Depletion of definitive gut endoderm in Sox17-null mutant mice

In the mouse, the definitive endoderm is recruited from the epiblast during gastrulation; following ingression through the primitive streak displaces the visceral endoderm to the extra-embryonic yolk sac (Lawson and Pedersen, 1987; Tam and Beddington, 1992). By early organogenesis, the definitive endoderm forms the epithelium of the primitive gut tube from which the digestive tract, liver, pancreas and associated visceral organs develop (Wells and Melton, 1999). In contrast to the wealth of information on the developmental fate and movement of the endodermal progenitors and the morphogenesis of their derivatives, little is known of the genetic determinants that regulate the specification and differentiation of the definitive endoderm. Mutations that cause the loss of function of genes such as Fgf8 and Fgfr1 results in defects in the migration and formation of not only the endoderm but also the mesoderm (Deng et al., 1994; Yamaguchi et al., 1994; Sun et al., 1999). Gata4 which encodes a zinc-finger transcription factor in the visceral endoderm and foregut endoderm is found to be required for gut closure and heart morphogenesis through the activity of other downstream genes (Molkentin et al., 1997; Narita et al., 1997). The loss of Hnf3β (Foxa2 – Mouse Genome Informatics) activity has resulted in the loss of not only the endoderm of the fore- and midgut as well as other embryonic structures such as the notochord (Ang et al., 1993; Sasaki and Hogan, 1993; Dufort et al., 1998). Mutations of the nodal signalling factors such as Smad2 (Madh2 – Mouse Genome Informatics) (Tremblay et al., 2000) and Fast1 (acts upstream of Hnf3β) (Hoodless et al., 2001) lead to a deficiency of definitive endoderm. However, like other mutants, the loss of definitive endoderm is accompanied by severe disruption of the differentiation of other embryonic tissues, which hampers the interpretation of the exact functional requirement for the endodermal lineages.

The Sry-related HMG box gene, Sox17, belongs to the Sox-subgroup F with Sox7 (Taniguchi et al., 1999) and Sox18 (Dunn et al., 1995). Sox17 was originally identified as a stage-specific transcription activator during mouse spermatogenesis (Kanai et al., 1996). A detailed analysis of the expression of this gene during embryogenesis reveals an earlier phase of activity in the visceral and definitive endoderm of the post-implantation embryo (this study). To date, about 30 Sox genes are identified in various vertebrate species (Pevny and Lovell-Badge, 1997; Wegner, 1999; Bowles et al., 2000). Members of this gene family encode transcription factors that regulate the specification of cell types and tissue differentiation, such as the differentiation of the Sertoli cells and the formation of the testis [Sry (Koopman et al., 1991); Sox9 (Foster et al., 1994; Wagner et al., 1994; Bishop et al., 2000; Vidal et al., 2001)], lens formation [Sox1 (Kamachi et al., 1998; Nishiguchi et al., 1998)], pro-B-lymphocyte proliferation [Sox4 (Schilham et al., 1996)], the differentiation of the neural crest cells that are progenitors of the enteric neurones [Sox10 (Pingault et al., 1998; Southard-Smith et al., 1998)] and the differentiation of the chondrocytes that form the cartilage [Sox5 and Sox6 (Smits et al., 2001), Sox9 (Foster et al., 1994; Wagner et al., 1994; Bi et al., 1999)].

Consistent with the general role of Sox genes in lineage differentiation, Sox17 orthologues in Xenopus (Xsox17α, Xsox17β) (Hudson et al., 1997; Clements and Woodland, 2000) and zebrafish (Zsox17) (Alexander and Stainier, 1999) are expressed specifically in the endoderm during gastrulation and play a key role in endoderm formation. The expression of a dominant negative Xsox17 construct, containing the repressor domain of Drosophila engrailed, suppresses endoderm differentiation of vegetal blastomeres, while over-expressing Xsox17, in conjunction with the upstream homeobox mixer gene, activates the expression of endoderm-specific genes in animal cap tissues (Hudson et al., 1997; Henry and Melton, 1998). In the zebrafish, casanova, a novel member of Sox subgroup F that acts upstream of Zsox17 and mediated by nodal signalling activity, can induce endodermal fate in a cell-autonomous manner, thus indicating a similar role for the Sox17 orthologue (Dickmeis et al., 2001; Kikuchi et al., 2001; Sakaguchi et al., 2001; Aoki et al., 2002).

