The Embryonic Function of Germ Cell Nuclear Factor Is Dependent on the DNA Binding Domain*

Germ cell nuclear factor (GCNF), an orphan nuclear receptor, is essential for mouse embryogenesis. GCNF specifically binds to a DR0 response element via its DNA binding domain (DBD) in vitro and functions as a repres-sor of gene transcription. To further study the role of GCNF during embryogenesis, we have employed a Cre/ loxP strategy and generated a line of GCNF mutant mice ( GCNF lox/lox ) in which the 243-base pair DBD-encoding exon has been deleted in the germline. However, the ligand binding domain (LBD) of GCNF is still expressed at the mRNA and protein levels in the GCNF lox/lox mice. GCNF lox/lox mice die at 9.5–10.5 days postcoitum. The tailbuds of these mutant embryos protrude outside the yolk sac. Expression of Oct-4 in the somatic cells of GCNF lox/lox embryos at 8.25 days postcoitum was not silenced as in the GCNF (cid:1) / (cid:1) embryos. Therefore, GCNF lox/ lox mice phenocopy the GCNF (cid:2) / (cid:2) mice. Our results indicate that the DBD is essential for the function of GCNF during early mouse embryogenesis, and that the LBD does not mediate any function independent of the DBD at this stage of embryonic development. Our results also suggest that GCNF is indeed a transcriptional factor that represses gene transcription mediated via its DBD. of the offspring resulting from the heterozygous intercross by Southern blot analysis and PCR. Digestion with Acc 65 I and Xho I yielded a 20 kb band for the wild type GCNF allele and a 10-kb band for GCNF lox allele. Using Primer and Primer 2, a 264-bp wild type DBD exon DNA fragment was detected in the GCNF (cid:4) / (cid:4) and GCNF (cid:4) /lox animals, but not in the GCNF lox/lox embryos. Primer 3 and Primer 4 were used to detect a 400-bp DNA fragment for the GCNF lox allele the

Germ cell nuclear factor (GCNF, 1 NR6A1) is a novel orphan member of the nuclear receptor superfamily as it is more distantly related to other members and forms a sixth and separate subbranch of the family (1,2). GCNF was initially cloned by our laboratory using low stringency screening with a DNA binding domain (DBD) probe (3) and subsequently cloned by other laboratories and given other names, e.g. RTR (retinoid receptor-related testis-specific receptor (4)) and NCNF (neuronal cell nuclear factor (5)). To date, homologs of GCNF have been cloned from several other species including human, Xenopus, and zebrafish (2, 6 -8). The mouse GCNF gene contains 11 exons (9) and is located on chromosome 2, 2 while the human GCNF gene is located on chromosome 9 at the locus q33-34.1 (10,11). Sequence analysis has shown that GCNF homologs have high homology in the DBD and the ligand binding domain (LBD) among different species (2,8). This high amino acid identity within the DBD and LBD across species indicates the functional conservation of GCNF protein in binding to its DNAresponse element and its putative ligand during evolution. Indeed, the DNA binding specificity of GCNF is conserved among mouse, human and Xenopus (3,(12)(13)(14)(15).
GCNF can specifically bind to either a DR0 element, a direct repeat of the estrogen receptor half-site (AGGTCA) with zero base pair spacing between the half-sites, as a homodimer (3,12,13,(15)(16)(17), or an extended DR0 half-site (TCAAGGTCA) either as a monomer (12), a homodimer (13,17), or both a monomer and homodimer (17). Recently, two dimerization motifs in mouse GCNF, one located in the DBD including the adjacent TA box and the other in ␣-helix 3 of the LBD, have been characterized (17). The TA box is critically involved in homodimeric interactions on the DR0 element (16,17), while both the TA box and the ␣-helix 3 are required for homodimeric binding of mouse GCNF to the extended DR0 half-site (17). Characterization of the DNA binding motif of GCNF led to the identification of several potential target genes for GCNF including the protamine 1 and 2 genes in the testis (12,18) and the Oct4 gene in the mouse embryo (19). In the absence of a ligand, GCNF functions to repress the transcription of these target genes in vitro and in vivo (7,19,20).
