Hox Genes and Regional Patterning of the Vertebrate Body Plan

Abstract

Several decades have passed since the discovery of Hox genes in the fruit fly Drosophila melanogaster. Their unique ability to regulate morphologies along the anteroposterior (AP) axis (Lewis, 1978) earned them well-deserved attention as important regulators of embryonic development. Phenotypes due to loss- and gain-of-function mutations in mouse Hox genes have revealed that the spatiotemporally controlled expression of these genes is critical for the correct morphogenesis of embryonic axial structures. Here, we review recent novel insight into the modalities of Hox protein function in imparting specific identity to anatomical regions of the vertebral column, and in controlling the emergence of these tissues concomitantly with providing them with axial identity. The control of these functions must have been intimately linked to the shaping of the body plan during evolution.

Anterior to posterior expression and function of individual Hox genes

Hox genes encode transcriptional regulatory proteins that control axial patterning in all bilaterians (Garcia-Fernandez, 2005). While these genes have been organized in a cluster since the ancestral situation (Duboule, 2007), genome duplications during evolution have given rise to four Hox clusters in most vertebrates, HoxA, B, C and D, the zebrafish possessing 7 clusters as a result of an additional genome duplication and subsequent loss of one of the clusters (Woltering and Durston, 2006). At corresponding positions within the four clusters, therefore, are genes with particular sequence similarity, called paralogous genes. The 39 mammalian Hox gene family members are thus subdivided into 13 paralogous groups (PGs).

In general, Hox genes within a cluster are expressed from 3' to 5', with the earliest genes expressed in the primitive streak at pre-somite stages, and more 5' genes expressed in the posterior part of the growing embryo at progressively later time points (Deschamps and van Nes, 2005; Dressler and Gruss, 1989; Duboule and Dolle, 1989; Gaunt, 1991; Gaunt and Strachan, 1996; Graham et al., 1989; Iimura and Pourquie, 2006; Izpisua-Belmonte et al., 1991). The spatio-temporal features of Hox gene expression onset are thus to a large extent coupled to the growth and elongation of the embryonic axis (Duboule, 1994), resulting in posteriorly overlapping Hox expression domains with spatially staggered anterior boundaries of expression. 3' genes (Hox1 through Hox4) display rostral expression limits in the hindbrain region of the embryo and more 5' genes exhibit increasingly posteriorly restricted expression domains along the AP axis. The most 5' genes are activated in the tail bud around mid-gestation.

Vertebrate Hox genes confer axial positional information to emerging embryonic tissues from the three germ layers. Loss-of-function mutations in individual mouse Hox genes have been found to alter the identity of tissues located within the expression domain of the genes, most often in the rostral part of that domain. Historically, two models were proposed to account for the Hox patterning along the mouse AP axis. Kessel and Gruss postulated the existence of a Hox code whereby varying combination of Hox genes functioning at any given axial level resulted in the specific morphologies along the AP axis (Kessel and Gruss, 1990). On the other hand, a mechanism of posterior prevalence was proposed to account for the observation that despite the broad, overlapping patterns of axial Hox expression, single loss-of-function Hox mutants only displayed phenotypes in the most anterior regions of their expression domains (Duboule and Morata, 1994). This concept, based on a mechanism of interference called “phenotypic suppression” in Drosophila, proposed that more posterior Hox genes are functionally dominant over anterior genes. While a thorough evaluation of these models is beyond the scope of the review, none of the models can fully account for all phenotypes produced by gain-of-function of individual Hox genes, or by loss-of-function of whole Hox paralogous groups of genes. Clearly, only a deeper molecular understanding of the mechanism of Hox gene activity will provide the keys to better account for and predict Hox patterning along the AP axis.

Patterning alterations due to Hox mutations have been reported to affect neural tissues, neural crest, endodermal or mesodermal derivatives, depending on the gene(s) mutated (Krumlauf, 1993; Mallo et al., 2009; Manley and Capecchi, 1998; Trainor and Krumlauf, 2000; Wellik, 2009; Wellik et al., 2002). However, the most abundantly studied phenotypic modifications in single Hox mutants have been changes in axial identity of particular derivatives of the segmented paraxial mesoderm (Iimura et al., 2009; Wellik, 2009). Somites provide the metameric pattern that prefigures the axial skeleton (reviewed in (Hirsinger et al., 2000)). During development, somites differentiate progressively into distinct compartments, each of them closely related to the production of specific tissues. The dermo-myotome of each somite gives rise to the dermal layer of the skin and the muscles of the body and limbs. The sclerotomal region of the somites differentiates into the axial skeleton, and the tendons that connect muscle and skeletal tissues derive from the somitic syndetome (Brent and Tabin, 2002; Christ et al., 2007).

