Mixl1 is required for axial mesendoderm morphogenesis and patterning in the murine embryo

In Xenopus, members of the Mix/Bix family of Paired-like homeobox genes have been implicated as regulators of gastrulation. Mix.1, the founding member of this family, was originally identified as an immediate-early response gene to activin induction and was later rediscovered during a screen for ventralizing molecules (Rosa, 1989; Mead et al., 1996). Currently, there are seven family members: Mix.1, Mix.2, Mix.3/Mixer, Bix1/Mix4, Bix2/Milk, Bix3 and Bix4 (Rosa, 1989; Vize, 1996; Ecochard et al., 1998; Henry and Melton, 1998; Mead et al., 1998; Tada et al., 1998). Transcription of all the members of the Mix/Bix family begins soon after the mid-blastula transition in presumptive mesendoderm in the equatorial region of the embryo and ceases after gastrulation. Expression requires maternal VegT, a T-box transcription factor that induces the Xenopus transforming growth factor (TGF) β-like molecules Xnr1, Xnr2, Xnr4 and derriere (Xanthos et al., 2001).

Studies of MIX/BIX homeoprotein function in Xenopus have shown that they are downstream transcriptional targets in the TGFβ superfamily pathway that regulates mesendodermal patterning. For Mix.2, the pathway of activin-induced transcriptional activation has been determined. Signals from TGFβ family members are transduced to the nucleus by intracellular SMAD proteins. SMAD proteins are recruited to DNA by other DNA-binding proteins, the prototype of which is FOXH1 (FAST1). FOXH1 forms a complex with activated SMAD2/SMAD4 dimers to bind to the activin-responsive element of the Mix.2 promoter (reviewed by Massague and Wotton, 2000). Other Xenopus MIX/BIX proteins, MIXER, BIX2/MILK and BIX3, recruit active SMAD2/SMAD4 complexes directly to an activin-response element found in mesendodermal genes such as Gsc (Germain et al., 2000).

Mix.1 has been implicated in endoderm development (Hudson et al., 1997; Lemaire et al., 1998; Latinkic and Smith, 1999). Expression of dominant negative Mix.1 constructs prevents endoderm formation and, while overexpression of Mix.1 does not induce endoderm formation in animal cap explants, it can act synergistically with the Paired-like homeobox genes Gsc or Siamois to upregulate transcription of endodermal markers (Henry and Melton, 1998; Lemaire et al., 1998; Latinkic and Smith, 1999). Moreover, Mix.1 has been shown to repress the expression of the mesodermal marker Xbra, the Xenopus homolog of Brachyury (Latinkic et al., 1997; Latinkic and Smith, 1999). Bix2/Milk, Mix.3/Mixer and Bix4 have also been shown to induce endoderm formation (Ecochard et al., 1998; Henry and Melton, 1998; Casey et al., 1999). Depending on the level of overexpression, Bix1/Mix4 can induce the expression of endoderm or ventral mesoderm markers in animal cap explants (Tada et al., 1998). However, the roles of individual Mix/Bix genes in the regulation of mesendodermal patterning in Xenopus remain unclear, especially as homeobox proteins are known to homo- or heterodimerize with DNA (Wilson et al., 1993; Wilson et al., 1995a; Mead et al., 1998).

In the chick, a Mix-like gene (Cmix) is expressed in epiblast and endoderm of the posterior marginal zone prior to formation of the primitive streak. Thereafter, expression is seen in epiblast and nascent mesoderm of the primitive streak but not in Hensen’s node. Cmix transcripts are no longer detectable after formation of the head process (Peale et al., 1998; Stein et al., 1998). In zebrafish, the gene associated with the bonnie and clyde (Bon) mutation is a Mix-type gene with similarity to Xenopus Mixer (Kikuchi et al., 2000). Bon is induced in the margin of late blastula stage embryos by NODAL signaling and is required for early endoderm formation and, like Mixer in Xenopus, regulates expression of the endodermal gene Sox17 (Henry and Melton, 1998; Alexander and Stainer, 1999). Bon mutants exhibit cardia bifida and have fewer Sox17-expressing endodermal cells (Alexander et al., 1999; Kikuchi et al., 2000). Genetic studies have shown that Bon lies downstream of the NODAL-related proteins encoded by the Cyclops and Squint genes (Kikuchi et al., 2000).

