|Home | About | Journals | Submit | Contact Us | Français|
The dynamic embryonic expression of germ cell nuclear factor (GCNF), an orphan nuclear receptor, suggests that it may play an important role during early development. To determine the physiological role of GCNF, we have generated a targeted mutation of the GCNF gene in mice. Germ line mutation of the GCNF gene proves that the orphan nuclear receptor is essential for embryonic survival and normal development. GCNF−/− embryos cannot survive beyond 10.5 days postcoitum (dpc), probably due to cardiovascular failure. Prior to death, GCNF−/− embryos suffer significant defects in posterior development. Unlike GCNF+/+ embryos, GCNF−/− embryos do not turn and remain in a lordotic position, the majority of the neural tube remains open, and the hindgut fails to close. GCNF−/− embryos also suffer serious defects in trunk development, specifically in somitogenesis, which terminates by 8.75 dpc. The maximum number of somites in GCNF−/− embryos is 13 instead of 25 as in the GCNF+/+ embryos. Interestingly, the tailbud of GCNF−/− embryos develops ectopically outside the yolk sac. Indeed, alterations in expression of multiple marker genes were identified in the posterior of GCNF−/− embryos, including the primitive streak, the node, and the presomitic mesoderm. These results suggest that GCNF is required for maintenance of somitogenesis and posterior development and is essential for embryonic survival. These results suggest that GCNF regulates a novel and critical developmental pathway involved in normal anteroposterior development.
Germ cell nuclear factor (GCNF; also called NR6A1) is an orphan member of the nuclear receptor gene superfamily (12, 15). The nuclear receptor gene superfamily includes a group of ligand-dependent transcription factors that bind to steroids and other lipophilic molecules, such as retinoic acid, which function to regulate many types of differentiation, homeostasis, and developmental processes (11, 23, 38). In addition, several members of this superfamily are orphan receptors for which ligands have yet to be identified (23, 56). Homologs of mouse GCNF have been cloned from several other species, including human, Xenopus, and zebrafish (9, 30).
GCNF is an orphan nuclear receptor that specifically binds to a novel nuclear receptor response element known as a DR0 (5, 12, 16, 70), a direct repeat of the estrogen receptor half-site sequence (AGGTCA) with no nucleotides between the half-sites. Binding to this element appears to be evolutionarily conserved, since Xenopus GCNF has the same DNA binding specificity as mouse GCNF (30). Unlike many other orphan receptors, GCNF does not heterodimerize with the retinoid X receptor; rather, it binds to DR0 elements as homodimers (5, 8, 16, 70).
We and others have shown that GCNF is a transcriptional repressor (5, 16). Like thyroid hormone receptor and chicken ovalbumin upstream promoter transcription factor, this repressor function was localized to the ligand binding domain, which is transferable to the Gal4 DNA binding domain (DBD) (4, 16–18); thus, in the absence of a ligand, GCNF represses target genes. Among the candidate GCNF-responsive genes that have been identified are two mouse genes, protamine 1 and protamine 2, which contain DR0 elements within 400 bp of the transcriptional start site.
GCNF was initially described to be predominantly expressed in germ cells of the adult mouse (12, 26, 34, 74) and human (1, 31, 35); further analysis revealed expression in embryonic carcinoma cells as well (24, 35). In the mouse, GCNF expression is detected in the embryonic ectoderm at 6.5 days postcoitum (dpc) by section in situ hybridization (62); after gastrulation at 7.5 dpc, it is expressed in all three germ layers. GCNF expression is strong in the developing nervous system by 8.5 dpc but decreases significantly after 9.5 dpc. Studies of Xenopus embryos have shown that the GCNF gene is expressed between the gastrula and mid-neurula stages (19, 30, 57). Because of these significant patterns of expression during gastrula and neurula stages of embryonic development and in specific organs of the adult, GCNF may regulate multiple developmental processes, particularly during embryonic development. Indeed, using a dominant negative GCNF transcript, normal frog embryonic development was disrupted (19, 30, 57). GCNF has also been shown to be involved in midbrain-hindbrain development in Xenopus embryos (19, 30, 57).
Recently, the targeted mutation of nuclear receptor genes in embryonic stem (ES) cells has demonstrated that many orphan receptors are essential for embryonic development (13, 27, 32, 33, 36, 37, 42, 47, 48, 51, 61, 74). To date, little is known of the role of the nuclear receptors during the postgastrulation and neurulation stages. Based on its embryonic expression pattern, GCNF is likely to play a key regulatory role during this critical developmental period.
