We report the isolation of a new homeobox gene from Xenopus laevis genomic DNA. The homeodomain sequence is highly diverged from the prototype Antennapedia sequence, and contains a unique histidine residue in the helix that binds to DNA. The homeo-domain is followed by a 65 amino acid carboxy-terminal domain, the longest found to date in any vertebrate homeobox gene. We have raised specific antibodies against an X1Hbox 8-β-gal fusion protein to determine the spatial and temporal expression of this gene. The nuclear protein first appears in a narrow band of the endoderm at stage 33 and develops into expression within the epithelial cells of the pancreatic anlagen and duodenum. Expression within the pan-creatic epithelium persists into the adult frog. This unprecedented restriction to an anteroposterior band of the endoderm suggests that vertebrate homeobox genes might be involved in specifying positional information not only in the neuroectoderm and mesoderm, but also in the endoderm. Our data suggest that X1Hbox 8 may therefore represent the first member of a new class of position-dependent transcription factors affecting endodermal differentiation.

The majority of homeobox-containing genes in Drosophila specify or interpret positional information along the anteroposterior axis of the fly (Gehring, 1987; Akam, 1987; Ingham, 1988). Vertebrate homeobox-containing genes share a number of structural features with those of Drosophila (Carrasco et al. 1984; Fienberg et al. 1987; Dressier & Gruss, 1988; Boncinelli et al. 1988; Scott et al. 1988). One of the most compelling arguments in favour of vertebrate homeobox genes having a role similar to those of Drosophila is that many of them are expressed in restricted and precisely defined regions along the anteroposterior axis of the embryo (Awgulewitsch et al. 1986; Holland & Hogan, 1988; Oliver et al. 1988), rather than in a tissue-specific or cell type-specific manner. To date, all vertebrate homeobox genes have been found to be expressed in the neuroectoderm and sometimes also in the mesoderm (e.g. Utset et al. 1987; Dony & Gruss, 1987; Toth et al. 1987; Graham et al. 1988), but none of the genes reported are expressed in the endoderm. In Drosophila, a small amount of transient expression of fushi tarazu, engrailed, Ultrabithorax and caudal has previously been noted in small areas of the embryonic gut (Akam & Martinez-Arias, 1985; DiNardo et al. 1985; Fjose et al. 1985; Ingham et al. 1985; Kornberg et al. 1985; Mlodzik et al. 1985; Krause et al. 1988). However, in all of these cases, the analysis did not extend to showing whether the stained cells are endodermal in origin or derived from another germ layer. What-ever the case, it is definite that no Drosophila or vertebrate gene isolated thus far is expressed exclusively in the endoderm.

In this paper, we report the isolation of a new homeobox gene, called X1Hbox 8, which is expressed solely in a narrow band of the endoderm in early Xenopus embryos. As development proceeds X1Hbox 8 protein is restricted to the nucleus of endodermal cells of the duodenum and the developing pancreas. The homeodomain sequence and the unique pattern of expression suggest that previously undiscovered classes of homeodomain protein may be involved in the specification or interpretation of anteroposterior position within the vertebrate endoderm. In vertebrates coculturing of tissue explants has shown that the differentiation pathway followed by the embryonic endoderm (e.g. choice between lung or intestinal epithelium) is induced by the underlying mesoderm (reviewed by Gurdon, 1987; Wessels, 1977). Thus we believe that the gene isolated here promises to be a valuable marker for studying endodermal differentiation along the anteroposterior axis and its earliest induction by mesoderm.

DNA sequence

An XlHbox 8 genomic clone was isolated by screening a Charon 4A library (Wahli & Dawid, 1980) with Drosophila homeobox probes as described previously (Carrasco et al. 1984). The EcoRI fragments were subcloned and a 1·8 kb insert that showed the only weak hybridization with the probe was cut separately with AluI and HaeIII and the EcoRI ends filled in, which allowed subcloning into M13 mp8 restricted with SmaI. Two subclones were used further: an 860 bp AluI fragment and a 350 bp EcoRI /HaeIII fragment (see Fig. 1A). The AluI and EcoRI /HaeIIIfragments are inserted in M13 in the opposite orientation relative to each other. Use of the M13 universal and homeobox-specific primers allowed the determination on both strands of the sequence between the homeobox EcoRI and the HaeIII site downstream of the translation stop codon. Sequencing was by the dideoxynucleotide chain-terminator method of Sanger et al. (1977). After the subcloning of all detectable EcoRI fragments, the ends of which were all double-strand sequenced in a search for the 5′ portion of the homeodomain sequence, the original genomic phage stock was inadvertently lost in a trans-Atlantic laboratory relocation. For this reason, no reliable sequence data extending 5′ to the EcoRI site in the homeobox is presently available. Completion of this sequence must await isolation of new genomic and cDNA clones.

Fig. 1

X1Hbox 8 encodes a novel type of homeodomain protein. (A) Construction of X1Hbox 8 fusion protein. The top line shows the genomic EcoRI fragment (E–E) that carries the 3′most two-thirds of the homeobox and downstream sequences. The far left end shows the rest of the homeobox to indicate the position of the EcoRI site within the homeobox. The middle line shows the AluI/Sau3AI fragment (A–S) used in constructing the β-gal fusion protein in pTRB 2 (bottom line). The proteincoding region is indicated by an open box, the homeobox by a shaded portion within this. The translation termination codon is asterisked. (B) Nucleotide and amino acid sequence of X1Hbox 8. The sequence between the EcoRI and the Heel 11 site was determined on both strands (see Methods). Some additional downstream sequence is shown to indicate the Sau3AI site used in constructing the fusion protein. A PstI site used in determining orientation in pTRB 2 is also shown. The homeodomain is indicated in bold type and a sequence of eight amino acids also found in the ‘potentiator’ region of human glucocorticoid receptor is underlined. (C) Comparison of X1Hbox 8 with other homeodomain sequences. The relevant region of the prototype Antp homeodomain (amino acid position within the 60 residues is indicated above the sequences) is compared first with X1Hbox 2 (Wright et al. 1987) and then with X1Hbox 8. The position of the helix-tum-DNA-recognition helix motif is indicated (De Robertis et al. 1985; Scott et al. 1988). The corresponding part of five other major classes of homeodomain are also presented.