To date, one murine Mix-like gene, Mixl1 (Mml, Mix) (Pearce and Evans, 1999; Robb et al., 2000) and one human homolog (Robb et al., 2000) have been identified. The amino acid sequence of the homeodomain and a conserved C-terminal acidic region of the murine MIXL1 is similar to MIX.1 and other proteins of Xenopus MIX/BIX family (Pearce and Evans, 1999; Robb et al., 2000). Mixl1 RNA was first detected in the visceral endoderm of the pre-gastrula embryo. By 6.5 days post coitum (dpc), expression was seen in nascent primitive streak and emerging mesoderm. Mixl1 expression was maintained in emerging mesoderm in mid- to late-streak embryos, becoming restricted to the posterior primitive streak by the head-fold stage. Expression was not seen in the node or in endoderm flanking the node. In early somite stage embryos, expression was detectable in the crown cells of the caudal notochord and in primitive streak (Pearce and Evans, 1999; Robb et al., 2000). A single human Mixl1 homolog has been identified (Robb et al., 2000).

To establish the role of Mixl1 in murine development, we created a mouse strain bearing a targeted mutation of the Mixl1 gene. Gastrulation stage Mixl1-null embryos had abnormalities in primitive streak and node formation. Post-gastrulation, Mixl1 mutants exhibited complex defects in axial mesodermal and endodermal structures, suggesting a role for Mixl1 in the regulation of morphogenetic cell movements associated with gastrulation. Despite the defects in early morphology and terminal differentiation of trunk organizer cells, gene marker studies indicated that their inductive properties were maintained. Moreover, in chimera studies, Mixl1-null embryonic stem cells were found to contribute poorly to gut endoderm, thereby implicating Mixl1 in endoderm patterning.

A 5′ 4 kb fragment of the Mixl1 genomic locus which included the Mixl1 translational start was cloned in frame with EGFP in the vector pEGFPKT1LOXNEO (Godwin et al., 1998). A 1.9 kb fragment of the Mixl1 genomic locus incorporating intron 1 and exon 2 was inserted at the 3′ end of the loxP-flanked neomycin selection cassette to complete the Mixl1 targeting vector. The W9.5 embryonic stem (ES) cell line was maintained, electroporated and selected with G418 as described (Robb et al., 1995). Homologous recombinants were identified by Southern analysis of HindIII-digested DNA isolated from individual clones. Southern blots were hybridized with the 3′ genomic DNA probe (Fig. 1A). The targeted deletion was confirmed by further analysis of Southern analysis of DNA from selected clones with the 5′ genomic DNA probe (Fig. 1A). Two targeted clones were injected into C57BL/6 blastocysts and chimeric offspring were bred to 129Sv or C57BL/6 mice to derive mice heterozygous for the Mixl1 mutation (Mixl1^+/–) or to Cre transgenic mice (Schwenk et al., 1995) to create mice heterozygous for a targeted Mixl1 allele lacking the neomycin selection cassette (Mixl1^+/Δneo).

Adult heterozygote Mixl1^+/– and Mixl1^+/Δneo mice were genotyped by Southern analysis of HindIII-digested tail DNA using the 3′ probe. Adult mice and embryos were also identified by PCR analysis using tail and embryonic yolk sac DNA (Moens et al., 1993). Tissue was digested in a buffered non-ionic detergent solution (100 ng/ml proteinase K, 50 mM KCl, 10 mM Tris.HCl (pH 8.3), 2.0 mM MgS0₄, 0.45% Nonidet P-40, 0.45% Tween-20) for 3 hours to overnight followed by inactivation of proteinase K activity at 100°C for 10 minutes. Samples (1 μl) were analyzed in a 10 μl Taq DNA polymerase PCR reaction, as described by the manufacturer (Invitrogen). To amplify the wild-type and Mixl1^– alleles, a mixture of three primers was used in the same reaction: In1a, 5′-CTTCCTGCTGGGACCTTTCAATGG-3′; Ex1a, 5′-TCGTCTTCCGACAGACCATGTACC-3′; and Neo1a, 5′-CTGACCGCTTCCTCGTGCTTTACG-3′. To amplify the wild-type and Mixl1^Δneo alleles, the primer GFP1a (5′-ATGGACGAGCTGTACAAGTAAAGCG-3′) replaced primer Neo1a. Embryos between 6.75 and 8.5 dpc were also genotyped by analysis of EGFP fluorescence using a Leica MZFLIII dissecting microscope using a mercury lamp and GFP2 filter system.

