A chicken caudal homologue, CHox-cad, is expressed in the epiblast with posterior localization and in the early endodermal lineage

Numerous homeobox-containing genes (homeobox genes) have been cloned from vertebrate genomes and, in several instances, the same or closely related genes have been cloned from different organisms (Scott et al. 1989; Graham et al. 1989; Duboule and Dollé, 1989). The interest in this vertebrate gene family arose as a result of the discovery of the homeobox sequence in Drosophila developmental genes (see Gehring, 1987a,b for reviews), which was subsequently shown to have been conserved during evolution, in animals including vertebrates (McGinnis et al. 1984). The large number of vertebrate homeobox genes cloned to date are found in one of four major Hox clusters which, in mice and humans, contain 6–10 homeobox genes (Acampora et al. 1989; Graham etal. 1989; Duboule and Dollé, 1989; Njplstad et al 1990; Fritz et al. 1989; Wedden et al. 1989). The organization of vertebrate homeobox genes in clusters is reminiscent of Drosophila homeobox gene organization. Comparison of sequence and genomic organization of the clusters in mouse and man led to the suggestion that the vertebrate Hox clusters arose as a result of duplication of an ancestral gene complex (Hart et al. 1987; Kappen et al. 1989; Schughart et al. 1989). The similarity between the Drosophila and vertebrate homeobox gene clusters extends beyond the arrangement of homeobox sequences in a gene complex. Comparison of vertebrate genes in the clusters to the Drosophila genes in the Antennapedia and Bithorax complexes revealed a colinear arrangement of cognate genes (Acampora et al. 1989; Duboule and Dollé, 1989; Graham et al. 1989). Furthermore, as in Drosophila, the order of the genes along the cluster represents the anteroposterior order of their rostral boundary of expression (Gaunt et al. 1988; Graham et al. 1989; Duboule and Dollé, 1989).

The high degree of homology between the vertebrate homeobox sequences and their Drosophila counterparts has resulted in the organization of the genes into families based on their homeodomain sequences (Scott et al. 1989). In some instances, a homeobox gene family contains several sequences from the same organism due to the duplication of the Hox complexes. The homology between cognate genes is not restricted to the homeo-box region but it extends throughout the genes (Graham et al. 1988; Nj01stad et al 1988, 1990). In addition to the genes in the Hox clusters, a few vertebrate homeobox genes have been cloned for which there is no evidence that they belong to a homeobox gene complex. In mice, the reported genes that belong to this category are the Hox 7.1 (Hill etal. 1989; Robert et al. 1989), Cdxl (Duprey et al. 1988), Evx1 and Evx2 (Bastian and Gruss, 1990), En-1 and En-2 (Joyner and Martin, 1987) genes. The Drosophila genome also contains a number of homeobox genes that are not members of the known clusters. Some of these genes have been identified by genetic analysis, as in the case of even-skipped (eve;Macdonald et al. 1986), while others, cloned solely by homeobox homology, had no known mutations for them, such as caudal (cad;Mlodzik et al. 1985).

Of the Drosophila homeobox genes that do not belong to one of the complexes, several are of particular interest as they appear to have vertebrate homologues. The Drosophila homeobox genes are msh (Robert et al. 1989), eve and cad whose vertebrate homologues are Hox 7.1, Xhox3 together with Evx1 and Evx2 (Bastian and Gruss, 1990; Ruiz i Altaba and Melton, 1989) and Cdx1, respectively. The cad-Cdx1 pair is of great interest as Cdx1 appears to be expressed in a pattern that resembles, in part, the cad expression pattern as it relates the organ and tissue layer in which they are expressed. In the course of embryonic development, the Drosophila cad gene transcripts are first detected as maternal transcripts that form a gradient whose maximal levels are localized to the posterior pole of the embryo (Mlodzik et al. 1985). This gradient is replaced at later stages of development by posterior transcript localization; in this case, it is made of transcripts of zygotic origin. In developmental stages after gastru-lation, the cad mRNA can be seen to be localized to the posterior midgut and Malpighian tubules, the posterior midgut being of endodermal origin (Mlodzik and Gehnng, 1987). Expression of the Cdx1 homeobox gene during the course of murine development is first detected by in situ hybridization in embryos 14 days post coitum (p.c.), and from this stage on its transcripts are localized to the epithelial lining of the intestine, which in mice is of endodermal origin (Duprey et al. 1988). Expression of the Cdx1 gene could not be detected in earlier embryos or in adult ovaries. Cdx1 and cad not only share high homology in their homeodomain sequences but Cdx1 is expressed in what appears to be part of the cad expression pattern.

