ABSTRACT
We have replaced part of the mouse homeogene Otx2 coding region with the E. coli lacZ coding sequence, thus creating a null allele of Otx2. By 9.5 dpc, homozygous mutant embryos are characterized by the absence of forebrain and midbrain regions. From the early to mid-streak stages, endomesodermal cells expressing lacZ fail to be properly localized anteriorly. In the ectodermal layer, lacZ transcription is progressively extinguished, being barely detectable by the late streak stage. These data suggest that Otx2 expression in endomesoderm and ectoderm is required for anterior neuroectoderm specifi-cation. In gastrulating heterozygous embryos, a post-tran-scriptional repression acts on lacZ transcripts in the ectoderm, but not in the external layer, suggesting that different post-transcriptional mechanisms control Otx2 expression in both layers.
INTRODUCTION
The identification of the murine Hox genes, and the demon-stration that inactivation of some of them leads to homeotic transformations, have provided major clues to understand the establishment of the body pattern in the vertebrate trunk and hindbrain (Krumlauf, 1993, 1994). By contrast, the genetic mechanisms underlying the development of more anterior regions, fated to the adult forebrain and midbrain, are still obscure. However, a number of regulatory genes, specifically expressed in morphogenetically defined regions of the devel-oping forebrain and midbrain, have been identified (Rubenstein et al., 1994). As in more posterior regions, most of them were isolated as vertebrate homologs of Drosophila genes involved in insect head formation.
In Drosophila, the head region consists of seven segments, identified by both the repetitive pattern of expression of two segment polarity genes, engrailed and wingless, and morpho-logical considerations (Diederich et al., 1991; Schmidt-Ott and Technau, 1992). Three genes, orthodenticle (Otd), empty spiracles and button head, whose mutations specifically affect some of these segments, were isolated by large-scale genetic screens (Finkelstein and Perrimon, 1990, 1991; Finkelstein et al., 1990; Cohen and Jürgens, 1990). The mouse gene Otx2 was identified as one of the two murine homologs of Otd. Both Otx2 and Otd contain highly related homeodomains belonging to the bicoid class and, like its Drosophila cognate, Otx2 shows an expression pattern mainly restricted to specific head regions during embryogenesis (Simeone et al., 1992a). Thus, by 10.5 dpc, the Otx2 transcription domain covers most of the forebrain and midbrain neuroepithelium, with a sharp boundary at the midbrain-hindbrain junction. Moreover, Otx2 is transcribed in the epiblast as early as 5.75 dpc and its expression domain pro-gressively regresses to anterior regions during gastrulation (Simeone et al., 1993). This pattern of expression is largely consistent with the results of whole-mount hybridization experiments recently reported in chick and Xenopus (Bally-Cuif et al., 1995; Pannese et al., 1995). Additionally, in these two latter species, two distinct phases of Otx2 transcription can be distinguished during gastrulation. In chick, during primitive streak elongation, the transcripts become progressively restricted to anterior regions but appear mainly associated with Hensen’s node. At the streak maximal extension (stage HH4), Otx2 expression is restricted to the node. In a second phase, transcripts become highly abundant in anterior mesendoderm and, at a slightly later stage, in anterior neuroectoderm (Bally-Cuif et al., 1995). Likewise, in Xenopus gastrulae, Xotx2 is expressed at stage 10.25 in dorsal bottle cells and in cells fated to prechordal mesendoderm at the dorsal lip of the blastopore. The transcripts can be detected in the presumptive anterior neu-roectoderm only later, at stage 10.5 (Pannese et al., 1995).
Both the chick Hensen’s node and the Xenopus dorsal blastopore lip, or Spemann’s organizer, show a remarkable property : when transplanted to other locations of the embryo, they are able to induce the formation of a complete second neural axis (Spemann and Mangold, 1924; Waddington, 1932).
Furthermore, in avians, transplantation experiments have established that the inductive properties of the node change throughout gastrulation. Both anterior and posterior nervous system can be induced by young nodes (from stages HH2 to HH4), whereas older nodes have a reduced inductive capabil-ity and generate only posterior neural structures (Storey et al., 1992). The competence of the epiblast to respond to neural induction also declines after stage HH4 (Storey et al., 1992). The stages at which Hensen’s node transplants are able to induce a neural axis exhibiting anterior characteristics therefore correspond precisely to the time period when Otx2 is expressed in the node (up to stage HH4). These Otx2 expression patterns in chick and Xenopus gastrulae suggest a conserved role for this gene in the specification of anterior neu-roectoderm in vertebrates, in line with its highly conserved nucleotide sequence during evolution.
To address directly the role of Otx2 during embryogenesis, we have inactivated Otx2 in mice, replacing part of its coding region by a lacZ reporter gene. At 9.5 dpc, the Otx2−/− mutation results in the absence of anteriormost regions of the neural tube, corresponding to the midbrain and forebrain regions. Fur-thermore, homozygous Otx2−/− embryos exhibit marked abnor-malities at the early to mid-streak stages of gastrulation, sug-gesting that early steps in anterior neuroectoderm specification are affected.
MATERIALS AND METHODS
Construction of an Otx2 targeting vector
A λ recombinant containing the mouse Otx2 gene was isolated from a 129/Sv genomic library using as a probe an Otx2 cDNA fragment previously characterized (Simeone et al., 1992a). A 6.5 kb SmaI fragment, located 220 pb upstream from the translation initiation site, and a 1.2 kb NsiI fragment, located 240 pb upstream from the stop codon, were excised from this recombinant phage and cloned respec-tively in the XmnI and ApaI sites of the mutagenesis pGN vector (Le Mouellic et al., 1990). Both fragments were inserted in the same ori-entation as the neomycin resistance gene and the lacZ reporter gene.
Transfection of ES cells and selection of targeted clones
HM-1 embryonic stem (ES) cells (Magin et al., 1992) were cultured on neomycin-resistant mouse embryonic fibroblasts, according to Robertson, 1987. 10 μg of the pGN31 targeting vector were linearized by digestion of the unique KpnI restriction site, and electroporated into 2×107 ES cells resuspended in 750 μl HeBS medium (20 mM Hepes pH 7.05, 137 mM NaCl, 5 mM KCl, 0.7 mM Na2HPO4, 6 Mm glucose), at 200 V, 960 μF. Positive selection was carried out for 11 days at 350 μg/ml G418. Resistant colonies were picked and DNA was extracted from a fraction (1/5) of the cells for a PCR test of homologous recombination event. The primers, specific respectively for the targeting vector (sense primer : 5′-TGCTGTGTTCCA-GAAGTGTT-3′, located immediately downstream from the neomycin resistance gene) and for the genomic Otx2 locus (antisense primer : 5′-CTGATTGAGATGGCTGGTAACAGC-3′, located 50 pb downstream from the 3′ NsiI fragment) were used. As shown in Fig. 1a, only homologous recombination events result in the amplification of a 1.6 kb band, whose identity was confirmed by hybridization with the NsiI 3′ fragment. 40 cycles were performed (denaturation : 1 minute, 95°C, hybridization : 2 minutes, 65°C, elongation : 6 minutes, 74°C) in a 50 μl volume containing 50 mmol KCl, 2 mM MgCl2, 1 mmol DTT, 15 mM Taps-HCl, 0.2 mM dATP, dCTP, dGTP, dTTP each pH 9.3, 2.5 u Taq DNA polymerase. Positive clones were expanded before freezing or DNA extraction for Southern blot analysis. The probes used included fragments internal (1.3 kb BamHI-HindIII fragment containing the pGN31 neo gene) or external (120 pb NsiI-HindIII located immediately downstream from the 3′ NsiI 1.6 kb fragment) to the targeting vector.