All tetrapod vertebrates possess a number of occipital somites generating the base of the skull, followed by cervical, thoracic, lumbar, sacral and caudal somites. Over the 350 million year tetrapod vertebrate history, the total number of segments and their distribution among the different regions has diverged among species, but common morphological traits for each region have been preserved. The vertebra that forms an articulation with the skull, the atlas, is the first of the elements of the cervical region that constitutes the skeleton of the neck. Just posterior to this area, rib-bearing thoracic vertebrae form the rib cage. A variable number of these vertebrae grow ribs around the body that join at the sternum, while the ribs associated to the rest of the thoracic vertebrae remain as floating ribs. Immediately posterior to the ribcage we find the lumbar vertebrae, which in land vertebrates constitute the load bearing skeleton and are generally the largest and densest of the vertebrae. They are followed by the sacral vertebrae, which grow lateral protrusions that fuse and are the site of pelvic attachment. The caudal vertebrae constitute the most posterior part of the axial skeleton, and their number varies widely between tetrapods, from the long, prehensile tail of some monkeys that have more than 30 caudal vertebrae (Schultz and Straus, 1945) to the set of 3 to 5 fused vertebrae of the human coccyx. The typical axial formula of mice consists of 7 cervical, 13 thoracic (with 7 sternal and 6 floating ribs), 5 or 6 lumbar (depending on the genetic background), 4 sacral and approximately 28 caudal vertebrae. Squamate reptiles, such as snakes, exhibit a much-diverged vertebral formula with an enormously elongated thorax (sometimes containing more than 200 thoracic vertebrae), and extremely reduced cervical, lumbo-sacral and caudal regions. A larger total number of somites in these animals was suggested to result from a faster segmentation process compared to the developmental rate (Gomez et al., 2008).

Inactivation of single Hox genes often resulted in transformations in the identity of specific vertebral elements. Those affecting anatomical boundaries have been the most commonly studied, maybe because they usually contain the clearest morphological changes. From these studies, it was clear that several adjacent Hox genes were often found to contribute to the identity of vertebrae at particular axial level (Mallo et al., 2009; Wellik, 2007; Wellik, 2009). Analysis of compound Hox mutants, including several members of the same paralogous group, invariably resulted in stronger phenotypes, which became extreme when the whole paralogous group was simultaneously inactivated (Horan et al., 1995; McIntyre et al., 2007; Wellik and Capecchi, 2003). These studies confirmed the functional redundancy among members of a given paralogous group, an idea that had been already suggested on the basis of similarities in their expression domains along the AP axis (Burke, 2000; Burke et al., 1995; Gaunt et al., 1989; Gruss and Kessel, 1991; Kessel and Gruss, 1990; McGinnis and Krumlauf, 1992; McIntyre et al., 2007). These studies on mutants of complete Hox paralogous groups (HoxPGs) also revealed an additional level of complexity in the function of Hox genes in patterning the vertebrate body plan. They indicated that particular HoxPGs control patterning of full anatomical axial domains, and they associated specific functions to individual paralogous groups, although the mechanisms by which this patterning occurs is only starting to be understood.

The control of regional patterns in the axial skeleton by Hox genes

Hox Paralogous 10 genes and the lumbar column

Recent data suggest that Hox paralogous groups control the specification of characteristic morphologies along regions of the vertebral column. This concept has shed new light on the patterning mechanisms by Hox genes. It has been shown that discrete paralogous groups of Hox genes play dominant roles in the morphogenesis of specific anatomical regions in the vertebral column. The first data that demonstrated that Hox activity controls regional patterning along the AP axis was provided by the analysis of compound mutants of HoxPG10 (Wellik and Capecchi, 2003). Simultaneous inactivation of all three HoxPG10 genes resulted in mice in which the prospective lumbosacral region had acquired thoracic-like characteristics, as these mutant vertebrae displayed associated ribs all along the thoraco-lumbar region (Fig. 1). While these phenotypes could also be classified as identity changes, they differed from previously described phenotypes for single or compound Hox mutants in two ways: they affected a larger number of segments, and all the segments affected belonged to a distinct anatomical region. The rib-suppressing activity of the HoxPG10 genes was further demonstrated by ectopic expression of one of the Hox10 genes in the paraxial mesoderm at axial levels that included the thoracic segments, resulting in completely ribless animals (Carapuco et al., 2005)(Fig. 1).

Fig. 1.

Regional patterning of the axial skeleton by Hox paralog groups 6 and 10. Skeletal preparations showing the effects of global inactivation (LOF) or of ectopic activation (GOF) of genes in the paralogous Hox groups 6 (red labels) and 10 (green labels). HoxPG10 genes have rib blocking activity as shown by the ectopic ribs observed in the lumbar area of mice with complete inactivation of this paralog group (D) and by the complete absence of ribs (green asterisk) after precocious activation of Hoxa10 in the presomitic mesoderm (C). HoxPG6 genes are able to induce ectopic ribs at cervical (R* in F) and lumbar (G) levels. Complete inactivation of HoxPG6 genes (E) results in smaller rib cages and loss of specific ribs (first rib in this specimen is attached to the second thoracic vertebra, T2), indicating that other genes must also cooperate in the rib-inducing process. A schematic representation of the Hox clusters is shown for reference. Skeletons of wild type embryos are shown in A and B. The lumbar and sacral domains, the position of the first thoracic (T1), last thoracic (T13), and first lumbar (yellow asterisks) vertebrae is indicated for reference.

