The mouse Ulnaless mutation deregulates posterior HoxD gene expression and alters appendicular patterning

The vertebrate limb serves as an excellent experimental system to elucidate the molecular mechanisms underlying patterning of the embryo. A combination of embryological, molecular and genetic approaches have provided a framework for identifying genes associated with morphogenetic signaling centers that coordinate initial patterning along the three axes of the limb (Tabin, 1991; Cohn and Tickle, 1996). Translation of this initial patterning information into the final limb structure may be modulated through the action of Hox genes.

The mouse Hox gene family consists of 39 members, which are organized into four clusters, HoxA, HoxB, HoxC and HoxD, located on different chromosomes. All share a highly conserved DNA-binding motif, the homeodomain. Each cluster contains 9-11 genes transcribed in the same orientation (McGinnis and Krumlauf, 1992). The expression of a Hox gene in the embryo is colinear with its position in the cluster. In vertebrates, both temporal and spatial colinearity is observed: expression of 3′ genes precedes expression of 5′ genes and 3′ genes have more anterior limits of expression than 5′ genes (Duboule and Dolle, 1989; Graham et al., 1989; Izpisua-Belmonte et al., 1991). The most 5′ genes (groups 9 to 13) of the HoxA and HoxD clusters are related to the Drosophila Abd-B gene and are activated in the forelimb and hindlimb buds consistent with temporal and spatial colinearity, although their subsequent expression is dynamic and complex (Dolle et al., 1989; Haack and Gruss, 1993; Nelson et al., 1996). The molecular basis of colinearity remains an outstanding question. However, there is evidence to suggest that colinearity requires local regulatory interactions, such as enhancer sharing, coupled with higher order regulation (Gerard et al., 1996; van der Hoeven et al., 1996; Gould et al., 1997).

The colinear activation of HoxA and HoxD genes along the proximal-distal axis of the limb can be correlated with the prospective stylopod, zeugopod and autopod (Davis et al., 1995; Nelson et al., 1996). In the forelimb, the stylopod consists of the humerus, the zeugopod consists of the ulna and radius, and the autopod consists of the carpals and digits. Expression patterns of HoxA and HoxD genes within these regions are critical for their patterning because gene-targeted mutations in individual HoxA and HoxD genes lead to specific reductions of limb elements. The more 3′ proximally expressed genes (group 9) alter the stylopod, and the more 5′ distally expressed genes (groups 11, 12 and 13) alter the autopod. Combinations of double and triple loss-of-function mutations within the HoxD cluster, and between paralogous genes in the HoxA and HoxD clusters, have led to the conclusion that the overall dosage of Hox genes within a domain is important for patterning. However, some genes appear to have a more dominant role in the patterning of specific limb elements (Dolle et al., 1993; Small and Potter, 1993; Davis et al., 1995; Davis and Capecchi, 1994, 1996; Favier et al., 1995, 1996; Fromental-Ramain et al., 1996a,b; Herault et al., 1996; Kondo et al., 1996; Zakany and Duboule, 1996).

In addition to mutations designed by gene targeting, the collection of existing mouse and chick limb mutants offers a parallel approach to elucidate the mechanisms of limb patterning. Characterization of these mutants can lead to the identification of novel genes, as well as structural or regulatory information for previously identified genes (Woychik et al., 1990; Schimmang et al., 1992; Storm et al., 1994). With respect to Hox genes, structural mutations in mice and human have recently been identified. The human synpolydactyly (SPD) mutation results in reductions, fusions and duplications of digits in hands and feet and is associated with a polyalanine expansion in the N terminus of the HOXD-13 gene (Muragaki et al., 1996). This phenotype is similar to a targeted deletion of Hoxd-11, Hoxd-12 and Hoxd-13 in the mouse; therefore, the SPD mutation may result in loss of Hoxd-13 function, and a gain-of-function by suppressing Hoxd-11 and Hoxd-12 activity in the autopod (Zakany and Duboule, 1996). The Hypodactyly (Hd) mutation in mice and the hand-foot-genital (HFG) syndrome in humans lead to reductions in the autopods and are associated with coding mutations in the Hoxa-13 gene (Mortlock et al., 1996; Mortlock and Innis, 1997). Hd/Hd forelimbs and hindlimbs resemble limbs of Hoxa-13 and Hoxd-13 double mutant mice; therefore, the Hd and HFG mutations may result in a primary loss of Hoxa-13 activity and a secondary loss of Hoxd-13 activity (Fromental-Ramain et al., 1996b; Mortlock et al., 1996).

