Posterior axis formation requires Dlx5/Dlx6 expression at the neural plate border

Neural tube defects (NTDs), one of the most common birth defects in human, present a multifactorial etiology with a poorly defined genetic component. The Dlx5 and Dlx6 bigenic cluster encodes two evolutionary conserved homeodomain transcription factors, which are necessary for proper vertebrate development. It has been shown that Dlx5/6 genes are essential for anterior neural tube closure, however their role in the formation of the posterior structures has never been described. Here, we show that Dlx5/6 expression is required during vertebrate posterior axis formation. Dlx5 presents a similar expression pattern in neural plate border cells during posterior neurulation of zebrafish and mouse. Dlx5/6-inactivation in the mouse results in a phenotype reminiscent of NTDs characterized by open thoracic and lumbar vertebral arches and failure of epaxial muscle formation at the dorsal midline. The dlx5a/6a zebrafish morphants present posterior NTDs associated with abnormal delamination of neural crest cells showing altered expression of cell adhesion molecules and defects of motoneuronal development. Our findings provide new molecular leads to decipher the mechanisms of vertebrate posterior neurulation and might help to gather a better understanding of human congenital NTDs etiology.


Introduction
Neural tube defects (NTDs) correspond to a wide spectrum of common congenital disorders resulting from total or partial failure of neural tube closure during early embryogenesis. NTDs affect from 0.3 to 200 per 10 000 births worldwide [1] and vary in type and severity depending on the neural tube levels affected along the antero-posterior axis. Anterior and posterior neural tube defects lead respectively to brain (ie. exencephaly, anencephaly) or spinal cord malformations (ie. spina bifida); complete antero-posterior defect in neural tube closure is at the origin of a more severe form of NTD termed craniorachischisis (reviewed in [2,3]). The origins of NTDs have been associated to genetic and/or environmental factors and more than 200 mutant mice have been reported to present different forms of neural tube malformations [4,5]. However, given the complexity of the NTD spectrum, there has been limited progress in defining the molecular basis of these conditions. PLOS  In vertebrates, neural tube defects originate from a failure in morphogenetic events taking place during the neurulation process. In mammalian embryos, neurulation involves two distinct morphogenetic processes along the rostro-caudal axis, known as primary and secondary neurulations. Primary neurulation refers to neural tube formation originating from folding of an open neural plate that forms the central lumen in the anterior part of the embryo. In contrast, secondary neurulation is characterized by mesenchymal condensation and cavitation in the posterior axis caudal to the tail bud [6,7]. In zebrafish, neurulation occurs homogeneously along the rostro-caudal axis by epithelial condensation forming the neural plate, followed by cavitation as observed during secondary neurulation in mammals [8,9]. Zebrafish neurulation has been linked either to primary or to secondary neurulation of higher vertebrates [6,7]. However, the morphogenetic similarities observed between neurulation in teleosts and other vertebrates indicate that zebrafish neural tube formation rather correspond to primary neurulation and constitutes a viable model to study vertebrate neural tube development [6,10].
The general primary neurulation dynamic seems conserved among vertebrates and is characterized by convergent movement of the neural plate borders (NPB) toward the dorsal midline to generate the neural tube with a central lumen [6,9]. NPB cells constitute a competence domain, established between neural and non-neural ectoderm, that delineates the presumptive domain at the origin of migratory neural crest cells (NCCs) and responsible for neural tube closure [11,12].
Dlx genes, the vertebrate homologues of distal-less (dll) in arthropods, code for an evolutionary conserved group of homeodomain transcription factors. The mouse and human Dlx gene system is constituted by three closely associated bigenic clusters located on the same chromosomes as Hox genes clusters. In teleost, the dlx clusters are arranged on chromosomes similarly to their tetrapod Dlx counterparts [13]. The most probable scenario suggests that Dlx genes have arisen from an ancestral dll gene as a result of gene duplication events [14]. Data indicate that Dlx genes from a same cluster, such as Dlx5 and Dlx6 paralogs, present redundant functions during vertebrate development [15][16][17][18].
