ABSTRACT
Background During tetrapod limb development, the HOXA13 and HOXD13 transcription factors are critical for the emergence and organization of the autopod, the most distal aspect where digits will develop. Since previous work had suggested that the Dbx2 gene is a target of these factors, we set up to analyze in detail this potential regulatory interaction.
Results We show that HOX13 proteins bind to eutherian-specific sequences at the vicinity of the Dbx2 locus that have enhancer activity in developing digits. However, the functional inactivation of the DBX2 protein did not elicit any particular phenotype related to Hox genes inactivation in digits, suggesting either redundant or compensatory mechanisms. We report that the neighboring Nell2 and Ano6 genes are also expressed in distal limb buds and are, in part, controlled by the same Dbx2 enhancers despite being localized into two different topologically associating domains (TADs) flanking the Dbx2 locus.
Conclusions We conclude that Hoxa13 and Hoxd genes cooperatively activate Dbx2 expression in developing digits through binding to eutherian specific regulators elements in the Dbx2 neighborhood. Furthermore, these enhancers can overcome TAD boundaries in either direction to co-regulate a set of genes located in distinct chromatin domains.
Bullet points
Hoxa13 and Hoxd genes cooperatively regulate Dbx2 expression in developing digits via eutherian specific enhancers.
Dbx2 is expressed in different digit joint precursors but its function there is not essential.
Dbx2 enhancers also control the expression of the Nell2 and Ano6 genes, which are located in different TADs, thus overcoming the boundary effect.
Dbx2 chromatin architecture and enhancers evolved in the mammalian lineage.
Grant Sponsor and Number Swiss National Research Foundation #310030B_138662.
European Research Council grants RegulHox #588029
INTRODUCTION
For many decades, the vertebrate limb has been an efficient experimental paradigm to study the basic principles and concepts underlying developmental processes. The main reason is the congruence between the definition of specific signaling regions in the developing limb buds, on the one hand, and their association with specific molecular markers, on the other hand. Classical experimental embryology indeed led to a fairly precise cellular definition of those regions in the limb bud, which have a particular activity and function, such as the limb apical ectodermal ridge and the zone of polarizing activity. John Fallon made seminal contributions in this early phase and was one of the pioneers of this field1–3; see also references in4,5, as well as6–9. Subsequently, transcription factors were cloned, which could be superimposed to such cellular landmarks, such as Hox genes 10 (see 11), followed by the key signaling molecules12–14. Soon after, gain- and loss-of-function experiments helped ascertain the central roles of these genes in controlling limb patterning and morphogenesis. In this view, the developing limb was the first vertebrate system where a bridge was established between cellular models and their molecular components.
Amongst these key factors are the Hox genes belonging to both the HoxA and HoxD clusters. They are transcribed into precisely delimited domains within the incipient limb buds 10,15 and they specify the proximal and distal limb segments as well as some anterior to posterior features 16–18; reviewed in 19. Hoxa13 and Hoxd13 are essentially required for the specification and development of the autopod, the distal-most limb domain that will give rise to the digits and part of the wrist. They control the size, shape, and number of autopod bones by regulating mesenchymal cell aggregation, chondrification and ossification17,20–22. In fact, the inactivation of both Hoxa13 and Hoxd13 in mice leads to an agenesis of the distal limb and the formation of a chondrogenic blastema at the distal extremity of the ulna/fibula and radius/ humerus 17,18. Different studies have addressed the identification of the HOXA13/HOXD13 downstream target genes in distal limb development. These included genes controlling cell adhesion, morphology, and proliferation/ survival (eg. Hand2, Shh, EphA7, EphA3, Bmp2/Bmp7; 23–29). The regulatory relationships and functional roles of many such target genes nevertheless remain poorly understood.
We previously reported that the transcription factor Dbx2 (Developing Brain Homeobox protein 2) is strongly downregulated in distal forelimb cells lacking Hox13 function 23. Dbx2 belongs to the Dbx subfamily of homeobox-containing proteins and is expressed in the mouse embryonic brain and neural tube, as well as during limb development 23,30,31. However, while it is well established that Dbx2 contributes to the specification of the V0 spinal cord interneurons 32–35, its potential role in limb development has remained elusive. However, a heterozygote deletion spanning the genomic region comprising the human loci NELL2, DBX2 and ANO6 was associated with intellectual retardation, skeletal and dental anomalies, reduced hand and feet size and clinodactyly of the fifth digit, suggesting that Dbx2 could be involved in digit development 36, where it may mediate part of the functions of HOX13 proteins.
In this study, we characterized the regulation of the mouse Dbx2 gene in developing digits. We show that the HOX13 factors directly regulate Dbx2 expression in digits, in part by binding to eutherian-specific regulatory elements located within 30Kb 5’ to the Dbx2 locus, as well as within its introns. Furthermore, we show that 5’ Hoxd genes also contribute to Dbx2 regulation by acting cooperatively and redundantly with HOX13 proteins. However, Dbx2 null mice do not display any of the major limb skeletal abnormalities displayed by any combinations of HOX mutations, suggesting either that Dbx2 is not a major downstream Hox effector or that its function is compensated for in this particular situation.