Our prior genetic and physical mapping has suggested that the mouse mutation, Ulnaless (Ul), may represent an allele of the HoxD cluster (Peichel et al., 1996). Ulnaless is a semidominant, radiation-induced mutation resulting in reductions and delays in growth of limb elements, similar to targeted mutations in Hox genes (Davisson and Cattanach, 1990; Peichel et al., 1996). However, Ulnaless differs from loss-of-function mutations in single Hox genes because there are severe reductions of both forelimb and hindlimb zeugopods, and no axial skeletal defects. We show that Ulnaless does not result from a mutation in a HoxD-coding region. Rather, posterior HoxD gene expression is altered in Ulnaless limbs. In the prospective Ulnaless zeugopod, proximal misexpression of the 5′ Hoxd-12 and Hoxd-13 genes results in the inactivation of the more 3′ group 11 genes, which are required for the formation of the radius and ulna (Davis et al., 1995). In the prospective Ulnaless autopod, the reductions of Hoxd-13, Hoxd-12, Hoxd-11 and Hoxa-13 expression are consistent with digit reductions in Ulnaless limbs. Consistent with the absence of axial skeletal defects, expression of HoxD genes is unaltered in the primary axis of Ulnaless embryos. Taken together, these results suggest that the Ulnaless mutation identifies and alters a cis-acting regulatory element(s) that controls HoxD gene expression in the appendicular, but not the primary axis of the embryo.

Ulnaless embryos were generated by one of four mating schemes: (Ul/+ × FVB/N) × FVB/N or C57BL/6J × (Ul/+ × MOLF/Ei) to generate +/+ and Ul/+ embryos; and (Ul/+ × MOLF/Ei) F₁ or (Ul/+ × FVB/N) F₁ intercrosses to generate +/+, Ul/+ and Ul/Ul embryos. For skeletal analysis, Ulnaless homozygote mice were produced from (Ul/+ × MOLF/Ei) F₁ intercrosses. For skeletal analysis of Hoxd-11 trans-heterozygotes, an (Ul/+ × MOLF/Ei) F₁ female was mated to a Hoxd-11^+/− male (Davis and Capecchi, 1994) and a resulting (Ul/+; Hoxd-11^+/−) female and (+/+; Hoxd-11^+/−) male were intercrossed. For skeletal analysis of Hoxd-12 trans-heterozygotes, an (Ul/+ × MOLF/Ei) F₁ female was mated to a Hoxd-12^−/− male (Davis and Capecchi, 1996), and a resulting (Ul/+; Hoxd-12^+/−) female and (+/+; Hoxd-12^+/−) male were intercrossed. All embryos and mice were genotyped at the Ulnaless locus (Peichel et al., 1996) and for the presence of the Hoxd-11 and Hoxd-12 targeted alleles (Davis and Capecchi, 1996) using established PCR assays on DNA from the yolk sacs or tails.

Embryonic, neonatal and adult skeletal stains were prepared as described (Selby, 1987; Jegalian and De Robertis, 1992; Peichel et al., 1996). Left and right forelimbs and hindlimbs of each animal were either measured on photographs taken at fixed magnification, or using an eyepiece micrometer on a Nikon dissecting microscope. Reductions in bone lengths were analyzed using the one-tailed t-test: two sample assuming unequal variances on Microsoft Excel 5.0. A P-value of <0.025 was considered significant.

PCR products were amplified from Ul/Ul and 101/H genomic DNA. All primer sequences and PCR conditions used are available on our laboratory homepage (www.molbio.princeton.edu/vogt/vogt.html). The products were cloned into the pCR2.1 vector and sequenced using the vector-specific primers TAF and TAR (Peichel et al., 1996).

