A conserved retinoic acid responsive element in the murine Hoxb-1 gene is required for expression in the developing gut

 
• RA

  retinoic acid

 
• Hox

  homeobox gene

 
• RAIDR

  retinoic acid RARE direct repeat₅ enhancer

 
• RAR

  retinoic acid receptor

 
• RXR

  retinoid X receptor

Retinoic acid (RA) is required during normal embryonic development in vertebrates. Both a deficiency and an excess of RA lead to a variety of developmental defects in organs including the brain, heart, pituitary, thyroid and limbs (Gudas, 1994; Hofmann and Eichele, 1994). The potent teratogenic effects of exogenous RA during development have implicated endogenous retinoids in the patterning of the anteroposterior axis of the embryo (Kessel and Gruss, 1991; Krumlauf, 1994). In vertebrates, 39 genes related to those of the Drosophila homeotic gene complex (HOM-C) have been identified. In mammals, many of the Hox genes are organized in four clusters, HoxA, HoxB, HoxC and HoxD. The conservation of gene organization, expression and function of the Hox genes is well documented in all vertebrate species studied (reviewed in Krumlauf, 1994; McGinnis and Krumlauf, 1992). Hox genes encode transcription factors containing the homeobox domain. Experiments involving the targeted disruption or misexpression of Hox genes suggest that Hox genes specify regional identity along the anteroposterior body axis (Krumlauf, 1994; McGinnis and Krumlauf, 1992). Hox genes are turned on sequentially from the hindbrain to the posterior end of the embryo in a nested fashion, colinear with their positions in the clusters. Hox genes at the 3′ ends of the chromosome clusters are expressed earlier and in more anterior regions of the embryo, while Hox genes at the more 5′ ends of the clusters are expressed at later times in embryogenesis and in more posterior regions of the embryo (Duboule and Morata, 1994; Krumlauf, 1994; Lewis, 1978). Alterations in normal expression patterns lead to homeotic transformations and malformations (McGinnis and Krumlauf, 1992).

The murine Hoxb-1 gene is located at the 3′ end of the Hoxb gene cluster on mouse chromosome 11 (Frohman et al., 1990; Wilkinson et al., 1989). The Hoxb-1 gene shows two phases of expression in the developing embryo. Hoxb-1 mRNA is first detectable at 7.5 d.p.c. in the posterior half of the embryo, in the neuroectoderm, in the primitive streak and in newly formed mesoderm rostral to the node. By the early somite stage at 8.5 d.p.c., Hoxb-1 expression becomes divided into two domains: the prospective rhombomere 4 in the hindbrain and the posterior half of the embryo, which includes the posterior neural tube, paraxial (somite) mesoderm and gut (Frohman et al., 1990; Murphy and Hill, 1991; Wilkinson et al., 1989). The anterior expression of Hoxb-1 induced by exogenous RA transforms rhombomere 2 to a rhombomere 4 identity (Marshall et al., 1992). Therefore, in the developing hindbrain, the Hoxb-1 gene, like its Drosophila counterpart, labial, is involved in specifying the identities of segments. Consistent with this, a Hoxb-1 null mutation results in neonatal lethality with alterations in rhombomeric identity and failure to form a functional facial (VIIth) nerve (Goddard et al., 1996; Studer et al., 1996).

Several lines of evidence implicate RA as one of the key signaling molecules in determining the anteroposterior patterning through the regulation of Hox gene expression. First, during the RA-induced differentiation of teratocarcinoma cells, most Hox genes are activated by RA sequentially, colinear with their positions in the clusters (Simeone et al., 1990, 1991). Second, in early development 3′ Hox genes such as Hoxa-1 and Hoxb-1 are expressed in the neural tube with distinct anterior boundaries in the hindbrain. In embryos treated with exogenous RA, an anterior expansion of the expression of these genes is observed and this expansion is associated with the homeotic transformation of segments in the hindbrain (Cho and De Robertis, 1990; Conlon and Rossant, 1992; Holder and Hill, 1991; Kessel, 1992; Kessel and Gruss, 1991; Marshall et al., 1992; Morriss-Kay et al., 1991; Papalopulu et al., 1991; Sive and Cheng, 1991). Third, RA has been detected in the Xenopus dorsal lip, in Hensen,s node of the chick and in the primitive streak in the mouse, an area responsible for organizing the vertebrate embryo and likely the site of Hox induction (Chen et al., 1992; Thaller and Eichele, 1987; Wagner et al., 1992). Thus, RA and potentially other bioactive retinoids are likely to be the natural morphogens. Fourth, the discovery of retinoic acid-response elements (RAREs) in the Hoxa-1 (Dupé et al., 1997; Frasch et al., 1995; Langston and Gudas, 1992; Langston et al., 1997), Hoxb-1 enhancers (Langston et al., 1997; Marshall et al., 1994; Studer et al., 1994), Hoxa-4 promoter (Doerksen et al., 1996) and Hoxd-4 promoter (Popperl and Featherstone, 1993) suggests that transcriptional activation of at least some Hox genes by RA is via a direct mechanism involving the RARs (Conlon, 1995; Gudas, 1994).

