Ectopic Hoxa-1 induces rhombomere transformation in mouse hindbrain

Genetic manipulation of Hox genes in chicken and mouse embryos indicates that vertebrate Hox genes share a similar role with the Drosophila HOM-C genes in specifying posi-tional identity. Homeotic transformations have been described in transgenic mice and chick embryos that exhibit ectopic expression of Hox genes (Kessel et al., 1990; Kessel and Gruss, 1991; Morgan et al., 1992; Lufkin et al., 1992; Jegalian and De Robertis, 1992; Pollock et al., 1992). Loss-of-function alleles in mice produce either homeotic transformations (Le Mouellic et al., 1992; Ramirez-Solis et al., 1993; Gendron-Maguire et al., 1993) or distinct developmental defects in regions corresponding to their rostral domains of expression (Chisaka and Capecchi, 1991; Lufkin et al., 1991; Chisaka et al., 1992).

The developing hindbrain is well suited to study the role of Hox genes in pattern formation and to examine the relation-ship between retinoid signaling pathways and the Hox gene expression. The vertebrate hindbrain is segmented during early development into rhombomeres, a series of transient structures along its length (Vaage, 1969). Rhombomeres are lineage restriction compartments. Cells in each rhombomere do not cross established boundaries and are programmed to form only that precise part of the hindbrain (Fraser et al., 1990; Lumsden, 1990). Thus, motor neurons of the cranial nerve V derive from rhombomere (r) 2 and r3, those of VII from r4 and r5 and those of IX from r6 and r7 (Lumsden and Keynes, 1989). Rhombomere transplantation experiments have shown that the establishment of positional identity is intrinsic to the hindbrain from an early stage of development (Guthrie et al., 1992). Many Hox genes and the zinc finger gene Krox-20 are expressed in the hindbrain with distinct rhombomere boundaries and the Hox genes also show ordered domains of expression in cranial neural crest-derived structures. The importance of these genes in rhombomere development have been illustrated by gene targeting experiments. Disruption of Krox-20 results in marked reduction or loss of r3 and r5 (Swiatek and Gridley, 1993; Schneider-Maunoury et al., 1993). Null mutation of the Hoxa-1 gene shows that the r5 and a large part of r4 are missing in the embryos (Mark et al., 1993; Carpenter et al., 1993). However, direct evidence is lacking to show a role for these genes in determining the positional identity for rhombomeres.

Retinoic acid (RA) treatment has been shown to alter the expression boundaries of homeobox genes and to cause homeotic transformations within the vertebrae (Kessel and Gruss, 1991) and hindbrain (Marshall et al., 1992; Kessel, 1993; Wood and Morriss-Kay, personal communication). We have found that expression of the Hoxa-1 gene is dramatically altered during gastrulation and early neurulation stages within mouse embryos following treatment of pregnant dams with all-trans RA (t-RA). In fact, after treatment with t-RA for only 4 hours at 7.5 days post coitum (d.p.c.), Hoxa-1 is expressed over the entire embryonic ectoderm and mesoderm, exhibiting clear ectopic expression in the developing headfold region (H.-J. K. and J. F. G., unpublished observations; Armstrong et al., 1992). Since this RA-induced Hoxa-1 expression was associated with the appearance of defects in head development, we were interested in determining the role that the Hoxa-1 gene itself may play in the process of RA-induced hindbrain transformation. Here we show that widespread ectopic expression of the Hoxa-1 gene directed by a human β-actin promoter in transgenic mice produces phenotypes in a subset of the ectopic domains primarily located in anterior regions. Associated with the morphological alterations is a specific change in the pattern of expression of another gene, Hoxb-1. No changes in the segmental patterns of expression of Hoxd-1, Hoxa-3 and Krox-20 are seen in the hindbrain while Hoxb-1 is ectopically activated in select hindbrain regions. Using lacZ fusion con-structs that are expressed in either r2 or r4, we show that after ectopic Hoxa-1 expression, markers normally expressed in r4 also appear in r2 and r3 while r2-specific markers disappear. Additionally, we find that neurons specific for the r4 region now appear in r2 and r3. This observation is supported by the finding that the Hoxa-2 and Hoxb-1 genes are expressed in a subset of neural crest cells migrating from r2 in the Hoxa-1 transgenic mice. In control embryos, Hoxa-2 and Hoxb-1 are only expressed in neural crest cells migrating from r4 but not r2. Taken together, we present evidence that when ectopically expressed, the Hoxa-1 gene itself can induce a reorganization of the mouse hindbrain, including a partial transformation of several rhombomeres into an r4-like identity. These data are consistent with a possible role of the Hox genes in conferring the positional identity of rhombomeres and point to the Hoxa-1 gene as a possible molecular mediator of the retinoic acid-responsive alteration of hindbrain identity.