Inactivation of the mouse Otx2 locus by homologous recombination. (a) The cDNA structure (second line) is shown above the restriction map of the wild-type locus (third line), open boxes and broken lines corresponding to exons and introns respectively. The homeodomain is shaded. In the targeting vector (fourth line), two regions homologous to genomic DNA (hatched boxes) flank a lacZ reporter gene and the neomycin resistance gene (black boxes), in the same orientation (indicated by a thin arrow below the coding sequences). Homologous recombination results in the deletion of 2.8 kb including most of the homeodomain (fifth line). The transcription initiation site of Otx2 was mapped in 10.5 dpc embryos and is located about 700 pb upstream from the Otx2 translation initiation site (thick arrow). Vertical arrowheads point to the positions of the PCR primers used to detect homologous recombination events. Thin lines (first and last lines) show HindIII fragments detected by Southern blot analysis using probes external to the vector or in the neomycin gene (black bars above the HindIII fragments) and corresponding to the wild-type (5.3 kb) or mutated (3.0 kb) allele. N, NsiI; S, SmaI; H, HindIII. (b) Southern blot analysis of a targeted cell line (+/−) or of wild-type (wt) HM-1 ES cells using a probe external to the mutagenesis vector. (c) Genotyping of heterozygous (+/−) or wild-type (wt) mice by PCR amplification of fragments specific for the wild-type (223 bp) or the mutated allele (429 bp) with allele-specific primers.
Inactivation of the mouse Otx2 locus by homologous recombination. (a) The cDNA structure (second line) is shown above the restriction map of the wild-type locus (third line), open boxes and broken lines corresponding to exons and introns respectively. The homeodomain is shaded. In the targeting vector (fourth line), two regions homologous to genomic DNA (hatched boxes) flank a lacZ reporter gene and the neomycin resistance gene (black boxes), in the same orientation (indicated by a thin arrow below the coding sequences). Homologous recombination results in the deletion of 2.8 kb including most of the homeodomain (fifth line). The transcription initiation site of Otx2 was mapped in 10.5 dpc embryos and is located about 700 pb upstream from the Otx2 translation initiation site (thick arrow). Vertical arrowheads point to the positions of the PCR primers used to detect homologous recombination events. Thin lines (first and last lines) show HindIII fragments detected by Southern blot analysis using probes external to the vector or in the neomycin gene (black bars above the HindIII fragments) and corresponding to the wild-type (5.3 kb) or mutated (3.0 kb) allele. N, NsiI; S, SmaI; H, HindIII. (b) Southern blot analysis of a targeted cell line (+/−) or of wild-type (wt) HM-1 ES cells using a probe external to the mutagenesis vector. (c) Genotyping of heterozygous (+/−) or wild-type (wt) mice by PCR amplification of fragments specific for the wild-type (223 bp) or the mutated allele (429 bp) with allele-specific primers.
Generation and genotyping of chimeric and mutant mice
10-15 ES cells were microinjected into C57Bl/6 blastocysts. Injected blastocysts were reimplanted in the uterine horn of pseudopregnant recipient females. Chimeric animals were back-crossed to B6/D2 mice and germ-line transmission was scored by the presence of agouti coat pigmentation. Heterozygous offspring were identified by two PCR reactions, one to detect the presence of the lacZ gene, and one to detect the presence of Otx2 sequences lost during homologous recombination. Tail tips were incubated in lysis buffer (50 mM Tris pH 8.0, 100 mM EDTA, 100 mM NaCl, 1% SDS, 0.6 mg/ml proteinase K) overnight at 55°C, phenol-chloroform extracted, ethanol precipitated and redissolved in 10 mM Tris-HCl, 1 mM EDTA pH 8.0 at a final concentration of 0.2-1.0 μg/μl. The presence of a mutated allele was detected using two primers specific for the lacZ sequence (sense primer : 5′-GCGTTGATTTAACCGCC-3′, antisense primer : 5′-CAGTTTACCCGCTCGCTAC-3′) and the presence of a wild-type allele was detected using two primers located in deleted portions of the first and the second intron (sense primer : 5′-GTCACTGA-GAAACTGCTCCC-3′; antisense primer : 5′-GTCTCTACATCTGC-CCTACC-3′). 30 cycles (denaturation : 1 minute, 95°C, annealing : 1 minute, 65°C; elongation : 30 seconds, 74°C) were performed and the amplified products, respectively 429 bp and 223 bp long, were separated by 2% agarose gel electrophoresis.
Histology
Mid-day of the day of the vaginal plug was considered as 0.5 dpc in the timing of embryos. When removed from the decidua, embryos were staged according to Ang and Rossant (1993) and genotyped as described above, after DNA extraction from the yolk sac or from the embryo itself after photographing. For in situ hybridizations, up to 7.5 dpc, the decidua were sectioned and the genotypes were determined by hybridizations using probes located in the lacZ sequence (lacZ probe) and in the deleted part of the Otx2 gene (Otx2-del probe). For wax sections, embryos were fixed in 4% PFA at 0°C for 10-30 minutes depending on the stage, dehydrated in ethanol and transferred to 100% xylene. After clearing, they were incubated in a mixture of 50% xylene/50% paraplast at 60°C for 30 minutes-1.0 hour, followed by three 1.0 hour incubations at 60°C in paraplast, mounted and sectioned (12 μm). The sections were collected on gelatin-coated slides, dewaxed, rehydrated and stained with phloxin or thionin.
In situ hybridization of sections and whole-mount hybridization
Whole-mount hybridizations of 7.5 dpc embryos were carried out following the protocol described by Wilkinson, 1992, using digoxy-genin-UTP labelled single-stranded RNA probes. In situ hybridiza-tions on sections of embryos were carried out as described by Wilkinson and Green, 1990, using [35S]CTP labelled single-stranded RNA probes.
The following probes were used in the in situ hybridization exper-iments : Otx2, Otx1, Evx1, Emx2 and cripto probes were previously described (Simeone et al., 1992b, 1993; Dono et al., 1993); the Hoxb-1 probe is a PCR-fragment spanning the region between nucleotides 911 and 1263 of the sequence reported in Frohman et al., 1990; the Shh probe is a PCR-fragment spanning the region between aminoacid 258 and the stop codon (Echelard et al., 1993); the Brachyury probe cor-responds to the one described in Herrmann (1991); the Krox-20 probe is a PCR-fragment spanning the region between the aminoacids 178 and 350 of the sequence reported in Chavrier et al. (1989); the En-2 probe is a gift from R. M. Alvarado-Mallart.
