Hox genes and growth: early and late roles in limb bud morphogenesis

Comparative molecular analysis suggests that the genes that gave rise to the modem Hox clusters have been specifying regional differences in the animal body for possibly more than one billion years (Kappen and Ruddle, 1993; Shubert et al., 1993). Sequence comparisons suggest that, at the time of the origin of the chordate, nematode and arthropod lineages, there were between 5 and 7 members of the Hox complex (Shubert et al., 1993). Since that time, these complexes have undergone expansion and duplication independently in different lineages ultimately generating, for example, 38 members in the four clusters of Hox genes observed in gnathostome vertebrates. During the course of this expansion, these genes retained their original functions in patterning the anterior-posterior (A-P) body axis and also acquired new functions in regulating other aspects of morphogenesis. When novel developmental innovations arise in evolution, modifying preexisting body plans, the derived embryonic steps tend to occur late in ontogeny (“the general precedes the specialized” in von Baer’s formulation; Gould, 1977). In performing patterning roles that are phylogenetically old, and therefore occur developmentally early, there may be a high degree of overlap in function between paralogous genes in different Hox clusters, and even between sequentially arranged genes within a cluster (see Fig. 1 for nomenclature). Continued function in these early roles constrains their divergence (as discussed by Holland et al., 1992). However, in roles that arose later in evolution and are frequently observed later in development of the embryo, the differences in function between paralogue groups and between different genes in a paralogue group are more pronounced, taking advantage of divergence in regions not required for primitive function. This concept is generally useful in attempting to decipher the role of the Hox genes in morphogenesis and is illustrated in our consideration of the role of the Hox genes in limb development.

The limb arises from an accumulation of cells in the lateral plate mesenchyme. These cells induce a specialized structure in the overlying ectoderm, the apical ectodermal ridge (AER). Subsequent proliferation and outgrowth of the limb mesenchyme is dependent on signals from the AER; in turn the AER is dependent on the underlying mesenchyme for its maintenance. Mesenchymal cells in the region subjacent to the AER (the so-called progress zone) remain in a highly proliferative and undifferentiated state. As the limb bud grows, cells are continuously displaced from this zone. Displaced cells decrease their rate of proliferation and subsequently begin to differentiate, leading to a proximal-to-distal wave of differentiation. Cells that will participate in the formation of distal structures are still being generated when cells participating in the formation of proximal structures are beginning to differentiate.

The Hox genes are expressed in intriguing patterns during morphogenesis of the wing and leg. These patterns evolve in complex ways as development proceeds. In earlier work in the mouse limb, these changing patterns of expression were described as the continuing evolution of a single expression domain (Dolle et al., 1989). Examination of Hox gene expression in the chick limb bud reveals that the evolving pattern of gene expression described for each gene actually represents temporal and spatial overlap of several distinctly regulated expression domains.

Members of the Hoxd cluster are sequentially expressed during the development of the chick limb bud (Fig. 2). In the early phases of limb bud development, transcripts of the Hoxd-9 and Hoxd-10 genes are expressed across the A-P extent of the nascent wing and leg bud. Activation of the nine paralogues is synchronous with the initial outgrowth of the bud. It is during this phase of development (stages 17-19) that the presumptive stylopod is displaced from the progress zone (Saunders, 1948).

The next phase of limb outgrowth involves the activation of the Hoxd-11 and Hoxd-12 genes in progressively restricted domains in the posterior half of the bud. This occurs at a stage when the presumptive zeugopod is being displaced from the progress zone (stage 19 through 21/22). Finally, Hoxd-13 is activated in the posterior distal region of the limb bud at stage 20. As development proceeds, expression of Hoxd-10, Hoxd-11, Hoxd-12 and Hoxd-13 come to occupy very similar domains in the distal segment (autopod) and remain strongly expressed in this region while the nested domains of expression in the proximal segments of the limb are fading. For most of these genes, a clear separation of proximal and distal expression domains is observed. Hoxd-13 is only weakly expressed in the posterior zeugopod at this stage, but is strongly expressed in the autopod in a domain that extends slightly anterior to that of the other Hoxd genes.

