In the vertebrate blastula, a small group of several hundred endoderm cells gives rise to the epithelial lining of the respiratory and gastrointestinal tract as well as to the liver, lungs, pancreas, thyroid, and thymus (Wells and Melton, 1999). Recently, work in Xenopus laevis has elucidated the framework of a conserved molecular pathway that initiates vertebrate endoderm development (reviewed in Stainier, 2002).
In the current model, the maternal T-box transcription factor VegT, which is localized to the vegetal region of the Xenopus embryo (Zhang and King, 1996; Stennard et al., 1996; Lustig et al., 1996; Horb and Thomsen, 1997), initiates endoderm development. VegT activity results in the transcription of zygotic endodermal genes (Zhang et al., 1998; Clements et al., 1999; Xanthos et al., 2001), which include the following: nodal-related proteins (Xnr1,2,4,5,6) and derriere, which are members of the transforming growth factor-beta (TGFβ) growth factor family (Jones et al., 1995; Joseph and Melton, 1997; Sun et al., 1999; Takahashi et al., 2000); homeodomain proteins of the Mixer/Mix/Bix family (Rosa, 1989; Vize, 1996; Henry and Melton, 1998; Tada et al., 1998; Casey et al., 1999); the zinc finger factors Gata 4, 5, and 6 (Jiang and Evans, 1996; Weber et al., 2000; Xanthos et al., 2001); and two closely related HMG domain transcription factors, Sox17α and Sox17β (Husdon et al., 1997).
VegT appears to function primarily by means of TGFβ signaling (Kofron et al., 1999; Xanthos et al., 2001), which is required for activating and/or maintaining the expression of the Mix-like genes, Gata4/5/6, and Sox17α/β (Henry et al., 1996; Clements et al., 1999; Yasuo and Lemaire, 1999; Xanthos et al., 2001). These transcription factors promote endodermal cell fate and the expression of endoderm marker genes such as the hepatic nuclear factors (HNF1β and FoxA1/HNF3; Husdon et al., 1997; Henry and Melton, 1998; Alexander and Stainier, 1999; Weber et al., 2000; Xanthos et al., 2001; Stainier, 2002). As a result, vegetal cells are committed to form endodermal tissues by the midgastrula stage (Wylie et al., 1987).
Early endodermal patterning occurs simultaneously with general endoderm specification, as evident by the asymmetric expression of the homeodomain transcription factor Hex (Newman et al., 1997) and the secreted growth factor antagonist Cerberus (Bouwmeester et al., 1996) in the anterior endoderm. The available evidence indicates that the overlapping activities of endoderm-inducing nodal signals and the maternal Wnt/β-catenin signals induce the expression of Hex and Cerberus in the anterior endoderm (Zorn et al., 1999).
Although the genes described above have been implicated in endoderm specification, their precise roles are unclear. The majority of functional studies to date have relied on overexpression experiments in Xenopus laevis. To rigorously define the function of these genes and to resolve the epistatic relationships between them requires “loss-of-function” analysis. Antisense morpholino oligos, used to block the translation of specific target RNAs, should provide a rapid and efficient way to achieve “loss-of-function” analysis in Xenopus (Heasman et al., 2000; Heasman, 2002). The related diploid species Xenopus (Silurana) tropicalis may be better for antisense studies because, unlike the pseudotetraploid Xenopus laevis, it does not have redundant copies of each gene (Amaya et al., 1998; Nutt et al., 2001). In laevis, the different mRNA copies often have divergent sequences in the 5′ untranslated regions (UTRs) where the antisense oligos bind; for this reason, depleting one copy with a morpholino leaves the other untouched. For example, in laevis there are two copies of Sox17α (Sox17α1 and Sox17α2) (Hasegawa et al., 2002) and two copies of Gata6 (Jiang and Evans, 1996; Gove et al., 1997) both with divergent 5′ UTR sequences. When examining genes in laevis, one must target both copies of a gene to ensure efficient depletion by the morpholino oligo. In addition, the diploid tropicalis system has the promise of forward genetic analysis while retaining the advantages of easy embryologic manipulation available in laevis (Amaya et al., 1998; Khokha et al., 2003).
