Knockdown of the complete Hox paralogous group 1 leads to dramatic hindbrain and neural crest defects

A vital aspect of vertebrate development is the organisation and patterning of different tissues along the various axes of the embryo. Along the anteroposterior (AP) axis this occurs partly under the control of the Hox family of transcription factors. These homeobox-containing genes have been conserved throughout evolution and are responsible for the patterning of various tissues at specific axial levels(McGinnis and Krumlauf, 1992). In vertebrates, the Hox genes are organised in multiple clusters, which contain 13 paralogous groups, identified on the basis of DNA sequence and the position on the chromosome. These clusters are believed to be derived from the successive duplications of an ancestral Hox cluster. The subsequent loss of certain genes, and divergence and function shuffling in the remaining genes resulted in the clusters as they are now(Prince and Pickett, 2002). One interesting aspect of the Hox genes is that they are expressed in a colinear fashion. This means that genes which are located more 3′ in the cluster are generally expressed at an earlier developmental time point[temporal colinearity (reviewed by Deschamps et al., 1999)] and also at a more anterior position in the embryo [spatial colinearity (reviewed by Duboule and Dolle, 1989; Graham et al., 1989)]. The mechanism behind this colinearity is not yet delineated but the final effect is to produce a complex pattern of Hox gene expression, known as the `Hox code', which defines particular AP axial levels(Deschamps et al., 1999; Maconochie et al., 1996).

Often however, the removal of the function of one particular Hox gene, or even one complete cluster, does not have dramatic consequences for the embryo(Spitz et al., 2001; Suemori and Noguchi, 2000). This, together with the observation that paralogous genes often have similar functions as well as similar expression domains, points to the possibility of functional redundancy between genes from the same paralogous group (PG). This indeed appears to be the case. The loss of function of complete paralogous groups have been shown to be more severe than knockouts of single Hox genes in both zebrafish [PG2 (Hunter and Prince,2002)] and mouse [PG8 (van den Akker et al., 2001)]. Thus the knockdown of a single Hox gene may not reveal its complete function, and entire paralogous groups may need to be abrogated before their shared role can be illuminated.

Here we investigate the function of the PG1 genes, which are the homologues of the Drosophila labial gene. These are the earliest of the Hox genes, with expression starting during gastrulation, and they eventually have their anterior boundary in the hindbrain at the level of rhombomeres (r) 3/4 in most vertebrates (Frohman et al.,1990; Frohman and Martin,1992; Godsave et al.,1994; Kolm and Sive,1995; Murphy and Hill,1991; Sundin et al.,1990; Wilkinson et al.,1989). Hoxa1 is also expressed at a later stage in the fore/midbrain (McClintock et al.,2002; Shih et al.,2001).

Knockout studies in mouse have concentrated on the Hoxa1 and-b1 genes, and have implicated these genes in hindbrain and craniofacial development. In the Hoxa1 null mutant, r5 is either absent or severely reduced, r4 is reduced and there are defects in the hindbrain and associated nerves in the region between r3 and r8(Carpenter et al., 1993; Chisaka et al., 1992; Dolle et al., 1993; Lufkin et al., 1991; Mark et al., 1993). In the Hoxb1 knockout the identity of r4 is altered, but segmentation is not affected (Goddard et al.,1996; Studer et al.,1996). When both Hoxa1 and Hoxb1 are deleted, a more severe phenotype is observed than in either of the single knockouts, with both r4 and r5 being mis-specified(Gavalas et al., 1998; Studer et al., 1998) and eventually deleted (Rossel and Capecchi,1999) as well as the 2nd pharyngeal arch and its derived tissues being lost. In zebrafish, knockdown of the Hoxb1b gene (thought to be the functional equivalent of the mouse Hoxa1 gene) leads to disruption of r4 and Hoxb1a, like mouse Hoxb1, is involved in the specification of nerves originating in r4. The double knockdown of Hoxb1b and Hoxb1a also implies some degree of functional redundancy between these genes. However, the phenotype of the double knockdown is not as severe as that observed in mouse, with r4 and r5 always present,albeit reduced (McClintock et al.,2002).

