The formation of the thumb requires direct modulation of Gli3 transcription by Hoxa13

Significance

To understand the special morphology of the thumb (digit 1), an important adaptation of the human hand, we studied Hoxa13-null mice that specifically lack this digit. We show that Hoxa13 mutant limbs specifically lose Hoxd13 expression in the presumptive digit 1 region, correlating with increased Gli3 transcription and Gli3 repressor activity. We also show that Hox13 paralogs regulate Gli3 transcription by negatively modulating the activity of Gli3 enhancers. Our results indicate that mutual antagonism between Gli3 and Hox13 paralogs determines the anterior−posterior asymmetry of the handplate and that Hoxa13 acts hierarchically to initiate formation of a digit 1 territory by attenuating Gli3 transcription and enabling second-phase expansion of 5′Hoxd expression.

Abstract

In the tetrapod limb, the digits (fingers or toes) are the elements most subject to morphological diversification in response to functional adaptations. However, despite their functional importance, the mechanisms controlling digit morphology remain poorly understood. Here we have focused on understanding the special morphology of the thumb (digit 1), the acquisition of which was an important adaptation of the human hand. To this end, we have studied the limbs of the Hoxa13 mouse mutant that specifically fail to form digit 1. We show that, consistent with the role of Hoxa13 in Hoxd transcriptional regulation, the expression of Hoxd13 in Hoxa13 mutant limbs does not extend into the presumptive digit 1 territory, which is therefore devoid of distal Hox transcripts, a circumstance that can explain its agenesis. The loss of Hoxd13 expression, exclusively in digit 1 territory, correlates with increased Gli3 repressor activity, a Hoxd negative regulator, resulting from increased Gli3 transcription that, in turn, is due to the release from the negative modulation exerted by Hox13 paralogs on Gli3 regulatory sequences. Our results indicate that Hoxa13 acts hierarchically to initiate the formation of digit 1 by reducing Gli3 transcription and by enabling expansion of the 5′Hoxd second expression phase, thereby establishing anterior−posterior asymmetry in the handplate. Our work uncovers a mutual antagonism between Gli3 and Hox13 paralogs that has important implications for Hox and Gli3 gene regulation in the context of development and evolution.

Many of the genes that play important roles in limb patterning have been identified, yet little is known about how differential gene expression patterns are implemented to lead to specific morphological traits. A critical event in the adaptive evolution of tetrapod limb function has been the establishment of a polarized digit pattern with a distinctive anterior digit enabling grasping motions, and ultimately leading to the formation of an opposable thumb. This has been explained, in part, by recruitment of Sonic hedgehog (Shh) expression to the limb, but the regulatory network ensuring formation of a distinct anterior digit 1 remains poorly understood.

In vertebrates, Hh signaling is transduced by Gli transcription factors (Gli1, Gli2, and Gli3), of which Gli3 plays a predominant role in the developing limb (1). Shh signaling stabilizes full-length Gli3, which works as a mild activator, and prevents its proteolysis to a truncated potent repressor (Gli3R) of Shh-regulated targets that dominates in the absence of Shh signaling (1–3). Thus, whereas digit formation is severely curtailed in the absence of Shh (4–6), complete loss of Gli3 function, as in the extratoes spontaneous mutation in mice, results in a dramatic expansion in number of nonpolarized digits reminiscent of ancestral tetrapod polydactylous limbs (7). This polydactylous phenotype is unaltered in compound Shh;Gli3-null mutants (2, 3), underscoring the dominant role Gli3R levels play in digit regulation. A gradient of Shh, arising from the zone of polarizing activity (ZPA) in the posterior limb bud, acts both to reduce anterior digit number and to polarize digits by modulating Gli3R. Because Shh suppresses Gli3 processing, Gli3R levels in the limb bud form an opposing gradient to that of Shh.

Other major regulators of digit patterning include the Hox genes, in particular, members of the HoxA and HoxD clusters (8, 9). During tetrapod limb development, Hoxa genes are sequentially activated in the distal limb bud (10), but their expression evolves to generate mutually exclusive expression domains of Hoxa11 and Hoxa13. Hoxa11 becomes confined to the zeugopod (forearm), while Hoxa13 is expressed in the autopod (hand). Hoxd gene expression progresses differently, in 2 successive phases (11, 12). The first phase occurs in the early limb bud, mainly involves Hoxd4 to Hoxd11, and correlates with the specification of the stylopod (upper arm) and zeugopod morphology. The second phase takes place in the autopod, mainly involves Hoxd10 to Hoxd13, correlates with autopod morphology, and separates from the first phase by a band of tissue devoid of Hoxd transcripts that corresponds to the wrist/ankle. These precise patterns of expression of both Hoxa and Hoxd genes rely on complex transcriptional regulation that involves multiple long-range enhancers located within the flanking genomic regions (13, 14). In addition, Hoxa13 has recently emerged as a major transcriptional regulator controlling the mutually exclusive Hoxa11 expression domain through a repressive mechanism (15), as well as the switch from the first to the second phase of Hoxd expression (16–18).

