ABSTRACT
Using gene targeting in mice, we have undertaken a systematic mutational analysis of the homeobox-containing 5′ HoxD genes. In particular, we have characterized the limb defects observed in mice with independent targeted disruptions of hoxd-12 and hoxd-13. Animals defective for hoxd-12 are viable, fertile, and appear outwardly normal yet have minor autopodal defects in the forelimb which include a reduction in the bone length of metacarpals and phalanges, and a malformation of the distal carpal bone d4. The limb phenotypes observed in hoxd-13 mutant mice are more extensive, including strong reductions in length, complete absences, or improper segmentations of many metacarpal and phalangeal bones. Additionally, the d4 carpal bone is not properly formed and often produces an extra rudimentary digit. To examine the genetic interactions between the 5′ HoxD genes, we bred these mutant strains with each other and with our previously characterized hoxd-11 mouse to produce a series of trans-heterozygotes. Skeletal analyses of these mice reveal that these genes interact in the formation of the vertebrate limb, since the trans-heterozygotes display phenotypes not present in the individual heterozygotes, including more severe carpal, metacarpal and phalangeal defects. Some of these pheno- types appear to be accounted for by a delay in the ossification events in the autopod, which lead to either the failure of fusion or the elimination of cartilaginous elements. Characteristically, these mutations lead to the overall trun- cation of digits II and V on the forelimb. Additionally, some trans-animals show the growth of an extra postaxial digit VI, which is composed of a bony element resembling a phalange. The results demonstrate that these genes interact in the formation of the limb. In addition to the previously characterized paralogous interactions, a multitude of inter- actions between Hox genes is used to finely sculpt the forelimb. The 5′ Hox genes could therefore act as a major permissive genetic milieu that has been exploited by evolutionary adaptation to form the tetrapod limbs.
INTRODUCTION
The vertebrate tetrapod limb has been a favorite system of developmental biologists since the turn of the century because it is dispensable for viability and accessible to experimental manipulation (detailed by Hinchliffe and Johnson, 1980). In addition, the variety in form and function of the limb observed in the extant animal kingdom and the fossil record allows for a merging between the disciplines of genetics, development, and evolution (Hinchliffe, 1994). We are interested in understanding the roles of a set of developmental genes called the Hox Complex in the patterning and formation of the vertebrate tetrapod limb.
Hox genes are defined by the presence of a conserved nucleotide sequence (the homeobox) that codes for an amino acid motif (the homeodomain) capable of binding DNA (reviewed by Duboule, 1994). There are at least 39 genes in the mammalian Hox Complex that are divided into four linkage groups called HoxA, B, C, and D located on four separate chromosomes. Each locus contains between nine and eleven closely linked individual Hox genes forming a genetic matrix believed to have arisen by a process of chromosomal duplication (Ruddle et al., 1994). Thus, corresponding genes in separate linkage groups are referred to as paralogues since they are most similar to each other in sequence information and expression patterns. The 5′ ends of the Hox loci carry the five paralogous gene families Hox9-13. Mutational analysis of some of these genes has demonstrated that they play an important role in limb development (Dollé et al., 1993; Small and Potter, 1993; Davis and Capecchi, 1994; Davis et al., 1995; Favier et al., 1995).
At early stages, the distal tip of the limb bud is composed of rapidly proliferating mesenchymal cells termed the progress zone (PZ), which is covered by a layer of epithelial cells called the apical ectodermal ridge (AER; Tickle and Eichele, 1994). Together these two components maintain each other to allow the limb to grow out of the body in a proximodistal direction (Niswander et al., 1993; Fallon et al., 1994; Cohn et al., 1995; Tabin, 1995). A zone of polarizing activity (ZPA) located at the posterior margin of the limb bud influences the anteroposterior patterning of the limb (Tickle and Eichele, 1994). It is during this early stage of limb bud growth that the Hox9-13 genes are consecutively activated (Dollé et al., 1989, 1991a,b; Izpisúa- Belmonte et al., 1991). The completed appendicular skeleton is divided into three zones called the stylopod, zeugopod, and autopod. In the forelimb, the stylopod is made up of the humerus, the zeugopod contains both the radius and ulna, and the autopod comprises carpals, metacarpals, and phalanges. These bones are derived from prechondrogenic condensations that are laid down during the early stages of limb growth by cells that have left the PZ. The bifurcation and segmentation of these condensations followed by endochondrial ossification produces the skeletal elements (Shubin and Alberch, 1986).