It has been shown that Dlx5 is one of the earliest NPB markers defining the limit of the neural plate during neurulation of mouse, chick, frog and zebrafish [18][19][20][21][22][23][24]. Inactivation of Dlx5 in mouse results in a frequent exencephalic phenotype suggesting defects of anterior neural tube closure [16], however the mice do not present obvious posterior axis malformations. As Dlx5 and Dlx6 have partially redundant functions, it has been necessary to simultaneously inactivate both genes to fully reveal their roles during development. Functional analyses of Dlx5/Dlx6 inactivation in mice, avian and fish have demonstrated evolutionary conserved roles in appendage morphogenesis, in neurogenesis, in the development of the face and of the reproductive system [16][17][18][25][26][27][28][29][30][31]. Dlx5/6 -/mice also present midline-fusion abnormalities including hypospadias, failure of anterior neuropore closure and tail malformations [17,31,32]. However, the origin of the latter phenotype has never been described.
Here we show that simultaneous inactivation of Dlx5 and Dlx6 in zebrafish and mouse results in posterior NTDs. Our data indicate a conserved role for Dlx5/6 in posterior neurulation in vertebrates and suggest that genetic pathways involving these genes might be implicated in syndromic forms of human midline defects.

Ethical statement
All experiments with zebrafish were performed according to the guidelines of the Canadian Council on Animal Care and were approved by the University of Ottawa animal care committee (institutional licence #BL 235 to ME). All efforts were made to minimize suffering; manipulations on animals were performed with the anaesthetic drug tricaine mesylate (ethyl 3-aminobenzoate methanesulfonate; Sigma-Aldrich, Oakville, ON, Canada). Embryos were killed with an overdose of the latter drug.
Procedures involving mice were conducted in accordance with the directives of the European Community (council directive 86/609) and reviewed and approved by the "Cuvier" ethical committee of the Muséum National d'Histoire Naturelle and the French Ministry of Agriculture (council directive 87-848, 19 October 1987, permissions 00782 to GL).

Animal maintenance
Zebrafish and their embryos were maintained at 28.5˚C according to methods described in [33]. Wild-type adult zebrafish were kept and bred in circulating fish water at 28.5˚C with a controlled 14-h light cycle. Wild type (WT), controls and morphant embryos were raised at similar densities in embryo medium in a 28.5˚C incubator. Embryos were treated with 0.0015% 1-phenyl 2-thiourea (PTU) to inhibit melanogenesis and were killed with an overdose of tricaine mesylate for analysis.
Mice were housed in light, temperature (21˚C) and humidity controlled conditions; food and water were available ad libitum. WT animals were from Charles River France. The mouse heterozygous strain Dlx5/6 +/was maintained on a hybrid genetic background resulting from crosses between C57BL/6N females and DBA/2JRj males (B6D2N; Janvier Labs, France); the mice are viable and present a normal phenotype [25,27,28,32]. The Dlx5/6 -/homozygous foetuses were obtained by crossing Dlx5/6 +/mice and our controls have been either WT or Dlx5/6 +/littermates.
As controls, we injected water or 1.6 mM of control standard MO (Gene Tools) that targets a human beta-globin intron mutation that causes beta-thalessemia. (Gene Tools 5'-CCTC TTACCTCAGTTACAATTTATA-3'). To ensure MOs specificity, rescue of morphant phenotypes was performed by co-injecting the corresponding dlx5a/dlx6a transcripts mutated on the MO target site (dlx5a MO binding site T(ATG)ACTGGAGTATTCGACAGAAGGA, mutdlx5a sequence C(ATG)ACGGGTGTTTTTGATAGGAGGA; dlx6a MO binding site AAAGCAGAAA ATTTG(ATG)ATGACCA, mutdlx6a sequence ATTGCAAATAATATG(ATG)ATGACCA). Mutated transcripts were synthesized using the SP6 mMessage mMachine kit (Ambion). The quantitative analysis of control, knockdown and rescue experiments are detailed in (S2 Fig) and [18]. Based on experimental results, co-injection of dlx5a/dlx6a translation-blocking MOs (0.8 mM each) and resulting severe phenotype specimens were selected for analysis compared to control embryos injected with water.

