A targeted mouse Otx2 mutation leads to severe defects in gastrulation and formation of axial mesoderm and to deletion of rostral brain

The patterning and development of the vertebrate neural tube is a complex process involving both cell extrinsic and cell intrinsic events. Among the cell extrinsic events, the mesoderm in the organizer region, namely the dorsal blastopore lip in amphibians, Hensen’s node in birds and the node in the mouse, is able to induce neural differentiation in the surrounding ectoderm tissue (Spemann, 1938; Waddington, 1933; Beddington et al., 1994). Furthermore, in vitro experiments have demonstrated that the prechordal mesoderm and notochord, descendants of the organizer, can also induce and pattern the neural tube along the anteroposterior axis in Xenopus embryos (reviewed in Slack and Tannahill, 1992; Doniach, 1993; Ruiz i Altaba, 1994). Support for a role of the prechordal mesoderm in the induction of forebrain and midbrain has also come from the analysis of the phenotype of homozygous Lim1 mutant embryos, where lack of prechordal mesoderm cells in these mutant mice is suggested to be responsible for the subsequent loss of anterior brain tissues (Shawlot and Behringer, 1995). However, the role of notochord in anteroposterior (A-P) patterning of the neural tube is still in question, since embryos mutant for the winged-helix domain gene, HNF-3β, lack an organized node and notochord but show relatively normal A-P patterning of the CNS (Ang and Rossant, 1994; Weinstein et al., 1994).

A major contribution to the identification of cell intrinsic molecules responsible for neural tube regionalization has come from cloning of genes homologous to homeobox-containing genes within the HOM-C complex in Drosophila (reviewed in Lawrence and Morata, 1994), namely the Hox genes in mice. The function of these genes in the regional specification of hindbrain and vertebrae has been demonstrated using both loss-of-function and gain-of-function genetic experiments (reviewed by Krumlauf, 1994). Hox genes are not expressed in the forebrain and midbrain, suggesting that other classes of homeobox genes are involved in the development of these rostral regions. Recently, two new classes of homeobox genes, related to the Drosophila orthodenticle (otd) (Finkelstein et al., 1990) and empty spiracles (ems) gene (Dalton et al., 1989), have been cloned. These genes, otx1, otx2, emx1 and emx2, are expressed in nested domains in the forebrain and midbrain regions (Simeone et al., 1992a). Since otd and ems have been shown to participate in a regulatory network required for head formation in flies, it has been suggested that the conserved murine genes serve similar roles in the patterning of rostral brain in mice (Finkelstein and Boncinelli, 1994).

The mouse otd-related genes, Otx1 and Otx2, belong to the bicoid-class of homeobox genes and the amino acid sequence of their homeobox differ by two and three amino acids from that of the otd gene respectively (Simeone et al., 1993). These genes are expressed in overlapping domains in the anterior CNS with the domain of Otx1 expression encompassed within the Otx2 domain. Otx2 is already expressed by embryonic day 5.5 (E5.5), while the expression of Otx1 mRNA is not detected until early E8.0. Otx2 expression at E5.5 is widespread in the epiblast, which gives rise to the embryo proper. From the early primitive streak to headfold stages, Otx2 expression in the ectoderm becomes restricted to the anterior end of the embryo (Simeone et al., 1993, Ang et al., 1994). We and others have previously shown that this Otx2 expression in the anterior ectoderm depends on interactions with the underlying mesoderm at the anterior end of the embryo in mice and Xenopus (Ang et al., 1994; Pannese et al., 1995; Blitz and Cho, 1995). Otx2 expression was also found in anterior mesendoderm tissues, including axial mesoderm tissues, such as notochord and prechordal mesoderm (Ang et al., 1994; Simeone et al., 1995). The earlier expression of Otx2, compared to Otx1, and its expression in axial mesoderm tissues that possess neural patterning capabilities suggests that this gene could be involved in the patterning of anterior neural tissues.

To begin to dissect the roles of Otx2 in vivo, we have generated a homeobox deletion in the gene using homologous recombination in ES cells. This mutation results in early gastrulation defects. Homozygous Otx2^hd mutant embryos also lack the prechordal mesoderm and notochord precursors by the headfold stage. By E8.25, rostral deletion of the neural tube anterior to rhombomere 3 was clearly apparent. These defects are consistent with multiple roles for Otx2 in gastrulation and the patterning of rostral brain in mice.