Despite the intense interest in these anterior Hox genes, a complete knockdown of all PG1 genes has not yet been performed. Therefore any function that is shared between all of the genes may still be hidden. To address this question we knocked down all the Xenopus laevis Hox PG1 genes. In Xenopus, the early expression of the three PG1 genes (Hoxa1,Hoxb1 and Hoxd1) is highly overlapping and they are all expressed in the presumptive hindbrain region. Here we use the morpholino (MO)knockdown technique to demonstrate that the complete loss of PG1 gene function has deeper implications for the development of the embryo than the loss of the individual genes.

Xenopus laevis embryos were staged according to Nieuwkoop and Faber (Nieuwkoop and Faber,1956). Culture of embryos and buffers was as described(Winklbauer, 1990).

The complete open reading frame of Xenopus Hoxd1(Sive and Cheng, 1991) was amplified using primers incorporating BamHI and XhoI restriction sites (for:5′CCGGGATCCGCCGCCACCATGAATTCCTACCTAGAATACACTTCTTGCGGG; rev:5′TGCACTCGAGCTAGGGTGAAGCGTCCTTGGATGGCG). After ligation of the PCR product into the pGEM-T Easy vector (Promega), Hoxd1 was excised by BamHI/XhoI digestion, ligated into the CS2+vector (Turner and Weintraub,1994), and checked by sequencing.

Two morpholinos were designed for each Hox PG1 gene (Gene-Tools Inc.) The pseudo-tetraploidy of Xenopus laevis was taken into consideration and the morpholinos were all designed to be effective against both alleles present- determined by 5′ RACE analysis for Hoxd1 and comparison with available EST data for Hoxa1 and Hoxb1. Sequence of morpholinos is as follows: Hoxa1,MO1-5′CTCATCCTCCTCGCATAGTCCATCT, MO2-5′CTCGCATAGTCCATCTATCACTAGG; Hoxb1, MO1-5′AGGAACTCATTCTGCTATTGTCCAT,MO2-5′ATCTGCCAGTGATGGAGGAGGGTCA; Hoxd1,MO1-5′AGGAACCTGCTGATCCCTCAATCTT, MO2-5′TAGGTAGGAATTCATCTCTAGGGAG. Morpholinos and mRNAs were diluted in Gurdon's buffer (15 mM Tris pH 7.5, 88 mM NaCl, 1 mM KCl) and injected at the four-cell stage into both left blastomeres with GFP mRNA co-injected as a lineage tracer. Before fixation the GFP was checked to ensure that the injections were on the correct side. Amounts injected ranged from 5 ng to 30 ng of PG1 morpholinos and 7.5 ng to 60 ng of control morpholino. Whenever a comparison was made between single,double or triple MO injections the total amount of MO was equalised by adding control morpholino. Several combinations of the 1st and 2nd MOs were analysed with the Krox20 and Engrailed-2 probes and gave the same result. In all experiments control morpholino (standard control, Gene-Tools Inc.) was also injected and the embryos included in subsequent analyses, but this never gave different results from the non-injected controls. CS2+GFP (25 pg) and CS2+xHoxd1 (100 pg) were linearised with NotI and transcribed with Sp6 polymerase.

The whole mount in situ hybridisation protocol used was described previously (Wacker et al.,2004b), as modified from a previous protocol(Harland, 1991). Antisense,digoxigenin-labelled probes were: Hoxd1(Sive and Cheng, 1991); Hoxa1, Hoxb1, Hoxc6, Hoxb9(Wacker et al., 2004a); Krox20 (Bradley et al.,1993); Engrailed-2(Hemmati-Brivanlou et al.,1991); Nrp1 (Richter et al., 1990); Gbx2(von Bubnoff et al., 1996); Xslug (Mayor et al.,1995); Xsnail (Mayor et al., 1993); dll4(Papalopulu and Kintner,1993); Otx2 (Pannese et al., 1995); Hoxa2(Pasqualetti et al., 2000); Myod (Hopwood et al.,1989); EST clones from the I.M.A.G.E. Consortium [LLNL] cDNA library (Lennon et al., 1996), Hoxa3 (IMAGE4405749), or the NIBB library, Hoxd3 (XL012i13); Hoxd4 (XL094l20); Hoxa5 (XL045g13).