Histological analyses
In situ hybridization on whole-mount zebrafish embryos were performed as previously described [18].
For whole-mount immunostaining on zebrafish embryos, dechorionated embryos were fixed in 4% paraformaldehyde (PFA) in 1X phosphate buffered saline (PBS) overnight at 4˚C, washed in PBST (PBS 0.1% Tween), dehydrated in methanol and stored in methanol 100% at -20˚C. The samples were then rehydrated in graded methanol-PBST series and treated with PBDT (PBS 1% DMSO 0.1% Tween). Cells were immunodetected on whole-mount embryos with mouse anti-SV2 monoclonal antibody diluted in PBDT an incubated overnight at 4˚C (1/ 100, AB 2315387, DHBS). After 5 rounds of 30 min washes in PBST, embryos were incubated overnight at 4˚C with secondary anti-mouse HRP-conjugated antibody diluted in PBST (1:200, Jackson Immuno), washed 5 times 30 min in PBST and revealed with DAB chromogenic substrate.
Skeletal preparations on mouse foetuses were performed as previously described [34].

Dlx5/6 inactivation induces posterior axis malformations in zebrafish and mouse
Axial phenotypes were analysed in dlx5a/6a zebrafish morphants and Dlx5/6 -/mouse embryos; a curly-shaped tail phenotype was observed in both species (Fig 1A-1D, white arrowheads). 80% mutant mouse embryos presented a curly tail associated with varying degrees of exencephaly (Fig 1, blue arrowhead); the latter phenotype is known to result from defective anterior neural tube closure [17,19,25,35]. The curly tail phenotype was invariably associated to a medio-dorsal split in the thoracic/lumbar region (Fig 2A and 2B, red arrowheads). Skeletal preparations and immunostainings revealed open vertebral arches dorsally at both thoracic and lumbar levels, characteristic of NTDs, and failure of epaxial muscle formation at the dorsal midline ( Fig 2C-2F, red arrowheads).

Expression of Dlx5 during zebrafish and mouse posterior axis formation
To understand the origin of the axis phenotype observed in Dlx5/6-inactivated specimens, we then compared the spatio-temporal expression of Dlx5 during zebrafish and mouse posterior neurulation. We previously showed in zebrafish that dlx5a-expressing ectodermal cells are laterally connected to the neural ectoderm to form the presumptive median fin fold [18]. At 15.5 hpf, dlx5a-expressing NPB cells along the neural keel follow a medial convergence toward the dorsal midline during neural rod formation at 16 hpf ( Fig 3A and 3B, black arrowheads). At later stages, dlx5a expression is limited to median fin fold ectodermal cells at 24 hpf and 48 hpf, and gradually decreased until 72 hpf [18]. Similarly, in mouse embryos, Dlx5 transcripts were detected in NPB cells surrounding the posterior neuropore and at the dorsal midline after neural tube closure at E8.25 and E9.5 (

Posterior neurulation defects in dlx5a/6a zebrafish morphants
To further investigate the role of Dlx5/6 during posterior neurulation, we next performed molecular analyses in early dlx5a/6a zebrafish morphants during neural keel-rod transition at 16 hpf, shortly after the onset of dlx5a expression at the neural plate border (Fig 3B) [18]. Using translation-blocking MOs against dlx5a and dlx6a, we obtained a high proportion (74%) of severe curly tail phenotypes compared to control and rescued embryos (S2 Fig) [18]. Given the key role for cell adhesion molecules (CAMs) in neural tube morphogenesis, neural tube closure and epithelial-to-mesenchymal transition [10,[36][37][38][39], we analysed the expression of ncad (cdh2) and ncam3, members of the CAM family involved in cell-cell adhesion. We also analysed expression of msx1b, marker of NPB cells and premigratory neural crest cells (NCCs) [40,41], and expression of foxd3, marker of premigratory and early migratory NCCs [42].