It has been shown that GCNF is not only expressed in mouse and Xenopus embryos after the onset of gastrulation (14,(21)(22)(23)(24) but also in the gametogenic cells of adult vertebrates (3,4,11,18,(25)(26)(27). This expression pattern indicates that GCNF may play a role in gametogenesis and normal embryonic development during gastrulation. Recently, we have generated a GCNF knockout mouse model (GCNF Ϫ/Ϫ ) using a conventional embryonic stem cell strategy to address the functions of GCNF during mammalian embryogenesis (24). Insertion of PGK-neomycin (PGK-neo) into the GCNF locus causes embryonic lethality at 9.5-10.5 dpc. The most remarkable phenotype of these GCNF mutant embryos is that the posterior tailbud develops outside of the yolk sac (24). These mutant embryos have serious defects in posterior and trunk development and somitogenesis (24). In addition, expression of the POU domain transcription factor, Oct4, is not repressed in the somatic cells of these GCNF mutant embryos at 8.25-8.75 dpc when it is silenced in normal embryos (19). It appears that GCNF is required for normal anteroposterior development, somitogenesis, and Oct4 expression in mouse embryos. However, recent studies in several other laboratories have shown that the introduction of the PGK-neo into the mouse genome by conventional ES cell targeting can cause unexpected phenotypes in the resulting animal models (28 -32). Therefore, generation and phenotypic analysis of a line of GCNF mutant mice, in which the PGK-neo cassette is removed, will definitely clarify whether the pheno-types observed in the conventional GCNF knockout mice are due to the ablation of the GCNF gene or to the insertion of the PGK-neo cassette in the GCNF allele, causing misexpression of neighboring genes such as the steroidogenic factor 1 (SF-1) gene, which is located 3Ј to the GCNF gene.
As a member of the nuclear receptor superfamily, the DBD of GCNF is required to bind to its response elements and regulate gene transcription. Deletion of the 55 amino acid residues containing the DBD zinc finger region in the N terminus of Xenopus GCNF abolishes its DNA binding activity, and overexpression of this truncated GCNF protein in Xenopus embryos causes abnormal head development (14). How much of the function of GCNF is mediated by the DBD in mouse embryonic development remains to be determined. GCNF may have DNA binding-independent activities, similar to other nuclear receptors such as glucocorticoid receptor (33) and TR3 (34).
In this study, we have generated a line of GCNF mutant mice lacking the 243-bp DBD-encoding exon 4 of the GCNF gene (9) using the Cre/loxP system (35). We found that these mutant mice have the same phenotypes as the conventional GCNF knockout mice (19,24). Unlike the conventional GCNF knockout mice, the LBD of GCNF was still expressed in these mutant embryos at both the mRNA and protein levels. Therefore, the DBD of GCNF is essential for mediating the function of GCNF during mouse embryonic development.

EXPERIMENTAL PROCEDURES
Construction of the Targeting Vector-Genomic clones of the murine GCNF gene have been isolated from a 129Sv strain mouse genomic library (24). The targeting vector was constructed from a genomic clone containing exon 4 encoding the GCNF DBD. First, oligonucleotides containing an XhoI restriction site and a loxP site were synthesized and then inserted into a neo/tk selection cassette (pNeoTKLOX, generously provided by Dr. Allan Bradley) flanked by two loxP sites to generate the plasmid ZJ-1. A second plasmid, ZJ-2, was constructed when the 3.6-kb BglI/ApaI fragment, downstream of the DBD exon from the genomic clone, was ligated into the BstXI site of the plasmid ZJ-1 in the presence of linkers. Then, the 1.3-kb BsiEI/BglI fragment flanking exon 4 from the genomic clone was inserted into the SalI site of the plasmid ZJ-2 to generate the plasmid ZJ-4. Finally the targeting vector, ZJ-5, was constructed when the 5.1-kb EcoRV/BsiEI fragment from the genomic clone 5Ј of the DBD exon was inserted into the KpnI site of the ZJ-4 plasmid in the presence of linkers. This targeting vector (ZJ-5) contains 6.4 kb of homologous DNA on the 5Ј-long arm and 3.6 kb of homologous sequence on the 3Ј-short arm flanking the neo/tk cassette (Fig. 1A).