After bleaching (80% methanol, 6% H₂O₂, 15 mM NaOH),stage 46 embryos were washed (4 ×30 minutes PBS+0.2% Tween), blocked (30 minutes PBS+ 0.2% Tween, 3% BSA) and incubated overnight at 4°C with the neural antibody 2G9 (Jones and Woodland,1989). The embryos were then washed and incubated overnight at 4°C with a secondary antibody conjugated to the Cy5 fluorophore. After washing, the embryos were dehydrated and fixed step-wise in methanol (25%,50%, 75%, 100%). Before analysis embryos were cleared in Murrays and the hindbrain was visualised using scanning confocal microscopy (Leica TCS-NT). To visualize the cartilage, stage 49 embryos were fixed and stained with Alcian Blue (Pasqualetti et al.,2000).

To gain as complete a picture as possible of the expression of the PG1 genes, and to determine the possibilities for overlap of function, we performed an extensive whole mount in situ hybridization study of Hox PG1 gene expression (Fig. 1). This is the first comparative study of all three genes, although there is data on aspects of the individual genes (Godsave et al., 1994; Kolm and Sive,1995; Wacker et al.,2004a). All three PG1 Hox genes begin to be expressed during gastrulation, where they are seen in a ring around the blastopore, with a gap on the dorsal side. When this expression is compared with the expression of the forebrain/midbrain marker, Otx2, it can be seen that whilst none of the Hox genes' anterior expression reaches the midbrain, Hoxb1appears to be expressed more anteriorly than Hoxa1 and Hoxd1, almost fusing with the Otx2 domain(Fig. 1A,I,Q). As elongation procedes, differences in the gene expression patterns begin to arise. As early as stage 14, a stripe of Hoxb1 expression in the future hindbrain is observed and this becomes further defined by stage 18(Fig. 1K-N). Hoxa1 and Hoxd1 on the other hand do not exhibit this early stripe of expression, although they are both expressed in the hindbrain region(Fig. 1E,U). Expression of Hoxd1 thereafter decreases and from stage 26 there only remains faint expression in the pharyngeal arches and the tailbud(Fig. 1W,X). The hindbrain stripe of Hoxb1 expression persists until at least stage 30,accompanied by faint expression in the pharyngeal arches(Fig. 1O,P). Hoxa1remains expressed in a comparatively large domain along the axis, including the pharyngeal arches. In addition, as has been previously reported(McClintock et al., 2002), we observed midbrain expression of Hoxa1 at stage 26 and 30(Fig. 1G,H). Thus, the Hox PG1 genes are expressed in expansive and overlapping domains throughout early development, and all three of them are expressed in the hindbrain while it is being specified and patterned (Melton et al., 2004), and at a later stage in the pharyngeal arches.

To analyse the role of the Hox PG1 genes we used the morpholino knockdown approach. Two morpholinos were designed for each PG1 gene and the effect on hindbrain patterning was investigated. To have an internal control, injections were performed at the four-cell stage into the left-hand side (lhs) of the embryo only (Fig. 2). For all three genes, both morpholinos gave the same phenotype, albeit at different concentrations (within a range of 10-30 ng), confirming the specificity of these morpholinos. The knockdown of each of the PG1 genes led to a defect in hindbrain patterning. For all three of them, r3 and r5 (shown by Krox20 expression) were closer together on the injected side,indicating the reduction of r4, with the greatest reduction seen with the Hoxa1 morpholinos (Fig. 2B,C). The expression domain of Engrailed-2 however, a marker of the midbrain-hindbrain boundary(Hemmati-Brivanlou et al.,1991), was generally unaffected, although the width of this stripe often appeared to be smaller on the injected side.