To elucidate the function of the mouse Sox17 gene, the phenotype of the Sox17-null mutant embryo was examined during post-implantation development. Mutant embryos are found to be deficient of the definitive endoderm lining the embryonic gut. Sox17-null ES cells are unable to compete with the wild-type cells in colonising the gut endoderm of the chimeras. Sox17 therefore plays a crucial role in the differentiation of the definitive endoderm in the mouse, which is consistent with the concept of evolutionary conservation of Sox17 function in endodermal development in vertebrates.

We used the mouse Sox17 genomic DNA clones which were previously isolated from a 129/SvJ genomic library (Kanai et al., 1996). The 1.4 kb EcoRI fragment corresponding to the fourth exon, including C-terminal 53 amino acids and 3′-UTR region of Sox17, was inserted into the EcoRI site between neo and TK cassettes in pKSloxPNT targeting vector. A 5.4 kb EcoRI-ClaI fragment, which includes the first exon of the Sox17 5′-UTR region was inserted downstream of a neo cassette. The targeting vector linearised by SalI was electroporated into a 129/Sv W9.5 ES cell line as described previously (Tajima et al., 1998). After selection with 350 μg/ml G-418 in ES cell media, 12 Sox17^+/– ES clones were recovered from 182 G418-resistant colonies analysed, and germline chimeras were generated from two correctly targeted ES lines (line numbers 2-7, 32-4). As mutant embryos of both lines exhibited the same phenotype, the findings of the 2-7 line and 32-4 line are pooled.

We isolated the Sox17^–/– ES clone from the 2-7 line by selection with 1 mg/ml G-418 in ES cell media (right panel in Fig. 2B), and generated chimaeric embryos by blastocyst injection into ROSA26 (Sox17^+/+) mice (Jackson Labs). Embryos were fixed with 1% PFA/0.2% glutaraldehyde/0.02% NP40-PBS and subjected to whole-mount X-gal staining. After the whole-mount samples were photographed under the dissecting microscope, transverse frozen sections were prepared for histological analysis.

Genomic DNA of ES cells and the tail tip of mice was isolated using a Wizard genomic DNA purification kit (Promega, WI). For embryo DNA, part of the yolk sac or allantois was dissected from each embryo, and genomic DNA was prepared using the same kit. To determine the genotype by Southern blot analysis, DNAs were digested with BamHI, electrophoresed and transferred to a nylon membrane. The blots were hybridised with the ³²P-labeled probe of the 1.5 kb ClaI-BamHI fragment corresponding to the first exon (its position is indicated by a green bold bar in Fig. 2A). To analyse the genotype of the yolk sac or allantois by PCR, the wild-type Sox17 DNA (235 bp) and the mutated allele DNA (275 bp) were amplified using the following primers: Sox17 forward, 5′-CTC TGC CCT GCC GGG ATG GCA CGG AAT CC-3′; Sox17 reverse, 5′-AAT GTC GGG GTA GTT GCA ATA GTA GAC CGC TGA-3′; the target allele forward (pgk promoter reverse), 5′-GCA GGG GCC CTC GAT ATC AAG CTT GGC TG-3′ (their positions are indicated by red arrowheads in Fig. 2A).

Embryos from a heterozygous cross were dissected at 8.0 and 8.5 days post coitum (dpc) and incubated in Dulbecco’s modified Eagle medium containing 10% BSA and horseradish peroxidase (HRP; Sigma type IV, 2 mg/ml) for 30 minutes. After separation of the allantois or a part of the visceral endoderm for genome typing, the embryos were fixed and developed by DAB reaction.

Whole-mount and section in situ hybridisation were performed essentially as described previously (Gad et al., 1997; Kanai-Azuma et al., 1999). RNA probes for Sox17 (Kanai et al., 1996), Sox7 (Taniguchi et al., 1999), Shh, Ihh (Echelard et al., 1993), Ptch (Motoyama et al., 1998), Gata4 (Molkentin et al., 1997), Hnf3α (Foxa1 – Mouse Genome Informatics) (Sasaki and Hogan, 1993), Hnf3β (Ang and Rossant, 1994), Hex (Keng et al., 2000), Hnf4 (Chen et al., 1994), Afp (Constam and Robertson, 2000) (Accession Number, AA1172792), Cer1 (Shawlot et al., 1998), Cdx2 (Beck et al., 1995) (Accession Number, AI327188) and Pdx1 (Guz et al., 1995) were used in this study. Embryos were photographed using a dissecting microscope and then 7 μm transverse frozen sections were prepared. To assess the population of definitive endoderm in the embryonic gut, cell counts were performed on histological sections. 8.5 dpc embryos, prepared for whole mount in situ hybridisation for Hnf3α mRNA, were cryosectioned serially at 7 μm. Four samples each of normal and mutant embryos were processed. The number of Hnf3α-expressing cells per transverse section was scored on 7-10 randomly selected sections of the anterior and posterior gut endoderm. The data for each sample were expressed as mean and standard deviation.