A 1 kb mouse Otx2 partial cDNA probe was used to isolate two overlapping genomic clones, containing the entire coding region of the Otx2 gene from a 129SV/J genomic library. To construct the targeting vector, pPNTO2, a 1.7 kb BglII-Xba1 piece of 5′ homology was first subcloned into the BamHI-XbaI site of pPNT (Tybulewicz et al., 1991). The 3′ 4.4 kb homology StuI fragment was subsequently cloned into the XhoI site of the above vector, with the PGKneo and PGKtk cassettes in the opposite transcriptional orientation compared to the endogenous Otx2 gene.

The culture, selection and electroporation with NotI-linearized pPNTO2 of R1 ES cells was carried out as described (Wurst and Joyner, 1993). ES cell colonies that were resistant to 150 μg/ml of G418 and 2 μm gancyclovir were analyzed by Southern blotting for homologous recombination events. Genomic DNA from these cell lines was digested with SpeI and EcoRI and probed with a 2.0 kb HindIII-XbaI 3′ flanking probe and a 1.7 kb BglII-XbaI 5′ internal probe. Hybridization was carried out in 50% formamide, 200 mM phosphate buffer (pH 7.3), 1 mM EDTA, 10 mg/ml of BSA, and 7% SDS at 63°C.

Chimeras were generated by blastocyst injection and ES-morulla aggregation with targeted ES lines. Chimeric males were bred to CD1 females to establish F₁ heterozygotes. Embryos from F₁ heterozygous intercrosses were typed either by Southern analysis or by PCR of yolk sac DNA. To detect a 1.3 kb band present in the wildtype allele but deleted in the Otx2^hd allele, the following primers were used: sense strand (5′-ATGATGTCTTATCTAAAGCAAC-CGCCTTACG-3′) and antisense strand (5′-TCATTGGGT-CATCAGTATAAACCA-3′). The Otx2^hd mutant allele was detected as a 650 bp neo fragment using a set of primers corresponding to sense strand oligonucleotide in the neo gene (5′-ATCTCCTGT-CATCTCACCTTGC-3′) and antisense PGK poly(A) sequence (5′-ACCCCACCCCCACCCCCGTAGC-3′). Samples were amplified for 35 cycles (94°C 40 seconds; 55°C 1 minute; 72°C 1.5 minutes) for the wild-type allele and for 40 cycles (94°C for 1 minute; 65°C for 1 minute; 72°C for 2 minutes) for the mutant allele. Amplified bands were visualized by agarose gel electrophoresis and ethidium bromide staining.

Whole-mount in situ hybridization was performed as described previously (Conlon and Herrmann, 1993). For histology and section in situ hybridization, embryos were fixed overnight in 4% paraformaldehyde in PBS. They were then processed and embedded in wax and sectioned at 7 μm. Slides were then dewaxed, rehydrated and stained with hematoxylin and eosin. In situ hybridization of sections was performed as described by Guillemot and Joyner (1993). The RNA probes used were:-Flk-1 (Yamaguchi et al., 1993), a 1.0 kb Pem-1 cDNA containing the entire protein coding sequence, gsc (Blum et al., 1992), Lim1 (Barnes et al., 1994), Brachyury (Hermann, 1991), Mox-1 (Candia et al., 1992), HNF-3β (Ang et al., 1993), Emx-2 (Simeone et al., 1992b), a 2.1 kb BF-1 cDNA containing the entire protein coding sequence, Netrin-1 (Serafini et al., 1994); Krox-20 (Wilkinson et al., 1989b) and Hoxb-1 (Wilkinson et al., 1989a).

Whole-mount immunocytochemistry was performed according to the method of Davis et al. (1991), except that embryos were fixed in 4% paraformaldehyde in PBS for 2 hours. The En-2 antiserum was used at a dilution of 1:50. The secondary antibody was peroxidase-coupled goat anti-rabbit (Jackson laboratory) which was used at a dilution of 1:200.