Whole-mount in situ hybridization using digoxigenin-labeled probes was performed as described by Riddle et al. (1993). The probes were Hoxd-1 (Frohman and Martin, 1992), Hoxd-3 (Bedford et al., 1995), Hoxd-9 and Hoxd-10 (Duboule and Dolle, 1989), Hoxd-11 and Hoxd-12 (Izpisua-Belmonte et al., 1991), Hoxd-13 (Dolle et al., 1991), Hoxa-11 and Hoxa-11-as (Hsieh-Li et al., 1995), Hoxa-13 (L. Post and J. Innis), Hoxc-11 (Peterson et al., 1994) and Shh (Chang et al., 1994). The Hoxd-4 probe was constructed by amplifying a 510 bp fragment in the homeodomain and 3′ UTR (Featherstone et al., 1988). The fragment was cloned into the pCR2.1 vector. To make a 348 bp antisense probe, the plasmid was linearized with BglII and transcription was performed with T7 polymerase.

The refined genetic mapping of the HoxD cluster to the Ulnaless locus and the similarities of the Ulnaless limb phenotype to gene-targeted mutations in Hox genes, prompted a molecular search for an alteration in HoxD cluster gene structure or expression (Fig. 1A,B; Peichel et al., 1996). A survey across the HoxD cluster of +/+, Ul/+ and Ul/Ul DNA by genomic Southerns showed no observable change in the structure of the locus (Peichel et al., 1996; data not shown). PCR amplification of the entire coding regions of Hoxd-1, Hoxd-3 and Hoxd-4 from genomic DNA of Ul/Ul and the 101/H parental chromosome revealed no alterations. The 5′ HoxD genes were examined for a structural change by DNA sequencing. Each exon and adjoining splice junctions of Hoxd-8, Hoxd-9, Hoxd-10, Hoxd-11, Hoxd-12, Hoxd-13 and Evx-2 were sequenced in both parental 101/H and Ul/Ul genomic DNA and no significant alterations were found (Fig. 1A).

The absence of a structural change in HoxD-coding sequences suggested that the Ulnaless mutation may be regulatory. To date, three local cis-acting regulatory elements for 5′ HoxD genes have been defined (Zappavigna et al., 1991, 1994; Renucci et al., 1992; Gerard et al., 1993, 1996; Beckers et al., 1996). These elements were sequenced in 101/H and Ul/Ul DNA, with no differences (Fig. 1A). Despite these results, evidence of altered HoxD gene expression would support the notion that Ulnaless affects an undefined regulatory element(s) required for the temporal and spatial coordination of HoxD gene expression.

Because alterations of the Ulnaless limb chondrogenic structures are evident at embryonic day 14, we focused our expression analysis of HoxD genes on earlier embryological stages (Fig. 1C). Examination of the 3′ HoxD genes at embryonic day 11 in Ul/+ limb buds revealed that relative to wild type, expression of Hoxd-1, Hoxd-3 and Hoxd-4 was not altered (Fig. 1D).

In the developing limb bud, the 5′ HoxD genes (Hoxd-9 through Hoxd-13) are activated in a colinear fashion that reflects their relative position in the cluster (Dolle et al., 1989; Nelson et al., 1996). In wild-type forelimb and hindlimb buds, Hoxd-13 expression is restricted to the distal region of the prospective autopod of the limb. Examination of Ul/+ and Ul/Ul embryos revealed that Hoxd-13 expression is altered in two ways. Proximally, in the prospective zeugopod, there is ectopic Hoxd-13 expression in mutant limb buds, such that the domain of Hoxd-13 expression is diffusely expanded to proximal sites (Fig. 2A). This proximal limb bud expression is indicative of a Hoxd-13 gain-of-function in the region that will give rise to the radius and ulna. Distally, in the prospective autopod, there is a reduction in the levels of Hoxd-13 expression in mutant limb buds (Fig. 2A). This reduced distal limb bud expression is indicative of a Hoxd-13 reduction-of-function in the region that will give rise to the digits.

Having detected an alteration in the most 5′ HoxD gene, we next extended our examination to include the other 5′ HoxD genes. In wild-type limb buds, Hoxd-12 is strongly expressed in a distal domain, corresponding to the prospective autopod, with weaker expression in a proximal and posterior domain. Examination of Hoxd-12 expression in Ulnaless embryos revealed it to be altered in a fashion similar to Hoxd-13. The Hoxd-12 domain is expanded proximally and anteriorly in Ul/+ and Ul/Ul limb buds (Fig. 2B). Distally, there is a reduction in Hoxd-12 expression in Ul/Ul limb buds (Fig. 2B). These alterations in Hoxd-13 and Hoxd-12 expression are more readily discerned at later stages of limb development (Herault et al., 1997).