In 16 hpf control embryos, ncad is constitutively expressed in the neural keel and the presomitic/somitic mesoderm. In contrast, ncam3 expression is limited to the dorsal part of the neural keel (Fig 4A, 4A', 4C and 4C'). In dlx5a/6a morphants, we observed a loss of ncad and ncam3 expression in aberrant protruding cells at the dorsal midline of the neural keel ( Fig 4B' and 4D', black arrowheads). Moreover, whole-mount expression patterns of msx1b and foxd3 Posterior axis formation requires Dlx5/Dlx6 expression at the neural plate border in morphants revealed a defect of neural tube formation, with bifid stripes of expression in the caudal-most part of the axis, characteristic of a delay in neural keel-rod transition (Fig 4E-4H, grey arrowheads). On sections, msx1b showed a decrease of expression at the midline where protruding cells were detected in morphants (Fig 4E'-4F', black arrowhead). At 16 hpf, foxd3 expression in the dorsal neural tube labelled premigratory NCCs (Fig 4G'). The analysis of dlx5a/6a morphants revealed that the aberrant protruding cells at the dorsal midline of the neural keel were positive for foxd3 expression (Fig 4H', black arrowhead). The results indicated that disrupted dlx5a/6a function affects msx1b and CAMs (ncad/ncam3) expression at the roof plate of the neural keel in aberrant delaminating foxd3-positive NCCs.
We then studied the effect of dlx5a/6a inactivation at later stages of neural tube formation. In 24 hpf controls, dorsal foxd3 expression in migratory NCCs was observed in the caudal neural tube (Fig 5A and 5A'). In dlx5a/6a morphants, foxd3-positive cells showed abnormal asymmetric profile (Fig 5B', black arrowhead) associated with a reduced neural tube and defect of lumen formation (Fig 5A'-5B', dashed lines), the latter phenotype being well revealed by the Posterior axis formation requires Dlx5/Dlx6 expression at the neural plate border constitutive ncad expression in the neural tissue (Fig 5C'-5D'). Beside NCCs, foxd3 and ncad transcripts were also detected in somites (Fig 5A and 5C); their expression was strongly increased in dlx5a/6a morphants. In control embryos, somites were arranged as V-shaped chevrons, while in dlx5a/6a morphants the somite boundaries presented abnormal U-shaped chevrons (Fig 5A-5D, dashed lines). A similar defect of somitic segmentation was also observed at earlier stages of development thanks to the analysis of fgf8a expression, a marker of anterior somitic mesoderm (S3 Fig). The expression profile of msx1b along the neural tube was also altered in dlx5a/6a MO embryos at 24 hpf (Fig 5E and 5F, black arrowhead), while expression in the median fin fold compartment seemed unaffected.  We next analysed the primary motoneuron (PMN) population in dlx5a/6a morphants; these cells originate from NCCs and are known to be affected in NTDs [43,44]. In 24 hpf controls, PMNs are characterized by synaptic vesicle SV2 expression; axonal projections start to elongate from the neural tube to reach their myotomal targets, forming the neuromuscular junctions at 30 hpf and 36 hpf (Fig 5G, 5I and 5K). In contrast, dlx5a/6a morphants showed defect of motoneuronal outgrowth at 24 hpf and PMNs failed to connect the myotomes at 30 hpf (Fig 5H and 5J). At 36 hpf, axonal projections were completed but PMNs showed defective neuromuscular junctions with aberrant synaptic arborescence (Fig 5L). PMN defects were associated with SV2 protein accumulation in the dorsal neural tube at 30 hpf and 36 hpf ( Fig  5J and 5L). Abnormal neuromuscular innervation was also present in Dlx5/6 mutant mice as shown by immunostainings for the neuronal marker Tuj1 in trunk epaxial muscles positive for Tnnt3 (S4 Fig, white arrowheads).