A positive/negative targeting vector containing 1.7 kb of 5′ and 4.4 kb of 3′ genomic sequences (Fig. 1A), was designed to delete the homeodomain region of Otx2, and replace it by the PGKneo cassette from the pPNT vector (Tybulewicz et al. 1991). Therefore the PGKneo insertion should truncate the Otx2 coding sequence immediately after the first exon which contains 32 amino acids. This allele is referred to as the Otx2^hd allele.

The linearized targeting vector was electroporated into ES cells (Nagy et al., 1993) and clones were selected for resistance to G418 and gancyclovir. 350 double resistant clones, obtained after electroporation of the pPNT02 targeting vector, were analyzed by Southern blot using a 5′ internal and a 3′ external probe. Five cell lines, including 5-23, and TE123, yielded the 3.5-kb SpeI band and 13.2-kb EcoRI band expected from a homologous recombination event (Fig. 1B). Using a genomic probe spanning exon 2 (Fig. 1A), the deletion of the homeobox region was confirmed in the genomic DNA of homozygous Otx2^hd mutants by Southern blot analysis (data not shown).

The 5-23, TE123 and C-12 targeted ES cell lines were used to generate chimeras by either ES cell-morulla aggregation (Nagy et al., 1993) or by blastocyst injection. These chimeras transmitted the mutation to their progeny. Mice generated from all three lines showed identical phenotypes. The analysis reported here is primarily from the 5-23 line. All analyses were carried out on a mixed CD1/129 background.

When heterozygous animals were crossed with wild-type CD1 females and their progeny was genotyped at 3 weeks of age, heterozygous mice were obtained with at a frequency of 141/311, less than the expected 50% ratio. These heterozygous mice appeared normal and were fertile. However, when offsprings from heterozygous intercrosses were harvested at E9.5 ane E10.5, a small fraction (8/149) of normal size embryos showed an open neural tube defect at the forebrain and midbrain levels. Caudal to the midbrain region, these embryos looked identical to wild-type embryos (data not shown). All these embryos have been genotyped to be heterozygous animals. The Otx2^hd mutation thus result in a heterozygous phenotype that is weakly penetrant on the CD1/129 background, and this heterozygous phenotype may explain the slightly lower number of heterozygous animals obtained at 3 weeks of age. This phenotype will be analyzed in more detail elsewhere.

When intercrosses of Otx2^hd heterozygous animals were analyzed at birth, no homozygous newborn animals were found, indicating that Otx2 is required for embryonic development (Table 1). To characterize the embryonic lethality, we analyzed litters from heterozygous intercrosses from E7.25 to E10.5. Homozygous mutant embryos were present between E7.25 and E9.5 at roughly the expected frequency of 25%, however, at E10.5, the proportion of mutants obtained declined to 17% and these mutant embryos were either severely growth retarded or being resorbed (Table 1). Thus, the Otx2^hd mutation leads to embryonic lethality around E10.5. The results from a typical genotyping analysis of yolk sacs from E8.5 embryos generated from Otx2^hd heterozygous intercrosses is illustrated in Fig. 1C.

By E7.25-E7.5 (mid-to late-streak stage), homozygous embryos were morphologically distinguishable from normal embryos. The abnormal looking embryos were smaller in the embryonic portion than their littermate (Fig. 2A,B), and these embryos were confirmed to be homozygous mutant by PCR analysis (Table 1). Sections through mutant embryos at this stage demonstrated clearly that these embryos had initiated gastrulation and mesoderm had formed all around the embryo. Embryonic ectoderm, mesoderm and endoderm cells are present in mutant embryos; however some of the ectoderm cells appeared disorganized and lost their characteristic epithelial structure (Fig. 3A-C).

In the extraembryonic region, the yolk sac looked abnormal and displayed a characteristic ruffled appearance in the endoderm layer (Fig. 3B,C), which reflected a reduced, discontinuous underlying extraembryonic mesoderm layer. The yolk sac was also constricted in the position normally occupied by the amnion, thus isolating the extraembryonic from the embryonic region (Fig. 3A-C). The allantois was poorly developed and formed a round ball of cells, in contrast to its fairly long and extended appearance in normal embryos at this stage. It was also often disconnected from the embryonic portion of the conceptus (Fig. 3C and data not shown).