Gene targeting experiments suggested a primary role for group 11 genes in the patterning of the zeugopod (Davis et al., 1995). The HoxD group 11 paralogue, Hoxd-11 is expressed in two domains, with a stronger proximal domain in the prospective zeugopod and a relatively weaker distal domain in the prospective autopod. We examined expression of Hoxd-11 in +/+, Ul/+ and Ul/Ul embryos, and observed a reduction of Hoxd-11 in the proximal domain of Ul/+ and Ul/Ul forelimbs and Ul/Ul hindlimbs. The distal domain of Hoxd-11 is also reduced in Ul/Ul forelimbs (Fig. 2C). Therefore, the Ulnaless mutation appears to result in Hoxd-11 reduction-of-function in both the prospective zeugopod and autopod.

In wild-type forelimbs and hindlimbs, Hoxd-10 is expressed in a pattern similar to Hoxd-11, except the proximal and the distal domain are of equal intensity. In Ul/Ul forelimbs and hindlimbs, the proximal expression of Hoxd-10 is reduced relative to the unaltered distal expression (Fig. 2D). In contrast to the more 5′ HoxD genes, Hoxd-9 expression is indistinguishable in +/+ and Ul/+ proximal limb buds, in the region of the prospective stylopod (Fig. 2E).

The alterations in the expression of multiple HoxD genes in the limb buds of Ulnaless mutants may be due to a cis-acting mutation that affects each HoxD gene independently, or alternatively, a primary effect on one HoxD gene (e.g. Hoxd-13), with secondary consequences for the other Hox genes. These alternatives are difficult to distinguish by examining expression within the HoxD cluster. Therefore, we examined the potential trans-acting consequences of the Ulnaless mutation on the unlinked paralogous group 11 genes, Hoxa-11 and Hoxc-11, which are also expressed in the prospective zeugopod. Hoxa-11 is expressed in a proximal stripe in both forelimbs and hindlimbs. Expression in this domain is highest at its anterior and posterior aspects, and weaker in the middle of the limb. In Ul/+ and Ul/Ul forelimbs, Hoxa-11 anterior expression is reduced and posterior expression is barely detectable (Fig. 3A, data not shown). Therefore, the Ulnaless mutation results in altered expression of an unlinked HoxA gene. Antisense transcription of the Hoxa-11 gene in the distal limb has been reported (Hsieh-Li et al., 1995). There is no change in the pattern or levels of expression of antisense transcripts in Ul/+ limbs (Fig. 3B). The effects on Hoxa-11 and Hoxd-11 transcription in Ulnaless are consistent with the reported defects of radius and ulna in the Hoxa-11; Hoxd-11 double mutant (Davis et al., 1995). However, the effects of Ulnaless on the hindlimb zeugopod are more severe than in the Hoxa-11; Hoxd-11 double mutant, suggesting that Ulnaless may also interfere with Hoxc-11 function in the hindlimb. In wild-type embryos, Hoxc-11 is expressed in the proximal hindlimb, but not the forelimb. Unlike the effects of Ulnaless on Hoxd-11 and Hoxa-11 RNA distribution, Hoxc-11 RNA expression appears unaltered in Ul/+ hindlimbs relative to wild type, suggesting that any effects of Ulnaless on Hoxc-11 are most likely at the functional level (Fig. 3C).

Because we observe a reduction in the expression of Hoxd-13, Hoxd-12 and Hoxd-11 in the prospective autopod of Ulnaless limb buds, we also examined expression of Hoxa-13, which is the only gene of the HoxA, HoxB and HoxC clusters expressed in the prospective autopod mesoderm (Haack and Gruss, 1993; Peterson et al., 1994; Nelson et al., 1996). In wild-type limb buds, Hoxa-13 is expressed along the entire anterior-posterior length of the distal limb. In Ul/+ forelimbs and hindlimbs, the proximal extent of the Hoxa-13 expression domain is unaltered relative to wild type; however, the levels of expression within the distal domain appear reduced in the prospective forelimb and hindlimb autopod (Fig. 3D).