Hoxd-12 and hoxd-13 constitute the last two genes of the HoxD locus on Chromosome 2 of the mouse. These genes are activated in a strictly regulated temporal and spatial pattern during development such that the more 3′ gene (hoxd-12) is expressed earlier and in a more proximal domain than the more 5′ gene (hoxd-13; Duboule, 1992; Izpisúa-Belmonte and Duboule, 1992). This feature of temporal and spatial colinearity of Hox genes may play an important role in limb patterning by determining regional identity along the appendicular axis and controlling the properly timed cellular proliferations required for the building of the limb (Davis et al., 1995; Duboule, 1995). To analyze the role of these genes in limb development, we have created targeted disruptions in mice (Capecchi, 1989). Inactivation of hoxd-12 results in subtle phenotypes restricted to the autopod: specifically, the reduction in length of certain metacarpal and phalangeal bones and the malformation of a distal carpal. The hoxd-13 mutant mouse, however, displays severe defects of the forepaw including the strong reduction or elimination of certain phalanges and metacarpals, wrist defects, and the growth of a rudimentary postaxial digit.
With only 39 individual genes in the Hox Complex, it is believed that these products interact with each other to produce numerous novel combinations that may allow for the vast diversity seen in higher vertebrate body plans. In support of this idea, it has been demonstrated that Hox paralogues function together to specify both the axial and appendicular skeleton of the mouse (Condie and Capecchi, 1994; Davis et al., 1995; Horan et al., 1995). Additionally, it has been shown that Hox genes within the same linkage group also work together to control axial patterning (Rancourt et al., 1995). To determine whether genetic interactions within the same linkage group might also control appendicular patterning, we generated a series of mice trans-heterozygous for the hoxd-11, hoxd-12, and hoxd-13 mutations. These animals show more severe autopodal phenotypes than do the individual heterozygotes, demonstrating a genetic interaction. Finally, we examined the interaction between hoxa-11 and hoxd-12 and show that they are required in combination to obtain full outgrowth of the radius and ulna.
MATERIALS AND METHODS
Construction of hoxd-12 and hoxd-13 targeting vectors
A DNA fragment that contains both hoxd-12 and hoxd-13 was isolated from a genomic library made from mouse CC1.2 embryo-derived stem (ES) cell DNA by using a probe 5′ to the hoxd-11 locus. The identity of the genes was confirmed by sequencing the homeoboxes (Izpisúa-Belmonte et al., 1991). Replacement-type gene targeting vectors (Thomas and Capecchi, 1987; Thomas et al., 1992; Deng et al., 1993) were constructed individually for hoxd-12 and hoxd-13 using the same 12.3 kilobases (kb) of genomic DNA flanked by two thymidine kinase genes in a Bluescript-based plasmid. The previously described neo expression cassette, KT3NP4 (Davis and Capecchi, 1994), was inserted into the homeodomain of each of the individual genes. For the hoxd-12 targeting vector, the XmnI site in the homeobox was replaced by a ClaI linker into which KT3NP4 was inserted in the orientation opposite to that of hoxd-12 transcription. This site corresponds to amino acid 18 of the homeodomain. For the hoxd-13 targeting vector, the PvuII site in the homeobox was replaced by a ClaI linker into which KT3NP4 was inserted in the same orien- tation as hoxd-13 transcription. This site corresponds to amino acid 36 of the homeodomain.
Electroporation and ES cell line analysis
The targeting vectors were linearized and electroporated into ES cells. Cell selection and DNA extraction was performed as described (Davis and Capecchi, 1994).
To identify disruption of hoxd-12, genomic DNA isolated from individual cell lines was digested with XhoI and hybridized with a 1.0 kb SalI-XhoI 3′ flanking probe (D). In this case, the wild-type DNA reveals an 8.6 kb DNA fragment and the targeted DNA reveals an extra 5.3 kb DNA fragment. Targeted cell lines were also confirmed by Southern analysis using a NotI digest followed by hybridization with a neo-specific probe, which reveals an 18 kb band unique to targeted cell lines, a ClaI digest hybridized with a 0.3 kb EcoRI-XhoI 5′ flanking probe (Z) to detect a band shift from 25 kb (wild-type) to 19 kb (mutant), and a NotI + ClaI double digest hybridized with two independent internal probes: 1.0 kb EcoRV-XhoI fragment (probe N) to detect a band shift from 14 kb (wild-type) to 8 kb (mutant) and a 3.4 kb XbaI-SalI fragment (probe C) to detect a band shift from 14 kb (wild-type) to 6.3 kb (mutant).