A recent paper on the expression patterns of Hox genes in snakes and caecilians raised a question regarding the role of HoxPG10 genes in the inhibition of rib development in vertebrates (Woltering et al., 2009). In this report it was shown that expression of Hoxc10 in both “snake like” animals extended well into the rib-forming domain of the paraxial mesoderm, which is at odds with the genetic data obtained in mice (Carapuco et al., 2005; Wellik and Capecchi, 2003). This unexpected expression was also shown in a recent report for Hoxa10 but not for Hoxd10 in squamate reptiles (Di-Poi et al., 2010). A possible molecular explanation to this enigma was proposed on the basis of the strong relaxation in the Hoxa10 and Hoxc10 coding sequences of snakes, which was not shared by the Hoxd10 protein (Di-Poi et al., 2010), suggesting that the former two proteins might have lost their rib-suppressing activity and contributed to an extended rib cage. Functional analysis of HoxPG10 proteins of squamate reptiles and other vertebrates will surely help to understand the role that these Hox genes played in the evolution of the vertebrate body plan.

HoxPG11 and the sacrum and tail

Formation of the sacrum absolutely depends on HoxPG11 genes, as demonstrated by the complete absence of this structure in mutants missing HoxPG11 (Wellik and Capecchi, 2003). In these animals the morphology of the vertebrae normally forming the sacrum was identical to that of the anterior lumbar vertebrae. The ability of HoxPG11 to generate sacral characteristics has also been shown in overexpression experiments. In particular, transgenic embryos expressing Hoxa11 in the presomitic mesoderm (PSM) throughout development, exhibit fusions between adjacent vertebrae that could be interpreted as a “sacralization” phenotype (Carapuco et al., 2005). However, proper formation of the sacrum also requires the activity other Hox genes, most particularly those of the HoxPG10. This conclusion stems from the observation that in the absence of HoxPG10 genes sacral vertebrae still fuse laterally but modify their morphology to appear as rib-like appendages (Wellik and Capecchi, 2003). This suggests that formation of the sacrum requires both inhibition of the rib-forming activity provided by HoxPG10 genes and an additional activity of HoxPG11 genes to promote generation and fusion of lateral outgrowths from adjacent vertebrae to form the sacral wing. It is noteworthy that HoxPG11 genes are also required for proper patterning the first several caudal vertebrae (Wellik and Capecchi, 2003). What makes HoxPG11 genes produce sacral or anterior caudal vertebrae at different axial levels is still unknown. Transgenic experiments suggest that specific characteristics of the tissues in which the HoxPG11 genes operate might play a role in this process, as somitic expression of Hoxa11 resulted in caudal-like phenotypes instead of the “sacralization” resulting from Hoxa11 expression in the PSM (Carapuco et al., 2005). Whether this is the case during physiological development of these areas, and which molecular mechanism underlies this specificity, remains to be determined.

HoxPG5-9 and the ribcage

HoxPG9 genes also have a regional patterning function. Genetic experiments indicate that the production of sternal versus floating ribs is under the control of HoxPG9 (McIntyre et al., 2007). In particular, the activity of these genes seems to be required to produce floating ribs as the complete inactivation of this paralogous group resulted in rib cages with 13 –14 ribs attached to the sternum instead of the typical 7. Expression patterns of genes in this paralogous group suggest that gene activity is likely to be required in the lateral plate mesoderm rather than in the somites (McIntyre et al., 2007). It has been shown that development of the distal rib requires migration of somitic mesoderm into the somatopleura (Sudo et al., 2001), possibly requiring the interaction of rib primordia with sternal precursors. The control of this migration seems to depend on a variety of signals from the surface ectoderm and the somatopleura itself (Aoyama et al., 2005; Sudo et al., 2001). Thus, it is possible that HoxPG9 modulates the response of the lateral plate mesoderm to the migration signals or the production of the signals themselves, so that penetration of rib precursors into the somatopleura is properly organized at the different rostro-caudal levels of the rib cage.

Regarding rib-inducing activity, no single Hox paralogous group inactivation so far resulted in rib loss in the entire rib cage, suggesting that more than one Hox paralogous group is involved in promoting rib formation (or that the process is controlled by as yet undiscovered genes). However, defects in the rib cage have been noted in mutants for Hox genes from HoxPG5-9. More recently, it has been shown that transgenic mice expressing a HoxPG6 gene in the PSM throughout axial extension develop ribs associated to vertebrae in the prospective neck and lumbar areas, in addition to those of the thoracic region (Vinagre et al., 2010) (Fig. 1), indicating that Hox genes in this paralogous group are able to confer a thoracic character to vertebral elements anterior and posterior to the thorax.