In this paper, we report the isolation of a new chicken homeobox gene that contains a homeobox sequence belonging to the cad family of homeodomains and was thus termed CHox-cad. Transcripts of this gene are first detected as gastrulation begins and, by day five of incubation, no transcripts are detectable. The onset of CHox-cad transcription correlates with the onset of gastrulation suggesting a role for this gene during the establishment of the three germ layers. In situ hybridization analysis revealed that the early CHox-cad expression is localized mainly to the caudal region of the embryo and is restricted to the epiblast, primitive streak and definitive endoderm, while at later stages the expression is localized to the endodermal lining of the developing gut.

A genomic clone of the CHox-cad gene, λGG4, was isolated by screening a chicken oviduct genomic library in EMBL4 kindly provided by Dr B. W. O’Malley as previously described (Rangini et al. 1989).

The cDNA library of stage 12-13 chicken embryos (2 days of incubation) was constructed from poly (A)⁺ RNA that was purified twice through oligo (dT)-cellulose columns. cDNA was prepared according to the procedure of Okayama and Berg (1982) as modified by Gtibler and Hoffman (1983). After EcoRI methylation of the cDNA, EcoRI linkers were added and the cDNA was cloned into the λgt10 (Huynh et al 1985). About 10⁶ phage were screened with a genomic fragment containing the CHox-cad homeobox. Screening was per-formed under high-stringency conditions Briefly, the filters were prehybndized at 65 °C in a solution containing 50mM phosphate buffer pH6.9, 5×SSC, 5×Denhardt’s solution, 0.1 mg ml^-1 Salmon sperm DNA as earner and 0.1% SDS. After prehybndization the probe was added to a similar solution and the filters were hybndized under the same conditions for a further 20 h. The final washes of the library were performed in 0 1×SSC, 0.1 % SDS at 65°C

For the preparation of embryonic RNA, fertilized chicken eggs were obtained from local farms and incubated for the desired time. Embryos were extracted in ice-cold PBS and staged according to Hamburger and Hamilton (1951) Preparation of RNA was according to Chirgwin et al (1979) as previously descnbed (Rangini et al 1989) Northern blot analysis was performed under high stnngency hybridization conditions as described (Rangini et al. 1989). In all RNA gels, 28S, 23S, 18S and 16S ribosomal RNAs were used as size markers and the estimation of transcript sizes was performed according to Lehrach et al (1977) 

Sequencing was performed on both strands of the DNA by the dideoxy chain-termination method (Sanger et al 1977) using the Sequenase II kit (US Biochemicals) The genomic sequence was determined on specific subclones prepared using known restriction sites. Sequencing of the C33 cDNA clone was accomplished by performing a series of nested unidirectional deletions as descnbed by Hemkoff (1984) using the Erase-a-Base kit (Promega). In some instances, specific oligonucleotide pnmers were used to sequence through regions where no deletions were obtained.

The protocol employed for in situ hybndization was adapted to chicken embryos from Wilkinson et al. (1987) Briefly, embryos at the appropnate developmental stages were extracted onto ice-cold PBS and staged. Fixation was performed in ice-cold 4 % paraformaldehyde in PBS for time penods ranging from 30 min to 2h depending on their size. Following fixation, the embryos were embedded in paraffin and 5–8 μm sections were collected on TESPA (3-amino-propyltriethoxysilane)-treated glass slides. The slides were treated with xylene to remove the wax and then rehydrated through an ethanol senes. Further treatments included saline for 5 min, PBS 5 min, fixation 20 min, twice PBS for 5 min, 20 μg ml^-1 proteinase K in 50 mM Tns-HCl, 5ITIM EDTA pH8.0 for 5 min, PBS 5 min and fixation 20 min. After acetylation for 10 min, PBS and saline for 5 min each, the slides were dehydrated through an ethanol series and air dned. The slides were hybndized in 50% formamide, 0 3M NaCl, 20 mM Tris-HCl, 5mM EDTA pH8.0, 10% dextran sulphate, lxDenhardt’s solution, 0.5mgml^-1 yeast RNA, 10 mM dithiothreitol and 2×10⁵cts min^-1μl^-1 of ³⁵S-UTP labelled probe The hybridizations were performed for about 18 h at 50°. Washes of the slides included 5 × SSC, 10 mM DTT at 50°C for 60min, 50% formamide, 2×SSC, 20mM DTT at 65°C for 30 mm, three 10 min washes in 0.5 M NaCl, 10 mM Tris-HCl, 5mM EDTA at 37°C, an extra 30min wash in the same buffer containing 20μgml^-1 RNAse A followed by a 15 min wash in the same buffer excluding RNAse. The slides were washed in 50 % formamide, 2 × SSC, 20 mM DTT at 65 °C for 30 min and the final washes were for 15 mm each at 65 °C in 2× and 0.1×SSC, respectively. After dehydration, the slides were dipped in photographic emulsion for exposure.