Most of the retinoic acid responses in the embryo are mediated through retinoic acid receptor proteins (RARs and RXRs) that are members of the steroid/thyroid nuclear receptor superfamily (reviewed in Leid et al., 1992; Mangelsdorf et al., 1994). RARs form heterodimers with RXRs and bind cooperatively to cis-acting RAREs on many of the responsive genes to activate transcription. Several different RARE-containing enhancers that contribute to the expression of the human and murine Hoxb-1 genes have been identified (Langston et al., 1997; Marshall et al., 1994; Ogura and Evans, 1995a,b; Studer et al., 1994). Marshall et al. (1994) described a functional DR₂-type RARE, conserved in the mouse, chick and pufferfish, which was able to mediate the RA responsiveness of the early expression of the Hoxb-1 gene in transgenic animals. We identified an RA-inducible enhancer (RAIDR₅) containing a DR₅-type RARE as part of a DNAse I hypersensitive site located approximately 6.5 kb 3′ of the murine Hoxb-1-coding region. This DR₅ RARE is required for RA-induction of Hoxb-1 in teratocarcinoma cells (Langston et al., 1997). In addition to this DR₅ RARE, the RAIDR₅ enhancer contains two other blocks of highly conserved sequences called CE1 and CE2. We have shown that the CE2 element in the RAIDR₅ enhancer of the Hoxa-1 gene, a paralog of Hoxb-1, contributes to Hoxa-1 expression in somites and adjacent mesenchymal tissue during development (Thompson et al., 1998).

There is accumulating evidence that suggests that Hox genes may be involved in the regional specification of the digestive tract (gut) along the anteroposterior axis (Roberts et al., 1995; Yokouchi et al., 1995). The digestive tract is composed of various organs including the esophagus, stomach, liver, pancreas, intestine and colon. These organs originate from a simple tube of endoderm, surrounded by visceral mesoderm, via sequential induction between these two germ layers (Haffen et al., 1987). The region-specific differentiation of gut epithelium is dependent on signals from the surrounding visceral mesoderm (Haffen et al., 1983). Conversely, primitive endoderm signals the adjacent mesenchymal tissues to undergo gut-specific mesodermal differentiation (Haffen et al., 1983; Kedinger et al., 1986). For gut morphogenesis, one endoderm invagination from the anterior end, the anterior intestinal portal (AIP), extends posteriorly to form the foregut at ∼7.5 d.p.c., and another endoderm invagination from the posterior end, the caudal intestinal portal (CIP), extends anteriorly to form the hindgut at around 8.5 d.p.c. The midgut is then derived by fusion of the foregut and hindgut at the midline around 9 d.p.c. (Rugh, 1996). Simple ducts subsequently begin to protrude from the primary gut tube to form structures such as the lung buds, hepatic diverticulum and pancreas evaginations. At later stages, the mesodermal layer of the tube differentiates into muscle and connective tissue, and the endoderm layer differentiates into epithelium with specific histological and biochemical features. For example, the foregut gives rise to organs such as the esophagus, lung, stomach, liver and pancreas. The molecular mechanisms underlying the generation of positional cues to specify duct budding and regional specification of the digestive tract in vertebrates are still unknown. Transcripts of many Hox genes are expressed in a restricted region in the developing gut (reviewed in Roberts et al., 1995; Yokouchi et al., 1995). Moreover, targeted disruptions of some 5′ Hox genes result in malformations of the posterior structures in the digestive tract such as the anus sphincter (Kondo et al., 1996). These findings suggest that Hox genes play significant roles in the anteroposterior patterning of gut.

In the studies presented here, we examined the role of the DR₅ RARE in the Hoxb-1 3′ RAIDR₅ enhancer in vivo through analysis of transgenic mice. We show that a 15 kb fragment of murine Hoxb-1 genomic DNA, which includes this DR₅ RARE, directs the expression of a lacZ reporter gene in a temporal and spatial pattern that reflects that of the endogenous Hoxb-1 gene. We also demonstrate that the 3′ DR₅ RARE is required for the regulation of Hoxb-1 region-specific expression in the gut and for the RA-induced anteriorization of Hoxb-1 expression in the gut. This suggests that the RA signaling pathway is involved in the specification of the regions of Hox gene expression and consequently in the anteroposterior patterning of gut.

The genomic DNA fragments containing the Hoxb-1 locus were isolated from a lambda FIXII mouse genomic library. A 15 kb NotI fragment containing the 3′ DR₅ RARE was subcloned into pBluescript KS, generating pHoxb-1. lacZ was inserted into the Hoxb-1-coding sequence as follows. The E. coli lacZ gene from pMC1871 (Pharmacia) was excised by digestion with BamHI and ligated to a BamHI-EagI adaptor; it was purified and ligated to the partially digested pHoxb-1 in the EagI site in exon I, creating an in-frame fusion at the 34th amino acid of Hoxb-1 (pHoxb-1RAIDR₅WT/lacZ). A stop codon from pMC1871, located downstream of the lacZ open reading frame, is included.

The DR₅ RARE mutation was made by PCR-mediated mutagenesis as described (Prelich, 1993). The RARE was mutated by PCR amplification of the subcloned enhancer fragment with the following oligonucleotides, which were paired with primers to outer flanking plasmid sequences (mutated nucleotides are in lower case letters, underlined sequence is the DR₅ RARE):

RARE top strand: 5′-tcT agc TAG AGA aTT acG CTC TGA AAT GCT TGC AGC-3′

RARE bottom strand: 5′-AGC ATT TCA GAG Cgc AAt TCT CTA gct Aga AGG AAG GAA AGG G-3′.

Sequences of the 5′ and 3′ outer flanking primers are: 5′-TGG TGG CTC ACA ATC TC-3′ and 5′-CGC TCT AGA ACT AGT GGA TC-3′ (SK primer of pBluescript vector), respectively.

