A function for all posterior Hoxd genes during digit development?


  • Saskia Delpretti,

    1. National Research Centre “Frontiers in Genetics,” School of Life Sciences, Ecole Polytechnique Fédérale, Lausanne, Switzerland
    Search for more papers by this author
  • Jozsef Zakany,

    1. National Research Centre “Frontiers in Genetics,” Department of Genetics and Evolution, University of Geneva, Geneva, Switzerland
    Search for more papers by this author
  • Denis Duboule

    Corresponding author
    1. National Research Centre “Frontiers in Genetics,” School of Life Sciences, Ecole Polytechnique Fédérale, Lausanne, Switzerland
    2. National Research Centre “Frontiers in Genetics,” Department of Genetics and Evolution, University of Geneva, Geneva, Switzerland
    • Department of Genetics and Evolution, University of Geneva, Geneva, Switzerland

    Search for more papers by this author


Background: Four posterior Hoxd genes, from Hoxd13 to Hoxd10, are collectively regulated during the development of tetrapod digits. Besides the well-documented role of Hoxd13, the function of the neighboring genes has been difficult to evaluate due to the close genetic linkage and potential regulatory interferences. We used a combination of five small nested deletions in cis, involving from two to four consecutive genes of the Hoxd13 to Hoxd9 loci, in mice, to evaluate their combined functional importance. Results: We show that deletions leading to a gain of function of Hoxd13, via regulatory re-allocation, generate abnormal phenotypes, in agreement with the dominant negative role of this gene. We also show that Hoxd10, Hoxd11, and Hoxd12 all seem to play a genuine role in digit development, though less compelling than that of Hoxd13. In contrast, the nearby Hoxd9 contributed no measurable function in digits. Conclusions: We conclude that a slight and transient deregulation of Hoxd13 expression can readily affect the relative lengths of limb segments and that all posterior Hoxd genes likely contribute to the final limb morphology. We discuss the difficulty to clearly assess the functional share of individual genes within such a gene family, where closely located neighbors, coding for homologous proteins, are regulated by a unique circuitry and all contribute to shape the distal parts of our appendages. Developmental Dynamics 241:792–802, 2012. © 2012 Wiley Periodicals, Inc.


The development of the vertebrate limb has been a model of choice in experimental molecular biology over the past 20 years. Because of the accessibility of this structure, even in its earlier stages, the possibility to manipulate it either manually or genetically, a collection of spontaneous mutations and the huge phenotypic variations observed amongst tetrapods, progress in understanding the underlying molecular mechanisms and signaling pathways has been spectacular (see reviews by, e.g., Allard and Tabin, 2009; Towers and Tickle, 2009; Zeller et al., 2009). The presence and roles of signaling centers (the ZPA and AER) in the coordination of the anterior to posterior (AP) and proximal to distal (PD) axes have been largely described, as well as those molecules that mediate these functions (e.g., Zeller, 2010).

In particular, the formation of different segments along the PD axis (from the shoulder to the digits), usually referred to as “proximo-distal patterning,” appears to depend on the continuous presence, from the earliest stages, of opposed signaling molecules, the combination of which instruct particular cells upon their fates along this axis (Mariani et al., 2008; Cooper et al., 2011; Rosello-Diez et al., 2011), rather than from a more deterministic, cell-autonomous process (Dudley et al., 2002; discussed in Tabin and Wolpert, 2007). These different PD segments are labeled by the expression of specific homeobox genes, such as Meis1/2 for the stylopod (arm; Mercader et al., 2009) and two members of the Hox gene family, Hoxa11 for the zeugopod (forearm; Nelson et al., 1996) and Hoxa13 for the autopod (hand; Fromental-Ramain et al., 1996b).

The function of Hox genes during vertebrate limb development has been long documented (reviewed in Zakany and Duboule, 2007), in particular for genes belonging to the HoxA and HoxD clusters. While Hoxd genes are necessary for proper formation of both the AP and PD axes (Dolle et al., 1989; Tarchini and Duboule, 2006), Hoxa genes mostly impact upon the PD patterning mechanism. Their critical function along the PD axis was demonstrated genetically, as combined deletions of both Hoxa11 and Hoxd11 were shown to delete most of the zeugopod (Davis et al., 1995), whereas functional abrogation of both Hoxa13 and Hoxd13 induced an agenesis of the autopod, the most distal piece of the limb (Fromental-Ramain et al., 1996b; Kondo et al., 1997).

This latter function of group 13 Hox genes in the developing digits has been particularly thoroughly studied, as it was proposed to be closely associated with the emergence of autopods (in particular of digits) in the course of tetrapod evolution (Sordino et al., 1995). More recently, studies of Hox gene expression in non-model species have brought new elements and various views on this issue (see Wagner and Vargas, 2008; Shubin et al., 2009; Woltering and Duboule, 2010). On the other hand, studies on the regulatory mechanisms underlying the transcription of Hoxd13 have revealed its key function in the morphogenesis (and hence the identification) of digits (Montavon et al., 2008; Wagner and Vargas, 2008). These studies have also highlighted the strong quantitative component to be considered in this system (Montavon et al., 2008, 2011).