We also examined the expression of Sonic hedgehog (Shh) in Ulnaless embryos, as perturbation of Shh expression would suggest that the primary defect in Ulnaless acts upstream of the HoxD genes (Riddle et al., 1993; Laufer et al., 1994; Chan et al., 1995; Parr and McMahon, 1995; Nelson et al., 1996). At all stages examined, no alterations in Shh expression in Ul/+ fore or hindlimbs were observed relative to wild type, consistent with the Ulnaless mutation altering limb patterning at the level of HoxD genes (data not shown).

Functional effects of the deregulation of Hox genes in Ulnaless limb buds on skeletal patterning were investigated in two ways: by comparing the skeletal defects in Ulnaless to those of known Hox mutations and by placing Ulnaless in trans to genetargeted alleles of Hoxd-11 and Hoxd-12 (Davis and Capecchi, 1994, 1996).

In the zeugopod, the Ulnaless mutation appears to result in a gain-of-function. Misexpression of Hoxd-13 in mouse by gene targeting and in chicken by retroviral infection also results in reductions of the zeugopod, although not as severe as in Ulnaless (van der Hoeven et al., 1996; Goff and Tabin, 1997). In addition, mice lacking both Hoxa-11 and Hoxd-11 have drastic reductions of the ulna and radius, and more subtle reductions of the fibula and tibia (Davis et al., 1995). Targeted mutations of Hoxd-11 or Hoxd-12 alone do not result in reductions of the zeugopod (Davis and Capecchi, 1994, 1996; Kondo et al., 1996). In (Ul/+, Hoxd-11^+/−) and (Ul/+, Hoxd-12^+/−) trans-heterozygotes, the Ul/+ zeugopod phenotype, not the more severe Ul/Ul phenotype, is observed (Fig. 1B; data not shown). Therefore, the Ulnaless zeugopod phenotype cannot simply be the result of a loss-of-function mutation in either Hoxd-11 or Hoxd-12, suggesting that Ulnaless results in a gain-of-function in the zeugopod.

In contrast to the putative gain-of-function effects on the zeugopod, we found evidence for allelism of Ulnaless and HoxD loss-of-function mutations in the autopod. We performed alizarin red skeletal stains on adult limbs and measured individual bones of the digits of both forelimbs and hindlimbs (Table 1). Both Ul/+ and Ul/Ul limbs had reductions of specific bones of the digits, consistent with loss-of-function Hox phenotypes created by gene targeting (Fig. 4; Table 1; Dolle et al., 1993; Davis and Capecchi, 1994, 1996; Fromental-Ramain et al., 1996b; Kondo et al., 1996). There is a severe reduction or absence of phalange 2 (P2) of digits II and V in both the forelimbs and hindlimbs of Ul/Ul mice (Fig. 4A). These bones are the last to form in the limb (Shubin and Alberch, 1986) and they are reduced or absent in all Hox targeted mutations. In addition, there is a reduction in the overall size of the Ul/Ul hindlimb autopod (Fig. 4C; Table 1). In particular, there is a reduction of the metatarsal of digit I in both Ul/+ and Ul/Ul hindlimbs (Fig. 4C; Table 1). This is a bone that is specifically affected in Hoxd-13^−/− hindlimbs (Dolle et al., 1993; Davis and Capecchi, 1996). Observation of Hox-like phenotypes in Ulnaless autopods suggest that the reduction of distal Hox RNA expression has phenotypic consequences. Both Hoxd-11, Hoxd-12 and Hoxd-13 triple mutant mice and Hoxa-13; Hoxd-13 double mutant mice, created by gene targeting, have extreme reductions of forelimb and hindlimb autopods, consistent with Ulnaless resulting in partial loss-of-function of these genes (Fromental-Ramain et al., 1996b; Zakany and Duboule, 1996).

Analysis of the autopod phenotypes of (Ul/+, Hoxd-11^+/−) and (Ul/+, Hoxd-12^+/−) trans-heterozygotes is consistent with the Ulnaless mutation leading to a reduction of function in Hoxd-11 and Hoxd-12 in the autopod. There are two key features to note in both the forelimbs and hindlimbs of these mice. First, the reductions in trans-heterozygotes encompass all of the phenotypes observed in Hoxd-11^−/− or Hoxd-12^−/− limbs (Fig. 4B,D; Table 1). This suggests that the Ulnaless mutation results in a reduction of both Hoxd-11 and Hoxd-12 function in the autopod. Second, there are additional reductions in trans-heterozygotes that are not present in either Ul/+, Hoxd-11^−/−, or Hoxd-12^−/− limbs alone (Fig. 4B,D; Table 1). Therefore, the Ulnaless mutation is likely to result in a partial loss-of-function of multiple Hox genes in the autopod, consistent with the observed reduction of Hoxd-13 and Hoxa-13 in the distal limbs of Ulnaless embryos.