Following amplification of both top and bottom strands, the PCR products were mixed and another round of amplification was carried out using only the outer flanking primers. The resulting 1 kb PCR product containing the RARE mutation was cut with SphI and BglII, purified, and ligated to the compatible vector, which was prepared by digestion of a 2 kb SphI-KpnI subclone of Hoxb-1 3′ flanking sequence in pBluescript with SphI and BglII; this created pKS-DR₅m. Then, an approximately 2 kb Hoxb-1 3′ sequence containing the DR₅ RARE mutations was cut from the pKS-DR₅m with SphI and KpnI and ligated to the 3′ portion and the 5′ portion of the Hoxb-1 chimeric gene as follows. The 3′ portion was derived from pHoxb-1RAIDR₅WT/lacZ by digestion with ClaI and KpnI. The 5′ portion of the chimeric gene was prepared by digestion of pHoxb-1RAIDR₅WT/lacZ by ClaI and SphI. This three fragment ligation generated pHoxb-1RAIDR₅mu/lacZ, which is identical to pHoxb-1RAIDR₅WT/lacZ except for the DR₅ RARE mutation. DNA sequencing analysis confirmed that pHoxb-1RAIDR₅mu/lacZ contained the desired RARE mutations but no other sequence changes.

The pHoxb1RAIDR₅ 3-del/lacZ construct was prepared as follows. The 5′ portion ClaI-SphI fragment containing the Hoxb-1 3′ proximal DR₂ RARE, but not the RAIDR₅ RARE, was ligated to a SphI-KpnI adaptor, and was then ligated to the Hoxb-1 5′ portion, the ClaI-KpnI fragment derived from the pHoxb-1RAIDR₅WT/lacZ. This 3′ deletion construct is identical to pHoxb-1RAIDR₅WT/lacZ except for a ∼2 kb deletion of Hoxb-1 3′-flanking sequence.

F9 teratocarcinoma cells were cultured in DMEM with 10% calf serum as previously described (Langston et al., 1997). F9 cells were transfected with Hoxb-1/lacZ plasmids by calcium phosphate coprecipitation. β-galactosidase activities were normalized to reporter activity of a plasmid containing the β-actin/CAT. Transient transfections, CAT assays and β-galactosidase assays were performed as described previously (Langston et al., 1997). Quantitation of CAT assays was carried out using a PhosphorImager (Molecular Dynamics).

DNA was prepared as a linearized NotI fragment with vector sequence removed and microinjected into 1-cell mouse embryos (C57B/6×CBA/J F₁) at the Rockefeller University Transgenic Mouse Facility. To identify the transgenic offspring, tail DNA was digested with EcoRI, Southern blotted to Hybond N nylon membrane (Amersham) and probed with a random-primed [³²P]dCTP-labeled lacZ fragment as described (Sambrook et al., 1989). Production of transgenic embryos was performed by mating transgenic males with non-transgenic females (CBA57/J6) in the evening and monitoring the plugs the following morning; a plug found the following morning is considered 0.5 d.p.c. Embryos were removed at the appropriate gestational ages and stained for β-galactosidase activity. The expression patterns of the Hoxb-1RAIDR₅WT/lacZ, Hoxb-1RAIDR₅mu/lacZ, and Hoxb-1RAIDR₅ 3′del/lacZ transgenes were confirmed at all stages of development in at least two independent founder lines of transgenic mice. The two transgenic lines of Hoxb-1RAIDR₅WT/lacZ are Tg11 and Tg21; the lines of Hoxb-1RAIDR₅mu/lacZ are Tg49, Tg50, Tg63 and Tg75, and the lines of the Hoxb-1 RAIDR₅ 3′-del/lacZ are Tg103 and Tg109.

Embryos were fixed as previously described (Conlon and Rossant, 1992; Means and Gudas, 1997) in 1% formaldehyde, 0.2% glutaraldehyde, 2 mM MgCl₂, 5 mM EDTA, and 0.02% NP-40 in PBS, for 30-90 minutes at 4°C. The embryos were then washed three times in PBS plus 0.02% NP-40, and stained in 1 mg/ml X-gal, 5 mM K₃Fe(CN)₆, 5 mM K₄Fe(CN)₆, 2 mM MgCl₂, 0.01% sodium deoxycholate and 0.02% NP-40 in PBS, at room temperature and in the dark. Stained embryos were washed with several changes of PBS + 0.02% NP-40. Photographs were taken on a Leica dissecting microscope.

For sectioning, the embryos were dehydrated and paraffin-embedded as previously described (Means and Gudas, 1997). Sections were cut (7 μm) and mounted on glass slides. Slides were deparaffined and counterstained with eosin as described previously (Means and Gudas, 1997).

Hoxb-1 plasmid was kindly provided by Dr Joe Grippo. The plasmid DNA contains a 435 bp EcoRI-HindIII fragment of Hoxb-1 cDNA. Digoxigenin-labeled antisense probes were synthesized from linearized plasmid with HindIII digestion and transcribed in vitro using T7 RNA polymerase as described by the manufacturer (Boehringer Mannheim Biochemicals, St Louis, MO). The Hoxb-1 plasmid was linearized with EcoRI and transcribed with SP6 RNA polymerase in order to obtain the sense probe.

Whole-mount in situ hybridization was performed as previously described (Conlon and Rossant, 1992) with the following changes. Embryos were treated with 10 μg/ml proteinase K for 8 minutes at room temperature and prehybridized for at least 1 hour at 68°C in hybridization buffer (50% formamide, 0.75 M NaCl, 1 mM EDTA, 50 μg/ml tRNA, 0.05% heparin, 1% sodium dodecyl sulfate). The hybridization buffer was replaced and single-stranded RNA probes labeled with digoxigenin were added to 1 μg/ml; embryos were hybridized overnight at 68°C.

Pregnant females were treated with RA at various times before embryo dissection. RA was administrated at a dose of 20 mg per kg of maternal body weight by oral gavage (Conlon and Rossant, 1992). A stock solution of 25 mg/ml of all-trans-RA (Sigma) in dimethylsulfoxide was dissolved just before use in corn oil so that a 0.2 ml dose of oil contained the requisite amount of RA. Control mice received corn oil alone. Embryos were isolated 12-16 hours later and stained for lacZ expression for 12-18 hours unless mentioned specifically in the text.

