PCR reactions were performed using the LightCycler-FastStart DNA Master SYBR Green kit and the Light-Cycler Thermal Cycle system (Roche Applied Science) according to the manufacturer's instructions.
Tissue Culture, Transfection, Extract Preparation, Western Blot, Luciferase Assays, Gel Retardation Assays, Immunoprecipitation, and GST Pull-downs-COS7, HEK293, and JEG-3 cells were maintained in Dulbecco's modified Eagle's medium containing 10% (v/v) fetal calf serum. HEK293 and JEG-3 cells were transfected in 60-mm dishes with 1 g of luciferase reporter and 0.1 g of CMV-driven GCMa or 0.5 g of Pitx expression vector for luciferase assays using DNA calcium phosphate precipitates (19). COS7 cells were transfected in 100-mm plates for preparation of extracts using the DEAE-dextran technique followed by chloroquine treatment. The total amount of plasmid was kept constant using empty CMV vector. At 48 h post-transfection, cells were harvested for luciferase assays or extract preparation as described (19). A monoclonal antibody directed against the T7-epitope (Novagen) or against the myc-epitope (Cell Signaling) and a polyclonal serum against the carboxyl-terminal half of GCMa (3) served as primary antibodies (1:10,000 dilution for monoclonal antibodies, 1:3,000 dilution for polyclonal ones), horseradish-peroxidase-coupled-protein A or anti-mouse-Ig-coupled-horseradish-peroxidase as secondary detection reagents in Western blots using the ECL detection system (Amersham Biosciences).
For electrophoretic mobility shift assays, 0.5 ng of 32 P-labeled probe were incubated with COS7 cell extract for 20 min on ice in a 20-l reaction mixture as described using poly[d(IC)] as unspecific competitor (5). For supershift studies 0.5 l of mouse anti-myc antibody (9E10, Cell Signaling) recognizing myc-tagged GCMa or mouse anti-T7 antibody (Novagen) binding to T7-tagged Pitx2 was added to the reaction mixture. Samples were loaded onto native 4% (w/v) polyacrylamide gels and electrophoresed in 0.5ϫ TBE (45 mM Tris, 45 mM boric acid, 1 mM EDTA, pH 8.3) at 120 V for 1.5 h. Gels were dried and exposed for autoradiography.
For immunoprecipitations, transiently transfected COS7 cells grown on 100-mm plates were lysed in the presence of 2 g/l leupeptin and aprotinin each in ice-cold 10 mM Hepes (pH 7.9), 0.2 mM EDTA, 2 mM dithiothreitol, and 1% Nonidet P-40. Immediately after lysis, NaCl was added to a final concentration of 400 mM. After incubation for 15 min under constant rotation, cell debris was removed from the extract by centrifugation. One-fifth of the extract was used for immunoprecipitation after adding 1 l of a monoclonal antibody against the T7-(Novagen) or myc-epitope (Cell Signaling) in 500 l of HNTG buffer (50 mM Hepes, pH 7.5, 150 mM NaCl, 1 mM EDTA, 10% glycerol, 0.1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride) under constant rotation at 4°C overnight. Next, 20 l of protein A-Sepharose CL-4B beads (Amersham Biosciences) were added, and incubation continued for another 1 h. After washing the beads five times with HNTG buffer, the precipitated proteins were analyzed by SDS-PAGE and Western blot. Alternatively, adult mouse kidneys were homogenized in lysis buffer (50 mM Tris, pH 7.6, 100 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, 0.5% Nonidet P-40, complete protease inhibitor mixture (Roche Applied Science)). Homogenization was carried out at 4°C using a Dounce homogenizer with a typical tissue to buffer ratio of 8 ml of buffer per gram of tissue. After 20 strokes of homogenization, homogenates were centrifuged at 10,000 ϫ g for 5 min, and the supernatants were collected. To bind antibodies to protein A-Sepharose, 40 l of protein A beads were first incubated overnight with an equal mixture of goat-anti-Pitx2-C16 (Santa Cruz Biotechnology) and goat-anti-Pitx2-N17 (Santa Cruz Biotechnology) at 4°C in PBS. After removal of unbound supernatant, the antibody-loaded protein A-Sepharose was washed three times with PBS. The following steps were the same as for immunoprecipitation of cell extracts (see above). For detection of endogenous GCMa protein in Western blots a rabbit anti-GCMa antibody (3) was used.