Normal and mutant embryos at 7.75-8.5 dpc were fixed in 4% PFA-PBS for 2 hours, and washed in PBS. The PFA-fixed embryos were treated with TdT enzyme reaction, and then detected with anti-FITC-HRP conjugates by using an in situ apoptosis detection kit (Takara Biomedicals, Japan). After the whole-mount samples were photographed using a dissecting microscope, frozen sections were prepared. TUNEL-positive cells per transverse section were counted on randomly chosen sections to determine the level of cell apoptosis in the anterior or posterior definitive endoderm.

Embryos were isolated from a cross between heterozygous parents at 7.5 dpc and a portion of the extra-embryonic region was removed for genome typing. Each embryo was then dissolved in sample buffer and each protein sample (one embryo/lane) was used for SDS-PAGE and immunoblotting analysis using anti-Sox17-C17 antibody against synthetic peptides corresponding to amino acids 403 to 419 of the SOX17 protein (Kanai et al., 1996).

Isolated embryos were fixed in 2.5% glutaraldehyde/0.1 M phosphate buffer (PB) for 4 hours at 4°C. After washing in PBS, they were postfixed in 1% OsO₄ in 0.1 M PB for 2 hours at 4°C. The embryos were then dehydrated and embedded in EPON 812. Serial semi-thin sections (1 μm) were cut and stained with 1% Toluidine Blue.

During post-implantation mouse development, Sox17 mRNA is detected first in the visceral endoderm nearest to the ectoplacental cone of the pre-streak stage embryo (Fig. 1A,B) and expression progressively spreads to the entire extra-embryonic visceral endoderm by the early-streak stage (Fig. 1C,D). Sox17 is not detected in the embryonic visceral endoderm. At the mid-streak stage, Sox17 is also expressed de novo in the endoderm subjacent to the anterior end of the primitive streak (the broken arrow in Fig. 1E), which coincides with the recruitment of the definitive endoderm (Lawson and Pedersen, 1987; Tam and Beddington, 1992). Concomitant with the displacement of definitive endoderm to the anterior region of the gastrula, the Sox17 expression domain expands to encompass the endoderm underlying the neural plate of the early-bud-stage embryo (broken arrows in Fig. 1F-H,I,L). Sox17 is therefore an early marker that is specifically expressed in the definitive endoderm of the gastrula, in contrast to genes such as Hex, Hnf3β and Cer1, which are expressed simultaneously in the anterior visceral endoderm and the definitive endoderm in the embryonic compartment of the gastrulating mouse embryo. In the neurulating embryo, Sox17 is expressed in the endoderm of the prospective foregut (Fig.1M). However, by the early headfold stage, expression in the foregut endoderm (Fig. 1P) and in the visceral endoderm near the embryonic region (Fig.1N-P) is reduced. By contrast, endodermal cells in the prospective mid- and hindgut show strong Sox17 expression (Fig. 1Q). At the three to five somite stage, Sox17 is expressed in the posterior endoderm (Fig. 1R,S) but by the seven to eight somite stage, Sox17 is expressed mainly in the endoderm of the midgut and the hindgut invagination of the embryo (Fig. 1T,V,W). Sox17-expressing cells are no longer found in the prospective foregut endoderm at these stages (Fig. 1R,U). After 9.0 dpc, Sox17 expression is also absent from the posterior gut endoderm. The expression of the closely related Sox7 gene (Taniguchi et al., 1999), overlaps with that of Sox17 in the extra-embryonic endoderm in the gastrula and neurulation stages (Fig. 1H′,K′,M′). However, Sox7 is not expressed in the definitive endoderm (Fig. 1H′,M′) or the embryonic gut (data not shown). Sox17 and Sox7 function may therefore be redundant in the visceral endoderm, but not in the definitive endoderm.

A null mutation of Sox17 was generated by replacing exons 2-4 of the gene with a neomycin-resistant cassette, which resulted in the complete loss of SOX17 proteins (Fig. 2A-D). Sox17^+/– mice were normal and fertile. During embryonic development (until 10.5 dpc), Sox17^–/– mutant embryos were found in the expected Mendelian ratio: 23.3% (29/124 embryos) at 9.5 dpc and 17.8% (19/107) at 10.5 dpc. No homozygous mutant embryos were found after 10.5 dpc and no live homozygous offspring were recovered among 206 newborn.