For hoxd-13, cell line genomic DNA was digested with EcoRV in a primary screen and hybridized with the 5′ flanking probe Z to detect a band shift from 6.2 kb (wild-type) to 4.5 kb (mutant). Targeted cell lines were confirmed with an EcoRI digest hybridized with a neo- specific probe to detect a unique 5.5 kb band in targeted cell lines only, a NotI digest hybridized with a neo-specific probe to detect a unique 18 kb band in targeted cell lines only, and a ClaI digest hybridized with the 3′ flanking probe C to detect a band shift of 25 kb (wild-type) to 6.3 kb (mutant).
Breeding, genotyping and skeletal analysis
Mice that carry the targeted mutations were obtained by injection of the appropriate ES cell lines into C57BL/6J (B6) blastocysts. Germline transmission of the mutant alleles from chimeras was confirmed by Southern analysis, and individual breeding colonies were established for the hoxd-12 and hoxd-13 mutants. The hoxa-11 and hoxd-11 mutations have been described previously (Small and Potter, 1993; Davis and Capecchi, 1994). Genotypes were determined by PCR analysis of tail DNA. PCR was performed in 12.5 μl reactions amplified for 25 cycles at 94°C (30 seconds), 60°C (20 seconds), and 72°C (60 seconds). The hoxd-11 primers included one that was specific to the intron 5′ of the homeobox (5′ primer) and the other specific to the 3′ untranslated region downstream from the homeobox (3′ primer). The neo primer was located at the 5′ end of the KT3NP4 cassette. The hoxd-12 and hoxd-13 primers were similarly designed with a 5′ primer and a 3′ primer. The oligonucleotide sequences are provided:
hoxd-11 5′ primer: 5′-CCTTTTTCCTATCTCAGTGCCAG-3′;
hoxd-11 3′ primer: 5′-GGGGTACATCCTGGAGTTCTCA-3′;
hoxd-12 5′ primer: 5′-AGTTGATCTGAGCGAGCTGACAT-3′;
hoxd-12 3′ primer: 5′-CTTGGTCCAAAAGGGCAGGCTT-3′;
hoxd-13 5′ primer: 5′-GCTTAGGTGTTCCAAGTATCCAG-3′;
hoxd-13 3′ primer: 5′-CCACATCAGGAGACAGTGTCTTT-3′;
and
KT3NP4 neo primer: 5′-TTCAAGCCCAAGCTTTCGCGAG-3′.
The PCR products were analyzed using Trevi-Gel agarose elec- trophoresis. For hoxd-11, the products are 507 bp (wild-type) and 367 bp (mutant); for hoxd-12: 334 bp (wild-type) and 226 bp (mutant); and for hoxd-13: 311 bp (wild-type) and 235 bp (mutant).
For adult skeletal analysis, mice were collected at 7-12 weeks of age and stained with alizarin red. Skeletons of newborn mice were stained with both alizarin red S and alcian blue 8GX. Embryos from timed pregnancies were stained with alcian blue only. All staining and bone length measurements were done as previously described (Condie and Capecchi, 1994; Davis and Capecchi, 1994).
RESULTS
Creating hoxd-12 and hoxd-13 mice
Replacement-type gene targeting vectors were constructed by insertion of a neo-resistance cassette into the homeodomain- encoding exon of hoxd-12 and hoxd-13 respectively (Fig. 1A; Materials and Methods for details). The vectors were independently electroporated into ES cells and cultured under conditions that enrich for homologous recombination events (Mansour et al., 1988). Genomic DNA was isolated from selected cells and characterized by Southern blotting to confirm a targeting event for each gene (Materials and Methods for details). A representative targeted ES cell line for each mutation was injected into B6-derived blastocysts to produce chimeric males that passed the mutation through the germline. All subsequent progeny were confirmed by Southern blotting or PCR analysis using tail DNA, and independent breeding colonies were established for each mutant strain.