In this paper we describe the dynamic expression of GCNF in postgastrulation and neurulation mouse embryos, showing that it is regulated differently from the Xenopus GCNF expression. In addition, we describe the results of a germ line mutation of the GCNF gene. Interestingly, the GCNF null mutation results in embryonic lethality, probably due to cardiovascular complications. Prior to death, the GCNF−/− embryos stall in development, with open neural tubes, failure to turn, and an absence of posterior ventral structures. GCNF−/− embryos also show a halt in somitogenesis after 13 somites, leading to a posterior truncation. Interestingly, the tailbud and the posterior of the embryo develop ectopically outside the yolk sac, a phenotype not previously reported for any germ line mutation. We propose a mechanism to account for the ectopic development of the tailbud. The posterior defects and failure to turn may be explained in part by the deregulation of genes required for normal somitogenesis and mesodermal development. These data suggest that the GCNF signaling pathway is required during postgastrulation and neurulation stages of mouse development.
Genomic clones containing GCNF sequences were obtained from a mouse genomic library (129Sv) in the Lambda FixII vector (Stratagene) by using GCNF cDNA as a probe (data not shown). The targeting vector was constructed from two clones that overlapped with the exon encoding the DBD. A 4-kb ApaI/EcoRV fragment of the 5′-most clone upstream of the DBD exon was subcloned into pBluescript KS (Stratagene). This plasmid was ligated with the 3-kb ApaI/Acc65I fragment of a clone 3′ downstream of the DBD exon. After partial ApaI digestion, the resulting vector was linearized between the two fragments, and ApaI/XhoI adapters were ligated in place. The herpes simplex virus (HSV) tk gene from pSP72 (Promega), lacking all sites of the polylinker except BamHI, was created by digestion with XhoI, followed by filling-in of the ends using the Klenow fragment of DNA polymerase I and further digestion with BamHI and calf intestinal phosphatase treatment. The two homologous arms, cut with BamHI and EcoRV, were then ligated into pSP72 containing HSV tk (39), thus placing the HSV tk gene downstream of the 3′ arm. The resulting vector was cut with XhoI, treated with calf intestinal phosphatase, and ligated with the 1.6-kb XhoI fragment of PGKneo (58), thus adding the neo (neomycin resistance) gene between the two homologous sequences. Restriction analysis was used to determine the orientation in which the neo gene was inserted.
A clone with the neo gene in the same orientation as the homologous sequence was selected for targeting. This plasmid was linearized at a unique XmnI site in the ampicillin resistance gene and electroporated (Gene Pulsar; Bio-Rad) into AB 1.2 ES cells (41) (see Fig. Fig.2A).2A). ES cells were maintained on STO fibroblasts to retain their undifferentiated phenotype (49). G418 and FIAU [1-(2′-deoxy-2′-fluoro-1-β-d-arabinofuranosyl-5-iodo)uracil] (59, 60) drug selection began 2 days after electroporation and continued for approximately 8 days, when 768 ES cell colonies were picked and grown individually in 96-well plates.
The colonies were screened by Southern analysis after digestion with enzymes Bsp106I and Acc65I, which produced a 21-kb band from the wild-type allele and a 9-kb band for the targeted allele when a 700-bp SacI/AccI fragment, derived from sequences 5′ of the homologous arms, was used as a probe (see Fig. Fig.2B).2B). Additionally, the neo gene was used with the same restriction enzyme digestion strategy to confirm that the correct recombination event had occurred and there was only a single insertion site. In this case, the targeted allele produced a 5.5-kb band. Hybridizations and washes were performed as previously described (34).
ES cell screening revealed six correctly recombined clones, which were expanded and injected into C57BL/6 blastocysts to produce chimeric mice. Four clones produced chimeric males that demonstrated germ line transmission of the targeted allele. The embryos analyzed are the F2 and F3 progeny from a mixed 129/C57 genetic background. Similar phenotypes were obtained with homozygous mutant embryos from the F1 generation of mice in an inbred 129Sv background.
DNA was extracted from tail samples, tissues samples scraped from serial sections, or embryos. Genotypes of weaned mice or embryos were determined by Southern analysis using the strategy described above or by PCR analysis of DNA samples. Two separate sets of primers were used. The primer set for the mutant allele (5′TCGATGCGATGTTTCGCTT3′ and 5′ATATGGGATCGGCCATTGA3′) was derived from the sequence in the neo gene; the primer set for the wild-type allele (5′CAGTGCTGACTTATCCATG3′ and 5′TTCCTGTTCATGCCCATCT3′) was from sequences within the DBD exon deleted in the mutated allele. PCR of the wild-type allele produced a band of 239 bp, while PCR of the mutant allele produced a band of 416 bp. In each of 30 cycles of PCR, DNA was denatured at 95°C for 3 min, primers were annealed at 58°C for 1 min, and extension was carried on at 72°C for 1 min.