Fig. 1

X1Hbox 8 encodes a novel type of homeodomain protein. (A) Construction of X1Hbox 8 fusion protein. The top line shows the genomic EcoRI fragment (E–E) that carries the 3′most two-thirds of the homeobox and downstream sequences. The far left end shows the rest of the homeobox to indicate the position of the EcoRI site within the homeobox. The middle line shows the AluI/Sau3AI fragment (A–S) used in constructing the β-gal fusion protein in pTRB 2 (bottom line). The proteincoding region is indicated by an open box, the homeobox by a shaded portion within this. The translation termination codon is asterisked. (B) Nucleotide and amino acid sequence of X1Hbox 8. The sequence between the EcoRI and the Heel 11 site was determined on both strands (see Methods). Some additional downstream sequence is shown to indicate the Sau3AI site used in constructing the fusion protein. A PstI site used in determining orientation in pTRB 2 is also shown. The homeodomain is indicated in bold type and a sequence of eight amino acids also found in the ‘potentiator’ region of human glucocorticoid receptor is underlined. (C) Comparison of X1Hbox 8 with other homeodomain sequences. The relevant region of the prototype Antp homeodomain (amino acid position within the 60 residues is indicated above the sequences) is compared first with X1Hbox 2 (Wright et al. 1987) and then with X1Hbox 8. The position of the helix-tum-DNA-recognition helix motif is indicated (De Robertis et al. 1985; Scott et al. 1988). The corresponding part of five other major classes of homeodomain are also presented.

Fusion protein

The fusion protein was constructed as follows. The AluI fragment that had been cloned into the SmaI site was excised from M13 mp8 using the flanking EcoRI and BamHI restriction sites. The fragment was gel-purified and then digested further with Sau3AI (see Fig. 1A). The mixture was ligated to BamHI-cut pTRB 2 (Burglin & De Robertis, 1987), such that only the fragment with BamHI-compatible ends gave viable plasmid. Hence, the last one-third of the homeodomain and all of the carboxy-terminus of X1Hbox 8 was linked in frame with the lac Z gene. The orientation of X1Hbox 8 in pTRB 2 was determined using an internal PstI site (Fig. 1A), and the maintenance of the correct frame over the junction was determined by sequencing double-stranded DNA with a primer hybridizing just 5′ to the pTRB 0/pTRB 1/pTRB 2 polylinker. Fusion proteins were induced in, and purified from, bacterial strain F′11recA, essentially as described by Oliver et al. (1988).

Antibody procedures

Antisera were raised in NZW female rabbits using the protocol given in Oliver et al. (1988), using approximately 1·5–2 mg of fusion protein per inoculation. The first immune serum of reasonable titre was taken after boosting the animal twice. Subsequent boosts yielded antisera of higher titre and specificity. The anti-X1Hbox 8 antibodies were purified by depletion and affinity purification as described by Oliver et al. (1988). In most cases, the X1Hbox 8 antiserum was additionally depleted against a μ -gal fusion protein matrix containing an Antennapedia-like vertebrate homeodomain (X1Hbox 1 fusion protein A in Oliver et al. 1988). This step removed residual anti-E. coli/μ -gal anti-bodies and nonspecific cross-reaction of the antibody with other homeodomains. Immunoblotting experiments with several homeodomain-containing fusion proteins (X1Hbox 1, X1Hbox 2, human Hox 5.3) and fusion protein competition experiments during the immunostaining of sections (data not shown) clearly show that the staining patterns shown in this paper are specific for X1Hbox 8 proteins, and not due to cross-reaction with the homeodomain of other proteins. It should be noted that the distribution of the protein antigen coincides with that of the X1Hbox 8 RNA (Fig. 2). Sections were prepared and stained with affinity-purified antibodies and alkaline phosphatase-coupled second antibodies (Promega) as described in Oliver et al. (1988).

Fig. 2

Spatial distribution of XlHbox 8 transcripts. (A) Stage 38 Xenopus embryos were dissected into five parts (1–5) and RNA extracted from each set of pooled fragments for analysis by RNAse protection as shown below (for full description see Methods). (B) RNAse protection of XlHbox 8 probe by total RNA from whole tadpoles or fragments thereof. Lane assignments are: P, probe incubated alone without RNAse; –, no RNA control; K, 10μg adult kidney RNA control; 1–5, total RNA from 15 fragments of each type indicated in panel A; W, total RNA from 10 whole tadpoles. Probe length is indicated; fulllength protection is at 139 nucleotides (nt). The other major bands (about 8–10 nt shorter) probably represent artefactual cutting within the RNA duplex (see Wright et al. 1987), but could be due to transcripts from other copies of the XlHbox 8 gene within the tetrapioid X. laevis genome with minor allelic variations (see Fritz et al. 1988). The film was deliberately over-exposed as a stringent test of the restriction of XlHbox 8 RNA to fragment 4. (C) RNAse protection of cytoskeletal actin probe by total RNA from the same samples. Lane assignments are exactly as for panel B, except that total RNA from only 9 fragments (lanes 1–5) or 6 whole tadpoles (lane W) was analysed. Probe length is indicated and the position of the cognate cytoskeletal actin band is shown. The film was exposed for approx. 2h to allow the band in fragment 5 to show properly - for this reason the intensities of the bands in lanes K, 1 and W are well outside the linear range of the film. From other exposures, the band intensities are in good agreement with the recovery of total RNA as judged by agarose gel electrophoresis. Markers are single-strand sequence ladders from an M13 subclone of XlHbox 2 (Wright et al. 1987).