Expression of HoxD genes in both limbs and genitalia appear to be coordinated (Beckers et al., 1996; van der Hoeven et al., 1996). Although both Ul/+ and Ul/Ul females are fertile, there is a reduction in Ul/+ male fertility and Ul/Ul males have not successfully bred (Peichel et al., 1996). The expression of Hoxd-11 and Hoxd-12 in genital buds is unaltered relative to wild type in Ul/+ and Ul/Ul; however, Hoxd-13 expression is greatly reduced in Ul/+ and undetectable in Ul/Ul (Fig. 5). Consistent with this reduction in Hoxd-13 expression in embryos, penian bones from adult Ul/+ males are reduced in width and clefted at the proximal end (Fig. 5D). The Ul/+ phenotype is reminiscent of penian bone defects in Hoxd-13^−/− males (Dolle et al., 1993).

In contrast to the limbs and genitalia, we observed no alteration of Hox gene expression along the main body axis of the Ul/+ and Ul/Ul embryos (Figs 2, 3, 5, data not shown). This is consistent with the absence of any changes in the number or morphology of vertebrae along the axial skeleton of Ul/+ or Ul/Ul mice (data not shown). In addition, (Ul/+, Hoxd-11^+/−) and (Ul/+, Hoxd-12^+/−) trans-heterozygotes did not uncover any abnormalities of the axial skeleton (data not shown). However, in this genetic background in Hoxd-11^−/− mice, we find the previously reported homeotic transformations in the lumbar-sacral region (Davis and Capecchi, 1994).

The Ulnaless mutation provides insight into the complex interactions of Hox genes within the limb. We propose that a functional consequence of the ectopic misexpression of Hoxd-13 and Hoxd-12 in the presumptive zeugopod is the reduction of Hoxd-10, Hoxd-11 and Hoxa-11 transcripts (Fig. 6). A functional hierarchy of Hox genes has previously been observed both in Drosophila (phenotypic suppression) and vertebrates (posterior prevalence). Both transcriptional and functional suppression of anterior Hox genes by posterior Hox genes occurs in Drosophila (Bachiller et al., 1994; Duboule and Morata, 1994; Gould et al., 1997). However, there is evidence that transcriptional repression is a secondary effect of functional inactivation due to autoregulation (Bachiller et al., 1994; Duboule and Morata, 1994). Although we observe a reduction in Hoxd-10, Hoxd-11 and Hoxa-11 transcripts, we are unable to determine if it is due to a direct or indirect mechanism. In the accompanying paper, reductions in Hoxd-11 and Hoxd-10 transcripts are not observed in Ul/+ and (Ul/+, HoxD^Del/+) trans-heterozygote limb buds and Hoxa-11 RNA levels were not examined (Herault et al., 1997). Differences in genetic constitution (Ul/+ versus Ul/Ul genotypes), background modifiers, or staging of embryos may be responsible for these results.

The reductions of the Ulnaless forelimb zeugopod resemble those obtained in Hoxa-11; Hoxd-11 mutant mice (Davis et al., 1995). Reductions of the radius are less severe in Ulnaless forelimbs than in the Hoxa-11; Hoxd-11 mutant forelimbs, which may be accounted for by the remaining Hoxa-11 expression in the anterior of Ulnaless limb buds (Fig. 3A). In contrast to the forelimbs, reductions of the Ulnaless hindlimb zeugopod are more severe than in the Hoxa-11; Hoxd-11 double mutants (Davis et al., 1995), an observation that we attribute to functional suppression of Hoxc-11 by Hoxd-13 and Hoxd-12.

