The 5′ members of the HoxD gene cluster (paralogous groups 9–13) are crucial for correct vertebrate limb patterning. Mutations in the HOXD13 gene have been found to cause synpolydactyly (SPD) and other limb malformations in human. We report the identification in a Greek family of a variant form of SPD caused by a novel missense mutation that substitutes glycine for valine in position 220 (G220V) of HOXD13. This mutation represents the first substitution of an amino acid located outside of the HOXD13 homeodomain that causes autopodal limb malformations. We have characterized this mutation at the molecular level and found that the G220V substitution causes a significant impairment of the capacity of HOXD13 to bind DNA and regulate transcription. HOXD13(G220V) was found to be deficient in both activating and repressing transcription through HOXD13-responsive regulatory elements. In accordance with these results, a comparison of the activities of HOXD13 and HOXD13(G220V) in vivo, using retrovirus-mediated misexpression in developing chick limbs, showed that the G220V mutation impairs the capacity of HOXD13 to perturb the development of proximal limb skeletal elements and to ectopically activate the transcription of the Hand2 target gene. We moreover show that the G220V mutation compromises the stability of the HOXD13 protein within cells and causes its partial accumulation in the cytosol in the form of subtle aggregates. Taken together, our results establish that the G220V substitution does not produce a dominant-negative effect or a gain-of-function, but represents a dominant loss-of-function mutation revealing haploinsufficiency of HOXD13 in human.
Hox genes encode highly conserved transcription factors controlling cell fates and regional identities along the primary body and limb axes in metazoans (1). Mammalian genomes contain 39 Hox genes arranged in four clusters (HoxA–D) (1). Hox proteins bind specific DNA sequences via the homeodomain, and are thought to regulate overlapping sets of target genes, although the molecular pathways controlled by Hox proteins are only starting to be unveiled (2–9). In developing vertebrate limbs, 5′ HoxA and HoxD genes (paralogous groups 9–13), which are related to the Drosophila Abd-B gene, are expressed in dynamic regionally restricted patterns, and are crucial for correct limb patterning (10–12). Targeted mutagenesis and misexpression of individual and multiple 5′ Hoxa and/or Hoxd genes in the mouse (reviewed in 13–15) and chick (16–19) showed that the perturbation of Hox gene expression does not produce homeotic transformations in the limbs, but alters instead the size, shape and number of specific skeletal elements. Hox genes are thus held to determine the formation and region-specific growth of limb skeletal elements, acting combinatorially at both early and late stages of limb development, to direct processes such as mesenchymal cell proliferation, aggregation, chondrification and ossification (reviewed in 13–15).
Two HOX genes have been found to be mutated in human malformation syndromes, HOXD13 in synpolydactyly (SPD) and HOXA13 in hand–foot–genital syndrome (reviewed in 20). SPD is a dominantly inherited limb malformation, characterized by syndactyly between the third and fourth fingers and between the fourth and fifth toes, with a partial or complete extra digit in the syndactylous web. Typical SPD is caused by expansions of a 15-residue polyalanine tract in the N-terminal region of HOXD13 (21–23). The mutant protein is held to act as a dominant negative, interfering with the function of wild-type HOXD13 and possibly with other 5′ HOXD proteins expressed in the autopod (24–26). Atypical forms of SPD, characterized by a distinctive foot phenotype, have been found to be associated with deletions and missense mutations within HOXD13 (27–30). Missense mutations within the HOXD13 homeodomain have been also found to cause brachydactyly–polydactyly or brachydactyly types D and E (16,31). The deletions in HOXD13 are predicted to produce truncated proteins unable to bind DNA, as they remove the homeodomain, and are thus anticipated to behave as dominant loss-of-function mutations revealing functional haploinsufficiency of HOXD13. The characterized homeodomain missense mutations have been shown to cause either complete (32) or partial (16,31) loss-of-function of HOXD13. Missense mutations in regions other than the homeodomain causing human limb malformations have not been molecularly characterized to date, and are in principle more difficult to interpret, due to the lack of information on the functional domains of the HOXD13 protein. The characterization of this class of HOXD13 mutations thus provides a unique opportunity to gain insight into the functional domains of the HOXD13 protein and therefore to provide a framework to interpret the effect of new mutations within the HOXD13 protein that cause limb malformations.
In this work, we report the functional characterization of a novel missense mutation within HOXD13 that substitutes a glycine in position 220 for a valine (G220V). This mutation, identified in a Greek family, causes a variant form of SPD with features similar to ‘classical’ SPD caused by moderate alanine tract expansions. We have characterized the effects of this mutation at the molecular level and found that the G220V substitution impairs the capacity of HOXD13 to regulate transcription. HOXD13(G220V) was found to bind DNA less efficiently and to be deficient in both activating and repressing transcription of target gene promoters. We moreover show that the G220V mutation compromises the stability of the HOXD13 protein within cells and causes its partial accumulation in the cytosol in the form of subtle aggregates. In accordance with the biochemical properties of the HOXD13(G220V) protein, a comparison of the activities of HOXD13 and HOXD13(G220V) via retrovirus-mediated misexpression in developing chick limbs showed that the G220V mutation impairs the capacity of HOXD13 to perturb the development of proximal limb skeletal elements and to ectopically activate the transcription of the Hand2 target gene. Taken together, our results establish that the G220V substitution does not produce a dominant-negative effect or a gain-of-function but represents a dominant loss-of-function mutation confirming haploinsufficiency of HOXD13 in human.