Generation of GCNF lox Mice-GCNF lox mice were generated by homologous recombination in the AB1.2 ES cell line (provided by Dr. Allan Bradley) using the Cre/loxP system. The KpnI-linearized targeting vector ZL-5 (20 g) was electroporated into 10 7 ES cells, and stably transfected clones were isolated after selection with 400 g/ml G418 for 10 days. Homologously recombined clones were identified by mini-Southern blot analysis using 5Ј-GCNF and neo probes (24). Subsequently, correctly recombined clones were amplified and then transiently transfected with a Cre expression vector (pOG231, 1 g) (36). After negative selection in FIAU (1-(2Ј-deoxy-2Ј-fluoro-1-␤-D-arabinofuranosyl-5-iodo)uracil), surviving clones were picked and analyzed for recombination of the two loxP sites using mini-Southern blot analysis with the 5Ј-GCNF and neo probes. Correctly recombined clones were expanded and microinjected into mouse C57BL/6 blastocysts to generate chimeric animals. Male chimeric mice (greater than 90% agouti coat color) were bred with C57BL/6 female mice to generate heterozygous GCNF lox mice. Two clones produced chimeric males that demonstrated germ line transmission of the GCNF lox allele. Heterozygous GCNF lox mice were intercrossed to generate homozygous GCNF lox/lox embryos.
Whole-mount in Situ Hybridization and Histological Analysis-Wild type and GCNF lox/lox embryos at the same somite stage were obtained on 7.5-8.5 dpc of pregnancies. The embryo yolk sacs were removed for genotyping. Whole-mount in situ hybridization was performed as previously described (24). The cRNA probe used for Oct4 was as described by Scholer et al. (38). Histological analysis of embryonic sections was performed as previously described (24).

RESULTS
Targeted Deletion of the DBD-encoding Exon of the GCNF Gene-A Cre/loxP targeting strategy was used to delete the 243 bp DBD-encoding exon of the GCNF gene. As shown in Fig. 1A, three loxP sites were introduced into a GCNF allele of ES cells by homologous recombination with the targeting vector. Mini-Southern blot analysis was performed to identify the ES clones that had gone under correct homologous recombination using 5Ј-GCNF and neo probes (Fig. 1B). Nineteen ES cell clones carrying three loxP sites were obtained from the screening of 160 ES cell clones. Subsequently, two independent ES cell clones carrying three loxP sites were amplified and then transiently transfected with a CMV-Cre expression plasmid to delete the GCNF DBD-encoding exon and the neo/tk cassette. ES cell clones carrying the recombined GCNF lox allele were identified by the Southern blot analysis using the 5Ј-GCNF and neo probes (Fig. 1C). Sixty-seven (of 80) ES cell clones carrying the GCNF lox allele were obtained, two of which were used to generate chimeric mice. Chimeric mice from each clone transmitted the recombined GCNF lox allele to their offspring. Heterozygous (GCNF ϩ/lox ) and homozygous (GCNF lox/lox ) embryos were identified by Southern blot analysis and PCR (Fig. 1D).