Section in situ hybridization of normal GCNF was performed with a 33P-labeled riboprobe of either the full-length GCNF cDNA or the DBD cDNA as described previously (34). Probes used for whole-mount in situ hybridization (48) were cRNA probes for brachyury T (69), HNF3β (2), Hoxb-13 (72), lunatic fringe (29), mCer-1 (7), myogenin (54), nodal (14), Otx2 (40), paraxis (10), RALDH-2 (44), and Wnt-3a (63). For histological analysis, the embryos were embedded in 3% agarose, dehydrated, processed for paraffin embedding, sectioned at 7 μm, and stained with hematoxylin and eosin.
Using whole-mount in situ hybridization, we examined expression of the GCNF gene in mouse embryos between 7.5 and 10.5 dpc. A 3′ untranslated fragment of the murine GCNF cDNA was used to generate the antisense probe to ensure specificity and lack of cross-reaction with other members of the nuclear receptor superfamily. GCNF transcripts were detected as early as 7.5 dpc in the anterior neuroepithelium of the head fold and throughout the primitive streak in the posterior (Fig. (Fig.1A).1A). At 8.5 dpc, GCNF expression was stronger in the posterior of the embryo than in the anterior (Fig. (Fig.1B).1B). This result is consistent with the pattern of expression found by Susens et al. for tissue sections (62). Section in situ hybridization analysis also showed that GCNF expression became significant in the posterior proliferating neuroepithelium and in the underlying mesoderm of the primitive streak by 8.5 dpc (62). GCNF expression was up-regulated again in the anterior at 8.75 dpc (Fig. (Fig.1C)1C) and down-regulated in the neuroepithelium by 9.5 dpc. GCNF expression remained abundant in the anterior but significantly decreased in the posterior (Fig. (Fig.1D).1D). Finally at 10.5 dpc, GCNF expression was markedly reduced, as determined by either whole-mount or section in situ (not shown). Thus, the dynamic expression of GCNF during the postgastrulation and neurulation stages suggests that GCNF may be involved in several different processes during early embryonic development.
To determine the physiological role of GCNF in the mouse, a targeted mutation of the GCNF gene was undertaken using homologous recombination in ES cells. Genomic clones from the GCNF locus were isolated from a mouse 129Sv library by hybridization with the full-length mouse GCNF cDNA. The initial screen produced three partial but overlapping clones that covered approximately 40 kb of genomic sequence (Fig. (Fig.2A).2A). The 5′-most cDNA sequence in these genomic clones was that of the DBD exon. Unlike most other nuclear receptor genes, both zinc fingers are encoded in a single exon.
To construct the targeting vector (see Materials and Methods), sequences surrounding the zinc finger exon were subcloned to serve as the homologous upstream (4 kb) and downstream (3 kb) arms (Fig. (Fig.2A).2A). The six positive ES cell clones identified (targeting efficiency of 1 to 2%) were injected into C57BL/6 blastocysts, and two produced male chimeric mice that transmitted the mutant allele to their offspring. The heterozygous offspring of chimeric males and C57BL/6 females were mated to produce mice homozygous for the mutant GCNF allele. At weaning, these progeny were screened to determine their genotype by PCR analysis (Fig. (Fig.2C).2C). None of over 500 mice screened were homozygous for the mutant allele (Table (Table1),1), indicating that the absence of a wild-type GCNF allele caused neonatal or perinatal lethality.
To precisely determine when homozygous GCNF mutants die, heterozygous females from timed matings with heterozygous males were sacrificed at various days postcoitum, and the embryos were genotyped by PCR analyses. At 7.5, 8.5, and 9.5 dpc, the ratios of wild-type, heterozygous, and homozygous mutant embryos were similar to the expected Mendelian ratios (Table (Table1).1). By 10.5 dpc, though, there were fewer homozygous mutants than predicted. At 11.5 dpc, the homozygous mutant tissues that could be genotyped were remnants of resorbing embryos. Therefore, lethality due to the lack of an intact GCNF gene occurred at midgestation, around 10.5 dpc.
To ensure that embryonic lethality was due to the targeted mutation of GCNF leading to a loss of gene expression, we evaluated expression at 7.5 dpc by section in situ hybridization using an antisense probe to the DBD. GCNF was strongly expressed in all three germ layers of the wild-type epiblast of GCNF+/+ embryos (Fig. (Fig.3A),3A), as well as in the extraembryonic tissue of the ectoplacental cone. GCNF−/− embryos, as determined by PCR of tissue scraped from serial sections, showed no positive hybridization signal, indicating that the GCNF gene had been functionally disrupted (Fig. (Fig.3B).3B).