To construct β-actin/Hoxa-1, a 7.3 kb SalI-XbaI fragment containing the 4.3 kb human β-actin promoter with intervening sequence 1, an SV40 polyadenylation signal (0.3 kb) and additional vector sequences was excised from the plasmid Lβ-actin(IVS)lacZ (supplied by J. Mann), blunt-ended and ligated to a blunt-ended 1.9 kb BamHI fragment from the plasmid pERA-1 (LaRosa and Gudas, 1988) to generate pLβAHox1.6. A 6.5 kb EcoRI-HindIII fragment was excised from pLβAHox1.6 and subcloned into the EcoRI-HindIII sites of pGEM-7Zf(+) (Promega) to generate pG7bAHox1.6.

Two lacZ reporter constructs were generated to follow the expression of a rhombomere-2-specific marker. The PstI-EcoRV fragment from pMC1871 (Pharmacia) was cloned into PstI and EcoRV sites of the vector pSL1190 (Pharmacia) to provide the 5′ one-third coding sequence of the lacZ gene with multiple cloning sites. Plasmid pSLlacZpA was constructed by digesting the above plasmid with EcoRV and SmaI and ligated with Klenow-blunted EcoRV-BamHI fragment from LβActin(IVS)lacZpA, which furnishes the remaining lacZ coding sequence and an SV40 polyadenylation signal. The EcoRI-AatII 7 kb genomic DNA fragment, which contains 660 bp exon 1 of Hoxa-1 and the presumptive Hoxa-1 promoter and the rhombomere-2-specific regulatory elements of Hoxa-2 (unpublished data, M. Z. and J. F. G.) was blunted with T4 DNA polymerase and cloned into the SmaI site of pSLlacZpA to generate plasmid pHox1.6lacZpA. The in-frame fusion of lacZ to Hoxa-1 was confirmed by DNA sequencing. The second lacZ reporter construct is a modification from the pHox1.6lacZpA. A 5.1 kb genomic sequence (AatII-EcoRI fragment) which contains the remaining exon 1, intron, exon 2 and 2.8 kb 3′ genomic sequence was added to the 3′ end of previous fusion gene to generate plasmid p12k-Z.

To purify the fusion genes for microinjection, The pG7bAHox1.6 was digested with XbaI to release the insert containing the human β-actin promoter, the Hoxa-1 coding sequences and the SV40 polyadenylation signal. The pHox1.6lacZpA and p12k-Z were digested with SalI to free the lacZ reporter constructs (10.3 kb and 15.4 kb, respectively). The inserts were then purified on a 10-40% sucrose gradient and adjusted to 2 μg/ml in 10 mM Tris-HCl (pH 7.5), 0.1 mM EDTA for microinjection.

Transgenic mice were produced essentially as described by Hogan et al. (1986). 4-week-old B6CBAF1 (The Jackson Laboratory) female mice were superovulated by injecting intraperitoneally 5 IU pregnant mare’s serum followed by injection of 5 IU human chorionic gonadotropin 46 hours later. Fertilized eggs were collected from the oviducts of the females after mating to C57BL (The Jackson Laboratory) males. DNA fragments (2 ng/μl) purified from the prokaryotic vector sequences were microinjected into the pronuclei of the zygotes, which were then reimplanted to the oviducts of pseudopregnant CD-1 (Charles River) female mice. Dual transgenic mice carrying β-actin/Hoxa-1 and Hoxb-1/lacZ were generated by injection of β-actin/Hoxa-1 into the eggs fertilized by Hoxb-1/lacZ transgenic male. Dual transgenic mice shown in Fig. 7 were created by co-injection of the β-actin/Hoxa-1 and a corresponding lacZ fusing gene. Equal moles of the two transgene constructs were mixed for injection.