For GST pull-downs, bacteria were grown in 200-ml cultures until A 600 reached 0.4 and induced by 1 mM isopropyl 1-thio-␤-D-galactopyranoside for 4 h to express the GST fusion proteins. Bacteria were collected by centrifugation, incubated in sonification buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 25 units/ml benzonase, 10 g/ml leupeptin, 10 g/ml aprotinin, 1 l/ml Triton X-100, 10 g/ml DNase I, 15 units/l lysozyme) at 4°C for 30 min, lysed by sonication, and centrifuged. The supernatants containing the GST fusion proteins were supplemented by 30 l of equilibrated glutathione beads and incubated under constant rotation at 4°C. After washing three times with washing buffer (4.3 mM Na 2 HPO 4 , 1.47 mM KH 2 PO 4 , 1.37 mM NaCl, 2.7 mM KCl) an aliquot of the beads, now carrying the GST fusion protein, was incubated together with 25 l of extract from COS7 cells or mouse kidney tissue expressing the desired protein after transient transfection. After washing three times with washing buffer, proteins bound to the beads were analyzed by SDS-PAGE and Western blot.
Tissue Sections and Immunohistochemistry-Mouse mating and genotyping were performed as described previously (14). For immunohistochemical analysis, kidneys were fixed overnight at 4°C in 4% paraformaldehyde, cryoprotected in 30% sucrose, and embedded in Tissue-Tec (Leica Instruments). Kidneys from adult mice were cryotome-sectioned to 12-m slices. Cryotome sections were incubated with rabbit or goat anti-lacZ antibodies and polyclonal goat-anti Pitx2-C16 (Santa Cruz Biotechnology). Secondary antibodies conjugated to Cy2 and Cy3 immunofluorescent dyes (Dianova) were used for detection.

RESULTS
Pitx Homeodomain Proteins Interact with GCMa-To isolate cofactors interacting with GCMa, we set up a yeast two-hybrid screen. Because fusions of the Gal4 DNA-binding domain to carboxyl-terminal GCMa regions containing the protein's transactivation domains (44) induced Gal4-dependent reporters in transformed yeast, even in the absence of a Gal4-activation domain, we fused different amino-terminal portions of GCMa to the Gal4 DNA-binding domain. The amino-terminal part of GCMa up to the first transactivation domain exhibited tolerable autonomous transactivation and could be used as bait for the Gal4-based yeast two-hybrid system (GCMa220, Fig. 1,  a and d). Expression of the bait (GCMa220) was verified by Western blot. Furthermore, growth kinetics of transformed yeast were not disturbed indicating the absence of toxicity for the bait (data not shown). After screening roughly 750,000 independent yeast colonies from a Matchmaker HeLa cDNA library (Clontech) in the AH109 yeast strain, 20 colonies grew on selective media lacking adenine and histidine. Of these, five clones could be reconfirmed as true positives after retransformation and filter assays for ␤-galactosidase activity. Among the clones identified, two consisted of the carboxyl-terminal part of the open reading frame of human Pitx1 (prey, Fig. 1b) coding for residues 77-315 and corresponding to the homeodomain plus the following regions until the regular stop codon (Fig. 1, b and d).