Unlike wild-type embryos at 10.5 dpc (Fig. (Fig.3C),3C), the GCNF−/− embryos were severely malformed, showing trunk and posterior defects, open neural tubes, failure of axis rotation and hindgut closure, and distended pericardia (Fig. (Fig.3D).3D). The GCNF−/− embryos died probably as a result of cardiovascular distress resulting from the defects in trunk and posterior development. Similar phenotypes were observed in homozygous GCNF−/− embryos in two independent lines of founder mice.
At 8.25 dpc, no gross morphologic differences between GCNF+/+ and GCNF−/− embryos were evident (data not shown); by 8.5 dpc, however, GCNF−/− embryos were distinguishable from GCNF+/+ embryos by several morphologic criteria. One difference was the curvature of the spine within the yolk sac, which was much wider in GCNF−/− embryos than in GCNF+/+ embryos at 8.5 dpc (Fig. (Fig.4A4A and B). In addition, the allantois, which was often enlarged, was not always attached to the chorion. The lack of proper chorioallantoic development probably also contributed to embryonic lethality. Also at 8.5 dpc, a small protrusion of tissue (Fig. (Fig.4B4B and D) was seen at the base of the allantois of GCNF−/− embryos, which continued to develop outside the yolk sac (Fig. (Fig.4F4F and J). Unlike GCNF+/+ embryos that turned (Fig. (Fig.4G4G and K), GCNF−/− embryos never turned and remained in a lordotic position with open neural tubes (Fig. (Fig.4H4H and L).
At 9.5 dpc the GCNF−/− embryos were clearly developmentally malformed, with a significant reduction of trunk and posterior structures. No more than 13 somites were observed in GCNF−/− embryos, instead of the 20 to 25 somites present in GCNF+/+ embryos at 9.5 dpc (Fig. (Fig.4I4I to L). The posterior of the GCNF−/− embryos, including some somites, protruded outside the yolk sac. When 9.5-dpc GCNF−/− embryos were compared to 8.75-dpc GCNF+/+ embryos with a similar somite number, the continued growth of anterior neural tissue and tailbud was clear. The gross morphology of mutant embryos suggested that growth had continued beyond 8.5 dpc at the extreme anterior and posterior ends but not in the intervening trunk region. Thus, development of GCNF−/− embryos was not completely halted.
Analysis of sagittal sections of GCNF+/+ embryos showed the position of yolk sac attachment around the node (Fig. (Fig.5A);5A); however, the yolk sac was attached at the base of allantois in the GCNF−/− embryos at 8.5 dpc (Fig. (Fig.5B).5B). In addition, the neural epithelium formed a large invagination within the primitive streak, a phenomenon also seen in cross sections (Fig. (Fig.5B5B and D).
At 9.5 dpc the posterior neuropore was still open in GCNF+/+ embryos (Fig. (Fig.5E),5E), and the neural tube was closed. In addition, cross sections of 9.5-dpc GCNF+/+ embryos also showed the hindgut situated in the middle of the tailbud (Fig. (Fig.5G).5G). In contrast, no posterior neuropore was observed in the tailbuds of GCNF−/− embryos (Fig. (Fig.5F),5F), and the neural tube remained open (Fig. (Fig.4L).4L). Closure of the neural tube, however, was observed in the middle of the ectopic tailbud (Fig. (Fig.5H);5H); in addition, ventral structures such as hindgut and ventral body wall were poorly developed in the GCNF−/− embryos. It is important to note that the notochord was present throughout the length of the neural tube (Fig. (Fig.5H)5H) of GCNF−/− embryos. In summary, the tissues in the primitive streak and the GCNF−/− tailbud were histologically disorganized, suggesting that loss of GCNF function disrupted normal posterior development in the embryo.
To examine more closely the extent of the observed defects and to determine the molecular defects in GCNF−/− embryos, we probed mutant embryos with genes that serve as markers for different structures involved in development at this time. The first set of marker genes examined the gross development of the GCNF−/− embryos to determine which tissues were present.
First, the homeobox gene Otx2 was used to analyze anterior development in GCNF−/− embryos (55). Otx2 was strongly expressed in the anterior neural folds, with a sharp limit at the presumptive midbrain and hindbrain junction in GCNF+/+ and GCNF−/− embryos by 8.5 dpc (Fig. (Fig.6A).6A). This result implied that anterior neural development had been initiated in GCNF−/− embryos.