The expression of Hoxd13 during the early development of the autopod is accompanied by an almost identical transcription pattern for three other neighboring Hox genes (Hoxd12, Hoxd11, and Hoxd10). The co-expression of these four genes into the same presumptive digit domain, except for the future thumb where Hoxd13 is the only gene expressed, is due to specific regulatory modalities, which leads to the sharing of remote enhancers function by the four contiguous promoters (Montavon et al., 2008), with decreasing intensity from Hoxd13 to Hoxd10. However, it has been difficult to evaluate whether or not these three neighboring genes exert any function (similar or different from Hoxd13) during digit development, precisely due to both their close “regulatory clustering” and the demonstrated existence of redundant or compensatory mechanisms at work in this gene family. While the classical inactivation of Hoxd13 generated a clear phenotype in digits (Dolle et al., 1993; Davis and Capecchi, 1996), the inactivation of either Hoxd12 (Davis and Capecchi, 1996; Kondo et al., 1996) or Hoxd11 revealed only a minor morphological alteration in digits. On the other hand, the loss of function of Hoxd10 had no visible impact upon digital morphologies (Davis and Capecchi, 1994; Carpenter et al., 1997). In all these cases, however, the selection cassette was left in place within the endogenous locus, which may have interfered with regulatory fine-tuning (e.g., Olson et al., 1996).

Using a different approach, combined inactivation in cis have indicated that Hoxd12 and Hoxd11 may have an effect upon digit morphology, since the triple deletion in cis of the Hoxd13 to Hoxd11 transcription units elicited a phenotype clearly stronger than the mere inactivation of Hoxd13 (Zakany and Duboule, 1996). However, it was subsequently shown that removing genes, rather than inactivating them by DNA insertion, would lead to regulatory reallocations, making functional inferences difficult. For instance, the deletion of the Hoxd13 gene did not induce a strong phenotypic alteration, unlike its inactivation by insertional mutagenesis, a phenomenon due to the concurrent up-regulation of Hoxd12, in the deleted allele, which could rescue the phenotype (Kmita et al., 2002a). Therefore, in the absence of Hoxd13, Hoxd12 was able to functionally substitute. However, in the same study, the over-expression of Hoxd11 did not rescue the phenotype induced by deleting both Hoxd13 and Hoxd12 (Kmita et al., 2002a).

An additional parameter to consider in this context is that the Hoxd13 product acts as a dominant negative over its neighboring genes, as shown for example by the mutant mice Ulnaless, where a gain of function of Hoxd13 in the forearm phenocopies the effect of a loss of function of group 11 genes (Davis et al., 1995; Herault et al., 1997). Therefore, any mis-regulation of Hoxd13, either induced by regulatory re-allocations or by the presence of a selectable marker in the gene cluster, may elicit a phenotype not directly associated with the genetic modification under investigation, but due to combined loss of function of other Hox genes.

In this work, we used a set of deletions within the HoxD cluster, to try and assess whether or not posterior Hoxd genes other than Hoxd13 have a function during the development of digits, under physiological conditions. We used deletions where the Hoxd13 locus always remains in place such as to avoid phenotypic alterations due to strong regulatory re-allocations of other genes (Kmita et al., 2002a; Montavon et al., 2008). We report abnormal phenotypes in adults, involving the proximal parts of the limbs and associate them with an early gain of function of Hoxd13, even when this gain of function is transient and occurs very early on. Digital anomalies, however, are obtained even in the absence of variation in Hoxd13 transcription, suggesting a genuine function for other posterior Hoxd genes during digit development, even for Hoxd10, whose inactivation by insertion had not previously revealed such an effect.


Deletions Within the HoxD Cluster Result in Hoxd13 Gain of Expression in Early Limb Buds

In order to evaluate a potential function either for Hoxd12 or for more 3′-located neighbors (Hoxd11, Hoxd10) during digit development, we screened our collection of deletion mutants (Fig. 1) for configurations where the associated deletion would not elicit a concomitant gain of expression of Hoxd13, which would make any phenotypic interpretation difficult, due to its dominant negative role (Duboule and Morata, 1994). To this aim, we performed in situ hybridization using both Hoxd13 and Hoxd12 probes at embryonic day 9.5 (E9.5), when the early limb bud starts to protrude from the lateral mesoderm of the main body axis. In wild-type specimens of the same age, these two genes are not yet transcribed at a detectable level (Fig. 2A), as expected from their position at the centromeric (“posterior”) extremity of the HoxD complex (Nelson et al., 1996; Tarchini and Duboule, 2006). Expression is scored soon after, with transcripts appearing in cells located at the postero-distal aspect of the growing bud, and is fully established in E11.5 limb buds. At this stage, the limb bud has transformed into two morphologically distinct sub-structures: the proximal and distal parts, which will further generate the arm and the hand-plate, respectively.

Figure 1.

Mutant lines used in this work. The HoxD cluster is depicted on top with a color code for Hoxd genes, schematized here as two-exon genes. Genes orthologous to the Drosophila Abd-B gene (“posterior” genes), from Hoxd13 to Hoxd9, are shown in gray, whereas Hoxd8 to Hoxd1 are in black. The various mutant strains used in this work are shown below.