Embryos obtained from timed matings of Mixl1^+/– or Mixl1^+/Δneo pairs were harvested and maintained at room temperature in Hepes-buffered Eagles Medium without Phenol Red during imaging with a Leica TCS-SP2 Confocal Imaging System connected to a Leitz Arisoplan microscope. Thick sections for confocal imaging were cut using disposable steel blades (Probing and Structure).

For histology, embryos were fixed in 4% paraformaldehyde, dehydrated through an ethanol series and embedded in paraffin wax. Blocks were sectioned at 6 μm, dewaxed and stained in Hematoxylin and Eosin. Embryos for whole-mount in situ hybridization were fixed in 4% paraformaldehyde and stored at 4°C. Hybridization was performed as described (Belo et al., 1997). After hybridization, stained embryos were refixed in 4% paraformaldehyde. Selected embryos were dehydrated, embedded in paraffin wax and sectioned at 10 μm.

Wild-type R26.1 ES cells, which are derived from a mouse carrying the ROSA-26 lacZ gene trap (Varlet et al., 1997), were injected into blastocysts obtained by mating Mixl1^+/– and Mixl1^+/Δneo animals. Chimeras derived from Mixl1^–/Δneo blastocysts were identified by PCR of yolk sac DNA, using the primers pairs described above. PCR of Mixl1^–/Δneo yolk sac DNA gave rise to products specific for the Mixl1^– allele and the Mixl1^Δneo allele, in addition to the wild-type (maternally derived) allele. Mixl1^–/Δneo ES cell lines were generated by first using CRE-mediated recombination to remove the loxP-flanked neomycin resistance cassette from a Mixl1^+/– ES cell line, thereby creating Mixl1^+/Δneo cell lines. The second allele was then replaced by a further round of gene targeting with the original targeting construct. Two Mixl1^–/– ES cell lines were identified by Southern analysis of HindIII-digested DNA using the 3′ probe to detect the 4.7 kb and 3.7 kb mutant alleles. The correct targeting of the locus was confirmed by further Southern analysis. Chimeric embryos were generated by injection of Mixl1^–/– ES cells into blastocysts carrying the ROSA-26 lacZ transgene. β-Galactosidase activity in chimeric embryos was detected by X-gal histochemistry (Elefanty et al., 1999). Stained embryos were post-fixed and sectioned at 10 μm.

Nodal^+/lacZ mice, generously provided by Liz Robertson, were bred with Mixl1^+/– mice. The offspring were genotyped for Mixl1 as above and for Nodal as described (Collignon et al., 1996). Compound heterozygotes were interbred and offspring genotyped by PCR of yolk sac DNA. β-galactosidase activity was detected as described elsewhere (Elefanty et al., 1999).

To investigate the role of Mixl1 in mammalian development, a loss-of-function mutation was generated. The Mixl1-coding sequence in exon 1 was replaced, in frame, by a cassette encoding the enhanced green fluorescent protein (GFP) and neomycin resistance (Fig. 1A). An ES cell line that had undergone homologous recombination was used to generate chimeric mice. Chimeras were bred to 129Sv or C57BL/6 mice to derive offspring heterozygous for the Mixl1 mutation (Mixl1^+/–) or to Cre transgenic mice (Schwenk et al., 1995) to create offspring heterozygous for a targeted Mixl1 allele lacking the neomycin selection cassette (Mixl1^+/Δneo) (Fig. 1B). Both Mixl1^+/– and Mixl1^+/Δneo mice appeared normal and were fertile. Heterozygous intercrosses never yielded viable homozygous offspring, indicating embryonic lethality. In litters produced by mating Mixl1^+/– or Mixl1^+/Δneo mice, homozygous Mixl1 mutant embryos were present in the expected Mendelian ratio at 7.5 dpc, but were absent after 10.5 dpc. The phenotype of homozygous embryos (hereafter referred to as Mixl1^–/– or mutant embryos) that were derived from matings of either Mixl1^+/– or Mixl1^+/Δneo heterozygous mice was indistinguishable and results from the analyses were therefore pooled. Examination of Mixl1 expression in wild-type and Mixl1^–/– 7.5 dpc embryos by whole-mount in situ hybridization showed a complete lack of Mixl1 transcripts (data not shown).