The possibility that HoxPG6 genes are involved in the specification of the boundary between cervical and thoracic regions was initially suggested by studies showing that their anterior limits of expression in the paraxial mesoderm correlated with the cervical to thoracic transition rather than to absolute somite number in vertebrate species with a different number of cervical vertebrae (Burke et al., 1995; Gaunt, 1994). In the mouse, HoxPG6 rostral expression limit is at the level of somite 13, which correspond to the cervico-thoracic transition at vertebra 8, while the anterior level of Hoxc6 expression in chicken was mapped to somite 20, fitting the rostral border of the thoracic region at vertebra 15 (Burke et al., 1995). The phenotype of the transgenic embryos expressing Hoxb6 in the mouse PSM at all axial levels seems to confirm that HoxPG6 can indeed promote formation of ribs from the developing somites and could thus be an integral part of the genetic network controlling formation of the rib cage (Vinagre et al., 2010). However. inactivation of HoxPG6 did not produce mice without a rib cage, although the mutant rib cages were smaller in size than those in wild type embryos and displayed abnormal rib phenotypes, including the absence of the first rib and distal fusions of ribs associated to thoracic vertebrae 2 to 4 (McIntyre et al., 2007) (Fig. 1). This indicates that additional factors must operate in this process. HoxPG5 genes are among the candidates for an additional role in rib induction, since their expression domains in the paraxial mesoderm are similar to those of HoxPG6 genes (Burke et al., 1995), and their regional inactivation also has phenotypic impact on the anterior ribcage (McIntyre et al., 2007). Proof for a redundant function for HoxPG5 and PG6 in rib induction processes would require simultaneous complete inactivation of both paralogous groups, which has not been performed to date.

Hox genes and the patterning of the neck

Much less is known about the mechanisms of regional patterning by Hox genes leading to proper morphogenesis of the cervical vertebrae. Loss-of-function experiments of genes of the HoxPG3, PG4 and PG5 have demonstrated a role for these genes in establishing morphologies in the cervical skeleton. Horan and colleagues showed that the HoxPG4 genes are critical for establishing the appropriate morphology of the cervical skeleton (Horan et al., 1995). With loss of three of the four HoxPG4 genes, the vertebrae in the cervical region were transformed into morphologies typical of the first (atlas) and second (axis) cervical vertebrae. Loss of HoxPG5 resulted in defects in both the cervical and thoracic skeleton, including a transformation of cervical vertebrae 3 to 7 towards axis morphology (McIntyre et al., 2007), and Hoxa3/Hoxd3 control the generation of the atlas (Condie and Capecchi, 1994). However, complete transformation of the cervical region into a different vertebral domain has not yet been reported for any combination of loss-of-function Hox mutants.

In summary, it is clear that Hox genes functionally define regional anatomical domains along the axial skeleton. It is likely that Hox function in axial patterning results from an inherent ability to promote the generation of different structures from tissue progenitors, and that it is the combination of inputs from the different HoxPG genes that leads to the properly regionalized vertebral column.

Molecular mechanisms of Hox gene control of rib development

While the involvement of Hox genes in axial development has been extensively documented, relatively little is known regarding the mechanisms underlying their activity. A recent report has shed light on how HoxPG6 and PG10 modulate somite differentiation to produce rib-bearing and rib-less areas in the vertebral column (Vinagre et al., 2010). Although ribs derive mostly from the sclerotome (Burke and Nowicki, 2003; Huang et al., 2000), this compartment may not to be the primary target of Hox genes in their control of rib development. Instead, the activity of these Hox genes seems to be mediated to a large extent by their ability to modulate expression of Myf5 and Myf6 (previously known as MRF4) in a specific area of the myotome, the hypaxial (lateral) compartment (Vinagre et al., 2010). In particular, HoxPG6 genes promote this expression in interlimb somites, which are those generating the thoracic region, and HoxPG10 blocks it in somites adjacent to the hindlimb bud, which are those generating lumbosacral vertebrae. Hypaxial modulation of Myf5 and Myf6 expression seemed to be also responsible for the “all-rib” or completely rib-less phenotypes resulting from ectopic activation of Hoxb6 and Hoxa10, respectively. The Hox patterning information gathered in the hypaxial myotome by Myf5/Myf6 seems to be transmitted to the adjacent sclerotome by PDGF and FGF signals to promote skeletogenesis. This paracrine mechanism is consistent with the correlated myotomal expression of Pdgfa and Fgf4 with that of the Myf5/Myf6 genes and with genetic data showing the involvement of the FGF and PDGF signaling pathways in rib formation (Grass et al., 1996; Huang et al., 2003; Soriano, 1997; Tallquist et al., 2000).