For Hoxa1 and Hoxb1 this reflects the phenotypes observed in zebrafish and mice (Carpenter et al.,1993; Goddard et al.,1996; McClintock et al.,2002), but there is less information on the Hoxd1 loss function phenotype. Therefore we decided to further check the specificity of the Hoxd1 morpholino. To this end we tried to rescue the hindbrain phenotype with a Xenopus Hoxd1 mRNA construct which lacks the 5′ UTR and therefore is not recognised by any of the morpholinos. Co-injection (in the lhs) of Hoxd1 mRNA rescued the Hoxd1morpholino phenotype, as shown by Krox20 expression(Fig. 2K,L). Rhombomeres 3 and 4 were restored and in some cases were even larger than on the control side,as also seen in the simple Hoxd1 overexpression phenotype(Fig. 2J). This demonstrates that the Hoxd1 morpholino is specific and that in Xenopus laevis,Hoxd1 has a role in hindbrain patterning.

Having established the specificity of the morpholinos we wanted to examine the effect of knocking down all three PG1 genes simultaneously. The morpholinos were injected (on the lhs) either in double combinations, or in the triple combination to remove all Hox PG1 function(Fig. 3). In all the double morpholino combinations the Krox20 stripes are still present,although somewhat reduced and closer together(Fig. 3D,E,F). However, with the triple MO combination, Krox20 expression in the hindbrain is either severely reduced or completely absent, with only a vestige of expression in the neural crest region (Fig. 3C). This indicates that an area encompassing at least r3-r5 is affected when all three PG1 genes are knocked down. The same effect was noted when the morpholinos were injected into the whole embryo (data not shown).

We attempted to rescue this phenotype with the morpholino insensitive xHoxd1 expression construct (Fig. 4). When this was co-injected with the triple MO combination the majority of embryos changed from having no Krox20 expression, or only one faint narrow stripe, to having two stripes or one broad stripe spanning the r3-r5 region (Fig. 4E-G). This indicates that xHoxd1 can partially rescue the triple knockdown phenotype and bring back some degree of patterning to the r3-r5 region of the hindbrain. We also attempted to rescue the phenotype with Hoxa1 and Hoxb1 constructs, but as the overexpression of these genes leads to a loss of the r3 Krox20 stripe, interpretation of the rescue was difficult.

From these results, it appears that the phenotype of the triple PG1 MO embryos is not just an additive combination of the individual morpholino phenotypes, but is instead a synergistic effect. This indicates that there is a degree of redundancy between these genes in their hindbrain patterning function, and that any of the three genes can fulfil a basic hindbrain patterning role, even in the absence of the others. Therefore we concentrated on the triple PG1 MO combination for a more in depth analysis of Hox PG1 function.

To further analyse the effect on hindbrain patterning of losing all PG1 function we carried out a detailed marker analysis of embryos injected with all three PG1 morpholinos (Fig. 5). It is clear that although hindbrain patterning is severely affected, as shown by the loss of Krox20 stripes(Fig. 5G,H), the cells still adopt a neural fate, as the pan neural marker, Nrp1(Richter et al., 1990), is unaffected (Fig. 5A,B). In addition the most anterior domain of the embryo, expressing Otx2(Pannese et al., 1995) is not altered (Fig. 5C,D), and neither is the midbrain/hindbrain boundary, as shown by Engrailed-2expression (Fig. 3C). However,as we move posteriorly along the axis, a transformation of posterior hindbrain into anterior hindbrain becomes apparent. This is shown by the expansion of Gbx2. This gene has a stripe of expression at the boundary between the midbrain and hindbrain and in r1 (von Bubnoff et al., 1996), and on the side of the embryo injected with the PG1 morpholinos this stripe is expanded posteriorly(Fig. 5E,F). This posterior expansion was only partially rescued by co-injection of the Hoxd1mRNA (75% phenotype, reduced to 58%), indicating that at least one of the other Hox PG1 genes may be necessary for the restriction of Gbx2expression. More posterior hindbrain markers, such as Krox20 and certain Hox genes (see following section), are either severely reduced or lost completely with the triple MO combination. Later analysis of the hindbrain using a neural specific antibody (Jones and Woodland, 1989) and confocal imaging indicated that not only was the patterning of the hindbrain affected, but also its morphology. In the triple MO injected side there were no clear rhombomere boundaries and the hindbrain appeared to be shorter and thinner compared with the control side(Fig. 5I,J). The midbrain also looked slightly affected but this could be a secondary effect due to the shortened hindbrain. Thus PG1 function is necessary for the correct segmentation of the hindbrain, as well as being involved in its patterning.