We also observed that the Dbx2 neighboring genes Nell2 (Neural EGFL Like 2) and Ano6 (Anoctamin 6) are expressed in the distal limb as well. Analysis of chromatin interaction profiles revealed that at the 3D level, the Nell2 and Ano6 genes are organized into distinct Topologically Associating Domains (TADs), which are regions of the genome where gene-enhancer interactions are favored 37. Interestingly, the boundary between these two TADs maps at the proximity of the Dbx2 locus and of its limb enhancers, which seem to be able to control the transcription of the three genes, regardless of in which TAD they reside.
RESULTS
Dbx2 expression during distal limb development
We characterized Dbx2 expression at different stages of mouse forelimb development by whole-mount in situ hybridization (WISH) and quantitative PCR (qPCR) and compared it with that of Hoxa13 and Hoxd13 (Fig. 1). Its transcripts were first scored during early limb development (E9.5-E10) throughout the limb bud mesenchyme with the exception of mesodermal cells underlying the distal-most limb ectoderm (Fig. 1A). This expression poorly correlates with that of Hox13 genes and, overall, the Dbx2 mRNA levels at this stage remained very low, as confirmed by qPCR analysis (Fig. 1B, C). By E11.5, Dbx2 expression in the proximal limb was confined to a small and posterior moon-shaped domain of cells (Fig. 1A, asterisk). Dbx2 transcripts were also detected in the anterior portion of the autopod (Fig. 1A, arrowhead). Thus, the distal limb expression of Dbx2 was delayed by approximately 24h when compared to the onset of Hoxa13 and Hoxd13 in the autopod (Fig. 1A-C; 10,38). At E12.5, Dbx2 mRNA spread to the entire distal-most portion of the autopod, both in digit and interdigit mesoderm and in a nested domain within the Hoxa13/Hoxd13 expressing cells (Fig. 1A, B). However, Dbx2 transcripts were rapidly downregulated in the interdigit region and, from E13 onwards, they were detected in sequentially formed domains reminiscent of the forming digit joints.
To assess which cell type(s) express the Dbx2 gene, we used known markers of tendon and cartilage precursors expressed in digit joints. We also re-analyzed single cell-RNA sequencing (scRNA-seq) datasets from E11, E13 and E15 mouse hindlimbs 39 (Fig. 1D, E; Fig. S1). This analysis revealed that Dbx2 is expressed in different cell populations within the developing limb. Some positive cells did not express Hoxa13/ Hoxd13 at the stages analyzed and displayed the Col2a1 marker of mature cartilage precursors (Fig. 1D; Fig. S1A). Dbx2 transcripts were also detected in a subpopulation of Hoxa13 and Hoxd13 positive cells, which expressed the Gdf5, Mkx, Scx and Col1a1 genes as well (Fig. 1D, E; Fig. S1A). Gdf5 is transcribed in different cartilage and tendon-ligament precursors of the joint interzone, whereas Mkx, Scx and Co1a1 mark tendon cell precursors 39–43. Of note, although Dbx2/Col2a1/Sox9+ cartilage cells did not express Hox13 genes at E13/E15, they derive from a common population of precursors expressing Hox13 genes (Fig. S1B)39.
These results showed that Dbx2 is expressed during digit development in Hoxa13/ Hoxd13 expressing cells corresponding to tendon and cartilage precursors of developing digit joints. Therefore, it supports the possibility that HOX13 proteins could act as direct regulators of Dbx2 expression in these cells, in agreement with the reported function of 5’Hoxd and Hoxa13 genes in digit joint development 40,44,45.
Identification of HOX13-bound sequences regulating Dbx2 expression in digits
Dbx2 expression in distal limbs is strongly compromised in the absence of HOX13 proteins 23. To further evaluate whether HOX13 paralogs could act as direct regulators of Dbx2 expression, we set up to characterize the extent of the Dbx2 regulatory landscape both by analyzing available Hi-C datasets and by performing 4C-seq experiments (Fig. 2). TADs are megabase-scale structures that constitute a unit of 3D genome organization 37,46. Thus, they often coincide with and delimit the extent of gene regulatory landscapes 47,48. TADs are generally independent from the transcriptional status of the gene(s) inside and can be identified across different cell types or tissues.
We analyzed high-resolution (5Kb bin size) Hi-C data from ES cells and embryonic cortex 49, as well as 40Kb-resolution Hi-C profiles from E12 mouse distal limbs 50 (Fig. 2A; Fig. S2). We observed that, in all cases, the Dbx2 genomic region is organized in two large TADs, which span the neighboring loci Tmem117 and Nell2 (5’TAD) and Ano6, Arid2 and Scaf11 (3’TAD), respectively. Although with some variation between tissues or cell types and depending on the TAD-separation score calculation parameters, the border between these two TADs consistently falls in the close vicinity of the Dbx2 gene. Accordingly, a region of approximately 150Kb spanning the Dbx2 locus forms a micro-domain of higher contact frequency spanning the TAD boundary (hereafter referred to as interTAD domain) and the Dbx2 interactions were mostly restricted to this interTAD domain.