At E7.75 (headfold stage), the headfolds forming at the rostral end of wild-type embryos were not apparent in homozygous mutant embryos (Fig. 2C, D). At E8.5 (somite stage), variations in the phenotype of the mutants were found. We have divided the mutants at this stage into a less severe class, which includes embryos enclosed in the yolk sac, and a more severe class whereby the embryos are completely excluded from the yolk sac (Fig. 2G). In the less severely affected class of mutants, there was also fairly good development of somites posteriorly and a rudimentary heart (Fig. 2F and data not shown). At the anterior end, the neural tube looked extremely abnormal with either numerous folds or a single fused structure instead of open neural folds (Figs 2F, 5E). Compared to wild-type or heterozygous embryos, no forebrain or midbrain tissue could be identified suggesting that deletion of neural tissue had occurred anteriorly. The somites were not always normal; in some embryos they were fused at the midline or were irregularly shaped (Fig. 5E). The more severely affected homozygous mutant embryos were smaller than the first class of mutants, wrapped up anteriorly in definitive endoderm and showed improper segmentation of mesoderm into somites (Fig. 2G). Some of the mutants in this class were very thin and spiral-shaped and showed no sign of organogenesis and segmentation of mesoderm into somites (data not shown). The two different classes of mutants occurred at approximately the same frequency.

The small size and abnormal morphology of E7.25 embryos demonstrated that defects had already occurred in homozygous mutants at this early stage. To determine if defects occurred in the primitive streak, we analyzed the expression of the genes goosecoid (gsc) and Brachyury. goosecoid is a homeobox gene expressed in the anterior primitive streak region in mid-to late-streak stage embryos (Fig. 4A and Blum et al., 1992). goosecoid-expressing cells were found ectopically located in the proximal region of homozygous mutants in one case and was absent in three other embryos analysed (Fig. 4B and data not shown). Previous studies have demonstrated that gsc is expressed transiently in the anterior primitive streak in wild-type embryos (Blum et al., 1992). Since it is difficult to accurately stage the homozygous mutant embryos due to their abnormal morphology, we cannot distinguish whether the mutant embryos that fail to express gsc do so because they are older than the late-streak stage or because Otx2 is required for the maintenance of gsc expression at this early stage. The Brachyury gene is expressed along the entire proximal-distal extent of the primitive streak at the mid-to late-streak stage (Hermann, 1991). In some homozygous mutants, Brachyury expression was only found in the proximal region in a similar position as gsc-expressing cells in Otx2^hd homozygous mutants (data not shown), while others showed almost complete extension but a thickening of the region of the primitive streak labeled (Fig. 4F). Together, these results demonstrate that Otx2^hd homozygous mutants show severe early gastrulation defects, characterized by a lack of proper primitive streak organization.

To assess the presence in mutants of the different extraembryonic tissues, we analysed the expression of the extraembryonic markers Flk1 and Pem. Flk1 is expressed in wildtype E7.5 embryos in nascent mesoderm of the proximal part of the primitive streak, extraembryonic yolk sac mesoderm, and proximoanterior embryonic mesoderm (Fig. 3E and Yamaguchi et al., 1993). Flk-1 expression in homozygous mutant embryos at this stage was detected in all these tissues indicating that specification of mesoderm derived from the proximal primitive streak is occurring normally. However, in some cases there was abnormal accumulation of mesoderm cells in the proximal region between the embryonic and extraembryonic regions (Fig. 3H). Pem is a homeobox gene expressed specifically in extraembryonic tissues (Fig. 3F and Lin et al., 1994). All the Pem-expressing tissues (yolk sac, chorion, primitive endoderm and ectoplacental cone) could be recognized in the mutant embryos, but the yolk sac looked abnormal in the mutants between the embryonic and extraembryonic regions (Fig. 3I).