Another interesting aspect of the triple PG1 knockdown is the loss of segmented MyoD expression (Fig. 5H), indicating that the somites are also affected. This effect is specific and the mechanisms underlying it are currently under investigation.

To investigate whether the hindbrain defects that we observe could be due to a disruption in the `Hox code', we analysed the expression of several Hox genes from different paralogous groups in embryos injected with the triple PG1 MO combination (Fig. 6). It is known that Hox-Hox cross-regulation occurs(Hooiveld et al., 1999), and as the PG1 genes are the first to be expressed in the embryo, and are expressed in a wide domain, they could play a role in the regulation of the other Hox genes. When we examined the expression of the PG1 genes themselves, Hoxb1 and Hoxa1 expression was downregulated (shown for Hoxb1, Fig. 6A,B), but the expression of Hoxd1 was more variable, being reduced in only one third of embryos (n=24) and at other times being unaffected. This indicates that auto- and cross-regulation of these genes play a role in the maintenance of expression, but that external factors are also involved,particularly for Hoxd1.

When the other Hox genes were analysed it became apparent that the function of the PG1 genes is necessary for the establishment of the `Hox code'. Both PG2 genes, Hoxa2 and Hoxb2, normally expressed up to the anterior boundary of r2 and r3, respectively(Pasqualetti et al., 2000)were almost completely absent (shown for Hoxa2, Fig. 6C,D). For PG3 the effect was less uniform; Hoxa3 and b3 [expressed up to the r4/r5 boundary and in the neural crest derived cells of the 3rd pharyngeal arch(Godsave et al., 1994)] were severely reduced, with only the most posterior expression remaining (shown for Hoxa3, Fig. 6E,F). The effect on Hoxd3 was less severe; its lateral stripe of expression was lost and its hindbrain and neural tube expression was fainter and shifted posteriorly, but slightly expanded laterally(Fig. 6G,H). This could be due to the inability of the neural crest to migrate (see following section) and thus the faint, expanded domain could be expression in the premigratory neural crest cells. Hoxd4 is expressed up to the r6/r7 boundary and when PG1 function is lost the anterior expression domain is shifted posteriorly and again the neural crest stripe is lost (Fig. 6I,J). Thus, the expression of Hox genes in r2-r7 is either reduced or completely absent.

We also investigated several more posterior Hox genes, which have their anterior expression boundaries in the spinal cord. For all of these genes(Hoxa5, Hoxc6 and Hoxb9) the most anterior, spinal cord expression was lost in the triple PG1 knockdown, but the diffuse mesodermal expression (in particular of Hoxa5 and Hoxc6), and more posterior spinal cord expression, was unaffected(Fig. 6K-P).

To establish the specificity of this effect of the morpholinos, and to determine the degree to which reintroducing just one of the Hox PG1 genes would restore the Hox code, we utilised the Hoxd1morpholino-insensitive mRNA (Fig. 7). When this was injected alone, it induced anterior expansion of the expression of Hoxb2 (Fig. 7A), but the other Hox genes analysed (Hoxa3, Hoxd4 and Hoxc6) were unaffected. When Hoxd1 mRNA was co-injected with the triple PG1 morpholino combination, however, it rescued the expression patterns of all of these Hox genes (Fig. 7C,F,I,L). This indicates that there may be a degree of redundancy between the PG1 genes and this is borne out by preliminary analyses of injections of the single morpholinos (Fig. 8). These showed that Hoxa2 and Hoxb2 expression was still present in all of the single knockdowns (shown for Hoxb2, Fig. 8A-D), although Hoxb2 was slightly reduced with the Hoxa1 and Hoxd1morpholinos. Hoxa3 hindbrain expression however was reduced with the Hoxa1 morpholino, and not affected by the other individual morpholinos despite being rescued in the triple PG1 knockdown by the Hoxd1 mRNA (Fig. 8E-H, Fig. 7D-F). Hoxd4 was affected by the Hoxd1 morpholino, which appeared to give a similar phenotype to the triple combination (Fig. 8I-L). Thus different Hox genes appear to be dependant on different Hox PG1 genes, or a combination thereof, but these interactions will require further investigation before they are fully elucidated. However, these data do indicate that PG1 function in general is necessary for the establishment or maintenance of the expression patterns of both anterior and posterior Hox genes. For the more posterior genes this requirement is restricted to the anterior, ectodermal expression domains, and the expression in the mesoderm and posterior ectoderm is unaffected.