To confirm this, we performed 4C-seq experiments in mouse proximal and distal forelimb cells at E12 using the Dbx2 promoter as a viewpoint (Fig. 2B). As expected, the vast majority of Dbx2 interactions were observed within the 150Kb region, matching the interTAD domain identified in the Hi-C data analysis. However, some diffuse Dbx2 interactions were also detected over the entire lengths of the 5’ and 3’ TADs flanking the Dbx2 locus, while Dbx2 contacts dramatically dropped down outside of these domains. Overall, the Dbx2 interaction profiles remained very similar in both proximal and distal forelimbs (PFL and DFL, respectively). Nonetheless, we scored a DFL-specific increase in contacts over a narrow region located 55Kb away from the Dbx2 transcription start site (TSS) on the 5’ side of the locus, as well as with a broad region comprised between 82Kb and 236Kb 3’ to the Dbx2 transcription start site located and encompassing part of the neighboring Ano6 gene (Fig. 2B, asterisk and Fig S3A). These results suggested that Dbx2 expression in the developing limbs is mostly driven by mid and short-range regulatory interactions within its immediate 150Kb surroundings.
To identify putative regulatory sequences controlling Dbx2 expression in developing digits, we analyzed H3K27 acetylation datasets 23, a histone modification specifically enriched in active enhancers and promoters 51. In mouse E12 PFL and DFL cells, within the 800Kb spanned by Dbx2 and its flanking TADs, we identified only five non-coding regions specifically enriched in H3K27ac (Fig. 2C and Fig. S3B). These were located within the Dbx2-interTAD domain, suggesting that they could correspond to Dbx2 regulatory elements. Of these, two were located in the intergenic region on the 5’ side of the Dbx2 locus, two others mapped within Dbx2 intronic sequences and one overlapped with the first Dbx2 exon and TSS. All these sequences were strongly contacted by the Dbx2 promoter (Fig. 2B). Other H3K27ac-positive regions were identified within the Ano6/Arid2/Scaf11 TAD, yet they were not specifically enriched in this epigenetic mark in DFL cells, arguing against a specific involvement of these sequences in Dbx2 regulation.
We also analyzed HOXA13 and HOXD13 ChIP seq datasets 28 to determine whether these proteins would directly interact with the Dbx2 locus (Fig. 2C and Fig. S3). We observed HOXA13/HOXD13 binding at several locations within the Dbx2 genomic region. While most of these HOX13 bound sequences were not located in H3K27ac-positive and Dbx2 interacting regions, we nevertheless observed strong binding of these proteins in three of the DFL-specific H3K27ac-positive regions showing an interaction with Dbx2. One of these HOX13-bound sequences partially overlapped with a Vista enhancer (mm1571) previously characterized to drive LacZ reporter expression in the neural tube and developing limbs 52,53 (arrowhead in Fig. 3B). We quoted the other sequences as putative Distal Limb Enhancers (DLE) and numbered them based on their 5’to 3’ position within the Dbx2 interacting domain (DLE1 to DLE3). These sequences are conserved across the different mammalian species analyzed, with the exception of DLE1, which is absent from the Dbx2 genomic region in ungulate species, suggesting a specific loss of this element in this taxon (Fig. 3A). Instead, the Vista mm1571 enhancer was conserved in all tetrapod species analyzed. Furthermore, we could identify evolutionarily conserved HOX binding sites within the DLE1 and DLE2 sequences (Fig. 3A).
To assess the functional role of these DLEs, we cloned the DLE1 and DLE2 sequences in a LacZ reporter vector and tested them in transient transgenic experiments. The two elements displayed activity in E13 DFLs in a domain reminiscent of Dbx2 expression in the last forming joint of the phalanges (Fig. 3B). Interestingly, DLE1 and DLE2 displayed mirror-imaged stainings, with DLE1 active in the posterior portion of the handplate and DLE2 anteriorly. Besides, DLE1 displayed weak yet reproducible activity in a narrow stripe of cells within the mesopod (Fig. 3B, asterisk), possibly related to the initial expression of Dbx2 at E11.5 (Fig. 1A, asterisk). Likely, this was maintained until E13 due to the stability of the beta-galactosidase protein. Neither DLE1 nor DLE2 displayed transgene activity in any embryonic structure other than the developing digits. To corroborate the functional role of the identified elements on Dbx2 regulation, we used CRISPR/Cas9 genome editing to produce mice lacking the DLE1 regulatory element (DLE1-/-) and analyzed Dbx2 expression. As expected, DLE1-/- mice displayed a significant decrease in Dbx2 expression in the E13 developing digits, as compared to control littermates (Fig. 3C,D). In agreement with the DLE1 transgenesis results, this effect was even more pronounced in the posterior digits, where the DLE1 transgene displayed LacZ activity.
