Mammalian-specific ectodermal enhancers control the expression of Hoxc genes in developing nails and hair follicles.

Vertebrate Hox genes are critical for the establishment of structures during the development of the main body axis. Subsequently, they play important roles either in organizing secondary axial structures such as the appendages, or during homeostasis in postnatal stages and adulthood. Here, we set up to analyze their elusive function in the ectodermal compartment, using the mouse limb bud as a model. We report that the HoxC gene cluster was co-opted to be transcribed in the distal limb ectoderm, where it is activated following the rule of temporal colinearity. These ectodermal cells subsequently produce various keratinized organs such as nails or claws. Accordingly, deletion of the HoxC cluster led to mice lacking nails (anonychia), a condition stronger than the previously reported loss of function of Hoxc13, which is the causative gene of the ectodermal dysplasia 9 (ECTD9) in human patients. We further identified two mammalian-specific ectodermal enhancers located upstream of the HoxC gene cluster, which together regulate Hoxc gene expression in the hair and nail ectodermal organs. Deletion of these regulatory elements alone or in combination revealed a strong quantitative component in the regulation of Hoxc genes in the ectoderm, suggesting that these two enhancers may have evolved along with the mammalian taxon to provide the level of HOXC proteins necessary for the full development of hair and nail.

In most bilateria, genes members of the Hox family of transcription factors are important for the proper organization of structures along the main body axis, during early development. On top of this major task, vertebrate Hox genes are also necessary for the morphogenesis of a variety of secondary structures such as the appendages or the external genitalia (1). In amniotes, Hox genes are organized in four clusters (HoxA, B, C, and D), all of them being involved in the development of the trunk axis. Subsequently however, subgroups of gene clusters display some specificities for particular tissues or structures. For example, both the HoxA and HoxD clusters are essential for the organization of the tetrapod limb morphology (2–7), while the deletion of both HoxA and HoxB clusters leads to severe cardiac malformations, not detected in any of the single deletion mutants. Likewise, deletions of both the HoxA and HoxC clusters induced a complete renal aplasia, which was not detected with either deletion alone (8).

Such a functional redundancy between Hox clusters can be explained by their patterns of duplication during the evolution of vertebrates, with particular functional traits being conserved in all or some gene clusters after duplication (8). In this context, Hox gene function associated with recent vertebrate features, i.e., features that must have appeared late, after the two rounds of duplication, can be expected to involve single gene clusters and thus to show less functional redundancy. Examples of this situation may be found either in the function of some Hoxa genes in uterine physiology (9–11) or in the apparent specialization of some Hoxc genes for epidermal derivatives (12, 13), while other clusters apparently do not play any detectable function there.

Indeed, while the single deletion of the HoxC gene cluster did not reveal any gross phenotype (14), several studies have suggested that Hoxc genes are involved in the development of ectodermal organs (12, 15). The fact that such organs usually become fully developed after birth explains why the full HoxC cluster deletion, which is lethal at birth because of respiratory problems (14), did not unmask these functional contributions. However, both expression and inactivation studies have revealed potential contributions, in particular at late fetal stages as well as during postnatal development and even adulthood (12, 16, 17). In such instances, some Hoxc genes display complex patterns of expression in relation to the differentiation of epidermal organs such as the hairs and nails (12, 15, 18, 19), two ectodermal derivatives that form through epithelial–mesenchymal interactions between the surface ectoderm and the underlying mesoderm, the latter being responsible for the development of site-specific skin derivatives (17, 20).

The morphogenesis of both hairs and nails begins with the induction of ectodermal thickenings or placodes by a subjacent dermal condensate (21). Once the placode forms, signaling events between the ectodermal and mesodermal components drive proliferation of the ectoderm to produce the hair peg or the nail matrix and eventually lead to a mature hair follicle (HF) or nail organ. Both the morphogenetic steps and the signaling molecules involved are remarkably similar during the development of these two integumentary structures (reviewed in refs. 22 and 23), and there is evidence suggesting that Hoxc genes may participate in these processes. This is mostly supported by the inactivation of Hoxc13, which induced specific hair and nail defects in adult homozygous mutant animals (12, 13). These mice showed long and fragile nails; they lacked vibrissae at birth and subsequently displayed a nude appearance with no external pelage. Although the morphogenesis of HFs appeared normal, hair differentiation was defective with very fragile shafts that broke as soon as they emerged through the skin surface, leading to alopecia. Both the hair and claw defects were associated with defective keratin gene expression, as Hoxc13 was reported to be an important regulator of various keratin genes (15).

In order to evaluate the regulatory mechanism at work at the HoxC locus, i.e., to assay first how many Hoxc genes are involved into the development of such epidermal organs and, second, whether a single epidermal regulatory module may control their expression there in a coordinated manner, we selected the developing mouse limb bud as model system, for it allows to microdissect out the ectodermal component from where nails and hairs subsequently arise. We used RNA-seq on this isolated ectodermal jacket during both the early patterning phase and also the subsequent phase of tissue-specific differentiation. Here, we show that Hoxc genes, unlike both Hoxa and Hoxd genes, are specifically expressed in a colinear manner in the limb bud ectoderm. At later fetal stages, expression becomes localized in the developing nail region and HFs. Accordingly, we report that mice lacking the HoxC cluster suffer from anonychia (absence of nail or claw) at birth. We show that this global transcriptional control is largely associated with two cis-regulatory regions and that only the deletion of both enhancers together led to a phenotype involving a disheveled fur and hypoplastic nails. Finally, the removal of both enhancers in a weakened HoxC genetic background caused a striking cyclic alopecia. We conclude that the HoxC cluster was evolutionary co-opted for exerting an essential function during the development of mammalian the integuments. As reported for other Hox clusters in different contexts, this co-option was at least partly achieved through the emergence of remote enhancers with ectodermal specificities.

Results

Colinear Expression of Hoxc Genes in the Limb Bud Ectoderm.