Only One Pitx Member, Pitx2, Is Present in the Developing Placenta and in the Kidney-To analyze whether Pitx genes are transcribed in the same tissues as GCMa, we performed RT-PCR studies (Fig. 2a). First we used a mismatch primer pair capable of amplifying the homeodomains of all three Pitx members. Using these primers, we detected signals in tissues of adult mice, namely in brain, kidney, and spinal cord. We failed to detect any mRNA for Pitx members in adult liver and thymus (Fig. 2). Additionally, we observed Pitx transcripts in placenta of GCMa ϩ/ϩ and GCMa null animals (GCMa lacZ/lacZ , see Fig. 2) at E9.5. Cloning and sequencing of the PCR product from kidney and placenta yielded fragments corresponding to Pitx2. No other member of the Pitx family could be detected in the cloned PCR products. By repeating RT-PCRs with specific primers for the three Pitx family members, we detected Pitx2 in all of the tissues previously identified as expressing a Pitx member, but neither Pitx1 nor Pitx3, indicating that Pitx2 is the predominantly expressed Pitx protein in these tissues (Fig.  2b). Due to technical limitations, the presence of very low levels of Pitx1 and Pitx3 in kidney and placenta cannot be excluded. Because several Pitx2 isoforms have been reported, we also used isoform-specific primers for detection of Pitx2a, Pitx2b, or Pitx2c and mainly detected Pitx2c in kidney and placenta (data not shown). For the following experiments we cloned fulllength Pitx2 from mouse kidney.
GCM and Pitx Proteins Interact in Vivo through Their DNAbinding Domains-To confirm the interaction between GCMa and Pitx proteins, we transiently transfected COS7 cells with GCMa alone or with GCMa in combination with Pitx1 or Pitx2. We immunoprecipitated proteins from these transfected cell extracts with an antibody reactive against GCMa and observed coprecipitation of Pitx1 as well as Pitx2 proteins (Fig. 3a). In complementary immunoprecipitations we used an antibody against the T7 tag fused to the amino termini of the Pitx proteins. Now, we transiently transfected COS7 cells with Pitx members alone or with Pitx members in combination with GCMa. Full-length GCMa coprecipitated with both Pitx1 and Pitx2 (Fig. 3b).
To analyze whether the interaction between GCMa and Pitx2 also occurred in a physiological context, we used tissue extract from kidney, where GCMa and Pitx2 are coexpressed (Fig. 2). First, we performed immunoprecipitations. When precipitating endogenous Pitx2 from kidney extract, we were able to coprecipitate endogenous GCMa (Fig. 3c). Concurrent electrophoresis of an extract of transiently transfected COS7 cells by GCMa served as a size-control (Fig. 3c). These data demonstrate that endogenous amounts of GCMa and Pitx2 are sufficient to interact with each other. We also successfully used a GST-Pitx2 fusion protein to pull down endogenous GCMa from kidney extracts (Fig. 3d).
To map the interacting epitopes we used truncated versions of GCMa and Pitx2 proteins (Fig. 1, a and c). The original GCMa bait is composed of the DNA-binding domain and the region between the DNA-binding domain and the first transactivation domain (GCMa220). Similar to full-length GCMa, GCMa220 can be coprecipitated with full-length Pitx2 (Fig. 4a). Coprecipitation of GCMa220 was also observed with the homeodomain of Pitx2 (Fig. 4a). In contrast, carboxyl-terminal parts of Pitx2 corresponding to amino acid residues 211-324 or 268 -324 were not able to coprecipitate GCMa220 (Pitx2-⌬211 and Pitx2-⌬268 in Fig. 4a). Having identified the homeodomain of Pitx2 as the region interacting with GCMa220, we asked which area of GCMa220 might be responsible for this contact. We created a shorter version containing only the DNA-binding domain of GCMa (GCM167, Fig. 1a). Even GCMa167 was sufficient for interaction with Pitx proteins (Fig. 4b). Because the DNA-binding domain is a conserved structural element of GCM family members, we investigated if the DNA-binding domain of GCMb is also able to interact with Pitx2. Indeed we observed that the DNA-binding domain of GCMb interacted with Pitx2 (Fig. 4b).