Figure 2.

Early gain of expression of posterior Hoxd genes is not maintained. Wild-type and mutant configurations are depicted on top, with the gene investigated in red. Genes orthologous to the Drosophila Abd-B gene (groups 9 to 13) are shown in gray. A–F: Early expression pattern of Hoxd13 in wild-type and mutant lines. While Hoxd13 is not detected in early growing wild-type limb buds (A), it is expressed prematurely in most mutant lines (B–F). A strong transcription of Hoxd13 is observed in del(12-9) (B), del(11-9) (C), and del(11-10) (D), whereas a weaker signal is seen in del(12-11) (E). By contrast, Hoxd13 is not detected in del(12-10) (F). G–L: Expression patterns of both Hoxd13 (G–I) and Hoxd12 (J–L) in control and mutant E12.5 limbs. In wild-type conditions, Hoxd13 is expressed in the developing distal limb exclusively (G). This pattern is maintained in mutant lines, even in those showing a strong gain of expression in the early bud such as del(12-9) (H) and del(11-9) (I). Therefore, the early gain of function of Hoxd13 in these mutant configurations is not maintained at later stages of development. Wild-type Hoxd12 is expressed in the distal limb, with the exception of presumptive digit I, but is not detected in the proximal limb (J). Similarly to Hoxd13, Hoxd12 is not detected in the proximal limb of mutant lines, as shown in del(11-9) (K). Rather, it is down-regulated in the distal limb, as observed in del(11-9) (K) and del(10-9) (L).

When the del(12-9), del(11-9), and del(11-10) mutant stocks were analyzed, a strong expression of Hoxd13 was detected in limb buds, as early as E9.5, i.e., at a stage where this gene would normally remain silent. This abnormal ectopic expression of Hoxd13 was strong and involved most of the incipient bud (Fig. 2B–D). In the del(12-11), however, ectopic Hoxd13 transcripts were present in much lower amounts (Fig. 2E, arrowhead). Interestingly, ectopic Hoxd13 transcripts were not detected in the del(12-10) configuration, at least by using in situ hybridization (Fig. 2F, arrowhead). In the two deletions where Hoxd12 was still present, i.e., the del(11-9) and del(11-10) mutant stocks, Hoxd12 transcripts were also detected prematurely, as for Hoxd13.

Ectopic Hoxd13 Transcription Is Not Maintained

At later stages of wild-type limb development (E12.5), Hoxd13 is strongly expressed in presumptive digits I to V (Fig. 2G). The Hoxd12 expression territory is similar, yet it does not include presumptive digit I (Montavon et al., 2008) (Fig. 2J) and neither Hoxd13 nor Hoxd12 is expressed in the proximal part of the limb, i.e., in the presumptive domain for the development of the zeugopod (Fig. 2G and J). At this stage, in the mutant lines where Hoxd13 and Hoxd12 were gained in the early limb bud, the proximal limb was devoid of posterior Hoxd gene transcripts, as either with Hoxd13 in del(12-9) and del(11-9) (Fig. 2H-I), or with Hoxd12 in del(11-9) (Fig. 2K). This indicated that the de-regulation of posterior gene expression observed in the early limb bud was not maintained subsequently, as had been already described in other mutant configurations where Hoxd13 was activated too early (Zakany et al., 2004).

Unexpectedly, the steady-state level of Hoxd12 mRNAs in presumptive digits was reduced to approximately 20% of its initial level in del(11-9) (Fig. 2K). A similar (yet not as severe) down-regulation of Hoxd12, also affecting the levels of more anterior Hoxd gene transcripts, was scored in presumptive del(10-9) mutant digits, where the expression of both Hoxd12 and Hoxd11 dropped down to 60 (Fig. 2L) and 40% percent of their initial expression levels, respectively.

Shorter Limbs in Deletions Removing Hoxd12 to Hoxd9

In order to evaluate a potential impact of this ectopic expression of posterior Hoxd genes upon the morphogenesis of the limbs during early development, we prepared and compared neonate and adult limb skeletons of either wild-type (Fig. 3A, C, E, G), or mutant (Fig. 3B, D, F, H–L) animals. Upon visual inspection, the mutant humerus, radius, and ulna looked slightly shorter, and hence we measured their precise lengths, both at birth and in adults, using fixed morphological landmarks (see Experimental Procedures section). We found the humeri of del(12-9), del(11-9) and del(11-10) mutant specimens significantly shorter than their wild-type counterparts, whereas this effect was observed neither in del(12-10) nor in del(12-11) animals (Table 1), i.e., in those mutant stocks where Hoxd13 was either poorly gained or not gained at all in the early bud. In 8-week-old mice, the reduction in size of the humerus was even greater, reaching only 85% of wild-type length in the del(12-9) stock (Table 1). Again, this reduction remained undetectable either in del(12-10) or in del(12-11) mice and thus only concerned mutant lines displaying a strong up-regulation of Hoxd13 in the early limb bud.

Figure 3.