Initially, we set out to study in detail how Hoxc genes are regulated in the developing limb by carefully isolating the ectodermal layer from the mesoderm component. To this aim, we generated RNA-seq datasets of both microdissected mouse distal forelimb mesenchymal progenitors and the overlying ectodermal cells at four consecutive developmental stages. We selected embryonic day 9.5 (E9.5), E10.5, E11.5, and E12.5 limb buds because they illustrate a wide range of processes, from the early emergence of the limb bud to the specification of the digits. Due to the progressive proximo-distal development of the limb bud, our analysis was restricted to the distal 150 μm, which is considered a good approximation of the progress zone (24–26). As a consequence, we microdissected a 150-μm-thick band of the distal limb bud, and the mesoderm and ectoderm components were separated by mild trypsin digestion and processed separately ( and ). Two biological replicates were collected per stage, and each pair of ectoderm and mesoderm replicates was obtained from the same batch of embryos.

Because a few mesoderm cells could occasionally remain attached to the ectoderm during dissection, potentially contaminating the ectodermal samples, we evaluated the purity of samples by analyzing the expression of genes known to be expressed exclusively in the ectoderm or mesoderm components of limb buds. Fgf8, Sp6, Trp63, Perp, Krt14, Sp8, and Wnt7a were selected as ectoderm specific genes, whereas Prrx1, Tbx5, Hoxa9, Hoxd10, Fgf10, Pecam1, and Cdh5 were selected as mesoderm specific genes (1, 27–31). A heatmap visualization of the transcription profiles of these genes across samples showed that the separation of the ectoderm and the mesoderm was clean, with a negligible level of contamination between both limb components (), except in replicate 1 of the E11.5 ectoderm samples where a slight contamination with mesodermal cells was scored (Fig. 1).

Fig. 1.

Colinear activation of Hoxc genes in the limb ectoderm. (A) Heatmap showing the log2(1 + FPKM) values of all Hox genes across the samples. Hoxc genes are activated in the ectoderm following a temporal sequence. The ectoderm and mesoderm samples were obtained as indicated in . (B) Expression profiles of HoxC cluster genes in the four ectodermal (ECT) stages analyzed. Data were normalized to the million uniquely mapped reads, and mean of duplicates is presented. (C) mRNA ISH in tissue sections showing expression of Hoxc5 and Hoxc8 in the early E9.5 limb bud ectoderm and of Hoxc12 and Hoxc13 in the late E12.5 limb bud ectoderm. Arrowheads indicate the ectodermal restriction of expression (n = 2). (Scale bars: 100 µm.) (D) Whole-mount ISH showing restriction of Hoxc12 and Hoxc13 expression to the forelimb digit tips at E14.5 (n = 3). (Scale bar: 500 µm.) (E) Hoxc12 and Hoxc13 transcripts become progressive confined to the dorsal ectoderm over the last phalanx (E16.5) and finally to the nail matrix in newborn (n = 2). For components of the nail organ, please see Fig. 2. (Scale bars: 200 µm.) In all sections, dorsal is up, and distal to the right.

Fig. 2.

Congenital anonychia in HoxC mutant mice. (A) Photographs showing lateral views of the hand of wild-type and HoxC newborn mice. Note the pronounced ventral flexion of digits in the mutant specimen. (Scale bar: 1 mm.) (B) Hematoxylin-eosin–stained longitudinal section of a forelimb finger, which shows the absence of the nail organ in HoxC mutant animals. (Scale bar: 200 µm.) In A and B, the framed area is magnified on the Right (n = 3). (C) Schematic representation of the nail organ showing its various components. (D) Immunohistochemistry for the detection of Krt5, Krt10, and hard keratins (AE13 antibody) in consecutive longitudinal sections of wild-type and HoxC newborn mice. For Krt5 immunostaining, a lower magnification is also shown to frame the area under study (n = 3). (Scale bar: 200 µm.) (D′) Schematic representation of the differentiation state of the nail region in HoxC mutants. In all sections, dorsal is up, and distal is to the right.

To evaluate the proximity between samples, we performed a principal-component analysis (PCA). When samples were plotted along the first two components and in two dimensions, the first dimension (PC1), accounting for 75% of the variance, strongly separated the samples according to the ectodermal and mesodermal nature of the tissue. The second dimension (PC2), accounting for 12% of the variance, separated the samples according to the developmental stage (). Hierarchical cluster analysis (HCA) of the 16 samples showed a strong clustering of samples by tissue and between consecutive stages (). Both the PCA and the HCA showed a high consistency between replicates.

We then looked at the tissue specific gene expression dynamics of all Hox family members and generated a heatmap using their normalized expression values [log2(1 + FPKM)]. While this plot expectedly highlighted the known temporal colinear activation of Hoxa and Hoxd genes in the limb mesoderm (Fig. 1), it revealed a previously overlooked distinctive expression of Hoxc genes in the limb ectoderm, with a comparable progressive temporal activation, starting with 3′-located genes and with 5′-located Hoxc12 and Hoxc13 gene expression being maximum at the last stage examined (E12.5) (Fig. 1). The visual inspection of coverage from RNA-seq across the HoxC cluster confirmed a transition from an initial higher expression of Hoxc5, Hoxc6, and Hoxc8 (E9.5 and E10.5) to a higher expression of Hoxc12 and Hoxc13 at later stages (E11.5 and E12.5) (Fig. 1). Genes with an intermediate position in the cluster such as Hoxc9, Hoxc10, and Hoxc11 were also sequentially activated, but their level of expression did not reach that of the genes located at both extremities of the cluster, at least during the period under scrutiny. Therefore, the expression dynamics of Hoxc genes in the limb bud ectoderm follows the typical temporal colinear pattern of Hox genes expression (32).