Gross anatomical examination of the Sox17^–/– embryos revealed no discernible defects prior to the early-somite stage (Fig. 3A-C). From 8.5 dpc onwards, Sox17^–/– embryos can be distinguished by the lack of axis rotation and deteriorating growth (Fig. 3D-F). Development of anterior structures such as neural tube, optic evagination, otic vesicle, branchial arches and heart tube was generally normal until 9.5 dpc (Fig. 3E, and data not shown). However, growth and morphogenesis of the posterior trunk was severely affected and progressively became disorganised, especially after 9.5 dpc (Fig. 3E,F).

Histological analysis of Sox17^–/– mutant embryos at 7.0-9.5 dpc revealed that the first morphological defect was found in the definitive endoderm at 8.0 dpc (headfold stage; Fig. 3G-I). No histologically abnormality was detected at the median plane of the mutant embryos (Fig. 3G). However, at the paramedian level, the cuboidal and vacuolated visceral endoderm-like cells were found to extend into the embryonic region in the posterior endoderm, but not in the anterior endoderm, of the mutant embryos (Fig. 3H,I). This suggests a reduced definitive endoderm region at the posterior and lateral regions of the mutant embryos. At 8.25 dpc, the cells in the medial region of the definitive gut endoderm (underneath the notochord and somites) displayed the squamous epithelial morphology typical of the definitive endoderm. However, the cells in the lateral region (beneath the intermediate and lateral plate mesoderm) acquired the cuboidal and vacuolated epithelial appearance reminiscent of that of the visceral endoderm (Fig. 3J,K,N,O). Furthermore, this endodermal layer was juxtaposed directly with the mesothelium of the splanchnopleure and not, as in the wild-type embryo, separated from the mesothelium by a layer of splanchnic mesenchyme (red arrows in Fig. 3O). In the foregut and hindgut portals, visceral endoderm-like cells were found in the lateral and ventral epithelium of the gut (Fig. 3L,M). No obvious histological defects were found in other embryonic structures such as the notochord, somites, neural tube and heart. At 9.5 dpc, Sox17^–/– mutant embryos display the apparently normal development of the craniofacial structures, the heart (Fig. 3P) and the primordia of the liver and the thyroid gland (Fig. 3Q,R). However, the foregut is reduced in size and deficient of endoderm cells (Fig. 3R), and the mid- and hindgut degenerate into a cord-like structure (Fig. 3S).

The histological changes in the epithelium of the embryonic gut point to the deficiency of definitive endoderm during gut development in the Sox17^–/– embryo. To characterise the cell types present in the gut of the mutant embryo, the expression of definitive endoderm markers was examined by in situ hybridisation studies.

Hnf3α and Hnf3β are expressed in the definitive endoderm (Ang et al., 1993; Monaghan et al., 1993; Sasaki and Hogan, 1993), and the activity of both genes is critical for the formation of the definitive endoderm and its derivatives (Ang and Rossant, 1994; Dufort et al., 1998; Duncan et al., 1998; Kaestner et al., 1999). In Sox17^–/– mutant embryos, Hnf3α was expressed in the definitive endoderm at 8.5 and 9.5 dpc (Fig. 4A-D). However, in contrast to the normal (+/+ and +/–) embryos, Hnf3α-expressing cells were confined to a much narrower domain in the medial and paraxial region of the embryonic gut along the entire anteroposterior axis of the mutant embryo (Fig. 4A-D). The number of Hnf3α-positive cells (per transverse section of the gut of 8.5 dpc embryo, n=4) was significantly reduced (P<0.01) in the endoderm in both the anterior (normal, 40.0±2.1; mutant, 18.0±1.5) and the posterior parts of the embryo (normal, 38.5±3.0; mutant, 20.3±4.2). Hnf3β expression was weak in the cells in the degenerating segments but was maintained in the floor plate of the neural tube of the 9.5 dpc mutant embryo (Fig. 4E). Cdx2-expressing population (Beck et al., 1995) was also reduced in the hindgut of the mutant embryo (Fig. 4F,G). Hex, which is normally expressed in the definitive endoderm (Keng et al., 2000; Martinez Barbera et al., 2000), was expressed in a more restricted domain in the endoderm of the foregut portal (8.5 dpc; Fig. 4H). These findings suggest that the cell population that displays markers of definitive endoderm is significantly reduced in the embryonic gut of the mutant embryo.