Fig. 2

Spatial distribution of XlHbox 8 transcripts. (A) Stage 38 Xenopus embryos were dissected into five parts (1–5) and RNA extracted from each set of pooled fragments for analysis by RNAse protection as shown below (for full description see Methods). (B) RNAse protection of XlHbox 8 probe by total RNA from whole tadpoles or fragments thereof. Lane assignments are: P, probe incubated alone without RNAse; –, no RNA control; K, 10μg adult kidney RNA control; 1–5, total RNA from 15 fragments of each type indicated in panel A; W, total RNA from 10 whole tadpoles. Probe length is indicated; fulllength protection is at 139 nucleotides (nt). The other major bands (about 8–10 nt shorter) probably represent artefactual cutting within the RNA duplex (see Wright et al. 1987), but could be due to transcripts from other copies of the XlHbox 8 gene within the tetrapioid X. laevis genome with minor allelic variations (see Fritz et al. 1988). The film was deliberately over-exposed as a stringent test of the restriction of XlHbox 8 RNA to fragment 4. (C) RNAse protection of cytoskeletal actin probe by total RNA from the same samples. Lane assignments are exactly as for panel B, except that total RNA from only 9 fragments (lanes 1–5) or 6 whole tadpoles (lane W) was analysed. Probe length is indicated and the position of the cognate cytoskeletal actin band is shown. The film was exposed for approx. 2h to allow the band in fragment 5 to show properly - for this reason the intensities of the bands in lanes K, 1 and W are well outside the linear range of the film. From other exposures, the band intensities are in good agreement with the recovery of total RNA as judged by agarose gel electrophoresis. Markers are single-strand sequence ladders from an M13 subclone of XlHbox 2 (Wright et al. 1987).

RNAse protection

RNA was isolated from embryos by a modified proteinase K/SDS method. Stage 38 embryos were anaesthetized in 0·02% MS222 (Sigma) and dissected as shown in Fig. 2. The dissected pieces (Fig. 2A) from thirty stage 38 embryos were pooled. RNA was also extracted from twenty whole tadpoles of the same stage. A pooled sample was immediately added to 400 μ l of 40 mm-Tris-HCl pH7·5, 4mm-EDTA, 1·7% SDS, 209mm-NaCl, 2 mg ml −1 proteinase K (freshly added - no predigestion), homogenized, and incubated at 37°C for 30 min. The samples were then phenol extracted, chloroform extracted, and ethanol precipitated after adding sodium acetate to 0·3 M. The nucleic acids were pelleted, redissolved and reprecipitated overnight by addition of an equal volume of 5 M-LiCl at −20°C. The RNA was pelleted, washed with cold 70% ethanol, redissolved in 200 μl of RNAse-free 40mm-Tris-HCl pH7·9, 10mm-NaCl, 6mm-MgCl2, heated at 65°C for 10 min, then cooled to room temperature. Contaminating DNA was removed thus: 2 μ l (80 units) of RNAsin (Promega Biotec) was added, mixed well and then 4 μl (4units) of RQ1 DNAse (Promega) added and thoroughly mixed before incubating at 37°C for 20 min. The solution was phenol-chloroform extracted and then ethanol precipitated. Absence of DNA and the absolute integrity of ribosomal RNA was checked on an agarose gel before the RNA was subjected to RNAse protection.

For the X1Hbox 8 protection an antisense probe encompassing the AluI to PstI sequence (Fig. 1B) was prepared from a Bluescribe (Stratagene) subclone (construct BS139), while for the actin protection an antisense probe of the coding region of X. laevis cytoskeletal actin was made from a subclone of M13 construct B9.118 (gift of J. B. Gurdon) in Phagescript (Stratagene). RNAse protection was carried out exactly as described in Wright et al. (1987), and products resolved on sequencing-type gels containing the appropriate concentration of acrylamide, with sequencing ladder as markers.

Isolation of a highly divergent Xenopus homeobox gene

A genomic library of X. laevis DNA was screened with an Antennapedia homeobox probe as described previously (Carrasco et al. 1984). A 1·8 kb EcoRI fragment of the X1Hbox 8 genomic clone which hybridized weakly with the probe contains the last 126 nucleotides of the homeobox and sequences 3′ to it (Fig. 1A). This fragment was subcloned (see Materials and methods) and most of the resulting sequence is shown in Fig. 1B. The homeodomain is followed by an additional 65 amino acids before the translation termination codon is reached. This is the longest carboxyl terminal domain yet reported in any vertebrate homeodomain protein, and its unusual length permitted the generation of high-titre specific antibodies from a β -galactosidase fusion protein derived from this genomic subclone (Fig. 1A).

The only similarity to other proteins arising from a computer search using the carboxy-terminal domain lies in a string of eight amino acids (underlined in Fig. 1B) that is also present in the ‘potentiator’ region of the human glucocorticoid receptor (Hollenberg et al. 1985; Seveme et al. 1988). The function of this sequence, if any, is unknown.

The sequence of only the last 42 amino acids of the homeodomain is available because the X1Hbox 8 homeobox contains an EcoRI site (see Materials and methods for explanation). Of the nine X. laevis homeodomains isolated so far (De Robertis et al. 1988; Harvey et al. 1986), X1Hbox 8 is most diverged from the ancestral Antennapedia sequence. As shown in Fig. 1C, the new X1Hbox 8 homeodomain does not tend to resemble any of the other homeobox families described for Drosophila. One amino acid change that is particularly noteworthy is a change in amino acid 44 of the homeodomain (Gln to His), which is located in the ‘recognition helix’ (Fig. 1C) that determines DNA-binding specificity (Desplan et al. 1988; Hoey & Levine, 1988). No other homeodomain isolated in any organism has histidine in this position (see Scott et al. 1988, for a compilation of 87 homeodomain sequences). Therefore, it is possible that this change may endow the X1Hbox 8 homeodomain protein with unique DNA sequence recognition properties.

We conclude that X1Hbox 8 is a novel type of homeobox gene that differs from those characterized previously in several respects. When the pattern of X1Hbox 8 expression was analysed in detail, its uniqueness became even more apparent.