Reduction in bone length in deletion mutants. A–D: Skeletal preparations of wild-type and mutant limbs at birth. The length of the humerus, radius, and ulna was measured in newborns (A,B), as well as that of the metacarpal and first phalanx of digit II (C,D, see also Table 1). A reduction in the length of the humerus, radius, and ulna is scored in those mutant lines where an early gain of function was observed, whereas the metacarpal (M) and first phalanx (P) of digit II are reduced in length in all mutant lines, regardless whether or not a gain of function had occurred. The autopods of both del(12-9) and del(12-10) were particularly affected, with a delay in the ossification of metacarpals and phalanxes and a soft tissue fusion between digits II and III (D). E–L: Limb skeletons of wild-type and mutant stocks at 8 weeks of age (see Table 1). The phenotype observed in neonates persisted in adults, with a re-enforcement in the decrease of bone lengths. As for newborns, the humerus, radius, and ulna were shorter in those mutant lines specifically displaying an early gain of function, whereas the metacarpal and first phalanx of digit II were reduced in length in all mutant lines (G–L). Again, del(12-9) and del(12-10) mutant animals were the most severely affected, with an additional bony element between the first and second phalanx of digit II and a soft tissue fusion between digits II and III (J, L).

Table 1. Newborn and Adult Long Bones Are Shorter in Mutant Limbs Carrying Hoxd Deletionsa
 HumerusP valueRadius/ulnaP valueMeta IIP valueP1 IIP value
  • a

    The humerus, radius and ulna, metacarpal and first phalanx of digit II were measured both in mutant neonates and in 8-week-old adult skeletons, using precise morphological landmarks (see Experimental Procedures section) and compared to their wild-type counterparts. In both newborns and adults, the length of the humerus, radius, and ulna was significantly reduced in those mutant lines showing an early gain of function of posterior Hoxd genes, whereas the length of metacarpal and first phalanx of digit II was reduced in all mutant lines, whether or not they displayed this early gain of function. Values significantly different from wild-type controls are typed with bold characters.

Newborn forelimb        
 wt100 ± 2.8 100 ± 3.7 100 ± 4.4 100 ± 4.9 
 del(12-11)97.7 ± 5.20.493.2 ± 6.50.0387.8 ± 7.30.0290.4 ± 4.70.02
 del(11-10)94.0 ± 3.00.00192.5 ± 2.90.00691.6 ± 3.50.0589.8 ± 5.20.01
 del(12-10)97.9 ± ± ± 3.6<0.00187.8 ± 4.2<0.001
 del(11-9)91.7 ± 4.00.0595.9 ± ± 4.80.0593.9 ± 3.40.05
 del(12-9)93.8 ± 3.60.00197.8 ± ± 4.40.00392.7 ± 4.1<0.001
Adult forelimb        
 wt100 ± 2.5 100 ± 3.0 100 ± 4.2 100 ± 4.9 
 del(12-11)98.7 ± ± 1.9<0.00174.2 ± 6.0<0.00194.3 ± 1.7<0.001
 del(11-10)96.4 ± 2.00.00289.8 ± 1.9<0.00170.0 ± 4.5<0.00189.1 ± 3.4<0.001
 del(12-10)98.6 ± ± 3.00.0160.2 ± 3.2<0.00184.0 ± 4.0<0.001
 del(11-9)86.8 ± 4.1<0.00190.2 ± 3.2<0.00170.3 ± 4.3<0.00186.0 ± 1.8<0.001
 del(12-9)85.0 ± 3.2<0.00187.1 ± 3.7<0.00158.9 ± 3.0<0.00182.8 ± 2.0<0.001

Likewise, both the radius and the ulna were affected. At birth, the reduction in bone length was significant only in the two del(11-10) and del(12-11) lines, whereas a slight effect was detected in adult mice of all mutant lines (Table 1). In this case, bones were shorter by 14 and 4% in the most and the least affected mutant line, respectively (Table 1). This abnormal phenotype thus involved those deletions where a weak Hoxd13 up-regulation was scored in the early limb bud, but also the del(12-10) animals where a normal expression of Hoxd gene was scored early on.

In many cases, the length of digits was also affected, both at birth and in adulthood, an effect particularly prominent for digit II, where the metacarpal bone (M) and the first phalanx (P1) were markedly reduced in neonate mutants from all deletion stocks (Fig. 3 and Table 1), as previously reported for the single mutations in either Hoxd11, Hoxd12, or Hoxd13 (Dolle et al., 1993; Davis and Capecchi, 1996). Some additional and weak phenotypic traits were also scored in both del(12-9) and del(12-10) animals, such as a delay in the ossification of metacarpals and phalanxes and a fusion between digits II and III at the level of soft tissues (Fig. 3C,D). These alterations in digits became altogether stronger in adult specimen (Fig. 3G–L), with a reduction in length reaching up to 41% for the metacarpal bone II (M) and 17% for its first phalanx (P1) in del(12-9) (Fig. 3L), or 39 and 16% in del(12-10), respectively (Fig. 3J; see also Table 1). In these latter two most affected mutant lines, a soft tissue fusion was also observed between digit II and III with a high penetrance. Finally, the first and second phalanxes were fused together.