Hex is expressed in the liver and thyroid primordia from 9.0 dpc onwards, and is essential for the development of the liver and thyroid (Keng et al., 2000; Martinez Barbera et al., 2000). In Sox17^–/– mutant embryos, Hex was expressed on time in both thyroid and liver primordia (Fig. 4I), which provides additional evidence that the initial development of these foregut derivatives is not affected (Fig. 3Q,R). Pdx1, which is necessary for the pancreatic outgrowth and the differentiation of the rostral duodenum (Jonsson et al., 1994; Offield et al., 1996), is expressed in the dorsal and ventral pancreatic regions of the gut endoderm after 8.5 dpc (Guz et al., 1995). In the Sox17^–/– mutant embryos, Pdx1 expression was not detected in the gut tissues at the sites of the primordia of either the dorsal or the ventral pancreatic buds (Fig. 4J). Therefore, in contrast to the independence of liver and thyroid on normal Sox17 function to initiate early differentiation, Sox17 activity may be essential for the formation of the pancreas.

During gut development, Sonic hedgehog (Shh) and Indian hedgehog (Ihh) are expressed in the endoderm and play crucial roles in the regionalisation of the primitive gut tube (Apelqvist et al., 1997; Hebrok et al., 1998; Ramalho-Santos et al., 2000; Sukegawa et al., 2000). In the 8.5-9.5 dpc Sox17 mutants, Shh was expressed at a low level in the anterior endoderm (open arrows in Fig. 4L,N,O), but was absent from the posterior endoderm (Fig. 4K,M) and the mid- and hindgut epithelia (Fig. 4N,P). Similarly, Ihh expression was absent in the endoderm of the mutant embryonic gut (8.5 dpc; Fig. 4Q) and the mid- and hindgut epithelia (9.5 dpc; Fig. 4R). However, Ihh expression in the visceral endoderm was not altered.

Besides the deficiency of the definitive endoderm and abnormal morphogenesis of the mid- and hindgut, no other obvious abnormality was found in other germ-layer derivatives and organs of the Sox17 mutant embryos, except for the absence of the mesenchymal tissue that normally intercalates between the splanchnic mesothelium and the endoderm in the lateral region of the embryonic gut. Ptch, which encodes a receptor that modulates Hedgehog (Hh) signalling (Goodrich et al., 1996; Motoyama et al., 1998), was not expressed in the splanchnic mesothelium (Fig. 4S,T); however, it was expressed appropriately in the floor plate and the ventral part of the paraxial mesoderm. These findings suggest that the disruption of Hh signalling activity revealed by the loss of Shh and Ihh expression in the endoderm of the embryonic gut and Ptch expression in the neighbouring mesothelium may cause the loss of the splanchnic mesenchyme in Sox17^–/– mutant embryo (Fig. 3O).

Correct tissue- and region-specific expression of several genes was observed in the cranial neural tube [Otx2 (Simeone et al., 1992); data not shown], somites [Meox1 (Candia et al., 1992); data not shown], the cardiogenic mesoderm (Gata4, Fig. 6F,G), the notochord (T, data not shown; Shh, Fig. 4K-P) and the tail bud mesenchyme (T, Wnt3α and Wnt5a, data not shown; Cdx2, Fig. 4F) of the mutant embryos before any disruption of the trunk and tail morphology became evident.

The histological and in situ hybridisation analyses demonstrated that the loss of Sox17 activity causes the depletion of the definitive endoderm in the embryonic gut. The occurrence of apoptosis of the endoderm of the Sox17^–/– mutant was examined by TUNEL analysis. Mutant embryos did not show any more TUNEL-positive cells than the normal embryos prior to the neural plate stage (Fig. 5A). There was, however, a significant increase of the number of TUNEL-positive cells in the prospective foregut endoderm underneath the anterior and lateral margin of the neural plate by the late neural plate to early headfold stages (Fig. 5B,D-F), but no appreciable increase of apoptotic cell death was found in the posterior definitive endoderm (Fig. 5C,D). Apoptotic cells were found more frequently in the endoderm in the medial region of the prospective foregut of the mutant embryo (Fig. 5G-I; normal, 0.33±0.17; mutant; 1.13±0.23; n=4). Interestingly, in the mutant embryo, normal Cer1 expression was found in the prospective foregut endoderm at the early-bud to headfold stages (Shawlot et al., 1998) (Fig. 5J,K). This suggests that the initial formation of the anterior definitive endoderm is therefore not affected by the loss of Sox17 activity, but increased cell death may underpin the deficiency of endoderm in the foregut. By contrast, there was no significant increase in the number of TUNEL-positive cells in the prospective hindgut of the mutant embryo (Fig. 5H; normal, 0.13±0.11; mutant, 0.14±0.12; n=4). The depletion of the definitive endoderm in the posterior segment of the embryonic gut may be caused by factors other than apoptosis.