XlHbox 8 is expressed in a narrow endodermal band

We analysed the spatial expression of X1Hbox 8 initially by using a sensitive RNAse protection assay (Materials and methods). Fig. 2 shows an experiment in which hatching tadpoles (stage 38) were dissected into five fragments. X1Hbox 8 RNA expression was restricted to a single fragment (Fig. 2, lane 4) containing the anterior digestive tract, branchial arches, developing heart, and lateral body wall. No transcripts are detected in fragments containing central nervous system. The distribution of X1Hbox 8 transcripts determined by RNAse protection is in strict agreement with the protein distribution pattern determined by immunolocalization which is described below.

Northern blots of RNA from various stages of development failed to detect XlHbox 8 transcripts specifically presumably because of the small number of cells that express this gene. However, we analysed the spatial and temporal expression of the XlHbox 8 homeodomain protein more precisely by collecting embryos at various stages of development, serially sectioning the entire animal, and immunostaining with anti-X1Hbox 8 antibodies. Antibodies were raised in rabbits using a β-gai fusion protein (Fig. 1A) as antigen. Antibodies reacting with E. coli proteins or with other Antennapedia-type. homeodomains were removed by depletion over E. coli-Sepharose and X1Hbox 1 homeodomain-Sepharose columns. The antiserum was then affinity-purified over an XlHbox 8 fusion protein column (Oliver et al. 1988). By several criteria, the resulting antibodies react specifically with the carboxy-terminal region of the XlHbox 8 protein (see Materials and methods).

Immunolocalization of X1Hbox 8 protein in Xenopus embryos of various stages (staged according to Nieuwkoop & Faber, 1967) is summarized in Fig. 3. At all stages of development, the protein is nuclear and localized exclusively in cells of endodermal origin. During early development immunostaining is undetectable until stage 33 (tailbud) when a narrow band of endodermal nuclei becomes weakly stained (bracketed in panel A). Examination of serial sections showed that this band spans the region of the embryonic endoderm located between the third and fifth postotic somites. The unstained endodermal regions anterior to this band will give rise to pharynx, oesophagus and stomach, while the stained region will give rise to duodenum (see below). At stage 38 (hatching tadpole, panel B) and stage 41 (swimming tadpole, panels C to F) the narrow band of nuclear staining, which spans the entire endoderm dorsoventrally, is more intensely stained. Examination of sections counterstained with the DNA stain Hoechst 33258 reveals that all endodermal nuclei within this narrow region contain the X1Hbox 8 protein (panels D and F). Intense nuclear staining is also detected in the dorsal and ventral pancreatic anlagen which evaginate from this endodermal region during development. The anatomical location of the Xenopus pancreas can be determined readily in serial sections thanks to the detailed description of normal development provided by Nieuwkoop & Faber (1967). Whole-mount staining of tadpoles with X1Hbox 8 (see Fig. 4) confirms this assignment. The dorsal and ventral pancreas can be seen best in Fig. 3, panel B. The liver, which is localized more anteriorly, is entirely negative (panels B, C, G and H). The pancreatic anlagen are connected to the duodenum via an excretory duct system, the epithelium of which strongly expresses X1Hbox 8 throughout its length (panels D, E and G).

Fig. 3

Immunolocalization of XlHbox 8 protein. Xenopus embryos staged according to Nieuwkoop & Faber (1967) were serially sectioned and stained with anti-XIHbox 8 antibody. (A) Frontal section of a stage 33 embryo. A narrow band of expression of the XlHbox 8 protein (bracketed) is faintly visible in the endoderm slightly posterior to the pharynx. (B) Sagittal section through the left half of a stage 38 embryo. In this section the anterior border where expression of the XlHbox 8 antigen starts is clearly visible. The dorsal and ventral pancreatic rudiments are positive for the antigen as indicated. The stained endodermal tube in the centre will later mature into duodenum (note the cavity at its anterior tip). The liver is entirely unstained. (C) Sagittal section through a stage 41 embryo. Note the band of expression in the endoderm (dark stain) while the remainder of the embryo is unstained. (D) Transverse section (slightly oblique angle) through a stage 41 embryo at low magnification. Only endodermal cells are stained; neuroectoderm and mesoderm are unstained. (E) Same section as in D but at higher magnification. Note XlHbox 8 expression in the gut epithelium as well as in the pancreatic excretory duct. All nuclei in the endodermal band are stained; the lower right comer appears unstained because of the slightly oblique plane of sectioning. (F) The same section as in E counterstained with the fluorescent DNA stain Hoechst 33258. (G) Section through a stage 49 embryo. All nuclei in the epithelium of the duodenum and pancreatic ducts are stained, but only a small percentage of cells within the pancreas proper contain XlHbox 8 protein at this stage. No XlHbox 8 protein is expressed in liver, stomach, or small intestine. (H) Same section shown in G at higher magnification. (I) Section through an adult Xenopus pancreas. All the nuclei in the pancreatic duct system are stained with anti-XIHbox 8 antibody. The nuclei of cells in the acini are weakly stained. Abbreviations; ph, pharynx; en, endoderm; p, pancreas; dp, dorsal pancreatic anlage; vp, ventral pancreatic anlage; I, liver; pd, pancreatic excretory duct; s, stomach; in, intestine; cns, central nervous system; pn, pronephros.