In addition to immunoprecipitations, we also performed GST-pull-down experiments using fusions between GST and full-length Pitx2 and parts thereof. As already observed in immunoprecipitations, GCMa220 bound to full-length Pitx2 and the homeodomain of Pitx2, but not to carboxyl-terminal parts of Pitx2 corresponding to amino acid residues 153-324, 211-324, or 268 -324 (Fig. 5a). We could also confirm in GSTpull-downs that the DNA-binding domain of GCMa is sufficient for the interaction between GCMa and Pitx2 as GCMa167 binds to both full-length Pitx2 and the homeodomain (Fig. 5b). We also inspected the binding ability of GCMb180 by GST-pulldown and confirmed that GCMb180 is able to interact with The detected PCR product represents all three mouse Pitx genes (a). This was made possible by using PCR primers recognizing conserved sequences within the homeobox domain. A glyceraldehyde-3-phosphate dehydrogenase-specific control product was amplified in parallel and is shown below. For detection of PCR products for Pitx1, Pitx2, or Pitx3, gene-specific primer pairs were used (b).

FIG. 3. GCMa and Pitx proteins coimmunoprecipitate. a and b,
tagged versions of GCMa and Pitx proteins were produced in transiently transfected COS7 cells. Cells extracts were prepared, and precipitations were performed. a, detection of T7-tagged Pitx proteins bound to GCMa precipitated with a GCMa-specific antibody. b, analysis of GCMa bound to T7-tagged Pitx proteins precipitated with a T7specific antibody. 1/5 of the input is shown for the coprecipitated proteins (c and d). Mouse kidneys were homogenized to produce extracts containing endogenous GCMa and Pitx2 proteins. c, coimmunoprecipitation of GCMa complexed to Pitx2 from kidney extract using Pitx2recognizing antibodies. As a size-control, GCMa from transiently transfected COS7 cells is shown. d, endogenous GCMa from kidney extract bound to a GST fusion containing the full-length Pitx2. GST-pull-downs with GST alone are shown as negative control. For detection of endogenous GCM proteins by Western blots an antibody reactive against GCM proteins was used in c and d.
full-length Pitx2 and with the homeodomain of Pitx2 (Fig. 5c). Our GST-pull-downs and immunoprecipitations consistently reveal that the DNA-binding domains of the two respective transcription factors mediate their interaction.
Pitx2 and GCMa Bind Cooperatively to DNA-To assess whether the observed interaction between GCMa and Pitx2 might influence DNA binding characteristics of the two proteins, we performed electrophoretic mobility shift assays. Using an oligonucleotide containing GCMa and Pitx binding motifs separated by 6 bp (Fig. 6a), we first showed that both GCMa and Pitx2 are able to bind to this oligonucleotide. However, GCMa appears to bind to this oligonucleotide with higher affinity (Fig. 6b). When incubating a low amount of Pitx with the labeled probe, the amount of the dimeric complex was below the detection limit. But when increasing amounts of GCMa were added in the presence of this low amount of Pitx2, we detected a ternary complex consisting of GCMa, Pitx2, and the probe (Fig. 6b). The amount of this ternary complex strongly exceeded the amount of Pitx2-DNA complex formed with the same amount of Pitx2 in the absence of GCMa indicating that increasing amounts of GCMa allowed more Pitx2 protein to bind to the probe (Fig. 6b). To verify the composition of the ternary complex we performed supershift studies. An antibody that detects the myc tag fused to GCMa, recognized GCMa bound to DNA, and caused a significantly slower migration of this binary complex. Slower migration of a Pitx2-containing DNA complex was also be observed when an antibody was used that recognizes the T7 tag fused to Pitx2. If the antibody reactive against the myc tag of GCMa was added to reactions that contained both GCMa and Pitx2 the ternary complex was also supershifted. Additionally, part of the Pitx2 was released from the ternary complex resulting in re-appearance of a binary complex composed of Pitx2 bound to labeled probe. The Pitx2-recognizing anti-T7 antibody also supershifted the ternary complex. Additionally, most of the GCMa was released from the ternary complex and re-appeared in a binary complex with the labeled probe (Fig. 6c).