In order to validate these results and to determine the spatial distribution of Hoxc transcripts in the limb ectoderm, we used in situ hybridization (ISH) on tissue sections, which showed a clear restriction of expression to the ectodermal monolayer (Fig. 1, arrowheads). As the limb bud emerged (E9.5), the expression of 3′-located Hoxc genes covered the whole limb bud ectoderm, whereas, at later stages (E12.5), the expression of 5′-located genes appeared more restricted to the tip of the limb bud (Fig. 1). Subsequently, at E14.5, Hoxc12 and Hoxc13 transcripts were found concentrated in the digit tips, as detected by ISH both in whole-mount and tissue sections and progressively confined to the developing nail region (E16.5) (Fig. 1 ). In newborn mice, Hoxc13 expression persisted in the nail matrix, while only residual Hoxc12 transcripts remained in this region (Fig. 1).

Function of Hoxc Genes in Nail Development.

This expression of Hoxc genes in the forelimb ectoderm during these developmental stages and the concentration of mRNAs in the nail region suggested a function for these genes in nail morphogenesis. This was previously illustrated both by the misshaped nail phenotype displayed by mice lacking the function of Hoxc13 (12), and by the involvement of HOXC13 in the human ectodermal dysplasia 9 (ECTD9) (Online Mendelian Inheritance in Man [OMIM] 614931) (33) characterized by severe nail dystrophy. Consequently, we carefully examined the nail in mice lacking the entire HoxC cluster (HoxC), even though these mutant mice die at birth, presumably due to respiratory problems (14).

Visual inspection of HoxC newborn limbs revealed a marked flexion of forelimb digits with a strong suspicion of nail agenesis, based on the lack of the typical light reflection caused by the nail plate, a rich keratinized structure (Fig. 2). Interestingly, congenital ventral contractures of the digits have also been reported in patients with microdeletions involving the HOXC cluster in heterozygosis (34). The complete absence of nails was confirmed by histological analyses. In addition to the lack of nail plate, there was no evidence of the nail matrix, neither of a nail bed, and the proximal fold was practically nonexistent, merely reduced to a slight groove (Fig. 2 , scheme). Therefore, in marked contrast to their wild-type littermates, HoxC mutant newborns showed no evidence of nail morphogenesis at the tip of their digits.

To evaluate the level of epithelial differentiation in the mutant nail region, we used a range of anti-keratin (Krt) antibodies, which altogether confirmed the complete absence of the nail organ in HoxC mutant specimens. For instance, we observed that Krt5, a specific marker of the basal layer, and Krt10, a suprabasal epithelial marker, unexpectedly persisted in the epithelium of the mutant nail region, while they were normally down-regulated in control animals (Fig. 2). In addition, the expression of hard keratins specific for hair and nail, as detected by using the AE13 antibody, was not found in the mutant nail region (Fig. 2). We thus concluded that at birth, the epithelium of the digit tip of HoxC mutants displayed a level of differentiation somewhat similar to that of the interfollicular epithelium in control samples, as generally illustrated by the lack of specific nail differentiation (Fig. 2, scheme). The analysis of mutant hindlimbs uncovered a similar although slightly milder phenotype (). Again, the expression of hard keratins was not detected while Krt10 abnormally persisted in the digit tip epithelium indicating the impairment of nail differentiation, even though Krt5 was reduced in suprabasal layers. The difference between forelimb and hindlimb could possibly reflect the expression of additional Hox genes in the hindlimb ectoderm, a possibility that remains to be assayed.

This previously unnoticed congenital anonychia observed in HoxC mutant animals indicates that HoxC cluster genes not only participate to shape the nails, as shown by the Hoxc13 loss-of-function mutation (12), but are also necessary for their morphogenesis, at least at the proper time. Indeed, we cannot exclude a developmental delay in the differentiation of the nail organ, for the perinatal death of HoxC mutant mice precluded further analysis. In Hoxc13 mice, twisted and fragile nails were reported. At birth, the epithelial differentiation of the nail region, according to Krt5, was similar to normal, but the expression of Krt10 persisted in the suprabasal layers of the nail matrix and hard keratins were not detected (). Therefore, it seems that, in the absence of Hoxc13 function, other Hoxc genes can activate the nail differentiation program, although the expression of the hard keratins specific to the hair and nails appears to depend on Hoxc13.

Transcriptional Regulation of Hoxc Genes in the Ectoderm.

The apparent coordinated function of Hoxc genes in limb ectodermal derivatives echoed other well-described cases where several Hox genes located in cis are controlled by batteries of common long-acting enhancers positioned outside the respective Hox cluster (35–37). We investigated this possibility by using ATAC-seq (assay for transposase-accessible chromatin with high-throughput sequencing) to identify accessible chromatin regions in isolated ectodermal hulls of E14.5 digit tips (Fig. 3, first and second tracks). We selected this stage because of the high expression levels of both Hoxc12 and Hoxc13. We used two forelimb replicates and two hindlimb replicates with each dataset individually processed and looked for open chromatin regions within and around the HoxC cluster. We identified several open chromatin regions in the HoxC cluster itself, which correspond to the activity of Hoxc genes, expectedly absent from the control forebrain ATAC-seq dataset. In addition, two focal regions of significant ATAC-seq enrichment were clearly scored outside the HoxC cluster in the distal limb ectoderm, whereas they were not detected in the E13.5 forebrain ATAC-seq control dataset (Fig. 3, third track). Furthermore, these two loci, as well as the HoxC cluster, were enriched in H3K4me2 in E10.5 ectoderm, a chromatin mark indicative of transcriptional and enhancer activity (Fig. 3, Bottom track). The presence of other chromatin marks indicative of active enhancer regions using published limb bud datasets was not possible due to the high dilution of ectodermal cells in this material.

Fig. 3.