As a prelude to examining endoderm specification by “loss-of-function,” we have isolated full-length X. tropicalis homologues of the transcription factors involved. We found the endoderm genes are conserved in tropicalis and consistent with its diploid genome, there is only one copy of each gene, which should simplify functional analysis. By using in situ hybridization, we present a detailed expression analysis in tropicalis embryos, with an emphasis on gastrula stages when endoderm specification occurs. The expression patterns of the genes we examined were almost identical to those described in laevis, suggesting tropicalis experiments are likely to be relevant to laevis and vice versa. We found that detection of gene expression in the deep endoderm was more efficient in tropicalis than laevis. As a result, we observed some previously unreported expression domains. Thus, we have assembled most of the molecular components of the endoderm specification pathway in X. tropicalis.
RESULTS AND DISCUSSION
To isolate X. tropicalis orthologues of X. laevis endoderm genes, we generated two arrayed cDNA libraries from gastrula and neurula tropicalis embryos. We isolated X. tropicalis clones from these cDNA libraries, either by high stringency screening of arrayed filters with laevis probes or by searching the publicly available Sanger Centre tropicalis expressed sequence tag (EST) and National Center for Biotechnology Information (NCBI) databases. We only examined cDNAs and ESTs that appeared to encode full-length clones, that is, they had the putative start of translation based on the laevis sequence. Table 1 summarizes the ESTs and full-length clones that we have identified in this study.
To examine gene expression in tropicalis embryos, we used the same whole-mount in situ hybridization protocol we routinely use for endodermal genes in laevis (Sive et al., 2000, with modifications detailed in the Experimental Procedures section), producing excellent results (Fig. 1A). In fact, we observed superior penetration of probes in tropicalis embryos, allowing us to visualize gene expression in the deep endoderm tissue by sectioning after the whole-mount hybridization procedure (Fig. 1D). In contrast, with laevis, probes never penetrated into the deep endoderm tissue by using whole-mount in situ hybridization (compare Fig. 1B with D). In laevis, hybridization to paraffin-sectioned embryos is required to fully detect gene expression in the deep endoderm (Fig. 1C; Butler et al., 2001). The smaller size of the tropicalis embryo most likely accounts for the more efficient penetration. As a result, for experimental analysis of endoderm formation by in situ hybridization, tropicalis embryos are easier to work with than laevis.
The vegetally localized, maternal T-box transcription factor VegT initiates endoderm formation in Xenopus laevis (Zhang et al., 1998; Xanthos et al., 2001). Vempati and King had previously deposited the full-length sequence of tropicalis VegT (tVegT) in Genbank, which is 93% identical to laevis VegT at the amino acid level. To examine the expression of tVegT which, has not been reported, we searched the Sanger Centre tropicalis expressed sequence tag (EST) databases and identified a full-length tVegT clone (Table 1), which we used for in situ hybridizations. As expected, tVegT is vegetally localized in tropicalis oocytes and is expressed in the presumptive endoderm and mesoderm at early gastrula (Fig. 2). As in laevis, tVegT expression is down-regulated in the endoderm by the end of gastrulation but maintained in the lateral and ventral mesoderm of the blastopore lip (data not shown).
In Xenopus laevis, the “Mix-like” homeodomain transcription factors Mixer, Mix1/2, and Bix1-4 have been implicated in endoderm differentiation (Henry and Melton, 1998; Ecochard et al., 1998; Casey et al., 1999). By using all of the different laevis Mix-like sequences, we searched the EST database to identify cDNAs encoding full-length Mix-like homeodomain proteins in tropicalis. From approximately 100,000 gastrula and neurula stage ESTs, we only found three different tropicalis Mix-like sequences which appear to correspond to one tMixer gene, one tMix gene, and only one tBix gene. In total, we identified one tMixer EST, eight nearly identical tMix ESTs, and six nearly identical tBix ESTs, all of which correspond to full-length cDNAs (Table 1). In contrast, the pseudotetraploid laevis has a more complex set of genes, with one Mixer (Henry and Melton, 1998), two Mix genes (Rosa, 1989; Vize, 1996), and four Bix genes (Tada et al., 1998; Ecochard et al., 1998). The slight variation in the EST sequences within the tMix and tBix clusters (on average 98% identical) was in some cases due to sequencing errors. The minor sequence differences may also be due to allelic variation as the cDNA libraries used for the ESTs were generated from several different outbred animals (A.Z. and E.A. unpublished observations). By comparison, the different laevis copies of the genes, arising from its pseudotetraploidy, vary considerably more: X. laevis Mix1 and Mix2 are only 85% identical, whereas Bix1–4 are on average only 80% identical to each other at the nucleotide level.