Interestingly, the Shh/Gli3 signaling system and Hoxd transcription factors are in constant interplay during limb development. Initially, Hoxd proteins contribute to activate Shh transcription (19, 20); subsequently, Shh function is essential to relieve Gli3R repression of Hoxd gene second-phase expression, enabling digit formation (2, 3, 21–23). In addition, endogenous limb bud 5′Hoxd proteins interact physically with Gli3 in vivo, which could serve to sequester Gli3R and provide a mechanistic basis for the observed stoichiometric functional antagonism between Gli3 and 5′Hoxd genes at the phenotypic level (24). Additionally, high Hoxd12 levels were able to convert Gli3R from a repressor to a transactivator of Gli-regulated targets in transfections, which may also contribute to the regulation of digit patterning (24).

Genetic studies indicate that both Hoxa/d paralogs play overlapping and partially redundant roles in regulating limb morphogenesis; for example, Hoxa11/d11 act in concert in the zeugopod, and Hoxa13/d13 act in the autopod (25, 26). Yet, surprisingly, such functional overlap is absent in the regulation of digit 1, which is uniquely affected by the sole absence of Hoxa13, in contrast to other digits (25, 27, 28).

Because of the importance of digit 1, the thumb in the human hand, we have here investigated the mechanistic basis for this absolute Hoxa13 requirement. We report that, in the absence of Hoxa13, the expression of Hoxd13 does not extend into the anterior limb bud, consequently leaving the presumptive territory of digit 1 devoid of any distal Hox expression, a circumstance sufficient to prevent digit condensation (25, 29). We also detect increased Gli3R activity in the anterior mesoderm resulting from the loss of the negative regulation exerted by Hoxa13 and Hoxd13 through direct binding to several enhancers in the Gli3 locus. Our results support a model in which Hoxa13 acts hierarchically to reduce Gli3 transcription and enable expansion of 5′Hoxd second phase, thereby establishing anterior−posterior (AP) asymmetry in the handplate.

Results

A Gene Dosage Effect in Hoxa13 Mutants.

Hoxa13^−/− mutants die at midgestation, due to placental and vascular defects (25, 28, 30). Although it has been reported that a small percentage of Hoxa13 homozygous mutants survive to adulthood in the C57BL/6J genetic background (27), the oldest homozygous embryos that we recovered in our colony were at embryonic day 16.5 (E16.5), despite being maintained in this genetic background. At this stage, the typical Hoxa13-null limb phenotype, consisting of absent digit 1 and syndactyly, was prominent (SI Appendix, Fig. S1A and refs. 25, 27, and 28). The phenotype becomes noticeable between E11.5 and E12.5 in association with the failure of digit 1 chondrogenic condensation to form as detected by Sox9, the earliest marker of chondroprogenitors (SI Appendix, Fig. S1B). Interestingly, a gene dosage effect for Hoxa13 was observed in heterozygous embryos as a poorly defined digit 1 condensation that eventually resulted in a mild but consistent digit 1 hypoplasia in adult heterozygotes (SI Appendix, Fig. S1 B and C).

Altered Hox Code Expression in the Anterior Limb Bud Mesoderm of Hoxa13^−/− Mutants.

The hallmark of digit 1 is the expression of Hoxd13 but not the other 5′Hoxd genes (31). This unique combination of Hox products is achieved during the second phase of Hoxd transcription when the expression of Hoxd13 spreads into the anterior handplate mesoderm, surpassing the anterior limit of Hoxd11 and Hoxd12 domains, a situation referred to as reverse collinearity (31). Due to the pivotal role played by Hoxa13 in the transition from phase 1 to phase 2 of Hoxd gene expression (16–18), the first phase of expression of Hoxd11 is abnormally prolonged in Hoxa13 mutants, especially in digit 1 territory and, the gap between the 2 phases is distally displaced (Fig. 1A). This also occurs with the first phase of expression of the more 3′Hoxd genes Hoxd4, Hoxd9, and Hoxd10, and with the Hoxa11 expression domain, which is normally restricted to the zeugopod, but becomes distally displaced in the absence of Hoxa13 (SI Appendix, Fig. S2) (18).