Identification of ectodermal enhancers. (A) Coverage around the HoxC genomic locus of ATAC-seq signals obtained with E14.5 forelimb distal ectoderm (FDE) and hindlimb distal ectoderm (HDE) in gray (each one is the mean of two replicates) with the peak calling of each replicate as well as the ATAC-seq profile of E13.5 forebrain cells (FB) in black used as a negative control and H3K4me2 ChIP-seq in blue. The two peaks highlighted in green were identified by ATAC-seq only in the distal digit ectoderm but are absent from the control profile in forebrain cells. The ATAC-seq coverage were normalized by million reads in peaks. (B) Results of the mVISTA tool (shuffle-LAGAN) overlaid with the ATAC-seq coverage track of forelimb distal ectoderm (FDE) (top track). Conservation of EC1 and EC2 is assessed for seven different species. For each track, the identity over 100 bp is shown between 50% and 100%, and the pink color indicates over 70% of sequence identity. Note that EC1 is conserved in placental mammals only (light green box), whereas EC2 is conserved in all mammals (dark green box). The poor quality of the opossum genome around the HoxC cluster and the presence of gaps in both cobra and frog genomes must be kept in mind. (C–E′) Reporter activity of EC1, EC2, and EC1+EC2 in E16.5 transgenic fetuses is shown by representative whole-mount LacZ expression patterns. Note the prominent LacZ staining in the tail and limb bud ectoderm. (Scale bars: 2 mm.) (F–H) Longitudinal sections of limb buds (dorsal up and distal to the right) showing expression restricted to the ectoderm. Magnifications of the hair follicles (HFs) in the proximal limb are also shown. (Scale bars: 150 µm.)

These two accessible regions, termed EC1 and EC2 (Enhancer of HoxC cluster 1 and 2), are located 165 and 81 kb upstream of the HoxC cluster, respectively, and are highly conserved among mammals in their DNA sequences, while absent from other vertebrate species (Fig. 3). Based on this phylogenetic conservation, the EC1 and EC2 sequences were characterized as 1.1 and 4.1 kb large, respectively. Of note, the EC1 sequence was conserved in placental mammals only, whereas EC2 was conserved throughout mammals. We set up to assess the activity of these two sequences in transgenic mice either individually or in tandem, using an enhancer LacZ reporter construct. Transient transgenic fetuses were harvested at E16.5 and the activity of the reporter LacZ gene revealed in whole-mount β-gal staining (Fig. 3 , Table 1, and see ). This time point was selected because at E16.5 the primitive nail matrix is already discernible as well as the primary and secondary HFs.

Table 1.

Summary of transgenic analyses

Transgene	Total no. of embryos analyzed*	No. (%) of transgenic embryos†	No. (%) of β-gal–stained embryos
EC1	23	7 (30.43%)	5 (71.43%)
EC2	112	28 (25%)	7 (25%)
EC1+EC2	39	14 (35.90%)	10 (71.43%)

Embryos recovered from PNI.

As determined by PCR on DNA from yolk sacs.

The EC1 enhancer displayed a very strong, specific, and reproducible activity (five out of seven transgenic specimens) in the distal limb bud ectoderm and the developing tail. At the level of the mid zeugopod or the proximal tail, a dense pattern of staining progressively changed to mark HF exclusively. The HFs were stained over broad areas of the body, although transgene activity was not scored in the whisker pad (Fig. 3 and ). The EC2 sequence elicited a similar activity, yet with a little more variability and smaller penetrance (7 out of 28 transgenic specimens). In general, the ectoderm area showing a dense staining was more restricted to the distal limb, and in most cases, staining was not detected in the tail. While variable areas of head ectoderm were also stained, the transgene did not seem to be active in whiskers, as for EC1 (Fig. 3 and ).

In order to evaluate a potential collaborative or synergistic effect of EC1 and EC2, we assayed their functional activity when introduced in tandem into the enhancer LacZ reporter construct. For this, EC1 and EC2 were fused one after the other with EC2 positioned 5′ to EC1. In this combined situation, the EC1–EC2 transgene displayed very robust and consistent reporter activity in the tail and distal limb ectoderm, as well as in HFs throughout the body surface (10 out of 14 transgenic specimens) (Fig. 3 and ). Reporter activity was also prominent in patches of head ectoderm, in the whisker pad, and eyelashes (Fig. 3, Table 1, and ). Of note, the dorsal and ventral midlines and the genital tubercle were also stained ().

Longitudinal sections of the limbs of stained embryos for each of the three transgenes confirmed the restriction of their activity to the ectodermal component and illustrated the transition from a distal stronger staining pattern, to a more proximal activity restricted to HFs (Fig. 3 ). Altogether, these results confirmed that both EC1 and EC2 control important aspects of Hoxc genes transcription in the ectoderm, both during early embryonic steps and during late phases of differentiation.

Deletions of EC1 and EC2 Enhancers in Mice.

To further validate the functional contribution of these two potential enhancer sequences, we used CRISPR–Cas9 genome editing to generate mice lacking each of them, separately (Fig. 4). Pairs of guide RNAs (gRNAs) were used to induce deletions covering either the EC1 or the EC2 sequences () and zygotes were transformed through electroporation. Stable lines were generated for both the HoxC and HoxC alleles and bred to homozygosity. Mice homozygous for each deletion were indistinguishable from their wild-type littermates in all gross aspects such as developmental growth, fertility, and viability. Likewise, no significant difference was detected in their ectodermal derivatives and their hair looked healthy and glossy, as well as the nails, which were apparently normal also in the histological analysis (Fig. 4 ). While more detailed molecular analyses revealed a slightly abnormal ectopic expression of Krt10 in some cells of the nail matrix in HoxC homozygous mutants (Fig. 4), the differentiation of the nail in HoxC homozygous mutants appeared completely normal (Fig. 4). From these results, we concluded that neither EC1 nor EC2 was necessary for the transcription of HoxC cluster genes in the ectoderm, at least in the presence of the other enhancer sequence. This raised the possibility that the two sequences carried a redundant specificity and may thus compensate for one another, as was described for a number of related cases in distinct mammalian gene regulatory landscapes (35, 38, 39).

Fig. 4.