Hoxa13 and 5’ Hoxd genes directly regulate Dbx2 expression
To assess the relative contribution of Hox13 paralogs to Dbx2 regulation, we measured its expression in the forelimb autopods of either Hoxa13-/- or Hoxd13-/- mice, and of compound mutants carrying different combinations of Hoxa13 and Hoxd13 null alleles (Fig. 4A), by using both WISH and qPCR. As previously described, Dbx2 was almost completely abrogated in double Hox13 mutant mice (Fig. 4B, C). Instead, only a weak reduction in Dbx2 expression was observed in Hoxd13-/- single mutants. There, transcripts were maintained in the distal forelimb, with the exception of the posterior-most portion of the autopod, where Dbx2 expression was sharply reduced (Fig. 4B, C). In contrast, Dbx2 mRNA levels strongly decreased in Hoxa13-/- embryos and transcripts remained detectable at low levels only in the distal portion of the central digits. Dbx2 expression was not detected in either the Hoxa13-/-Hoxd13+/- or Hoxa13+/-Hoxd13-/- compound mutants (Fig. 4B), indicating that a single allele of either genes was not sufficient to activate Dbx2 in the DFL, despite the fact that in these mutants, a reduced but correctly specified autopod is still observed 17,23. Of note, a very faint and spatially ill-defined Dbx2 signal was scored in the Hox13 double knock-out mice (Fig. 4B), reminiscent of the early expression of Dbx2 in the incipient limb bud at E9.5 to E10 (Fig. 1A). This expression was not observed in either Hoxa13-/-Hoxd13+/- or Hoxa13+/-Hoxd13-/- mutant embryos. This may reflect the inability of Hox13 mutant limbs to properly terminate the early limb developmental program and to initiate the transcriptional network operating at later stages in the autopodial domain 23,28.
Because Hoxd genes exert largely overlapping functions in the development of the distal limb domain 54, we also assessed whether other Hoxd paralogs could contribute to Dbx2 regulation. We thus analyzed Dbx2 expression in series of mutant mice carrying deletions of different combinations of Hoxd genes (Fig. 4A, D). HoxDDel(Hoxd9-Hoxd12) knock-out mice, hereafter referred to as Del(9-12), carry a deletion including all Hoxd genes normally expressed in the autopod but Hoxd13. They displayed normal levels of Dbx2 expression as compared either to control or to Hoxd13-/- mutant autopods. Instead, mice carrying a homozygote deletion including from Hoxd8 to Hoxd13 (HoxDDel(Hoxd8-Hoxd13)-/-; hereafter Del(8-13)) displayed a drastic downregulation of Dbx2 mRNA levels, which was significantly stronger than that observed in Hoxd13-/- mutant and comparable to that of Hoxa13-/- mice. This reduction was also observed in mice carrying a large genomic deletion removing the HoxD centromeric gene desert, which contains all the elements controlling Hoxd gene expression in the autopod 55. Altogether, these data indicate that although Hoxd13 is the main Hoxd gene regulating Dbx2 expression in digits, other Hoxd genes cooperatively contribute to this activation along with Hoxa13 (Fig. 4E).
Dbx2 does not significantly contribute to the distal limb skeleton development
To determine the extent to which Dbx2 contributes to HOX13 functions in distal limbs, and also to assess its importance in the hand/foot phenotype associated with the deletion of the human NELL2/DBX2/ANO6 genomic region 36, we used CRISPR/Cas9 genome editing to disrupt the Dbx2 homeodomain (Fig. 5A). We designed specific sgRNAs targeting the flanking region of the Dbx2 third exon, which encodes two out of the three alpha-helices (H1-H2) of the homeodomain and part of the third (H3). We thus produced mice carrying a 377 bp large deletion, which removes the H1-2 coding sequence and produces a frameshift mutation, thus disrupting also H3 and the DBX2 C-terminal domain. This mutation is expected to prevent the binding of the protein to DNA and hence inactivate its function (Fig. 5A, B). The frequency of mouse pups carrying this Dbx2 mutant allele, either heterozygous or homozygous, was significantly reduced when compared to the expected Mendelian ratio (Fig. 5C), suggesting that the Dbx2 mutation led to embryonic or perinatal lethality. However, no clear skeletal or hand/ foot phenotype was observed in the Dbx2-/- mice, neither in the length or number of phalanges, nor in their degree of ossification or in their phalangeal joints (Fig. 5D, F). Therefore, although Dbx2 operates downstream of HOX13 genes in distal limb development, it is not the main contributor to the effects observed in these structures upon the loss of Hox13 and other Hoxd genes 17,40,45.
Nell2/Dbx2/Ano6 coregulation in developing limbs
The absence of limb alterations in Dbx2 null mice raised the question of whether the neighboring Nell2 and Ano6 genes may contribute to limb development. In fact, the entire Dbx2 genomic region has a syntenic interval in humans and other tetrapods (Fig. 6A) and the deletion involved in hand-foot defects in humans also contains the NELL2 and ANO6 genes 36. WISH analysis as well as mining a scRNA-seq dataset 39 revealed that Nell2 and Ano6 are specifically expressed in the distal portion of mouse developing limbs, in a population of Hoxa13/Hoxd13 double-positive cells, part of which also express Dbx2 (Fig. 6B, C). In both cases, their transcripts were distributed on both sides of the developing digits, displaying an indentation (Nell2) or a faint band (Ano6) corresponding to the joints of the forming phalanges (Fig. 6B and Fig. S3B). However, we could not identify any DFL-specific H3K27ac positive region in the Nell2 or Ano6 TADs (Fig. 2B; Fig. S3A, B), suggesting that their expression in developing limbs could be driven by the regulatory elements of the Dbx2-containing interTAD region.