As Hox PG1 genes are expressed in the cranial neural crest, and PG1 function is clearly also necessary for the expression of certain other Hox genes in this region, we decided to investigate the effects of PG1 knockdown on neural crest and its derivatives. When we examined the expression of Xslug, which is expressed in all premigratory and migrating neural crest cells (Linker et al.,2000), we observed that on the triple MO injected side, the Xslug domain in the hindbrain region appeared to be upregulated compared with the control side (Fig. 9C). The expression, however, did not extend laterally but was restricted to an area close to the hindbrain. When the Hoxd1morpholino insensitive RNA was co-injected with the triple MO combination, the normal expression pattern was restored, with an additional expansion anteriorly (Fig. 9D). This is in keeping with the Hoxd1 overexpression scenario, where the Xslug domain is expanded anteriorly(Fig. 9B). Xsnail,another neural crest marker, was also restricted to the area close to the neural tube but in this case the expression level was not significantly altered (Fig. 9E,F). These changes in expression could indicate a role for PG1 genes in neural crest cell migration, as in the knockdown situation the cells seem to remain close to the hindbrain, whereas when Hoxd1 is overexpressed the Xslugpositive cells are more widespread. It has been shown that Xslug, but not Xsnail, is downregulated when the neural crest cells reach their target (Linker et al., 2000). Therefore the apparent upregulation of Xslug compared with the control could be because the cells stay close to the hindbrain and thus Xslug expression is not downregulated. In later embryos expression of dll4 in the pharyngeal arches(Papalopulu and Kintner, 1993)is severely reduced, whilst the expression of this gene in the forebrain and eye is still present (Fig. 9G,H). This indicates that PG1 gene function is vital for the development of the pharyngeal arches and this is confirmed by later analysis of the craniofacial structures using alcian blue staining. It can clearly be seen that, on the side injected with the triple PG1 MO combination, the gill cartilages, derived from the 3rd and 4th pharyngeal arches (the branchial arches) are completely missing (Fig. 9K). The ceratohyal, derived from the 2nd pharyngeal arch is still present, but reduced in size. Meckel's cartilage, derived from the 1st pharyngeal arch, is less affected but also appears to be slightly reduced. When the morpholino-insensitive Hoxd1 RNA is co-injected, the gill cartilages are restored, but remain smaller than in the control side(Fig. 9L).

Thus, in the absence of Hox PG1 function, cranial neural crest cells are specified but they cannot migrate away from the hindbrain. Subsequent development of the pharyngeal arches and their derivatives is severely affected, with the complete loss of the gill cartilages.

The complexity of the Hox gene family, and its members' redundancy and cross-reactivity, make it difficult to establish precisely what the function of each individual Hox gene is. However, there is increasing evidence that paralogous groups of Hox genes have shared functions. Therefore, rather than investigate individual Hox genes, we have taken a more global approach and studied the role of a complete paralogous group. The morpholino knockdown approach allowed us to block simultaneously the translation of the three Xenopus laevis PG1 Hox genes, Hoxa1, Hoxb1 and Hoxd1, and assess the roles of this paralogous group in the development of the embryo.

The literature abounds with information about the effects on hindbrain patterning of single and double knockouts of Hox PG1 members. The data from mouse indicate that Hoxa1 and b1, but not Hoxd1,are vital for the correct patterning and formation of the hindbrain and its derivatives. In this study we have shown that, in Xenopus laevis, all three Hox PG1 genes are expressed in the presumptive hindbrain region at a time when they could influence hindbrain patterning. This is in contradiction to the situation in mouse, where Hoxd1 is only expressed in the mesoderm and neural crest (Frohman and Martin, 1992). However, the observed Hoxd1 expression in the presumptive hindbrain region in Xenopus is very low, and is downregulated earlier than the other two paralogous genes, and it is therefore possible that a comparable low level expression in mouse was below detection limits.