Ribless phenotypes closely resembling those of the transgenics ectopically expressing Hoxa10 resulted from mutations in the Myf5 and Myf6 genes, which seem to have redundant roles in rib formation (Braun and Arnold, 1995; Braun et al., 1992; Kassar-Duchossoy et al., 2004; Patapoutian et al., 1995; Tajbakhsh et al., 1996; Yoon et al., 1997; Zhang et al., 1995). Myf6 expression using an hypaxial promoter proved quite active in rescuing the Hoxa10-mediated ribless phenotype (Vinagre et al., 2010), revealing that regulation of Myf5/Myf6 expression accounts for the rib-modulating Hox activity to a very large extent. This indicates that, while the involvement of additional factors cannot be ruled out, explaining Hox activity to control rib-containing versus rib-less areas in the skeleton mostly requires understanding how Hox genes regulate expression of Myf5 and Myf6 in the hypaxial myotome. Some evidence indicates that Hox proteins bind to an enhancer involved in the control of hypaxial expression of Myf5 (Buchberger et al., 2007) (and probably Myf6), both in vivo and in vitro (Vinagre et al., 2010), suggesting that the control of Myf expression by Hox proteins may be direct. Confoundingly, there is an apparent spatial and temporal gap between when and where Hox gene activity is required for vertebral patterning, and the effects on Myf5/Myf6 transcription. In transgenic experiments, the rib-inducing or repressing activities of HoxPG6 and PG10 were only observed when these genes were ectopically expressed using an enhancer active in the presomitic mesoderm, but not using a driver that was activated only later in the somites (Carapuco et al., 2005). However, the regulated Myf5/Myf6 expression is only observed at later stages, in the developing somites. Whether this results from the carry over of Hox proteins expressed in the PSM into the developing somites or whether modulation of the Myf enhancer requires sequential input from Hox proteins in the PSM and later from additional factors in the developing somite remains to be determined.

Regardless of the mechanism that eventually explains the spatial offset and temporal gap in this Hox-mediated control, it is clear that Hox genes must interact with other factors to control Myf5/Myf6 expression (Vinagre et al., 2010). Pax3 and Six1/Six4 are prime candidates for a cooperative role because they are also required for the activity of the hypaxial Myf5 enhancer bound by Hox gene products (Bajard et al., 2006; Giordani et al., 2007) and abnormal rib phenotypes have been described in mutant mice for these genes (Grifone et al., 2005; Henderson et al., 1999). Interestingly, specific functional interactions between members of the Hox and Pax families have been described in other apparently unrelated processes in the developing embryo (Gong et al., 2007; Wellik et al., 2002; Yallowitz et al., 2009), indicating that interactions between these genes might belong to general developmental mechanisms. It should be noted that the interaction between Hox proteins with Pax3 and Six1/4 to modulate Myf5/Myf6 expression might follow some non-conventional mechanism because the latter are not present in the PSM, which is where Hox activity is required.

Hox control of rib formation and evolution of the body plan

The control of rib development through regulated Myf5/Myf6 expression suggests a possible mechanism for coordinated evolution of the vertebrate body plan under the control of Hox genes, which would derive from a dual function of these Myf factors. While Myf5 and Myf6 seem to have a non-myogenic activity responsible for rib inducing processes, they still exert their classical myogenic role in these tissues (Pownall et al., 2002). Wherever ribs are formed adjacent to the Myf5/Myf6-expressing hypaxial myotome, the corresponding set of intercostal muscles will be produced by the activity of the same genes in the corresponding myotome. Indeed, cell-tracing experiments have indicated that precursors of ribs and intercostal muscles are intimately associated in the developing somite (Evans, 2003). In addition, muscles with intercostal characteristics were observed associated to the ectopic ribs that developed upon Hoxb6 ectopic expression, indicating that this mechanism can operate efficiently in vivo (Vinagre et al., 2010). Interestingly, formation of the turtle carapace, which derives from a distinct course of rib development, has been shown to correlate with both turtle-specific patterns of Myf5 expression in the hypaxial myotome (Ohya et al., 2006) and turtle specific muscle connections (Nagashima et al., 2009), further reinforcing the relevance of aspects of Myf5 expression in the production of species-specific skeletal body plans.

A function for Hox and caudal-related genes in axial growth

Posterior axial elongation of the vertebrate trunk occurs from progenitors in the primitive streak area before somite formation and at early somite stages, and in the growth zone of the tailbud thereafter (Cambray and Wilson, 2002; Cambray and Wilson, 2007; Tzouanacou et al., 2009). The Hox genes and their relatives, the Drosophila Caudal-related Cdx genes are highly expressed in this posterior embryonic region (Fig. 2), and have been suggested to confer patterning information to tissues emerging from the growth zone (Deschamps and van Nes, 2005). It was recently shown that, in addition to this patterning function, Cdx and Hox genes also play a role in posterior tissue generation (Young et al., 2009).

Fig. 2.

Cdx and central Hox genes are expressed in the primitive streak area, where the progenitors for axial tissues for trunk and tail reside (see references in the text). Expression of Cdx2 and Hoxb8 is shown on the left at E7.5 (Head fold stage). Axial tissues generated by these areas is schematically indicated in purple on the E9.5 embryo on the right.