Deletions of enhancers. (A) Schematic representation of the deletion alleles generated by CRISPR-Cas9. (B) HoxC homozygous mice were phenotypically like wild-type controls (Left) and showed normal nail development as can be observed in the hematoxylin–eosin (HE) longitudinal section (Middle) and anti-Krt10 staining (Right) (n = 3). (C) HoxC homozygous mice appeared normal as well (Left) and HE analysis did not show any alteration of the nail (Middle), although a minor ectopic expression of Krt10 was detected in the nail matrix (Right) (red arrowhead in magnified Inset) (n = 3). (Scale bars: B and C, 400 µm.) (D and E) Mice homozygous for the deletion of both enhancers (HoxC) displayed disheveled fur and hypoplastic nails. Histological comparison of HFs in HE-stained sections of dorsal skin from P14 control and HoxC homozygous mice showed some hair shafts distorted at the level of the sebaceous glands (E). Over time, however, an acceleration in the cycling rate became evident as illustrated by mutant dorsal HFs having precociously entered the next anagen at P97, while control HFs remained in telogen as in P91 (n = 2). (Scale bars: D and E, 200 µm.) At birth, the mutant nails were hypoplastic as indicated by the abnormal expression of Krt10 remaining in the proximal nail matrix, and the substantial reduction in Hoxc13 and hard keratins expression (red arrows in E), compared with wild-type control (black arrows in D) (n = 2). The nail defect was highly attenuated over the first 2 wk of postnatal development (P14) despite continued reduction in Hoxc13 mRNA level (red arrowheads).

To investigate whether EC1 and EC2 may indeed share parts or all of their spatiotemporal regulatory specificities, we deleted a 86-kb-large DNA sequence containing the two sequences to produce the HoxC allele (Fig. 4). Mice homozygous for this allele were viable and fertile. While they globally looked healthy, they exhibited mild but consistent fur and nail phenotypes (Fig. 4 ). When compared to control littermates, homozygous HoxC mice showed the persistence of a disheveled fur (a rough fur phenotype) with some of the mice developing bald patches on their backs. Histological analyses of the skin did not reveal any obvious defect in the morphogenesis of the HFs but showed some hair shafts distorted at the level of the sebaceous glands, reminiscent but less marked than those observed in Hoxc13-null and Foxn1-null mice and that associate with defective keratinization of the hair shaft (40). Over time, after the second or third hair cycles, an acceleration in the hair cycle became obvious. Therefore, when compared to wild-type littermates, homozygous HoxC mutants showed a precocious reinitiation of the next cycle with shortened telogen and premature anagen development, as shown in the analysis of postnatal day 91 (P91) and P97 littermates (Fig. 4 ). This is possibly driven by the loss of hair inducing a plucking effect similar to the situation described in Foxn1-null mice (41). The nails were unusually hypoplastic at birth, showing abnormal Krt10 expression in the proximal nail matrix as well as a substantial reduction in hard keratins expression (Fig. 4 ). Accordingly, HoxC mutant specimens displayed a substantial reduction in Hoxc13 level of mRNAs (Fig. 4 ). The nail defects became attenuated during the first 2 wk of postnatal life, and by P14, the histology of the nail was close to normal. These observations suggested that the presence of both EC1 and EC2 is required to reach a threshold in Hoxc genes mRNAs that is needed for the normal morphogenesis of epidermal organs. Alternatively, sequences other than EC1 and EC2 in the deleted 86-kb-large region may also contribute to the phenotype (see below).

To further document a possible role of Hoxc gene dosage in the development of ectodermal organs and to examine the function of EC1 and EC2 in this context, we looked at the phenotypic consequence of deleting the two EC1 and EC2 enhancers in a sensitized genetic background, where the expression of the presumptive target Hoxc genes was further reduced. We generated transheterozygous mice harboring either one of the single or combined enhancer deletions, over the full deletion of the entire HoxC cluster. Animals of the HoxC or HoxC genotypes showed no particular phenotype, indicating that the individual functional contribution of either the EC1 or the EC2 enhancers was dispensable even under conditions of sensitized background.

However, the removal of both enhancers on this same background (HoxC) led to a striking phenotype characterized by the total absence of body hair (Fig. 5). Transheterozygous HoxC mice were born at the expected Mendelian ratio and were viable and fertile. The phenotype became detectable shortly after birth due to their nude aspect (Fig. 5). In contrast to heterozygous or wild-type control mice, no fur coat emerged from the skin of HoxC specimens during the first postnatal week, reminiscent of the phenotype of Hoxc13 homozygous mutant animals (12). Interestingly, despite their lack of hairs, they displayed vibrissae at birth, in contrast to Hoxc13 mutant mice. These mice remained nude until the new anagen phase, at about 5 wk of age, when they developed hairs in a cranio-caudal wave. The hairs were rapidly lost in the next couple of days, however, and the mice remained nude until the next hair cycle when a new wave of hairs emerged and was rapidly lost again. Thereafter, the waves of hair regrowth and loss became more irregular. This iterative pattern of hair formation and loss resulted in a typical cyclic alopecia phenotype (Fig. 5).

Fig. 5.

Effects of EC1 and EC2 combined deletion in a sensitized background. (A) HoxC transheterozygous mice lack external hair during the first hair cycle and develop cyclic alopecia as can be observed in the displayed sequence (n = 3). (B) The characterization of HoxC nails revealed an underdeveloped organ at birth (top row), with only minor focal areas of specific nail differentiation marked by the down-regulation of Krt10 (arrowhead). Hard keratin expression was not scored (compare with Fig. 4) (n = 2). The phenotype progressively attenuated with nail differentiation, as revealed by the expression of hard keratins (red arrow), observed in P7 mutant specimens, associated with a reduced level of Hoxc13 mRNAs (red arrow) (n = 2). (Scale bars: 200 µm.) (C) Analysis of mutant HoxC HFs. On P7 dorsal skin, mutant mice showed a number of HFs that became distorted when entering the epidermis. This was accompanied by a down-regulation of Hoxc13 expression (red arrows) and by a marked reduction of hard keratins detected by AE13 immunohistochemistry (n = 2). In B and D, the red arrows denote mutant altered expressions, while black arrows denote normal expression in controls. (D) At E14.5, Hoxc13 expression was bellow detection level in the ectoderm of the digit tips of HoxC mutants, yet it persisted at a normal level in the tail mesenchyme (n = 2). (Scale bars: 400 µm.)