Fig. 1.

The 5′Hoxd gene expression in Hoxa13 mutant limb buds. Forelimb autopods of wild type and Hoxa13 heterozygous and Hoxa13 homozygous mutants hybridized with (A) Hoxd11, (B) Hoxd12, and (C) Hoxd13 at E 12.5. The arrowheads point to altered expression patterns in homozygous mutants. (Scale bar: 500 μm.)

In addition, in Hoxa13-null autopods, Hoxd13 transcription failed to extend into digit 1 territory but remained similar to that of the unperturbed Hoxd12 (Fig. 1 B and C). Thus, digit 1 progenitors in Hoxa13 mutants express an altered Hox code that corresponds to the zeugopod, rather than the autopod, as it includes Hoxa11 and the first-phase Hoxd expression but lacks the characteristic Hoxd13 expression (15–18). The fact that the prospective digit 1 cells in Hoxa13-null embryos are devoid of both Hoxa13 and Hoxd13 products may account for the loss of digit 1, as all digit chondrogenic condensations fail to form in the absence of both Hox13 paralogs (25, 29).

No differences in the pattern of expression of other markers of the anterior mesoderm such as Tbx2, Tbx3, and Alx4, which are expressed primarily at a proximal level, were observed in Hoxa13-null limb buds (SI Appendix, Fig. S3). Thus, the gene expression perturbations in Hoxa13 mutants are a late event, restricted to the autopod, as expected for a gene that is first activated in the forelimb around E10.5 (32).

Increased Gli3R Activity in Hoxa13^−/− Anterior Mesoderm.

Our results indicate that Hoxa13 is required, directly or indirectly, for the normal anterior spread of Hoxd13. However, we previously showed that Hoxd13 is strongly and uniformly expressed all along the AP extent of the handplate in compound Hoxa13;Gli3 mutants (ref. 29 and SI Appendix, Fig. S4). Indeed, in the absence of Gli3, the second phase of expression of all 5′Hoxd genes occurs symmetrically all along the AP axis of the handplate, providing a similar Hox code, and presumably a similar amount of Hox products, to all digits (SI Appendix, Fig. S4) (2, 3). These considerations suggest that Hoxa13 regulation of Hoxd13 transcription may be mediated by modulating Gli3 and prompted us to analyze the state of the Gli3R in Hoxa13 mutants.

RNA in situ hybridization showed up-regulation and distal expansion of the expression domain of Pax9, a GLI3R activated target gene (33), while Jag1, a GLI3R repressed target gene (33), was absent from the anterior mesoderm of Hoxa13 mutants at E11.5 and E12.5 (Fig. 2 A and B). Bmp4, a gene whose expression correlates positively with the level of Gli3R (34), showed a more robust and extended domain in the anterior mesoderm (Fig. 2C). Overall, these expression pattern modifications are consistent with increased Gli3R activity in the anterior mesoderm of Hoxa13 mutants. Accordingly, RT-qPCR in dissected E11.75 anterior mesoderm (Fig. 2D) showed a 1.7-fold increase in Gli3 expression (n = 4, P < 0.05) in Hoxa13 mutants that was accompanied by a 90% decrease in the level of Hoxd13 expression (n = 4, P < 0.001), confirming our in situ hybridization results (Fig. 1C). In line with this quantitative increase in transcription, Western blot analysis confirmed a slightly higher level of Gli3R in the anterior limb mesoderm of Hoxa13 mutant limb buds (n = 10, P < 0.05; Fig. 2E).

Fig. 2.

Expression of Gli3R target genes and quantification of Gli3 mRNA and protein levels in Hoxa13 mutant limb buds. Pattern of expression of (A) Pax9, (B) Jag1, and (C) Bmp4 in E11.5 and E12.5 forelimb buds of wild type and Hoxa13 homozygous mutants. Altered expressions are indicated by arrowheads. (Scale bar: 500 μm.) RT-qPCR quantification of (D) Gli3 and Hoxd13 mRNA and (E) Gli3R protein levels in the anterior mesoderm of wild type and Hoxa13 homozygous mutants. (*P < 0.05; ***P < 0.001).