In addition, we used two probes to genotype sectioned embryos and to follow the lacZ transcripts, consisting respectively of a 0.8 kb EcoRV-SacI fragment located in the lacZ coding sequence (lacZ probe) of the PGN vector and of a 0.15 kb fragment contained in the portion of the Otx2 cDNA deleted in the mutated allele (Otx2-del probe) and including aminoacids 1 to 33 of the Otx2 coding sequence.
Histochemical staining
For β-galactosidase staining, embryos were dissected in PBS, fixed on ice in 4% formaldehyde, 0.2% glutaraldehyde for 5-20 minutes, depending on their size. After 3 washes in PBS (phosphate-buffered saline) for 5 minutes, the embryos were incubated at 37°C for 4 hours to overnight, in 2 mM MgCl2, 4 mM K3Fe(CN)6, 4 mM K4Fe(CN)6, 1 mg/ml X-Gal in PBS. Embryos were then washed in PBS, pho-tographed and embedded in paraffin and sectioned as described above.
RESULTS
Disruption of an Otx2 allele in mice by homologous recombination in ES cells
Construction of the targeting vector
One allele of the Otx2 gene was inactivated by homologous recombination in ES cells. A 6.5 kb SmaI fragment, located approximately 220 bp upstream from the Otx2 translation initiation methionine, and a 1.2 kb NsiI fragment, spanning the 3′ terminal 500 bp of the second intron and the 5′ terminal 300 bp of the third exon, were inserted in the same orientation, into the polylinker of the pGN vector (Le Mouellic et al., 1990, 1992). In the resulting targeting vector, pGN31, these two fragments flank both a neomycin resistance gene, and the lacZ coding sequence fused to regions of the Otx2 endogenous locus lying 213 pb upstream from the translation initiation site (Fig. 1a). Homologous recombination events therefore result in the deletion of a large part of the Otx2 coding sequence, including most of the homeodomain, and its replacement by the lacZ reporter sequence. Since the transcription initiation site of the Otx2 gene was mapped at position −684 (A. S., unpublished data), the lacZ reporter gene should be transcribed as a composite RNA, consisting of 471 bp of the Otx2 mRNA 5′ terminal sequence, fused to the bacterial lacZ coding sequence.
Characterization of recombined ES cell lines and generation of Otx2+/− mice
The targeting plasmid, pGN31, was linearized and electropo-rated into HM-1 ES cells. Neomycin-resistant clones contain-ing a disrupted Otx2 allele were selected by PCR analysis using two primers, respectively specific for the targeting vector (located downstream from the neomycin resistance gene in the pGN vector) and for the Otx2 locus (located in the genomic sequence, downstream from the Otx2 homologous 3′ DNA fragment cloned in pGN31) (Fig. 1a). From 250 independent ES clones analyzed, 13 produced the expected 1.6 kb band indicative of a disrupted Otx2 allele (data not shown). The presence of a recombined Otx2 allele in these 13 clones was further verified by Southern blot analysis. Genomic DNAs, digested with HindIII and probed with an Otx2 fragment external to the targeting vector, showed the expected pattern of two positive bands, 5.3 kb and 3.0 kb long, corresponding to the wild-type and recombined loci respectively (Fig. 1b). Two clones had additional integration events, as showed by hybridization with a probe from the neomycin resistance coding sequence (data not shown).
Cells from three ES clones showing a correct recombination event in the Otx2 locus were injected into C57BL/6 blastocysts and generated chimeras transmitting the mutation to their offspring (Fig. 1c).
lacZ expression in Otx2+/− embryos
lacZ expression in heterozygous mice was examined through-out embryogenesis, both at the protein level, by histochemical localization of β-galactosidase activity, and at the mRNA level, by in situ hybridization of histological sections (Fig. 2).
lacZ and Otx2 expression in Otx2+/− embryos. (a-d) 5.75 dpc embryos. lacZ staining of whole-mount embryos (a) or sections (b) show the presence of the β-galactosidase activity both in the visceral embryonic endoderm and in the embryonic ectoderm. Some mosaïcism is observed especially in the ectoderm layer (b). In situ hybridization of sections show that lacZ (c) and Otx2 (d) transcripts display identical patterns of expression. (e-j) 6.5-7.0 dpc embryos. (f) A 6.5 dpc lacZ-stained Otx2+/− embryo. As shown on sections of 6.5 dpc (g) or 7.0 dpc (e, j) lacZ-stained embryos, the β-galactosidase activity is almost completely restricted to the external layer. In situ hybridization of sections at 6.5 dpc show that the lacZ (h) or Otx2 (i) transcripts are equally distributed in the external layer and in the ectoderm. (k-n) 7.75 dpc embryos. (k) A whole-mount lacZ-stained embryo at the head-fold stage. Sections of lacZ-stained embryos (l) and in situ hybridization with a lacZ (m) or Otx2 (n) probe show that the β-galactosidase activity, and both the lacZ and Otx2 transcripts display identical patterns of expression, in the anterior part of all three germ layers. (o-q) 16-20 somites stage. In a whole-mount 9.25 dpc embryo (o), the lacZ-staining is mainly restricted to the head region but is also detectable in the foregut (arrow in o). (p-q) Sagittal sections of a lacZ-stained 9.25 dpc embryo, with arrowheads pointing to a strong site of labelling including the thyroid rudiment (p) or to the notochord (q). (r) Whole-mount 10.5 dpc lacZ-stained embryo. The limit between midbrain and hindbrain regions is indicated by a thin arrow, an arrowhead points to the labelling lining the oral cavity. Scale bar : 50 μm, (a-j); 100 μm (k-n, p, q); 250 μm (o); 500 μm (r); ee, embryonic ectoderm; ve, visceral embryonic endoderm ; en, endomesoderm; mes, mesoderm; ne, presumptive anterior neuroectoderm; nt, notochord; tr, thyroid rudiment.