Since some posterior structures developed poorly in GCNF−/− embryos, HNF3β was used as a marker to analyze the development of the midline and ventral tissues. HNF3β is expressed in the node, the floor plate, the developing foregut, and the hindgut at 8.5 dpc (3, 53) and in the primordial liver at 9.5 dpc (3, 43) (Fig. (Fig.6B).6B). In GCNF−/− embryos, HNF3β expression was clearly seen in the midline, the foregut, and in the tailbud; however, there was no expression in posterior presumptive hindgut tissues. This suggests that there is a lack of hindgut development in GCNF−/− embryos. Interestingly, there is also a loss of anterior HNF3β expression, suggesting that the anterior neural development is not entirely normal.
As a marker of midgestation posterior development, we used Hoxb-13 because it is involved in patterning along the anteroposterior axis of the mouse (10, 72). Hoxb-13 is not expressed, though, until 9.0 dpc, after turning is complete. In GCNF+/+ 9.5-dpc embryos (Fig. (Fig.6C),6C), expression of Hoxb-13 extends anteriorly from the tailbud. Hoxb-13 was expressed in GCNF−/− embryos, even though the embryos had not turned (Fig. (Fig.6C).6C). Interestingly, this suggests that Hoxb-13 expression may be independent of turning. Expression of Hoxb-13 was more limited in the GCNF−/− embryos, though, possibly due to the lack of a hindgut and malformation of structures anterior to the tailbud. The presence of Hoxb-13 indicated that certain aspects of posterior development in the GCNF−/− embryos had progressed beyond 8.75 dpc, even though others, such as somitogenesis, had not.
We analyzed the expression of Wnt-3a, which is expressed in the tailbud, to ascertain that the extreme caudal portion of the GCNF−/− embryos is indeed the tailbud. A mutation of Wnt-3a produces mice that are truncated in tissues posterior to the forelimbs (63). In GCNF+/+ embryos, Wnt-3a was expressed in the most posterior region, the tailbud (Fig. (Fig.6D).6D). Interestingly, the GCNF−/− embryos at 9.5 dpc had Wnt-3a expression at the tip of the presumed tailbud (Fig. (Fig.6D),6D), consistent with the fact that the lack of GCNF did not result in a loss of tailbud development and truncation of all caudal tissues.
Since GCNF−/− embryos have a reduced number of somites, we examined development of the presomitic mesoderm (PSM). Analysis of the PSM marker brachyury T (68) on embryos within their yolk sacs showed the presence of PSM in the ectopic tailbud (Fig. (Fig.6E6E and F). This result is consistent with the development of some somites outside the yolk sac.
In summary, the expression patterns of these marker genes in the anterior and posterior of GCNF−/− embryos were similar to those for wild-type embryos, indicating that certain aspects of development of the mutant embryos had progressed beyond 8.75 dpc, even though others, such as somitogenesis, had stalled. The defects in the trunk of GCNF−/− embryos were not the result of a truncation of the entire posterior and might be due to molecular defects in the primitive streak, the tailbud, or the node.
At 9.5 dpc, the maximum number of somites in GCNF−/− embryos was 13, similar to the somite number of GCNF+/+ embryos at 8.75 dpc. This suggests that GCNF might affect somitogenesis after 8.75 dpc. Having demonstrated the presence of PSM in the GCNF−/− embryos (Fig. (Fig.6F),6F), we proceeded to determine the underlying molecular mechanism that resulted in a halt in somitogenesis in GCNF−/− embryos.
We first analyzed the extent of somite differentiation using myogenin, a late myotome muscle marker (54). In GCNF+/+ embryos, myogenin was expressed in all differentiating somites by 9.5 dpc (Fig. (Fig.7A).7A). In GCNF−/− embryos, myogenin was expressed in the anterior seven differentiating somites at 9.5 dpc but not in the posterior six somites (Fig. (Fig.7B).7B). Thus, although somite differentiation had initiated in the GCNF−/− embryos, it may have been delayed.
We next used paraxis as a marker gene for the formation and epithelialization of somites (10). At 8.5 dpc, we detected paraxis expression in both newly formed and preexisting somites of GCNF+/+ and GCNF−/− embryos (Fig. (Fig.7C7C and D). Thus, formation of these early anterior somites was not affected by the loss of GCNF expression. However, the intersomitic boundaries of paraxis expression were not well demarcated in GCNF−/− embryos as in the GCNF+/+ embryos. The failure to form well-defined epithelial somites in GCNF−/− embryos suggests that early somite formation may be affected. Surprisingly, examination of paraxis expression in GCNF−/− embryos at 9.5 dpc revealed a separate ectopic domain in the PSM, more posterior to its normal domain (Fig. (Fig.7I7I and J). This result suggested that there was an altered differentiation of the PSM within the ectopic tailbud, which failed to remain in an undifferentiated state.