Phenotypes of hoxd-12 and hoxd-13 mice
Mice heterozygous for hoxd-12 were intercrossed to produce homozygous mutant animals. This mutation was not lethal as the ratio of genotypes observed in adult mice fulfilled the Mendelian expectation. Homozygous mutants of both sexes were fertile and showed no outwardly apparent abnormalities. However, alizarin red stained skeleton preparations revealed subtle malformations in the forelimb. Defects include reductions in the length of several autopodal bones. The metacarpals of digits II and V are shortened (Table 1 and Fig. 2A,B). Phalange 1 (P1) of digit II is slightly diminished in length (80% normal bone length) while P2 of digit V is the most strongly affected (55% normal bone length; Table 1 and Fig. 2G,H). Thus, digits II and V appear significantly shorter than control limbs (Table 1). In addition, the wrist carpal bone d4 shows a subtle, characteristic indentation at its distal border (Fig. 2A,B) that could be the result of improper fusion of the prechondrogenic condensations (discussed below). The autopods of mice heterozygous for hoxd-12 appear normal except for the appearance at low penetrance (approx. 6%) of a very small bony element located postaxially to digit V (not shown).
In contrast, mutations in hoxd-13 cause extensive limb defects when homozygous. This phenotype has been previously described (Dollé et al., 1993), and the mutant animals reported here show an identical phenotype. Appendicular skeleton defects include a strong diminution of the metacarpal length and the reduction or absence of P2 on digits II and V (Fig. 2F,I). Phalangeal bones on other digits are also shortened due to the failure of normal segmentation. Additionally, a webbed effect is sometimes seen when bony elements fuse the interdigital regions (Fig. 2C). In the wrist, the prechondrogenic condensations that normally fuse together to form the d4 carpal bone fail to do so, yielding an extra wrist bone that often articulates with a rudimentary digit VI (Fig. 2I). Minor carpal defects in d4 (including the indentation) and the same type of postaxial bone as found in hoxd-12 heterozygotes are seen in hoxd-13 heterozygotes at a penetrance of 36% (Table 2). The higher penetrance of limb phenotypes in hoxd-13 heterozygotes with respect to those of hoxd-12−/+ mice (6%) or hoxd- 11−/+ mice (0%) is not totally unexpected. If Hox gene products function quantitatively to control skeletal patterning, then the dosage of gene products is likely to be an important factor in proper limb formation. Since the homozygous mutant phenotype for hoxd-13 is more severe than that for hoxd-11 or hoxd-12 mutant mice (Fig. 1B), it is not surprising to see a partial phenotype for hoxd-13 heterozygotes.
Analysis of trans-heterozygotes
The sites of limb defects in hoxd-11, hoxd-12, and hoxd-13 individual mutant mice are summarized in Fig. 1. The absence of any prominent limb abnormalities in hoxd-12 mutant mice suggested that perhaps another gene of the Hox Complex could be compensating for the loss of hoxd-12, as has been demonstrated previously for hoxd-11 (Davis et al., 1995). Since the limb phenotypes seen in hoxd-11 and hoxd-13 mutant mice occur in overlapping domains (Fig. 1C), one or both of these genes might compensate for the absence of the hoxd-12 gene. Ideally one could test this possibility by creating mice doubly mutant for (hoxd-11/hoxd-12) and (hoxd-12/hoxd-13). However, these types of strains cannot be generated readily by breeding mice that carry the individual mutations because the number of meioses required to generate the appropriate cross- overs is prohibitive. Instead, we decided to create mice trans- heterozgyous for the hoxd-11, hoxd-12, and hoxd-13 mutations to look for additional limb phenotypes not apparent in mice heterozygous for the individual HoxD mutations. The appearance of new phenotypes would suggest an interaction between these gene products. For the analysis described below, trans- heterozygotes were generated by breeding a female homozygous mutant for one gene with a male heterozygous for the second mutation. The resulting progeny will be an equal distribution of trans-heterozygotes and control littermates (single heterozygotes).
(Hoxd-11/hoxd-13) trans-heterozygotes (hoxd-11−/+, hoxd-13+/−) show several severe defects in the autopod. The base of metacarpal III consistently shows a bony outgrowth. Carpal bones d3 and d4 fuse together at their proximal borders, and d4 shows the same indentation seen in hoxd-12−/− mice (Fig. 3A,B). Digits II and V are shorter than normal due to the absence or strong reduction in length of P2 (Table 1 and Fig. 3C-F). The phenotype of P2 on digit II shows variable expressivity, from total absence of the bone (Fig. 3D) to a strong reduction in its length (Table 1). Overall, these defects lead to the characteristic shortening of digits II and V of the forelimb.