We isolated one of each of the full-length tMixer, tMix, and tBix clones for further analysis (Table 1). At the amino acid level, tMixer was 78% identical to laevis Mixer, tMix was 74% and 75% identical to laevis Mix1 and Mix2, and tBix was 78%, 74%, 75%, and 73% identical to laevis Bix1–4, respectively. In situ hybridization with tMixer, tMix, and tBix probes to gastrula stage tropicalis embryos (Fig. 3) revealed the expression pattern predicted from laevis. tMixer is expressed throughout the deep and superficial endoderm of the gastrula, with expression strongest across the endoderm/mesoderm boundary. By comparison, tMix and tBix transcripts are predominantly localized to the mesoderm, with relatively lower levels in the deep endodermal mass at early gastrula. By late gastrula, their endodermal expression has declined to undetectable levels, but the mesodermal expression persists.
In summary, we have found that tropicalis has only three “Mix-like” genes: tMixer, tMix, and tBix; and their expression patterns are identical to their laevis counterparts. In contrast, laevis has at least seven different “Mix-like” genes, suggesting that functional analysis of this gene family will be more straightforward in tropicalis.
We searched the ESTs databases for clones encoding tropicalis members of the Gata4/5/6 family of zinc finger transcription factors, which have been implicated in vertebrate endoderm development. We found five nearly identical ESTs encoding full-length tGata6 (Table 1). We did not find any Gata5-like ESTs and only three tGata4 ESTs, all of which encoded partial clones. Focussing on one of the tGata6 clones, we found that tGata6 uses an alternative upstream ATG like mouse and human GATA6, which has not been reported in the laevis Gata6 cDNAs (Brewer et al., 1999). In laevis, a downstream, internal ATG is assumed to be used; therefore, the tGata6 has a longer open reading frame than the laevis Gata6 cDNAs. Over the common regions, tGata6 was 94% identical to laevis Gata6 at the amino acid level. By in situ hybridization, we found that tGata6 was expressed in the presumptive deep endoderm of the gastrula (Fig. 4A–C) but not in the superficial epithelial endoderm of the blastopore lip (Fig. 4B,C red arrows). At tail bud stages (Fig. 4D–G), tGata6 mRNA was localized to the deep midgut endoderm, with levels highest near the liver diverticulum and weaker in the head and posterior endoderm. The dorsal midgut endoderm underlying the axial mesoderm also expressed high levels of tGata6 (Fig. 4E,F). In addition, tGata6 mRNA was detected in the cardiac and lateral plate mesoderm at tail bud stages (Fig. 4E,F).
Although Gata6 mRNA has been detected in the laevis gastrula endoderm by reverse transcriptase-polymerase chain reaction (Jiang and Evans, 1996), the details of its early expression pattern are undocumented. The tGata6 expression pattern that we observe is reminiscent of that described for laevis Gata5 (Jiang and Evans, 1996; Weber et al., 2000). Gain-of-function experiments suggest that Gata4 and Gata5 are more important than Gata6 in laevis endoderm specification; however, we have not found any tropicalis Gata5 sequences in the gastrula and neurula ESTs. Perhaps tGata6, rather than tGata5, plays a more important role in tropicalis endoderm development. It will be important to test this, as well as to isolate full-length tGata4 to examine the functional importance of each of these genes in endoderm development.
To isolate tropicalis orthologues of the endoderm-specific, HMG-box transcription factor Sox17, we screened our arrayed cDNA libraries at high stringency with laevis Sox17α and Sox17β probes. We isolated three tSox17α full-length cDNAs and three full-length tSox17β cDNAs (Table 1). Sequence comparison showed that tSox17α was 93% and 89% identical at the amino acid level to laevis Sox17α1 and Sox17α2, respectively, whereas tSox17β was 80% identical to laevis Sox17β.
In situ hybridization showed that tSox17α and tSox17β are identically expressed throughout all the deep and superficial endoderm during gastrulation (Fig. 5A–C,F–H), which is consistent with their laevis counter parts (Hudson et al., 1997). During neurula and early tail bud stages, tSox17α expression declines significantly in the anterior endoderm, except in the endoderm behind the cement gland. Strong tSox17α expression persists in the endoderm posterior to the liver diverticulum until late tail bud stages (Fig. 5D). By late tail bud, tSox17α transcripts are undetectable in most of the endoderm but expression is maintained in the presumptive gall bladder region and the extreme posterior region, as has been observed in laevis (Zorn and Mason, 2001). In addition, we observe tSox17α transcripts in what appear to be endothelial cells in the head and along the flank of the embryo (Fig. 5E). In contrast to tSox17α, tSox17β expression declines rapidly after gastrulation. However, like tSox17α, a small patch of endoderm cells behind the cement gland maintains expression even into tail bud stages.