To address this question, we performed 4C-seq experiment in E12 DFLs using either the Nell2 or the Ano6 promoters, as well as DLE1 and DLE2 as viewpoints (Fig. 6D; Fig. S3D, E). As expected, the Nell2 and Ano6 promoters displayed strong interactions with sequences located in their own TADs, while they showed reduced contacts with the neighboring TAD. However, in both cases, we observed significant contacts of both the Nell2 and Ano6 promoters with the DLE1-3 region (Fig. 6D and Fig. S3C). In the reverse experiment, DLE1 interacted not only with the Dbx2 promoter, but also with the close neighborhood of the Nell2 and Ano6 TSSs. Such interactions were also scored when using DLE2 as a viewpoint, though with reduced in intensity, likely due to the fact that the DLE2 contacts remained overall strongly confined to the Dbx2 interTAD domain. Nonetheless, DL2 contacts with Nell2/Ano6 were still higher than those displayed by the Dbx2 promoter, which contacted predominantly the interTAD region (Figs. 2B, 6D and Fig. S3C). We did not observe any significant enrichment of H3K27me3, a histone modification usually associated with inactive enhancers and promoters 56, over the DLE1 to 3 regions, ruling out the possibility that the interactions would represent contacts between H3K27me3 islands, as reported in other instances 57.
To further document that part of the transcription of both Nell2 and Ano6 could be driven by elements shared with the Dbx2, we analyzed their expression in mice lacking the DLE1 sequence by WISH and qPCR. We observed that Nell2 and Ano6 transcript levels were significantly decreased in the autopods of DLE1-/- embryos when compared to control littermates (Fig. 6E, F). This decrease was more pronounced for Nell2 than for Ano6, yet it remained proportionally lower than that observed for Dbx2 (Fig. 3C, D), in agreement with the differences observed in contact frequency between DLE1-2 and the promoters of these three genes. These results strengthened the hypothesis that the regulatory elements located in the genomic vicinities of Dbx2 also control part of the Nell2 and Ano6 expression in developing digits.
Structural differences at the Nell2/Dbx2/Ano6 locus between birds and eutherian mammals
While the DLE1-3 regulatory elements are broadly conserved across eutherians, they could neither be identified in non-eutherian mammals, nor in any other vertebrate (Fig. 3A). Instead, a large syntenic region around the Dbx2 locus is conserved across all tetrapods, except in monotremes, where the Ano6 gene was specifically lost (Fig. 6A). Hi-C interaction profiles produced from embryonic chicken limbs 58 revealed that the TAD organization of the chicken Dbx2 region is similar to that of the mouse (Fig. 2A, 7A and Fig. S2), with Dbx2 located in the close vicinity of the boundary between the Nell2 and Ano6 containing TADs (Fig. 7A), in agreement with the syntenic correspondence. This suggests that the Dbx2 TAD architecture is maintained across vertebrates and arised before the emergence of mammals.
Therefore, we asked whether Dbx2, Nell2, and Ano6 were expressed in the developing wings of chick embryos. We did not observe the expression of either Dbx2, Nell2 or Ano6 in the distal domain of embryonic chick limb buds (Fig. 7B, C) by WISH. Although weak expression levels of Dbx2, Ano6 and Nell2 transcripts were detected in RNAseq data, they were not specifically increased in the chicken autopod 58, in agreement with the idea that digit-specific expression of these genes was acquired after the emergence of the eutherian lineage. Instead, Dbx2 was expressed in the developing chicken neural tube (Fig. 7C), in agreement with its expression in the mouse CNS and with the presence of the evolutionary conserved, neural tube-specific, mm1571 regulatory element located within the second Dbx2 intron (Fig. 3A, B; Fig. S3). Likewise, Nell2 was expressed in the neural tube and somites of both species, in agreement with the function of this gene in sensory and motor neurons differentiation 59,60. Ano6 transcripts were also detected in the paraxial and lateral mesoderm of both species. Therefore, Dbx2, Nell2, and Ano6 expression in embryonic structures others than the developing digits is common to different tetrapod lineages, yet it is associated with different populations of neural and mesodermal precursors, suggesting that their transcription in these structures likely relies on gene-specific regulatory elements (Fig. 7B, C; Fig. S3). These results suggest an evolutionary scenario whereby the acquisition of distal limb enhancers within an ancestral TAD organization led to the co-option of Nell2, Dbx2 and Ano6, in the developing mouse digits. The functional consequences of this co-option remain to be established.