Since Gli3 processing depends on Shh activity (35), we investigated the state of the ZPA in Hoxa13 mutants. In situ hybridization failed to detect any difference in Shh expression or signaling (SI Appendix, Fig. S5 A–C). Analysis of the expression of Fgf8, the best marker of the apical ectodermal ridge, showed a mild anterior down-regulation in homozygous limb buds that was reflected in reduced Spry4, but not Dusp6 (SI Appendix, Fig. S5 A, D, and E). This situation, probably secondary to the increase in Bmp4, was also accompanied by a slight down-regulation in the anterior propagation of Grem1 (SI Appendix, Fig. S5F). However, these changes did not reach statistical significance when analyzed by RT-qPCR (P > 0.05; SI Appendix, Fig. S5G).

Altered Gli3 Expression Pattern in the Absence of Hoxa13.

The previous results revealed a moderate excess of Gli3 transcription and GLI3R protein, and corresponding Gli3R activity in the anterior mutant autopod. To examine in detail how Hoxa13 impacts the overall pattern of Gli3 expression, we performed a systematic temporal in situ hybridization analysis that exposed a highly dynamic expression pattern. Because of the similarity with 5′Hoxd genes, we describe it as evolving in 2 phases (Fig. 3A). As reported (23), at E10.5, Gli3 was expressed in the distal mesoderm, except for the posterior border where Shh is expressed. Between E10.5 and E11.5, Gli3 expression in the distal mesoderm became progressively down-regulated from posterior to anterior until being confined first to the anterior autopod (prospective digit 1 territory) and then to the wrist. By E11.5, a second domain of expression started in the autopod, roughly overlapping digit 4 primordium and progressively propagating anteriorly to cover all of the digit primordia. Therefore, by E12.5, 2 domains of expression, proximal and distal, were clearly distinct and separated by a gap of tissue devoid of transcripts. The proximal domain, a remnant of the first phase of expression, remained as a transverse band at the zeugopod−autopod boundary. The distal domain in the digital plate, corresponding to the second phase of expression, became progressively confined to the interdigital tissue as the digit condensations differentiated (E12.5) and then to the interphalangeal joints (E13.5).

Fig. 3.

Dynamics of Gli3 expression pattern in wild type and Hoxa13 mutants. (A) Expression of Gli3 during forelimb development in control and Hoxa13 homozygous mutant limb buds. (B) Schematic representation of Gli3 expression pattern showing phase 1 in lighter green and phase 2 in darker green. (C) Comparison of Gli3 and Hoxd13 domains of expression in the 2 limb buds of the same wild-type embryo at E10.5 and E11.5 accompanied by an explanatory drawing depicting the abutting of expression domains. (Scale bars: 500 μm.)

In the absence of Hoxa13, the dynamics of Gli3 expression were dramatically altered from E11, as the down-regulation of the first phase of expression was delayed and incomplete, never reaching the digit 1 territory (Fig. 3A). As a consequence, the digit 1 territory remained within the first phase of Gli3 expression even at later stages, while the second distal domain was restricted to the posterior digits 2 to 4 (schematic summary shown in Fig. 3B). This analysis uncovers a previously unidentified hierarchical effect of Hoxa13 in Gli3 regulation.

The similarity in the expression pattern of Gli3 and Hoxd10-11 in E12.5 Hoxa13 mutants (compare Figs. 3A and 1A) suggested similar regulation and, considering the cooperation of Hoxd13 with Hoxa13 in regulating Hoxd transcription (16–18), pointed to 5′Hoxd genes also playing a role in regulating Gli3 expression. Notably, the analysis of Gli3 and Hoxd13 expression in each of the 2 limbs of the same embryo showed that the anterior boundary of Hoxd13 always abutted the posterior boundary of first-phase Gli3 expression (Fig. 3C), supporting Hoxd13 implication in modulating Gli3 transcription.

Hox13 Negatively Modulates the Activity of Gli3 Enhancers.

To explore the mechanism for Hoxa13 and Hoxd13 modulation of Gli3 transcription, we screened the Gli3 genomic locus for Hoxa13 and Hoxd13 (hereafter conjointly referred to as Hox13) binding sites. We used published datasets of Hox13 DNA binding in limb buds, as well as chromatin state and transcriptome changes between wild type and Hoxa13;Hoxd13 double mutants (17). We identified several Hox13 binding regions, 4 of which (highlighted in Fig. 4A) were also enriched in H3K27ac marks, a mark associated with active enhancers. In agreement with this possibility, 2 of the Hox13 peaks (highlighted in blue in Fig. 4A) overlapped with previously reported VISTA enhancers, mm1179 and hs1586, which are active in the limb (36). These observations suggest that Hoxa13 and Hoxd13 could directly regulate Gli3 transcription, and, based on our in situ and RT-qPCR analyses, this regulation is expected to be negative, consistent with elevated Gli3 transcript levels (1.7-fold, false discovery rate [FDR] ≤ 0.05) observed in the RNA sequencing (RNA-seq) profiles generated in Hoxa13;Hoxd13 double-mutant autopods (Fig. 4A, 2 bottom tracks) (17).