At birth, the nails of HoxC mice were severely underdeveloped with only minor focal areas of specific nail differentiation and a total absence of nail plate and hard keratin expression (Fig. 5; compare with control in Figs. 2 and 4), reminiscent of the HoxC nail phenotype (Fig. 2). Over time (P7), however, the defect became less pronounced, although the expression of Hoxc13 and hard keratins remained lower than normal. The nail and hair defects observed in HoxC mice were strikingly similar to that reported for mice lacking the Foxn1 function, a gene known to be a transcriptional target of Hoxc13 and, in turn, a transcriptional regulator of hard keratins of the hair shaft (40, 42). Accordingly, the histological analysis of HFs showed a phenotype similar to that described in Foxn1-null mice. HFs formed normally during fetal development but the hair shaft was distorted when entering the epidermis (Fig. 5), a phenotype that was considered to derive from the lack of hard keratins in Foxn1 mutants. Hard keratins were reduced substantially in the HFs of HoxC mice, correlating with a lower level of Hoxc13 mRNAs.

The CRISPR-Cas9 approach also triggered an inversion of the targeted region producing the HoxC allele. Mice homozygous for this allele (HoxC) were indistinguishable from wild-type littermates and displayed normal fur and nails indicating that the inverted rearrangement of the enhancers had no functional consequence (). In addition, mice transheterozygous for this inverted configuration and the deletion of the HoxC cluster (HoxC) did not show any abnormality ().

These results suggested that both EC1 and EC2 contribute to the normal level of transcription of Hoxc13 in the ectoderm. In contrast to both Hoxc13 (12, 13) and HoxC (14) mutant mice, the hereby described HoxC mice were viable and healthy, indicating that the induced defects were likely localized to the reported ectodermal components. We further tested this hypothesis by looking at the expression of Hoxc13 at E14.5, both in the limb ectoderm and in the growing tail mesenchyme, another site where Hoxc13 is normally transcribed. While no detectable Hoxc13 transcripts were scored in the limb ectoderm by whole-mount ISH, expression of this gene in the tail mesenchyme remained as in controls (Fig. 5), supporting the definition of EC1 and EC2 as enhancers carrying a predominant ectodermal specificity.

To try to better quantify the functional impact of the EC1 and EC2 enhancers on the expression level of Hoxc13, we performed RT-qPCR using different mutant alleles (Fig. 6). We dissected out the distal aspects of E15.5 digits to isolate material corresponding to the region with the highest level of Hoxc13 expression (Fig. 1). HoxC homozygous showed a significant decrease of 40% in the steady-state level of Hoxc13 mRNAs. A stronger decrease was observed for the HoxC homozygous samples, as only ∼30% of the total amount of Hoxc13 mRNAs were maintained, while HoxC homozygous samples displayed only 5% of the total amount of Hoxc13 mRNAs. Other genetic combinations, such as the HoxC that displayed only 2% of Hoxc13, confirmed this range of effects (Fig. 6). Because neither HoxC nor HoxC homozygous embryos showed any clear phenotypic alteration, we conclude that the presence of 30% of the steady-state level of Hoxc13 mRNAs is enough to ensure normal development of these ectodermal organs. Our results also reveal that a reduction in Hoxc13 expression level from 5 to 2% of the wild-type level, determines the change from the disheveled phenotype of HoxC to the cyclic alopecia of HoxC mutants. These results confirmed the functional importance of both enhancers, although with a more important contribution of EC2. Also, the respective effects of the two single enhancer deletions suggested that these two sequences may account for the global effect observed when deleting the large EC1-EC2 DNA fragment. Accordingly, we consider it unlikely that other DNA sequences within this fragment have a strong involvement in regulating Hoxc genes during ectoderm development.

Fig. 6.

Relative steady-state levels of Hoxc13 mRNA in the various mutant alleles or combinations thereof. Comparative RT-qPCR analyses of Hoxc13 mRNA levels in forelimb digit tips of E15.5 embryos of HoxC, HoxC, HoxC, HoxC, HoxC, and HoxC, as indicated in the x axis. The y axis represents the level of expression of Hoxc13 relative to control set to 1. The dissection in HoxC samples included a larger digit segment than in the rest of genotypes, and therefore they were normalized by using similar dissections of control samples (gray dots). The addition of the negative effects observed independently in the single EC1 and EC2 deletion alleles approximately corresponds to the effect obtained by deleting the two enhancers at the same time. *P ≤ 0.05; ***P ≤ 0.001; ****P ≤ 0.0001.

Discussion

We report here that the HoxC cluster has evolved a general functional specificity for some ectodermal organs, the nails and hairs, and that, at an earlier stage, Hoxc genes are activated in the embryonic limb ectoderm following a colinear time sequence. We also show that part of this subsequent specificity is achieved through the presence of two enhancers positioned at a distance from the gene cluster itself and acting together to raise the transcription level of a subgroup of Hoxc genes.

Colinear and Noncolinear Expression of HoxC Cluster Genes in the Ectoderm.

During the formation and elongation of the major body axis, Hox genes belonging to all four gene clusters are activated in a time sequence that reflects their topological order along their respective clusters (32). While the impact of this temporal process upon the colinear distribution of Hox transcripts in space is still a matter of discussion (see, e.g., refs. 43–46), the mechanism underlying this phenomenon is accompanied by a progressive transition from a negative to a transcriptionally permissive chromatin structure (47). This phenomenon is however observed only during the first wave of activation of a given Hox cluster. Indeed, once the cluster open, global enhancers located in the surrounding regulatory landscapes can subsequently regulate subgroups of genes located in cis simultaneously, as was shown for example for the transcription of Hoxd genes in the developing digits or external genitals (48).