DISCUSSION
HOX13 mediated activation of Dbx2 in digits and its function in distal limb development
In this study, we show that Hoxa13 and posterior Hoxd genes directly activate Dbx2 expression by binding to different regulatory elements located either within the Dbx2 introns or in the 30Kb 5’ to Dbx2. This is supported by the expression of these former genes in the autopod anlage, which precedes that of Dbx2 by approximately 24h. Because HOX13 proteins have been proposed to act as pioneer factors 24 (see also 60), their binding at the Dbx2 locus may facilitate the access to other transcription factors, thus explaining why some Dbx2 expressing cells do not express any Hox13 genes in E13 and E15 distal limbs whereas mice lacking all Hox13 functions completely loose Dbx2 expression. Besides binding to the DLE1 to DLE3 sequences, HOXA13 and HOXD13 also bind to various locations within the Nell2/Dbx2/Ano6 genomic region. Many such sequences are only partially conserved across the eutherian lineage, in contrast with the high conservation of the DLE regions. While the functional significance of this large HOX13 coverage has not been addressed, it is clearly reminiscent of what was described for the TAD flanking the HoxD cluster, suggesting that HOX13 proteins may global regulate the TAD activities at the Dbx2 locus 23.
Mice lacking Dbx2 function did not show any major skeletal anomaly, neither in the number of phalanges, their length or their ossification pattern, nor in their joints. Also, we did not observe any major limb alterations in the offspring of mice carrying a deletion of the DLE1 sequence, thus suggesting that Dbx2 is likely not a major mediator of HOX13 function during distal development. The observed embryonic/perinatal lethality of Dbx2-/- mice could result from defects in the specification of neuronal type in the CNS. However, Dbx2-/- mice could display as yet undetected anomalies in the development and/or function of digital tendons and/or ligaments, as suggested by the expression of this gene in precursors identified by the presence of transcripts from Scx, a known marker of tendon and ligament progenitor differentiation 42,62. Therefore, our results do not support the possibility that the loss of function of DBX2 alone leads to the hand/foot defects observed in humans carrying a heterozygote deletion of the NELL2/DBX2/ANO6 genomic region 36. However, the observation that these genes are expressed during embryonic limb development indicates that their combined loss may generate these severe limb alterations. Accordingly, Ano6 inactivation in mice was reported to affect bone formation and to result in micromelia 63. In this context, the loss of the shared DLE1 regulatory element in ungulates may illustrate the flexibility of distal limb structures and their spectacular morphological diversification. Accordingly, we cannot rule out the existence of a human-specific function of the DBX2/NELL2/ANO6 genes in hand/foot development.
Contribution of chromatin architecture and enhancer activity in Nell2, Dbx2, and Ano6 coregulation in developing digits
We used a comprehensive set of scRNA-seq, ChIP-seq and Hi-C data to identify regulatory elements controlling Dbx2 expression in digits and two such elements (DLE1 and 2) displayed enhancer activity in transgenic mice. The deletion of DLE1 leads to a strong downregulation of Dbx2 transcripts. The DLE1-3 sequences are evolutionarily conserved across eutherians, while they were not identified in other vertebrate species. Together with our observation that Dbx2 is not expressed in developing embryonic chicken extremities, this suggests that limb-specific Dbx2 expression evolved in the mammalian lineage. Also, the comparison of Hi-C interaction profiles at the Dbx2 loci between mouse and chick revealed a similar TAD organization (Fig. 8B), which likely originated early in the tetrapod lineage, although we cannot exclude that more subtle changes in TAD architecture also contributed to the evolution of Nell2/Dbx2 and Ano6 expression in digits. However, while in chick the Dbx2, Nell2 and Ano6 promoters are regulated mostly by locus-specific short or mid-range regulatory interactions, the mouse DLE can co-regulate these three genes despite their location in different TADs.
This organization is reminiscent of that observed at the HoxD cluster, yet with an inversion of functionalities. At the HoxD locus, two flanking TADs contain distinct enhancers, which act in an exclusive manner upon Hoxd genes located at the TAD boundary (e.g.64–66). At the Dbx2 locus, the regulatory elements are located at the TAD boundary and can interact with target genes located within the two adjacent TADs, thus providing a first example of such a regulatory architecture. However, the functional contribution of this organization, as well as the mechanisms whereby the DLE sequences can differentially interact with the Nell2/ Dbx2 and Ano6 promoters, remain to be determined. Recent reports have used chromosome engineering to analyze the insulating effect of TAD boundary regions 67–69, supporting the conclusion that TADs are domains where enhancer-promoters contacts are favored, if not constrained. Our results suggest that, in some cases, enhancers located in between TADs may be selected to interact with either TAD depending on the context. Accordingly, it was recently shown that boundary elements can play an important role in allowing the establishment of interTAD promoter-enhancer interactions in drosophila embryos 70, yet with a mechanism substantially different from the one proposed here. Another non-exclusive possibility is that Dbx2 would be expressed in developing digits as a bystander effect due to the activity of the neighboring limb enhancers71.
EXPERIMENTAL PROCEDURES
Mouse strains and transgenesis essays
Mice were kept and handled following good laboratory practices. Mutant strains were maintained in heterozygosis. The Hoxd13, Hoxa13, Del(1-13), De(8-13), and Del(9-12) mutant mouse lines (Fig. 4) were previously described 17,20,72–74.