The dynamic expression of tSox17α/β that we observed between neurula and tail bud stages is undocumented in laevis. We were able to observe these expression domains in tropicalis because of the superior penetration of probes with the in situ procedure. When we checked laevis embryos more carefully, we observed an identical expression pattern (data not shown). The role of Sox17 in these tissues later in development will be interesting to examine.
Members of the FoxA forkhead family of hepatic nuclear factors (also known as HNF3 α/β/γ) have long been studied as key regulatory molecules and valuable markers of endoderm tissue. To identify tropicalis FoxA family members, we screened our arrayed cDNA library filters with a laevis FoxA2 (HNF3β) probe at moderate stringency. Although we did not isolate tropicalis FoxA2, we isolated several tropicalis clones of the related FoxA1 (HNF3α/FKH2). One of the tFoxA1 clones was full-length (Table 1) and 92% identical at the amino acid level to laevis FoxA1 (Bolce et al., 1993). In tail bud and larval stage laevis embryos, FoxA1 is expressed throughout the endoderm, in the notochord, floor plate, and in ciliated cells of the epidermis. In addition, FoxA1 mRNA has been detected in gastrula endoderm by Northern blot but was undetected by whole-mount in situ hybridization (Bolce et al., 1993; R. Harland, personal communication). The superior penetration of probes into tropicalis embryos allowed us to visualize tFoxA1 mRNA in the deep and superficial endoderm of the midgastrula (Fig. 6A,B). At later stages, tFoxA1 is expressed in a pattern identical to that described for laevis (Fig. 6C–F).
Hex and Cerberus
The initial stages of endodermal patterning in laevis occur simultaneously with endoderm specification as observed by the asymmetric expression of the genes Hex and Cerberus in the anterior endoderm (Zorn et al., 1999). To isolate tropicalis orthologues of Hex, we screened our arrayed cDNA library filters with a laevis Hex probe and isolated two putative full-length tHex clones (Table 1). However, one clone (tGas023e18) appeared to be an unspliced transcript or had a cloning artifact in the open reading frame. Sequence analysis of the remaining tHex cDNA indicated that it was 96% identical to laevis Hex and the amino acid level. We identified six full-length tropicalis Cerberus sequences by database searching (Table 1). One of these was completely sequenced and was 84% identical to laevis Cerberus and the amino acid level. In situ hybridization analysis with tHex and tCerberus probes demonstrated that the tropicalis genes were expressed in a pattern identical to their laevis counterparts (Bouwmeester et al., 1996; Newman et al., 1997; Zorn et al., 1999). tHex is expressed in the most dorsoanterior endomesoderm of the blastula and gastrula embryo (Fig. 7A–C; data not shown) and later is restricted to the forming liver diverticulum (Fig. 7D). tCerberus is also expressed in the anterior endomesoderm of the early gastrula, but its expression is also expanded laterally around the margin at the endoderm/mesoderm boundary. By late neurula, tCerberus mRNA is undetectable.
Recently, a conserved molecular pathway that regulates endoderm specification has been deduced (Stainier, 2002). Here, we have isolated most of those genes from the diploid Xenopus tropicalis, because we believe that “loss-of-function” analysis by antisense oligos and possibly forward genetics will work better in this species than in the pseudotetraploid Xenopus laevis. In addition, tropicalis retains all of the classic embryologic advantages enjoyed by Xenopus laevis (Amaya et al., 1998; Khokha et al., 2003). We show that the key components of the endoderm development pathway are conserved in tropicalis, but as expected, there are not multiple redundant copies of each gene as found in laevis. For example, there are only three Mix-like genes in tropicalis, whereas there are seven in laevis. This reduced complexity should make functional analyses significantly easier in tropicalis. We show that the standard laevis in situ hybridization protocol gives superior results for endodermal genes in tropicalis. The tropicalis endoderm genes that we isolated are expressed in a manner almost identical to their laevis counterparts, suggesting that they will have very similar functions. Results obtained in the tropicalis system, therefore, are likely to be relevant to the laevis system and vice versa. Finally, the high sequence conservation between laevis and tropicalis, coupled with the in-depth EST sequencing, allowed us to find most of the genes that we looked for in the databases. As a result, we were able to easily and quickly assemble an entire molecular pathway in tropicalis and this is likely to hold true for any genes one wishes to study.