Pitx2 Changes GCMa-dependent Promoter Activation in a Cell-specific Manner-GCM proteins bind to the octamer AT-GCGGGT motif on DNA and can thus activate luciferase reporter genes whose expression is under the control of a promoter containing six tandemly arranged GCM-binding sites (6xgbs-luc) (1). This reporter was used in transfections of two different cell lines to examine the influences of cotransfected Pitx proteins (Fig. 7). Whereas Pitx1 and Pitx2 do not influence FIG. 4. Mapping of interacting epitopes of GCMa and Pitx proteins. GCM and Pitx cDNAs were transiently transfected into COS7 cells, cells extracts were prepared, and precipitations were performed. a, detection of T7-tagged Pitx proteins bound to GCMa precipitated with a GCMa-specific antibody. b, analysis of GCMa bound to T7-tagged Pitx proteins precipitated with a T7-specific antibody. c, Western blot shows myc-tagged GCMa220 that bound to T7-tagged full-length Pitx2 or indicated fragments thereof and hence was precipitated with a T7specific antibody. d, detection of T7-tagged full-length Pitx2 bound to myc-tagged GCMa220, GCMb180, or GCMa167. Precipitation was performed with a monoclonal myc-specific antibody (9E10). 1/10 of the input is shown for the coprecipitated protein.

FIG. 5. GCM and Pitx family members interact in GST-pulldowns.
GST fusions were purified from bacteria and incubated with cell extracts of COS7 cells transiently transfected with different GCM cDNAs. a, analysis of GCMa220 binding to GST fusions containing either full-length Pitx2 or fragments Pitx2-HD, Pitx2-⌬153, Pitx2-⌬211, or Pitx2-⌬268 as indicated above the lanes. b, GCMa167 binding to full-length Pitx2 and the Pitx2 homeodomain (Pitx2-HD). c, GCMb180 binding to full-length Pitx2 and the Pitx2 homeodomain (Pitx2-HD). For detection of GCM proteins by Western blots an antibody reactive against the myc tag of GCM proteins was used. As control pull-downs with GST alone were always performed in parallel. 1/10 of the input is shown.
transcription from the 6xgbs reporter in HEK293 cells, a robust 151-fold increase of luciferase activity was detected with GCMa alone (Fig. 7a). In the presence of either Pitx1 or Pitx2, the GCMa-dependent activation rates declined to 2.7-or 43.9-fold, respectively, arguing that Pitx proteins inhibit GCMa-mediated transactivation of the reporter (Fig. 7a).
Transfection of GCMa, Pitx1, or Pitx2 together with the same 6xgbs-luc luciferase reporter into JEG-3 yielded a different picture. GCMa increased the promoter activity by 33-fold, whereas Pitx1 and Pitx2 even repressed promoter activity by 10-and 5-fold, respectively (Fig. 7b). Surprisingly, GCMa together with either Pitx1 or Pitx2 increases the promoter activ-ity up to three times more than GCMa alone (Fig. 7b). This Pitx-mediated increase of GCMa-dependent promoter activation is all the more noteworthy, because Pitx1 and Pitx2 alone have a repressive effect (Fig. 7b). The results demonstrate that Pitx proteins increase the capacity of GCMa to turn on promoters in JEG-3 cells, whereas in HEK293 cells Pitx proteins inhibit GCMa-dependent promoter activation.
Additional transactivation assays were performed using a reporter containing in its promoter five tandemly repeated composite elements each consisting of a GCM and a Pitxbinding site (5x(gbsϩpbs)-luc, see Fig. 7, c and d). This reporter was designed in analogy to Lamolet et al. (45), who described the requirement of a T-box-binding site for Pitx1 activation of POMC transcription. The spacing of 10 bp between the centers of GCM-and Pitx-binding sites ensures close apposition of GCMa and Pitx2 on DNA, because the motifs will be presented one turn apart on the same site of the DNA double helix. Whereas GCMa itself was unable to change promoter activity of 5x(gbsϩpbs)-luc in JEG-3 cells, Pitx 1 and Pitx2 elicited 35or 17-fold inductions of the reporter, respectively (Fig. 7d). Cotransfection of GCMa with either Pitx1 or Pitx2 in JEG-3 cells exhibited robust 162-or 165-fold activations (Fig. 7d). Again, synergistic activation was only observed in JEG-3 but not in HEK293 cells. In HEK293 cells GCMa, Pitx1, and Pitx2 activate the 5x(gbsϩpbs)-luc reporter 5.1-, 5.5-, and 1.7-fold, respectively (Fig. 7c). In the presence of both Pitx and GCMa proteins, induction corresponds to the additive activation rates (Fig. 7c).