We also studied the expression of shha and bmp4 that are organizers of tail development [11,[45][46][47][48]. In 24 hpf controls, shha is expressed in the notochord, the neural tube floor plate and in the tail stem cell pool, namely the chordoneural hinge ( Altogether, our data indicated that disrupted dlx5a/6a function in zebrafish led to a loss of CAM expression in protruding NCCs at the dorsal midline, resulting in posterior neural tube defects and mispatterned NCC-derived primary motoneurons. Posterior axis formation requires Dlx5/Dlx6 expression at the neural plate border

Discussion
Dlx5 is one of the earliest NPB markers during gastrulation [20, 21, 24]. Particular attention has been paid to its role in anterior neural tube formation and in the specification of the border between non-neural and neural plate territories [11,21,35,49,50]. However, the implication of Dlx5/6 during posterior neurulation has never been reported.
Our analysis highlights the role of Dlx5/6 in NPB cells during posterior neurulation in zebrafish and mouse (Figs 1-5) [18]. In both models, Dlx6 expression was undetectable by in situ hybridization during early neurulation. In zebrafish, dlx6a expression is observed in the presumptive median fin fold that is related to dlx5a-expressing NPB cells during unpaired fin development [18]. Dlx6 expression appears delayed and weaker compared to expression of the Dlx5 paralog, a difference that was observed by others and us in vertebrate embryos using a variety of probes for these genes. The inactivation of both genes is however required to observe a fully penetrant phenotype, underlying the redundant functions of Dlx paralogs in vertebrates (S2 Fig) [16-19, [51][52][53].
Expression analyses for Dlx homologs in various models including chick, xenopus, lamprey and amphioxus suggest that Dlx expression in NPB cells has been established early during chordate evolution [21,24,49,54,55]. However, we can still notice differences in Dlx5 expression along the rostro-caudal axis of zebrafish and mouse. In mouse, Dlx5 is expressed in NPB cells during both anterior and posterior neurulation (Fig 3) [20]. In amphioxus, amphiDll is expressed in NPB cells all along the axis, suggesting a conserved role of Distal-less-related genes in anterior and posterior neurulation in chordates [55]. In contrast, in zebrafish, the anterior limit of dlx5a-expressing NPB cells is located at the 8 th somite level and appears closely related to the establishment of the presumptive median fin fold (Fig 3) [18]. According to our expression analysis, we did not find evidence of anterior neurulation defects in dlx5a/6a morphants [29]. In teleosts, neurulation is characterized by uniform epithelial condensation and cavitation, which give rise to both anterior and posterior neural tube [6,8]. It has been suggested that zebrafish might present primitive mechanisms of neurulation, however basal chordates show primary and secondary neurulation as observed in higher vertebrates [6,56].
Our results indicate that, while the cellular morphogenetic basis of neural tube formation is uniform along the rostro-caudal axis of zebrafish, the anterior and posterior neurulation processes do not involve same molecular mechanisms, suggesting evolutionary divergence of Dlx5/6 function during anterior neural tube formation in teleosts. Our data bring new insights into the genetic and evolutionary origins of neural tube formation in chordates. Special attention should be paid in future studies to elucidate the genetic requirement differences between anterior and posterior neurulation in teleosts.
Our analysis reveals that Dlx5/6 inactivation in zebrafish and mouse leads to early curly tail phenotypes in both models (Fig 1). In Dlx5/6 -/mice, the tail phenotype is associated with midline axis defects and brain malformations characteristic of NTDs (Figs 1 and 2) [17,19,25]. Intriguingly, the axis defects observed in Dlx5/6 -/mice was similar to the phenotype of CT "curly tail" mutants, a historical model of NTDs [5,57,58]. The data thus suggested that the axis phenotype observed in Dlx5/6-inactivated zebrafish and mice resulted from defects of posterior neurulation, an aspect that we confirm through our functional analysis in zebrafish.
We demonstrate that dlx5a/6a morphants present neural tube defects associated with aberrant dorsal delamination and migration of NCCs (Figs 4 and 5). The protruding NCCs observed at the dorsal midline of the neural keel show loss of CAM expression (Fig 4). Cell adhesion molecules are important actors during neurulation [7]. In particular, ncad is required for NPB convergence, neural tube closure, maintenance of neural tube integrity and epithelialto-mesenchymal transition [10,[36][37][38][39]. In zebrafish, ncad inactivation represses neural tube formation due to defect of convergence and intercalation of NCCs [10,36,37]. Our data thus indicate that dlx5a/6a genes act in NPB cells for adhesion integrity of NCCs during neural tube formation.