A new missense mutation (G220V) N-terminal to the HOXD13 homeodomain causes a variant form of SPD
We identified a new dominantly inherited limb autopodal malformation in a family of Greek origin (Fig. 1). The proband (III.15) belongs to a three-generation family and is the second son of two non-consanguineous parents (Fig. 1). At birth SPD and clinodactyly of the fifth finger was noted at the right hand. Camptodactyly was observed at the fifth left finger. His feet were normal and no other abnormalities were noted. Radiographs of the hands (Fig. 1B) demonstrate a type II SPD with complete bony fusions of the three phalanges of the fourth finger with the phalanges of the extra digit of the right hand. There is clinodactyly of the right fifth finger and camptodactyly of the left fifth finger. Further investigation of this family revealed ten other family members with hand- and/or foot abnormalities (Fig. 1): unilateral camptodactyly (fixed flexion of the PIP joint) of the fifth finger was noted in individual I.1.; bilateral camptodactyly of the fifth fingers in individuals II.1, II.3, II.5, II.13; a unilateral fixed flexion of the DIP joint of the fifth finger in individual II.12 and a bilateral fixed flexion of this joint in individuals II.10, III.5, III.7 and III.12. In the feet, the most obvious malformation was variable degrees of broadening and external rotation of the fifth toe in individuals I.1, II.1, II.3, II.13, III.5. Partial cutaneous syndactyly of toes 2–3 and toes 3–4 was seen in individual II.13. We could only obtain radiographs of the hands and feet of the proband's mother (individual II.12). Apart from the unilateral flexion deformity of the DIP joint at the right hand no other bony abnormalities were noted (data not shown).
Direct sequencing of HOXD13 revealed a heterozygous G-to-T transition in exon 1 at position 659 of the coding sequence in the proband (III.15) and his mother (II.12), but not in his father (II.11). This base change, which converts amino acid 220 from glycine to valine (G220V) destroys a KpnI restriction site (Fig. 1C). The same base change was identified by direct sequencing in eight other affected family members (I.1, II.1, II.3, II.5, II.13, III.5, III.7, III.12), but not in six unaffected family member (II.8, III.3, III.4, III.6, III.9, III.10) or in 50 unrelated unaffected controls. No DNA sample was obtained from individual II.10 who is clinically affected.
The G220V mutation is located 48 amino acids N-terminal to the homeodomain within a region of the protein that has never been previously functionally characterized. An alignment of all available Hoxd13 protein sequences, however, showed that the glycine in position 220 is evolutionarily highly conserved, as it is present in all analysed sequences, belonging to 16 different species (Fig. 1D). This further suggested the possible relevance of this amino acid in the structure and/or function of the HOXD13 protein.
The G220V mutation impairs both the transcriptional activation and repression functions of HOXD13
To test whether the G220V missense mutation affects the capability of the HOXD13 protein to regulate transcription, the activities of HOXD13 wild-type and of HOXD13(G220V) were compared in transient transfection assays of P19 embryonal carcinoma cells on a set of HOXD13 responsive elements. A luciferase reporter construct was generated, pT81(TTAT)6 containing one of the two optimal DNA-binding consensus sites of HOXD13 (16). Three additional luciferase reporter constructs were tested, one, pT81HCR, containing the HCR sequence, an evolutionarily conserved HOX binding region within the HOXD9 promoter (16,33), another, pT81Hand2, containing a regulatory region within the Hand2 locus bound in vivo by HOXD13 (8), and the third one, pT81Barx1, containing the HOXD13 responsive region within the Barx1 locus (8). HOXD13WT effectively enhanced the activities of the pT81(TTAT)6 (Fig. 2A), the pT81HCR (Fig. 2B) and the pT81Hand2 (Fig. 2C) reporters. HOXD13(G220V), on the contrary, displayed a significantly reduced capability to activate transcription from the pT81(TTAT)6, the pT81HCR and the pT81Hand2 reporters (Fig. 2A–C, respectively).
As described previously (8), the expression of HOXD13 wild-type caused repression of the pT81Barx1 reporter activity (Fig. 2D). Also in this case, HOXD13(G220V) showed a reduced functionality compared with HOXD13 wild-type, as it failed to repress the activity of the pT81Barx1 reporter (Fig. 2D).
Thus, our results indicate that the G220V mutation affects both the capacity of HOXD13 to activate, as well as to repress transcription.
The N-terminal portion of HOXD13 contains two separable transcriptional activation functions affected by the G220V mutation
Next we wanted to characterize the transcriptional activation domain of HOXD13, and to verify whether the G220V mutation would be located within it, thus impinging on its function. To this end, the N-terminal region of HOXD13 was divided in two portions (1–131 and 132–267), each of which was fused to the DNA-binding domain of the yeast GAL4 protein (GAL4DBD) to produce the GAL4DBD-D13(1–131) and the GAL4DBD-D13(132–267)WT or GAL4DBD-D13(132–267)G220V chimeric proteins (Fig. 3A). Both the GAL4DBD-D13(1–131) and the GAL4DBD-D13(132–267)WT chimeric proteins significantly increased the activity of the pT81UAS reporter (Fig. 3B), indicating that both the 1–131 and 132–267 regions of HOXD13 contain sequences sufficient to mediate transcriptional activation. The GAL4-D13(132–267) chimera bearing the G220V mutation (GAL4DBD-D13(132–267)G220V), conversely, showed only a modest increase in the activity of the pT81UAS reporter (Fig. 3B). To further confirm the existence of a transcriptional activation domain within the HOXD13 132–267 region and to verify its activity in the presence of the G220V mutation, we generated additional GAL4DBD fusions in which the GAL4DBD was located C-terminally to the HOXD13 132–267 portion (Fig. 3A). As in the case of the N-terminal fusions, the D13(132–267)WT-GAL4DBD fusion stimulated the pT81UAS reporter activity, while the D13(132–267)G220V-GAL4DBD fusion displayed only a modest transcriptional activation (Fig. 3B).