In all other mutant lines (i.e., del(12-11), del(11-10), and del(11-9), a significant reduction in the length of digit II was also observed (Fig. 3H, I, K). In these cases, however, the soft tissue fusion was not observed and a correct number of distinct phalanxes was scored. Digit V was slightly affected as well in many configurations, in particular the second phalanx (P2), as previously described for the single inactivation of either Hoxd11, Hoxd12, or Hoxd13 (Dolle et al., 1993; Davis and Capecchi, 1996; Kondo et al., 1996). While it was difficult to precisely associate one particular loss of function to this phenotype, all configurations lead to the reduction in the length of this phalanx, between 20 to 30% reduction when compared to wild-type.

Alterations in the Carpus

We examined the carpus of mutant and control animals in details, in both newborns (Fig. 4A,B) and adults (Fig. 4D–K), and scored a number of phenotypic alterations (Fig. 4; Table 2). In some mutant lines, supernumerary bones were observed. For example, a small bone located ventrally, immediately proximal to the navicular and lunate bones, was detected in 14 out of 16 del(11-10) mutant carpus (Fig. 4D, arrowhead; Table 2). This additional bone was sometimes observed in del(11-9) carpus as well, although smaller in size. Another case was a small bone sometimes apparent at the basis of metacarpals in del(11-9) specimen (Fig. 4E, arrowhead). Otherwise, such supernumerary bones were rarely observed in either the del(12-9), the del(12-11), or the del(12-10) carpus.

Figure 4.

Abnormal mesopod in deletion mutants. Carpal elements are organized in two rows. The triquetrum-pisiform (TP), the lunate (L), and the navicular (N) form the proximal row, whereas the hamate (H), capitate (C), trapezium (Tzi), and trapezoid (Tzo) form the distal row. A–C: Skeletal preparation of carpus from either wild-type or del(12-11) mutant specimen at birth. The number and shape of carpal bones were analyzed in newborn mutant specimen and compared to wild-type (A). For example, in del(12-11) (B), the central bones of the distal row are fused, as well as the navicular and lunate (C, in black). D–K: Skeletal preparations of wild-type and mutant carpus at 8 weeks of age. Some supernumerary bony elements are observed in both del(11-10) and del(11-9) proximal to the navicular and lunate (D, E; arrowheads) (D), but also at the basis of metacarpals in del(11-9) (E; arrowhead). Both the proximal and distal rows of bones display abnormal fusions. In all mutant lines, the navicular and lunate are fused, to some extent (H–K). In the distal row, the central bones are fused in all mutant lines (F–K; see also Table 2).

Table 2. Fusions of Carpal Bones in Mutant Linesa
Adult carpusAnimal numberDistal bone fusionProximal bone fusionProximal supernumerary bone
  • a

    Supernumerary bones are observed in adult mutant carpus. For example, in the del(11-10) condition (and sometimes in del(11-9) carpus), a small bone was observed proximally to the navicular and lunate in 14 out of 16 cases. Supernumerary bones were rarely observed in the del(12-9), del(12-11), and del(12-10) mutant limbs. In the proximal row, all mutant lines showed a fusion of the navicular and lunate. Yet this phenotype was 100% penetrant only in the del(11-9) stock. In the distal row, mutant animals showed a fusion between the central bone and the bone located directly proximal, a phenotype which was almost fully penetrant in all mutant lines.

wtn = 200/400/400/40
del(12-11)n = 510/108/103/10
del(11-10)n = 815/1613/1614/16
del(12-10)n = 59/107/100/10
del(11-9)n = 1326/2626/2613/26
del(12-9)n = 612/126/120/12

Fusions between carpal bones, on the other hand, were frequently observed in all deletion mutant stocks, such as the fusion between the navicular and lunate, in the proximal row of bones. While this phenotypic trait was fully penetrant in del(11-9) animals (Table 2), other mutant lines displayed more variations. For instance, only half of del(12-9) animals exhibited this proximal bone fusion (Table 2) and the fusion was of different aspects, from the mere ossification of the junction between the navicular and the lunate (e.g., in del(12-9) or del(12-10); Fig. 4I,K, arrowheads), up to a complete fusion of the two bones into a single proximal bone, as observed in del(11-9) and del(11-10) animals (Fig. 4H,J, arrowheads). A similar fusion was scored in animals trans-heterozygous for Hoxd11 and Hoxd12 mutations (Davis and Capecchi, 1996). In the distal row, mutant animals showed a fusion between the central bone and the bone located directly proximal, a phenotype that was close to 100% penetrant for all mutant lines (Fig. 4F–K, arrowheads; Table 2). These phenotypes were already visible at birth as fusions of cartilaginous condensations, as illustrated in the del(12-11) stock (Fig. 4A,B).