To determine if the different populations of embryonic mesoderm tissues were present in mutants at E8.5, the markers Brachyury, Mox-1 (Candia et al., 1992) and Lim1 (Barnes et al., 1994) were used to identify axial, paraxial and lateral mesoderm cells, respectively. Both Mox-1 and Lim-1 expression could be detected in mesoderm cells of mutant E8.5 embryos, indicating that the homozygous mutants contain paraxial and lateral mesoderm cells (Fig. 4J,L). Mox-1 expression in the somites in posterior regions was similar in homozygous mutants and their littermates (Fig. 4I,J). Anteriorly however, Mox-1 expression spread across the midline, as expected from the observed fusion of somites. In four out of five embryos analyzed, Brachyury expression in homozygous mutants was absent in the anterior midline, at the normal position of the notochord (Fig. 4M,O). In one case, Brachyury-expressing cells were present anteriorly but appeared to bud off and diverge from the axial notochord (Fig. 4N). Thus, axial mesoderm cells are severly affected in Otx2^hd homozygous embryos, while paraxial and lateral mesoderm cells do develop more normally. The appearance of fused somites is consistent with axial mesoderm defects since it has also been seen in other notochordless mouse mutants and notochordless chick embryos (Dietrich et al., 1993; Ang and Rossant, 1994; Teillet and le Douarin, 1983; Rong et al., 1992).

We next tested whether axial mesoderm defects occurred at earlier stages than those examined above. Lim1, Brachyury, and HNF-3β (Ang et al., 1993; Monaghan et al., 1993; Sasaki and Hogan, 1993) are all expressed in the node and headprocess in wild-type embryos at E7.5 (Fig. 4C,E,G). HNF-3β and Lim1 are also expressed in midline cells anterior to the notochord known as prechordal mesoderm cells (Fig. 4C,G – arrows). In homozygous mutants, all three genes were expressed in the node and in a few cells extending anteriorly at a short distance from the node (Fig. 4D,F,H). This was in sharp contrast to wild-type embryos in which labelled headprocess cells had migrated much further anteriorly (Fig. 4C,E,G). In particular, the anterior-most midline expression of Lim-1 and HNF-3β in the prechordal mesoderm cells were missing in the mutants (arrows in 4D and H). These results suggest that the defects in axial mesoderm cells observed at E8.5 are due to a failure in the generation of these cells at earlier stages. Prechordal mesoderm fail to develop and notochord development is incomplete in E7.5 Otx2^hd homozygous embryos.

To determine precisely how much rostral brain tissue was deleted in mutant embryos, several A-P region-specific neural markers were used to characterise Otx2^hd homozygous embryos. For these studies, we used only E8.25-E8.5 mutant embryos from the less severe class, where the head is free of underlying yolk sac and is not surrounded by endoderm tissue. At E8.5, the forkhead-related gene, BF1 (Tao and Lai, 1992), and the homeobox gene, Emx-2, which are normally specifically expressed in the telencephalon, and part of the diencephalon for Emx-2 (Simeone et al., 1992b), were not expressed in the neural tube of homozygous mutant embryos, thus demonstrating the absence of these tissues in mutant embryos (data not shown). Lack of HNF-3β and netrin1 (Serafini et al., 1994), in the rostral portion of the neural tube of mutants embryos demonstrated the absence of ventral midbrain tissue (Fig. 5B,E and data not shown). Furthermore, we analyzed mutant embryos for En protein expression by whole-mount antibody staining. No expression of En was observed in three E8.5 mutant embryos, indicating that the midbrain and anterior hindbrain regions were missing in homozygous mutants (Fig. 5A,D). These results demonstrate that forebrain, midbrain and anterior hindbrain are missing in Otx2^hd homozygous mutants.

To further characterize at what level in the hindbrain deletion had occurred, homozygous E8.5 mutants of the less severely affected group were analyzed for expression of two hindbrain markers, Hoxb-1 (Wilkinson et al., 1992a) and Krox-20 (Wilkinson et al., 1992b). Hoxb-1 was expressed in rhombomere 4 and the posterior spinal cord in both mutant and wild-type animals (Fig. 5C,F). Krox-20 was expressed in two bands across the neural tube in mutants as in wild-type or heterozygous embryos. However in mutants the first band of expression occurred close to or at the anterior-most end of the embryo (Fig. 5B,E and data not shown). These results demonstrate that in homozygous mutants, neural tissues anterior to rhombomere 3 have been deleted.