To generate Dbx2+/- and DLE1+/- mutant lines pairs of specific sgRNA targeting both sides of the Dbx2 third exon (CTGCTGTTGAAAGTAGGACT; CCACTGTTCTGAGAGTCCGA) and the DLE1 enhancer (GAAAAGGAAGACCACCCGTG; AGGGGCTAGAGATCTCCCAG) were co-electroporated, together with the Cas9 protein (TruecutV2; Thermofisher), in fertilized mouse oocytes. To screen for each mutant allele, we designed specific primer pair flanking the Dbx2 third exon and enhancer (DLE1_F: ACACACAGATAAATGCACGTGAAGTG; DLE1_R: GGAGGGCCACTCTTAGGTGTG). In each case, we selected F0 mouse mutant carrying a deletion of 377bp (chr15: 95632232-95632608, mm10) spanning the whole Dbx2 third exon, and a 1015bp deletion (chr15:95600674-95602176, mm10) encompassing the DLE1 sequence. Mutants mice were backcrossed with Wt B6CBAF1 mice. Mutant F1 and F2 mice were selected using specific genotyping primers for their respective Wt and mutated alleles (Dbx2_F: GGAACTCCCACCTTCGACTGACTG/ ACTGTTGATTAGGGCTGGGCTTTGA; Wt alleles: 756/ mutant allele 388bp; DLE1_Wt: GGAGTGAGGTTGTGCCAAGA/ ACCTGTAAGCCAACCCCTAC; DLE1_Mut: ACACACAGATAAATGCACGTGA/ GAGGGCCACTCTTAGGGTGG).
For the transgenesis essays, the TgDLE1::LacZ and TgDLE2::LacZ plasmids were linearized with NotI and KpnI. The fragment encoding the enhancer, b-globin minimal promoter and LacZ reporter was gel purified and injected in the masculine pronucleus of fertilized oocytes. Transgene injections were performed by the transgenesis platform of the University of Geneva. F0 embryos were dissected at E12-E13 and stained for LacZ activity.
Probe, transgene and sgRNA cloning
The sgRNA targeting guides were generated by annealing complementary pairs of oligonucleotides and cloned into the pX330 vector as described in 65.
The plasmid encoding for the mouse Dbx2 RNA probe was a gift from Thomas Jessell (Addgene plasmid 16288; 33). Instead, specific primers were used to amplify a portion of the transcribed region of the mouse Nell2, Ano6, Gdf5 and Mkx genes as well as of the chicken Dbx2, Nell2 and Ano6 orthologs. In each case, the PCR products were cloned into the pGEM-Teasy plasmid and sanger sequenced. For the chicken Dbx2, Nell2 and Ano6 genes, primers were designed based on the exon/intron structure predicted from the UCSC Non-Chicken Refseq genes, spliced EST, Chicken mRNA and Ensembl gene prediction.
For the transgenesis assays, the DLE1 /DLE2 sequences were amplified with specific primers (DLE1 Fw: ATCCTGCTGTCTCTGGCTTTCAT/ GGGATCTGATGCATGTAGTGGAATTC; DLE2 Fw: TCCAAGTTCTGTCTTCTAGGGCA/ GGATTGTGTATTAACCAGGACCGA) and cloned into the pSK-bglob::LacZ reporter plasmid 23, generating the TgDLE1::Lacz and TgDLE2::LacZ reporter vectors.
Probe and sgRNA preparation
For the sgRNA transcription, we PCR amplified the sgRNA sequence cloned into the px330 plasmid using a T7 promoter containing primer and a universal reverse oligonucleotide (TAATACGACTCACTATAG). PCR products were gel purified and transcribed in vitro using the HiScribe™ T7 High Yield RNA Synthesis Kit (NEB). The transcribed sgRNAs were purified using the RNeasyTM mini kit (Qiagen).
Specific probes for the different genes analyzed were synthesized in vitro by linearizing the respective coding plasmids using specific restriction enzymes and by in vitro transcribed with either T7/T3 or Sp6 RNA polymerase. The probes were purified using the RNA easy mini kit.
Gene expression analysis
For the qPCR analysis, pairs of E10/E11/E12/E13 mouse DFLs, as well as HH30-31 chicken distal wings, were microdissected and stored in RNAlater. Toral RNA was extracted from each pair of DFLs/ distal wings using the Qiagen microRNA extraction kit and retro-transcribed using the Promega GOscript reaction mix with random primers. Gene expression levels were measured by real-time qPCR using the SYBR® Select Master Mix for CFX (Thermofisher), and specific primer pairs for the mouse Dbx2, Hoxa13, Hoxd13 genes as well as for the chicken orthologs Hoxd13 and Dbx2 (Table I). The mouse Hmbs and chicken Gapdh housekeeping genes were used as internal controls for the normalization of gene expression levels (2-(ΔCt)). WISH experiments were performed as described in 75.
Skeletal preparation
Alcian blue and alizarin staining was performed as described in 54. Briefly, P7 mouse pups cadavers were eviscerated and skin and fat tissues were removed as much as possible. After 48h fixation in ethanol, cadavers were stained alcian blue solution (150 mg/l alcian blue 8 GX in 80% ethanol and 20% acetic acid) for two days and washed in 100% ethanol overnight. Subsequently, they were cleared for at least 3h in 2% KOH solution and stained for 2h in 50 mg/l alizarin red / 2% KOH solution. Finally, they were washed in 2% and 1% KOH solution and progressively dehydrated to 100% glycerol solution.