Hindlimb Versus Forelimb

Upon superficial examination, the phenotypic alterations appeared less evident in hindlimbs than in forelimbs. The precise lengths either of the mutant femur, tibia, and fibula (Fig. 5A, B), or of the metatarsal bones and the first phalanx of digit II (Fig. 5C–H) confirmed this general assessment (Table 3). While a significant reduction in the length of the femur was scored in both del(12-9) and del(11-9) mutant stock, yet not in del(11-10) as observed in the forelimbs, the extent of this reduction was not as important as in forelimbs. In addition, the tibia, fibula, and metatarsal of digit II were also reduced in length in all mutant lines and the first phalanx in digit II was truncated specifically in the del(12-9) and del(11-9) stocks (Fig. 5). We examined the tarsus and found no bone fusion, except in some of the del(12-10) mutants, where two cuneiforms appeared to be fused together with a low frequency (Fig. 6).

Figure 5.

Reduction in hindlimb bone length in mutant lines. Hindlimb skeletons of 8-week-old wild-type and mutant specimen. The length of the femur, tibia, and fibula (A, B), as well as the metatarsal and first phalanx of digit II (C–H), were measured and compared to wild-type (A–C, see Table 3). A reduction in the length of the femur was scored in both del(12-9) and del(11-9) animals (A, B), but not in del(11-10), unlike in forelimbs. In addition, the tibia, fibula, and metatarsal of digit II were also significantly reduced in length, in all mutant lines (B–H). Finally, the first phalanx in digit II was slightly shorter in del(12-9) and del(11-9), yet not in other mutant animals.

Figure 6.

Fusion of tarsal bones in mutant lines. A: In the tarsus, the talus (T) and calcaneus (Ca) are found in the proximal row, whereas the navicular (N), cuboid (Cu), and three cuneiforms (C1–3) are in the distal row (C). B,C: Fusion of bones was not observed in the tarsus, except in del(12-10) mutants, where two cuneiforms fused together in rare occasions.

Table 3. Reduction in the Length of Hindlimb Long Bones in Mutant Linesa
Adult hindlimbFemurP valueTibia/fibulaP valueMeta IIP valueP1 IIP value
  • a

    The femur, tibia, fibula, metatarsal, and first phalanx of digit II were measured in mutant adult skeletons and compared to wild-type. In del(12-9) and del(11-9) mutants, the femur was reduced in length whereas it was not significantly reduced in del(11-10), unlike in forelimbs. On the other hand, the tibia, fibula, and metatarsal of digit II were significantly shorter in all mutant lines. Finally, the first phalanx in digit II was shorter in del(12-9) and del(11-9), yet not in other mutant animals. Values significantly different from wild-type controls are shown in bold.

wt100 ± 3.1 100 ± 2.1 100 ± 3.5 100 ± 5.7 
del(12-11)97.5 ± 1.00.0496.0 ± 2.40.00486.7 ± 1.0<0.00198.4 ± 3.90.5
del(11-10)97.9 ± ± 1.6<0.00186.0 ± 1.9<0.00195.4 ± 2.80.06
del(12-10)101.6 ± ± 3.00.0288.5 ± 4.1<0.00197.6 ± 1.80.3
del(11-9)93.6 ± 4.70.00191.8 ± 3.0<0.00185.8 ± 2.5<0.00191.8 ± 5.40.006
del(12-9)92.2 ± 3.6<0.00188.6 ± 2.3<0.00180.8 ± 3.3<0.00190.3 ± 6.0<0.001


Of the four Hoxd genes expressed in a coordinated manner during the development of the autopod (Hoxd10 to Hoxd13), only Hoxd13 was shown to exert an essential function there, since its inactivation via insertional strategies generated strong phenotypic alterations in digits (Dolle et al., 1993; Davis and Capecchi, 1996). Loss of function mutations of either Hoxd12 or Hoxd11 only moderately affected the limb extremities (Davis and Capecchi, 1996; Kondo et al., 1996). In addition, the presence in all these alleles of the selection cassette along with its promoter made a definitive functional conclusion difficult to arrive at, due to potential interferences, e.g., with Hoxd13. However, the combined inactivation in cis of Hoxd13, Hoxd12, and Hoxd11 generated a distal phenotype stronger than the mere inactivation of Hoxd13 (Zakany and Duboule, 1996), suggesting a substantial function for either Hoxd12, Hoxd11, or both, in the making of digits. This conclusion was re-enforced by the phenotype of mice lacking the entire HoxD gene cluster (Spitz et al., 2001), which was expectedly slightly more severe than that of mice lacking from Hoxd13 to Hoxd11.

However, classical loss of function mutations within these “posterior” Hoxd genes (insertional mutants, for example) generated phenotypic effects distinct from those obtained after the full deletion of the same locus (Kmita et al., 2002a). This apparent paradox was explained by the fact that the deletion of genes from the HoxD cluster sometimes leads to the re-allocation of regulation and hence to the gain of function of neighboring genes. For example, the deletion of the Hoxd13 locus induced a strong over-activation of Hoxd12 expression in developing digits, which was able to almost fully compensate for the loss of function of Hoxd13 (Kmita et al., 2002a). In contrast, a similar gain of function of Hoxd11 had deleterious effect upon digit morphology as it triggered supernumerary chondrogenic condensations. In addition, the deletion of the entire HoxD cluster also included those genes involved in shaping more proximal parts of the limb, and hence it may have interfered with an early phase of Hoxd gene expression (Tarchini et al., 2006), with a potential subsequent impact upon digit morphology without any relationship with the specific functions of the Hoxd13 to Hoxd10 genes.