Tail DNA was isolated as described (Walter et al., 1989) and used in a PCR assay to determine the genotype of new born mice. Yolk sac DNA was used to assay the genotype of founder embryos. Yolk sac was separated from each individual embryo, digested with 100 μg/ml proteinase K overnight at 55°C in 50 mM KCl, 10 mM Tris-HCl (pH 8.3), 2 mM MgCl₂, 0.45% Tween 20 and 0.45% NP40, and then incubated at 95°C for 20 minutes. A pair of primers from exon 1 (5′-GGACAAGAGCAGCAGACTCTG-3′) and exon 2 (5′-GGGTCT-CATTGAGCTGTAGGG-3′) of the Hoxa-1 gene were used in the PCR amplification to determine whether a pup or embryo was trans-genic. Using this PCR analysis, both the transgene and the endogenous Hoxa-1 (internal control) gene are amplified, yielding 320 bp and 770 bp products, respectively.

RNA expression was examined by in situ hybridization analysis in 6 μm-thick paraffin sections of mouse embryos as described (Wilkinson and Green, 1990). Sections were hybridized to ³⁵S-labeled riboprobes prepared from RNA expression vectors. The ³⁵S-labeled riboprobes were adjusted to approximately 150 bp in size by limited alkaline hydrolysis. Generally, hybridized slides were exposed to Kodak NTB-2 emulsion for 3-4 days.

The Hoxa-1 probe contained a 700 bp EcoRI-BamHI fragment from the Hoxa-1 cDNA. The Hoxb-1 probe contained a PCR-cloned 450 bp fragment from total mouse genomic DNA, representing nucleotides 901-1350 of the Hoxb-1 gene (Frohman et al., 1990). The Hoxa-2 probe contained a 280 bp ApaI-PstI fragment of the Hoxa-2 gene. The Hoxa-3 probe was isolated from a 410 bp PvuII-EcoRV fragment from Hoxa-3 genomic DNA containing some homeobox and 3′ flanking sequences (McGinnis et al., 1984; Baron et al., 1987). The Krox-20 probe (nucleotides 2417-2789 in the 3′ untranslated region (Chavrier et al., 1988)) was obtained by PCR-cloning from total mouse genomic DNA. All these probes were cloned into suitable RNA expression vectors.

Histochemical staining of the mouse embryos with X-gal was conducted as described by Kothary et al. (1989) with slight modifi-cation. Embryos were fixed in 0.2% glutaraldehyde (Sigma) in PBS with 2 mM MgCl₂ and 5 mM EGTA for 15 minutes and then rinsed three times in PBS with 2 mM MgCl₂, 0.02% NP40 and 0.01% sodium deoxycholate. Embryos were stained in 0.1% X-gal (Sigma) in PBS with 2 mM MgCl₂, 0.02% NP40 and 0.01% sodium deoxy-cholate, 20 mM K₃Fe(CN)₆ and 20 mM K₄Fe(CN)₆ at 37°C for 12 hours and stored in PBS containing 10 mM EDTA at 4°C.

Mouse embryos were collected at 9.5 d.p.c. and fixed in 3.5% paraformaldehyde in PBS (Ca²⁺- and Mg²⁺-free) overnight at 4°C. Whole-mount immunohistochemistry with the anti-neurofilament (155,000 M_r) monoclonal antibody 2H3 (Dodd et al., 1988) was conducted as described (Swiatek and Gridley, 1993). The hindbrains of the stained embryos were dissected, opened along the midline of the ventricle roof, flat-mounted with dorsal side up on slides in 10% PBS, 90% glycerol and 0.02% sodium azide, and sealed with plastic. Photographs were taken on a Zeiss Axioplan microscope.

To express Hoxa-1 ectopically in the mouse, we constructed a β-actin/Hoxa-1 expression vector, in which the Hoxa-1 coding sequence (LaRosa and Gudas, 1988) was placed under the control of a human β-actin promoter plus intervening sequence 1 (Gunning et al., 1987). Transgenic mice carrying this vector were generated by DNA microinjection into mouse zygotes. In experiments designed to establish stable lines, no transgenic mice were born from eggs microinjected with the β-actin/Hoxa-1 construct. This result suggested that ectopic Hoxa-1 expression could be embryolethal and we began to examine founder embryos at earlier stages of development. We identified 28 Hoxa-1 transgenic embryos at approximately 9.5 d.p.c. by PCR analysis of yolk sac DNA. Six of these trans-genic embryos were analyzed for RNA expression by in situ hybridization. This analysis clearly shows that Hoxa-1 is ectopically expressed in the transgenic embryos (Fig. 1). In control embryos at 9.5 d.p.c., Hoxa-1 expression is found in the foregut epithelium and in posterior neuroectoderm (Fig. 1D,E) (also see Murphy and Hill, 1991). However, when sections of Hoxa-1 transgenic embryos harvested at 9.5 d.p.c. are probed for Hoxa-1, a strong, generally ubiquitous pattern of expression is detected (Fig. 1F,G). Thus, under control of the human β-actin promoter, Hoxa-1 is ectopically expressed in anterior regions, rostral to the normal boundary of expression seen in control embryos.