Hox genes of the central paralogous groups have the potential to stimulate axial growth

Loss-of-function studies of the ParaHox Cdx genes have shown that these genes are required for posterior axial extension of both vertebrates and invertebrates (Chawengsaksophak et al., 2004; Copf et al., 2004; Faas and Isaacs, 2009; Isaacs et al., 1998; Shimizu et al., 2005; Shinmyo et al., 2005; van den Akker et al., 2002; Young et al., 2009; Savory et al., 2009). In the mouse the importance of Cdx2 in this process was already clear from the analysis of embryos developing in the absence of one or both Cdx2 alleles, as they exhibited mild or severe posterior truncations, respectively (Chawengsaksophak et al., 2004; Chawengsaksophak et al., 1997). The participation of the other two mouse Cdx genes, Cdx1 and Cdx4, in posterior elongation of axial tissues was not clear from initial genetic studies as single or double inactivation of these genes did not have an effect on axial extension (Subramanian et al., 1995; van Nes et al., 2006), However, inactivation of either of these two genes was able to increase the severity of the axial growth arrest of Cdx2 heterozygotes, indicating that these two genes also participate in posterior elongation of axial tissues (van den Akker et al., 2002; Young et al., 2009). The fact that Hox genes are evolutionarily closely related to the Cdx genes, and that Hox genes of the anterior and central paralogous groups are expressed similarly to Cdx genes in the posterior embryonic growth zone throughout axis elongation of the trunk (Young and Deschamps, 2009) incited Young and colleagues to test whether these Hox genes could compensate for missing Cdx alleles in compound mutant embryos. Direct experimental tests of this hypothesis showed that this was indeed the case as two “central” Hox genes, Hoxa5 and Hoxb8, were found to rescue the posterior truncation phenotype of compound Cdx mutants when expressed in the spatio-temporal window of Cdx activity (Young et al., 2009) (Fig. 3). The length and morphology of the sacro-caudal part of the embryonic axis of Cdx mutants was significantly restored by each of these Hox transgenes, indicating that Hox genes are capable of stimulating well-balanced posterior axial growth. Whether these Hox genes act downstream or in parallel to Cdx when compensating for the compromised axial growth resulting from the loss of Cdx genes is not yet clear. Previous work showing the regulatory capacity of Cdx gene products on Hox promoters (Charite et al., 1998; Tabaries et al., 2005), and the shifts of some Hox expression domains observed in Cdx mutants (van den Akker et al., 2002) suggested that Hox genes are downstream targets of Cdx transcription factors. However, study of the expression of endogenous Hoxa5 and Hoxb8 in time and space in the truncated Cdx mutants (Young et al., 2009), favors a mechanism whereby Cdx and transgenic Hox gene products would act in concert rather than hierarchically to stimulate axial growth. It will be important to understand the nature and functional relevance of the growth stimulating activity of central Hox genes. Does it represent an atavistic function from an ancestral time when Hox and Cdx gene products were even more closely related than they now are? And do mouse Hox gene products regulate the same transcriptional target program as Cdx proteins to promote axial growth? Future experimental work should shed light on these questions. It will also be interesting to evaluate the capacity of “Hox genes only” to sustain posterior axial growth and patterning in the absence of any active Cdx genes.

Fig. 3.

Axial skeleton of new born mice upon partial loss-of-function of Cdx genes and upon precocious HoxPG13 expression (A–E) and schematic representation of positive and negative regulators of axial extension before and after the trunk-tail transition (F). A, Cdx loss-of-function mutations (loss of one allele of Cdx2 and of both alleles of Cdx4) arrest axial extension prematurely. B, Wild type. C, Gain of function of the central Hox gene Hoxb8 partially rescues axial extension in Cdx mutants. D, Hoxc13 early gain of function leads to posterior axial truncation. E, Posterior activation of the canonical Wnt pathway by Lef1 constitutive expression partially rescues axial elongation in Cdx mutants. The posterior axial skeleton of new borns is shown with anterior on the left. See text for more details. F, Schematic representation of axial extension of trunk and tail between early somite stage (E8.0, X axis) and the end of posterior extension by tissue addition (E13.5). Shown are positive (“Central” Hox genes, and Cdx genes), and negative (HoxPG13) regulators of axial extension, and some of the effectors of axial growth between developmental stages E8.0 and E13.5. Expression levels (not to scale) are schematically indicated in the Y dimension. In green, growth stimulatory central Hox and Cdx expression, with Wnt signaling (blue) and RA clearance (orange) showing a parallel course. In red, growth inhibitory HoxPG13 expression. Decrease of axial growth stimulation and increase of inhibition correspond to the trunk-tail transition (around E10.5).

Regardless of the hierarchical relationship between Cdx and rescuing Hox genes in the experiments of Young and colleagues, the fact that two “central” Hox genes of paralogous groups 5 and 8 can rescue the Cdx mutant defects in axial growth (Young et al., 2009) suggests that the potential to stimulate trunk axial extension may be a generic property of central Hox genes. This also remains to be fully investigated.