lacZ and Otx2 expression in Otx2+/− embryos. (a-d) 5.75 dpc embryos. lacZ staining of whole-mount embryos (a) or sections (b) show the presence of the β-galactosidase activity both in the visceral embryonic endoderm and in the embryonic ectoderm. Some mosaïcism is observed especially in the ectoderm layer (b). In situ hybridization of sections show that lacZ (c) and Otx2 (d) transcripts display identical patterns of expression. (e-j) 6.5-7.0 dpc embryos. (f) A 6.5 dpc lacZ-stained Otx2+/− embryo. As shown on sections of 6.5 dpc (g) or 7.0 dpc (e, j) lacZ-stained embryos, the β-galactosidase activity is almost completely restricted to the external layer. In situ hybridization of sections at 6.5 dpc show that the lacZ (h) or Otx2 (i) transcripts are equally distributed in the external layer and in the ectoderm. (k-n) 7.75 dpc embryos. (k) A whole-mount lacZ-stained embryo at the head-fold stage. Sections of lacZ-stained embryos (l) and in situ hybridization with a lacZ (m) or Otx2 (n) probe show that the β-galactosidase activity, and both the lacZ and Otx2 transcripts display identical patterns of expression, in the anterior part of all three germ layers. (o-q) 16-20 somites stage. In a whole-mount 9.25 dpc embryo (o), the lacZ-staining is mainly restricted to the head region but is also detectable in the foregut (arrow in o). (p-q) Sagittal sections of a lacZ-stained 9.25 dpc embryo, with arrowheads pointing to a strong site of labelling including the thyroid rudiment (p) or to the notochord (q). (r) Whole-mount 10.5 dpc lacZ-stained embryo. The limit between midbrain and hindbrain regions is indicated by a thin arrow, an arrowhead points to the labelling lining the oral cavity. Scale bar : 50 μm, (a-j); 100 μm (k-n, p, q); 250 μm (o); 500 μm (r); ee, embryonic ectoderm; ve, visceral embryonic endoderm ; en, endomesoderm; mes, mesoderm; ne, presumptive anterior neuroectoderm; nt, notochord; tr, thyroid rudiment.
In pregastrulating embryos, by 5.75-6.0 dpc, lacZ and Otx2 mRNAs were evenly distributed throughout the epiblast and the embryonic visceral endoderm of heterozygous and wild-type embryos respectively (Fig. 2c,d). At the protein level, however, the β-galactosidase staining appeared slightly fainter in the embryonic ectoderm than in the embryonic endoderm layer and various degrees of mosaïcism could be observed in the epiblast (Fig. 2a,b).
From early to mid-streak stages, lacZ mRNA accumulated both in the ectoderm and the external layer of heterozygous embryos with a progressive anteriorization (Fig. 2h). This pattern corresponds to the one previously reported for Otx2 mRNA in wild-type embryos, as shown in Fig. 2i. By contrast, a clear difference could be observed between the domains of expression of the β-galactosidase activity, repre-sentative of the protein accumulation and of the corresponding transcript. The β-galactosidase activity displayed a strong signal essentially confined to the external layer and was almost undetectable in the ectoderm layer at 6.5, 6.75, 7.25 dpc, whatever the detection protocol used (Fig. 2e,f,g,j), despite the presence of lacZ mRNA. However, like Otx2 and lacZ mRNAs, the enzymatic activity in the external layer was progressively displaced to the anterior part of the embryo (Fig. 2e,f,g and j).
From the headfold stage (7.75 dpc) onwards, the discrep-ancy between the domains of expression of the β-galactosidase activity and the lacZ messenger RNA was no longer observed in heterozygous embryos. At the headfold stage, the patterns of hybridization of wild-type or heterozygous embryos with Otx2 or lacZ probes, respectively, were perfectly superimpos-able (Fig. 2m,n), and also corresponded to the distribution of the β-galactosidase activity, which was detectable in all three germ layers and restricted to the most anterior one-third of the embryos (Fig. 2k,l). In the midline, a strong staining was also present at the oral plate level, immediately adjacent to the embryonic-extraembryonic junction (Fig. 2l), in line with the expression of Otx2 in the foregut at later stages. This pattern closely reflects the pattern of Otx2 transcription, previously described by in situ hybridization (A. S., unpublished data; Ang and Rossant, 1994; Simeone et al., 1992a, 1993).
At the 16-20 somites stage (Fig. 2o), the β-galactosidase activity was mainly restricted to the forebrain and midbrain regions, but additionally labeled regions included the foregut, the first branchial arch, the thyroid rudiment, the anterior half of the notochord and the rostral-most head ectoderm including the olfactory placodes (Fig. 2p,q). A sharp posterior limit of lacZ expression in the rostral neural tube became visible by 10.5 dpc (arrow in Fig. 2r). At this stage, the regions showing β-galactosidase activity included the entire forebrain and midbrain, except the regions of the optic chiasma and the optic recess. A clear labelling was also detectable in the developing eye and in the epidermal layer lining the oral cavity. This pattern of expression again corresponds to the one previously reported for Otx2 mRNA (Simeone et al., 1992a, 1993).
Embryonic lethality of the Otx2−/− phenotype
Matings between Otx2+/− mice failed to produce progeny pos-sessing two disrupted alleles, indicating the embryonic lethality of the homozygous mutant phenotype (Table 1).
Moreover, when B6/D2 wild-type females were mated to Otx2+/− males, the percentage of heterozygotes that reached weaning among the progeny did not exceed 35% (Table 2). However, the segregation of the mutated allele showed a Mendelian distribution in 9.5 dpc embryos and at birth (Table 1) and no obvious abnormalities were noted among heterozy-gous embryos up to 12.5 dpc. This suggests that a dominant effect of the mutation, probably depending on the genetic back-ground, affects the survival of some heterozygous animals during the first weeks following birth. This heterozygous phenotype was not further analyzed.
Phenotype of Otx2−/− mice at 8.5-9.5 dpc
The morphology of homozygous mutant embryos was first examined at 8.5-9.5 dpc. At 8.5 dpc, all Otx2−/− embryos showed a strong reduction in size and marked morphological abnormalities compared to their heterozygous or wild-type counterparts (Fig. 3a). As immediately apparent, the pheno-types observed displayed variations, possibly due to the mixed genetic background of the progenitors.
Morphology of 8.5 dpc and 9.5 dpc Otx2-/− embryos. (a) Comparison of Otx2+/− or Otx2−/− whole-mount lacZ-stained embryos at 8.5 dpc. (b-f) Whole-mount Otx2−/− embryos at 9.5 dpc. (b) Side-view and (d) dorsal view of an Otx2−/− embryo showing the strongest phenotype. The heart is indicated by an arrowhead. Somites are clearly visible in d. (c) Whole-mount Otx2−/− embryo at 9.5 dpc showing a less severe phenotype. The heart is indicated by an arrowhead. (e-f) Enlarged views of the head region of Otx2−/− embryos showing the strongest phenotype. An arrowhead points to the anterior limit of the neural tube, either at the presumptive hindbrain/midbrain junction (e), or at the hindbrain level (f). h, heart. Scale bar: 100 μm.
Morphology of 8.5 dpc and 9.5 dpc Otx2-/− embryos. (a) Comparison of Otx2+/− or Otx2−/− whole-mount lacZ-stained embryos at 8.5 dpc. (b-f) Whole-mount Otx2−/− embryos at 9.5 dpc. (b) Side-view and (d) dorsal view of an Otx2−/− embryo showing the strongest phenotype. The heart is indicated by an arrowhead. Somites are clearly visible in d. (c) Whole-mount Otx2−/− embryo at 9.5 dpc showing a less severe phenotype. The heart is indicated by an arrowhead. (e-f) Enlarged views of the head region of Otx2−/− embryos showing the strongest phenotype. An arrowhead points to the anterior limit of the neural tube, either at the presumptive hindbrain/midbrain junction (e), or at the hindbrain level (f). h, heart. Scale bar: 100 μm.