Deletion of the homeobox region of the Otx2 gene produced an embryonic lethal phenotype in mice. The phenotype was characterized by severe gastrulation defects, absence of prechordal mesoderm, reduced notochord production and severe anterior truncations. All mutant embryos were severely growth retarded or resorbed by E10.5, presumably because the separation of embryonic and extraembryonic regions resulted in defective yolk sac circulation. Given that Otx2 is broadly expressed at the pre-streak and early streak stage embryos, and is later restricted to anterior structures, this phenotype implicates Otx2 in several different aspects of early postimplantation patterning.

In the mid-to late-streak stage mutant embryos, the incomplete elongation of the primitive streak, the accumulation of mesoderm cells between embryonic and extraembryonic region of the embryo and the ectopic location of gsc-expressing cells in the proximal region, indicate that Otx2 is required in some manner for the normal organization of the streak. Otx2^hd homozygous mutants also showed defects in extraembryonic structures. The ruffling in the yolk sac is most likely caused by defects in the migration and accumulation of extraembryonic mesoderm.

The smaller size of the embryonic portion of mutant embryos at the mid-streak stage suggests that proliferation of the epiblast tissue could also be affected or delayed in absence of Otx2. To examine whether cell proliferation in the epiblast is affected at this early stage, BrdU incorporation experiments will be performed. In addition, the severe constriction observed between embryonic and extraembryonic regions in Otx2^hd mutants suggest that other processes besides proliferation are also affected in the mutants. A very similar extraembryonicembryonic constriction was also observed in HNF-3β mutants (Ang and Rossant, 1994; Weinstein et al., 1994) and to a lesser extent in Lim-1 mutants (Shawlot and Behringer, 1995), both of which also affect node and notochord development, suggesting that alteration of these processes is somehow involved in the generation of the constriction.

We have demonstrated that the prechordal mesoderm is missing in Otx2^hd homozygous mutant embryos at the headfold stage using Lim-1 and HNF-3β genes as markers for this tissue. These results demonstrate that Otx2 is an essential regulator of prechordal mesoderm development. Recent phenotypic studies on Lim1 homozygous mutants have also demonstrated an essential role for the Lim1 gene in prechordal mesoderm development. Since Otx2 and Lim1 are both expressed in prechordal mesoderm in headfold stage embryos and are required for its development, it will be interesting to study whether these two genes might function in the same genetic pathway.

Defects in notochord development were also observed in Otx2^hd homozygous embryos. In the late streak and headfold stage embryos, there was limited midline extension of the head process. In later embryos, the notochord formed only at the most posterior end in some cases, while in other cases, notochord cells were present but misplaced lateral to the midline. Anterior notochord normally express Otx2 at the midbrain level (Ang et al., 1994), perhaps explaining the anterior notochord defects observed in mutant embryos. However abnormalities in the development of notochord cells, which normally do not express Otx2 at the hindbrain and trunk levels, cannot be readily explained. It seems likely that the failure of later notochord development reflects earlier defects in headprocess development. Interestingly, studies of Otx2 in Xenopus and chick have demonstrated Otx2 expression in the organizer tissue of these species, namely dorsal blastopore lip and the Hensen’s node respectively (Pannese et al., 1995; Blitz et al., 1995; Bailly-Cuif et al., 1995). Thus, by analogy to the situation seen in these species, Otx2 might be expressed in the presumptive node at the anterior end of the primitive streak in early streak stage mouse embryos (Lawson et al., 1991), this expression being obscured by the simultaneous widespread expression of Otx2 in the epiblast at this stage. Lost of Otx2 in this structure may be responsible for the notochord defects at later stages.

Otx2^hd homozygous mutants that failed to become enclosed by visceral yolk sac by E8.5 due to the severe constriction between embryonic and extraembryonic regions, were extremely abnormal and in some cases seemed to lack any axial organization. However, the less severely affected embryos that were still enclosed in the yolk sac, allowed delineation of specific anterior defects in the mutants.

By E8.25, deletions in the anterior neural tube rostral to rhombomere 3 were clearly apparent in these homozygous mutants. Loss of forebrain, midbrain and anterior hindbrain was demonstrated using early molecular markers for these tissues such as BF1, Emx-2 and En. This phenotype could have been predicted, given the anterior expression of Otx2 and its relationship to the Otd gene in Drosophila. Loss of function mutations in otd in flies lead to deletion of anterior head structures (Finkelstein et al., 1990). These results support evolutionary conservation of the function of these genes in head development in flies and mice.