The mechanism through which Cdx mutations impede posterior body axis extension includes, at least in part, the maintenance of canonical Wnt signaling in the growth zone, as shown by the rescue of the truncation phenotype of Cdx mutants by posterior expression of an activated downstream effector of Wnt signaling, Lef1 (Young et al., 2009) (Fig. 3). Cdx proteins also control the clearance of retinoic acid (RA) from the growth zone by the RA-degrading enzyme Cyp26a1, contributing to growth by protecting the axial progenitors from the differentiating effect of RA (Savory et al., 2009; Young et al., 2009).

HoxPG13 genes arrest posterior axis extension: posterior Hox genes as selective substrate for body length variation

Inactivation of Hoxb13 in the mouse resulted in a small extension of the vertebral column (Economides et al., 2003), already suggesting that Hox PG13 genes may exert a negative instead of positive control on posterior axial elongation. In agreement with this observation, transgenic animals expressing Hoxa13, Hoxb13 or Hoxc13 under the control of the Cdx2 promoter prematurely terminated axial growth (Young et al., 2009) (Fig. 3). This termination was preceded by a premature down-regulation of Wnt signaling and of retinoic acid clearance, two events that also precede “natural” axial termination in wild type embryos, and truncation in Cdx mutants (Fig. 3). Interestingly, the premature axial termination by HoxPG13 genes is prevented by co-expression of a Cdx2-driven central Hox gene, providing evidence that HoxPG13 genes may control the same genetic pathways as Cdx genes in mediating axial growth (Young et al., 2009). These results obtained with mouse mutants suggest that modulation of the Cdx/Hox pathways, either by prolonging the Cdx/central Hox-mediated maintenance of Wnt signaling in the growth zone, or by delaying or weakening the inhibitory action of HoxPG13 genes on posterior growth, can positively affect axial length. Interestingly, it has been recently shown that Hoxa13 and Hoxd13 are not expressed at post-cloacal levels in snake embryos (Di-Poi et al., 2010). The authors conclude from their findings that snakes possess a simplified posterior Hox code, having eliminated the contribution of two of the HoxPG13 genes normally involved in axial termination. It is therefore possible that reduction of HoxPG13 activity was a key event for the evolution of elongated body plans.

Given the strong evolutionary conservation of the A–P patterning function of Hox genes, a question arises regarding whether central and posterior Hox genes of short germ band insects also participate in posterior axial growth. This question is becoming an exciting issue, as several components of the Cdx/Hox/Wnt pathway have been proven to play a crucial role in axial extension in these insects and in crustaceans that, like vertebrates, extend their axes by posterior addition of tissues (Bolognesi et al., 2008; Martin and Kimelman, 2009; McGregor et al., 2008). Therefore, conservation of Hox and Cdx-mediated functions between vertebrates and invertebrates could extend beyond the classical AP patterning to include a shared genetic network underlying posterior axial growth.

Is axial extension governed by a single genetic system from anterior trunk to the tip of the tail?

The analysis of an allelic series of Cdx mutants has demonstrated the requirement of all three Cdx genes for extension of the trunk and tail (van den Akker et al., 2002; van Nes et al., 2006; Young et al., 2009). Posterior axial truncation in the most severely impaired Cdx loss-of-function mutants studied so far occurs at a level beyond the forelimbs. The anterior trunk develops normally in these mutants. It will be necessary to await the analysis of Cdx null embryos to know whether these genes are responsible for driving extension of the complete trunk and tail axis, as it seems to be the case in intermediate and short germ band insects. Injection of Caudal RNAi in the cricket Gryllus imaculatus resulted in axial truncation at a very rostral level (Shinmyo et al., 2005). An alternative hypothesis is that invertebrate Cdx has a realm of action covering a more extensive part of the axis than vertebrate Cdx genes. In one possible scenario, vertebrate Cdx genes would control posterior trunk and tail extension whereas anterior trunk extension would be under separate genetic control. In this regard, it will be challenging to study the functional relationship between Cdx genes and T/Brachyury, a unique gene required for posterior axial growth in mice (Wilkinson et al., 1990; Wilson and Beddington, 1997). Homozygous mutant embryos for T/Brachyury still generate 6–8 somites, indicating that the gene is not required for the formation of the most anterior somites. At early developmental (primitive streak) stages T/Brachyury and Cdx work independently as shown by the fact that expression of T/Brachyury is unaffected in early Cdx mutants, and vice versa (Young et al., 2009). Whether the T/Brachyury and Cdx pathways are hierarchically linked at later stages of axial extension is still an unsolved question. Transgenic reporter experiments suggest that T/Brachyury is a Cdx target (Savory et al., 2009). However, T/Brachyury is expressed in the growth zone much longer than Cdx genes, and thus cannot work as a Cdx-dependent gene after E11.0. Regardless of how these genes interact, they must combine their effect in order to regulate posterior growth, since they both impact on the canonical Wnt signaling in a dosage-dependent way (Martin and Kimelman, 2008; Martin and Kimelman, 2009; Young et al., 2009).