The accompanying embryolethality made it difficult to analyze morphological phenotypes associated with ectopic Hoxa-1 expression in detail. However, 20 of the Hoxa-1 trans-genic embryos examined at 9.5 d.p.c. clearly exhibited developmental abnormalities that manifested predominantly as defects in anterior neural tube closure (Fig. 1C) or alterations in the general morphology of hindbrain (Fig. 1B) and forebrain regions. Approximately 40% of the Hoxa-1 transgenic embryos showed alterations in hindbrain development including, in some cases, loss of the typical rhomboid shape (Fig. 1B). Although the observed phenotypes varied in detail, it is interesting to note that the developmental defects generally appear in anterior regions of the embryos that do not normally express the Hoxa-1 gene.

As the most dramatic morphological alterations were detected in anterior regions of the Hoxa-1 transgenic embryos, we wanted to know if these dysmorphologies were accompanied by changes in the expression of other genes in the Hox network. Thus, we examined the expression patterns of the homeobox genes Hoxa-2, Hoxa-3, Hoxb-1 and Hoxd-1. Hoxa-2 and Hoxa-3 reside approximately 6 kb and 20 kb, respectively, upstream of Hoxa-1 on chromosome 6 (Tan et al., 1992; McGinnis et al., 1984). Hoxb-1 and Hoxd-1 are paralogous to Hoxa-1 and reside on mouse chromosomes 11 and 2, respectively (Krumlauf, 1992). We also examined the expression pattern of Krox-20, a gene encoding a zinc finger-containing protein thought to play a role in regulating segmentation within the hindbrain (Wilkinson et al., 1989; Sham et al., 1993; Swiatek and Gridley, 1993).

Of particular interest, ectopic Hoxa-1 dramatically alters the regulation of Hoxb-1 expression in select regions of the hindbrain. In control embryos at 9.5 d.p.c., Hoxb-1 is seen in r4 and in cells populating the VII/VIII ganglion complex (Murphy and Hill, 1991) (Fig. 2A,B). Within the hindbrains of Hoxa-1 transgenic embryos at 9.5 d.p.c., the boundaries of Hoxb-1 expression have been extended both rostrally and caudally, but it is interesting that Hoxb-1 is not activated in all of the sites of ectopic Hoxa-1 expression (Fig. 2C,D). The altered pattern of Hoxb-1 expression can be categorized by a second major domain in addition to that seen in r4 which extends from r2 into r1. Rhombomere 5 is always negative for Hoxb-1 expression and there is a variable or weak expression in r3 and r6 (Fig. 2C,D). These changes in Hoxb-1 expression appear in anterior regions and we have not detected alterations in its posterior domains. A similar pattern of altered Hoxb-1 expression was seen in all of the Hoxa-1 transgenic embryos examined.

These results were intriguing and led us to examine the mechanisms by which Hoxa-1 could regulate Hoxb-1 expression in the hindbrain. As a first step in determining if this cross-regulatory relationship between Hoxa-1 and Hoxb-1 was a direct effect, we asked if ectopic Hoxa-1 expression would activate a Hoxb-1/lacZ transgene (Marshall et al., 1992). Stable lines bearing the Hoxb-1/lacZ transgene alone reiterate the endogenous pattern of Hoxb-1 expression seen in control embryos (compare Fig. 2A with A′, and B with B′), exhibiting β-galactosidase activity that in the hindbrain becomes confined to r4 and to the neurogenic neural crest cells migrating from this region at 9.5 d.p.c. (Fig. 2A′, B′). In the dual transgenic embryos made from the Hoxb-1/lacZ lines by microinjecting the β-actin/Hoxa-1 construct, however, there is a significant induction of β-galactosidase activity in the r1 and r2 regions (Fig. 2C′, D′). Weaker signals are also detected in the region of r3 and r6 (Fig. 2C′, D′). No induction of β-galactosidase activity is detected in r5 and posterior domains of Hoxb-1/lacZ are unaffected. As the response of the Hoxb-1/lacZ transgene to ectopic Hoxa-1 closely matches that of the endogenous Hoxb-1 gene (compare Fig. 2D and D′), this Hoxb-1/lacZ transgene appears to contain the elements necessary for Hoxb-1 regulation by Hoxa-1. This data also indicates that induction of Hoxb-1 at the translational level (β-galactosidase activity) is comparable to that at the transcription level (in situ hybridization).