To examine genetic interactions of hoxd-12 with either hoxd-11 or hoxd-13, we produced both (hoxd-11/hoxd-12) trans-heterozygotes (hoxd-11−/+, hoxd-12+/−) and (hoxd- 12/hoxd-13) trans-heterozygotes (hoxd-12−/+, hoxd-13+/−). In the wrist of (hoxd-11/hoxd-12) trans-animals, a fusion between the navicular lunate and triangular proximal carpal bones is often seen (Fig. 4A,B). This is a phenotype characteristic of hoxd-11−/− mice, but is never seen in the hoxd-11−/+ heterozygotes (Davis and Capecchi, 1994; Davis et al., 1995). Again, there is a significant reduction in the lengths of specific phalangeal and metacarpal bones that result in the overall truncation of both digits II and V (Table 1) with P2 of digit V showing the greatest reduction in length (Fig. 4E,F).
Lastly, (hoxd-12/hoxd-13) trans-heterozygotes (hoxd-12−/+, hoxd-13+/−) show the strongest interaction of these HoxD genes. The wrists of these animals show similar phenotypes to (hoxd-11/hoxd-13) trans-heterozygotes with a bony outgrowth of metacarpal III, a fusion of d3 and d4 carpal bones at their proximal borders, and the indentation of d4 is sometimes seen (Fig. 4C). Additional phenotypes are also evident. An extension of d4 that covers the entire base of metacarpal V is sometimes seen (Fig. 4C). Another animal had a gross fusion of metacarpals III and IV at their proximal ends (Fig. 4D). The P2 phalange of digits II and V are affected, with digit V having the most severe shortening in this bone (from 0% to 55% of normal P2 length; Table 1 and Fig. 4G). These defects account for the significantly shortened digits II and V relative to controls (Table 1).
Most noticeable in the (hoxd-12/hoxd-13) trans-heterozy- gotes, however, is the presence of a rudimentary phalange-like element seen as a postaxial digit VI (Fig. 4G). To follow the development of this bone, we collected timed embryos and newborn pups for skeletal analysis. The sixth digit is readily visible at embryonic day (E)14.5 as a cartilaginous outgrowth on the postaxial side of the forelimb (Fig. 5A,B), but the phalangeal phenotypes are not yet detectable. An extra postaxial bone is sometimes seen in individual hoxd-12−/+ and hoxd- 13−/+ heterozygotes. In these cases, however, the phenotype occurs both with a relatively low penetrance (6% for hoxd-12 heterozygotes and 30% for hoxd-13 heterozygotes; Table 2) and weak expressivity, where the bone is extremely small and does not resemble the phalange-like element seen in the trans- heterozygote mice. In contrast, (hoxd-12/hoxd-13) trans- animals have the extra digit at 86% penetrance and with stronger expressivity.
Analysis of subsequent developmental time points indicated that the (hoxd-12/hoxd-13) trans-heterozygotes experience a delay in the ossification events of the autopodal bones. In control littermates, the ossification center for metacarpal V has appeared by day 1 (D1) in newborns, yet in the trans-het- erozygotes, this same bone does not begin ossification until one full day later (Fig. 5C-E). Similarly by D5 the final limb ossification event at P2 in digit V has been initiated, but is delayed in the trans-heterozygotes (Fig. 5F,G). This may account for the severe reduction (and sometimes complete elimination) of P2 on digit V. Additionally, the indentation on the distal carpal bone d4 may be explained by a delay in the fusion of the two condensations that make up this bone. At D1 it can already be seen that this fusion has taken place in the control littermate, but in the trans-animal this growth appears to have been slowed down or delayed such that the two elements remain separate (Fig. 5C-E). By D5, these two condensations approach each other except at their most distal borders which maintain a scalloped outline and lead to the indentation seen in the final ossified bone (Fig. 5G).
Table 2 and Fig. 6 summarize all of the forelimb defects for these trans-heterozygous strains. We restricted our analysis to the forelimbs because the hindlimbs showed only mild defects including reductions in the length of P2 on digit V and an aberrant sesamoid bone at the base of metatarsal IV (not shown).
Genetic interaction between hoxa-11 and hoxd-12
Previously we have shown that in mice homozygous for mutations in both hoxa-11 and hoxd-11 only a vestige of the radius and ulna are formed (Davis et al., 1995). Since hoxd-12 does not have a paralogous family member in the HoxA linkage group, we tested for genetic interaction between hoxa- 11 and hoxd-12. Mice homozygous for both hoxa-11 and hoxd- 12 mutations show marked reductions in the length as well as curvature of both the radius and ulna (Fig. 7). Hoxd-12 mutant homozygotes alone do not have any apparent defect in the formation of the radius and ulna and intermediate genotypes including hoxa-11, hoxd-12 compound heterozygotes did not show any obvious phenotypes.