Hox/Cdx and the progenitors for the axis

Recent work has localized long-term progenitors for the axis with self-renewing properties in the area between the node and the anterior primitive streak in early somite mouse embryos, and in the remnant of this region in the posterior growth zone at subsequent stages (Cambray and Wilson, 2002; Cambray and Wilson, 2007; Tzouanacou et al., 2009; Wilson et al., 2009). Evidence for bipotent mesoderm and neurectoderm-generating progenitors had been acquired in the past from single cell labeling experiments in the mouse (Forlani et al., 2003; Lawson et al., 1991). The contribution of posterior stem cell-like cells to the notochord (Selleck and Stern, 1991; Stern et al., 1992), the somites (Selleck and Stern, 1991; Stern et al., 1992), and the neural tube (Mathis and Nicolas, 2000a; Mathis and Nicolas, 2000b; Mathis and Nicolas, 2003) had also been reported.

The transcription factors Cdx, central Hox and T/Brachyury exert their function at least in part by maintaining growth-favoring signals in the posterior growth zone. Impaired Cdx activity can be compensated by constitutively active canonical Wnt signaling, similar to what was observed in zebrafish T/Brachyury mutants (Martin and Kimelman, 2008; Martin and Kimelman, 2009; Young et al., 2009). It is therefore likely that Cdx, Hox and T/Brachyury are required to provide the progenitors with an adequate niche. These transcription factor-encoding genes, as well as the signaling factors Wnt3a and Fgf8 are expressed throughout the growth zone (Young et al., 2009). While a molecular signature of the axial progenitor cell population is still missing, it is possible that these cells owe their self-renewing properties to their localization at a very precise position within the growth zone, relatively to the node and streak, and/or to the area undergoing continuing active transcription of Fgf8 (Dubrulle and Pourquie, 2004).

The self-renewing progenitors in the growth zone represent a continuum of axial tissue delivery (Wilson et al., 2009) that gets exhausted when axial extension is terminated. Changes in Cdx and Hox dosage may affect the niche of the self-renewing axial progenitors for trunk and tail via a positive feedback loop on Wnt signaling.

Conclusion

The present overview of recent progress on the role of Hox genes in vertebrate embryonic morphogenesis (mostly the mouse) has focused on three major recent discoveries that shed new light on Hox function.

First, the recently uncovered concept that some Hox genes confer properties to a collection of structures within a given anatomical region has gained wider and wider support by both loss- and gain-of-function studies. In earlier studies, observations of local phenotypes, such as changes in the identity of an individual vertebra, were reported when a single Hox gene was inactivated. Work performed more recently, and reviewed above, has demonstrated that inactivation or overexpression of Hox genes can also result in the transformation of anatomical regions, not just individual elements. Accordingly, expression of HoxPG6 induces rib morphogenesis, expression of HoxPG10 inhibits rib formation, HoxPG11 proteins are necessary for development of the sacrum, and no floating ribs are formed in the absence of HoxPG9 genes. The regional patterning function of Hox genes is therefore dependent not only on the temporal and spatial regulation of their expression, but also on features unique to proteins of each particular Hox paralogous group. These parameters are key to understanding the molecular basis of body plan evolution.

A second important area of recent progress in the Hox field that we have covered in this review is an insight into the mechanism of action of regional patterning by Hox genes. The downstream program of HoxPG6 in inducing rib formation involves the stimulation of a non-myogenic activity of the myogenic genes Myf5/6 in the myotome that instructs the adjacent sclerotome, a program that is suppressed by HoxPG10 genes to produce the rib-less domains caudal to the ribcage. Even if the road to fully understand Hox control of rib morphogenesis is still long, the discovery of the Hox-Myf link is surely a milestone.

The third recent advance in the molecular genetics of Hox gene function during vertebrate development that we review here concerns the involvement of these genes in controlling body axis length. Hox genes thus appear as regulators of both body length and body shape. Evidence was gathered that Hox genes expressed in the trunk can participate in elongating the axis when Cdx (ParaHox) gene function is impaired by mutation and fails to allow completion of axial growth. Moreover, whilst these central Hox genes are capable of stimulating axial growth, genes of the last Hox paralogous group, HoxPG13, cause the arrest of axial extension. In a sense, this control activity of axial growth and termination are regional Hox activities as well, as they impart morphogenetic information to the anatomical regions rostral and caudal to the trunk-tail transition. Stimulation of axial growth by Hox genes of the central paralogous groups was found to rely on the maintenance of Wnt signaling in the embryonic growth zone, and on clearance of retinoic acid away from this area. Future work is expected to further elucidate how central Hox genes favor axial growth, and how Hox genes of the last paralogous group counteract this action. Here as well, regulation of HoxPG13 expression in time and space appears to have been an important substrate for body length variation during evolution of animals. Recent work provides a nice demonstration that a decrease in the number of active alleles of HoxPG13 in post-cloacal paraxial mesoderm is likely to have allowed the delay in axial growth arrest in squamate reptiles. These data, and the newly uncovered regional control of morphogenesis by Hox genes thus supports the notion that Hox genes have been very important players in modifying the animal body plan during evolution.