Mesodermal derivatives
Although at 9.5 dpc, Otx2−/− embryos displayed variable phe-notypes, two main groups could be distinguished. Embryos showing the strongest phenotype (Figs 3b,d-f, 4g) had a com-pletely abnormal body plan, with no more than 10 somites (Figs 3d, 4g). At 9.5 dpc, embryos showing a less severe phenotype (Figs 3c, 4a,c) were delayed in their development and resembled 8.5 dpc embryos. They displayed a number of somites ranging between 10 and 13, often irregular in shape and disorganized (Fig. 4a,c).
RNA in situ hybridization of 9.5 dpc Otx2−/− embryos. (a,c,f,g) Bright-field pictures of b, d-e, i and h, respectively. (b) A sagittal section of an embryo displaying a mild phenotype hybridized to Emx2. Only the posterior domain of expression of Emx2 is detected (arrowheads). (d,e) Sagittal sections of the embryo shown in Fig. 3c, hybridized to Hoxb-1 and Krox20 probes, respectively. Arrowheads in d and e indicate rhombomere 4 (d), or rhombomeres 3 and 5 (e). (h) Sagittal section of an embryo showing the strongest phenotype (Fig. 3b), hybridized to a Hoxb-1 probe. (i) Transverse section of an embryo showing the strongest phenotype (Fig. 3b) hybridized to Shh. The thin arrow points to the notochord, the arrowhead points to the gut. nt, notochord : g, gut; fp, floor plate. Scale bar, 100 μm
RNA in situ hybridization of 9.5 dpc Otx2−/− embryos. (a,c,f,g) Bright-field pictures of b, d-e, i and h, respectively. (b) A sagittal section of an embryo displaying a mild phenotype hybridized to Emx2. Only the posterior domain of expression of Emx2 is detected (arrowheads). (d,e) Sagittal sections of the embryo shown in Fig. 3c, hybridized to Hoxb-1 and Krox20 probes, respectively. Arrowheads in d and e indicate rhombomere 4 (d), or rhombomeres 3 and 5 (e). (h) Sagittal section of an embryo showing the strongest phenotype (Fig. 3b), hybridized to a Hoxb-1 probe. (i) Transverse section of an embryo showing the strongest phenotype (Fig. 3b) hybridized to Shh. The thin arrow points to the notochord, the arrowhead points to the gut. nt, notochord : g, gut; fp, floor plate. Scale bar, 100 μm
The heart was present, even in the strongest phenotypes, but markedly reduced in size (Fig. 3b,c). Although the presence of the prechordal plate was difficult to assess, due to the absence of specific markers, the notochord was recognized in all cases both by morphological criteria, in transverse sections of Otx2−/− embryos (Fig. 4f), and by its hybridization to the Shh probe (Fig. 4i; Echelard et al., 1993).
Gut development
Non-neural head structures
Non-neural anterior structures (optic vesicles, branchial arches), normally visible at 8.5 dpc (6-9 somites stage), were not individualized, even in the weaker phenotype of Otx2−/− embryos (Fig. 3b,c). The presence of olfactory placodes, normally present at 9.0 dpc (8-10 somites stage), was not revealed, either by histological observations, or by in situ hybridization with Otx1 or Emx2 probes (data not shown). The
absence of these structures is unlikely to result merely from a delay in development and probably represents specific features of the Otx2−/− phenotype.
Morphology and antero-posterior patterning of the neural tube
In all Otx2−/− embryos isolated at 9.5 dpc, the neural tube was recognizable and morphologically normal at the spinal cord level (Fig. 3d-f). At this axial level, the tube was closed and the floor plate cells were differentiated as assessed by their positive hybridization signal with the Shh probe (Fig. 4i).
At the hindbrain level, the morphology of the neural tube was variable. In the majority of embryos showing the strongest phenotype (Fig. 3b), the neural tube proceeded anteriorly to the rostralmost somites but did not show the typical hindbrain mor-phology, thus resembling the spinal cord (Fig. 3d,e). However, a minority (less than 20%) of these highly affected embryos showed a hindbrain-like structure at the level of the anterior-most somite, but the neural tube was abruptly interrupted immediately anterior to the hindbrain/spinal cord junction (Fig. 3f). No evidence of hindbrain segmentation was obtained in either of these two types of embryos, which displayed a Hoxb-1 transcription pattern reminiscent of that reported for wild-type embryos at 8.0 dpc, with two broad domains of expression flanking the segmented mesoderm (Fig. 4g,h). By contrast, embryos displaying a less severe phenotype (Fig. 3c) showed an almost normal hindbrain morphology. A clear evidence for the presence of the hindbrain and its segmented organization was also obtained by in situ hybridization (Fig. 4c-e). At the stage analyzed, in wild-type embryos, Krox20 is activated in two transverse stripes, corresponding to rhombomeres 3 and 5 (Wilkinson et al., 1989), whereas expression of Hoxb-1 is restricted to rhombomere 4 (Frohman et al., 1990). This pattern of hybridization of Hoxb-1 and Krox20 was clearly recogniz-able on the sections shown in Fig. 4d,e, respectively. However, the distance between the rostral end of the embryos and the anterior border of Krox20 and Hoxb-1 expression domains was strongly reduced, relative to wild-type mice (Fig. 4d,e). Beyond the presumptive hindbrain level, the neuroectoderm was either abruptly interrupted (Fig. 4c) or replaced by a thin layer of epidermal cells in the head region (Fig. 4a).
To delineate more precisely the missing part of the neural tube, the expression pattern of several anterior or posterior markers, normally expressed at this stage in heterozygous or wild-type embryos, was analyzed by in situ hybridization of sections from the embryos shown in Fig. 4. At 9.5 dpc, Otx1, En-2 and Emx2 expression domains define specific territories in the developing midbrain and forebrain (Davis and Joyner, 1988; Davis et al., 1988; Simeone et al., 1992a,b). No positive hybridization signal was obtained with any of these probes in the head region of the Otx2−/− embryos analyzed, even after overexposure of the hybridized sections (data not shown). A posterior domain of expression of Emx2 has also been reported for wild-type embryos at this stage, including the coelomic epithelium covering the mesonephric vesicle and the final part of the mesenteric attachment. Although these structures were difficult to unambiguously identify in Otx2−/− mutants, a clear signal was apparent at the corresponding level in one of the weakest Otx2 embryonic phenotype (Fig. 4b). Finally, no lacZ signal could be detected in the head region of any Otx2−/− 9.5 dpc embryo.
Otx2−/− phenotype during gastrulation
Before the onset of gastrulation, defined by the appearance of mesoderm and primitive streak, no morphological difference could be reliably detected between homozygous, wild-type or heterozygous embryos.