We have defined an 1.5 kb DNA fragment at the 5′ flanking region of the Hoxb-1 gene that directs β-galactosidase activity predominately to r4, with some posterior expression (Fig. 3A). This is achieved with a series deletion experiments on the 18 kb Hoxb-1/lacZ transgene. Again, as with the 18 kb Hoxb-1 genomic fragment shown in Fig. 2, ectopic Hoxa-1 expression can produce a similar duplication of β-galactosidase activity in the r2 and r4 regions in transgenic mice containing this r4-specific lacZ construct (Fig. 3B,C). Preliminary data using lacZ fusion gene with smaller fragments within this Hoxb-1 regulatory region suggest that this region contains an enhancer that directs the late pattern of Hoxb-1 expression in r4. It is interesting that all fragments directing r4 expression were capable of inducing β-galactosidase activity in r2 following ectopic Hoxa-1 expression (H. M. and R. K., unpublished observa-tion). At present, we have not completed experiments to demonstrate whether the Hoxa-1 gene product can directly bind to regulatory regions in the Hoxb-1 gene. We were impressed with the striking similarity in the response of the Hoxb-1/lacZ transgenic embryos that had been treated with RA (Marshall et al., 1992) and that exhibited ectopic Hoxa-1 expression. RA has been shown to cause the r4-specific expression of Hoxb-1/lacZ constructs in the r2 region, suggesting that a transformation of r2 into an r4-like identity has occurred. In our results, since ectopic Hoxa-1 expression also resulted in ectopic Hoxb-1 expression in r2, we looked more closely for other markers that would suggest a similar transformation.

During a study to characterize the cis-acting regulatory elements required for Hoxa-1 and Hoxa-2 gene expression, we identified an r2-specific element, which is one of the enhancers for the Hoxa-2 gene (M. Z. and J. F. G., unpublished observation). This enhancer directs expression of the reporter gene lacZ predominantly in r2 (Fig. 4A). Expression in the pericardium and some somites is also seen. If the genotypic and phenotypic changes observed in r2 represent a rhombomere transformation to an r4 identity in the Hoxa-1 transgenics, then r2-specific reporter genes should be suppressed in the r2 region in these embryos. Three dual transgenic mice were generated by co-injection of β-actin/Hoxa-1 and lacZ construct A (Fig. 4E). Two of the three animals fail to express β-galactosidase activity in r2 whereas the expression in the pericardium is unchanged (Fig. 4B). In the third animal, only a few cells in r2 still express the reporter (data not shown), compared to a prominent stained band in the control (Fig. 4A).

Another lacZ construct (lacZ construct B in Fig. 4E) containing an additional enhancer from the Hoxa-1 gene (M. Z. and J. F. G., unpublished observation) was used to generate transgenic animals in this study. Animals carrying this transgene exhibit β-galactosidase activity in another broad posterior domain in addition to r2 that more closely represents the posterior expression domain of the endogenous Hoxa-1 gene (Fig. 4C). Double transgenic mice were created by co-injection of β-actin/Hoxa-1 and lacZ construct B. Again, when Hoxa-1 is ectopically expressed, lacZ reporter gene activity in r2 is diminished, whereas the posterior expression domain is generally not affected (Fig. 4D).

It is clear that ectopic Hoxa-1 can upregulate expression of an r4-specific marker predominantly in r2 and downregulate an r2-specific marker. These data suggest that ectopic Hoxa-1 could transform r2 into an r4-like phenotype. To test the extent of such transformation in the hindbrain neuroectoderm and to determine the selectivity of this effect, we examined the expression patterns of other genes with unique expression domain in the hindbrain.