Embryonic Lethality of GCNF lox/lox Mutants-At weaning, no homozygous GCNF lox/lox mice were obtained from the intercross of heterozygous mice (Table I), indicating that deletion of the DBD-encoding exon caused embryonic lethality, similar to the GCNF null mutant mice reported previously (24). To determine when homozygous GCNF lox/lox mutants die, heterozygous females were sacrificed at various days postcoitum after mating with heterozygous males, and the embryos were collected and then genotyped by either PCR or Southern blot analysis (Fig. 1D). At 8.5-9.5 dpc, embryos obtained from the GCNF ϩ/lox X GCNF ϩ/lox mating showed the expected 1:2:1 FIG. 1. Generation of GCNF lox mice using ES cell homologous recombination and Cre/loxP technologies. A, targeting strategy for homologous recombination. GCNF 3lox allele represents the structure of the GCNF locus after homologous recombination with the targeting vector. This recombined 3lox allele contains three loxP sites (filled triangles), one upstream of the DBD-encoding exon (filled box) and the others downstream of the DBD-encoding exon flanking the neo/tk selection cassette. The Cre recombined allele (GCNF lox ) marks the structure after subsequent recombination of the two-loxP sites, one upstream of the DBD-encoding exon and the other downstream of the neo/tk cassette, by Cre recombinase. Restriction enzyme sites of Acc65I (Ac), ApaI (Ap), BglI (Bg), BsiEI (Bs), EcoRv (Rv) XhoI (X), and SalI (S) in each allele are included. B, mini-Southern blot analysis to identify ES clones containing the GCNF 3lox allele. Digestion with Acc65I and XhoI yielded a 20-kb band for the wild type GCNF allele and a 10-kb band for the targeted GCNF 3lox allele, indicating the correct homologous recombination in the 5Ј-arm. Digestion with Acc65I and SalI yielded a 12.6-kb band for the targeted GCNF 3lox allele, indicating the correct homologous recombination in the 3Ј-arm. C, Southern blot analysis to determine ES cell clones containing the GCNF lox allele. Digestion with Acc65I yielded a 20-kb band for the wild type GCNF allele and an 18.6-kb band for the GCNF lox allele. No band detected by the neo probe confirms the complete recombination between the loxP sites in the GCNF 3lox allele by Cre recombinase. D, genotyping of the offspring resulting from the heterozygous intercross by Southern blot analysis and PCR. Digestion with Acc65 I and XhoI yielded a 20 kb band for the wild type GCNF allele and a 10-kb band for GCNF lox allele. Using Primer 1 and Primer 2, a 264-bp wild type DBD exon DNA fragment was detected in the GCNF ϩ/ϩ and GCNF ϩ/lox animals, but not in the GCNF lox/lox embryos. Primer 3 and Primer 4 were used to detect a 400-bp DNA fragment for the GCNF lox allele in the embryos.
Presence of an Intact GCNF LBD in the GCNF lox/lox Mutants at Both the mRNA and Protein Levels-Since the size of the DBD-encoding exon of the GCNF gene is 243 bp and the loxP sites were inserted into the intron surrounding the DBD exon, we postulated that deletion of the 243-bp DBD-encoding exon should not cause a reading frameshift in the GCNF downstream LBD in GCNF lox/lox mutants ( Fig. 2A). To determine whether the DBD of GCNF is deleted and the LBD of GCNF is still expressed in the GCNF lox/lox embryos, we performed RT-PCR analysis and RNase protection analysis. As shown in Fig.  2B, the DBD of the GCNF transcript was completely deleted in the GCNF lox/lox mutant embryos. However, the LBD of the GCNF transcript was still expressed in the mutant embryos (Fig. 2B). To determine whether the LBD protein is present in the GCNF lox/lox embryos, we performed Western blot analysis using antibodies raised against the LBD of GCNF. As shown in Fig. 2C, a 58-kDa GCNF protein band was detected in GCNF ϩ/ϩ and GCNF ϩ/lox embryos but not in GCNF Ϫ/Ϫ embryos. A truncated GCNF protein band (50 kDa) was also detected in the heterozygous and homozygous GCNF lox embryos, indicating that the LBD is present in GCNF lox/lox embryos.