A chicken oviduct genomic library was screened under low-stnngency hybridization conditions as described by Rangini et al. (1989). From this library screen about 15 independent homeobox-containing phage were iso-lated. One that appeared to hybridize preferentially to the scr probe was selected for further analysis. This phage, λGG4, was restriction mapped and the position of the homeobox within this genomic fragment was established (Fig. 1A). The analysis of λGG4 by hybridization to a number of homeobox probes revealed that this phage probably contains only one homeobox sequence. This result suggests that the homeobox gene contained in λGG4 might not be a member of one of the known vertebrate Hox clusters unless the distance to the neighboring homeobox genes is greater than the DNA contained in the cloned phage.

Initial characterization of the cloned homeobox gene was performed by sequencing the genomic fragment from λGG4 that contains the homeobox itself. This region was sequenced on both strands after subcloning the appropriate restriction fragments. The sequence of the homeobox and flanking regions is shown in Fig. 2. Comparison of CHox-cad with homeoboxes from other organisms revealed 83.6% homology to Cdx1 (Duprey et al. 1988), 72.6 % to caudal (Mlodzik et al. 1985) and 61.7 % to the antp (Garber et al. 1983) homeoboxes at the nucleotide level. When the comparisons were performed between the putative protein translations, the extent of identity was 95 % with the Cdxl, 80.3 % with caudal and 62.3 % with the antp homeodomains. Due to its homology to the caudal and Cdx1 homeo-domains, we call this novel chicken homeobox gene CHox-cad. In addition, the homology between Cdxl, CHox-cad and caudal extends five amino acids upstream of the homeodomain (Fig 3). Downstream from the homeobox, five out of six amino acids are shared between Cdxl and CHox-cad but not with caudal. It is interesting to note that the homology upstream from the homeobox ceases at the position where the caudal gene is interrupted by an intron. Analysis of the genomic sequence upstream from the CHox-cad homeobox revealed the sequence 5′ CTCTCTCTGCCAGG (Fig. 2, overlined) which is a good consensus sequence for a splice acceptor site (Shapiro and Senapathy, 1987). This sequence in CHox-cad localizes the intron-exon boundary to the same position relative to the homeobox as in the case of the Drosophila caudal gene.

Interestingly, the genomic sequence of the CHox-cad homeobox reveals that the homeodomain itself is interrupted by an intron. The intron is localized 118 bp from the beginning of the homeobox sequence thus breaking the homeodomain sequence between amino acids 44 and 45. The length of the intron is 128 bp and it is flanked by the sequences 5′ AGGTGAGT and 5′TCTTCCCACAGG, which are in good agreement with the consensus splice sequences for the donor and acceptor splice sites, respectively (Shapiro and Senapathy, 1987).

In order to study the organization of the CHox-cad gene as well as its putative protein product, cDNA clones were isolated. A cDNA library prepared from embryos at stage 12–13 (H and H, 2 days of incubation) was prepared in λgt10. About 5×10⁵ –10⁶ recombinant phage were screened with probe A. Of the 7 recombinant phage isolated, the one with the largest insert of approximately 2.6 kb (C33), was selected for sequencing and further analysis (Fig. IB). The insert in C33 is 2486 bp long (Fig. 2), suggesting that this cDNA clone is almost full length in relation to the estimated size of the CHox-cad transcript from northern analysis. The cDNA contains a 744 bp long open reading frame capable of coding for a 248 amino acid protein which includes the CHox-cad homeodomain. The AUG of the putative initiator methionine is present in the sequence 5′CCAACAUG at position 246–253 of the cDNA which is a good consensus sequence for the initiation of translation (Kozak, 1986). This AUG is present 61 bp downstream from an in-frame stop codon. The stop codon in frame with the homeodomain that marks the end of the CHox-cad protein is present 51 amino acids downstream from the homeodomain (Fig. 2). The position of the putative initiator methionine and of the termination codon of the CHox-cad translation product reveals that the C33 cDNA contains 250 and 1492 bp of 5′ and 3′ untranslated sequences, respectively. In the 3′ untranslated sequence, 19 bp from the end of the cDNA there is an AATAAA polyadenylation signal (Bimstiel et al. 1985) and the cDNA clone ends with a short poly A tail.