Mouse Cerberus-related-1 (mCer-1), which belongs to the Cerberus/Dan-related gene family (46), was then used to analyze early somite formation. Expression of mCer-1 is evident at the onset of gastrulation in the anterior visceral endoderm (6, 7). From the early somite stage, mCer-1 expression was observed solely in the first and second newly formed somites of GCNF+/+ embryos (Fig. (Fig.7G),7G), as previously described (7). However, mCer-1 expression was greatly reduced in the newly formed somites in GCNF−/− embryos (Fig. (Fig.7H),7H), which suggests that the formation of new posterior somites is affected at 9.5 dpc.
Since lunatic fringe, a gene that participates in the Notch signaling pathway during segmentation and exhibits distinct dynamic expression during early embryonic development (21, 29), is required in the PSM for the formation of somites, we determined its expression in GCNF−/− embryos. At 9.5 dpc, lunatic fringe was expressed weakly in the first presumptive somitomere; however, it was strongly expressed in the second presumptive somitomere within the PSM of GCNF+/+ embryos (Fig. (Fig.7I).7I). In the posterior-most PSM, lunatic fringe was expressed in a broad swathe of cells. Unlike GCNF+/+ embryos, GCNF−/− embryos did not possess this broad posterior-most expression domain of lunatic fringe in the PSM (Fig. (Fig.7J).7J). In addition, the levels of lunatic fringe expression in the first and second presumptive somitomeres were often equal. These results suggest that there are molecular defects in the PSM present in GCNF−/− embryos which likely affect the formation of new somites.
Further evidence that differentiation of the PSM was altered in GCNF−/− embryos came from analysis of the expression of RALDH-2, an enzyme important for the generation of retinoic acid during early embryonic development (44). Interestingly, RALDH-2−/− embryos (45), like GCNF−/− embryos, have open neural tubes and posterior truncations. At 9.5 dpc, RALDH-2 was expressed within the cervical mesenchyme, newer and more caudal somites, and the cloacal region toward the base of the allantois in GCNF+/+ embryos (Fig. (Fig.7K).7K). In GCNF−/− embryos, RALDH-2 was expressed in the somites similarly to GCNF+/+ embryos, with highest levels detected in the anterior 12 somites (Fig. (Fig.7L).7L). However, an unexpected ectopic expression of RALDH-2 was also observed in GCNF−/− embryos at the tip of the tailbud, in a manner similar to the ectopic expression of paraxis at this stage.
In summary, GCNF is not required for the early initiation, epithelialization, and differentiation stages of somitogenesis since genes such as myogenin and paraxis were appropriately expressed up to 8.75 dpc. However, GCNF is required for the continued development of somites since in its absence, there was an altered differentiation of the PSM with ectopic expression of paraxis and RALDH-2 and a reduction of lunatic fringe and mCer-1. Inappropriate somite differentiation and altered potential of the PSM eventually led to a halt in somitogenesis.
Since the mouse node is also crucial for mesoderm formation and recruitment of cells required for somitogenesis, the next step was to determine whether the node was functioning appropriately in GCNF−/− embryos. The node, which is located at the anterior end of the primitive streak, is crucial for organizing and patterning the midline axis of the developing embryo (14). Expression of nodal, a transforming growth factor β family member, is required for mesoderm formation and axial rotation (14). In GCNF+/+ embryos, nodal was expressed around the node at the onset of gastrulation and during the early somite stages (Fig. (Fig.8A);8A); however, its expression was down-regulated after 9.0 dpc (Fig. (Fig.8C),8C), as expected (67). At 8.25 dpc, nodal was appropriately expressed at the node in GCNF−/− embryos (Fig. (Fig.8B);8B); however, its expression was not down-regulated even by 9.5 dpc (Fig. (Fig.8D).8D). This persistent and disorganized expression of nodal in the node is consistent with the disorganization and altered differentiation of the PSM that resulted in the halt in somite formation at later somite stages. This suggests that the loss of GCNF function and ectopic development of the tailbud led to a disruption of the normal function of the node.
Since the development and differentiation of the node and PSM were altered, the next question was whether the primitive streak was normal in GCNF−/− embryos. Brachyury T expression is found in the primitive streak at 7.0 to 7.5 dpc and persists in the PSM as long as new somitic mesoderm is being produced (68). At 8.5 dpc, brachyury T was expressed in the node and primitive streak in GCNF+/+ embryos (Fig. (Fig.8E)8E) and in GCNF−/− embryos (Fig. (Fig.8F);8F); however, the expression domain was broader in the PSM of GCNF−/− embryos than in the PSM of GCNF+/+ embryos. At 8.75 dpc, the brachyury T expression domain was significantly different in GCNF−/− embryos compared to GCNF+/+ embryos, although it was still expressed in the node region (Fig. (Fig.8H).8H). Brachyury T was expressed in a V-shaped domain in GCNF−/− embryos, compared to the linear domain in GCNF+/+ embryos (Fig. (Fig.8H).8H). At 9.5 dpc, the expression of brachyury T persisted in the tailbud of GCNF−/− embryos (Fig. (Fig.8J)8J) in a more exaggerated V shape, while it was down-regulated in the PSM of GCNF+/+ embryos (Fig. (Fig.8I).8I). The persistent, disorganized brachyury T expression in the tailbud of GCNF−/− embryos further suggests that loss of GCNF function resulted in an altered differentiation of the PSM.