Morphology of Otx2−/− mutants
Up to the mid-streak stage, Otx2−/− embryos did not show strong morphological defects either in shape or in size (Fig. 5a). Histological sections revealed the presence of the primitive streak elongating from the embryonic-extraembry-onic junction and of mesodermal cells, intercalating posteriorly between the ectoderm and the external layer (Fig. 5b,c).
Morphology, lacZ and goosecoid expression of Otx2−/− mutant embryos at early streak stage. (a) Whole-mount β-galactosidase-stained 6.5 dpc Otx2−/− embryo. The β-galactosidase activity is restricted to the distal half of the embryonic part. (b) Sagittal section of the embryo shown in a. The staining is only detectable in the external layer containing endomesoderm cells. (c-e) Sagittal sections of a 6.5 dpc Otx2−/− embryo; (c), bright field, (d,e) dark-field pictures; (d,e) hybridized to the Otx2-del and lacZ probes, respectively. The embryo shown is homozygous, as proved by the absence of signal with the Otx2-del probe. The lacZ transcript (e) shows a distribution identical to the one displayed by the β-galactosidase activity in b. (f-k) Sagittal sections of a 6.5 dpc Otx2+/− embryo (f,g), a 6.8 dpc Otx2+/− embryo (h,i) and a 6.5 dpc Otx2−/− embryo (j,k). (f,h,j) Bright-field pictures, (g,i,k) corresponding sections hybridized with a goosecoid probe. While Otx2+/− embryos show the expected signal (g,i), no signal is detected in the Otx2−/− embryo (k). ee, embryonic ectoderm; en, endomesoderm; mes, mesoderm. Scale bar, 50 μm.
Morphology, lacZ and goosecoid expression of Otx2−/− mutant embryos at early streak stage. (a) Whole-mount β-galactosidase-stained 6.5 dpc Otx2−/− embryo. The β-galactosidase activity is restricted to the distal half of the embryonic part. (b) Sagittal section of the embryo shown in a. The staining is only detectable in the external layer containing endomesoderm cells. (c-e) Sagittal sections of a 6.5 dpc Otx2−/− embryo; (c), bright field, (d,e) dark-field pictures; (d,e) hybridized to the Otx2-del and lacZ probes, respectively. The embryo shown is homozygous, as proved by the absence of signal with the Otx2-del probe. The lacZ transcript (e) shows a distribution identical to the one displayed by the β-galactosidase activity in b. (f-k) Sagittal sections of a 6.5 dpc Otx2+/− embryo (f,g), a 6.8 dpc Otx2+/− embryo (h,i) and a 6.5 dpc Otx2−/− embryo (j,k). (f,h,j) Bright-field pictures, (g,i,k) corresponding sections hybridized with a goosecoid probe. While Otx2+/− embryos show the expected signal (g,i), no signal is detected in the Otx2−/− embryo (k). ee, embryonic ectoderm; en, endomesoderm; mes, mesoderm. Scale bar, 50 μm.
By contrast, Otx2−/− embryos showed strong abnormalities at the late streak stage (Fig. 6a). The amnion was not clearly individualized, making the embryonic-extraembryonic junction difficult to delineate precisely (Fig. 6b). While extraembryonic structures appeared almost normal, with a clearly discernable chorion and allantoïs (Fig. 6b), the pre-sumptive embryonic part (containing the ectoderm) was markedly reduced in size and appeared as a sphere, linked to the proximal region by a more or less severe constriction. As shown by histological sections (Fig. 6b,c,f,i,k), mesodermal cells were present, intercalating between the ectoderm and the external layer and also accumulating in a disorganized fashion at the level of the constriction. The ectoderm itself was circu-larized and entirely contained in the spherical, distal part of the embryo.
Otx2−/− phenotype at 7.5 dpc. (a) β-galactosidase activity in a whole-mount lacZ-stained 7.5 dpc Otx2−/− embryo. An arrow points to the constriction observed in these embryos. (b) Sagittal section, showing the circularized ectoderm and the disorganization of the forming mesoderm. The amnion is not clearly individualized but the allantois and chorion are clearly visible. (c) Sagittal section of alacZ-stained Otx2−/− embryo, showing that the labelling is restricted to the external layer. (d) Sagittal section of an Otx2−/− embryo hybridized to the lacZ probe. The labelling is clear in the external mesendoderm layer but absent from the embryonic ectoderm. (e) Whole-mount hybridization of an Otx2−/− embryo with a Brachyury probe, showing the extent of the primitive streak. (f-l) RNA in situ hybridizations of sagittal sections of Otx2-/− 7.5 dpc embryos. (f,i,k) Bright-field pictures corresponding to the dark-field pictures shown in g,h,j,l, respectively. (g,h,j,l) Hybridized respectively to Brachyury, Evx1, Hoxb-1 and cripto probes. ee, embryonic ectoderm; al, allantois; ch, chorion; en, endomesoderm. Scale bar : 100 μm.
Otx2−/− phenotype at 7.5 dpc. (a) β-galactosidase activity in a whole-mount lacZ-stained 7.5 dpc Otx2−/− embryo. An arrow points to the constriction observed in these embryos. (b) Sagittal section, showing the circularized ectoderm and the disorganization of the forming mesoderm. The amnion is not clearly individualized but the allantois and chorion are clearly visible. (c) Sagittal section of alacZ-stained Otx2−/− embryo, showing that the labelling is restricted to the external layer. (d) Sagittal section of an Otx2−/− embryo hybridized to the lacZ probe. The labelling is clear in the external mesendoderm layer but absent from the embryonic ectoderm. (e) Whole-mount hybridization of an Otx2−/− embryo with a Brachyury probe, showing the extent of the primitive streak. (f-l) RNA in situ hybridizations of sagittal sections of Otx2-/− 7.5 dpc embryos. (f,i,k) Bright-field pictures corresponding to the dark-field pictures shown in g,h,j,l, respectively. (g,h,j,l) Hybridized respectively to Brachyury, Evx1, Hoxb-1 and cripto probes. ee, embryonic ectoderm; al, allantois; ch, chorion; en, endomesoderm. Scale bar : 100 μm.
lacZ expression in Otx2−/− mutants
At the early streak stage, analysis of the β-galactosidase activity, which, as shown for Otx2+/− embryos, is detected almost exclusively in the external layer, pointed to two main differences between heterozygous and homozygous mutant embryos.
First, the domain of lacZ expression appeared clearly altered both in the ectoderm and in the external layer. The lacZ tran-scription domain, which shows an anterior localization extending from the distal tip to the embryonic/extraembryonic junction in Otx2+/− embryos (Fig. 2h), remained confined to the distal third to half of the external layer, without clear ante-riorization in 6.5-6.75 dpc Otx2−/− embryos (Fig. 5e). The lacZ staining showed the same pattern in the external layer (Fig. 5a,b). Likewise, in the embryonic ectoderm, lacZ transcripts were not distributed from the distal tip to the junction between embryonic and extraembryonic ectoderm as observed in Otx2+/− embryos (Fig. 2h), but remained confined to the distal part of the ectoderm, adjacent to the β-galactosidase-express-ing external layer (Fig. 5e).