Comparison of the genomic and cDNA sequences corroborated our observations regarding the exonintron organization of the CHox-cad gene described earlier. As expected, the cDNA clone did not contain the intron that interrupts the homeobox sequence, giving rise to a contiguous 61 amino acid homeodomain. The CHox-cad putative protein product also contains the hexapeptide Pro-Tyr-Glu-Trp-Met-Arg (Fig. 3, underlined) which has been found in a number of homeodomain proteins (Schughart et al. 1988). In all cases described, the hexapeptide is present upstream to the homeodomain. Furthermore, the genomic sequence suggests that upstream from the homeodomain there is an intron, an observation supported by the cDNA sequence. Thus, the hexapeptide and the homeodomain in CHox-cad are also in different exons.

Detailed comparison between the CHox-cad, Cdx1 and cad putative protein products showed that they are most homologous in the region of the homeodomain (Fig. 3). This region of homology can be extended up to the hexapeptide sequence. The five ammo acids immediately upstream to the homeodomain, Gly-Lys-Thr-Arg-Thr, are identical among the three proteins. Upstream of these five amino acids CHox-cad and Cdx1 have 7 and 9 amino acid regions, respectively, which, though not identical, are composed mostly of conservative changes. Cad shows no similarity to the other two proteins in this region. In the region of the hexapeptide, Cdxl and CHox-cad share 12 identical amino acids including hexapeptide (Fig. 3). The cad protein shows some similarity to CHox-cad in this region with mostly conserved rather than identical amino acids. Upstream of the hexapeptide region, the homology between CHox-cad and the other two proteins decreases, exhibiting only small conserved regions of 4 – 5 similar amino acids (Fig. 3). A further region of similarity between CHox-cad and Cdx1 was observed when the Cdxl cDNA sequence was translated from its 5 ′ end. This region, Pro-Ala-Lys-Glu-Asp-Trp in CHox-cad, is, 56 amino acids downstream of the putative initiator methionine. In Cdx1 the sequence is Ala-Pro-Lys-Asp-Asp-Trp and it begins 9 amino acids upstream of the previously suggested putative initiation site (Duprey et al. 1988). This region of similarity between the two proteins suggests that the Cdx1 protein initiates further upstream.

Downstream from the cad homeodomain, the protein continues for 169 amino acids until the first in-frame stop codon. In the case of Cdx1 and CHox-cad, the stop codon is localized 54 and 51 amino acids, respectively, downstream from the homeodomain. Searching the cad downstream protein sequence revealed short regions of similarity that usually are shared with one but not the other vertebrate protein products. Comparison between CHox-cad and Cdxl in this region showed that the homeobox homology can be extended six amino acids downstream. The homology between the proteins then decreases and it increases towards the end of the proteins, where they share seven identical and three similar amino acids out of 11.

The temporal pattern of transcription of the CHox-cad gene during the first days of chicken embryonic development was studied by northern analysis. RNA was prepared from chicken embryos from the time the egg is laid, to 5 days of incubation. During this time period, the chicken embryo develops from the blastoderm stage (stage 1; Hamburger and Hamilton, 1951; H and H) to the middle stages of organogenesis (stage 26). For stages 1–2 and 4–5, total RNA was utilized; for all the later developmental stages, poly (A)⁺ RNA was prepared. The northern blot was probed under high stringency with a 673 bp long SacI genomic fragment from the CHox-cad gene (probe A, Fig. 1A). The results of this hybridization showed that probe A recognizes two transcripts, 1.6 and 2.6 kb in size (Fig. 4A). Accumulation of the two transcripts begins very early during embryonic development and reaches maximal levels at stages 4–5 (16h of incubation). At stage 5, the primitive streak in the embryo reaches its maximal length and the embryo begins neurulation (H and H). Following stages 4–5, the transcript levels begin to decrease until they become altogether undetectable, the rate of decrease of the two transcripts, however, is different. The 1.6 kb transcript is no longer present by day 4 of embryonic development (stage 21-22) and the 2.6 kb transcript is undetectable by 5 days of incubation (stage 26).