To analyse the function of the two N-terminal transcriptional activation domains in the context of the HOXD13 protein, we generated three deletion mutants, two lacking the first 131 amino acids of HOXD13 (HOXD13Δ(1–131)WT and HOXD13Δ(1–131)G220V), and one representing an internal deletion of amino acids 132–267 (HOXD13Δ(132–267)) (Fig. 4A). These were tested in transient transfection assays of P19 cells on the pT81HCR, pT81Hand2 and pT81Barx1 reporters. Both the HOXD13Δ(1–131)WT and the HOXD13Δ(132–267) mutant derivatives efficiently stimulated the activities of the pT81HCR (Fig. 4B) and of the pT81Hand2 (Fig. 4C) reporters, confirming the existence of two regions sufficient for transcriptional activation within HOXD13. The HOXD13Δ(1–131) deletion mutant bearing the G220V mutation, HOXD13Δ(1–131)G220V, conversely, showed a distinctly lower activity on both pT81HCR and pT81Hand2 (Fig. 4B and C, respectively). Neither HOXD13Δ(1–131)WT nor HOXD13Δ(132–267) could effectively repress the activity of the pT81Barx1 reporter. HOXD13Δ(1–131)WT only marginally reduced the pT81Barx1 activity, while HOXD13Δ(132–267) expression actually strongly stimulated it (Fig. 4D). Interestingly, the presence of the G220V mutation in the context of the HOXD13Δ(1–131) deleted protein caused a moderate but consistent increase in the capability of the HOXD13Δ(1–131)G220V protein to stimulate the reporter activity.
Thus, HOXD13 contains two separable activation functions, one located in the N-terminal portion of the protein (amino acids 1–131) and the other one adjacent to the homeodomain (amino acids 131–267). HOXD13 contains in addition a repression activity, which likely requires the integrity of the HOXD13 protein to function. All of these ‘effector’ domains are impaired by the G220V mutation.
The G220V mutation partially impairs the DNA-binding function of HOXD13
Despite the fact that the G220V mutation is located outside of the homeodomain, which is considered to be sufficient for effective DNA-binding (34,35), we tested whether HOXD13(G220V) would bind DNA as efficiently as wild-type HOXD13. Two different oligonucleotide probes were used in electrophoretic mobility shift assays (EMSA), corresponding to the two optimal DNA binding consensus sequences of HOXD13, TTTTATTGG and TTTTACGAG (16). The HOXD13(I47L) mutation, representing a missense mutation of amino acid 47 within the homeodomain (16), was used as a control (Fig. 5, lanes 9–11 and 20–22). The HOXD13, HOXD13(G220V) and HOXD13(I47L) proteins were produced by in vitro transcription/translation in rabbit reticulocite lysates and quantified by SDS–PAGE. Equivalent amounts of proteins were used in binding reactions. HOXD13(G220V) was found to bind less efficiently than HOXD13 both the TTTTATTGG and the TTTTACGAG sites (Fig. 5, lanes 6–8 and 17–19, respectively). Unlike HOXD13(I47L) (16), HOXD13(G220V) did not show any change in binding preference between the two consensus sites with respect to wild-type HOXD13 (Fig. 5, compare lines 9–11 and 20–22 with lanes 6–8 and 17–19). These results show that the G220V mutation, despite the fact that it is located N-terminally to the homeodomain region, causes an unexpected impairment of the DNA-binding ability of HOXD13.
The G220V mutation compromises the stability of the HOXD13 protein within cells
The data showing an impairment of HOXD13 DNA-binding by the G220V mutation led us to speculate that the mutation might alter indirectly the DNA-binding capability of HOXD13 by affecting the overall structure of the protein. As alterations in structure usually lead to an accelerated degradation of the protein (36), we verified whether the G220V mutation would affect the stability of HOXD13 within cells. To this end, COS cells, transiently transfected with the HOXD13 or the HOXD13(G220V) expression constructs, were treated with cycloheximide (CHX) to block protein synthesis, and the decays of the HOXD13 and HOXD13(G220V) proteins were compared by SDS–PAGE and immunoblot analysis. HOXD13 proved to be relatively stable within cells, as after 4 h protein levels were reduced only to 98 ± 7% of the initial amount (Fig. 6A). Conversely, the stability of the HOXD13(G220V) proteins was significantly reduced as only 66 ± 5% of the initial amount was present at 4 h after CHX treatment (Fig. 6A). We also compared the rates of decay of the HOXD13Δ(1–131)WT and HOXD13Δ(1–131)G220V deletion mutants. After 4 h of CHX treatment HOXD13Δ(1–131)WT levels were reduced to 83 ± 6% of the initial amount, while HOXD13Δ(1–131)G220V levels were down to 61 ± 4%. These results show that the G220V mutation significantly impairs the stability of HOXD13 within cells.