After homologous recombination, the GFP reporter and the neomycin selection cassette replaced most of Mixl1 exon 1, leaving potential 5′, 3′ or intronic regulatory sequences intact. In keeping with pattern of Mixl1 RNA expression, the visceral endoderm, primitive streak and nascent mesoderm of gastrulation stage Mixl1^+/– embryos expressed GFP. Fluorescence was first detected at 6.5 dpc, 1 day after Mixl1 RNA was detectable by whole-mount in situ hybridization (Pearce and Evans, 1999; Robb et al., 2000) and GFP signal in Mixl1^+/– embryos persisted in visceral endoderm up to the late bud stage. Co-incident with the Mixl1 mRNA expression at 7.5 dpc, maximal fluorescence was seen in the primitive streak and mesodermal wings at this time (Fig. 1C,F). At 8.25 dpc, fluorescence was confined to the posterior primitive streak of Mixl1^+/– embryos, and by 9.0 dpc was completely absent, mirroring the Mixl1 mRNA expression pattern (data not shown). In Mixl1^–/– embryos, the fluorescence was brighter than in Mixl1^+/– embryos and remained detectable until 9.5 dpc (Fig. 1D,F and not shown).

At 6.5 dpc, Mixl1^–/– embryos are morphologically indistinguishable from their littermates. However by 7 dpc, at the mid-streak stage, Mixl1^–/– embryos display a thickened primitive streak and lack a morphologically recognizable node at the anterior end of the primitive streak (Fig. 2A,F). An excess of mesoderm-like cells accumulates in the anterior streak and in the region immediately lateral to the primitive streak (Fig. 2B-E,G-J). The early-neural plate stage (7.75 dpc) Mixl1^–/– embryos are retarded in the elevation of neural folds and develop a shallow anterior intestinal portal (data not shown). At the early somite-stage (8.5 dpc), the embryos adopt a flattened disc morphology instead of a lordotic shape and the anteroposterior (AP) axis is markedly foreshortened (Fig. 3A-C). The neural folds are small and disorganized and very little head mesenchyme is found underneath the head folds (Fig. 3I-M). The heart tube is absent (Fig. 3I,J). The foregut invagination is rudimentary and no hindgut portal is formed (Fig. 3I,L,M). From two to five pairs of somites are formed in the paraxial mesoderm flanking a thick axial mass of flattened neurectoderm overlying a condensed core of mesodermal tissues that replaced the notochord.

Rarely, Mixl1^–/– embryos with severely disorganized head folds were recovered from 9-9.5 dpc litters. These embryos failed to turn and did not form a heart tube (Fig. 3D, Fig. 4A-E). In the 20 9.5 dpc Mixl1^–/– embryos examined, the axial tissue mass present at 8.5 dpc had grown in size, and developed into a huge mass of tail bud-like tissue protruding dorsally from the caudal region of the embryo (Fig. 3D, Fig. 4A,B,E). In some embryos, an additional bulbous tissue forms ectopically in the trunk of the embryo (Fig. 4B,D), giving an impression that the embryonic axis is bifurcated. Analysis of Brachyury expression shows that axial tissues in these mutants are discontinuous along the body axis. Brachyury is expressed in a poorly organized structure that resembles the notochordal plate and in a notochord-like structure that ends caudally, either in an ectopic protrusion (arrows in Fig. 4B,D) or in a bulbous mass (Fig. 4E).

In the mutants, the fusion of the posterior and anterior amniotic folds is delayed, leading to persistence of the proamniotic canal as late as the late bud/early head fold stage (Fig. 2F). The differentiation of extra-embryonic tissue in 8.5 dpc Mixl1^–/– conceptuses appears normal. Blood islands are plentiful in the yolk sac of Mixl1^–/– embryos (Fig. 3J). However, in comparison with wild-type littermates, the allantois of Mixl1^–/– embryos appeared disproportionately large in size (Fig. 3B).

Brachyury (T), which is expressed in the primitive streak and axial mesoderm of the wild-type 7.5 dpc embryo (Wilkinson et al., 1990), is expressed in an expanded domain in both the epiblast and mesenchymal components of the thickened primitive streak and in the adjacent mesodermal cells in the Mixl1^–/– embryos (Fig. 5A-D). In 8.25 dpc mutant embryos, Brachyury is expressed in an irregular and broad strip of mesodermal cells underlying the midline of the head fold and neural plate along the AP axis. (Fig. 5E-G). Histological examination reveals that Brachury is expressed in ventral tissues of the open neural plate that resemble the floor plate of the neural tube (Fig. 5H). In the posterior region of the mutant embryo, Brachyury is expressed in the superficial layer of cells of the primitive streak and in the ventral-most tissues that appear to have organized into a notochordal plate-like structure (Fig. 5I). In contrast to the uniform expression in the primitive streak of wild-type embryo, expression is excluded from the core of the primitive streak (Fig. 5I).