The visceral endoderm of the yolk sac characteristically displays active endocytosis, revealed by the uptake of horseradish peroxidase (HRP) for transporting macromolecules across the foetal membrane (Bielinska et al., 1999). The visceral endoderm in the yolk sac of the mutant embryos showed normal HRP-uptake activity (Fig. 6A-C). However, endodermal cells that actively incorporate HRP are also found in the prospective gut region of the embryo at late headfold to early somite stage, resulting in a significant reduction in the area of non-HRP-labelled definitive endoderm that is confined to the paraxial and medial region of the embryonic gut. Several genes that are expressed in the visceral endoderm [Hnf4 (Chen et al., 1994); Gata4 (Molkentin et al., 1997; Narita et al., 1997); Ihh (Maye et al., 2000)] also show an increase in their expression domain into the lateral region of the gut of the mutant embryo (Hnf4, Fig. 6D,E; Gata4, Fig. 6F,G; Ihh, Fig. 4Q). α-Fetoprotein (Afp), which is expressed specifically in the extra-embryonic part of the visceral endoderm (Dziadek and Adamson, 1978), restricts its expression to the visceral endoderm up to the early neural plate stage (Fig. 6H) but extends into the posterior and lateral regions of the embryonic gut of the headfold stage mutant embryo (Fig. 6I,J). By contrast, no significant expansion of Afp expression domain was found in the anterior region of the embryonic gut at this stage of development (Fig. 6I).

Our results show that deficiency in the definitive endoderm in different region of the embryonic gut may be brought about differently by apoptosis and inadequate population expansion during the neural-plate to headfold stages. The expansion of the visceral endoderm-like cells into the embryonic gut and the loss of the expression of markers for definitive endoderm all suggest that the reduced population of definitive endoderm could not effectively displace the visceral endoderm to the extra-embryonic sites and results in the occupation of the substantial region of the gut by an inappropriate type of cells.

To assess the impact of the loss of Sox17 on tissue potency, the differentiation of Sox17 mutant cells was examined in chimaeric mouse embryos comprised of Sox17^–/– ES cells and blastocyst derived from ROSA26 (Sox17^+/+) mice which constitutively express β-galactosidase (Tremblay et al., 2000). Twelve of the 31 chimaeric embryos harvested at 8.25 and 9.5 dpc showed more than 80% tissue contribution by Sox17^–/– ES cells. Five chimeras (one at 8.25 dpc and four at 9.5 dpc) with high mutant ES cell contribution displayed a reduction in embryonic gut size and failed to undergo axis rotation (Fig. 7A-D,E,F), which recapitulates the phenotype of the Sox17-null mutant embryos. The remaining seven chimaeric (seven at 9.5 dpc) embryos, containing moderate to high mutant ES cell contribution, were slightly retarded (Fig. 7E; one showed delayed axis rotation) but were otherwise morphologically normal.

Histological analysis of the chimaeric embryos showed that the Sox17^–/– ES cells can contribute extensively to the ectodermal and mesodermal tissues (such as skin, neural tube, notochord, somites, lateral plate mesoderm and yolk sac mesoderm; Fig. 7B-D,F-K). Of significant interest, however, is the marked reduction in the contribution by mutant ES cells to the foregut endoderm (Fig. 7B,C,H,I) and the almost complete exclusion of the mutant cells from the mid- and hindgut endoderm (Fig. 7D,F,J,K). This finding clearly points to the cell-autonomous activity of Sox17 in regulating differentiation of the definitive endoderm particularly in the mid- and hindgut of the embryo. The differential impact of the loss of Sox17 function on the endodermal potency of cells colonising the foregut versus those colonising the mid- and hindgut underpins the different defects between the anterior and posterior endoderm formation in the null mutant embryos. In the chimeras, only the wild-type host cells contributed to the extra-embryonic visceral endoderm. That the restoration of Sox17 activity in the visceral endoderm has neither enhanced nor rescued the morphological abnormality and the depletion of definitive endoderm displayed by the chimeras with the strongest mutant ES cell contribution indicates that Sox17 function in the extra-embryonic tissues is not crucial for embryogenesis. The phenotype of the null mutant embryo is therefore primarily caused by the defective differentiation of the derivatives of the Sox17-deficient epiblast.