Fig. 4.

Hox13 regulation of Gli3 expression. (A) University of California, Santa Cruz (UCSC) Genome Browser view of the regulatory landscape upstream Gli3 (16, 17). The black line indicates the extension of the topologically associating domain (TAD) in embryonic stem (ES) cells as in Dixon et al. (45). Chromatin immunoprecipitation (ChIP)-seq tracks for Hoxa13 and Hoxd13 are shown in purple at the top. The H3K27ac and H3K27me3 profiles of the digital plate of E11.5 limb buds are also shown in green and orange, respectively. Two replicates of the transcriptome profiling in wild type (blue) and Hox13^−/− mutant (red) limb buds at E11.5 are included at the bottom. Two Hox13 binding sites with potential enhancer activity are highlighted in yellow (RI and RII). Two other Hox13 binding sites, highlighted in blue, overlap with 2 previously validated VISTA (36) elements (mm1179 and hs1586), and their activity at E11.5 (Vista Enhancer Browser) is shown below. The activity of RII (LacZ reporter) is also shown, with a magnification of the forelimb. (B) Changes in eRNA levels between Hoxa13-null and wild-type anterior limb bud mesoderm. The eRNA levels were measured for the indicated enhancers using primers located both upstream (5′) and downstream (3′) of the corresponding Hox13 peaks. The eRNA levels were normalized using Vimentin as a housekeeping gene. The eRNA levels 3′ of RI were not measured, since appropriate primers could not be designed. Significance of differences was determined using the 2-tailed, Student’s t test (n = 3; *P < 0.05). (C) Deregulation of Gli3 expression in Hoxa13;Hoxd11-13 compound mutants. Genotype is indicated on the top, and stage is indicated on the left. Note that, while the removal of Hoxd11-13 has no consequences (second column), its removal in the absence of Hoxa13 (third column) has a stronger impact in Gli3 expression pattern than the sole removal of Hoxa13 (Fig. 3A). Finally, in the complete removal of Hoxa13 and Hoxd11-13, Gli3 expression remains in most of the distal autopod (fourth column). (D) FLAG immunoprecipitation of lysates from E12.5 heterozygous 3x-FLAG-Gli3 limb buds immunoblotted with the FLAG antibody showed the Flag-Gli3 band (a). Reprobing with an antibody specific for Gli3 additionally showed the wild-type Gli3 band (b). The FLAG-Gli3 immunoprecipitates failed to detect CoIP of Hoxa13 using the Hox13 antibody (c). Finally, immunoprecipitates using immobilized Hoxa13 antibody and immunoblotted with the FLAG antibody also failed to detect CoIP of the FLAG-Gli3 fusion protein (d). (E) Schematic diagram indicating the interactions between Gli3 and 5′Hox genes. (Scale bars: A and C, 500 μm.)

To evaluate whether Hox13 proteins can negatively regulate Gli3 expression through distal enhancer elements, we took advantage of the fact that enhancer activity positively correlates with the generation of short (50 to 2,000 nucleotides) bidirectional transcripts known as enhancer RNAs (eRNAs) (37). Therefore, we measured eRNA levels as a readout of enhancer activity around some of the Hox13 peaks identified within the Gli3 locus in wild-type and Hoxa13 mutant contexts. Namely, we analyzed the mm1179 VISTA enhancer and 2 previously undiscovered enhancers, that we called Region I (RI) and Region II (RII), using amplicons located upstream (5′) and downstream (3′) of the corresponding Hox13 peak (primer list in Table 1). Using RT-qPCR, we found that loss of Hoxa13 significantly increased eRNA levels at the 2 novel putative Gli3 enhancers analyzed, the 5′ region in RI, and the 3′ region in RII (n = 3, P < 0.05; Fig. 4B). Moreover, a general trend toward increased eRNA levels upon loss of Hoxa13 was also observed for the remaining amplicons. The hs1586 VISTA enhancer was not analyzed, because of its intronic localization that makes its transcription difficult to separate from Gli3 expression. To further examine the enhancer activity of these novel regions, we used a transient transgenic reporter assay expressing a lacZ cassette under the control of RI or RII. Whole-mount β-galactosidase staining for RII showed reporter activity in the E11.5 limb bud restricted out of the region of Hox13 expression, as expected for an enhancer whose activity is repressed by Hox13 (2 out of 6; Fig. 4A). The RI transgene did not show any activity in the limb, indicating that it may be more context-dependent, for example, requiring other cis regulatory elements missing in the transgenic assay, or it might represent a different type of distal regulatory element. Overall, these results indicate that Hoxa13 can negatively modulate the activity of enhancers controlling Gli3 expression in the autopod, which presumably leads to reduced Gli3 transcription.