We next verified by immunofluorescence the expression of HOXD13(G220V) and its subcellular localization, and compared them with those of HOXD13. COS cells were transiently transfected with the HOXD13 or the HOXD13(G220V) expression constructs, and stained by immunofluorescence. HOXD13 displayed, as anticipated, a predominantly nuclear localization, whereas HOXD13(G220V) showed, besides the correct nuclear localization, a significant fraction of cytosolic localization in ∼30% of the expressing cells (Fig. 6B). Cytosolic HOXD13(G220V) formed a weft of fine aggregates, analogous, even if lighter, to those observed in the case of HOXD13 polyalanine expansion mutants (26) (Fig. 6B). To confirm the partial cytoplasmic localization of HOXD13(G220V), fractionated protein extracts were prepared from the nuclear and cytoplasmic cell compartments. Immunoblot analysis revealed that indeed an appreciable part of the total HOXD13(G220V) protein (∼8.5%) localizes to the cytoplasm, whereas HOXD13 wild-type is located exclusively in the cell nucleus (Fig. 6C and D). As a comparison, we analysed a HOXD13 mutant protein carrying a polyalanine expansion of 14 extra alanines, HOXD13(+14A), which was previously reported to form cytoplasmic aggregates (26). A significant fraction of HOXD13(+14A) (∼19% of the total) displayed cytoplasmic localization (Fig. 6C and D).
Taken together, these results show that the G220V mutation affects the stability and the subcellular localization of the HOXD13 protein indicating that a fraction of the total expressed HOXD13(G220V) protein is in fact inactive within cells.
The misexpression of HOXD13(G220V) in developing chick limbs produces a milder phenotype compared with the misexpression of wild-type hoxd13
We used retrovirus-mediated expression in developing chick limbs to analyse the effect of the G220V mutation in vivo. HOXD13 misexpression in developing chick limbs has been previously reported to cause specific perturbations of limb development that affect mainly the skeletal elements of the stylopod, zeugopod and proximal autopod (16,17). No abnormalities were reported involving the phalanges, the normal site of expression of cHoxd13, indicating that elevated levels of HOXD13 do not interfere with normal development of the distal autopod (16,17). Misexpression in developing chick limbs thus represents a profitable system to test whether mutations in HOXD13 act by a dominant negative or other gain-of-function mechanism, or whether they result in a loss-of-function.
We generated an RCAS retroviral vector (37) for the expression of HOXD13(G220V). Concentrated retroviral suspensions of comparable titers of RCAS-HOXD13 (16) or RCAS-HOXD13(G220V) were injected in ovo in the prospective right hindlimb field of stage 10 chick embryos. A retroviral construct expressing GFP was used to verify the infection at high frequency of the entire right leg bud (data not shown). Embryos were harvested at stage HH34 and stained with Alcian Blue and Alizarin Red to visualize cartilage and bone elements, respectively. The resulting phenotypes were classified into three categories, mild, moderate and severe, according to the severity of the observed developmental defects. The measurement of the tibia length was considered as the main objective parameter, being it one of the principal hallmarks of the HOXD13 misexpression phenotype (16,17). Thus, mild phenotypes included discrete shortening of cartilages of the stylopod, zeugopod and/or autopod, sometimes associated with a mild delay in ossification, with a reduction of the tibia lenght to 90–75% of that of the contralateral uninjected control limb (100%). Embryos displaying moderate phenotypes presented more pronounced shortening of cartilaginous elements, a delay in ossification, a premature articulation or fusion of the fibula to the fibulare, and a reduction of the tibia length to 74–40% of the contralateral control tibia length. Finally, severe phenotypes were characterized by a strong shortening of skeletal elements, a delay in ossification, a premature articulation or fusion of the fibula to the fibulare and a shortening of the tibia length to 39–20% of the length of the contralateral control tibia. Of the harvested chick embryos injected with RCAS-HOXD13(WT) (53 embryos in total) 75% showed a distinct phenotype, whereas only 46% percent of the embryos injected with RCAS-HOXD13(G220V) (54 embryos in total) displayed a detectable phenotype. The embryos misexpressing HOXD13(WT) (Fig. 7A and B) showed essentially the phenotypes described previously (16,17). Of the embryos displaying limb defects, 22% showed a severe, 56% a moderate and 22% a mild phenotype (Fig. 7F). Embryos misexpressing HOXD13(G220V), on the contrary, displayed consistently milder phenotypes (Fig. 7C and D). Only 12% of the embryos displaying limb defects showed a severe phenotype, while 50 and 38% showed a moderate and a mild phenotype, respectively.
Thus, the distribution of the phenotypic classes differs markedly between wild-type or HOXD13(G220V) misexpressing chick limbs. The misexpression of HOXD13(G220V) resulted in an overall lower frequency of moderate-to-severe phenotypes indicating that the G220V mutation causes a significant reduction of HOXD13 activity in vivo.