The findings of the phenotypic analysis of the Sox17-null mutant embryo show that Sox17 plays a crucial role in the maintenance and differentiation of the definitive endoderm of the embryonic gut. Sox17 function may not be essential for the formation of the progenitor cells of the definitive endoderm during gastrulation but is required for sustaining the viability of the foregut endoderm and the elaboration of the mid- and hindgut endoderm. The loss of Sox17 function results in the apoptosis of the endoderm in the foregut and the failure of the expansion of the endoderm of the mid- and hindgut. The diminished contribution of Sox17^–/– mutant cells to the gut endoderm of the chimera further shows that Sox17 is essential for the embryonic cells to acquire an endodermal potency.

The lack of defects in the development of the extra-embryonic endoderm and the initial phase of the formation of anterior definitive endoderm in Sox17 mutant raise the possibility of potential redundancy of other gene activity. Members of the Sox gene family share similar DNA-binding specificity, especially within the same subgroup. Among the members, Sox7, Sox17 and Sox18, of mammalian Sox subgroup F, Sox18 has been shown to be expressed in yolk-sac blood islands, the developing endothelial cells and presumptive endocardial cells, but not in the endoderm cell lineage during early embryogenesis (Pennisi et al., 2000). Analyses of expression revealed that Sox7 and Sox17 genes are co-expressed in the extra-embryonic visceral endoderm, suggesting that there may be functional compensation by Sox7, which could account for the lack of discernible defects on the visceral endoderm in the Sox17-null mutant embryo. However, only Sox17 is expressed in the definitive gut endoderm, suggesting redundancy among the subgroup F members is unlikely in the definitive endoderm lineage. The severity of the posterior gut phenotype of the Sox17 mutant further implicate that although Sox2 is expressed throughout the gut endoderm, and Sox3 transcripts are also detected in the posterior region of the foregut in mouse early-somite stage embryos (Wood and Episkopou, 1999), the activity of these two Sox genes are not sufficient to compensate for the loss of Sox17 function. Recently, the casanova mutation, which causes defective endoderm formation in the zebrafish, has been found to involve a novel Sox gene of the subgroup F (Dickmeis et al., 2001; Kikuchi et al., 2001; Sakaguchi et al., 2001). Whether any casanova orthologue is present in the higher vertebrates is unclear.

During early embryogenesis, morphogenetic activity of the extra-embryonic visceral endoderm has been shown to be important in the establishment of the body plan of the embryo (Beddington and Robertson, 1998; Beddington and Robertson, 1999; Brennan et al., 2001; Episkopou et al., 2001; Hoodless et al., 2001; Kalantry et al., 2001; Lu et al., 2001; Yamamoto et al., 2001). Despite the expression of Sox17 in the extra-embryonic visceral endoderm of the embryo during gastrulation and early organogenesis, no early developmental defects are found in the Sox17 mutant embryo. The overlapping expression of the structurally homologous Sox7 may have provided the redundancy that functionally compensates for the loss of Sox17 activity. The defective differentiation of the definitive endoderm in the mutant embryo suggests that similar functional overlap of the activity of other Sox genes of the same subgroup or any other endoderm-specific genes are absent. However, the disorganised development of the posterior trunk and the interruption of axis morphogenesis may point to some secondary impact of either the loss of Sox17 function in the definitive endoderm or the abnormal formation of the embryonic gut that is deficient of the proper type of endodermal cells.

Several studies have suggested that the definitive endoderm may exert morphogenetic influence on the development of the mesoderm and ectodermal tissues. The anterior definitive endoderm and Hex activity specifically expressed in this tissue are required in tissues for normal forebrain formation (Martinez Barbera et al., 2000; Withington et al., 2001). In the Sox17^–/– mutant embryo, proper though spatially restricted, the expression of several genes such as Hnf3α, Hnf3β, Hex, Cer1 and Cdx2 is still present in the endoderm of the embryonic gut. This may suggest that whatever morphogenetic activity exerted by the definitive endoderm is still provided to other germ-layer derivatives before extensive loss of anterior definitive endoderm by apoptosis occurs. This may account for the apparently normal development of the cranial and upper trunk structures. Development of the more posterior trunk structures is becoming more severely affected with more extensive depletion of the definitive endoderm in the posterior segments of the embryonic gut. In this regard, the loss of Shh and Ihh activity in the endoderm and the concurrent downregulation of Ptch expression in the lateral plate mesoderm may be significant. Whether the specific loss of Hh signalling in the gut endoderm but not in other structures such as the notochord and the floor plate is associated with the retarded development and growth of posterior trunk structures of the Sox17^–/– embryos is still unclear. However, a more widespread deficiency in Hh signalling in the Shh/Ihh-double mutant embryo has also resulted in a similar developmental arrest phenotype with the lack of axis rotation at the early somite stage (Zhang et al., 2001). Consistent with this notion, Hh signalling has been shown to play a crucial role in the regionalisation of the primitive gut tube by acting through the adjacent mesoderm (Apelqvist et al., 1997; Hebrok et al., 1998; Ramalho-Santos et al., 2000; Sukegawa et al., 2000).