X. tropicalis embryos were obtained by either natural mating or in vitro fertilizations, according to the protocol from the University of Virginia tropicalis Web site (http://faculty.virginia.edu/xtropicalis/). X. tropicalis husbandry details can also be found in Khokha et al. (2002). Embryos were staged according to the normal table of development for Xenopus laevis (Nieuwkoop and Faber, 1994).
cDNA Library Screening
The details of the tropicalis cDNA library construction, arraying, and EST project will be described in detail elsewhere and are available upon request. Briefly, polyA+ mRNA was isolated from gastrula and neurula tropicalis embryos. Oligo-dT–primed cDNA was directionally cloned into pCS107 and electroporated into Escherichia coli,resulting in the gastrula (tGas) and neurula (tNeu) cDNA libraries. Clones (55,000) from each of these libraries were robotically arrayed by the RZPD, and high-density filters were produced. These same libraries were given to the Sanger Centre, UK, for the tropicalis EST project. X. tropicalis clones were either isolated by screening the filters or by searching the publicly available EST database at NCBI and the Sanger Centre (http://www.sanger.ac.uk/Projects/X_tropicalis/). For filter screening, the full-length coding regions of X. laevis Sox17α (pSK-Sox17α, KpnI/SacI), Sox17β (pSK-Sox17β, EcoRI/NotI; Zorn et al., 1999), Hex (pSK-Hex, EcoRI/XhoI; Newman et al., 1997), Cerberus (pBS-Cerberus, EcoRI/XhoI; Bouwmeester et al., 1996), and Hnf3β/FoxA2 (pCS2-XFD3, EcoRI/NotI, a gift from Dr. Knoechel), were gel-isolated, radiolabeled by random priming, and hybridized at high stringency in Church's buffer. We only pursued cDNAs and ESTs that appeared to encode full-length clones: that is, they had the putative start of translation based on the laevis sequence. A summary of the accession numbers and clone identification numbers of all the ESTs and full-length cDNAs identified in this study are presented in Table 1. The clones described in this study, as well as any clones from the Sanger Centre tropicalis EST project, will be available from the HGMP through the MCR Geneservice, UK.
In Situ Hybridization
Table 1 lists the restriction enzymes used to linearize each clone to produce antisense digoxygenin-11-UTP RNA probes using T7 RNA polymerase (Ambion T7 MEGA script kits). Whole-mount in situ hybridizations were performed in home-made baskets by using the standard laevis protocol (Sive et al., 2000) with minor modifications to improve penetration of endodermal probes. Embryos were incubated with anti-digoxigenin antibody (1/2,000) overnight at 4°C (overnight with the antibody is essential to get efficient penetration into the deep endoderm). The following day, embryos were washed 12 × 30 min in Maleic acid buffer with a final wash at 4°C overnight. The following day, embryos were equilibrated in alkaline phosphatase buffer and the chromogenic reaction, with either nitro blue tetrazolium/5-bromo-4-chloro-3-indoxyl phosphate (NBT/BCIP) or BM purple allowed to proceed for several days at 15°C. After photography of the whole-mounts, embryos were embedded in 7% low-gelling temperature agarose and 30-micron sections were cut on a Vibratome and mounted in 90% glycerol/Tris buffered saline.2
Table 2. Conservation between Xenopus tropicalis and X. laevis Genesa
Nucleotide Identity in the Open Reading Frame (%)
Amino Acid Identity (%)
The values shown only reflect the regions common between tropicalis and laevis.
tropicalis gata6 appears to use an alternative upstream ATG like mouse and human, the reported laevis cDNAs do not contain these sequences; therefore, it is assumed that the downstream ATG is used (Brewer et al., 1999).
We thank Jen Ashurst, Amanda McMurray, Liz Huckle, Ruth Taylor, Jane Rogers, and Richard Durbin at the Wellcome Trust Sanger Centre, UK, for enthusiastically undertaking the tropicalis EST project. This work was supported by grants from the Wellcome Trust UK (J.S., E.A., N.P., A.Z.) and the NIH (A.Z.).