Here, we show a clear collinear time sequence in the activation of Hoxc genes in the limb ectodermal component. This temporal sequence in mRNAs production does not correspond to any distinct ectodermal structure or compartment, which could have required the proper deployment of various HOXC proteins in time and space, for example to identify different cell types produced through a developmental iteration. We interpret this as an indication that the HoxC cluster was initially not transcribed at all in the ectodermal precursors and hence that its chromatin structure was not transcriptionally challenged before this precise spatiotemporal situation. As a consequence, upon activation, the gene cluster had to be processed like all Hox clusters during their initial phase of activation, following a 3′ (Hoxc4 in this case) to 5′ (Hoxc13) direction. In this view, the temporal colinear activation we describe in the limb ectoderm may not reflect any particular functional constraint other than allowing eventually the transcription of the most 5′ located genes that seem to be required for the development of ectodermal organs. As such, it may illustrate an intrinsic constraint of the system (49). Alternatively, temporal colinearity in limb ectoderm may underlie the slightly different distributions of Hoxc genes, which in turn may impact the morphology of the nails.

In contrast, expression of Hoxc genes in HFs occurs throughout the body, with no particular rostral to caudal colinear distribution. For this reason, it was initially proposed to “break” spatial colinearity (12). In fact, this deviation from the spatial colinearity observed during the development of the major body axis is only detected in the ectodermal compartment, since Hoxc gene expression in the dermal papilla seems to be colinear indeed, with Hoxc4 to Hoxc10 transcribed in dorsal skin, whereas Hoxc10 to Hoxc13 mRNAs are found in the tail (19). Altogether, our results further document the functional co-option of at least Hoxc13 (see below) for a late but critical function in the hair and nail epidermal organs. Of note, the two enhancers reported here are located outside of the HoxC cluster, a global regulatory strategy often associated to Hox loci and observed in many other developmental contexts (see refs. 37 and 50–52).

Subfunctionalization of Hox Clusters.

All four Hox gene clusters in amniotes are involved in the specification of structures during the development of the major body axis, which is likely their most ancestral function. In addition to this function, conserved in most animals with a bilateral symmetry, clusters were subsequently co-opted for specific functions, after the rounds of genomic duplications (53, 54). For example, both the HoxD and HoxA clusters are necessary to properly develop limbs and external genitalia. Such cluster-specific functions may be redundant between gene clusters (as is the case in limbs and genitals), and hence the full assessment of such global specificities has been difficult to evaluate due to the experimental problems to produce multiple full Hox cluster deletions (8).

In the case of HoxC, however, its deletion alone induced a drastic phenotype associated with the ectodermal compartment, thus suggesting that no other Hox cluster is equally functional in these important organs, which could have compensated for the absence of Hoxc genes function. This is reenforced by the fact that only Hoxc genes were found clearly expressed in the developing limb ectoderm (Fig. 1). This observation suggests that the functional co-option of the HoxC cluster into ectodermal organs occurred after the full set of genome duplications, potentially to take over the control of structural components of ectodermal tissue such as the hard keratins.

Without an extensive genetic analysis, it is difficult to determine how many Hoxc genes, and which ones, were involved in these two functional recruitments. Also, the situation may be slightly different for the nails and the HFs, indicating either distinct functional requirements for these two organs, or/and different modalities in the recruitment and implementation of enhancers. In the case of HFs, indeed, Hoxc13 function seems to be sufficient as Hoxc13-null mice are nude (12, 13) and the analysis of these mutant mice revealed the function of this gene in the production of hard keratins and keratin-associated-proteins that are direct targets of Hoxc13 (40, 55). In the absence of hard keratins, the hair shafts are brittle and readily break as they emerge through the skin surface, thus producing the nude phenotype. Likewise, the nails are fragile and twisted (12, 40). The Foxn1 gene, whose null mutation generates a nude phenotype, was also described as a Hoxc13 regulatory target during hair and nail differentiation (40, 42, 56).

In the case of the nail, the functional contribution of Hoxc genes seems to be slightly different from in the hair. Indeed, while nails of Hoxc13 mutant mice are fragile (12, 13), we show here that their abnormal phenotype is more severe in the complete absence of the HoxC cluster. HoxC mutant animals, which die as newborns due to defective lung development (14), show a congenital anonychia with no trace of nail-specific keratinocyte differentiation, a phenotype that thus far remained unnoticed. Therefore, other Hoxc genes, possibly Hoxc12, whose expression is similar to that of Hoxc13 (18), could either have a function during nail development, or partially substitute for Hoxc13 function when the latter is inactivated.

A Mammalian-Specific Up-Regulation?

The critical importance of Hoxc13 in mammalian hair and nail is also emphasized by the strikingly similar phenotypes resulting from loss-of-function conditions in different species including rabbits and pigs (33, 57, 58). In these cases, hair and nail differentiation is impaired in a way comparable to the murine situation. In addition, there are clear similarities with the human ECTD9 (OMIM 614931), also caused by the loss of function of HOXC13. ECTDs are a group of congenital disorders characterized by abnormal development of two or more ectodermal appendages, without other obvious systemic anomalies. Hairs are most commonly affected in association with alterations of the nails, teeth (anodontia or hypodontia), or sweat glands. Among the very rare ECTDs involving hair and nail exclusively is the ECTD9, which is caused by mutations in HOXC13 (33, 59, 60).

While hairs are specific for mammals, different skin appendages exist in other vertebrates, which also express Hox genes during their development, like in feather buds (61). Terminal keratinized digit structures like nails also exist in all tetrapods, under a variety of forms, and, in the chick, expression of Hoxc13 was observed in the tips of digits along with the development of claws (). Therefore, it is likely that a sustained level of Hoxc genes is required throughout the tetrapods lineage, in conjunction with several ectodermal organs that involve high amounts of keratins, regardless of the final morphologies of these organs (hairs or feathers, nails or claws). Accordingly, Hoxc genes could be necessary to activate a “keratinization program,” which would be interpreted differently depending on the phylogenetic context. The two global enhancer sequences we report in this study may be necessary for this ectodermal specific activation of Hoxc genes, as suggested by their behavior when introduced into transgenic mice and the phenotype resulting from their deletion.