The HOXD13 downstream target gene Hand2 is only moderately up-regulated by HOXD13(G220V) misexpression
We recently reported that the Hand2 gene is a direct transcriptional target of HOXD13, and that its expression in developing chick limbs is dramatically perturbed by HOXD13 retrovirus-mediated misexpression (8). We thus analysed by whole-mount in situ hybridization the effect of HOXD13(G220V) misexpression on Hand2 expression in embryonic chick limb buds. Hand2 is expressed initially throughout the flank mesenchyme and subsequently becomes restricted to the posterior mesenchyme of the developing limb buds (Fig. 8 and ref. 38). As previously reported (8), HOXD13 misexpression led, at HH stages 20 and 23, to a marked up-regulation of Hand2 expression in the injected limb buds within its normal expression domain, and caused in addition the appearance of an ectopic distal anterior domain of expression (12/12, 100% of the analysed embryos) (Fig. 8). The misexpression of HOXD13(G220V), conversely, did not affect Hand2 expression in 33% (5/15) of the analysed embryos (Fig. 8, upper right panels). In 46% (7/15) of the embryos analysed, we observed only a moderate upregulation of Hand2 expression with a slight anterior shift of its distal expression domain (Fig. 8, central right panels). Finally, in only 20% (3/15) of the analysed embryos we could detect a more pronounced upregulation of Hand2 expression (Fig. 8, lower right panels). None of the RCAS-HOXD13(G220V) injected embryos showed a marked ectopic anterior Hand2 expression domain as observed in limb buds misexpressing HOXD13.
These results show that also in vivo during chick limb development, HOXD13(G220V) displays a reduced capability to activate transcription, as it fails to strongly upregulate the expression of the Hand2 downstream target gene.
A number of different mutations in the HOXD13 gene have been shown to cause SPD in human. These include various degrees of polyalanine expansions, which cause ‘classical’ SPD (21–23), and frameshifting deletions, which are predicted to result in non-functional truncated proteins, lacking the homeodomain, that cause atypical forms of SPD (27,28,30). In this work, we report the identification and analysis of a novel missense mutation involving amino acid 220 of HOXD13 that results in a variant form of SPD. This mutation represents the first substitution of an amino acid located outside of the HOXD13 homeodomain that causes malformations of the limb autopod.
A variant form of SPD is caused by a G220V missense mutation within HOXD13
The new dominantly inherited limb autopodal malformation reported in this work presents only some of the canonical features of SPD observed in patients carrying polyalanine tract expansions and frameshifting deletions in the HOXD13 protein (39). The proband showed webbing of the 3/4 fingers with a partial digit duplication and clinodactyly of the fifth finger in the right hand, but lacked the typical 4/5 toes webbing. Other family members showed mostly camptodactyly of the fifth finger and occasionally of the fifth toe, while only one individual showed partial cutaneous syndactyly of toes 2–3 and toes 3–4. Thus, in addition to hallmarks of ‘classical’ SPD, the phenotype displayed by individuals carrying the G220V mutation presents also additional features, such as fifth finger clinodactyly, that are not always associated with canonical SPD. Interestingly, some traits of the phenotype caused by the G220V mutation correlate with those characterizing the mild end of the phenotypic spectrum of classical SPD, and are thus comparable with the milder and less penetrant phenotypes of patients carrying short (7–9 extra alanines) polyalanine tract expansions (22). This observation indicates that the G220V mutation alters only partially HOXD13 function, and moreover prompts the intriguing notion that mutations as different as a single missense mutation and polyalanine tract expansions, which furthermore affect distinct regions of the HOXD13 protein, yet produce phenotypes showing several common traits.
HOXD13(G220V) shows an impaired capability to activate and repress transcription
As in the case of many HOX proteins, the regions other than the homeodomain are poorly characterized as to their function (40). These regions are often divergent among paralogous Hox proteins and are held to endow the various Hox proteins with unique functional properties in terms of transcriptional regulation. The availability of clinically manifested mutations affecting protein regions outside the homeodomain thus offers an ideal opportunity to gain insight into the possible different functional domains of the HOXD13 protein. The mutated glycine residue at position 220 of HOXD13 is evolutionarily highly conserved. It is present in 100% of the available Hoxd13 protein sequences, belonging to 16 different species, and is moreover conserved in the paralogous Hoxc13 proteins. The high evolutionary conservation of this glycine residue indicates that it may play a relevant structural role within a functional domain of the HOXD13 protein. Indeed, we found that the large region of the HOXD13 protein N-terminal to the homeodomain can be divided in two portions that retain transcriptional activation capability. We show that the function of one of these, which spans amino acids 131–267, and includes glycine residue 220, is affected by the G220V mutation. Our data moreover indicate that the G220V mutation indirectly affects also the function of the other transcriptional activation domain located at the N-terminus of HOXD13 (amino acid 1–131), as the full length HOXD13(G220V) protein displayed essentially the same reduction in activity as truncated HOXD13 lacking the N-terminal activation domain (HOXD13(Δ1–131)G220V).
We previously reported that HOXD13 can also act as a transcriptional repressor, as it down-regulates the expression of the direct target gene Barx1 via a regulatory element located 3′ to its coding region (8). We found that the G220V mutation also affects the repression function of HOXD13. Interestingly, neither of the two portions of the HOXD13 N-terminal region we analysed appeared to reproduce the repression function displayed by full-length HOXD13, suggesting that the repression activity may require the entire N-terminus, or alternatively that it may span the deletion boundary. In this context, it is noteworthy that both activation domains, especially the one located at the N-terminus (amino acids 1–131), in isolation, displayed a stronger transactivation capability as compared with the complete HOXD13 protein. This indicates that the balance between the activation and repression functions within the HOXD13 protein may rest on the overall structure of the region N-terminal to the homeodomain, and may be regulated via intra-molecular contacts. Further work will be required to fully characterize and delimit the HOXD13 domains mediating activation and repression.