We also examined the expression of Nodal in Mixl1^–/– embryos. Expression of Nodal was assessed in Mixl1 mutant embryos that are heterozygous for Nodal^lacZ, a null allele generated by integration of a lacZ transgene into the Nodal locus that provides a reporter for Nodal expression (Collignon et al., 1996). Nodal expression in the wild-type embryo is very dynamic, being initially expressed throughout the epiblast at 6.5 dpc, then localizing to the primitive streak before being restricted to a horseshoe-shaped domain around the node by 8 dpc (Zhou et al., 1993; Conlon et al., 1994; Varlet et al., 1997). In the mid- to late-streak stage, Mixl1^–/– mutant embryo, the domain of Nodal expression in the epiblast and the primitive streak is significantly expanded (Fig. 5J). In older mutants, Nodal expression spreads more widely in the posterior midline, but is absent from lateral plate mesoderm (Fig. 5K,L).

In the wild-type embryo, Foxa2 (HNF3β) expression is localized to the anterior primitive streak, the node and the anterior mesendoderm at 7.5 dpc and at early somite stages marks axial midline structures such as notochord, floor plate and gut endoderm (Ang et al., 1993; Monaghan et al., 1993; Ruiz i Altaba et al., 1993; Sasaki and Hogan, 1993). In 7.5 dpc Mixl1^–/– embryos, an expanded domain of expression of Foxa2 is found in the mesendodermal tissues in distal region of the gastrula-stage embryo at the anterior end of the primitive streak (Fig. 6A,B). At 8.25 dpc, Foxa2 is present in the axial tissues over most of the length of the embryonic axis of the mutant embryos (Fig. 6C,F). Sonic hedgehog (Shh) is first detected at the late streak stage in the anterior mesendoderm, which extends rostrally from the primitive streak, and during organogenesis, in the node, head process, notochord, prechordal plate, floorplate and endoderm in the dorsolateral region of the embryonic gut (Echelard et al., 1993; Roelink et al., 1994). In the late-streak stage Mixl1^–/–embryos, Shh expression is restricted to cells clustered near the anterior end of the streak (Fig. 6G,H). In 8.25 dpc mutants, weak, poorly localized Shh expression is seen in the ventral part of the condensed axial tissue that resembles the notochordal plate (Fig. 6I-L). Gsc, which marks the prechordal plate in late head-fold-stage embryos, is expressed in the mesendoderm in the anterior region of the mutants (data not shown). Overall, the marker studies showed that tissues displaying the molecular characteristics of axial mesendoderm are specified in the Mixl1^–/– embryos, despite proper axial structures not being formed.

In the Mixl1^–/– embryos, Twist is expressed in the presumptive paraxial mesoderm in the posterior region of the embryo and the allantoic mesoderm. However, unlike wild-type embryos, Twist expression is not seen in the mesoderm underneath the presumptive cranial neural plate or in lateral plate mesoderm (Fig. 6M,N). Meox1, which marks the paraxial presomitic mesoderm and segmented somites of wild-type embryos, is also expressed in the paraxial mesoderm of the 8.5 dpc mutant embryos (data not shown). In early somite stage wild-type embryos, Cer1 is expressed in the anterior presomitic mesoderm and in two to three newly formed somites (see Fig. 8A). Cer1 expression is maintained in the paraxial mesoderm and the somites of the mutant embryo (see Fig. 8B,C). The absence of the heart tube in Mixl1^–/– embryos may reflect a defect in cardiomyocyte specification or a morphogenetic defect in heart tube formation. In wild-type embryos, Nkx2.5 is expressed strongly in the cardiogenic mesoderm and later in the myocardium (Lyons et al., 1995). In 8.0 dpc Mixl1^–/– embryos, Nkx2.5 is expressed ectopically in the mesoderm present bilaterally in the anterior region of the embryo. This finding suggests that cardiogenic mesoderm is specified in the absence of Mixl1 but the progenitors fail to undergo heart tube morphogenesis (Fig. 6O,P). Despite the abnormal development of the anterior region of the Mixl1^–/– embryos, expression of the prospective forebrain and the anterior neural ridge marker Six3 (Oliver et al., 1995) and Otx2 (Ang et al., 1993; Simeone et al., 1993) is correctly localized to the abnormal neural tube of the mutant embryos (Fig. 6Q,R and data not shown).