Germ line mutation of the GCNF gene proves that the encoded orphan nuclear receptor is essential for embryonic survival and normal development. GCNF−/− embryos cannot survive beyond 10.5 dpc, probably due to cardiovascular failure. Prior to death, GCNF−/− embryos suffer significant defects in posterior development, which are probably due to the loss of GCNF expression within cells of the posterior. The defects observed are not due to the neo cassette, as a GCNF knockout line with the neo cassette deleted phenocopies the GCNF knockout described here (data not shown). Defects in GCNF−/− embryos may be related to the retardation and/or halt in somitogenesis after 13 somites in mutant embryos and the abnormal positioning of the tailbud. Indeed, alterations in expression of multiple marker genes were identified in the posterior of GCNF−/− embryos, including the primitive streak, node, and PSM. These results suggest that GCNF is required for proper anteroposterior development, somite formation, and posterior development and that it is essential for embryonic survival.
The results of this study highlight some interesting similarities and differences in how GCNF functions in amphibians and mammals (19, 30). First, GCNF is important for both amphibian and mammalian embryonic survival. In Xenopus, the expression of a dominant-negative form of GCNF results in malformations of the head, eyes, ears, cement gland, and somites, resulting in embryonic lethality (19). Second, when wild-type GCNF is overexpressed in Xenopus embryos, it causes disturbances in somitogenesis and tail formation, with normal head development (19). Finally, failure of neural tube closure was also observed in Xenopus embryos after expression of the dominant-negative GCNF (19). Together with the results from our study, the findings indicate that the presence of GCNF is critical for normal posterior development, somitogenesis, and neural tube closure. Unlike the results of our study, expression of the dominant-negative form of GCNF in more posterior regions of Xenopus had no significant effect on posterior development. Alternatively, severe posterior defects in Xenopus may have been overlooked, as only embryos exhibiting less severe phenotypes were analyzed (19).
The results presented here suggest that the lack of trunk development in GCNF−/− embryos is distinct from the effects of other genes that have previously been shown to cause truncations of caudal structures. For example, brachyury T−/− embryos have little or no notochord and have somites only up to the region of forelimb buds, forming no trunk or tail (22, 71). The posterior truncation is due to a failure to form mesoderm in T−/− embryos. Inactivation of the HNF3β gene has demonstrated effects on notochord development (2, 20). Although the expression pattern of HNF3β in GCNF−/− embryos is somewhat different from the wild-type pattern, HNF3β is expressed in the midline structures, such as the notochord and floor plate, of GCNF−/− embryos. When Wnt-3a was mutated (63), the resulting truncation occurred caudal to the forelimb buds. Wnt-3a was expressed in the ectopic tailbud of GCNF−/− embryos, although forelimbs did not develop. In contrast to Wnt3a−/− embryos, which do not express brachyury T, GCNF−/− embryos do express brachyury T in their tailbuds, indicating that these embryos possess a tailbud structure similar to that of wild-type embryos. In addition, the development of the GCNF−/− tailbud progresses to the expression of later-stage markers such as Hoxb-13. Hoxb-13 is not expressed in the normal tailbud until 9.0 dpc; thus, its appearance in GCNF−/− embryos implies that the tailbud not only is present but has also sustained some degree of normal development. Other mouse models, such as paraxis, lunatic fringe, and Mesp2, have less severe posterior defects (10, 52, 73). Taken together, these data suggest that GCNF may regulate a unique signaling pathway distinct from the brachyury T, HNF3β, Wnt-3a, and Hoxb-13 pathways.