The apparent separation of Hoxd gene expression into several phases is confirmed by recent identification of Sonic hedgehog as a gene involved in the regulation of the Hoxd gene expression in distal mesenchyme. Early expression of Hoxd genes is independent of Sonic regulation. Hoxd-9 and Hoxd-10 appear in proximal regions before Sonic is expressed (Laufer et al., unpublished data). The posterior distal expression of Hoxd-11, Hoxd-12 and Hoxd-13 in the limb appears to be regulated by Sonic; ectopic expression of Sonic in the anterior region of the bud is sufficient to elicit ectopic expression of these genes in the anterior distal region (Riddle et al., 1993; Laufer et al., unpublished data).

Analysis of the sequences required to drive expression of the mouse Hoxd-11 gene suggests that Hox gene expression is also independently regulated in proximal and distal regions of the mouse limb bud. The degree of overlap between domains prevents their observation as distinct entities in the wild-type animal. However, transgenic mice bearing different segments of the Hoxd-11 fused to a β-galactosidase reporter construct demonstrate that expression of Hoxd-11 may be independently regulated in the proximal (stylopod), central (zeugopod), and distal (autopod) segments of the limb (Gerard et al., 1993). In particular, expression in the distal segment of the limb is separable from expression in the central segment.

As development proceeds, Hoxd expression fades in the mesenchyme with the exception of the perichondrial regions. The maintenance of Hoxd gene expression in the perichondrial regions correlates with the expression of a second vertebrate hedgehog homologue, Indian hedgehog. This late pattern of expression may represent an independently regulated phase of Hox gene transcription indicative of a separable role in limb morphogenesis.

The early pattern of Hoxd gene expression in the chick limb correlates with the outgrowth of the bud. The onset of expression of Hoxd-9 and Hoxd-10 throughout the bud is synchronous with initial limb outgrowth and at this stage bud outgrowth is symmetric along the A-P axis. Growth of the bud becomes markedly asymmetric along the A-P axis with a distinct posterior bias when the Hoxd-11 and Hoxd-12 genes are activated in the posterior half of the bud. This bias in outgrowth is observed during stages 19 to 23, a period when the presumptive sylopod and zeugopod constituents are being displaced from the progress zone. Subsequent growth of the bud becomes less asymmetric as the distal domains of Hoxd gene expression spread to the anterior regions of the bud.

This correlation between the timing and position of early Hoxd gene expression and the growth of the undifferentiated limb mesenchyme could reflect the response of both Hox gene expression and cel) proliferation to a common inducer, or even a requirement for additional cell divisions to achieve the sequential activation of increasingly 5’ Hox genes. Alternatively, one role of the 5’ Hoxd genes early in limb bud development may be to mediate the asymmetric growth of the bud. In such a case, the discrete regulation of these genes in the three segments of the limb may reflect a fundamental role of these genes in the evolution of the appendage; the sequential appearance of elements capable of activating these genes would correspond to the addition of segments along the proximal distal axis.

The effects of ectopic expression of Hoxd-11 suggest that regulation of the growth of undifferentiated mesenchyme is an early role of the Hoxd genes in the limb. These ectopic expression experiments were performed using a replication competent retroviral vector to express the mouse Hoxd-11 protein specifically in the developing chick limb bud (Morgan et al., 1992) Hoxd-11 is normally expressed in the posterior regions of the bud in a domain that encompasses the primordia of the fibula, posterior metatarsals and digits II, III and IV (see Fig. 3). The retrovirus was used to infect the entire limb bud and therefore expand that domain of expression into anterior regions. These infections led to complex and variable phenotypes which reflects the mechanistic constraint on virus-mediated ectopic expression (see Fig. 4 legend).

Perhaps the most striking phenotype associated with the mis-expression of Hoxd-11 in regions anterior to its normal domain of expression in the leg bud was the appearance of an additional phalange in digit I leading to a morphology similar to that of digit II (Fig. 4A,B). Anterior digits containing either an additional phalange or a single elongated phalange were observed, while no effect was observed on the bones of the posterior digits II, III and IV, which arise from the region in which Hoxd-11 is normally expressed.

A conceptually similar phenotype is observed in Hoxd-11- injected wings. Cells in the anterior region of the chick wing which do not express Hoxd-11 do not normally give rise to skeletal elements. Ectopic expression of Hoxd-11 in the anterior region of the wing results in the formation of an additional digit at the anterior edge of the wing which resembles digit II in structure.