We propose that the misexpression of Hoxd-13 and Hoxd-12 in Ulnaless results in suppression of group 11 genes, either through transcriptional repression or functional inactivation. These results are consistent with previous studies of Hoxd-13 misexpression in the limb. Phenotypic reductions of the forelimb zeugopod were observed by misexpression of Hoxd-13 in the Hoxd-11 pattern through gene transposition in mice (van der Hoeven et al., 1996). In addition, retroviral expression of Hoxa-13 or Hoxd-13 throughout chick limb buds results in reductions of the zeugopod (Yokouchi et al., 1995; Goff and Tabin, 1997). In these studies, the phenotypic effects are hypothesized to result from functional inactivation of group 11 genes. However, the Ulnaless zeugopod reductions are more severe than those obtained by Hoxd-13 misexpression through gene transposition or retroviruses (van der Hoeven et al., 1996; Goff and Tabin, 1997). This suggests that the timing, levels and patterns of misexpression in Ulnaless may be critical to cause the postulated transcriptional and functional inactivation, and for the severity of the zeugopod phenotype.

In Ulnaless, Hox targeted mutations and retroviral gain-of-function studies, altered Hox expression results in reduced and delayed growth of specific limb elements along the proximaldistal axis. In addition, X-irradiation of stage 20 chick limb buds depletes the overall cell population and the resulting limbs have reductions of the zeugopod, similar to Ulnaless (Wolpert et al., 1979). Therefore, a consequence of the Ulnaless mutation and other Hox mutations may be to reduce the number of cells available to condense into the cartilage precursors of the zeugopod and autopod and/or to alter the sub-sequent growth of these cells (Dolle et al., 1993; Davis and Capecchi, 1996; Zakany and Duboule, 1996; Goff and Tabin, 1997). Retroviral misexpression of Hoxa-13 in chick limb buds appears to alter initial precartilage condensation by changing the adhesive properties of cells, resulting in reductions of the normal precartilage condensations and ectopic cartilage condensations (Yokouchi et al., 1995). In contrast, we do not observe similar ectopic cartilage condensations in Ulnaless limbs at a comparative stage (Fig. 1C).

We demonstrate that Ulnaless alters the regulation of HoxD genes in three domains: (1) in the proximal limbs, there is mis-expression of Hoxd-13 and Hoxd-12, (2) in the distal limbs, there is reduction of Hoxd-13, Hoxd-12 and Hoxd-11 expression and (3) in the genital bud, there is reduction of Hoxd-13 expression. These observations lead us to propose that the Ulnaless mutation results in the loss-of-function of a cis-acting element(s) that regulates expression of the 5′ HoxD genes in the limbs and genitalia, but not the primary embryonic axis (Fig. 6). There are several lines of evidence to support this model. First, genes from both ends of the HoxD cluster do not recombine with the Ulnaless locus in 1564 N2 animals (Peichel et al., 1996), indicating that the Ulnaless mutation is tightly linked to the HoxD cluster. Second, in the Ulnaless back-ground, there are no mutations in the coding sequence of Hoxd-8 through Evx-2. In addition, there are no rearrangements of the HoxD gene cluster in Ul/Ul genomic DNA (data not shown; Peichel et al., 1996). Third, altered expression of 5′ HoxD genes (Hoxd-10 through Hoxd-13), but not 3′ genes (Hoxd-1 through Hoxd-9), is observed exclusively in the limbs and genitalia. Fourth, Hoxd-13 and Hoxd-12 are no longer expressed according to their position in the HoxD cluster. Although the initial activation of 5′ HoxD genes was not examined, we favor the model that the Ulnaless mutation leads to the deregulation of HoxD gene colinearity. Recent studies in the chick have shown that the later expression of HoxD genes in the limb is complex (Nelson et al., 1996; Vargesson et al., 1997). Therefore, it is possible that the misexpression of posterior HoxD genes in Ulnaless results from the alteration of a regulatory element required for a later phase of HoxD expression in the limb (Nelson et al., 1996), or from a failure of proximal cells to stop expressing Hoxd-13 (Vargesson et al., 1997).