Fig. 3

Immunolocalization of XlHbox 8 protein. Xenopus embryos staged according to Nieuwkoop & Faber (1967) were serially sectioned and stained with anti-XIHbox 8 antibody. (A) Frontal section of a stage 33 embryo. A narrow band of expression of the XlHbox 8 protein (bracketed) is faintly visible in the endoderm slightly posterior to the pharynx. (B) Sagittal section through the left half of a stage 38 embryo. In this section the anterior border where expression of the XlHbox 8 antigen starts is clearly visible. The dorsal and ventral pancreatic rudiments are positive for the antigen as indicated. The stained endodermal tube in the centre will later mature into duodenum (note the cavity at its anterior tip). The liver is entirely unstained. (C) Sagittal section through a stage 41 embryo. Note the band of expression in the endoderm (dark stain) while the remainder of the embryo is unstained. (D) Transverse section (slightly oblique angle) through a stage 41 embryo at low magnification. Only endodermal cells are stained; neuroectoderm and mesoderm are unstained. (E) Same section as in D but at higher magnification. Note XlHbox 8 expression in the gut epithelium as well as in the pancreatic excretory duct. All nuclei in the endodermal band are stained; the lower right comer appears unstained because of the slightly oblique plane of sectioning. (F) The same section as in E counterstained with the fluorescent DNA stain Hoechst 33258. (G) Section through a stage 49 embryo. All nuclei in the epithelium of the duodenum and pancreatic ducts are stained, but only a small percentage of cells within the pancreas proper contain XlHbox 8 protein at this stage. No XlHbox 8 protein is expressed in liver, stomach, or small intestine. (H) Same section shown in G at higher magnification. (I) Section through an adult Xenopus pancreas. All the nuclei in the pancreatic duct system are stained with anti-XIHbox 8 antibody. The nuclei of cells in the acini are weakly stained. Abbreviations; ph, pharynx; en, endoderm; p, pancreas; dp, dorsal pancreatic anlage; vp, ventral pancreatic anlage; I, liver; pd, pancreatic excretory duct; s, stomach; in, intestine; cns, central nervous system; pn, pronephros.

Fig. 4

Spatial expression pattern of XlHbox 8 protein visualized by antibody whole-mount staining. (A) A side view of a stage 40 tadpole immunostained with XlHbox 8 primary antibody and peroxidase-coupled anti-rabbit second antibody, visualized with diaminobenzidine/hydrogen peroxide, and cleared in benzyl benzoate/benzyl alcohol. Samples were processed exactly as described by M. Klymowsky (personal communication). (B) Diagrammatic interpretation of panel A. The regions labelled dp and vp are the dorsal and ventral pancreatic rudiment, respectively; duo represents the region of the duodenum. The epithelial nuclei of these regions together constitute the only specific XlHbox 8 staining in the embryo. The punctate appearance of the lower left edge of the embryo is a photographic artefact caused by the extreme curvature of the tadpole combined with nonuniform illumination by the microscope condenser. This whole-mount embryo staining allows us to conclude that the endoderm is the only place where XlHbox 8 antigen is found. Photomicrograph was taken in a BioRad confocal microscope using an image-enhancement program.

Fig. 4

Spatial expression pattern of XlHbox 8 protein visualized by antibody whole-mount staining. (A) A side view of a stage 40 tadpole immunostained with XlHbox 8 primary antibody and peroxidase-coupled anti-rabbit second antibody, visualized with diaminobenzidine/hydrogen peroxide, and cleared in benzyl benzoate/benzyl alcohol. Samples were processed exactly as described by M. Klymowsky (personal communication). (B) Diagrammatic interpretation of panel A. The regions labelled dp and vp are the dorsal and ventral pancreatic rudiment, respectively; duo represents the region of the duodenum. The epithelial nuclei of these regions together constitute the only specific XlHbox 8 staining in the embryo. The punctate appearance of the lower left edge of the embryo is a photographic artefact caused by the extreme curvature of the tadpole combined with nonuniform illumination by the microscope condenser. This whole-mount embryo staining allows us to conclude that the endoderm is the only place where XlHbox 8 antigen is found. Photomicrograph was taken in a BioRad confocal microscope using an image-enhancement program.

XlHbox 8 antigen distribution in the digestive tract was also analysed in serial sections of later embryos (stages 45, 47, 48 and 49 - which correspond to 4, 5, 7 and 12 days of development) and of adult pancreas. By day 12 the endoderm in the duodenal region has evolved into a single layer of columnar epithelial cells all of which strongly express X1Hbox 8 (Fig. 3, panels G and H), except for the initial third of the duodenum (ventral horizontal segment - see Nieuwkoop & Faber, 1967) which is unstained (not shown). At this stage, the dorsal and ventral pancreatic anlagen have fused to form a single organ in which the number of strongly stained nuclei has decreased substantially. Less than 10% of the cells express X1Hbox 8 strongly (Fig. 3, panels G and H), in contrast to earlier stages when the vast majority of presumptive pancreatic cells were stained (panel B). All the epithelial nuclei of the pancreatic excretory duct system are strongly stained. X1Hbox 8 protein is not expressed detectably in stomach, liver, or small intestine (panel G), or in gall bladder or bile duct (not shown).

Nuclear X1Hbox 8 protein is also detectable in the pancreas of adult frogs. Fig. 3, panel I, shows that the most intense staining is found in the nuclei of the excretory tubules. The nuclei of cells in the exocrine acini are also stained above background, although much more weakly. The cells in the Islets of Langer-hans are not stained (data not shown).

Fig. 4 depicts a stage 40 Xenopus tadpole stained for X1Hbox 8 in a whole-mount procedure. This is presented to help in reconstructing the pattern of X1Hbox 8 protein expression determined from the section stainings shown in Fig. 3. This type of staining confirms the view that this homeo protein is exclusively localized in epithelial nuclei of the region of the posterior foregut.

We conclude that X1Hbox 8 homeodomain-containing protein is found only in nuclei of cells lying across a narrow band of the endoderm in early Xenopus embryos. These cells continue to express the antigen as they mature into the epithelial lining of the duodenum, the pancreatic epithelium and the epithelium of the interconnecting duct system.

We report here the isolation of a new vertebrate homeobox-containing gene. Of the nine homeobox genes isolated to date from Xenopus (sequences listed in De Robertis et al. 1988), the XlHbox 8 homeodomain is most diverged from the prototype Antennapedia homeodomain. X1Hbox 8 has an unusual Gin to His change in the proposed DNA recognition helix. A histidine residue in this position is not found in a homeodomain from any organism (Scott et al. 1988). In addition, X1Hbox 8 has a carboxy-terminal domain of 65 amino acids downstream of the homeodomain, the longest of any vertebrate homeodomain protein.