Previously we identified a single RA-inducible DNAse I hypersensitive site located 6.5 kb 3′ of the murine Hoxb-1-coding sequence (Langston et al., 1997). Further analysis demonstrated that an enhancer, called RAIDR₅, at this site regulates the RA responsiveness of the Hoxb-1 gene in F9 teratocarcinoma and embryonic stem cells. This newly identified DR₅ RARE is different in location and sequence from the RA-regulated 3′ Hoxb-1 enhancers previously described (Marshall et al., 1994; Ogura and Evans, 1995a). Sequence comparison also shows a conserved DR₅ RARE in the chick Hoxb-1 3′ enhancer, and these Hoxb-1 DR₅ RAREs are identical to the DR₅ RAREs in the human and murine Hoxa-1 3′ RAIDR₅ enhancers (Langston and Gudas, 1992; Langston et al., 1997).

To elucidate the role of the Hoxb-1 3′ DR₅ RARE in animals, we constructed a wild-type Hoxb-1/lacZ transgene that includes 6 kb of 5′ and 7.5 kb of 3′ flanking sequences in the murine Hoxb-1 gene; this transgene includes the distal 3′ DR₅-type RARE (Fig. 1, RAIDR₅ WT). The lacZ gene was inserted into the Hoxb-1-coding sequence so that an in-frame fusion was generated that contained the first 34 amino acids of Hoxb-1 fused to the β-galactosidase-coding sequence. We made a second construct with point mutations in the DR₅ RARE (Fig. 1, RAIDR₅mu), and a third construct with the 3′-most 2 kb fragment containing the RAIDR₅ enhancer deleted (Fig. 1, 3′-deletion). The same DR₅ RARE point mutations in the conserved Hoxa-1 gene result in a loss of RAR binding in gel shift assays (Langston and Gudas, 1992).

To test the RA inducibility of the reporter constructs, these Hoxb-1/lacZ constructs were transfected into F9 cells. Transiently transfected F9 cells were treated with RA or left untreated. The pHoxb-1RAIDR₅WT/lacZ transgene was able to confer RA-responsiveness (∼5-fold) as measured by β-galactosidase activity (Fig. 2). In contrast, both the RAIDR₅mu and the RAIDR₅ 3′ deletion constructs showed only minimal RA induction. These results are consistent with our previous data that a ∼300 bp Hoxb-1 DNA fragment containing the DR₅ RARE is sufficient for RA inducibility in F9 cells (Langston et al., 1997) and they show that the 3′ DR₅ RARE is required for the RA-induced transcriptional activation of the Hoxb-1 gene in F9 cells.

To determine the role of the Hoxb-1 3′ DR₅ RARE in vivo, transgenic mice carrying the various Hoxb-1/lacZ transgenes were generated. Transgene expression patterns were analyzed at all stages of development during which the endogenous Hoxb-1 gene is expressed in the embryo.

The normal pattern of expression of Hoxb-1 mRNA in the gastrulation stage of development has been well defined (Frohman et al., 1990; Murphy and Hill, 1991; Wilkinson et al., 1989). Hoxb-1 mRNA is expressed in embryonic ectoderm and mesoderm in the posterior half of the embryo by late primitive streak stages. By the headfold stage, Hoxb-1 mRNA expression extends slightly anteriorly to the primitive streak in the neuroepithelium and the underlying mesoderm. At about the time of formation of the first somite (8.5 d.p.c.), the most anterior neural staining at the preotic sulcus becomes separated from the staining at the posterior of the embryo and the mesoderm expression retreats posteriorly to about the level of the primitive streak. The focus of this study is on Hoxb-1 transgene expression between 8 and 12 d.p.c.

The expression patterns of the Hoxb-1RAIDR₅WT/lacZ transgene (designated as WT) and the Hoxb-1RAIDR₅mu/lacZ transgene (designated as DR₅mu) at 8-8.5 d.p.c. are shown in Fig. 3. At 8.0 d.p.c., the expression of the WT transgene extends from the posterior of the embryo up to a well-defined anterior boundary in the headfolds. Expression of the WT transgene is restricted to the neural tube, the lateral mesoderm and the primitive streak (Fig. 3A). A similar pattern is seen with the DR₅mu transgene. The anterior boundary of expression retreats toward the posterior (Fig. 3C). In addition, expression of the WT transgene is observed in the foregut pocket (Fig. 3C).

A different, distinct pattern of expression is seen with the RAIDR₅mu transgene; β-gal staining is observed in the neural epithelium and the lateral mesoderm but not in the developing gut epithelium (Fig. 3D). These data suggest that the DR₅ RARE is required for initiating Hoxb-1 expression in the gut epithelium but is not required for setting the correct anterior boundary of Hoxb-1 expression in the neural tube and mesoderm.

By 9.5 d.p.c., the pattern of expression of the Hoxb-1RAIDR₅WT/lacZ transgene is similar to that at 8.5 d.p.c., but the intensity of the β-gal staining is higher. At this stage, expression in the neural epithelium is in a discrete band in rhombomere 4, which is one segment anterior to the otic vesicle and a domain in the posterior neural tube (Fig. 4A,D). WT transgene expression is also seen in the notochord and paraxial (somitic) mesoderm. It is worth noting that, in rhombomere 4, there is staining of the neural crest cells that have migrated from rhombomere 4 (Fig. 4D). The WT transgene expression is also detected in the gut region at and below the level of the heart and in the paraxial mesoderm in the posterior region of the embryo.