Second, in Otx2−/− embryos, the level of lacZ mRNA transcription in the ectoderm layer was markedly reduced relative to heterozygous embryos at the early streak stage (compare Figs 5e and 2h), and became almost undetectable at the late streak stage (compare Figs 6d and 2m). By contrast, the level of lacZ expression, both at the RNA and protein level, remained high in the external layer of 6.5 dpc Otx2−/− embryos (Fig. 5a,b,e) and a limited domain of strong expression was also still present in the presumptive anterior part of the external layer of 7.5 dpc embryos (Fig. 6c,d).
Expression of gastrulation markers
At 7.5 dpc, a strong expression of Brachyury (Herrmann, 1991) extending up to the distal tip of the embryos was observed, suggesting that the primitive streak was formed in Otx2−/− animals (Fig. 6e,g). At this stage, two other markers of the pos-teriorly forming mesoderm, Hoxb-1 (Frohman et al., 1990) and cripto (Dono et al., 1993), were also transcribed at high levels along the primitive streak, as expected (Fig. 6j,l). Evx1 is normally expressed in wild-type embryos in all three germ layers along the primitive streak, with a progressive decrease in the level of expression from the proximal to the distal part of mid-to late-streak embryos (Dush and Martin, 1992). Despite the strong abnormalities observed in mutant embryos, the main characteristics of this pattern of Evx1 expression were unaffected in Otx2−/− embryos (Fig. 6h). In several cases, however, an anterior displacement of some Evx1, Hoxb-1, cripto and Brachyury-expressing mesoderm cells was observed, in line with the disorganization of mesoderm also noticed in histological sections (data not shown, Fig. 6j).
By contrast, in the ectoderm, the respective anterior limits of expression of the different posterior probes were always respected. Hoxb-1, cripto or Evx1 were never expressed in anterior regions of the ectoderm in Otx2−/− 7.5 dpc embryos.
The expression of goosecoid, which is transiently expressed in 6.5 to 6.9 dpc wild-type embryos (Blum et al., 1992; Faust et al., 1995; Conlon et al., 1994) was also tested by in situ hybridization. As previously reported, in Otx2+/− or wild-type embryos, goosecoid transcripts were detected in the visceral embryonic endoderm and at the proximal posterior region of the ectoderm at 6.5 dpc (Fig. 5f,g), and in the anterior part of the primitive streak at 6.9 dpc (Fig. 5h,i). By contrast, hybrid-ization signals remained either undetectable or strongly reduced in Otx2−/− embryos from the same litters (Fig. 5j,k). Likewise, at 7.5 dpc, Shh expression was observed in the midline mesoderm of the head process in Otx2+/− or wild-type embryos but could not be detected in their Otx2−/− counterparts (data not shown).
DISCUSSION
Otx2−/− embryos are characterized by the absence of forebrain and midbrain regions
The inactivation of Otx2, a murine homolog of the Drosophila orthodenticle homeodomain gene, resulted in the deletion of anterior structures. In all homozygous embryos tested, anterior parts of the neural tube, corresponding to the presumptive forebrain and midbrain, were deleted. At 9.5 dpc, this deletion was clearly visible on histological sections and was further assessed by the absence of hybridization signal with a number of markers specific for the corresponding regions. The absence of other ectodermal derivatives, like the optic lens placodes and the olfactory placodes, was also consistent with the deletion of forebrain neuroectoderm, since the ventral anterior forebrain and the optic vesicles are required to induce the olfactory placodes and the lens placodes, respectively. The role of Otx2 appears therefore similar to that of the Drosophila Otd gene which was shown to function as a ‘gap gene’, since its mutation results in the deletion of specific head segments (Cohen and Jürgens, 1990; Finkelstein and Perrimon, 1990, 1991).
By contrast, regions of the neural tube fated to the spinal cord were always present. At the hindbrain level, the pheno-types observed were more complex, suggesting a variable expressivity of the mutation. In embryos showing the strongest phenotypes, the typical hindbrain morphology was not observed and evidence of segmentation was never obtained. However, in weaker Otx2−/− phenotypes, the hindbrain region was present and showed clear evidence of anteroposterior seg-mentation, based on the identification of rhombomeres 3, 4 and 5 with Krox20 or Hoxb-1 probes. We also detected the presence of floor plate cells by hybridization with a Sonic hedgehog probe, which indicated some dorsoventral pattern-ing. However, we could never detect En-2 expression in the first rhombomere at 9.5 dpc, which suggests that anteriormost regions of the hindbrain were also either missing, or improperly specified.
Regionalisation and specification of the embryonic ectoderm during gastrulation are defective in Otx2−/− embryos
By the end of gastrulation, Otx2 is transcribed in the pre-sumptive anterior neuroectoderm in mouse, chick and Xenopus (Simeone et al., 1993; Bally-Cuif et al., 1995; Pannese et al., 1995). The absence of lacZ transcripts in the ectoderm of 7.5 dpc Otx2−/− embryos therefore suggests that, at this stage, anterior neuroectoderm was not properly specified. However, expression of posterior neural markers was never observed in anterior regions of the ectoderm layer, also suggesting that anterior neuroectoderm did not acquire an alternative (more posterior) identity, but failed to be induced. At the early streak stage, cell lineage studies in mouse have shown that a rostro-caudal pattern is already established in the epiblast, in which cells fated to anterior neuroectoderm are localized at the distal anterior third of the embryonic ectoderm (Lawson and Pedersen, 1992; Quinlan et al., 1995). The abnormal distribu-tion of ectodermal cells expressing lacZ mRNA, observed in 6.5 dpc Otx2−/− embryos, together with the strongly reduced level of transcription relative to their Otx2+/− counterparts, shows that the regionalisation of the epiblast was already affected at this stage.
The requirement for an inductive signal to maintain or induce Otx2 transcription in the ectoderm layer is supported by several lines of evidence. In vitro, isolated explants of mouse ectoderm become committed to express Otx2 in a cell autonomous fashion by the mid-streak stage. At earlier stages, such explants are unable to stably express Otx2 and a signal from mesendoderm (which consists, in this in vitro study, in both the mesoderm and external layer from 7.5 dpc embryos) is required to maintain ectodermal Otx2 transcription (Ang et al., 1994). The Otx2 expression pattern in Xenopus gastrulae and exogastrulae also suggests that Otx2 transcription in the presumptive anterior neuroectoderm is dependent upon a direct contact with the underlying mesendoderm (Pannese et al., 1995). Consistent with these data, lacZ expression in Otx2−/− gastrulating embryos was strongly maintained in the external layer, but was almost undetectable in the embryonic ectoderm, suggesting that an inductive signal required to maintain, or induce, Otx2 expression in the ectoderm layer, but not in the external layer, could not take place in the absence of a func-tional Otx2 allele.