It has been previously demonstrated that Dlx5 specifies NPB cells during cranial neural tube formation in mouse and chick [20, 21, 24]. However, Dlx activity in xenopus is not necessary for induction of NCCs [49]. Our findings confirm that Dlx inactivation does not impact on NCC induction as foxd3 expression is maintained in the protruding NCCs observed in dlx5a/6a morphants (Fig 4). In addition, msxb1 expression is decreased in the protruding NCCs of dlx5a/6a morphants, associated with defects in primary motoneurons development (Figs 4 and 5). It has been shown that msx and ncad genes are required during zebrafish neurogenesis [37,40]. The neuromuscular deficiencies observed in dlx5a/6a morphants may result from altered expression of msx1b and CAMs in premigratory NCCs. In mouse, Dlx5 and Msx2 genes regulate anterior neural tube closure through expression of EphrinA5-EphA7 involved in cell adhesion [35]. This suggests that a common genetic network implicating Dlx, Msx and cell adhesion molecules is involved in neural plate border and neural crest cells during mouse and zebrafish neurulation.
Taken together, our findings reveal that Dlx5/6 genes are required during vertebrate posterior axis formation. The results show that Dlx5 expression is limited to NPB and VER cells (including the cloacal derivative) during posterior neurulation. The VER is known to act as a signalling centre during tail somitogenesis and elongation [59,60]. VER cells undergo epithelial-to-mesenchymal transition during tail development as observed dorsally in NCCs [61].
The continuous dorso-ventral ectodermal Dlx5-positive domain might reflect that the VER represents a ventral extension of dorsal ectodermal cells during tail morphogenesis. The inhibition of BMP signalling by Noggin suppresses the VER epithelial-to-mesenchymal transition process during tail morphogenesis [59,61]. In dlx5a/6a morphants, ectodermal Bmp4 expression is not affected during early posterior axis development [18], and the expression of Noggin has been shown to be independent of Dlx5 during craniofacial development [62]. The data thus indicate that BMP4 signalling is regulated independently of Dlx5 during early posterior axis formation. Moreover, Dlx5/6 expression in the VER does not appear required for tail elongation as both Dlx5/6 -/mice and dlx5a/6a morphants show equivalent number of axial segments compare to controls. However, as zebrafish morphants present abnormal somite boundaries associated to overexpression of foxd3 and ncad in posterior somites (S3 Fig, Fig 5), Dlx5/6 expression in the VER might ensure proper regulation of somitogenesis.
These new results give insights for a better understanding of the cellular and molecular processes that could be altered in some human congenital NTDs, such as craniorachischisis that originates from defects of both anterior and posterior neurulations. The anterior and posterior NTDs observed in Dlx5/6 mutant mice are also associated with hypospadias, characterized by midline urethral malformations, and limb ectrodactyly [17,31,32]. Expression of Dlx5/6 in the cloacal membrane, linked to the ventral ectodermal ridge [61], and in the genital tubercle is necessary for urethral formation. Moreover, it has been shown that Dlx5/6 expression in the apical ectodermal ridge is required for proper appendage formation in mouse and zebrafish [17,18,32]. The limb phenotype of Dlx5/6 -/mice resembles that of patients with congenital split hand-foot malformation type I (SHFM-I), linked to genomic deletion or rearrangement in the DLX5/DLX6 cluster locus. Altogether, the data unveil the role of Dlx5/6 in ectodermal cells for the proper development of the neural tube, the urogenital system and limbs. It has been reported that NTDs can be associated with limb malformations and other midline defects, including urogenital and diaphragmatic disorders, as observed in Czeizel-Losonci syndrome [63][64][65][66][67]. Our data bring new light on common etiology for a spectrum of idiopathic anomalies characterizing certain human congenital disorders.