This view is nevertheless difficult to reconcile with the fact that neither of these two ectodermal enhancers seems to be conserved outside mammals. For instance, it is not found in birds, neither at the syntenic position nor elsewhere, while one would expect the same ectodermal regulation to occur for this highly conserved gene cluster. In addition, the deletion of these two sequences led to a very severe decrease in Hoxc13 mRNA levels, although with a phenotype weaker than that produced by the deletion of the gene cluster itself. One possible explanation to this paradox may rely upon the quantitative versus qualitative aspect of this regulation. The hypomorph phenotype observed in newborns when deleting the two enhancers may indeed reflect a basal level of regulation, independent from these two sequences, which could be sufficiently performant in nonmammalian species. In this view, it is conceivable that the requirement for HOXC proteins to accompany the development of ectodermal organs be slightly different in various classes of vertebrata, with a higher protein level necessary in mammals than in birds. It is thus possible that the EC1 and EC2 enhancers evolved along with the mammalian lineage to progressively adjust the overall dose of HOXC proteins such as to achieve well-adapted ectodermal organs. In this scenario, these enhancers would thus be specific for mammals but involved in a function much more largely distributed within vertebrates.

At a more mechanistic level, it is difficult to distinguish between a simple additive effect of these two enhancers and a genuine synergistic effect. By using a single-cell ATAC-seq dataset recently generated from E11.5 mouse forelimb buds (62), we observed that most of the cells having these enhancers open were indeed ectodermal cells () and that all those cells but one showing ATAC-seq signals on both enhancers were ectodermal cells. These results thus confirm the conclusion that both EC1 and EC2 are specific ectodermal enhancers and indicate that the two sequences can function in the same cell. Further work will be necessary to investigate the nature of this collaborative regulation.

Materials and Methods

See extended methods provided in .

Mouse Strains and Animal Ethics.

We have used the four HoxC, HoxC, HoxC, and HoxC mutant lines generated by CRISPR/Cas9 and the previously published Hoxc13 (63) and HoxC cluster deletion (14) strains. All animal procedures were conducted according to the European Union regulations and the 3R principles and were performed in agreement with the Swiss law on animal protection, under license no. GE 81/14 (to D.D.) and reviewed and approved by the Bioethics Committee of the University of Cantabria. All analyses were carried out on at least two independent specimens, and in many cases, three or more samples were analyzed.

Histological Analysis, Immunohistochemistry, and ISH.

Dissected tissues were fixed in 4% paraformaldehyde and processed for paraffin embedding and microtomy. Hematoxylin–eosin staining, immunohistochemistry, and ISH were performed following standard procedures. The primary antibodies used were as follows: anti-Keratin 5 (Sigma; SAB4501651), anti-Keratin 10 (Biolegend; 905401), and anti-Pan-Cytokeratin (AE13; Santa Cruz; sc-57012).

RNA-Seq.

For each of the 16 samples, RNA libraries were prepared using the Truseq Stranded mRNA Library Prep Library Preparation kit (Illumina; 20020594) and 100-bp single reads were generated.

ATAC-Seq and Mouse Transgenic Enhancer Assays.

We performed ATAC-seq (64) on isolated ectodermal cells from the distal tips of wild-type E14.5 embryos. Two forelimb and two hindlimb biological replicates were generated, and each dataset was processed individually. ATAC-seq on E13.5 forebrain was used as control. The analysis of the conservation of the EC1 and EC2 enhancers was performed with mVISTA shuffle-LAGAN (65). The activity of EC1, EC2, and EC1+EC2 was assayed in mouse transgenesis by using a vector carrying a β-globin minimal promoter and the LacZ coding sequence (pSK- LacZ). For the analysis of both enhancers, the EC1 and EC2 DNA fragments were cloned together, separated by a stretch of 41 random nucleotides, with EC2 positioned 5′ to EC1.

CRISPR-Cas9 Modifications.

The gRNAs used to generate the delEC1, delEC2, del(EC1-EC2), and inv(EC1-EC2) mutant lines were designed with CHOPCHOP (https://chopchop.cbu.uib.no/). The gRNA sequences used in the CRISPR experiments are listed in . The CRISPR-Cas9 methodology used to generate the delEC1 and delEC2 mouse alleles was adapted from Qin et al. (66) and the del(EC1-EC2) and inv(EC1-EC2) mouse alleles were generated with the Alt-R CRISPR-Cas9 System from IDT (https://eu.idtdna.com/pages/products/crispr-genome-editing/alt-r-crispr-cas9-system). The mutant lines were generated by electroporation.

B6CBAF1/J Fertilized eggs were collected from the oviducts of E0.5 pregnant females. The collected eggs cultured in WM medium were washed with Opti-MEM (Gibco; 31985-047) three times to remove the serum-containing medium. The eggs were then lined up in the electrode gap filled with the electroporation solution, electroporated, and transferred into pseudopregnant foster mice (66, 67).

RT-qPCR.

Forelimb digits 2 to 5 (either the whole digit or the distal third phalanx region) from wild-type, HoxC, HoxC, HoxC, HoxC, and HoxC embryos were dissected out in cold RNase-free PBS. Total RNA was extracted with RNeasy Plus Micro Kit (Qiagen) and 500 ng of total RNA was reverse transcribed to produce first-strand cDNA with iScriptTM cDNA Synthesis kit (Bio-Rad) using standard conditions. RT-qPCR was carried out on an Applied Biosystems StepOnePlus using NZYSpeedy qPCR Green Master Mix, ROX plus (NZYTech). The primers used to amplify Hoxc13 were as follows: forward (Fwd), AGC​ACT​GGG​CTC​TTT​CCA​AT (68), and reverse (Rev), CGG​GCT​GTA​GAG​GAA​CCA​CGT. PCR efficiencies were measured using serial dilutions of cDNA. Relative transcript levels were normalized to GAPDH (Fwd, TGC​AGT​GGC​AAA​GTG​GAG​AT; Rev, ACT​GTG​CCG​TTG​AAT​TTG​CC). Between 5 and 10 biological replicates were analyzed for each genotype, with 3 technical replicates for each sample. The expression levels of mutant samples were calculated relative to wild-type controls (average set to 1). Significance of differences were determined using the two-tailed, Welsh t test.