During embryogenesis, definitive endoderm cells originate from the primitive streak and intercalate into the visceral endoderm layer, displacing these cells to the vitelline yolk sac (Lawson and Pedersen, 1987). We examined this morphogenetic process in Mixl1^–/– embryos by assessing the expression of Pem (Lin et al., 1994) and α-fetoprotein (Afp) (Dziadek and Adamson, 1978), which mark the formation of the visceral endoderm. At 7.5 dpc, expression of Pem (not shown) and Afp in Mixl1^–/– embryos was similar to that in wild-type embryos, indicating the visceral endoderm has been displaced to the extra-embryonic region by the definitive endoderm in the mutant embryo (Fig. 7A-D).

The expression of Cer1 and Sox17 were examined to assess differentiation and patterning of the endoderm. Cer1 and Sox17 are first expressed in the anterior visceral endoderm of early streak stage embryos and in anterior definitive endoderm in the late-streak stage embryo. At the head fold stage, both genes are expressed in the definitive endoderm of the prospective foregut pocket. By the early somite stage Cer1 expression in endoderm wanes and Sox17 expression becomes confined to hindgut (Belo et al., 1997; Biben et al., 1998; Shawlot et al., 1998; Kinder et al., 2001; Kanai-Azuma et al., 2002). At 7.5 dpc Cer1 and Sox17 are expressed in the anterior endoderm of Mixl1^–/– embryos, in a pattern comparable with that of wild-type littermates (Fig. 7E-H). However, there is an evident reduction in the population of Sox17-expressing cells in the prospective foregut endoderm (Fig. 7G,H). By 8.5 dpc, Cer1 expression is no longer detected in the anterior endoderm of wild-type littermates but is retained in endoderm in the region of the presumptive foregut invagination of the Mixl1^–/– embryos (Fig. 8A-D). Sox17, which is expressed in the mid- and hindgut of the wild-type embryo, is ectopically expressed in the endoderm in the rudimentary structure that is reminiscent of the foregut portal (Fig. 8E-H). Taken together, these results indicate that definitive endoderm has been specified in Mixl1^–/– embryos. However, the lethality of the mutant embryos precludes further analysis of the morphogenesis and differentiation of the endodermal derivatives.

As Mixl1 is expressed in both the epiblast and extra-embryonic tissues during gastrulation (Pearce and Evans, 1999; Robb et al., 2000), it is possible that the proper inductive interactions do not exist between these tissues in the Mixl1^–/– embryos. In order to identify the tissues in which Mixl1 is required for axial mesoderm patterning, we generated mouse chimeras containing wild-type epiblast cells and Mixl1-deficient visceral endoderm and trophectodermal derivatives. Descendants of the wild-type R26.1 ES cells in the epiblast can be identified by the constitutive expression of a ROSA-26 lacZ transgene (Friedrich and Soriano, 1991; Varlet et al., 1997). These ES cells were injected into Mixl1^–/Δneo blastocysts obtained by mating Mixl1^+/– and Mixl1^+/Δneo mice. (This mating strategy was employed to facilitate genotyping of the chimeras – see Materials and Methods.) In the chimeras, wild-type ES cells contribute efficiently to embryonic, but rarely to extra-embryonic, lineages (Beddington and Robertson, 1989; Varlet et al., 1997). If Mixl1 expression in visceral endoderm is essential, wild-type ES cells will not rescue the defects of Mixl1^–/– host blastocysts. Five 8.5 dpc Mixl1^–/– blastocyst-derived chimeras that contained a greater than 50% contribution from wild-type (lacZ-positive) ES cells were phenotypically indistinguishable from chimeric embryos derived from wild-type or Mixl1^+/– host blastocysts and wild-type ES cells. The rescue of the mutant phenotype by the presence of wild-type cells in the embryo containing Mixl1-deficient extra-embryonic ectoderm and visceral endoderm indicates that loss of Mixl1 function in the extra-embryonic tissues is unlikely to have caused the disruption in embryogenesis.