The most exceptional feature of X1Hbox 8 is that it is the only homeobox gene known to be expressed exclusively in the endoderm. The X1Hbox 8 protein is found only in the nuclei of a narrow band of endodermal cells as early as stage 33. Our detailed analysis demonstrates that X1Hbox 8 is expressed in the epithelium of the duodenum and of the pancreas, an endodermal organ that evaginates from this region. Expression in the pancreas and its excretory duct system continues into the adult. The distributions of gene products from a large number of vertebrate homeobox genes have been described. These were always found to be expressed in the neuroectoderm and/or in the mesoderm, usually in derivatives of both of these germ layers (for example, Oliver et al. 1988) but never in the endoderm.

What might be the function of the long carboxyterminal domain of X1Hbox 8? We might draw a comparison between this protein and the evenskipped protein of Drosophila. The eve protein has a long carboxy-terminal domain which has been found to change the DNA-binding specificity of the homeodomain lying upstream of it (Hoey & Levine, 1988; Hoey et al. 1988). It may be that the carboxy-terminal region of X1Hbox 8 acts synergistically with the His change in the recognition helix to give X1Hbox 8 proteins unique DNA-binding characteristics and hence unique target gene specificities.

XlHbox 8 is expressed exclusively in a very narrow slice of the gut endoderm - the posterior part of the foregut. One might speculate that other homeobox genes may be expressed in bands in other regions along the anteroposterior axis of the endoderm. Why have these putative homeobox genes not been detected yet? Hybridization of the standard Antenna-pedia-type DNA probes to the X1Hbox 8 homeobox sequence is weak. There is only 67% nucleotide sequence homology between the region of X1Hbox 8 reported here and Antp, and is probably even less in the first third of the homeobox (not sequenced here) because this region is usually less conserved than the region of the DNA-binding helices. Perhaps an X1Hbox 8 homeobox probe will allow the isolation of other genes contained within an X1Hbox 8-type family, expressed in endoderm, by low stringency screening of gene libraries. One potentially useful application of X1Hbox 8 probes may be in the diagnosis of the site of origin of intestinal tract carcinomas.

Vertebrate endoderm/mesoderm coculture experiments have shown that the type of epithelial differentiation undergone by many portions of the digestive tract is ‘instructed’ by induction by the underlying mesenchyme (Gurdon, 1987; Saxen, 1977; Wessels & Rutter, 1969; Wessels & Cohen, 1967; Rutter et al. 1964). For example, chick allantois (which originates from the posterior endoderm) is capable of differentiating into gizzard-, intestine- or lung-type epithelium when cultured in close contact with mesenchymes from early organs of these types. Similarly, lung epithelium that has not yet been instructed by mesoderm can be induced to form stomach or intestinal epithelium if apposed in culture together with mesoderm obtained from these regions of the gut (Wessels, 1977). X1Hbox 8 expression is detected in a narrow band of the gut long before the pancreatic anlagen are formed (Fig. 3A). If its expression is triggered by mesodermal induction, then its activation precedes any sign of pancreatic differentiation. After the pancreas is formed, its continued growth requires growth factors (‘permissive induction’) from mesoderm, but these can be provided by a variety of mesenchymes or even cell extracts (Wessells, 1977). Using X1Hbox 8 probes as sensitive markers should allow experiments addressing the question of whether duodenal differentiation requires ‘instructive induction’ from a specific type of mesoderm, as is the case in other parts of the endodermal tube.

Expression of several homeobox genes in the mesoderm that surrounds the endodermal epithelium has been observed in vertebrate embryos (Holland & Hogan, 1988; Dony & Gruss, 1987; Oliver et al. 1988). Clearly, different homeodomain proteins are expressed along the anteroposterior axis of the gut mesoderm (Oliver et al. 1988; Graham et al. 1988). It is also interesting to note that Slack (1985, 1986) has suggested that the most frequent homeotic transformation in man is digestive tract metaplasia. It is not uncommon, for example, for adult humans to have patches of intestinal epithelium in the stomach or patches of gastric epithelium in the oesophagus. XlHbox 8 is expressed exclusively in the endoderm, conceivably is involved in the earliest specification of epithelial identity, and might even be a target for an epithelial instruction signal from the mesoderm. This proposition might be tested experimentally by microinjecting X1Hbox 8 protein or synthetic mRNAs into the parts of Xenopus embryos (vegetal blastomeres) that give rise to endodermal structures, in an attempt to change the epithelial fate. Markers are available (e.g. specific digestive enzymes) which can be used to test the efficacy of such a transformation. Preliminary experiments in which X1Hbox 8 antibodies were injected into embryos in the hope of causing a phenotypic defect, as described for X1Hbox 1 antibodies by Cho et al. (1988), have so far failed to produce significant changes in the organization of the tadpole gut.

X1Hbox 8 may provide an excellent marker for the re-examination and characterization of the mesen-chymal/endodermal induction process at the molecular level. Faced with such an intriguing distribution and potentially powerful ways to explore and manipulate the expression of this gene we are now undertaking a complete characterization of the X1Hbox 8 gene, beginning with the cloning of full-length cDNAs.

We acknowledge Jane Hardwicke for expert technical assistance during the sectioning, staining and photographic work, and Clyde Lulham (BioRad Microsciences Division) for image analysis of whole-mount embryos. We thank Larry Zipursky and John Merriam for comments on the manuscript. We recognize the invaluable contribution of the laboratory of M. Klymkowsky for the Xenopus wholemount procedure. This work was supported by NIH grant HD 21502-03. CVEW is an ACS (California Division) Senior Fellow.