We looked for those deletions of posterior Hoxd genes, which would not be accompanied by a gain of function of any neighboring gene, in particular of Hoxd13 since this latter has been shown in many instances to exert a dominant negative effect upon other Hoxd gene functions (“posterior prevalence”; Duboule and Morata, 1994). Accordingly, cells expressing strongly Hoxd13 were shown previously to lose the function of, e.g., group 11 Hox genes (van der Hoeven et al., 1996), even in the presence of normal amounts of Hox group 11 transcripts.

Amongst the various deletion stocks we assessed in this work, several led to an early gain of function of Hoxd13. The mechanism leading to these gains of function, however, hardly follows any clear-cut logic; for example, the del(9-12) induces a strong gain of function of Hoxd13, whereas the del(10-12) has no effect. One may thus conclude that something associated with the breakpoint near Hoxd10 prevents Hoxd13 to be activated prematurely (see Fig. 1). However, Hoxd12 is gained in both the del(9-11) and the del(10-11), suggesting in this case that the Hoxd10breakpoint does not affect the gain of function in any way. Also, all breakpoints located immediately downstream Hoxd12 appeared to negatively affect the transcription of this latter gene, suggesting that sequences involved in this process must be located further telomeric in the gene cluster.

Yet regardless how and when the precocious activation of Hoxd13 was produced, it was not maintained subsequently. Consistent with this observation and with the concept of posterior prevalence, these transient gains of function were translated into a slight shortening of long bones, preferentially of the humerus. A large majority of those cells present in early limb buds from E9.5-day-old embryos are indeed considered as fated to become part of the stylopod, i.e., the proximal part of the limb from where the humerus will condense and develop (see references in Tabin and Wolpert, 2007). An early pulse of Hoxd13 may thus preferentially affect such cells and generate an abnormal phenotype in the length of the humerus. Accordingly, at birth, the radius and ulna were less affected than the humerus, with only marginal, but significant, reductions in lengths. However, adult limbs showed a similar reduction of the zeugopod, suggesting that the growth and development of limb bones may not follow the same dynamics in the various parts of the limb. It also indicated that bone phenotypes at birth may not fully reflect the adult conditions. While it is not yet fully clear as to how posterior HOXD proteins can influence bone length, their involvement in shaping long bones via cortical ossification has been previously documented, as well as their expression in growth plates (Zakany and Duboule, 1996; Villavicencio-Lorini et al., 2010).

Interestingly, the phenotypes were observed at birth, i.e., well after the abnormal transcription of either Hoxd13 or Hoxd12 was terminated, indicating that the negative effect of these proteins could not be compensated for, or corrected, during the rest of the pregnancy. Also, if anything, the phenotype was accentuated in adults, when compared to newborns, rather than being rescued. Therefore, a short transcription burst of Hoxd13, occurring as early as day 9.5 of development, will affect limb bud cells in a way that will be maintained throughout development, until adulthood. One possibility amongst others is that the size of the pool of cells available to develop the mesenchymal condensation precursor of the zeugopod was reduced by the dominant-negative effect of this protein over other HOX proteins necessary to entertain the growth and renewal of this pool. In any case, the fact that a short burst of ectopic Hoxd13 expression can lead to such phenotypic effects in the adults illustrates the absolute necessity to prevent its expression in the proximal part of the developing limb buds, at least in those tetrapods where a long zeugopod has an adaptive value. On the other hand, a slight mis-regulation of this gene may rapidly lead to a reduced zeugopod, a morphology that is found in many tetrapods, for example, in seals.

Out of all the deletions used in this work, only the del(10-12) did not show any trace of premature activation of Hoxd13 in the early limb bud. Accordingly, the adult humerus was normal in size. In contrast, a slight shortening was observed for the zeugopod bones (radius and ulna), which were close to wild-type at birth (98% of normal length) yet significantly shorter in adults (96% of wildtype). This is most likely due to the loss of function of both Hoxd10 and Hoxd11 in the growing limbs. Functional inactivation of either gene separately did not induce any strong reduction in the sizes of the radius and ulna (Davis and Capecchi, 1994; Carpenter et al., 1997). However, the combined inactivation of both Hoxd11 and Hoxa11 led to an almost complete abrogation of these two bones, thus indicating the key functional contribution of group 11 Hox genes to the building of the zeugopod (Davis et al., 1995). As in most other cases analyzed in these studies, the effects observed on the bones of the hindlimbs were comparable, though generally less pronounced.