The northern analysis shown in Fig. 4A suggests that early CHox-cad transcription correlates with gastrulation and the formation of the primitive streak. In order to map temporally in detail the first embryonic stages during which CHox-cad transcription is evident, RNA was prepared from blastoderm-stage embryos and older. The difference in this instance was that the embryos were staged according to Eyal-Giladi and Kochav (1976) for stages X to XIV which are prestreak stages (a subdivision of stage 1, H and H). Older embryos (stage 2 and older) were staged according to H and H. The blot was probed with a 1119 bp fragment from the CHox-cad cDNA (see below) specific for the 2.6 kb transcript (probe B, Fig. IB). Hybridization with this probe could not detect any C77ox-cad-specific transcripts in embryos from stages X to XIII. The first appearance of the 2.6kb transcript is in embryonic stages that correspond to the formation of the primitive streak (stages XIV-4, Fig. 4B). This result further strengthens the correlation between the onset of gastrulation and the onset of CHox-cad transcription.

To increase the sensitivity of detection of the CHox-cad transcript, the northern analysis of CHox-cad expression at stages 21–22 and 26 was repeated using 8qg of poly (A)⁺ RNA per lane instead of 1 μg as in Fig. 4A and probed with probe A (Fig. 4C). This result supports the previous observation that by day 4 (stage 21–22) the 1.6 kb transcript is absent and by day 5 (stage 26) no CHox-cad transcripts remain. Fig. 4C shows the control hybridization of the same blot to the CHox 3 probe which is expressed at constant levels during these developmental stages (Rangini et al. 1989). Northern analysis performed on mRNA prepared from embryos after 6 to 10 days of incubation and adult tissues, such as ovaries, brain, heart, liver and kidney showed no expression of CHox-cad (data not shown).

In order to establish the site of transcription of CHox-cad in the early embryo, stage 5 and 6 (H and H) chicken embryos were analyzed by in situ hybridization. Chicken embryos incubated for about 16 h (stage 5, H&H) were dissected out and processed for in situ hybridization as descnbed (Materials and methods). Serial cross sections of chicken embryos were probed with strand-specific RNA probes prepared from either probe A (Fig. 1A) or probe B, which lacks the homeobox sequence (Fig. IB), both yielding the same results. The in situ hybridization results show that the CHox-cad transcripts are localized to epiblast cells and to cells in the primitive streak (Fig. 5C and 5D). Once cells migrate out of the primitive streak and become mesoderm cells, they become negative for CHox-cad transcripts by in situ hybridization (Fig. 5D). Hybridization to serial sections of the same embryos revealed that maximal levels of CHox-cad transcripts are present in the caudal part of the embryo (Fig. 5C), and more rostral sections show decreasing transcript levels (Fig. 5A and 5B). In all cases, parallel sections were hybridized to sense RNA probes as a negative control. These results suggest that CHox-cad is expressed in the primitive streak and in cells from the epiblast that are being recruited into the primitive streak. In addition, the CHox-cad transcripts are restricted along the anteroposterior embryonic axis and their maximal level is in the caudal region of the embryo. Anterior to Hensen’s node where the neural plate has formed, no CHox-cad transcripts could be detected by this technique. Detailed analysis of the in situ hybridization results of sections of 10 embryos in 5 independent experiments revealed a thinly populated layer of cells under the mesoderm which also contains low levels of CHox-cad transcripts as judged by the grain density in each cell (Figs 5C and 5D; arrows). This CHox-cad- positive layer is one-cell thick and the cells are loosely packed suggesting that these cells are part of the definitive (‘gut’) endoderm as they migrate out of the primitive streak (Stem and Canning, 1988). This observation suggests that CHox-cad might be transcriptionally active in the early endodermal lineage from the onset of gastrulation.

Further evidence that CHox-cad transcript levels are different anterior or posterior to Hensen’s node was obtained by northern analysis. Chicken embryos were incubated to stages 5–6 (16–20 h of incubation) at which time the embryos are beginning neurulation. The embryos were dissected out and sectioned into anterior and posterior parts by cutting the embryo into two parts at right angles to the embryonic axis at the level of Hensen’s node. This dissection at Hensen’s node results in the developing neural plate and notochord being contained in the rostral section and the primitive streak in the caudal section irrespective of the precise position of Hensen’s node along the axis. RNA was prepared from the two regions of the embryos and probed for the 2.6 kb CHox-cad-specific transcript with probe B (Fig. 6). Posterior to Hensen’s node, the 2.6kb transcript is present in high levels, while in the anterior regions of the embryos there is a marked decrease in the level of the 2.6 kb transcript. These data further support the observations obtained by in situ hybridization of early embryos, which indicate the accumulation of CHox-cad transcripts in the caudal region of the embryo.