Akam
,
M.
(
1987
).
The molecular basis for metameric pattern in the Drosophila embryo
.
Development
101
,
1
22
.
Akam
,
M.
&
Martinez-Arias
,
A.
(
1985
).
The distribution of Ultrabithorajc transcripts in Drosophila embryos
.
EMBO J
.
4
,
1689
1700
.
Awculewitsch
,
A.
,
Utset
,
M. F.
,
Hart
,
C. P.
,
McGinnis
,
W.
&
Ruddle
,
F. H.
(
1986
).
Spatial restriction in expression of a mouse homoeo box locus within the central nervous system
.
Nature, Lond
.
320
,
328
335
.
Boncinelli
,
E.
,
Somma
,
R.
,
Acampora
,
D.
,
Pannese
,
M.
,
D’Esposito
,
M.
&
Simeone
,
A.
(
1988
).
Organization of human homeobox genes
.
Hum. Reprod. (in press)
.
Burglin
,
T. R.
&
De Robertis
,
E. M.
(
1987
).
The nuclear migration signal of Xenopus laevis nucleoplasmin
.
EMBO J
.
6
,
2617
2625
.
Carrasco
,
A. E.
,
McGinnis
,
W.
,
Gehring
,
W. J.
&
De Robertis
,
E. M.
(
1984
).
Cloning of an X. laevis gene expressed during early embryogenesis coding for a peptide region homologous to Drosophila homeotic genes
.
Cell
37
,
409
414
.
Cho
,
K. W. Y.
,
Goetz
,
J.
,
Wright
,
C. V. E.
,
Fritz
,
A. F.
,
Hardwicke
,
J.
&
De Robertis
,
E. M.
(
1988
).
Differential utilization of the same reading frame in a Xenopus homeobox gene encodes two related proteins sharing the same DNA-binding specificity
.
EMBO J
.
7
,
2139
2149
.
De Robertis
,
E. M.
,
Fritz
,
A. F.
,
Goetz
,
J.
,
Martin
,
G.
,
Mattaj
,
I. W.
,
Salo
,
E.
,
Smith
,
G. D.
,
Wright
,
C. V.
&
Zeller
,
R.
(
1985
).
The Xenopus homeo boxes
.
Cold Spring Harbor Symp. Quant. Biol
.
50
,
271
275
.
De Robertis
,
E. M.
,
Burgun
,
T. R.
,
Fritz
,
A. F.
,
Wright
,
C. V. E.
,
Jegalian
,
B.
,
Schnegelsberg
,
P.
,
Bittner
,
D.
,
Morita
,
E.
,
Oliver
,
G.
&
Cho
,
K. W. Y.
(
1988
).
Families of vertebrate homeodomain proteins
.
In DNA-Protein Interactions in Transcription
(ed.
J.
Gralla
).
UCLA Symposium
95
,
107
115
.
Desplan
,
C.
,
Theis
,
J.
&
O’Farrell
,
P. H.
(
1988
).
The sequence specificity of homeodomain-DNA interaction
.
Cell
54
,
1081
1090
.
Dinardo
,
S.
,
Kuner
,
J. M.
,
Theis
,
J.
&
O’Farrell
,
P. H.
(
1985
).
Development of embryonic pattern in Drosophila melanogaster as revealed by accumulation of the nuclear engrailed protein
.
Cell
43
,
59
69
.
Dony
,
C.
&
Gruss
,
P.
(
1987
).
Specific expression of the Hox 1.3 homeo box gene in murine embryonic structures originating from or induced by the mesoderm
.
EMBO J
.
6
,
2965
2975
.
Dressler
,
G. R.
&
Gruss
,
P.
(
1988
).
Do multigene families regulate vertebrate development?
Trends in Genet
.
4
,
214
219
.
Fienberg
,
A. A.
,
Utset
,
M. F.
,
Bogarad
,
L. D.
,
Hart
,
C. P.
,
Awgulewitsch
,
A.
,
Ferguson-Smith
,
A.
,
Fainsod
,
A.
,
Rabin
,
M.
&
Ruddle
,
F. H.
(
1987
).
Homeobox genes in murine development
.
Current Topics in Developmental Biology
23
,
233
257
.
Fjose
,
A.
,
McGinnis
,
W. J.
&
Gehring
,
W. J.
(
1985
).
Isolation of a homoeo box-containing gene from the engrailed region of Drosophila and the spatial distribution of its transcripts
.
Nature, Lond
.
313
,
284
289
.
Fritz
,
A. F.
,
Cho
,
K. W. Y.
,
Wright
,
C. V. E.
,
Jegalian
,
B.
&
De Robertis
,
E. M.
(
1988
).
Duplicated homeobox genes in Xenopus
.
Devl Biol, (in press)
.
Gehring
,
W. J.
(
1987
).
Homeoboxes in the study of development
.
Science
236
,
1245
1252
.
Graham
,
A.
,
Papalopulu
,
N.
,
Lorimer
,
J.
,
McVey
,
J. H.
,
Tuddenham
,
E. G. D.
&
Krumlauf
,
R.
(
1988
).
Characterization of a murine homeo box gene, Hox-2.6, related to the Drosophila Deformed gene
.
Genes Dev
.
2
,
1424
1438
.
Gurdon
,
J. B.
(
1987
).
Embryonic induction - molecular prospects
.
Development
99
,
285
306
.
Harvey
,
R. P.
,
Tabin
,
C. J.
&
Melton
,
D. A.
(
1986
).
Embryonic expression and nuclear localization of Xenopus homeobox (Xhox) gene products
.
EMBO J
.
5
,
1237
1244
.
Hoey
,
T.
&
Levine
,
M.
(
1988
).
Divergent homeo box proteins recognize similar DNA sequences in Drosophila
.
Nature, Lond
.
332
,
858
861
.
Hoey
,
T.
,
Warrior
,
R.
,
Manak
,
J.
&
Levine
,
M.
(
1988
).
DNA-binding activities of the Drosophila melanogaster even-skipped protein are mediated by its homeodomain and influenced by protein context
.
Mol. cell Biol
.
8
,
4598
4607
.
Holland
,
P. W. H.
&
Hogan
,
B. L.
(
1988
).
Spatially restricted patterns of expression of the homeobox-containing gene Hox 2.1 during mouse embryogenesis
.
Development
102
,
159
174
.
Hollenberg
,
S. M.
,
Weinberger
,
C.
,
Ong
,
E. S.
,
Cerelli
,
G.
,
Oro
,
A.
,
Lebo
,
R.
,
Thompson
,
E. B.
,
Rosenfeld
,
M. G.
&
Evans
,
R. M.
(
1985
).
Primary structure and expression of a functional human glucocorticoid receptor cDNA