Morphologic and Histologic Abnormalities in GCNF lox/lox Embryos Phenocopy GCNF Ϫ/Ϫ Embryos-Similar to GCNF Ϫ/Ϫ embryos, there was no gross morphological difference between GCNF ϩ/ϩ and GCNF lox/lox embryos before 8.0 dpc. From 8.75 dpc, a protrusion of tissue started to appear at the base of the allantois of GCNF lox/lox embryos (data not down). At 9.25-9.5 dpc, the tail bud continued to develop outside the yolk sac (Fig.  3A). This phenotype was also observed in the GCNF Ϫ/Ϫ embryos. In addition, other phenotypes of these mutant embryos resembled those in GCNF Ϫ/Ϫ embryos (Fig. 3A). The GCNF lox/lox embryos did not undergo turning and remained in a lordotic position. Although the anterior neural tissue continued to grow, the anterior neural tube in GCNF lox/lox embryos remained open. Similar to GCNF Ϫ/Ϫ embryos, the GCNF lox/lox embryos also had a significant reduction of the trunk and posterior structures, with the somite number not greater than thirteen. Hindgut was never observed in these mutant embryos, and the pericardium was often dilated. The allantois in GCNF lox/lox embryos was often enlarged and was not always attached to the chorion, resulting in a lack of chorioallantoic development that probably contributed to the embryonic lethality. Another dramatic phenotype of GCNF Ϫ/Ϫ embryos was the presence of a large invagination of neural epithelium within the primitive streak 8.5 dpc (Fig. 3B). This large invagination was also observed in GCNF lox/lox embryos but not in GCNF ϩ/ϩ embryos (Fig. 3B).
Loss of Repression of Oct4 in Somatic Cells of GCNF lox/lox Embryos-Oct4, a member of the POU homeodomain family of transcription factors, plays an essential role in the maintenance of embryonic stem cell potency and the establishment of the germ cell lineage (39). It has been shown that the Oct4 gene is expressed in the epiblast of mouse embryo at 6.5-7.5 dpc and then down-regulated from the anterior to the posterior during gastrulation (19). A previous report has shown that GCNF can bind to the DR0 elements in the promoter of the Oct4 gene and repress the transcription of this gene and that the expression of Oct4 gene in large parts of the GCNF Ϫ/Ϫ embryos at 8.25-8.5 dpc is not silenced (19). To determine whether the DBD of GCNF is essential for mediating GCNF-dependent repression of Oct4 gene transcription in vivo, we analyzed the expression of Oct4 in GCNF lox/lox embryos using in situ hybridization assay. As shown in Fig. 4, Oct4 was expressed in the GCNF lox/lox embryos at 7.5 dpc, similar to the GCNF ϩ/ϩ and GCNF Ϫ/Ϫ . By 8.25 dpc, Oct4 expression was restricted to primordial germ cells in the GCNF ϩ/ϩ embryos. However, Oct4 derepressed expression domains were detected in the anterior and posterior of GCNF lox/lox and GCNF Ϫ/Ϫ embryos at 8.25 dpc (Fig. 4). These results suggest that the DBD of GCNF is required for GCNF to silence the expression of the Oct4 gene in somatic cells of mouse embryos.

DISCUSSION
Nuclear receptors such as ER, PR, AR, TR, and RAR are well known transcription factors that bind their cognate DNA-responsive elements through their DNA binding domains and then regulate the expression of their target genes via their transactivation domains, which ultimately elicits their distinct physiological functions in response to their cognate hormones (40,41). Recently, DNA binding-independent functions of nuclear receptors such as GR (33) and TR3 (34) have been described. As a member of the nuclear receptor superfamily, GCNF has been shown to bind to DR0 elements and repress gene transcription. Similar to other orphan nuclear receptors, such as SF-1 (41) and TR3 (34), the DBD of GCNF is required for its DNA binding activity as deletion of 55 amino acid residues at the N terminus of Xenopus GCNF (14) or of 163 amino acid residues at the N terminus of mouse GCNF 3 abolishes its DNA binding activity. Replacement of the GCNF DBD-encoding exon by a PGK-neo cassette in the mouse germline using a conventional knockout strategy causes embryonic lethality at 9.5-10.5 dpc (24). These mouse embryos have defects in the correct formation of somites and anteroposterior axis postgastrulation. In