The spatial pattern of expression of CHox-cad was also studied in embryos at stage 19 (3.5-4 days of incubation). By this stage of development, the 2.6kb transcript is still present but it is in very low abundance (Fig. 4A). Stage 18 (Fig. 7A, 7C and 7E) and stagel9 (Fig. 7B, 7D and 7P) embryos were sectioned and analyzed by in situ hybridization with either probe A or probe B as described (Materials and methods). Both probes gave the same results. At these developmental stages, CHox-cad transcripts are limited to the developing gut. CHox-cad-specihc transcripts can be seen in the foregut (Fig. 7A and 7B) and in the hindgut (Fig. 7E and 7F), where expression is limited to the epithelial lining of the gut, which is of endodermal origin. In embryos at this stage, the gut is forming into a tube with the ventral side of the midgut opening into the yolk sac. The yolk sac is lined by an epithelia of endodermal origin which is also positive for CHox-cad expression (Fig. 7C and 7D). These results show that CHox-cad is expressed in the endodermal lining of the embryonic gut throughout its length.

In this paper, we describe CHox-cad, a novel homeobox gene isolated from the chicken genome. The expression of CHox-cad in primitive-streak-stage embryos is limited to cells of the epiblast, primitive streak and the definitive endoderm. In addition, analysis of serial sections of embryos in several experiments suggested that CHox-cad expression in the epiblast and primitive streak is also restricted with respect to the anterior-posterior axis of the embryo, with maximal transcription levels at the caudal end. Several vertebrate homeobox genes expressed during gastrulation have been reported, such as the murine Hox 1.5 (Gaunt, 1987), Hox 1.6 (Sundin et al. 1990) Hox 2.9 (Frohman et al. 1990) and Evx1 (Bastian and Gruss, 1990) genes, the Xenopus Xhox3 (Ruiz 1 Altaba and Melton, 1989) and Xhox1A (Harvey et al. 1986) and the chicken CHox 3 (Rangini et al. 1989), CHox 7 (Fainsod and Gruenbaum, 1989) and Ghox-lab (Sundin et al. 1990). For some of these genes, in addition to their temporal pattern of expression, the spatial localization of their transcripts in the gastrulating embryo is known. Hox 1.5 is expressed in the ectoderm and mesoderm of primitive streak embryos and exhibits a predominantly posterior localization. A very similar pattern was found for the Evx1 gene. Xhox3 is also expressed in a posteroanterior gradient, but its expression is restricted to the mesoderm. Ghox-lab is expressed in the epiblast, primitive streak and mesoderm of pnmitive-streak-stage embryos with a predominant posterior localization. Hox 2.9 is expressed along the length of the primitive streak and in the mesoderm in the posterior half of the embryo. Comparison of the patterns of expression reveals that CHox-cad shares with some of these genes, the rostrocaudal restriction of transcript accumulation. However, CHox-cad exhibits a novel spatial pattern of germ layer specificity, its expression being restricted to the epiblast, primitive streak and definitive endoderm. Northern analysis of the CHox-cad rostrocaudal gradient revealed that the relative abundance of the transcripts is substantially different in analysis of mRNA extracted from embryonic regions anterior or posterior to Hensen’s node, increasing noticeably posterior to Hensen’s node.

In addition to the expression of CHox-cad in the epiblast, primitive streak and definitive endoderm of early embryos, at later stages, the gene is expressed in the endodermal lining of the embryonic gut including the yolk sac. Several homeobox genes expressed in part or all of the gut of the embryo have been described, such as Hox 1.3 (Dony and Gruss, 1987), Hox 1.4 (Galliot et al. 1989), Hox 1.6 (Duboule and Doll6, 1989), Hox 2.1 (Holland and Hogan, 1988), Hox 5.1 (Featherstone et al. 1988), Hox 6.1 (Sharpe et al. 1988), Cdx1 (Duprey et al. 1988) and XlHbox 8 (Wright et al. 1988). Apart from Cdx1 and XlHbox 8, all the other homeobox genes mentioned are expressed in mesoderm-derived tissues in the gut. In contrast, Cdx1 and XlHbox 8, like CHox-cad, are found in endoderm-derived tissues of the gut, but are expressed at different times during development. The cell type specificity and, in some instances, the spatial restriction of expression of these homeobox genes suggest that several homeobox genes are involved in the differentiation of the vertebrate gut.