When we knock down the function of all three PG1 genes we observe a more severe effect than in either the single or double knockdowns. Fig. 10 shows a schematic for the effect on the hindbrain of the stepwise elimination of Hox PG1 genes, with increasingly more severe phenotypes occurring when there is less PG1 function present. In Xenopus, the hindbrain expression of the r3/r5 marker, Krox20, is completely lost in the triple knockdown, whereas its expression domain is merely altered (indicating a reduction in the r3-r5 region) when at least one PG1 gene remains functional. Preliminary results indicated that the triple knockdown phenotype is a synergistic, rather than an additive effect, which implies a degree of redundancy between the PG1 genes and, thus, we concentrated on the triple knockdown to elucidate the basic PG1 group function. We cannot therefore rule out effects being due to the double morpholino combinations. However, defects in the hindbrain of Xenopusembryos completely deficient in PG1 function were more extensive than any previously reported for single or double PG1 knockouts in other organisms. Expression of markers for r2-r7 was downregulated in the triple knockdown,whereas the expression of an r1 marker, Gbx-2 was expanded posteriorly, indicating a possible transformation of more posterior hindbrain to an r1 type identity. Later morphological analysis showed that the hindbrain is reduced in size and segmentation is perturbed. This phenotype is actually similar to the situation in zebrafish when the functions of Hox cofactors from the Meis and Pbx families are compromised(Fig. 10). In Pbx4(lazarus) mutants that have been injected with a Pbx2morpholino (to achieve a total block of early Pbx function) the hindbrain is not segmented and r2-r7 acquire an r1-like identity, referred to as the hindbrain ground state (Waskiewicz et al.,2002). Likewise, when the function of Meis is blocked,using two dominant-negative constructs, a similar, non-segmented hindbrain is produced (Choe and Sagerstrom,2004). The similarity of these phenotypes to the triple PG1 MO phenotype suggests that they could, at least partially, be due to the blocking of PG1 gene function. A recent study in Xenopus showed that hindbrain Krox20 stripes were eliminated when either a Meis morpholino or a dominant-negative Hoxd1 RNA construct (which may also block Hoxa1 and Hoxb1 function and, thus, be similar to the triple MO situation) were injected into the embryo(Dibner et al., 2004). It has also been noted in one Hoxa1 knockout study in mouse that segmentation is completely blocked, but again all three of the PG1 genes could be affected, as the truncated Hoxa1 splice variant still present could perhaps act as a dominant-negative(Chisaka et al., 1992).

The PG1 function in hindbrain could be mediated via other, downstream, Hox genes. For example, when both Hox PG2 genes are knocked out in mouse the rhombomere boundaries in the r2/r3 region are absent(Davenne et al., 1999). Thus the effect that we see with the triple PG1 knockdown on segmentation in this anterior region of the hindbrain (where PG1 genes are not expressed) could be due to the downregulation of Hoxa2 and Hoxb2.

Numerous interactions between various Hox genes have been shown to exist in the embryo (Deschamps et al.,1999; Hooiveld et al.,1999; Maconochie et al.,1997; Melton et al.,2004). Indeed one aspect of colinearity could be the sequential activation of anterior Hox genes that are, in turn, responsible for the activation of the more posterior ones(Hooiveld et al., 1999). Although the majority of the literature in mouse does not support this model,this could be due to a high degree of redundancy between different Hox genes masking the effect in single or double knockout mice. As the PG1 genes are the first to be expressed in the embryo we investigated the possibility that they could initiate a cascade of Hox expression, which would lead to the established `Hox code'. Although PG1 function is clearly not a prerequisite per se for Hox gene expression we did uncover a reliance on PG1 function for the correct expression of Hox genes from paralogous groups 2-9. This widespread effect on posterior Hox genes is more severe than that observed in the single or double PG1 knockouts in mouse(Dolle et al., 1993; Maconochie et al., 1997; Rossel and Capecchi, 1999). However, as the triple PG1 knockdown has never previously been investigated,it is possible that the situation in mouse is similar to that in Xenopus, with a high degree of functional redundancy between the genes.