Results of the chimera study reveal that the Sox17^–/– mutant ES cells are less competent to colonise the foregut endoderm and are excluded from the mid- and hindgut. In the null mutant embryo, Sox17-deficient cells can form the early endodermal progenitor and can contribute to the definitive endoderm in all regions of the gut. However, increased apoptosis of the mutant cells in the foregut and failure of mutant cells to differentiate or proliferate in the posterior gut leads to the depletion of the definitive endoderm in all segments of the embryonic gut. These findings strongly suggest that Sox17 may function differentially as a maintenance factor in the endoderm of the foregut and a differentiation regulator in the rest of the embryonic gut.

The different impact of the loss of Sox17 function on the potency of the cells to contribute to different parts of the gut of the chimera points to a variable requirement for Sox17 activity in the allocation of different populations of definitive endoderm. Like the Sox17 mutant cells, Smad2-deficient cells are able to contribute at a moderate level to the anterior definitive endoderm, but are absent from the hindgut (Tremblay et al., 2000). While Sox17-deficient cells are unable to colonise the mid- and hindgut of the chimera, Hnf3β-deficient cells can form the hindgut, but are not in the foregut and midgut, of chimeras generated from the null ES cells and the tetraploid embryo (Dufort et al., 1998). These findings strongly suggest a differential requirement of Sox17, Smad2 and Hnf3β in the endoderm of different parts of the embryonic gut. The similarity in behaviour of the Sox17- and Smad2-deficient cells in the chimera is consistent with the recent demonstration that Xsox17 activity may be required for the maintenance of Nodal signalling activity in the cells that are primed for endodermal differentiation in the Xenopus (Engleka et al., 2001; Aoki et al., 2002).

In the Xenopus, Xsox17α and Xsox17β direct the differentiation of embryonic cells towards the endodermal fate in a cell-autonomous fashion (Hudson et al., 1997; Clements and Woodland, 2000). Furthermore, the activation of Xsox17 by VegT enables the delineation of the endodermal progenitor from other germ-layer tissues and initiates endoderm differentiation (Engleka et al., 2001). Recently, the isolation and characterisation of other genes involved with endoderm formation in Xenopus and zebrafish has enabled the construction of a molecular pathway involving nodal-related signalling and the induction of Mix-type homeobox or Gata-type zinc-finger genes, which in turn activate the endoderm determining genes, Sox17 and Hnf3β (Alexander and Stainier, 1999; Reiter et al., 2001). In mouse early embryogenesis, Hnf3β (Dufort et al., 1998), Gata4 (Narita et al., 1997), Smad2 (Tremblay et al., 2000) and Fast1 (Hoodless et al., 2001), which are involved in Activin/Nodal-related signalling, play an important role in the formation of the definitive endoderm. The phenotypic consequences of the loss of Sox17 function in the mouse are consistent with the concept that Sox17 activity is required by the progenitor cells to acquire endodermal potency and to differentiate properly into definitive endoderm. Our findings highlight the conservation of Sox17 function in endoderm formation during vertebrate embryogenesis.

The authors thank Drs P. Koopman, H. Hamada, T Hara and Y. Saijoh for their kind and helpful advice and discussion regarding this work and their critical reading of this paper. They also thank Dr N. Fujisawa for his generous assistance, and Mr I. Tsugiyama and Mrs H. Hayashi for their technical and secretarial support. The authors are grateful to Drs E. Olson, A. McMahon, B. G. Herrmann, C. V. Wright, H. Hamada, I. Matsuo, S.-L. Ang, J. Motoyama, H. Sasaki, Y. Hiraoka, J. Barras, S. Dunn, T. Noguchi, F. Guillemot, D. Anderson and M. Wilkinson for providing the in situ probes. This work was mainly supported by financial grants from the Ministry of Education, Science, Sports and Culture of Japan to Y. Kanai (numbers 11660306 and 13660295). This work was also supported by the Program for Promotion of Basic Research Activities for Innovative Biosciences (Y. K.), the Sumitomo Foundation (M. K.-A.), the National Health and Medical Research Council (NHMRC) of Australia and Mr James Fairfax (P. P. L. T., J. M. G.). P. P. L. T. is a NMHRC Senior Principal Research Fellow. This paper is dedicated to the memory of Dr Toshiya Yamada (Institute for Molecular Bioscience, The University of Queensland, Australia), our dear friend, who died on 12th May 2001.