Analysis of the fate map of the chicken embryo at the time of gastrulation reveals that endodermal cells are of epiblast origin (Nicolet, 1970). At specific stages during gastrulation, the cells that migrate through the anterior regions of the primitive streak are in their majority destined to become endodermal cells, while in other regions of the primitive streak the contribution to the endoderm is smaller (Nicolet, 1970). These endodermal cells will initially form the lining of the gut, which eventually will give rise to the endodermal epithelia of other organs. The expression of CHox-cad in the epiblast, primitive streak and definitive endoderm during gastrulation and later in the epithelia of the gut correlates with the pathway that the precursor endodermal cells follow from gastrulation to the gut. These observations raise the possibility that CHox-cad becomes transcriptionally active at the onset of gastrulation in endoderm precursor cells in the epiblast. CHox-cad then remains active in the same cells as they migrate and begin differentiating up to day 5 of embryogenesis. This possibility is further supported by the observations of Stem and Canning (1990), which showed that precursor mesoderm and endoderm cells can be labelled with the HNK-1 antibody before the onset of gastrulation. At present, we cannot rule out the possibility that CHox-cad is activated also in mesodermal precursors, but if this is the case the gene is turned off as they migrate out of the primitive streak. This possibility arises from the fact that maximal CHox-cad expression is found in the caudal regions of the primitive streak, which gives rise mainly to mesodermal cells (Nicolet, 1970).

In Drosophila, cad expression begins as a maternal transcript gradient that is replaced by a zygotic transcripts localized to the posterior end of the embryo. At later stages of development, the cad transcripts are localized to the posterior midgut and Malpighian tubules, the posterior midgut being of endodermal origin. Several aspects of the CHox-cad pattern of expression are reminiscent of the cad pattern of expression. Early in embryogenesis, both CHox-cad and cad exhibit a pattern of expression in the form of transcript accumulation in the caudal region of the embryo. Somewhat later in embryonic development both genes are expressed in cells of endodermal origin in the gut and CHox-cad expression is turned off by day 5. Cdx1, on the other hand, is expressed in the differentiating epithelial lining of the intestine in older embryos when CHox-cad expression is undetectable. In contrast to cad maternal expression, neither Cdx1 nor CHox-cad are expressed in the ovary. Cdx1 and CHox-cad, therefore, appear to implement different aspects of the cad expression pattern.

Comparison of the putative protein product of CHox-cad with other homeodomain proteins reveals that it belongs to the cad family of homeobox genes (Scott et al. 1989). This family includes the murine Cdxl (Duprey et al. 1988) and the C. elegans ceh-3 (Burglin et al. 1989) homeobox genes as well as CHox-cad and cad. The highest degree of homology was localized in the region of the homeodomain but, in all cases, was also extended by a number of amino acids upstream to the homeobox. The presence of an intron that interrupts the CHox-cad homeodomain between amino acids 44 and 45 is a relatively rare observation particularly in vertebrate homeobox genes. Only three vertebrate homeobox genes, out of at least 70 members cloned, have been reported to contain a homeobox whose sequence is interrupted by an intron: Xhox3 (Ruiz i Altaba and Melton, 1989) Evx1 and Evx2 (Bastian and Gruss, 1990). These three genes belong to the eve subfamily of homeobox genes and the intron splits the homeodomain in all three genes between amino acids 46 and 47. In Drosophila, a number of homeoboxes have been identified whose sequence is interrupted by an intron. The homeobox introns in the fly have been localized to two locations. In the engrailed and invected genes, the intron is localized between amino acids 17 and 18 (Poole et al. 1985) A second location for homeodomain introns in Drosophila homeobox genes is between amino acids 44 and 45 as in Labial (Mlodzik et al. 1988), Abdominal-B (DeLorenzi et al. 1988), Distal-less (Cohen et al. 1989) and NK-1 (Kim and Nirenberg, 1989). Therefore, the intron in the CHox-cad homeo-box is in a position that is not uncommon for fly homeobox genes.