.
Nature, Lond
.
318
,
635
641
.
Ingham
,
P. W.
(
1988
).
The molecular genetics of embryonic pattern formation in Drosophila
.
Nature, Lond
.
335
,
25
34
.
Ingham
,
P. W.
,
Howard
,
K. R.
&
Ish-Horowicz
,
D.
(
1985
).
Transcription pattern of the Drosophila segmentation gene hairy
.
Nature, Lond
.
318
,
439
445
.
Kornberg
,
T.
,
Siden
,
I.
,
O’Farrell
,
P.
&
Simon
,
M.
(
1985
).
The engrailed locus of Drosophila: In situ localization of transcripts reveals compartment specific expression
.
Cell
40
,
45
53
.
Krause
,
H. M.
,
Klemenz
,
R.
&
Gehring
,
W. J.
(
1988
).
Expression, modification and localization of the fushi-tarazu protein in Drosophila embryos
.
Genes Dev
.
2
,
1021
1036
.
Mlodzik
,
M.
,
Fjose
,
A.
&
Gehring
,
W. J.
(
1985
).
Isolation of caudal, a Drosophila homeobox containing gene with maternal expression, whose transcripts form a concentration gradient at the pre-blastoderm stage
.
EMBO J
.
4
,
2961
2969
.
Nieuwkoop
,
P. D.
&
Faber
,
J.
(
1967
).
Normal Table of Xenopus laevis (Daudin
., 2nd edition.
Amsterdam
:
North Holland Publ. Co
.
Oliver
,
G.
,
Wright
,
C. V. E.
,
Hardwicke
,
J.
&
De Robertis
,
E. M.
(
1988
).
Differential antero-posterior expression of two proteins encoded by a homeobox gene in Xenopus and mouse embryos
.
EMBO J
.
7
,
3199
3209
.
Rutter
,
W. J.
,
Wessells
,
N. K.
&
Grobstein
,
C.
(
1964
).
Control of specific synthesis in the developing pancreas
.
Natl. Cancer Inst. Monogr
.
13
,
51
65
.
Sanger
,
F.
,
Nicklen
,
S.
&
Coulson
,
A. R.
(
1977
).
DNA sequencing with chain-terminating inhibitors
.
Proc. natn. Acad. Sei. U.S.A
.
74
,
5463
5467
.
Saxen
,
L.
(
1977
).
Directive versus permissive induction: a working hypothesis
.
In Cell and Tissue Interactions
(ed.
J. W.
Lash
&
M. M.
Burger
), pp.
1
9
.
New York
:
Raven Press
.
Scott
,
M. P.
,
Tamkun
,
J. W.
&
Hartzell
,
G. W.
(
1988
).
The structure and function of the homeodomain
.
BBA Reviews on Cancer (in press)
.
Severne
,
Y.
,
Wieland
,
S.
,
Schaffner
,
W.
&
Rusconi
,
S.
(
1988
).
Metal binding ‘finger’ structures in the glucocorticoid receptor defined by site-directed mutagenesis
.
EMBO J
.
7
,
2503
2508
.
Slack
,
J. M. W.
(
1985
).
Homeotic transformations in man: implications for the mechanism of embryonic development and for the organization of epithelia
.
J. theor. Biol
.
114
,
463
490
.
Slack
,
J. M. W.
(
1986
).
Epithelial metaplasia and the second anatomy
.
The Lancet, August
2
, pp.
268
271
.
Toth
,
L.
,
Slawin
,
K. L.
,
Pintar
,
J. E.
&
Nguyen-Huu
,
M. C.
(
1987
).
Region-specific expression of a mouse homeobox gene in the embryonic mesoderm and central nervous system
.
Proc, natn. Acad. Sci. U.S.A
.
84
,
6790
6794
.
Utset
,
M. F.
,
Awgulewitsch
,
A.
,
Ruddle
,
F. H.
&
McGinnis
,
W.
(
1987
).
Region-specific expression of two mouse homeo box genes
.
Science
235
,
1379
1382
.
Wahli
,
W.
&
Dawid
,
I. B.
(
1980
).
Isolation of two closely related vitellogenin genes, including their flanking regions, from a Xenopus laevis gene library
.
Proc. natn. Acad. Sci. U.S.A
.
77
,
1437
1441
.
Wessels
,
N. K.
(
1977
).
In Tissue Interactions and Development
, pp.
105
127
.
California
:
Benjamin
.
Wessells
,
N. K.
&
Cohen
,
J. H.
(
1967
).
Early pancreas organogenesis: morphogenesis, tissue interactions, and mass effects
.
Devi Biol
.
15
,
237
270
.
Wessells
,
N. K.
&
Rutter
,
W. J.
(
1969
).
Phases in cell differentiation
.
Scientific American
220
,
36
44
.
Wright
,
C. V. E.
,
Cho
,
K. W. Y.
,
Fritz
,
A. F.
,
Burglin
,
T. R.
&
De Robertis
,
E. M.
(
1987
).
A Xenopus laevis gene encodes both homeobox-containing and homeobox-less transcripts
.
EM BO J
.
6
,
4083
4094
.

While this paper was in press, the distribution of mouse gene Cdx-1, which is most similar to the Drosophila gene caudal, was reported (Duprey, P., Chowdhury, K., Dressier, G. R., Balling, R., Simon, D., Guenet, J.-L. & Gruss, P., 1988; A mouse gene homologous to the Drosophila gene caudal is expressed in epithelial cells from the embryonic intestine. Genes Dev. 2, 1647–1654.). The endoderm-specific expression of the Cdx-1 gene is different from that of XlHbox 8 in at least two respects. First, XlHbox 8 is expressed more anteriorly within the gastrointestinal tract than the Cdx-1 gene. Second, Cdx-1 mRNA is first found quite late in intestinal differentiation, when the villi are just about to form. In contrast, X1Hbox 8 protein is already found in a restricted endodermal band long before morphogenesis and differentiation of the duodenum and pancreatic anlagen.