Table 1.

List of primers used for analysis of eRNA expression

Primer name	Sequence F	Sequence R
RI eRNA5′	GCCCAAGCCCAGTTAATTGT	GAAAATGCACTGGAGGAGGC
RII eRNA5′	AGCAGCTTTTGAAGGCCATC	TGGATCAACTCAGAGCGTGT
RII eRNA3′	ACCCCAACAAGAATCCATGC	TGCAAGTGTTTTCCCCTCTG
mm1179_eRNA5′	TGGGAGGAAGAGTGTTACCG	AGTCCTGTTTTCCTGAGGGG
mm1179_eRNA3′	TGCAAAGTCACAGGCTTCAA	GACATCTTTCACAGCCCAGC

A negative regulation of Gli3 expression by distal Hox genes was also detected in the Hoxa13;Hoxd11-13 allelic series (Fig. 4C). In agreement with previous results showing that the absence of Hoxd11-13 had no major impact on Gli3 messenger RNA (mRNA) and protein expression (38), our analysis additionally confirmed that the pattern of Gli3 expression was normal in this mutant (Fig. 4C). However, the removal of one functional allele of Hoxd11-13 from the Hoxa13-null background had a stronger impact on Gli3 expression than the removal of Hoxa13 alone (Fig. 4C). Finally, in the total absence of distal Hox products (Hoxa13;Hoxd11-13 double homozygous mutants), the first phase of Gli3 expression persisted in the central autopod (Fig. 4C). These results support a dose-dependent function of Hoxa13 and 5′Hoxd proteins in negatively regulating first-phase Gli3 transcription. Notably, in preliminary RNA-seq experiments, we observed an increase in Hoxa13 mRNA, comparing Hoxd11-13^+/− and Hoxd11-13^−/− (E12.5 autopod, 1.6-fold, n = 3, FDR = 0.008); indeed, comparison of Hoxa13 protein levels (SI Appendix, Fig. S6 A and C) revealed an ∼7-fold increase in Hoxd11-13 mutants relative to wild-type controls. Given that Hoxa13 and Hoxd13 are known to have redundant functions, this increase in Hoxa13 could compensate for the loss of Hoxd11-13 and explain why the removal of Hoxd11-13 has no effect on Gli3 transcription (Fig. 4C) or on digit 1 formation.

No Evidence of Hoxa13−Gli3 Physical Interaction.

We also considered the possibility that Hoxa13 does not act simply to regulate Gli3 transcription, but by directly interacting with Gli3R protein, as has been shown for 5′ Hoxd11-13 (24). The binding of Hoxa13 protein to GLI3R could modify its activity either by sequestering Gli3R or by transforming its repressor activity into an activator. To test this hypothesis, we performed coimmunoprecipitation (CoIP) of E12.5 limb bud lysates from mice expressing an epitope-tagged allele of Gli3 [3XFLAG-BirA-Gli3 (39, 40)]. No interaction between Hoxa13 and Gli3 was detected, using either immobilized anti-FLAG to coimmunoprecipitate Hoxa13 or using anti-Hoxa13 (17) to coimmunoprecipitate Gli3 (Fig. 4D). Additional analysis with a different antibody further confirmed the lack of interaction between Hoxa13 and Gli3 proteins (SI Appendix, Fig. S6B), supporting the conclusion that the gain in Gli3R activity observed in the anterior autopod of Hoxa13 mutants results primarily from loss of transcriptional modulation by Hox13 proteins.

Discussion

The thumb, or digit 1, is a crucial adaptation for the functionality of the human hand. It is the last digit to form (41) and the one with higher risk of developmental disruption, with more than 1,000 syndromes in the Online Mendelian Inheritance in Man database, including thumb abnormalities (42). Here, we have used the Hoxa13-null mutant that specifically fails to form digit 1, to investigate the mechanisms controlling the formation of this digit.