The apparent homeotic change was observed in roughly 30% of the infected legs showing any phenotype. This level of penetrance indicates that the digit-affecting action of Hoxd-11 occurs early in limb development, when only 30% of injected limb buds are already fully infected by the spreading virus.

Both the appearance of an additional phalange in digit I in the leg and an additional anterior digit in the wing can be ascribed to the proposed effect of the Hoxd genes on outgrowth or survival of undifferentiated limb mesenchyme. We propose that procedures that increase the proliferation or survival of limb bud mesenchyme lead to the formation of additional cartilaginous condensations. We suggest that, when these condensations are sufficiently separated, they give rise to additional bones. Insufficiently separated condensations fuse to form a single enlarged bone. Hence the effects on distal elements of ectopic Hoxd-11 expression in the anterior region of the wing bud resemble those observed when elevated levels of bFGF are achieved either by implantation of an FGF-soaked bead or by implantation of FGF-secreting cells in the anterior wing bud (Riley et al., 1993). In both cases, increased proliferation or survival of undifferentiated limb mesenchyme leads to the formation of additional bones. Indeed, earlier activation of Hoxd-11 in the anterior region of the hind limb bud can have a more pronounced effect on limb growth leading to the formation of an additional digit anterior to the normal digit I.

A role for the Hox genes in stimulating growth of limb precursors is sufficient to explain the appearance of an additional bone in the anterior digit after ectopic expression of Hoxd-11 in the anterior region. However, this postulated function is not sufficient to explain other skeletal phenotypes observed in response to ectopic Hoxd-11. Most of these reflect a failure of specific bones to mature properly after apparently normal initial development (Fig. 4). Ectopic Hoxd-11 expression in the anterior region leads to an abnormally short second tarsometatarsal, while the third and fourth tarsometatarsals are comparatively normal. In a similar fashion, the tibia is reduced to half its normal length and is now very similar in length to the fibula. This phenotype was observed in more than half of the affected specimens as compared to the one third that showed an additional phalange in digit I. The higher penetrance suggests that these phenotypes result from a later effect of altered Hoxd-4 expression on the limb bud. (When the injected virus has had time to spread completely in a higher percentage of limbs.)

If the digital phenotypic effect were more prevalent than the proximal effect, one might ascribe the difference in penetrance to the fact that there is a proximal-to-distal wave of differentiation in the limb bud. Thus the Hox genes could act at a single developmental stage (e.g. cartilage condensation); and when that stage occurred proximally (early) there would naturally be less frequent complete infection than when that same stage occurred in distal regions (late). However this is not the case: the proximal phenotype is more frequent, and hence occurs later in time. This strongly implies that phenotypes in regions proximal to the digits are not due to the effect on early proliferation (that gives rise to the digit phenotypes), but rather that they result from a distinct later effect of altered Hoxd expression on the developing limb. Consistent with this hypothesis is the observation that at early stages of development, the initial condensations that will become the tibia and fibula appear normal in infected limb buds. Through stage 26, there is no obvious difference between the zeugopod of the infected and uninfected contra-lateral limbs. At this stage, the cartilaginous condensations that will contribute to the tibia and fibula are approximately equal in length. The normal disparity between the size of these bones at later stages arises as a consequence of greater subsequent growth of the tibia compared to the fibula. Hoxd-11 expression in anterior regions prevents this preferential growth leading to a phenotype in which both bones are similar lengths. In a similar fashion, the second matatarsal also fails to undergo normal elongation and is reduced to roughly half the size of the normal bone. Even when the anterior metatarsal shows a complete phenotypic conversion to resemble the posterior metatarsal, both are half the length of a normal metatarsal. (Fig. 4B).

These effects could be due to persistent expression of the Hoxd-11 gene in differentiating cells where it is normally turned off. However, such an explanation would predict that all bones in the limb would be affected equally since Hoxd genes are normally down regulated in all but the perichondrial regions within the diagrammed domain of expression. The fact that this inhibitory effect on bone elongation is observed only in bones from the regions where Hoxd-11 is not normally expressed suggests that it is more specific. Some other aspect of differential gene expression such as the expression of the endogenous gene or Hoxd-12 in the normal expression domain of Hoxd-11 prevents the inhibitory action of the exogenous gene in this region. Alternatively the exogenous Hoxd-11 interferes with the activity of factors that are only responsible for stimulating the preferential growth of the tibia and anterior metatarsal. Other Hox genes are likely targets for this interaction; protein-protein interactions between different Hoxd gene products which inhibit transcriptional activity have been described (Zappavigna et al., 1994). This late effect on bone growth has also been observed in gene ablation experiments with the mouse Hoxd-13 gene. Absence of the Hoxd-13 gene product leads to delayed and incomplete ossification of the proximal phalanges in the anterior and posterior digits (Dolle et al., 1993).