In embryo sections, expression of the WT transgene in the foregut epithelium can be seen, and expression is also observed in the gut-associated mesoderm and some ectoderm in close proximity (Fig. 5A-C). At ∼9.5 d.p.c., the gut is a simple tube, wider at both ends than in the middle. The foregut includes primarily the pharynx, pharyngeal pouches and esophageal region at this stage. Expression in the gut epithelium, gut-associated mesoderm and the adjacent ectoderm has an apparent anteroposterior restriction in the foregut from the third pharyngeal arch throughout the pharynx to the esophageal the DR₅mu transgene expression in the neuroepithelium and mesoderm is similar to that observed for the WT transgene, i.e. just up to the posterior edge of the preotic sulcus (Fig. 3B).

At 8.5 d.p.c. the most anterior neural staining of the Hoxb-1RAIDR₅WT/lacZ transgene at the preotic sulcus is separated from the staining in the posterior of the embryo. Meanwhile, the mesodermal region, with the most intense staining found in the epithelial cells in the lateral walls of the most posterior part of the foregut (Figs 4A, 5B). There appears to be a coordinated expression of the Hoxb-1 RAIDR₅WT/lacZ transgene in the endoderm, mesoderm and ectodermal tissue layers in the gut at this time. At 9.5 d.p.c., expression of the Hoxb-1 RAIDR₅mu/lacZ transgene is observed in the neural tube, notochord and somitic mesoderm at a level similar to that observed in animals carrying the Hoxb-1 RAIDR₅WT/lacZ transgene (Fig. 4B,E). The most striking change, as compared to the embryos with the WT transgene, is the lack of DR₅mu transgene expression in the gut region. Only very weak β-gal staining is seen in the foregut epithelium and in the most posterior region of the foregut of the DR₅mu transgenic animals (Fig. 4B). No staining is observed in the gut-associated mesoderm and the adjacent ectoderm epithelium (Fig. 5E-H). In the four independent lines of DR₅mu transgenic animals analyzed, three lines exhibited essentially no β-gal staining in the foregut epithelium (Fig. 4B). One DR₅mu transgenic line showed a low level of transgene expression in the gut epithelium, a level that was significantly lower than that observed in the embryos with the Hoxb-1RAIDR₅WT/lacZ transgene (data not shown).

The expression pattern of the Hoxb-1RAIDR₅ 3′-del/lacZ transgene (designated as 3′-del) is similar to the pattern of the Hoxb-1RAIDR₅mu/lacZ transgene (Fig. 4C,F). Expression of the 3′-del transgene was found in the neural tube, notochord and somitic mesoderm at a level similar to that observed in animals carrying the Hoxb-1RAIDR₅WT/lacZ transgene. However, staining in the foregut of the animals carrying the 3′-del construct is significantly reduced as compared to the Hoxb-1RAIDR₅WT/lacZ transgene (Fig. 4C versus A). We conclude that the DR₅ RARE is required for region-specific expression in the gut, including the gut epithelium, and in cells derived from all three germ layers. This is consistent with the notion that Hoxb-1 expression in different germ layers in the gut is coordinately regulated by bioactive retinoids such as RA.

In 10.5-11.5 d.p.c. embryos, expression of the Hoxb-1RAIDR₅WT/lacZ transgene is reduced relative to earlier stages, but expression is still observed in all of the same regions seen in earlier embryos, i.e. in rhombomere 4, the notochord, the most posterior somitic mesoderm, and in a very restricted region of the foregut epithelium and gut-associated mesoderm (data not shown). After 11.5 d.p.c., lacZ activity is significantly reduced, but is still detectable in small, diffuse groups of cells in the hindbrain, paraxial mesoderm, posterior neural tube, notochord and root of the forelimb of the embryos carrying either the Hoxb-1RAIDR₅WT/lacZ or the Hoxb-1RAIDR₅mu/lacZ transgenes (data not shown).

Taken together, our data show that a 15 kb DNA fragment of the Hoxb-1 gene is sufficient to recapitulate the expression of the endogenous Hoxb-1 gene both in the early and the late stages of development. We demonstrate that the conserved 3′ DR₅ RARE in the murine Hoxb-1 RAIDR₅ is essential for Hoxb-1 region-specific expression in the developing gut (Table 1).

The RA-induced expression pattern of the endogenous Hoxb-1 mRNA has been reported previously (Conlon and Rossant, 1992; Marshall et al., 1992; Morriss-Kay et al., 1991; Wood et al., 1994). RA treatment resulted in extension of the endogenous Hoxb-1 mRNA expression in the anterior direction relative to the expression seen in untreated embryos both prior to and after the formation of the headfolds. Ectopic anterior expression of Hoxb-1 was induced in both neuroectodermal and mesodermal cell layers of headfold embryos. The timing of RA treatment was critical for its effects on Hoxb-1 gene expression and for its teratogenic effects (Conlon and Rossant, 1992; Morriss-Kay et al., 1991; Popperl et al., 1995). Expression of Hoxb-1 in rhombomere 4 became largely refractory to the RA treatment if the embryos were treated with RA after 8 d.p.c. (Conlon, 1995; Wood et al., 1994).

We performed whole-mount in situ hybridization to show the expression pattern of the endogenous Hoxb-1 mRNA after RA treatment (Fig. 6). In 8.5 d.p.c. embryos, the endogenous Hoxb-1 mRNA was expressed in a pattern identical to that described above by other laboratories. For example, just 4 hours of RA treatment at 8.5 d.p.c. resulted in a dramatic anterior shift of Hoxb-1 mRNA expression in the foregut (Fig. 6A,B), as shown previously (Conlon and Rossant, 1992), indicating that the RA effect on Hoxb-1 gut expression is relatively rapid. By ∼9.5 d.p.c., exogenous RA can still cause anteriorization of endogenous Hoxb-1 mRNA expression in the gut (Fig. 6D,E).