DISCUSSION
Mice homozygous for the hoxd-12 mutation are viable and fertile but show mild defects in the forelimb autopod. In these animals, the lengths of metacarpals in both digits II and V are significantly reduced. In addition, P1 of digit II and P2 of digit V are shorter than normal. Taken together, these phenotypes result in characteristic shortening of digits II and V. In addition, hoxd-12−/− mice have a subtle defect in the carpal bone d4, which is marked by a reproducible indentation at its distal edge.
Hoxd-13 mutant mice have more extensive phenotypes in the autopod. In these animals, the entire paw is smaller than normal due to the strong reduction, complete absence, or improper segmentation of metacarpal and phalangeal bones in all of the digits, but especially in digits II and V. In the wrist, d4 is not properly formed and instead yields an extra carpal which often produces a rudimentary postaxial digit VI.
Interestingly, the hindlimbs of hoxd-12 animals appear relatively normal, as was seen for hoxd-11 mutant mice (Davis and Capecchi, 1994). Hoxd-13 animals, however, display severe hindlimb defects (not shown here but identical to those reported by Dollé et al., 1993). This difference suggests that the various HoxD genes function independently in the different limbs.
The relatively mild forelimb defects seen in hoxd-12 mice suggested that compensation by other Hox genes may be masking this genetic lesion. The notion of spatial and temporal colinearity would have predicted a hoxd-12 phenotype located between the sites showing a phenotype in hoxd-11 and hoxd-13 mice (Davis and Capecchi, 1994; Dollé et al., 1993). However, since mutations in hoxd-11 and hoxd-13 show over- lapping defects (Fig. 1C), it remains possible that either one or both of these genes’ products could provide sufficient compensation to allow the near normal development of the forelimb autopod in mice deficient for hoxd-12. Genetic redundancy and interaction within the Hox Complex have been demonstrated for paralogues involved in the formation of both the axial (Condie and Capecchi, 1994; Horan et al., 1995) and appendicular skeletons (Davis et al., 1995). In addition, non- allelic noncomplementation occurs within the HoxB linkage group during formation of the vertebral column (Rancourt et al., 1995). Here we show, through the analysis of trans-het- erozygous animals generated from mice mutant for hoxd-11, hoxd-12 or hoxd-13, that adjacent genes of the 5′ HoxD complex genetically interact to form the autopod.
The severity of phenotypes in these trans-animals increases from (hoxd-11/hoxd-12) through (hoxd-11/hoxd-13) to (hoxd- 12/hoxd-13) trans-heterozygotes. This pattern seems to reflect a dominant role for hoxd-13 in the final structure of the autopod. This is not surprising since the hoxd-13 individual homozygous mutant shows a stronger limb phenotype than that of hoxd-11 or hoxd-12 individual mutants (Fig. 1B). From an evolutionary standpoint, this hierarchy may have provided flexibility for modulating the autopod. A limb can be divided into three zones: the most proximal is the stylopod containing one large support bone; this is followed by the middle zeugopod with two similar bones, the radius and ulna, that usually function as a hinge to allow limb rotation; finally at the most distal end is the autopod with several carpal, metacarpal, and phalangeal bones. The most distal end of the limb holds the greatest potential for exploitation by natural selection (e.g., the opposable thumb in primates, the wrist modification in pandas that allows for grasping bamboo, extensive phalangeal growth in bats to produce a large wing, etc.). Thus, it is advantageous to let the last genes (the Hox-13 cognates) involved in producing the last structures (the autopod) have the greatest developmental dominance in order to influence the autopodal unit. This concept may underlie the phenomenon of ‘proximal stability/distal variability’ in limb structures (Hinchliffe, 1991).
It is important to note that in all three trans-animals we created, the phenotype involves a significant reduction in the total lengths of digits II and V. This is caused primarily by the shortening or total elimination of the P2 bone in each digit. This phenotype is characteristic of numerous Hox mutant mice and is consistent with the idea that these genes influence cellular proliferation in limb patterning. Mesenchymal cells form prechondrogenic condensations which subsequently bifurcate and segment to produce a primary pattern of the limb. Because of the ordered development of this primary pattern, a deficiency of limb bud cells invariably causes truncations in the most terminal elements (Shubin and Alberch, 1986). These terminal elements in mice are the P2 bones of digits II and V.