It has been proposed that the combination of Hoxd genes expressed in a digit primordium specifies the unique identity of each digit in a combinatorial code and mediates its characteristic morphogenesis (Tabin, 1992). The apparent homeotic nature of the morphological changes in the anterior digit and metatarsal resulting from ectopic expression of Hoxd-11 can be interpreted to support such a model. The separable early and late roles of Hox genes that emerge from the analysis presented here do not in themselves contradict this view. Rather they provide a two-phase mechanistic basis for the effect of Hox genes on limb (including digit) morphology.

However, the simple Hox-code model is excluded when one also takes into account the extremely dynamic expression patterns of the multiple domains of each Hox gene within the limb (Nelson, Morgan and Tabin, unpublished data). The Hoxd genes do appear to act early when there are nested expression patterns of the Hoxd genes along the anterior-posterior limb axis. However, they also clearly function later as well, when their relative domains are very different, and not at all aligned with unique digit primordia. Indeed, at this later time all of the Hoxd genes are expressed across the entire distal limb bud and may play partially redundant roles in this domain. This is consistent with the finding that deletion of the Hoxd-13 gene affects all the digits (not just the most posterior one), but results in a fairly subtle effect in mice whose other Hoxd genes are intact (Dolle et al., 1993). Thus, while the expression of Hox genes in the limb bud appear to regulate digit morphology, they do not encode digit identity by a simple combinatoral code.

An attempt was recently made to apply the Hox code model to the problem of understanding the origin of pentadactyly (five digits) in modern tetrapods (Tabin, 1992). A two part argument was proposed.

1. (1)

   It is know that the regulation of the number of digits produced in a limb field is independent of the regulation of the morphology of the digits. If there were a constraint on the number of distinct morphological types of digits that could be specified in a given limb field such that only five ‘different’ digit types were permissible, then more than five digits might never be able to evolve with a selective advantage in a population. Evidence presented in support of this hypothesis included the fact that polydactylous individuals arising spontaneously in many species do not typically have a novel extra digit but rather have a morphological duplicate of either the most preaxial or postaxial digit. Similarly, polydactylous primitive tetrapods such as Acanthostega (8 total digits) have only five or fewer morphological types of digits. Finally, when extra distal ‘digits’ arise in tetrapod evolution with unique morphologies, they are invariably produced by modification of a wrist bone and are not true extra digits (e.g. Panda Bears, some moles, some frogs). Thus, there does indeed appear to be some constraint, limiting number of potential digit ‘types’ to five.

2. (2)

   It was further argued that the Hox code could directly produce such a constraint since only five combinatorial codes are possible from the Hoxd genes expressed in the limb. However, as we have seen, the Hoxd genes do not act in a simple combinatorial code for ‘digit identity’. While they do contribute to the regulation of digit morphology, our current understanding of their action does not provide an indication of a constraint on potential morphologies. Either such a constraint remains to be discovered in the subtler aspects of Hox gene action or else one will have to look elsewhere for it.

In summary, rather than an alteration in Hoxd code producing the apparent homeotic effect of Hoxd-11 misexpression, we suggest that the effect on the anterior digit results from an early stimulatory effect on limb outgrowth which generates sufficient additional mesenchyme to generate an additional chondrification center. At this stage of development, all of the 5’ Hoxd genes may have qualitatively similar roles. Appropriately timed ectopic expression of another 5’ Hoxd gene might well achieve a similar result. Later effects on skeletal morphogenesis may reflect paralogue specific roles in regulating bone growth. The fact that these genes are having an effect at a stage when the expression domains of the Hoxd-10, Hoxd-11 and Hoxd-12 genes are very similar in the autopod renders the hypothesis that different combinations of Hoxd genes specify digit identity improbable. Rather, the combined early action on the accumulation of undifferentiated mesenchyme and later action on skeletal growth lead to the characteristic pattern of the chick limb skeleton.