GCMa and Pitx2 Colocalize in the kidney-Our RT-PCR data show the presence of Pitx2 in the kidney. The presence of GCMa in the kidney was shown previously (18). Therefore we examined if both transcription factors colocalize to the same FIG. 6. GCMa facilitates binding of Pitx2 to DNA. a, the sequence of the probe used for electrophoretic mobility shift assays is given. Binding motifs for GCM and Pitx proteins were marked. b, binding of increasing amounts of GCMa to the radiolabeled probe in the presence or absence of low amounts of human Pitx2a (Pitx2 low). Higher amounts of Pitx2a were used in the first lane (Pitx2 high). Note that the ternary complex appears with increasing amounts of GCMa. Incubation of radiolabeled probe with GCMa and Pitx2 in the presence of antibodies against the myc tag fused to GCMa or the T7 tag fused to Pitx2. c, the ternary complex composed of GCMa, Pitx2a, and the radiolabeled DNA displayed slower migration in the presence of each of the two antibodies whereas the binary complexes were supershifted with only one of the two antibodies. Arrows in a and b indicate positions of binary complexes containing either GCMa or human Pitx2a in addition to the probe and the position of the ternary complex. region in this tissue. First, we used first strand cDNA from cortex and outer and inner medulla of dissected rat kidneys. Quantitative and qualitative PCR amplification with rat-specific primers for GCMa and Pitx2 revealed predominant presence of transcripts for both genes in the outer medulla, the region with the highest density of proximal tubules (Fig. 8, a  and b). Second, we used previously generated GCMa ϩ/lacZ mice, where a lacZ marker replaces all GCMa coding sequence except the 32 amino-terminal residues to which it is fused. Hence these mice are perfect as GCMa-reporter mice (14). For cellular colocalization studies we prepared frozen sections of adult kidneys of these mice and immunostained them with antibodies reactive against Pitx2 and ␤-galactosidase. We demonstrated a strong overlap of ␤-galactosidase (representing GCMa expression) and Pitx2 in cells forming tubular structures in kidney. Given the fact that GCMa has been previously shown to specifically localize to S3, Pitx2 is most likely localized to the same S3 segment of proximal tubules. Additionally, a low number of cells stained positive for Pitx2 in adjacent regions of both cortex and inner medulla (Fig. 8c). DISCUSSION In this study, we present data demonstrating for the first time binding of Pitx proteins to transcription factors belonging to the family of GCM proteins. In particular, Pitx2, which is expressed together with GCMa in kidney and in placenta interacts with GCMa. No interactors have been described so far for GCMa and only a limited number for Pitx proteins. One of them is the pituitary-specific POU-homeodomain protein Pit-1 (28). The other two described Pitx-binding partners are either the basic helix-loop-helix proteins NeuroD1/Pan1 or Tpit, which belongs to the T-box transcription factor family (45,46). GST-pull-down data showed binding between Pitx2 and Pit-1, but the interaction was not narrowed down to any particular region (28,46). In the case of NeuroD1/Pan1, evidence for a direct physical interaction between the basic helix-loop-helix domain of Pan1 and the homeodomain of Pitx1 was presented (28,46). This latter interaction seems to be specific for the two protein classes rather than for the two proteins, because both proteins are exchangeable by other basic helix-loop-helix and Pitx proteins (46). In analogy, we observed binding between all analyzed GCM and Pitx members, obviously because the conserved DNA-binding domains of the two proteins mediate the interaction between them. Regardless of the ability of GCMa to bind all Pitx members, Pitx2 probably is the only interactor in vivo, because other Pitx paralogues are not expressed at sites of GCMa expression. On the other hand, the fact that all GCM and Pitx members are able to interact with each other makes an interaction between GCMb and Pitx protein members in the parathyroid gland possible. It remains to be seen which of the Pitx proteins, if any, are expressed in the parathyroid gland. If there really is an interaction between GCMb and a Pitx protein, it might also be of relevance for the function of GCMb in a number of human tumor cell lines and in parathyroid adenoma (47,48), where GCMb has been found to be expressed.