The requirement for Mixl1 in embryonic development was studied in a reciprocal experiment. Two independent Mixl1^–/Δneo ES cell lines were derived by a second round of targeted mutation of the normal Mixl1 allele of a Mixl1^+/Δneo ES cell line using the original targeting construct (see Materials and Methods). Chimeras were produced by introducing Mixl1^–/Δneo ES cells (identified by the lack of lacZ expression) into Mixl1^+/+ ROSA-26 transgenic blastocysts. A proportion of the chimeric embryos with a high contribution of Mixl1^+/Δneo ES cells phenocopied the defective development of the Mixl1^–/– mutants, demonstrating that normal Mixl1 activity in visceral endoderm is not sufficient to compensate for the deficiency in the embryonic tissues (Fig. 9A). Chimeras with 20% or more wild-type cells showed partial to nearly complete rescue of developmental defects. Histological examination of these chimeras reveals that Mixl1^–/– cells can colonize most embryonic tissues, including the neural tube and the notochord. However Mixl1^–/– cells consistently display low level of contribution to the heart and foregut and are almost completely excluded from the endoderm of the midgut and hindgut (Fig. 9B-F)

During mouse development, absence of Mixl1 activity results in abnormal germ layer morphogenesis during gastrulation, the absence of a morphologically distinct node and formation of a condensed core of mesodermal tissues in place of the notochord. The Mixl1^–/– embryos do not form a heart, fail to elongate along the AP axis, develop a vestigial foregut and are mostly arrested in development by 9 dpc. Analysis of the expression of molecular markers for germ layer derivatives in the mutant embryos reveals that the progenitors of the neural, mesodermal and endodermal lineages are all present and the precursor tissues are properly patterned during gastrulation and early neurulation. In a few Mixl1^–/– embryos that developed to 9.5 dpc, a poorly organized notochord-like structure was formed in the midline of the grossly malformed neural tube and a mass of T-expressing tissues that resembles an overgrown tail bud was found at the caudal end of the axis. In some embryos, an outgrowth of T-expressing tissues protruded from the ventral side of the trunk as if the embryonic axis had bifurcated. The phenotypic consequences of Mixl1 mutation strongly suggest that Mixl1 function is crucial for the morphogenesis of germ layer derivatives and axis development.

The defective morphogenesis of the embryonic gut in the Mixl1^–/– embryo and the low contribution of the Mixl1^–/– cells to the gut endoderm of the chimeras indicate that Mixl1 function is essential for gut morphogenesis and the potency of the embryonic cells for endoderm differentiation. Members of the Mix/Bix family have been implicated in endoderm formation in both Xenopus and zebrafish. In the mouse, the definitive endoderm is formed by recruitment from the epiblast to the endodermal layer and replaces the visceral endoderm during gastrulation (Lawson et al., 1987; Lawson et al., 1991; Tremblay et al., 2000). In Mixl1^–/– embryos, progenitors of definitive endoderm are properly specified at gastrulation. However, by the early somite stage, only a small population of presumptive definitive endodermal cells characterized by Sox17 and Cer1 expression is found in the poorly developed foregut portal. The reduction in definitive endoderm is consistent with the finding that fewer Mixl1^–/– cells colonize the foregut and almost none are found in the hindgut of the chimera. Mixl1^–/– cells can, however, colonize all other types of embryonic tissues, including the axial mesendoderm, neural tube and heart of the chimera. These results suggest that the abnormalities of gastrulation and organogenesis in the mutant embryo are probably caused by the disruption of the non-autonomous functions of Mixl1, but Mixl1 plays a tissue-specific and autonomous role in the differentiation of embryonic cells into definitive endoderm. As Mixl1 expression ceases in the definitive endoderm after its allocation to the embryonic gut, it is likely that Mixl1 activity impacts on the endodermal progenitors while they are in the anterior primitive streak.

Studies in Xenopus and zebrafish have shown that Mix/Bix genes are downstream transcriptional targets in the TGFβ superfamily pathway that regulates mesendodermal patterning. In the mouse embryo, the TGFβ family member Nodal is a key regulator of the formation of the anterior primitive streak, node, midline structures and definitive endoderm (reviewed by Schier and Shen, 2000). LEFTY2 (EGFB – Mouse Genome Informatics) is an atypical member of the TGFβ superfamily that antagonizes NODAL signaling and restricts the range and duration of NODAL activity (Meno et al., 1999; Meno et al., 2001). At gastrulation, Lefty2 mutants, like Mixl1 mutants, have an expanded primitive streak with an excess of axial mesoderm leading to a variety of associated patterning defects (Meno et al., 1999). In 7.5 dpc Mixl1^–/– embryos, Lefty2 expression in the primitive streak was unaltered (data not shown).