The Absence of Hoxd13 Expression in Digit 1 Progenitor Cells Explains Digit 1 Absence in Hoxa13 Mutants.

We show that, in the absence of Hoxa13, the Hoxd13 expression domain in the autopod remains similar to that of Hoxd12, without any evidence of the so-called reverse collinearity (31). This is in agreement with the crucial role of Hoxa13 in promoting the second phase of Hoxd gene expression (16–18). Thus, in Hoxa13 mutants, digit 1 territory is devoid of Hox13 paralogs, a situation equivalent to Hoxa13;Hoxd13 double mutants, that explains the phenotype and highlights the essential function of Hox13 paralogs in the formation of the digital condensations (25, 29). Currently, digit patterning, understood as the generation of a periodic digit−interdigit pattern, is explained by a Turing-type or reaction−diffusion mechanism in which 5′Hox genes modulate digit spacing, in a dose-dependent manner (29). This model predicts that, in the total absence of distal Hox genes (Hoxa13 and Hoxd11-13), the area of digit patterning is strongly reduced, precluding the formation of digital condensations, nicely fitting with the situation selectively observed in digit 1 territory of Hoxa13 mutants.

Hoxa13 Mutants Display Elevated Gli3R Activity in the Anterior Mesoderm.

Although Hoxa13 plays a role in the anterior propagation of Hoxd13, this function is not required in the absence of Gli3. Actually, in the absence of Gli3, regardless of whether or not Hoxa13 is present, the second phase of expression of 5′Hoxd genes, including Hoxd13, uniformly spans the AP axis of the autopod (refs. 2, 3, 23, and 29 and SI Appendix, Fig. S4). Therefore, during normal development, Hoxa13 may function to modulate the repressor function of Gli3R to permit a fully realized second phase of Hoxd13 expression. Because of its higher transcriptional efficiency, Hoxd13 is considered to be less sensitive than Hoxd12 and Hoxd11 to repression by Gli3R, and therefore is the only 5′Hoxd member normally expressed in digit 1, the area of maximum Gli3R activity (31, 35). Thus, Hoxa13 could potentially attenuate Gli3R activity sufficiently to permit the propagation of Hoxd13, but not the other 5′Hoxd genes, to digit 1 territory, yielding reverse collinearity.

Supporting this hypothesis, we provide compelling evidence of a moderate increase in Gli3R expression and activity in the anterior mesoderm of Hoxa13 mutants, including elevated Gli3 transcription and Gli3R protein levels, and altered expression of known Gli3R target genes Pax9 and Jag1 (33). Collectively, our results are consistent with an excess of Gli3R in the anterior mesoderm precluding the normal spread of Hoxd13 expression and subsequent expansion and differentiation of digit 1 progenitors. Because digit 1 is the digit that normally develops with higher Gli3 levels, our results reveal that a balance of Gli3R level is crucial for digit 1 formation. Too much Gli3R precludes anterior spread of Hoxd13, leading to the loss of digit 1, while too little also permits Hoxd12 and Hoxd11 extension, leading to polydactyly with loss of anteriorization.

Yet digit 1 is the only digit that reportedly forms in the hindlimb with high levels of Gli3R, as occurs in Shh and Ozd mutants (4–6). However, it should be noted that this hindlimb digit, as well as the small growth that occurs in the Shh-null forelimb, is also accompanied by late-stage expression of both Hoxa13 and Hoxd13. Due to the massive cell death that occurs in both Shh and Ozd mutant limb buds, a cell lineage study would be required to determine the origin of the progenitors of the single rudimentary digit that forms.

Hox13 Paralogs Modulate Gli3 Transcription.

Of note, our analysis also uncovered a previously unappreciated highly dynamic pattern of Gli3 expression that, due to similarity with 5′Hoxd genes, we describe as evolving in 2 phases. The first phase of expression starts in the early limb bud and is then progressively down-regulated from posterior to anterior as the handplate forms, becoming confined to a transverse band at the distal zeugopod. The second phase of expression occurs in the digital plate and is separated from the first phase by a band of tissue devoid of Gli3 transcripts that corresponds to the wrist/ankle. This dynamic pattern is highly altered in the absence of Hoxa13, as the down-regulation of the first phase is delayed and incomplete, resulting in Gli3 expression remaining over the digit 1 territory. Consequently, when the second phase of expression is established, the gap between the 2 phases lies between prospective digits 1 and 2.