To determine if the RAIDR₅ RARE was responsive to RA in vivo, we treated pregnant mothers carrying the Hoxb-1RAIDR₅WT/lacZ or Hoxb-1RAIDR₅mu/lacZ transgene with RA for 12-16 hours. The expression patterns of the WT and DR₅mu transgenes at 8 d.p.c. in untreated embryos and in embryos exposed to RA at ∼7.5 d.p.c. for 12 hours and harvested at 8 d.p.c. are shown in Fig. 7. RA treatment causes a dramatic anterior expansion of the WT transgene in the neural tube, somites and lateral plate mesoderm, such that even the most anterior structures such as the midbrain and forebrain express the WT transgene (Fig. 7B). There is no significant difference in expression between the WT and the DR₅mu transgenes in response to RA at this stage except that no activity of lacZ in the foregut pocket was observed in the DR₅mu transgenic embryos (Fig. 7B,C).

Treatment with RA at the early somite stage (8.5-8.75 d.p.c.) for 12-16 hours prior to embryo harvesting at ∼9.5 d.p.c. did not result in any apparent teratogenic effects on the central nervous system or the overall body structure. However, the expression pattern of the Hoxb-1RAIDR₅WT/lacZ transgene was altered only in the gut of the embryos following RA treatment (Fig. 7D versus E). Exogenous RA treatment dramatically anteriorized the expression of the WT transgene in the foregut, so that WT transgene expression was detected extensively in the branchial arches, along the anteroposterior axis to the first branchial arch (Fig. 7E,I; note difference in locations of transgene expression relative to rhombomere 4 in Fig. 7D versus E). The expression of lacZ was most abundant in the pharyngeal pouch, which is an extension of the pharyngeal foregut endoderm. Upon careful examination of the transgene staining in the sections from the RA-treated WT transgenic embryos, lacZ expression was detected in all three somitic mesoderm and other expression domains in the embryos by 9.5 d.p.c.; the only exception was the minor anterior extension in r4 (Fig. 7F,J). No ectopic transgene expression was observed in the foregut (Fig. 8C,D). No β-gal staining was seen in the gut-associated mesoderm and the adjacent ectoderm (Figs 7F,J, 8C,D); only the notochord expressed the transgene (Fig. 8D). No anterior extension of the transgene expression in the foregut was detected after 12-16 hours RA treatment (Fig. 7F,J). Therefore, in the early somite stage, the DR₅mu transgene does not respond to RA. Similarly, embryos carrying the RAIDR₅3′-del/lacZ transgene did not respond to exogenous RA (Fig. 7G,K). We conclude that the germ layers in the foregut region (Fig. 8A,B).

In contrast, when mice carrying the Hoxb-1RAIDR₅mu/lacZ transgene were treated with exogenous RA at the early somite stage (∼8.75 d.p.c.) for 12-16 hours, there were no detectable changes in the transgene expression in the foregut, RAIDR₅ RARE is essential for anterior extension of Hoxb-1 expression in the foregut in response to exogenous RA.

In 10-11.5 d.p.c. embryos, the lacZ expression patterns and the RA responses of both the Hoxb-1RAIDR₅WT/lacZ and Hoxb-1RAIDR₅mu/lacZ transgenic embryos are similar to those in earlier stage embryos. Embryos carrying the WT transgene still show restricted transgene expression in the gut and RA causes some detectable anterior expansion of the gut expression, whereas in the DR₅mu transgenic embryos no expression of lacZ is detectable in the gut. There is no ectopic transgene expression in the DR₅mu transgenic embryos after RA treatment (data not shown).

Collectively, our results show that the 3′ DR₅ RARE directly mediates part of the Hoxb-1 RA response and that it is an essential regulatory element for the correct Hoxb-1 expression pattern in the gut epithelium and the associated mesoderm and ectoderm during embryogenesis. Our data also strongly suggest that endogenous retinoids play a significant role in the regulation of Hoxb-1 expression in the foregut. Moreover, the observations that the Hoxb-1 gene is expressed in the developing gut and that this expression is regulated by RA through a DR₅ RARE suggest a role for Hoxb-1 in the anteroposterior patterning of the gut.

In the present studies, we have shown that the Hoxb-1 3′ DR₅ RARE is required for the expression of the Hoxb-1/lacZ transgene in the gut. Hoxb-1 expression in gut endoderm, gut-associated mesoderm and the adjacent ectoderm appears to be regulated in a coordinated manner corresponding to the anteroposterior axis at this stage (8.5-10.5 d.p.c.).

We demonstrate that Hoxb-1 expression in the gut responds to exogenous RA, resulting in anterior expansion of Hoxb-1 expression in the gut similar to the well-documented RA response in the neural tube at earlier stages. We have shown that this RA response of the lacZ reporter gene in the gut region is dependent on the DR₅ RARE in the Hoxb-1 3′ RAIDR₅ enhancer (Fig. 7). Therefore, the DR₅ RARE 3′ of the Hoxb-1 gene is the major determinant of the anteroposterior limit of Hoxb-1 expression in the three germ layers in the gut. However, the RAIDR₅ RARE is dispensable for Hoxb-1 neural tube expression and for setting the anterior expression boundary in the neuroepithelium.

These data are consistent with a model that Hoxb-1 expression in the gut is defined by a developmental field that is independent of the early neural tube expression. By analogy to the previously defined role of RA in the hindbrain, it is conceivable that a retinoid gradient in the early somite stages is required for the coordinated expression of the Hoxb-1 gene and subsequent pattern formation in the gut. The requirement for a DR₅ RARE for Hoxb-1 expression in the gut provides another illustration that retinoids such as RA are able to regulate the expression of Hox genes at the 3′ ends of the chromosomal clusters directly, thus suggesting that endogenous retinoids may be required for the appropriate spatial and temporal regulation of Hoxb-1 expression in the gut. We think that this regulation is an important component of the normal control of the Hoxb-1 gene, because the RAIDR₅ enhancer has been conserved from chick to mouse.