Various secondary modifications of this primary pattern can occur (Muller, 1991). For example, in mice, two wrist condensations fuse together to form the one element that will eventually ossify into the d4 carpal bone. Also, the skeletal analyses of different developmental time points for (hoxd-12/hoxd-13) trans-heterozygotes demonstrate a delayed timing in ossification events in the autopod, as was seen with hoxd-13 homozygous mutants (Dollé et al., 1993). This altered timing of development (heterochrony) also occurs in mice mutant for the two paralogues hoxa-11 and hoxd-11 (Davis et al., 1995). Thus, it appears that individual mutations in several different Hox genes result in a similar defect which could be explained by an insufficient number of cells to finish the autopod (truncating digits II and V by the loss of P2) and a resulting delay in ossification events.
Interestingly, the (hoxd-12/hoxd-13) trans-animals also produce an extra rudimentary postaxial digit VI that has been observed in other mutant mice (Dollé et al., 1993; Wurst et al., 1994; Lampron et al., 1995; Fawcett et al., 1995) and occurs spontaneously in a partially inbred strain (Cusic and Dagg, 1985). The production of this extra digit is preceded by a local outgrowth of mesenchymal cells on the postaxial side of the limb. The digit, therefore, may be the result of the increased number of mesenchymal cells participating in forming a condensation by overcoming an ‘anti-chondrogenic factor’ from the overlying ectoderm (Solursh, 1984; Hurle et al., 1991) or by initiating an intrinsic chondrogenic event (Hinchliffe and Horder, 1993). The frequent occurrence of this phenotype in a variety of mice suggests this particular posterior region of the limb bud is very sensitive to cell concentration and/or interaction with the ectoderm (Laufer et al., 1994; Parr and McMahon, 1995; Yang and Niswander, 1995). If Hox genes regulate cell proliferation and Hox-deficient mice are therefore perceived as having fewer cells to construct a normal limb, then it appears contradictory to have extra digits forming in some of these strains. In addition to regulating the number of limb cells, Hox genes may influence other cell properties, such as recruitment and aggregation. Therefore, polydactyly may occur in these strains as a result of the improper distribution of the few remaining limb cells. In support of this, it has recently been reported that hoxa-13 affects the adhesiveness of limb bud cells by controlling homophilic cell interactions (Yokouchi et al., 1995).
As noted, in (hoxd-11/hoxd-12) trans-heterozygotes, the phenotypes include an occasional fusion between the navicular lunate and triangular proximal carpal bones. This is an interesting defect because it was initially observed in animals homozygous for a hoxd-11 mutation (Davis and Capecchi, 1994). Neither hoxd-11−/+ nor hoxd-12−/− animals ever show this phenotype, yet the trans-combination of (hoxd-11−/+, hoxd-12+/−) is able to reproduce the defect, as if half the amount of the hoxd-12 gene product can phenocopy the amount of hoxd-11 gene product from half to zero (i.e., creates a phenotype seen in hoxd-11 null mice, but not in hoxd-11 het- erozygotes). Mice mutant for the paralogue hoxa-11 also occasionally show this same fusion, and remarkably, compound heterozygotes between these two Hox-11 cognate alleles (hoxa-11−/+; hoxd-11−/+) also show the phenotype (Davis et al., 1995). From these results we concluded that hoxa-11 and hoxd- 11 must be working together, since in order to produce this defect all that was required was any two mutant alleles of either hoxa-11, hoxd-11, or one from each locus. It would appear from this trans-heterozygote analysis that hoxd-11 and hoxd- 12 might also genetically interact in the proper development of the wrist. It is important to note that not every genotype [hoxa- 11−/− or hoxd-11−/− or (hoxa-11−/+; hoxd-11−/+) or (hoxd-11−/+; hoxd-12+/−)] shows this phenotype with 100% penetrance. This may suggest a degree of variability in the limb patterning program, where combinations of gene products can compensate, interact, or compete with each other.