The comparison between CHox-cad and Cdx1 is of particular interest as they are both vertebrate homeobox genes. It is important to establish whether they represent the same gene in two evolutionary distant organisms, or whether an ancestral homeobox gene underwent duplications and they represent different members of the vertebrate cad gene family. The CHox-cad and Cdx1 proteins were found to be the most similar in the region that extends from the hexapeptide to several amino acids downstream to the homeodomain. In this region, which in CHox-cad is 93 amino acids long and in Cdxl is 94 ammo acids long, both proteins share 82 identical amino acids and 5 conservative changes. From the putative initiator methionine to the hexapeptide region, the Cdx1 protein has 52 residues, of which 13 residues are identical and 12 are conservative changes when compared to CHox-cad. Downstream from the extended homeodomain, a similar level of homology is observed. Further information as to the relation between CHox-cad and Cdx1 comes from the analysis of their temporal patterns of expression. CHox-cad expression was found to be maximal during gastrulation and the beginning of neurulation by northern analysis and in situ hybridization, whereas, in situ hybridization of 7 and 8 day post-coitum (p.c.) mouse embryos, which represent gastrulation and neurulation stages, were found to be negative for Cdx1 expression (Duprey et al. 1988). Northern analysis of Cdx1 expression showed low levels at 10 days p.c. which disappeared until 14 days p.c. when it began to increase, reaching maximal levels at 17 days p.c. Between days 5 and 10 of chickeri embryo development (stages 26 through 36), no CHox-cad transcripts could be detected by northern analysis. Stage 36 chicken embryos (10 days of incubation) roughly parallel in development 16 day p.c. mouse embryos (Sundin et al. 1990). Therefore, the comparison of the temporal patterns of expression suggests that at developmental stages during which CHox-cad is maximally expressed, Cdx1 expression is undetectable. Also, the reverse situation holds true even though the establishment of parallel developmental stages is more complicated later in embryogenesis.

In summary, the comparison of the CHox-cad and Cdx1 protein sequences reveal two related proteins whose evolutionary relation is not clear. Comparison of their temporal patterns of expression showed nonoverlapping patterns. These observations suggest that the genes are involved in different processes during embryonic development. In addition, it is interesting to note the possible source of the two transcripts recognized by probe A. While probe A hybridized to two transcripts, 1.6 and 2.6kb in size, probe B recognizes only the larger of these. The CHox-cad cDNA clone that was isolated and the fact that probe B lacks the homeobox identify the 2.6 kb transcript as a real CHox-cad mRNA. Other regions of the CHox-cad cDNA were also utilized as probes for northern analysis; in all instances probes lacking the homeobox region only hybridized to the 2.6 kb transcript (data not shown). Regarding the 1.6 kb transcript, even though it is recognized under high stringency, and its temporal pattern of expression is similar to that of the larger transcript, its source remains unclear. The only probe that detects this transcript contains the CHox-cad homeobox sequence, raising the possibility of homeobox cross-hybridization. If this is the case then, the 1.6 kb transcript originates from a different homeobox gene whose homeobox sequence is very similar to CHox-cad judging from the hybridization stringency. Support for the possibility that the vertebrate cad family may contain several members comes from the analysis of other homeobox genes in vertebrates. Analysis of the murine Hox complexes suggested that they arose as a result of duplications during evolution (Schughart et al. 1989). Furthermore, two murine eve type genes have been cloned, Evx1 and Evx2, suggesting that duplication of homeobox genes during evolution was not limited to the Hox clusters (Bastian and Gruss, 1990). In a similar manner an ancestral cad may have undergone duplications and divergence and the different genes may have undertaken different functions, but only cloning and analysis of the different members will provide the ultimate answer.

The evidence presented here shows that early CHox-cad expression begins with the onset of gastrulation and reaches maximal levels when the primitive streak reaches its full length. This temporal correlation between the gastrulation process and CHox-cad expression raises the possibility that this gene may be involved in gastrulation. This suggestion is further supported by the expression of CHox-cad in the early endodermal lineage and evidence for a posterior localization of CHox-cad transcription. Several homeobox genes are expressed during gastrulation and they can be divided into those expressed in ectoderm and mesoderm, and those expressed only in the mesoderm. In addition, some of the homeobox genes exhibit a posterior pattern of expression. The expression of multiple homeobox genes during gastrulation exhibiting different germ layer restriction, raises the possibility that a network of homeobox genes is being employed at this early stage of embryonic development as is the case in the Drosophila embryo.

We wish to thank Hefzibah Eyal-Giladi for her help in the interpretation of the in situ hybridizations Rebecca Haffner for the invaluable discussions. Joel Yisraeli and Kenneth Robzyk for critically reading the manuscript. This work was supported by grants from the United States-Israel Binational Science Foundation (86–00014) and grants from the Fund for Basic Research, administered by The Israel Academy of Sciences and Humanities No. 241/87 to A F. and No 193/87 to Y.G. Z.R. was supported by the Israel Ministry of Research and Development and the National Council for Research and Development.

The sequences shown in Figs 2 and 3 have been submitted to the EMBL database and have the accession numbers X57760 and X57761.