addition, unsilenced expression of the Oct4 gene in somatic cells is observed in these mutant embryos at 8.25 dpc. Even though the DBD region of the GCNF mRNA is completely deleted in these mutant mice (24), it is still unclear whether the phenotypes observed previously are due to the loss of GCNF or to the insertion of the PGK-neo cassette in the GCNF allele. Several studies in other laboratories have suggested that the introduction of PGK-neo cassettes can interfere with RNA processing of targeted transcripts (even when inserted in an intron) (28) and cause position effects on neighboring genes, particularly when inserted within gene clusters and locus control regions (29 -31) leading to unexpected phenotypes in the resulting animal models. Additionally, bi-directional transcriptional activities of PGK-neo, normal sense PGK promoter activity, and aberrant antisense promoter activity in neomycin and PGK promoter regions, which drive the production of aberrant transcripts, have been reported in cultured cells (42) and in vivo in mice (32). Considering that the SF-1 gene, encoding another nuclear receptor, is immediately adjacent to the GCNF gene in the mouse genome and is known to be involved in adrenal and gonadal development (43), it is possible 3 P. Gu and A. J. Cooney, unpublished data. that insertion of the PGK-neo in the GCNF locus affects SF-1 expression, complicating the interpretation of phenotypes observed in our conventional knockout embryos. Therefore, generation of a mouse line without the PGK-neo cassette in the GCNF locus is necessary to determine the role of GCNF during embryonic development by comparison of the phenotypes of the conventional knockout embryos to those mice without the PGKneo insertion. In this study, we have successfully generated GCNF lox/lox mice that do not contain the DBD-encoding exon of the GCNF gene nor the selection marker gene, PGK-neo, using Cre/loxP technology. Similarly to the conventional GCNF knockout mice, GCNF lox/lox mice died in utero at 9.5-10.5 dpc with defects in posterior embryonic development and the formation of somites and the anteroposterior axis. Expression of the Oct4 gene in GCNF lox/lox embryos was not silenced in so-matic cells at 8.25 dpc. Therefore, GCNF lox/lox mice phenocopy the conventional GCNF knockout mice (19,24), indicating that the phenotypes observed in the conventional knockout embryos result from the ablation of GCNF, not from the insertion of the PGK-neo cassette into the GCNF locus, confirming that GCNF is essential for normal embryonic development.
As no RNA splicing acceptor consensus sequences, which are located upstream of the DBD-encoding exon of the GCNF gene, were deleted in our knockout strategy and no consensus RNA splicing acceptor elements or transcription termination signals were included in the loxP site and its surrounding DNA sequences, it was possible that transcription of the GCNF gene in GCNF lox/lox mice could be initiated and that normal RNA splicing can occur between exon 3 and exon 5 forming a truncated GCNF message, which encodes a protein that does not contain FIG. 2. The LBD of the GCNF gene is expressed in the GCNF lox/lox embryos. A, schematic representation of protein-coding regions of the wild type GCNF mRNA (GCNF ϩ/ϩ ) and GCNF mutant mRNA (GCNF lox/lox ). Numbers listed in the figure represent the amino acid residues of the wild type GCNF protein. Primer 5 and primer 6 recognize the DBD cDNA sequences derived from wild type GCNF mRNA, while primer 7 and primer 8 recognize the LBD cDNA sequences derived from the wild type GCNF mRNA or mutant GCNF mRNA. Localization of an antisense LBD cRNA probe and the epitope of an anti-GCNF polyclonal antibody (LBD pAb) are also included. B, loss of the DBD in the GCNF lox/lox embryos at 8.5 dpc by RT-PCR analysis and RNase protection assay. Total RNA from GCNF ϩ/ϩ , GCNF ϩ/lox , and GCNF lox/lox embryos at 8.5 dpc was reverse transcribed and then amplified by PCR using different sets of primers. Primers 5 and 6 were used to determine the presence of the DBD region in the GCNF mRNA, while primers 7 and 8 were used for detecting the LBD region of the wild