We recently reported that HOXD13 directly controls the expression of a number of genes that play crucial roles in limb development (8). In particular, we found that the Hand2 gene, which is required for the correct distal limb patterning along the anterior–posterior axis, was markedly up-regulated in its normal expression domain, and in addition displayed an ectopic distal–anterior domain of expression (8). In accordance with the results showing that HOXD13(G220V) has a reduced capacity to activate transcription via several regulatory elements, including that derived from the Hand2 locus, we found that HOXD13(G220V) misexpressed in developing chick limbs causes only a weak up-regulation of endogenous Hand2 expression, demonstrating that also in vivo HOXD13(G220V) displays an impaired transcriptional activation function.
The HOXD13(G220V) mutated protein fails to produce marked proximal limb defects if misexpressed in developing chick limbs
To verify the functions of the HOXD13(G220V) protein in vivo, we misexpressed it in developing chick limbs. The misexpression of HOXD13 in developing chick limbs causes specific perturbations of limb development that affect mainly the skeletal elements of the stylopod, zeugopod and proximal autopod (16,17). Minor or no defects are observed in the digits, showing that overexpression of HOXD13 wild-type does not perturb the development of regions, such as the distal autopod, that coincide with the normal HOXD13 expression domain (16,17). The marked shortening of proximal limb skeletal elements caused by the ectopic expression of HOXD13 have been interpreted as to be caused by a functional interference with endogenous Hox proteins, in particular those belonging to paralogous group 11 (15). Indeed, shortening and bowing of the radius and ulna have been reported also in mice having a Hoxd11/lacZ transgene integrated upstream of Hoxd13 (15,41) and in mice carrying the Ulnaless mutation (42,43). Moreover, the zeugopod shortening observed in all these cases is reminiscent of that found in Hoxa11−/−;Hoxd11+/− and Hoxa11+/−;Hoxd11−/− mice (44). Thus, misexpression in developing chick limbs represents a useful system to test whether mutations in HOXD13 act by a dominant-negative or other gain-of-function mechanism, or whether they result in a loss-of-function. The expression of dominant negative or gain-of-function mutants is anticipated to produce defects in the distal autopod, while the expression of loss-of-function mutants is expected to result in milder phenotypes in proximal limb regions. The misexpression of HOXD13(G220V) resulted in an overall weaker skeletal phenotype when compared with the misexpression of HOXD13 wild-type. In fact, while the majority of the HOXD13 misexpressing embryos displayed severe and moderately severe phenotypes, ∼40% of the embryos misexpressing HOXD13(G220V) showed a mild phenotype. No significant perturbations of normal digit development were observed in HOXD13(G220V) misexpressing chick limbs. Thus, HOXD13(G220V) expression in the distal autopod does not interfere with the function of endogenous Hoxd13, showing that HOXD13(G220V) does not act via a dominant-negative or other gain-of-function mechanism. These results, moreover, are in accordance with our data showing an impaired capability of HOXD13(G220V) to activate and repress transcription, to bind DNA and to activate in vivo a direct downstream target of HOXD13. Thus, taken together our data indicate that the G220V mutation does not act via a dominant-negative mechanism but instead represents a dominant loss-of-function mutation confirming haploinsufficiency of HOXD13 in human.
The G220V mutation alters the stability of HOXD13 within cells and causes its partial cytosolic accumulation
The G220V mutation represents the substitution, at an evolutionarily highly conserved position, of a structurally versatile amino acid (glycine) with a hydrophobic aminoacid (valine). The introduction of a hydrophobic aminoacid in a protein is likely to produce structural alterations, leading to the exposure of regions that are buried in the native state, thus possibly causing aggregation and the subsequent degradation of the protein (45). Usually misfolded proteins are held to form aggregates and become insoluble before degradation, especially when their production exceeds the capacity of the cell to degrade them (36). We found indeed that the HOXD13(G220V) protein has a significantly lower stability within cells, when compared with wild-type HOXD13, and that a fraction of the expressed HOXD13(G220V) protein is located in the cytosol where it forms subtle aggregates. Interestingly, HOXD13 mutant proteins containing polyalanine tract expansions have been also shown to form cytosolic aggregates, whose presence depends on the number of extra alanines (26). Our data on the DNA binding and transcription regulating functions of HOXD13(G220V), both in vivo and in cultured cells, indicate that this missense mutation does not cause a complete impairment of HOXD13 functions. Thus, the G220V substitution is likely either to cause stable structural alterations, which only partially affect HOXD13 functions, and result in the aggregation and subsequent degradation of only a proportion of the protein, or to cause the misfolding of a fraction of the total protein, which gives rise to cytosolic aggregates and is eventually degraded. Both the lack of a correct structure, or the misfolding of a fraction of total HOXD13(G220V), provide an explanation for our unforeseen results on the reduced DNA binding activity of the mutated protein. Thus, even if the G220V mutation is located outside of the homeodomain region, which is necessary and sufficient for DNA binding by homeodomain proteins (34,35), HOXD13(G220V) binds DNA less efficiently than HOXD13 wild-type due to the structural alterations produced by the mutation. However, given the fact that the expression of high levels of HOXD13(G220V), both in transfection and in chick limb experiments, does not compensate for the loss in HOXD13(G220V) activity, we favour the hypothesis that the G220V mutation produces stable structural alterations of the HOXD13 protein, which partially impair its function and cause its accelerated decay within cells.