A direct function of Hoxa13 in controlling Gli3 transcription is supported by the presence of several Hox13 binding sites in the Gli3 genomic landscape, particularly at 2 novel putative enhancers named RI and RII, identified in this study. Using eRNAs as surrogates of RI and RII activity, we show that Hoxa13 normally reduces RI and RII enhancer activity, providing a mechanism for Gli3R attenuation. Furthermore, the evaluation of RII activity in transient transgenic assays in mice showed limb activity complementary to the area of high Hox13 expression, in agreement with Hox13 negatively modulating Gli3 expression and RII activity. In Hoxa13 mutants, the attenuation of RI and RII enhancer activity is lost, allowing for elevated and prolonged first-phase Gli3 expression sufficient to disrupt digit I specification.

Notably, in Hoxa13 mutant autopods, Gli3 and Hoxd10 and Hoxd11 patterns of expression are strikingly comparable. Indeed, the dynamics of the down-regulation of the Gli3 first phase is totally coincident with the progression of the second phase of Hoxd13. This points to a role for Hoxd13 in attenuating Gli3 transcription, which is supported by the Hox13 binding sites in the Gli3 genomic landscape and confirmed by the study of the Hoxa13;Hoxd11-13 allelic series.

Overall, our study uncovers a level of interaction between Hox genes and the Shh/Gli3 pathway: the regulation of Gli3 transcription by Hox13 paralogs therefore establishing a mutual antagonism that determines the AP asymmetry of the handplate (Fig. 4E). Of the 2 Hox13 paralogs, Hoxa13 acts hierarchically to initiate the formation of digit 1 by reducing Gli3 transcription and by enabling expansion of the 5′Hoxd second expression phase (Fig. 4E). Consequently, Hoxa13 can compensate for Hoxd13 loss in digit 1 territory, but Hoxd13 cannot compensate Hoxa13 loss, because its expression in that territory is downstream of Hoxa13 activity. On the other hand, Hoxd13 and the 5′Hoxd proteins further modulate Gli3R function through physical interaction (24), an activity that we demonstrate is not shared by Hoxa13.

Materials and Methods

Embryos, Skeletal Preparations, and In Situ Hybridization.

The Hoxa13 (25), Hoxd^{Del (11–13)} (43), and Gli3^XtJ (7) mutant strains were genotyped by PCR as published. Whole-mount skeletal preparations were performed by staining with Alcian blue 8GX (Sigma Aldrich) and Alizarin red S (Sigma Aldrich) following standard protocols. Whole-mount in situ hybridization was performed according to standard procedures using digoxigenin-labeled antisense riboprobes.

All animal procedures were conducted according to the European Union regulations and 3R principles and reviewed and approved by the Bioethics Committee of the University of Cantabria, and according to the ethical guidelines of the Institutional Animal Care and Use Committee (IACUC) at NCI-Frederick under protocol #ASP-12-405.

Western Blot and qRT-qPCR.

For immunoblots, the following antibodies were used: 1) mouse monoclonal Gli3 clone 6F5 anti-Gli3-N antibody, kindly provided by S. Scales at Genentech (44), 2) the goat polyclonal Gli3 (AF3690-SP; R&D Systems), and 3) βactin (C4SC-4778; Santa Cruz) assessed as normalization control. RT-qPCR was carried out on an Applied Biosystems StepOnePlus using SYBR Green Supermix (Bio-Rad), and the data were analyzed with the delta-delta Ct method using the StepOne software (SI Appendix). For evaluation of eRNA levels, total RNA was isolated using Tripure reagent according to the manufacturer’s instructions (Ref. 11667157001; Roche). Then, total RNA was treated with TURBO DNase-free kit (AM1907; Life Technologies) before complementary DNA synthesis and RT-qPCR. The primers used are listed in Table 1.

CoIP Analysis of Hoxa13 and Gli3.

Interaction of Gli3 with Hoxa13 was evaluated by CoIP. A detailed description of the CoIP method is provided in SI Appendix, Supplementary Materials and Methods. Briefly, antibodies specific for Hoxa13 (17) and for the FLAG epitope (M2 Sigma F1804) were used to immunoprecipitate Hoxa13 or the FLAG-tagged Gli3 protein from E12.5 limbs heterozygous for a 3XFLAG-Gli3 allele. FLAG-Gli3 immunoprecipitates were evaluated for CoIP of Hoxa13 using the Hoxa13 antibody, and Hoxa13 immunoprecipitates were immunoblotted and assessed for CoIP of the FLAG-Gli3 fusion protein using the FLAG antibody.

Data Availability.

All data and materials are available in the main text or SI Appendix.