In the case of Pit1 it has been previously observed that, as a consequence of protein-protein interaction, binding of Pitx2 to its cognate DNA-binding site is facilitated (49). We also observed that GCMa helps recruiting Pitx2 to its binding site. This might argue that Pitx proteins generally need coactivators to bind to DNA. But it could also mean that in cases where binding of Pitx proteins is necessary to modulate gene expression, Pitx proteins only occupy their binding sites if their partner transcription factor is already present. Furthermore, the interaction between Pitx and its binding partners has been reported to result in a transcriptional synergism (30,45,46). This synergism was detected on either the prolactin-or pro-opiomelanocortin-promoter containing reporters. In line with these previous observations, the interaction between GCMa and Pitx2 synergistically induced transcription of promoters containing solely GCM DNA-binding sites or GCM and Pitx DNA-binding sites. However, in contrast to published cases the transcriptional synergism observed between GCMa and Pitx proteins is cell-specific and was only observed in JEG-3 cells. In HEK293 cells we even measured repressive effects of Pitx proteins on GCMa-dependent promoters of certain configurations. Such repressive effects have not been observed so far for Pitx proteins but might be more common than previously thought.
Pitx2-deficient mice have primarily been studied for their defects in laterality and organogenesis (50). Thus it is difficult to compare placental phenotypes of Pitx2-and GCMadeficient animals. The observed early mortality of a fraction of Pitx2 embryos around E10 is compatible with the existence of additional placental defects. These have, however, not been investigated so far. Similarly, functions of both proteins in the kidney cannot be compared in the currently available mouse mutants, because GCMa knockout mice do not survive long enough.
Recently we speculated that GCMa is involved in mediating physiological rather than developmental functions of the kidney, because GCMa is not expressed in the embryonic kidney, but begins to be expressed perinatally (18). Now that we identified GCM and Pitx family members coexisting in the same cells in the kidney, potential target genes for both transcription factors might be postulated to contain adjacent GCM and Pitxbinding sites in their regulatory elements. Therefore, computational searches for adjacent GCM and Pitx-binding sites FIG. 8. Colocalization of GCMa and Pitx2 in mouse kidney. a, PCR with first strand cDNA generated from RNA of dissected kidney regions. Abbreviations describe different sections of the kidney; CO, cortex; OM, outer medulla; IM, inner medulla. For amplification of GCMa-and Pitx2-specific products, 40 cycles of PCR were performed. b, quantitative evaluation of the amount of GCMa and Pitx2 transcripts were performed by real-time PCR (LightCycler). Amounts were determined relative to the inner medulla which arbitrarily was set to 1. c, immunohistochemistry on cryosections of adult kidney. 12-m cryosections of the adult mouse kidney were incubated with antibodies that specifically recognize ␤-galactosidase (green) or Pitx2 (red) as indicated. ␤-Galactosidase is expressed from the GCMa locus and mimics GCMa expression. An overlay of the two images is shown on the right. Magnification, ϫ200. might enable the identification of such target genes in the kidney.
The presence of Pitx2 transcripts in placenta is compatible with the assumption that GCMa and Pitx2 might also colocalize in the placenta and jointly regulate common target genes. Due to limitations in the quality of antibodies we could not directly investigate this. However, if true, comparison of GCMa-and Pitx2-dependent target genes in kidney and placenta might help to understand the biological significance of the interaction between these two transcription factors.