The initiation and differentiation of the first seven to nine somites are not affected by the absence of GCNF. Previous fate mapping studies suggest that by the late streak stage, the bulk of prospective cranial mesoderm has already exited the streak (64, 66). The cells in the anterior segment of the primitive streak at this stage are mainly destined for the first 6 to 10 somites and the PSM of the early-somite-stage embryo, a finding which was confirmed by ablation experiments (65). Thus, GCNF may not be required for the initiation of primary body formation or participation in somite formation during the early stages of development. The expression of paraxis at 8.5 dpc in these somites reinforces this speculation. A transition in cellular recruitment to the paraxial mesoderm from the primitive streak to the tailbud mesenchyme occurs at the stage of posterior neuropore closure at the end of the neural stage (65, 66). Since GCNF may be required in the PSM and/or tailbud, somitogenesis is affected after this stage in GCNF−/− embryos. Ectopic expression of paraxis and RALDH-2 in the posterior of GCNF−/− embryos indicates there is inappropriate differentiation within the PSM, ultimately leading to the halt in somitogenesis and the posterior truncation. The altered differentiation of the PSM is probably due to the loss of GCNF within cells of the posterior; however, the ectopic location of the tailbud also likely contributes to the phenotype.
The tailbud develops ectopically outside the yolk sac after the late streak stage, which could be due to the failure of the embryos to turn and envelop correctly in their embryonic membranes. The failure of mutant embryos to turn has been described for many gene knockout models, yet the positioning of the tailbud outside the yolk sac suggests that this is a novel phenotype related to the loss of GCNF function and not merely due to a failure of the GCNF mutant embryos to turn (25, 28, 45, 50). The halt in somitogenesis does not occur until after the posterior of the GCNF−/− embryo develops outside the yolk sac. Therefore, in addition to the loss of GCNF function contributing to the halt in somitogenesis and the posterior truncation, the ectopic location of the tailbud may play a role as well.
The exact mechanism that results in the formation of this protrusion remains to be investigated. Here we propose a model that accounts for its formation based on the data presented here (Fig. (Fig.9).9). In both GCNF+/+ and GCNF−/− embryos, as the somites, which express paraxis, are generated from the primitive streak, the neural plate forms at 8.0 dpc (Fig. (Fig.9A9A and B). As GCNF+/+ embryos continue to grow at 8.5 dpc, the node, which expresses nodal, regresses caudally (Fig. (Fig.9C),9C), with a resulting lengthening of the notochord that continues to induce the neural plate to form the neural folds. The edges of the neural plate begin to elevate, forming the neural groove. In GCNF−/− embryos, a neural groove has been observed histologically, but neural epithelium abnormally invaginates within the primitive streak region by 8.5 dpc (Fig. (Fig.9D;9D; Fig. Fig.5B5B and D). This invagination then pushes the primitive streak away from the original direction of posterior regression.
A factor contributing to ectopic development of the tailbud is the altered location of the posterior site of attachment of the yolk sac. The yolk sac is normally attached around the node, anterior to the primitive streak in GCNF+/+ embryos (Fig. (Fig.9C).9C). In GCNF−/− embryos, however, the yolk sac is attached more caudally, at the posterior of the primitive streak close to the allantois (Fig. (Fig.9D).9D). Consequently, the resulting PSM and undifferentiated somites are pushed out of the yolk sac by the invagination of the neuroepithelium and growth of the posterior. At 9.5 dpc, the posterior neuropore in GCNF+/+ embryos is closed by “zipping up” the neural groove caudally (Fig. (Fig.9E).9E). In contrast, no posterior neuropore was observed in GCNF−/− embryos. In addition nodal expression is down-regulated in wild-type embryos at 9.5 dpc (Fig. (Fig.9C9C and E); however, the disorganized and persistent expression of nodal (Fig. (Fig.9F)9F) is indicative of altered function of the node, the posterior organizer, which in turn may lead to the abnormalities observed in the posterior of GCNF−/− embryos. The neural tube elongates within the tailbud as it pushes through the yolk sac (Fig. (Fig.9F)9F) and forms a new posterior that continues to produce somites until the abnormal differentiation observed in the PSM, i.e., ectopic expression of paraxis, eventually leads to a halt in somitogenesis.
In summary, inactivation of the GCNF gene leads to embryonic lethality with major disruption of normal anteroposterior development. GCNF is likely to be a receptor for a novel ligand signaling pathway that is involved in regulating various aspects of embryonic development and normal anteroposterior development. Further work will be required to elucidate the GCNF signaling pathway.
The first two authors contributed equally to this work.
We thank R. R. Behringer, K. Mahon, and R. L. Johnson for many helpful suggestions on the manuscript. We also thank A. Bradley for providing the HSVtk plasmid and S. Aizawa, D. L. Ang, R. Beddington, M. Buckingham, P. Chambon, R. P. Harvey, N. Heintz, A. P. McMahon, R. Nusse, E. N. Olson, E. J. Robertson, and T. F. Vogt for providing plasmids to generate RNA probes for in situ hybridization.
This work was supported by NIH grant DK57743 to A.J.C.