Nodal expression is expanded in the Mixl1^–/– embryo, raising the possibility that Mixl1 may to regulate NODAL activity. Ectopic expression studies in Xenopus and zebrafish embryos have shown that NODAL and related factors can induce the formation of mesoderm and axis duplication (Jones et al., 1995; Toyama et al., 1995; Joseph and Melton, 1997). In mouse chimeras, Nodal^–/– cells are impaired in their ability to contribute to midline structures, indicating Nodal is required for midline morphogenesis (Varlet et al., 1997). Furthermore, Smad2, which is activated by Nodal, is required for definitive endoderm formation and embryos with a null mutation of the forkhead DNA-binding protein Foxh1, which binds Smad2 (and Smad3), exhibit abnormal patterning of the node, prechordal mesoderm, notochord and definitive endoderm (Tremblay et al., 2000; Hoodless et al., 2001; Yamamoto et al., 2001). Taken together, these observations showing disruption of the morphogenesis of mesendodermal structures following alteration of NODAL signaling suggest that Mixl1, like its Xenopus homologs, may influence the spatial or temporal pattern of NODAL signaling in the mouse embryo.

A role for the T-box gene Brachyury in notochord maintenance, axis elongation and specification of posterior mesoderm has been identified in many vertebrate species (Schulte-Merker et al., 1994; Wilson et al., 1995b; Conlon and Smith, 1999). In Xenopus, members of the Mix/Bix family have been shown to regulate expression of the Brachyury homolog Xbra (Latinkic et al., 1997; Tada et al., 1998; Casey et al., 1999; Latinkic and Smith, 1999). At the early gastrula stage, Xbra is expressed throughout the marginal zone of the embryo. In response to activin signaling, as gastrulation proceeds, transcripts are lost from involuting mesoderm, but persist in the notochord (Smith et al., 1991). Mix.1, which is induced by activin, downregulates Xbra transcription, in part through activation of Gsc (Artinger et al., 1997; Latinkic et al., 1997; Latinkic and Smith, 1999). In the Mixl1^–/– embryos, the Brachyury expression domain is strikingly expanded in the primitive streak and also in the abnormal midline structures present in older mutants, with weaker activity in the condensed ‘core’ cells of the midline structure and of the primitive streak. This suggests that, like Mix.1 in Xenopus, Mixl1 may be acting as a repressor of Brachyury expression in the murine embryo.

In zebrafish, the Mix-like gene, Bon, is responsible for the bonnie and clyde mutation (Kikuchi et al., 2000). Like Mixl1^–/– embryos, Bon mutants have cardiac bifida and a reduction in Sox17-expressing endodermal cells. However, they do not show the drastic disruption in axial morphogenesis seen in Mixl1-null embryos. Within the homeodomain, MIXL1 is more closely related to BON than any member of the Xenopus MIX/BIX family. Nevertheless, given that the overall the amino acid identity of BON and MIXL1 is only 25% and, taken together with the differences in phenotype of Bon and Mixl1 mutants, it is likely that the zebrafish homolog of Mixl1 remains to be discovered. Mixl1 is the only known mammalian member of the Mix/Bix family, and searching of the mouse and human genomes has not uncovered other potential homologs (A. H. H., A. G. E. and L. R., unpublished). By contrast, in Xenopus, seven Mix-like genes may be involved in the transcriptional regulation of mesoderm and endoderm development. Our study shows that, in the mouse, Mixl1 serves many functions of the Xenopus Mix/Bix genes to control the morphogenesis of mesendodermal structures, especially the node, notochord, axial mesoderm and the gut, and the differentiation of the definitive endoderm. We postulate that Mixl1 is the functional mammalian homolog of the members of the Xenopus Mix/Bix homeodomain family.

We are grateful to Peter Gruss, Hiroshi Hamada, Elizabeth Jones, Richard Harvey, Bernhard Herrmann, Gail Martin, Andrew McMahon, Virginia Papaioannou, Janet Rossant, Bill Shawlot Miles Wilkinson and Chris Wright for the provision of riboprobes; to Liz Robertson for generously providing R26.1 ES cells and Nodal^+/lacZ mice; to Steven Mihajlovic for assistance with histology; to Dennis Advani for the preparation of illustrations; to Bette Borobakas, Marilyn Ibrahim and Anita Steptoe for expert technical assistance; and to Peter Rowe for reading the manuscript. The project was supported by the National Health and Medical Research Council of Australia, the Cooperative Research Centre for Cellular Growth Factors, The Sylvia and Charles Viertel Charitable Foundation and the Anti-Cancer Council of Victoria. L. R. is a Viertel Foundation Senior Research Fellow, A. G. E. is an NHMRC Senior Research Fellow and P. P. L. T. is a NHMRC Senior Principal Research Fellow.