From our data we cannot distinguish when the PG1 genes are required for the expression of the Hox genes examined, although it is likely to be an early input, as the later domains of Hoxa2 and b2, at least,extend beyond the Hox PG1 expression domains. It is clear, however, that without Hox PG1 function the `Hox code' is severely compromised. Thus, via the control of the `Hox code', the PG1 genes can eventually exert their influence outside of their own expression domains. This has already been shown for Hoxa1, the knockout of which has defects in the r3 domain, which is anterior to the normal Hoxa1 expression domain(Helmbacher et al., 1998). This mechanism would account for the extent of the defects that we observe in the triple PG1 knockdown. It could also be due to a more anterior expression domain early in development of one or all of the Hox genes, which could be the case for Hoxb1, as its early expression is very close to the Otx-2 domain. Alternatively, these effects could be due to a non cell-autonomous effect. Several signalling pathways influence patterning of the hindbrain, such as the FGF and retinoic acid pathway(Roy and Sagerstrom, 2004; Walshe et al., 2002), and it is feasible that the extent of the phenotypes that we observe may be partly due to a disruption in these signalling pathways. It has also been shown in vitro that certain Hox proteins are able to cross the cell membrane(Amsellem et al., 2003; Chatelin et al., 1996) and thus it is possible that the Hox proteins themselves act non cell-autonomously to pattern the hindbrain.

A link between the Hox PG1 genes and craniofacial development has previously been demonstrated, with Hoxa1 and -b1 double mutants lacking the 2nd pharyngeal arch and its derivatives as well as some derivatives of the 1st pharyngeal arch(Gavalas et al., 1998; Rossel and Capecchi, 1999). Here we show that in the absence of any PG1 gene function, cranial neural crest cells are still specified but they appear to be unable to migrate away from the neural tube. Conversely, when Hoxd1 is overexpressed,migration seems to be enhanced, as seen by the widespread Xslugdomain.

It is interesting to note that in the Hoxa1/b1 double knockout in mouse, neural crest cells fail to delaminate from the presumptive r4 territory, although they delaminate normally from the other rhombomeres(Gavalas et al., 2001). This failure could be more widespread in the triple PG1 knockdown, leading to a wholesale block of cranial neural crest cell migration. This could be due to the effect on the Hox code, as many of the Hox genes downregulated are also expressed in the neural crest, or are known to be involved in their patterning(Trainor and Krumlauf,2001).

Major defects in the craniofacial structures are later seen in embryos lacking PG1 function, which are probably due to the inability of the cranial neural crest cells to migrate. Derivatives of all the pharyngeal arches are affected and most strikingly, the gill cartilages, derived from the 3rd and 4th pharyngeal arches, are completely missing. Effects such as this have not previously been seen in PG1 mutants, where generally only the 1st and 2nd pharyngeal arches are affected (Gavalas et al., 1998). Thus the complete abrogation of PG1 function has identified additional regions of the embryo which are ultimately dependant upon these genes for their formation.

In conclusion, we have knocked down the function of the complete Hox Paralogous Group 1 and illustrated that these genes have a degree of functional redundancy in their role of patterning the Xenopushindbrain. We have demonstrated that PG1 function is essential for the correct establishment of the `Hox code'. In the absence of PG1 function, and perhaps as a consequence of `Hox code' disruption, hindbrain segmentation is perturbed and r2-r7 are mis-specified. In addition we have identified a novel requirement for Hox PG1 function in the migration of the cranial neural crest and the development of the pharyngeal arches.

We thank Jacqueline Deschamps for helpful discussion of the manuscript. We also thank all the members of the Durston Group for valuable advice and technical assistance. We thank the European Union for support. JNP was funded by the Fundação para a Ciência e a Tecnologia, Portugal(SFRH/BD/2794/2000). WMRA was funded by the EU Life Sciences programme. C.M. and A.J.D. thank the EU HPRN, Quality of Life and EU Life Sciences, Genomics and Biotechnology for Health programmes.