The phenotype produced by the G220V mutation shares several features with the phenotypes displayed by moderate polyalanine tract expansion mutations that are placed at the mild end of the SPD phenotypic spectrum. Polyalanine tract expansion mutations are held to cause limb malformations, whose severity correlates with the length of the alanine expansions (22), via a dominant negative effect (24–26). The analysis of the biochemical properties of HOXD13 polyalanine tract expansion mutant proteins showed that these mutations result in misfolding, cytosolic aggregation and degradation in relation to the length of the repeats (26). Thus, mutations as different as polyalanine expansions and the G220V missense mutation produce phenotypes having common traits by causing structural alterations of the HOXD13 protein that result in similar degradative processes, leading to a dominant loss-of-function of the haploinsufficient HOXD13 gene.
MATERIALS AND METHODS
Patient evaluation and mutation analysis
Venous blood samples for DNA extraction were obtained from the proband, both parents, and 14 other family members, with their informed consent and the approval of the local research Ethics Committee. To search for mutations in HOXD13 (GenBank accession nos AF005219 and AF005220) the entire coding region of the gene was amplified by PCR in four segments, as described previously (22). Amplified fragments were either directly cycle sequenced (Applied Biosystems Prism Dye Terminator Kit) or subcloned into pCRScript (Stratagene) before being cycle sequenced and analysed on an ABI 377 automated sequencer (Applied Biosystems).
Expression constructs and reporter plasmids
Reporter constructs: the pT81(TTAT)6 and pT81(TTAC)6 reporter constructs were generated by cloning six copies of the 5′-GCCGTCCGTTTTATTGGGGACACA-3′, or of the 5′-GCCGTCCGTTTTACGAGGGACACA-3′, oligonucleotides (16) into the pT81luc reporter plasmid (46). The pT81HCR (33), pT81UAS (40) and pT81Barx1 (8) luciferase reporter constructs were described previously. Expression constructs: pSG5-HOXD13 wild-type (WT) was generated cloning the human HOXD13 coding sequence into the EcoRI–BglII sites of the SV40 promoter-based pSG5 mammalian expression vector (Stratagene). pSG5-FLAG-HA-HOXD13WT was generated by cloning the human HOXD13 coding sequence in frame with the FLAG-HA epitopes into the blunted BamHI site of pSG5. The G220V missense mutation was produced by PCR mutagenesis using the primer 5′-CTACATCTCCATGGAGGTGTACCAGTC-3′. The amplified fragment was sequence-verified and substituted for the corresponding wild-type NcoI–BglII fragment within pSG5-HOXD13WT and pSG5-FLAG-HA-HOXD13WT. Expression constructs for the HOXD13Δ1–131WT and HOXD13Δ1–131G220V deletion mutants were obtained by PCR mutagenesis using the 5′-CCAGAATTCATGTCGCACGGCGTGGGC-3′ and D13-3′: 5′-ATGCGGATCCGAATTCTCAGGAGACAGTATCTTTG-3′. The HOXD13Δ132–267 interal deletion mutant was produced via splicing by overlapping extension (SOE) PCR, using the following primers: D13-5′: 5′-ATGCGAATTCGGATCCATGGACGGGCTGCGGGCAG-3′; D13SOE-R: 5′-CTCTCTTCTTCCTCCCACGGCAGCTGTAGTAGCC-3′; D13SOE-F: 5′-GGCTACTACAGCTGCCGTGGGAGGAAGAAGAGAGTGCC-3′; D13-3′: 5′-ATGCGGATCCGAATTCTCAGGAGACAGTATCTTTG-3′. The amplified PCR fragments were sequence-verified and cloned into the BamHI site of pSG5 vector. The pSGGal4DBD vector was described by Vigano et al. (40). The GAL4 fusion protein-expression constructs were generated by cloning the N-terminal regions of HOXD13 upstream to the GAL4DBD coding sequence in the EcoRI–BamHI sites of the pSG5 vector, and downstream to the GAL4DBD coding sequence in pGAL vector (47). The N-terminal fragments of the HOXD13 coding sequence were obtained by PCR using the following primers: for HOXD13(132–267)WT or G220V: D13F(132–267): 5′-CCAGAATTCATGTCGCACGGCGTGGGC-3′; D13R(132–267): 5′-CGCGAGCTCTCTTCGGTAGACGCACATGTC-3′; for HOXD13(1–131): D13-5′ and D13R(1–131): 5′-CGCGAGCTCACGGCAGCTGTAGTAGCCG-3′. RCAS avian retroviral expression constructs for HOXD13WT and HOXD13G220V and for their HA-tagged versions were generated by cloning blunted EcoRI/BglII and EcoRI restriction fragments containing the respective coding regions into the shuttle vector pSLAX13 and then subcloning into the retroviral vector pRCAS(BP)A (37).