The nature and location of the elements required for regulation of the HoxD cluster in the axial and appendicular axes are undefined. Therefore, several possibilities exist to explain the deregulation of posterior HoxD expression caused by the Ulnaless mutation. There may be a distal control region that regulates all the 5′ genes in the HoxD cluster though long-range interactions, as in the case of the globin LCR (Hanscombe et al., 1991; Wijgerde et al., 1995; van der Hoeven et al., 1996). Alternatively, there may be intergenic chromatin domain boundaries that function to restrict the domains of expression of Hox genes, as in the Drosophila bithorax complex (Hagstrom et al., 1996; Zhou et al., 1996). We favor the first possibility for several reasons. First, the necessary elements for regulation of limb expression appear to lie outside of the HoxD cluster, downstream of Evx-2, because deletion of the region encompassing Hoxd-11, Hoxd-12 and Hoxd-13 results in proper expression of Hoxd-10 and Evx-2 in the limbs (Zakany and Duboule, 1996). Second, the region between Hoxd-13 and Evx-2 is not sufficient to direct limb expression of a Hoxd-9 or Hoxd-11 transgene (van der Hoeven et al., 1996). Third, our extensive Southern analysis of the genomic structure of the Ulnaless locus and sequencing of the intergenic regions (C. L. P. and T. F. V., unpublished data), has failed to yield any alterations of the HoxD intergenic regions.

Alternatively, it is possible that the Ulnaless mutation does not alter a cis-acting element in the HoxD cluster, but alters a yet unidentified tightly linked trans-acting gene that regulates expression of the HoxD cluster. We consider this less likely, in part because the alterations of HoxD expression in Ul/Ul limbs are stronger than in Ul/+ limbs, requiring a dosage effect of the putative trans-acting factor.

Recently, insight into the mechanism of HoxD regulation has been gained through transgenic mouse experiments. Random integration of Hoxd-9 or Hoxd-11 and their respective flanking sequences recapitulate many aspects of axial expression, but not appendicular expression. However, moving either Hoxd-9 or Hoxd-11 and flanking sequences to a new position within the HoxD cluster (between Hoxd-13 and Evx-2) results in limb expression of these genes (Renucci et al., 1992; Gerard et al., 1993; Beckers et al., 1996; van der Hoeven et al., 1996). In their new positions, both Hoxd-9 and Hoxd-11 are initially repressed in the proximal limb, but later are expressed in their normal pattern. In addition, they are also expressed in the distal limbs and genitalia, unlike endogenous Hoxd-9 or randomly integrated Hoxd-11 transgenes (Gerard et al., 1993; Beckers et al., 1996). Therefore, expression of HoxD genes in limbs and genitalia is initially dependent upon position within the cluster and is controlled by a global regulatory mechanism. Local cis-acting elements are initially subservient to this higher order regulation and can be shared between genes (Gerard et al., 1996; van der Hoeven et al., 1996).

Morphological differences between organisms may in part correspond to changes in Hox gene expression (Carroll, 1995). In the Ulnaless mutation, deregulation of HoxD gene expression also leads to morphological changes and may therefore be acting as an agent of evolutionary change. A feature of the Ulnaless mutation is that it alters expression of Hox genes in the limbs, but not in the primary embryonic axis. This suggests that the mechanisms regulating HoxD expression differ in the primary and secondary embryonic axes, for example through the action of different members of the Polycomb- and trithorax-group genes (Cohn et al., 1997; Schumacher and Magnuson, 1997). Uncoupling of the regulatory mechanisms of Hox gene expression in these two regions of the embryo may have allowed for evolutionary diversity in the limbs. Therefore, Ulnaless provides a valuable opportunity to define the molecular mechanisms contributing to the regulation of HoxD gene expression, the uncoupling of the regulatory mechanisms for axial and appendicular patterning, and the hierarchy of Hox gene function.

We thank D. Duboule and Y. Herault for helpful discussions and sharing of unpublished data. We are grateful to A. P. Davis and M. Capecchi for the gift of the Hoxd-11 and Hoxd-12 mutant mice and discussions; A. Awgulewitsch, P. Beachy, P. Chambon, D. Chang, M. Featherstone, M. Frohman, M. Bedford, J. Innis and S. Potter for probes; A. Arthur, J. Ehrlich and I. Ivanovska for assistance; and K. Hagstrom, V. Prince, P. Schedl, S. Tilghman and E. Wieschaus for critical reading of the manuscript. We thank all the members of the Vogt lab and C. Abbott for contributing ideas and encouragement throughout the course of this work. Special thanks to R. F. and T. R. Vogt for art supplies. This research was supported by grants HD-30707 from the National Institutes of Health and DB-143 from the American Cancer Society to T. F. V.