type and mutant GCNF mRNA. Actin primers were used as positive controls for the RT-PCR assays. For RNase protection assay, 10 g of total RNA from GCNF ϩ/ϩ , GCNF ϩ/lox , or GCNF lox/lox embryos at 8.5 dpc were hybridized with radiolabeled antisense GCNF LBD and GAPDH cRNA probes. A 260-bp GCNF LBD and a 110-bp glyceraldehyde-3-phosphate dehydrogenase band were observed in mouse embryo RNA samples. Yeast tRNA was used as a negative control for the assay. C, Western blot analysis showing the presence of a truncated GCNF protein (50 kDa) lacking the DBD in the GCNF lox/lox embryo. Protein extracts from GCNF ϩ/ϩ , GCNF ϩ/lox , and GCNF lox/lox embryos at 8.0 dpc or GCNF Ϫ/Ϫ embryos at 9.0 dpc, and from Cos-1 cells transfected with a CMVGCNF expression plasmid (CMVGCNF) or with an empty CMV vector (Cos-1) were immunoblotted with a specific antibody against a 19-amino acid peptide at the C terminus of the GCNF protein. A 58-kDa protein band represents the wild type GCNF protein, while a 50-kDa protein band represents the truncated GCNF protein, which lacks the 81-amino acid residues of the DBD. the DBD but does have an intact LBD. Indeed, we found that the LBD region, but not the DBD region, of the GCNF mRNA was expressed in the GCNF lox/lox mice detected either by RT-PCR or by RNase protection assay (Fig. 2). Since the size of the DBD exon of the GCNF gene is 243 bp (9), deletion of this exon should not cause a reading frameshift in the downstream LBD. Using a specific GCNF polyclonal antibody raised against a 19-amino acid peptide at the C terminus, we found that a 58-kDa GCNF protein band was detected in the GCNF ϩ/ϩ and GCNF ϩ/lox mice but not in the conventional GCNF knockout mice (Fig. 2) and GCNF lox/lox mice (Fig. 2). Instead, a 50-kDa-truncated protein band was detected in the GCNF lox (GCNF ϩ/lox and GCNF lox/lox ) embryos (Fig. 2). These studies suggest that the truncated GCNF transcript in the GCNF lox mice is translated into a 50-kDa protein that contains an intact LBD but not a DBD. These results clearly indicate that we have successfully deleted the DBD of GCNF in vivo. Although the truncated GCNF protein lacking only the DBD was observed in the GCNF lox mice (Fig. 2), phenotypes of these GCNF lox mice (Figs. 3 and 4) were the same as those of conventional GCNF knockout mice (19,24), which did not express GCNF protein (Fig. 2). These results suggest that the DBD of GCNF is essential for mediating the function of GCNF during early embryonic development. The results also suggest that at this stage of embryonic development the LBD of GCNF does not mediate any functions independent of the DBD. It should be emphasized that GCNF can bind to the DR0 element in the promoter of the Oct4 gene and repress the transcription of the Oct4 gene in vitro (19). Loss of GCNF in the conventional GCNF knockout mice causes loss of repression of Oct4 expression in somatic cells (19). Similar results were obtained in the GCNF lox/lox mice. Considering that the LBD of GCNF does not bind to DR0 element, 3 we can conclude that GCNF is indeed a transcription factor that represses gene transcription in mouse embryos. In addition, the expressed GCNF LBD does not have a dominant negative effect as the GCNF ϩ/lox mice are normal. The lack of a dominant negative effect is probably due to an inability of the wild type GCNF and the GCNF LBD to heterodimerize.
In summary, we have generated a line of mice that expresses a truncated GCNF protein lacking only the DBD. These mice phenocopy the conventional GCNF knockout mice. This study clarifies that the phenotypes observed in the conventional GCNF knockout mice are due to the loss of GCNF, not to the insertion of the PGK-neo cassette into the GCNF locus, and that GCNF is essential for normal mouse embryonic development. More importantly, our results suggest that the DBD of GCNF is essential for the function of GCNF during embryonic development, that GCNF does not have DNA-binding independent activity as reported for other nuclear receptors such as GR and TR3 (33,34) during mouse embryogenesis, and that GCNF is indeed a transcription factor that represses gene transcription in vivo.