During mouse embryo gastrulation, the foregut forms as early as 7.5 d.p.c. when differential growth produces a ventral invagination of the anterior definitive endoderm (Rugh, 1996). At 8.5 d.p.c., a second endodermal invagination starts from the posterior end to form the hindgut. The midgut is derived from both foregut and hindgut primordia and forms from the fusion of the primitive gut tubes. The foregut gives rise to organs such as the esophagus, lung, stomach, liver and pancreas. The midgut develops into the small intestine. The hindgut develops into the large intestine and forms part of the cloaca, the common gut-urogenital opening. The molecular mechanisms underlying the positional cues to specify regional differentiation of the digestive tract are poorly understood. Organ-specific differentiation of the endoderm-derived epithelium is induced by signals from the surrounding visceral mesoderm and partially depends on the reactive potential of the endoderm to the inductive signal (Haffen et al., 1983; Kedinger et al., 1986).

Transcripts of some Hox genes have been found in a restricted region in the developing digestive tract (Conlon and Rossant, 1992; Dollé et al., 1991; Gaunt et al., 1989, 1990; Holland and Hogan, 1988; Toth et al., 1987). Recently, Hox genes from paralog groups 9-13, located at the 5′ end of the Hox clusters, have been shown to be expressed in progressively posterior domains in the visceral mesoderm in the chick hindgut, colinear with their order on the chromosome (Roberts et al., 1995; Yokouchi et al., 1995). Moreover, Hoxd-12 and Hoxd-13 targeted disruptions result in malformations of the anus sphincter (Kondo et al., 1996). In addition, ectopic expression of Hoxa-4 and Hoxc-8 in transgenic mice causes abnormal development in the digestive system (Pollock et al., 1992; Wolgemuth et al., 1989). Our data that the DR₅ RARE is required for Hoxb-1 regional expression in the foregut and its anteriorization in the gut epithelium in response to RA are consistent with the concept of the classical Hox code applied to the developing gut. In such a model, Hox genes are coordinately expressed in the gut from the anterior to the posterior end in a colinear fashion, as they are expressed in the neural tube (Kondo et al., 1996; Roberts et al., 1995; Yokouchi et al., 1995).

In Drosophila, homeotic genes are expressed in restricted parts of the visceral mesoderm along the anteroposterior axis during gut morphogenesis. The restricted homeotic gene expression is known to determine the morphological borders of the gut (reviewed in Bienz, 1994). Labial, the Drosophila homolog of Hoxb-1, is expressed in a restricted domain in the epithelium in the midgut that is required for constriction of the endoderm (Panganiban et al., 1990). Decapentaplegic protein (DPP), produced and secreted from the visceral mesoderm cells, locally induces the region-restricted expression of labial. There are intriguing parallels between the expression patterns of Hoxb-1, sonic hedgehog and bone morphogenic protein (BMP) genes in the vertebrate gut and those of their Drosophila homologs, labial, hedgehog, and dpp during Drosophila morphogenesis (Roberts et al., 1995). Given the remarkable correlation of Hox gene expression with morphological borders, it is likely that Hox gene expression also regulates aspects of vertebrate gut morphology.

Our results also suggest that the retinoic acid receptors, RAR and RXR, may function in the anteroposterior patterning during gut development. RAR and RXR mRNA expression patterns during early embryogenesis have been well documented (reviewed in Mangelsdorf et al., 1994). For example, RARα is ubiquitously expressed while transcripts of RARβ and RARγ are temporally and spatially restricted (Ruberte et al., 1990). The expression patterns of RAR and RXR transcripts show some overlap with Hoxb-1, including their expression in the embryonic gut and many endoderm-derived tissues. More specifically, at ∼8.5 d.p.c., embryos express RARβ mRNA in the foregut endoderm and CRBPI transcripts were detected in the gut endoderm (Ruberte et al., 1991), CRABP II transcripts were observed in the foregut endoderm at ∼8.5 d.p.c. embryos, whereas CRABP I mRNA was not seen (Ruberte et al., 1992). Moreover, mice with RAR/RAR or RAR/RXR double mutations show abnormalities in several organs derived from the embryonic foregut during development (Kastner et al., 1997; Mendelsohn et al., 1994). Thus, RARα and/or RARβ are implicated in Hoxb1 regulation by RA in the developing foregut, whereas RARγ appears not to be involved. Consistent with these genetic and expression data, teratogenic doses of RA have been shown to cause abnormalities in the digestive tract in Syrian hamsters (Shenefelt, 1972). Therefore, it is likely that the endogenous retinoid signaling pathway plays an important role in gut development through the regulation of Hox gene expression. We are currently investigating whether there are any abnormalities in gut morphology, especially in the foregut region, after exogenous RA treatment of 8.5-9.5 d.p.c. mouse embryos.

It is worth noting that Hox gene knockout and misexpression experiments have primarily reported phenotypic abnormalities in the craniofacial region, hindbrain, axial skeleton and limbs (reviewed in Krumlauf, 1994). There have been few reports of abnormalities in other regions of the embryo, such as the gut, associated with alterations in Hox gene expression. This may be because in certain tissues, like the digestive tract, there is significant functional redundancy among many different Hox genes, especially paralogous genes, so that an alteration in the expression of any one gene would not significantly affect the development of the gut. Alternatively, there may be subtle functional abnormalities in many of these organs that have gone unrecognized because the evaluation of the Hox mutants thus far has been focused on more obvious abnormalities.