Despite the dramatic ectopic expression of the Hoxa-1 gene, no obvious changes in the expression patterns of Hoxa-3 and Krox-20 are found in the Hoxa-1 transgenic mice (Fig. 5C-F) Additionally, no ectopic expression of the Hoxd-1 gene is detected in the hindbrain of Hoxa-1 transgenics (Fig. 5A,B). These results contrast with the response of the Hoxb-1 gene and suggest that ectopic Hoxa-1 expression can cause select alterations in patterns of gene expression in the hindbrain. It is interesting to note that while Hoxd-1 is a labial family member, its expression pattern is not altered in a manner similar to its paralogous gene, Hoxb-1. This result is not surprising since, unlike Hoxa-1 and Hoxb-1, Hoxd-1 exhibits no expression in hindbrain neuroectoderm (Frohman and Martin, 1992; Krumlauf, 1992). Apparently, the regulatory elements directing the Hoxa-1-inducible ectopic expression of Hoxb-1 are not shared by Hoxd-1.

In control embryos at 9.5 d.p.c., strong expression of Hoxa-2 is detected in hindbrain neuroectoderm with an anterior boundary of expression at the level of r1/r2 boundary (Fig. 6A). The general pattern of Hoxa-2 expression in the Hoxa-1 transgenics resembles that of control embryos in the neuroectoderm of the hindbrain (Fig. 6B). In control embryos, Hoxa-2 is also expressed in neural crest cells in the facial/vestibuloacoustic (VII/VIII) ganglion complex, but is never expressed in the cell population associated with the trigeminal (V) ganglion (Fig. 6A, and H.-J. K., unpublished data). In the Hoxa-1 transgenic mouse embryos, however, the expression of Hoxa-2 is detected in neural crest cells lateral to r2, r3 and r4 (Fig. 6B). To confirm this observation, sagittal sections which run through both trigeminal and facial/vestibuloacoustic ganglia were used to examine the Hoxa-2 expression pattern by in situ hybridization. In addition to strong expression in the VII/VIII ganglion and branchial arch 2, Hoxa-2 is indeed expressed in the neural crest cells in the trigeminal ganglion area (Fig. 6C). In the cranial nerve ganglia of control embryos, Hoxb-1 exhibits a similar expression pattern as Hoxa-2. It is expressed in the neural crest cells populating the facial/vestibuloacoustic, but not the trigeminal ganglia (Frohman et al., 1990; Hunt et al., 1991). Interestingly, there is also an induction of Hoxb-1 in the area of the trigeminal ganglion in the Hoxa-1 transgenic mouse embryos (Fig. 6D). We also notice that the induction of both Hoxb-1 and Hoxa-2 is restricted in the posterior portion of the trigeminal ganglion. It is clear that only a small population of cells in these regions express Hoxb-1 and Hoxa-2 genes.

When Marshall et al. (1992) reported that retinoic acid induced transformation of r2/3 to an r4/5-like identity in the hindbrain of mouse embryos, they found that, in addition to alterations in Hoxb-1 expression, r2-derived neurons were transformed into neurons with r4-specific characteristics. The alteration of Hoxb-1 and Hoxa-2 expression in the hindbrain region of the Hoxa-1 transgenic mice suggest a possible rhombomere transformation induced by the ectopic Hoxa-1. To analyze the phenotypic changes associated with the putative rhombomere transformation induced by ectopic expression of Hoxa-1, antibody against the neurofilaments was used in whole-mount immunohis-tochemistry to visualize neurons in mouse embryos. The gross morphology of the cranial nerves in the Hoxa-1 transgenic mice appears indistinguishable from those in the control embryos (Fig. 7A). Further analyses of the neuron staining pattern in the rhom-bomeres of the hindbrain were then conducted.

Studies have shown the existence of a population of r4-specific efferent neurons in the hindbrain (Marshall et al., 1992; Simon and Lumsden, 1993). These neurons migrate contralaterally and their primary dendrites extend across the floor plate. We analyzed 9 Hoxa-1 transgenic embryos and 20 age-matched (by the number of somites) control embryos with neurofilament staining. In the control, there are neurons that cross the midline at the level of r4, but not at r2, as expected (Fig. 7B). However, in 7 out of the 9 transgenic embryos examined, neurons and their processes are detected across the midline at r2 as indicated by arrows in Fig. 7C. Thus, in accordance with the changes of expression pattern of Hox genes, r2 appears to undergo a phenotypic transformation towards an r4-like identity in the Hoxa-1 transgenic mouse embryos. Interestingly, many contralaterally migrating neurons are also found in r3 region. Taking account of the observation that ectopic Hoxa-1 also induces Hoxb-1 expression in r3, it is conceivable that r3 may also undergo a partial transformation to an r4-like phenotype.