The various effects of these deletions upon digital morphologies was exemplified by the size reductions of both the metacarpal bone of digit II and the first phalange (P1) of digit II (Table 1), which were affected in all three Hoxd11, Hoxd12, and Hoxd13 “conventional” mutant mice (Dolle et al., 1993; Davis and Capecchi, 1996). In this case, a potential gain of function of Hoxd13 should not interfere, as the autopod is the place where this gene is normally expressed. In addition, the variation of Hoxd13 transcripts in some of these deletions was marginal, if any (Montavon et al., 2008). The comparison between the del(10-12) and del(11-12) revealed that Hoxd10 likely plays a role in the presence of Hoxd13 since the effect on bone length was significantly aggravated. Likewise, the comparison between del(9-11) and del(9-12) confirmed that Hoxd12 does indeed participate in the growth of metacarpals (10% shorter when Hoxd12 is absent). Since Hoxd11 was shown to also affect digit development (Davis and Capecchi, 1994), it is tempting to conclude that all four Hoxd genes co-expressed during autopodial development participate in the final pattern.

However, as mentioned above, Hoxd10 does not elicit a comparable phenotype, when all other genes remain functional (Carpenter et al., 1997), and the loss of function of Hoxd12 induced a minor defect in digits II and V (Davis and Capecchi, 1996; Kondo et al., 1996). This could be due to a compensatory effect of the neighboring genes, due to the regulatory re-allocations (Kmita et al., 2002a; Montavon et al., 2008). In this case, Hoxd10 and Hoxd12, in addition to Hoxd11, would indeed have an essential function during digit development. Alternatively, these functional requirements may be revealed only when an artificial, though physiologically relevant, context is used. In other words, the fact that, in our mutant conditions, the deletions of both Hoxd11 and Hoxd12 in cis generated an abnormal digit phenotype, even in the presence of Hoxd13, does not necessarily mean that the inactivation of both genes in cis (for example, by introducing stop codons) would give the same phenotype. Yet these results, added to those obtained when Hoxd13 was deleted along with the neighboring genes (e.g., Zakany and Duboule, 1996) make it likely that all Hoxd12, Hoxd11, and Hoxd10 genes, co-expressed in distal limb buds during digit development, participate in the final morphology to complement the prominent function of Hoxd13.

Hoxd9, on the other hand, which is the last gene of the series to be expressed during digit development, does not seem to play a role there. Rather, like other group 9 Hox genes, it organizes more proximal structures, including part of the limb field (Xu and Wellik, 2011). The comparison of pairs of deletions with or without Hoxd9 present, such as del(9-12) with del(10-12), or del(9-11) with del(10-11), did not reveal any significant modification in digital morphologies, in agreement with the absence of a related phenotype in mice mutant for this gene, even when combined with an Hoxa9 allele (Fromental-Ramain et al., 1996a). This is not surprising, considering the very low amount of Hoxd9 mRNAs detected during digit development, when compared to those of either Hoxd13, Hoxd12, or Hoxd11 (Montavon et al., 2008).


Stocks of Mice

All mutant alleles were produced in vivo by targeted meiotic recombination (TAMERE) (Herault et al., 1998). The del(11-9) mutant line was obtained using targeted recombination between a first loxP site located between Hoxd9 and Hoxd8 and a second one between Hoxd12 and Hoxd11. All other mutant lines used in this work were described previously (Kmita et al., 2002b; Tarchini and Duboule, 2006; Di-Poi et al., 2007). Genotyping of mice and embryos was performed by PCR analysis. Mice were handled following the guidelines of the Swiss law on animal protection, with the requested authorization (to D.D.).

In Situ Hybridization

In situ hybridization was performed according to standard protocols. The Hoxd12 and Hoxd13 probes were described previously (Dolle et al., 1991; Izpisua-Belmonte et al., 1991).

Skeletal Preparation and Analysis

Newborn skeletons were stained with standard Alcian blue/Alizarin red protocols, adult skeletons with Alizarin red only. Skeletal elements were dissected, fixed flat onto a glass support, and captured with an Olympus Cell D imaging system. Images were analyzed with Adobe Photoshop CS5. Measurements were taken for both left and right limbs along a line, parallel to the main length of the skeletal element, from and to the perpendicular line passing at the following morphological marks: for the forelimbs, the head and lateral epicondyle of the humerus, the oleocranon process of the cubitus and styloid process of the ulna, the bases and condyles of metacarpals and phalanxes; for the hindlimbs, measurements taken were the head and lateral epicondyle of the femur, the intercondylar eminence of the tibia and lateral malleolus of the fibula, the bases and condyles of metatarsals and phalanxes. Measurements in newborns were normalized to the size of the scapula. Measurements of the scapula were taken parallel to the spine of the scapula, from the base of the spine to the acromion. Mean values were calculated. The Student's t-test using two-tailed distribution with two-sample unequal variance was used to calculate significance of distribution. For adults, the number n of specimen analyzed was the following: wild-type n=20; del(12-11) n=5; del(11-10) n=8, del(12-10) n=5; del(11-9) n=13; del(12-9) n=6. For newborns: wild-type n=5; del(12-11) n=5; del(11-10) n=6, del(12-10) n=5; del(11-9) n=5, del(12-9) n=8.


We thank M. Vieux-Rochas, P. Schorderet, and T. Montavon for critical comments and suggestions. We also thank B. Mascrez and E. Joye for help with mice and experiments, respectively, as well as members of the Duboule laboratories for discussions and reagents. S.D. and D.D. designed the experiments. J.Z. produced mouse strains. S.D. carried out all experiments. S.D. and D.D. wrote the paper.