Since mutations in hoxa-11 and hoxd-11 together almost entirely eliminate the formation of the radius and ulna, and since hoxd-12 is expressed in the early limb bud after hoxa-11 and hoxd-11, it may appear surprising that combined mutations in hoxd-12 and hoxa-11 should affect the length of the radius and ulna above that observed in hoxa-11 mutant homozygotes alone. A plausible explanation for this observation is that the Hox-11 paralogous members, which are expressed first and show the most dominant role in the formation of the zeugopod, are involved in initiating the outgrowth of the prechondrogenic condensations that mediate the formation of the radius and ulna. Hoxd-12 then cooperates with hoxa-11 and hoxd-11 (as well as other Hox genes) to regulate the final outgrowth of these long bones. This observation again emphasizes the complexity of the genetic interactions among these Hox genes in mediating the patterning of the vertebrate limb.
There are seven genetic clues that provide information as to the role of Hox genes in limb development. First, individual targeted disruptions of four of the 5′ Hox genes (hoxa-11, hoxd-11, hoxd-12, and hoxd-13) have not resulted in digit homeosis. Instead, regional malformations in the shapes, lengths and segmentations of bones are observed, suggesting that the Hox genes control the localized growth and/or recruitment of cells that contribute to the formation of the appendicular skeleton. Second, these four mutant strains show sites of overlapping limb defects. Third, double mutants for the par- alogues hoxa-11 and hoxd-11 revealed a major role for Hox genes in patterning the limb along the proximodistal axis. In these double mutants, only a vestige of the radius and ulna is formed (Davis et al., 1995). However, following this major dis- continuity, limb development resumed. Based on this observation, we proposed that cognate genes of the posterior set of Hox genes work in union to specify limb bone formation in the proximal to distal direction. Fourth, nonparalogous as well as paralogous Hox genes interact to regulate the final outgrowth of the long bones. A case in point is the shortening and curvature of the radius and ulna in hoxa-11, hoxd-12 double mutants. The changes in the expression patterns of Hox genes during development may account for this genetic interaction. This may result from overlapping expression domains at later developmental time points or by the recruitment of distal cells to interact during the final outgrowth of the limb bones. Fifth, the trans-genetic interactions reported here demonstrate that genes in the same linkage group also interact to pattern the limb. Sixth, a characteristic phenotype seen in the majority of these mutant strains is the strong reduction or elimination of P2 in digits II and V, leading to the overall shortening of these two digits. These are the last elements formed in the last digits specified. Such a common endpoint for all of these mutations could be explained if these mutations cause a reduction of cell number at the time of formation of the final elements of the autopod. This observation again emphasizes a common role for these genes in controlling the proliferation of the precursor cells, positively or negatively, needed for forming the limb. Seventh, also common to all of these mutations is a delay in the growth of chondrogenic and ossification patterns. This het- erochrony is certainly a major contributor to the final pheno- typic consequences of these Hox mutations. The observed het- erochrony may again result from a shortage of cells.
It is apparent from this genetic analysis that the formation of the vertebrate limb involves a complex set of interactions among multiple Hox genes. Due to the extensive overlap of function, the role of individual genes does not become apparent until they are combined with mutations in other members of the complex. A strong role in mediating the proper outgrowth of the limb along the proximodistal axis has been established. What needs to be resolved is how intimately these genes are involved in generating the final pattern of the tetrapod limb. Minimally, these genes appear to play a role in contributing the timed appearance of the precursor cells necessary for limb construction (Davis et al., 1995). However, a more extensive role is likely. Through multiple combinatorial interactions among Hox genes, the needed complexity for controlling, positively and negatively, the detailed growth of the limb prechondrogenic condensations could be readily attained. Through such a mechanism the Hox genes could be directly responsible for the fine sculpting of the limb bones. In such models the growth and patterning of the limb can be viewed as inherently linked such that the branching and segmentation patterns of the prechondrogenic condensations are a direct consequence of the ‘developmental constraints’ imposed by the changing limb bud geometry (Oster et al., 1988; Hurle et al., 1991; Hinchliffe and Horder, 1993). We can begin to address such potential roles for Hox genes by rearranging the sites and timing of Hox gene expression within the context of the native Hox Complex. Regardless of the extent of information that Hox genes contribute to patterning the limb, they clearly play an influential role in its formation. By varying the timing and expression domains of Hox genes in the limb, novel genetic combinations and interactions may have contributed profoundly to the amazing diversity of form and function visible in tetrapod limbs.
ACKNOWLEDGEMENTS
We thank M. Allen, S. Barnett, C. Lenz, E. Nakashima, G. Peterson, S. Tamowski, and M. Wagstaff for excellent technical assis- tance. L. Oswald helped with the preparation of the manuscript. A. P. D. was supported by an NSF predoctoral fellowship and an NIH genetics training grant.