We have generated transgenic mice that exhibit ectopic Hoxa-1 expression in anterior regions during early development. This ectopic Hoxa-1 expression is embryolethal and results in morphological and molecular changes in the hindbrains of the transgenic embryos. As we could not establish a line of mice ectopically expressing the Hoxa-1 gene, we focused our analysis on 9.5 d.p.c. mouse embryos. The most striking and consistent alteration in gene expression that we observed in the Hoxa-1 transgenic embryos is the activation of Hoxb-1 expression in neuroepithelial cells of r2. Thus in the Hoxa-1 transgenics, two stripes of Hoxb-1 expression are generally visible, one in r2 and one in r4. Weaker and more variable activation of Hoxb-1 expression is detected in r1 and r3 and in a small population of cells in r6. Curiously, no ectopic Hoxb-1 expression has ever been detected in r5 of the Hoxa-1 transgenics.

On the basis of morphological observations and patterns of gene expression within the mouse hindbrain, it appears that ectopic Hoxa-1 expression can cause partial transformations of several rhombomeres into an r4-like identity. In these experiments, the most striking effect is associated with r2. In this region, we find strong and consistent expression of Hoxb-1 and r4-specific Hoxb-1 reporter genes. In addition, we find that an r2-specific marker is suppressed in the Hoxa-1 transgenic embryos. Both these observations are consistent with an interpretation that the r2 region has been transformed to an r4-like identity. The r2-specific marker is not expressed in r4 in control embryos and is suppressed in the Hoxa-1 transgenics, suggesting that the identity of r2 has been altered. Morphological evidence also supports the hypothesis that ectopic Hoxa-1 can induce rhombomere transformation. Immunohistochemical staining for neurofilament reveals the appearance of neurons that cross the floorplate within the r2, r3 and r4 boundaries in the Hoxa-1 transgenic embryos. In age-matched control embryos, neurons project across the floor-plate in r4 but not in r2 and r3. The appearance of these neurons that cross the midline in r2 and r3 resembles those crossing the floorplate at r4 and support the hypothesis that ectopic expression of Hoxa-1 in anterior regions can lead to a repatterning within the developing hindbrain. It is important to note that these morphological and molecular changes seen in response to ectopic Hoxa-1 appear to represent a partial transformation of the rhombomeres and their associated structures into the r4-like identity. While we have found convincing evidence that Hoxa-1 could induce strong changes in the expression of Hoxa-2 and Hoxb-1 reporter genes in r2 neu-roectoderm, only a subpopulation of neural crest cells migrating from r2 expresses the Hoxa-2 and Hoxb-1 genes. In addition, we could not detect any obvious alteration in the overall morphology of the trigeminal ganglia, which might be expected in the Hoxa-1 trans-genic embryos exhibiting the additional stripe of Hoxb-1 expression in r2. Neural crest cells derived from r1 and r2 populate the trigeminal ganglion while those migrating from the level of r4 populate the facial ganglion (Lumsden et al., 1991). Complete transformation of neural crest cells migrating from r2 into an r4-like identity may have lead to noticeable alterations in the appearance of the trigeminal ganglion.

In Drosophila, homeobox genes impart distinct developmental fates to cells within the insect parasegments (Akam, 1989; McGinnis and Krumlauf, 1992). In vertebrates, segmentation occurs during the formation of the somites and in the development of the hindbrain, where the transient rhombomere structures form. A Hox code theory is hypothesized for the development of the vertebrae, in which the identity of a vertebral segment is specified by the combination of Hox genes expressed there (Kessel and Gruss, 1991). Thus, posterior transformations are found in gain-of-function mutations (Kessel et al., 1990; Kessel and Gruss, 1991; Lufkin et al., 1992; Jegalian and De Robertis, 1992; Pollock et al., 1992) and anterior transformations result from loss-of-function mutations (Le Mouellic et al., 1992; Ramirez-Solis et al., 1993). In hindbrain, many Hox genes are expressed with distinct anterior boundaries which coincide with rhombomere boundaries. Thus, the concept of Hox code has been extended to hindbrain segmentation (Hunt et al., 1991; Krumlauf, 1993). Kessel (1993) reported the trans-formation of r3 into an r4 phenotype following RA treatment of mouse embryos. Interestingly, this posterior transformation of rhombomere is associated with the change of an r3 code to an r4 code in r3 (Kessel, 1993). Here we show that overexpression of Hoxa-1 gene can also lead to a putative transformation of anterior structures (r2 and r3) into more posterior structure (r4) in the developing hindbrain.