Alterations of the node in Otx2−/− mutants
The anterior end of the mouse primitive streak, or node, of 7.5 dpc embryos, is able to induce neural structures in transplan-tation experiments and thus appears as a functional equivalent of the chick Hensen’s node or Xenopus organizer (Beddington, 1994). At early stages of gastrulation, its fate map is also largely similar to those described in avians and amphibians: in particular, endomesoderm cells, derived from regions located at the anterior end of the primitive streak, contribute to defin-itive endoderm, prechordal and chordal mesoderm (for nomenclature of embryonic cells in early mouse gastrulae: see Lawson and Pedersen, 1992, p. 3-36). Cells in the midline ectoderm of the neural tube are also derived from the node ectoderm (Lawson and Pedersen, 1992; Sulik et al., 1994).
In Otx2−/− embryos, the node structure could not be recog-nized at the end of gastrulation and its formation was altered as early as the early streak stage, as shown by the absence or strong reduction of goosecoid expression. Similarly Shh expression was delayed. However, the precise cell population involved in the anterior neuroectoderm specification remains unidentified. In the external layer of Otx2−/− embryos, the identity of lacZ-expressing embryonic cells cannot be defini-tively established for lack of early cellular markers. Accord-ingly, an endomesoderm subpopulation deriving from the node could be either missing or unproperly specified. Our data are also consistent with grafting experiments performed in chick since at stage HH4, sectors of Hensen’s node able to induce neural structures expressing anterior markers are characterized by the presence of progenitor cells to chordal and prechordal mesoderm, gut endoderm and neural tissue (Selleck and Stern, 1991; Storey et al., 1995).
In Otx2−/− embryos, the absence of anterior neuroectoderm therefore appears related to defects observed as early as the early streak stage, which provides genetic evidence that its specification occurs at early gastrulation. This timing is con-sistent with grafting experiments performed in chick, showing that later than stage HH4, Hensen’s node transplants become unable to induce anterior neural structures (Storey et al., 1992). Furthermore, the inactivation in mouse of another homeobox gene Lim-1, expressed in mesodermal derivatives, including the node region at the onset of gastrulation and prechordal mesoderm at the headfold stage, has been shown to result in a phenotype clearly related to the one displayed by Otx2−/− embryos (Shawlot and Behringer, 1995). Lim1−/− embryos show the same absence of forebrain and midbrain regions and abnormalities are also apparent in the node region shortly after the onset of gastrulation, as shown by the altered location of the subpopulation of cells expressing goosecoid. However, in contrast to Lim1−/− embryos, Otx2−/− embryos present severe defects in the trunk. We also never noticed the formation of any secondary axis in Otx2−/− mutants. The genetic interactions between these genes will have to be investigated by codetec-tion at the single cell level in the primitive streak and node region.
Anteriorization of lacZ expression domain in the embryonic external layer is defective in Otx2−/− embryos
Fate mapping studies have shown that, during gastrulation, the visceral embryonic endodermal cells are progressively replaced by a complex population of cells, including definitive endoderm cells that will contribute to the gut (Lawson and Pedersen, 1987), prechordal and chordal mesoderm cells. The dynamics of this process remain largely unknown. However, lineage analysis after single cell labelling experiments have shown that cells located in the external embryonic layer anteriorly to the node are progressively displaced anteriorly during gastrulation (Lawson and Pedersen, 1987, 1992). Such morphogenetic movements are strikingly similar, and could contribute, to the observed anteriorization of the lacZ expression domain in the Otx2+/− embryos. Furthermore, whereas the lacZ transcripts appeared uniformly distributed in the visceral embryonic endoderm layer, the anteriorization of the lacZ expression domain was clearly visible by the early streak stage. This suggests that cell displacements in the external embryonic layer might be initiated at, or even before, the onset of gastrulation. In line with this hypothesis, fate mapping studies of chick embryos have shown that several cell types, including presumptive gut endoderm, head process and notochord cells, converge to the posterior midline and then to the centre of the blastoderm, where the node will eventually form, even before primitive streak formation (Hatada and Stern, 1994). However, the dynamic pattern of lacZ expression in the external embryonic layer might also result, at least in part, from genetic mechanisms that maintain the Otx2 expression anteriorly and repress it posteriorly.
In the external layer, the anteriorization of lacZ-expressing cells was impeded in Otx2−/− mutant embryos as soon as the early streak stage. These data suggest that Otx2-positive cells located in the external embryonic layer are involved in the genetic cascade of regulatory events leading to the establish-ment of anteroposterior polarity.
Post-transcriptional regulation of lacZ expression in ectoderm during gastrulation
The patterns of lacZ and Otx2 transcription in Otx2+/− embryos appeared perfectly superimposable throughout embryogenesis, showing a progressive anteriorization in both the ectoderm and the external layer during gastrulation. By contrast, the β-galac-tosidase activity remained almost undetectable specifically in the ectoderm of gastrulating Otx2+/− embryos, even though the corresponding RNA was clearly present. Instability of the enzyme activity in such a stage- and tissue-specific fashion is unlikely since β-galactosidase activity can be detected in the ectoderm of other transgenic gastrulating embryos (Lallemand and Brûlet, 1990). The observed repression might reflect mech-anisms of post-transcriptional regulation exerted on the Otx2 locus specifically in the ectoderm, up to 7.25 dpc. This repres-sion might be exerted on β-galactosidase, but not Otx2, protein accumulation, due to the deletion of part of the 5′ leader sequence or of the introns in the lacZ transcript. Alternatively, the regulation exerted on the lacZ gene might reflect a post-transcriptional mechanism also controlling Otx2 protein accu-mulation and possibly involving cis-acting elements in the 5′ leader portion of Otx2 mRNA, which is present in the lacZ fusion transcript. Translational repression mechanisms during early embryogenesis are known to play a crucial role in the specification of the anteroposterior embryonic polarity in several species (reviewed in Kimble, 1994). The possibility of such a regulation in the case of Otx2 will have to be confirmed.
ACKNOWLEDGEMENTS
Philippe Brûlet thanks Dr Kristie Lawson for very helpful discus-sion and Dr D.W. Melton for the gift of the HM-1 ES cells; Dr R.M. Alvarado-Mallart for the gift of the En-2 probe; Dr G. Persico for the gift of the cripto probe; Prof F. Jacob, Drs H. Le Mouellic and J. Shellard for their critical comments on the manuscript; Mrs M. Compain and C. Sengmany for secretarial and technical assistance. D. A. was the recipient of an EMBO long term fellowship while in P. B. laboratory. This work was supported by grants from the Centre National de la Recherche Scientifique, the Association Française contre les Myopathies, the Association pour la Recherche sur le Cancer, the Institut National de la Santé et de la Recherche Médicale, the Ligue Nationale contre le Cancer, the Groupement de Recherches et d’Etudes sur le Génome as well as from the Italian Association for Cancer Research and the Italian Telethon Program.