Hi-C/ ChIP seq/ scRNA-seq data analysis 4C-seq interaction profiling
All scripts used to analyze data and generate figures are available at https://github.com/lldelisle/scriptsForBeccariEtAl2021. The calculations were performed using the facilities of the Scientific IT and Application Support Center of EPFL.
For the chicken Hi-C analysis, the raw forelimb and hindlimb data were extracted from GEO (see Table II) and processed independently using HiCUP v0.8.0 76 on galGal6. Valid pairs were obtained using a custom python script. Both valid pair files were merged before analysis. Valid pairs from Hi-C carried out on mouse material were downloaded from GEO (see Table II).
Valid pairs of each study were loaded in a cool file using cooler version 0.8.10 77 using a resolution of 5Kb, 20Kb or 40Kb. The TAD-separation score and the domains were obtained using hicFindTADs version 3.5.2 78–80 with --minBoundaryDistance 100000 and either default parameters for the choice of window size or a fixed window size of 120Kb. Plots were obtained using pyGenomeTracks version 3.678,81.
ChIP-seq paired-end (PE) fastq of HOXA13 and HOXD13 data, as well as single-read (SR) fastq from H3K27me3 and H3K27ac and corresponding inputs were downloaded from GEO (see Table II). Adapter sequences and bad quality bases were removed with Cutadapt82 version 1.16 with options -a GATCGGAAGAGCACACGTCTGAACTCCAGTCAC -A GATCGGAAGAGCGTCGTGTAGGGAAAGAGTGTAGATCTCGGTGGTCGCCGTATCATT -q 30 - m 15 (-A being used only in PE datasets). Reads were mapped with bowtie 2.3.583 with default parameters on mm10. Alignments with a mapping quality below 30 were discarded with samtools view version 1.984,85. For HOXA13 and HOXD13, coverage and peak calling were computed by macs2 version 2.1.1.20160309 with options --call-summits -f BAMPE -B. Coverage was then normalized by the number of million fragments used in macs2 coverage. For histone marks, coverage and peak calling were computed by macs2 with options -f BAM --nomodel --extsize 200 --broad using the BAM of input in -c. The coverage was then normalized by the number of million tags used in macs2 coverage. Plots were obtained using pyGenomeTracks version 3.678,81. For DFL_E12_H3K27ac, the two replicates were averaged.
For the scRNA-seq, matrices with counts were downloaded from GEO (see Table II). UMAP and expression plots were obtained using Seurat package version 3.2.286 on each dataset individually.
We performed our 4C-seq experiments according to 87. Briefly, 12 pairs of wildtype DFLs or PFLs were dissected, dissociated with collagenase (Sigma Aldrich/Fluka) and filtered through a 35 micron mesh to isolate single cells. Cells were fixed with 2% formaldehyde (in PBS/10%FBS) for 10 min at room temperature and the reaction was quenched on ice with glycine. Cells were further lysed with 10 mM Tris pH 7.5, 10 mM NaCl, 5 mM MgCl2, 0.1 mM EDTA, 1x Protease inhibitor cocktail to isolate nuclei and stored at -80°C. Nuclei from pools of at least 10 distal or proximal limbs were digested with DpnII (New England Biolabs) and ligated with T4 DNA ligase HC (Promega) in diluted conditions to promote intramolecular ligation. Samples were digested again with NlaIII (New England Biolabs) and ligated with T4 DNA ligase HC (Promega) in diluted conditions.
These templates were amplified using Expand long template (Roche) and inversed PCR primers flanked with adaptors allowing multiplexing (Table III). Barcodes (4bp) were added between the Illumina adaptor and the specific DpnII primers. Libraries were prepared by means of 8–10 independent PCR reactions using 70–100 ng of DNA per reaction. PCR products were pooled and purified using the PCR purification kit (Qiagen). Multiplexed libraries were sequenced on Illumina HiSeq 2500 at the Sequencug platform of the University of Geneva to obtain 100 bp single-end reads. Demultiplexing, mapping and 4C-seq analysis were performed using a local version of the pipeline described in 88, on the mouse assembly GRCm38 (mm10). The profiles were smoothened using a window size of 11 fragments and normalized to the mean score in +-5 Mb around the viewpoint. When multiple independent biological replicates were available, average 4C-seq profiles were calculated.
Data are available in GEO (accession number: GSE161386).
ACKNOWLEDGMENTS
We thank the transgenesis and sequencing platforms of the University of Geneva. We also thank Aurélie Hintermann for help in chick embryo dissection and for discussions.
Footnotes
Conflict of interest disclosure: The authors declare no competing interests.
REFERENCES
- 1.↵
- 2.
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.
- 8.
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.
- 22.↵
- 23.↵
- 24.↵
- 25.
- 26.
- 27.
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.
- 69.↵
- 70.↵
- 71.↵
- 72.
- 73.
- 74.
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