How the Hoxa-1 gene functions to help specifying rhombomere identity is not known. Hox/HOM-C genes encode DNA-binding homeodomain-containing proteins, which are believed to act as transcription factors. Only a few downstream targets other than the homeotic genes themselves have been reported. In Drosophila, autoregulation and cross-regulation of homeobox genes that is both direct and indirect has been documented (reviewed by Andrew and Scott, 1992). Accompanying both ectopic Hoxa-1 and RA-induced rhombomere trans-formations is a dramatic alteration of Hoxb-1 expression in specific hindbrain regions, with little change of the expression of linked Hox genes, including Hoxa-2 and Hoxa-3. These ectopic changes in the Hoxb-1 patterns of expression may be the result of a direct action of Hoxa-1 or RA on Hoxb-1 regulatory elements in permissive hindbrain regions. Clearly other components are required for modulating Hoxb-1 because it does not respond to Hoxa-1 in all sites of expression. To begin to approach the nature of the Hoxb-1 response to Hoxa-1, we performed a series of deletion experiments on the 18 kb Hoxb-1/lacZ transgene to map crudely the region that responds to Hoxa-1. In initial results, we have narrowed the regulatory activity to an 1.5 kb region upstream of the Hoxb-1 gene that functions as an enhancer directing β-galactosidase activity to r4 and also mediates the ectopic response to Hoxa-1 in r2. It is possible that Hoxa-1 works by directly binding to this 1.5 kb regulatory domain, as there are several TAAT motifs (Catron et al., 1993) of the type shown in vitro to be recognized by Hox/HOM-C proteins (H. M. and R. K., unpublished data). However, further analysis is required to determine if there is a direct or indirect response.

The action of Hoxa-1 on Hoxb-1 regulation in permissive hindbrain regions suggests that Hoxa-1 could also regulate Hoxb-1 expression at some time during normal development. Both genes are expressed in early gastrulation and for a time (7-8 d.p.c.) share similar expression domains. However, in Hoxa-1⁻/Hoxa-1⁻ mice, both the early phase and r4 restricted patterns of Hoxb-1 expression appear normal, with the exception that the r4/5 boundary is diffuse in association with the loss of r5 (Carpenter et al., 1993; Mark et al., 1993). Therefore in vivo Hoxa-1 is not required to establish and maintain Hoxb-1, but it still might be required for the ectopic response of Hoxb-1 to RA and this must be examined. Another possibility is that the cross-regulation between Hoxa-1 and Hoxb-1 that we have described in this paper represents a conserved property of the vertebrate labial-related genes. In Drosophila, labial is regulated by direct and indirect autoregulatory mechanisms (Chouinard and Kaufman; 1991) and the stimulation of Hoxb-1 by Hoxa-1 in transgenic mice may mimic an endogenous autoregulatory pathway of the Hoxb-1 gene itself.

Retinoic acid plays an important role in normal development and, in excess, can cause abnormal development (reviewed in Hofmann and Eichele, 1994). Regulation of Hox gene expression by retinoic acid has been demonstrated both in vitro (Simeone et al., 1990; 1991) and in vivo (Conlon and Rossant, 1992; Morriss-Kay et al., 1991), and an RA-responsive element has been identified in the Hoxa-1 gene (Langston and Gudas, 1992). We found that treatment with RA at 6.5 or 7.5 d.p.c. results in a dramatic induction of Hoxa-1 expression in the entire embryonic ectoderm and mesoderm (H.-J. K., unpublished data; Armstrong et al., 1992). It has been reported that RA can cause rhombomere transformation in mouse hindbrain (Marshall et al., 1992; Kessel, 1993; Wood and Morriss-Kay, personal communication). Taken together, it is conceivable that Hoxa-1 may act as a molecular mediator of the RA-induced rhombomere transformation. In other words, RA may first cause an induction of Hoxa-1 expression through the binding of its receptor to the RARE in Hoxa-1 gene. The ectopically expressed Hoxa-1 can then respecify the cells in r2 and r3 to a new r4-like identity.

The authors would like to thank Dr Jeff Mann for providing Lβ-actin(IVS)lacZ and Dr Janet Rossant for providing phspPTlacZpA.