Cell culture and transient transfections
P19 mouse embryonal carcinoma cells were cultured in Minimum Essential Medium Alpha (αMEM) (Gibco-Invitrogen), whereas COS7 and DF1 chick fibroblast cells, in Dulbecco’s Modified Eagle’s Medium (DMEM). Media were supplemented with 10% fetal bovine serum (Gibco-Invitrogen), 2 mm L-glutamine (Gibco-Invitrogen), 100 U/ml penicillin and 100 µg/ml streptomycin. Transfections were carried out by CaPO4 precipitation (48). In a typical experiment, P19 cells were transfected with 2.5-4 µg of reporter plasmid, 0.1-1 µg of expression construct and 0.1 µg of pCMV-βgal (Clontech), reaching a total amount of 10 µg DNA per 6 cm dish, with pBluescript vector. After 48 h, cells were harvested and used for luciferase and β-galactosidase assays.
CHX cell treatment, cell extracts, immunofluorescence staining and immunoblotting
COS7 cells were transfected with 5 µg of HOXD13 expression vectors. Twenty-four hours after transfection cells were trypsinized and replated to compensate for differences in transfection efficiencies. CHX was then added at a final concentration of 200 µg/ml. After 2, 4, 8, 10 and 12 h of treatment cells were harvested in TEN buffer (40 mm Tris pH7.9, 150 mm NaCl). Total cell extracts were prepared by lysing cells in 10 mm Hepes pH7.9, 400 mm NaCl, 0.1 mm EGTA, 0.5 mm DTT, 5% glycerol, 0.5%PMSF, 1% PIC (Roche). Fractioned nuclear and cytoplasmic cell extracts were obtained by harvesting cells in TB buffer [20 mm Hepes pH7.3, 110 mm K-acetate, 5 mm Na-acetate, 2 mm Mg-acetate, 1 mm EGTA, 1 µg/ml PIC (Roche)], followed by the addition of 40 µg/ml digitonine and centrifugation at 14 000 rpm. The nuclear fraction was then washed with 10 mm Hepes pH7.9, 1.5 mm MgCl2, 10 mm KCl, 1.5 mm DTT, 2 mm PMSF, followed by the addition of 20 mm Hepes pH7.9, 25% glycerol, 420 mm NaCl, 1.5 mm MgCl2, 0.2 mm EDTA, 2 mm PMSF, 1.5 mm DTT to extract nuclear proteins. Immunoblotting to detect HOXD13 or tagged HOXD13 were performed with a specific mouse monoclonal antibody against HOXD13 (kind gift of A. Kuroiwa), with an anti-FLAG (Sigma), or with an anti-HA antibody (Santa Cruz Biotechnology). α-NFY-B (kind gift of C. Imbriano) and α-actin (Sigma) antibodies were used as loading controls for nuclear and total extracts, respectively. Chemiluminescent detection was performed using the ECL kit (Amersham Bioscience). For immunofluorescence staining, COS7 cells were transfected with 10 µg of the HOXD13 expression vectors. Forty-eight hours after transfection, cells were fixed in pre-cooled methanol, blocked (PBS with 1% BSA) and incubated overnight with primary anti-HA antibody (Santa Cruz Biotechnology) at 4°C in a wet chamber. Fluorescein-conjugated secondary antibody (Molecular Probes) was incubated at room temperature for 2 h. Nuclei were counterstained with DAPI (Sigma).
Electrophoretic mobility shift assays
Full-length HOXD13WT, HOXD13G220V and their deletion mutant derivatives were synthesized in vitro using the TNT-coupled transcription/translation system (Promega) and used in EMSAs as described by Caronia et al. (16) together with the following oligonucleotide probes 5′-GCCGTCCGTTTTATTGGGGACACA-3′ and 5′-GCCGTCCGTTTTACGAGGGACACA-3′. The total amount of reticulocyte lysate in each binding reaction was adjusted to normalize for translated protein content.
Avian retrovirus production, microinjection, cartilage staining and whole mount in situ hybridization
DF1 chick fibroblast cells were transfected using FUGENE (Roche) with 1 µg of the RCAS-HA-HOXD13WT or RCAS-HA-HOXD13G220V retroviral constructs to generate virus stocks, subsequently harvested, concentrated and titrated as described by Morgan and Fekete (37). A titre of ∼1 × 108 cfu/ml was obtained for each virus. Microinjection of fertilized chick eggs was performed as described Caronia et al. (16). Whole-mount in situ hybridizations, using digoxigenin-labelled anti-sense RNA probes were carried out according to the study by Wilkinson (49). Hybridization probes were already described (8).
This work was supported by grants from the Italian Association for the Study of Malformations (ASM), the Italian Telethon and the Italian Association for Cancer Research (AIRC) to V.Z.; and by a grant (G.0199.07) from the Research Foundation Flanders to P.T.
We are grateful to the family for their participation in our study. Thanks are also due to Carol Imbriano for the anti-NF-YB antibody, to M.A. Ros for the kind gift of the cHand2 probe and to Valentina Salsi for sharing materials and for providing help with whole mount in situ hybridizations. N.B. is a PhD Fellow from the Fund of Scientific Research, Flanders, Belgium, and recipient of an EMBO Short-Term Fellowship. P.D. is a Clinical Research Investigator from the Fund for Scientific Research, Flanders, Belgium.
Conflict of Interest statement. None declared.