Previously several regulatory elements have been identified in both the 5′ and 3′ flanking regions of murine and human Hoxb-1 genes (Marshall et al., 1994; Ogura and Evans, 1995a,b; Studer et al., 1994). Marshall et al. (1994) found a DR₂-type RARE ∼2 kb 3′ of the murine Hoxb-1-coding sequence that is conserved in chick and pufferfish. Transgenes with this DR₂ RARE can partially recapitulate the early expression pattern of the endogenous Hoxb-1 gene and can mediate the response to exogenous RA in the neural tube at early times (∼7.5-8.5 d.p.c.). The Hoxb-1 3′ RAIDR₅ RARE was not included in the transgene constructs used in these prior studies. The restricted rhombomere 4 expression of the endogenous Hoxb-1 gene has been shown to be dependent on two enhancers in the Hoxb-1 5′ flanking region, an r4 enhancer with a Hoxb-1 autoregulatory loop and a 5′ DR₂ RARE repressor element (Popperl et al., 1995; Studer et al., 1994).

Our studies show that the RAIDR₅ RARE, located further 3′ of the previously studied DR₂ RARE, is also required for the lacZ transgene to recapitulate the correct regional expression of the endogenous Hoxb-1 gene in the gut in both early and late stages (Figs 3, 4). Therefore, the regulation of Hoxb-1 expression and its response to RA involves multiple RAREs and other regulatory elements.

From the studies described here and elsewhere (Marshall et al., 1994), it is clear that expression of Hoxb-1 in both neuroectoderm and endoderm is directly regulated by RA signaling. There are two phases of Hoxb-1 sensitivity to exogenous RA. The Hoxb-1 expression boundary in r4 exhibits a period of sensitivity to exogenous RA when embryos are treated before 8.5 d.p.c.. This initial RA-response is dependent on the Hoxb-1 3′ DR₂ RARE. During the second phase at ∼9.5 d.p.c., Hoxb-1 expression in r4 becomes insensitive to exogenous RA, but Hoxb-1 expression in the foregut is still sensitive to alteration by RA. This later RA response is dependent on the 3′ DR₅ RARE. Thus, Hoxb-1 expression is established by the sequential activation of different RAREs in different germ layers. The multiple RAREs found in the Hoxb-1 enhancers suggest that endogenous retinoids play an essential role in the regulation of Hoxb-1 and hence in the anteroposterior patterning of the embryo.

Hoxb-1, Hoxa-1 and Hoxd-1 constitute the paralog group 1, evolutionarily related to the Drosophila labial homeotic gene. Hoxb-1 shares several features of its early expression pattern with Hoxa-1, both spatially and temporally (Frohman et al., 1990; Murphy and Hill, 1991). In contrast to the complex regulatory elements found in Hoxb-1, only one functional DR₅ RARE has been identified in the 3′ flanking sequence of Hoxa-1 to date (Frasch et al., 1995; Langston and Gudas, 1992; Langston et al., 1997). We have previously shown that the Hoxb-1 3′ RAIDR₅ enhancer is conserved in the chick and murine Hoxb-1 genes, as well as in the human and murine Hoxa-1 genes. There are three blocks of conserved sequences in the enhancer: a DR₅ RARE, a conserved element 1 (CE1) and a conserved element 2 (CE2). The Hoxb-1 DR₅ RARE sequence is 5′-GGTTCA (N)₅ AGTTCA-3′ and is identical to the DR₅ RARE found in the Hoxa-1 3′ enhancers. The high degree of homology among the three conserved sequences suggests that the RAIDR₅ enhancers in the Hoxb-1 and Hoxa-1 genes have been conserved during the duplication and divergence of the Hox gene complexes. Frasch et al. (1995) demonstrated that the Hoxa-1 3′ enhancer directs lacZ transgene expression in the floor plate, notochord, gut epithelium and the posterior neural tube in transgenic mice. More recently, Dupé et al. (1997) used a cre-lox targeting mutation of the Hoxa-1 3′ DR₅ RARE to demonstrate that this enhancer plays an important role in the early establishment of the Hoxa-1 anterior expression boundary in the neural plate. These data show that the function of the Hoxa-1 DR₅ RARE appears to be different from that of the Hoxb-1 DR₅ RARE, as we have shown in this study. It is possible that the Hoxb-1 3′ DR₂ and DR₅ RAREs have some redundant functions in determining Hoxb-1 expression in the neural tube and notochord, but not in the gut. Alternatively, during the Hox gene divergence these two enhancers may have acquired some different functions in the control of the region-specific expression of these genes.

RARβ is known to be directly transcriptionally activated by RA through a DR₅ RARE identical to the Hoxb-1 DR₅ RARE (de Thé et al., 1990; Sucov et al., 1990). By 8 d.p.c., RARβ expression is seen in many regions of the embryo that overlap with regions expressing Hoxb-1, including the neural tube, lateral mesoderm and the foregut (Ruberte et al., 1990). The overlapping expression domains in the gut among RARβ, Hoxa-1 and Hoxb-1 may result from the presence of a common DR₅ RARE in their regulatory regions. Alternatively, the RARβ receptor itself may play a role in the responses of the Hoxb-1 and Hoxa-1 genes to bioactive retinoids in the gut. The presence of a DR₅ RARE in several genes expressed in the embryonic gut suggests that endogenous retinoids such as RA play a regulatory role in the developing gut.

We thank all the members of the Gudas laboratory for helpful discussions and Taryn Resnick for editorial assistance. This research was supported by NIH grant R01CA39036 to L. J. G., and in part by a Leukemia Research Foundation fellowship to D. H. and an NIH fellowship (1F32CA71153) to S. W. C.