This interpretation of the phenotype assumes that Otx2 acts autonomously in the anterior CNS and is required for the specification of these regions. However, this hypothesis cannot readily explain why structures posterior to the expression domain of Otx2 are also deleted. Otx2 expression caudally marks the mid-hindbrain boundary. However, deletions of the hindbrain region, containing the En-expressing domain of the metencephalon and close to the anterior border of rhombomere 3, occur in mutant embryos. This could either be due to an early expression of Otx2 in cells fated to become hindbrain at the late-streak stage or to a dependence of anterior hindbrain development on more rostral neural tube. Comparison of the fate-map studies of late streak stage embryos with the domain of Otx2 expression at this stage suggest that this domain does not include hindbrain territories (Tam, 1989). However a more extensive study is necessary to exclude this possibility.

Alternatively, loss of anterior neural tissue could be an indirect consequence of the loss of notochord and prechordal mesoderm in the Otx2^hd homozygous mutants. Classical embryological studies have demonstrated a role for both these tissues in the induction and patterning of the neural tube (reviewed by Slack and Tannahill, 1992; Doniach, 1993; Ruiz i Altaba, 1994). A role for the prechordal mesoderm was supported by the phenotype of Lim1 null homozygous mutants (Shawlot and Behringer, 1995). In contrast, a role for the notochord tissue in A-P neural tube patterning has not been confirmed by mutant studies in mice. Mouse embryos lacking the HNF-3β gene do not develop a notochord, yet they showed expression of rostral brain markers (Ang and Rossant, 1994; Weinstein et al., 1994). Together, these results suggest an essential role for the prechordal mesoderm, but not the notochord, in patterning of the anterior neural tube in mice. Furthermore, the Lim1 mutant phenotype is remarkably similar to that of the Otx2^hd homozygous mutants, in that deletions of the anterior neural tube occur at about the same anteroposterior level (Shawlot and Behringer, 1995). Thus, loss of rostral brain tissues observed in Otx2^hd homozygous mutants could be a consequence of the lack of prechordal mesoderm development.

To distinguish between these hypotheses, we are planning to recombine mesoderm tissue from homozygous mutant Otx2^hd embryos with ectoderm tissue from wild-type embryos to determine if the mutant mesoderm tissue can induce expression of anterior neural markers such as En genes. The reverse experiment of recombining ectoderm tissue from Otx2^hd homozygous embryos and mesoderm from wild-type embryos will be performed to determine if the Otx-2 negative ectoderm tissue can respond to inducing signals from the mesoderm. The use of the in vitro tissue recombination assay (Ang and Rossant, 1993, 1994) to analyze Otx2^hd mutants should allow us to dissect the roles of Otx2 in the ectoderm and mesoderm tissues at the late-streak stage.

In conclusion, Otx2^hd homozygous mutants show complex defects in gastrulation, axial mesoderm and rostral brain development that implicate Otx2 in the pathways of primitive streak organization, axial mesoderm development as well as anterior head development. Further experiments will be required to dissect out these different roles.

We would like to thank Dr André Dierich and Dr Marianne Lemur for generating the germline chimeras with the TE123 ES cell line. We are grateful to Dr François Guillemot for performing sectioned in situ hybridization experiments, and for critical reading of the manuscript and to Isabelle Tilly and Christine Engel for excellent technical assistance. We wish to thank Drs R. Behringer, B. G. Hermann, K. Kaestner, R. Krumlauf, A. P. McMahon, E. M. de Robertis, M. Tessier-Lavigne, D. G. Wilkinson and C. V. E. Wright for gifts of probes. S.-L.A. was supported by a fellowship from the Fondation de la Recherche Médicale, O. J. holds a Univerisity of Toronto Open fellowship, and J. R. is a Terry Fox Research Scientist of the National Cancer Institute of Canada and an International Scholar of the Howard Hughes Institute. This work was supported by funds from the Institute National de la Santé et de la Recherche Médicale, the Centre National de la Recherche Scientifique, and the Centre Hospitalier Universitaire Régional and grants from the Medical Research Council of Canada and Bristol-Myers Squibb Ltd.