Posterior neural tube closure depends on 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 neural tube has never been described. Here, we show that Dlx5/6 expression is required during vertebrate posterior neural tube closure. Dlx5 presents a similar expression pattern in neural plate border cells during zebrafish and mouse posterior neurulation. Dlx5/6- inactivation in 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. Similarly, dlx5a/6a zebrafish morphants show defects of posterior neural tube closure accompanied by aberrant delamination of neural crest cells with altered expression of cell adhesion molecules and defects of motoneuron formation. Our findings provide new molecular leads to decipher the mechanisms involved during vertebrate posterior neurulation for a better understanding of the etiology of human congenital NTDs and other midline field defects.


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 closure 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.
In vertebrates, neural tube closure 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 marker 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 neural tube malformations.
Here we show that simultaneous invalidation of Dlx5 and Dlx6 in zebrafish and mouse results in similar defects of posterior neural tube closure. 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.

Dlx5/6 invalidation induces posterior axis malformations in zebrafish and mouse
To analyse the effect of Dlx5/6 invalidation on axis formation, we generated dlx5a/6a zebrafish morphants and Dlx5/6 -/mouse embryos and compared the resulting axial phenotypes. The disruption of Dlx5/6 function led to similar early posterior malformations characterized by curly-shaped tails in both species (Fig. 1 A-D, white arrowheads). In Dlx5/6 mutant mice, 80% of embryos and foetuses presented a curly-shaped tail associated with varying degrees of exencephaly ( Fig. 1, blue arrowhead), the latter phenotype known to result from defect of anterior neural tube closure 17,19,25,34 . Moreover, E18.5 mutant mice displayed medio-dorsal split in the thoracic/lumbar region ( Fig. 2A

Expression of Dlx5 during zebrafish and mouse posterior axis formation
To understand the origin of the axial phenotype observed in Dlx5/6-invalidated 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  We also studied the effect of dlx5a/6a invalidation at later stages when the neural tube is We also studied the expression of shha and bmp4 that are organizers of tail development 11, [44][45][46][47] . 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 (Supp. Fig. 2A). At 48 hpf, shha expression is limited to the floor plate (Supp. Fig. 2C). In morphants, shha expression was maintained but well reveals the undulating phenotype of axial structures (Supp. Fig. 2A-D).
While bmp4 did not show obvious defect of expression at 16 hpf and 24 hpf 18 , expression in the spinal cord was altered at 48 hpf (Supp. Fig. 2E-F).
Altogether, our data indicated that disrupted dlx5a/6a function in zebrafish led to loss of expression of cell adhesion molecules in protruding NCCs at the dorsal midline, resulting in defects of posterior neural tube closure and of primary motoneuron formation.

DISCUSSION
It has been described previously that Dlx5 is one of the earliest NPB marker during gastrulation 20,21,24 . Particular attention has been paid on its role during anterior neural tube formation in defining the border between non-neural and neural plate territories 11,21,34,48,49 .
However, the role of Dlx5/6 during posterior neurulation has never been reported.
Our analysis in zebrafish and mice shows that Dlx5 is expressed in NPB cells during posterior neurulation (Fig. 3) 18 . Dlx5 expression is also detected in the VER, a source of midline ventral ectoderm known to act as a signalling centre during tail morphogenesis 50,51 .  54 . 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 neural tube closure 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 6,55 . 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 disrupted Dlx5/6 function in zebrafish and mouse leads to early curly-shaped tail phenotypes in both models (Fig. 1). In Dlx5/6 -/mice, the tail phenotype is associated with dorsal axis defects and brain malformations characteristic of NTDs (Figs. 1-2) 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 that show neural tube closure defects 5,56,57 . Our results suggested that the axis phenotype observed in Dlx5/6-invalidated zebrafish and mice resulted from defects of posterior neural tube closure, an aspect that we confirm through our functional analysis in zebrafish.
We show that dlx5a/6a morphants present defects of neural tube formation associated with aberrant dorsal delamination and migration of NCCs (Figs. [4][5]. The protruding NCCs observed at the dorsal midline of the neural keel show loss of cell adhesion molecule transcripts (Fig. 4). It has been shown that these latter are important actors during neural tube formation 7 . In particular, ncad is required for NPB convergence, neural tube closure, maintenance of neural tube integrity and epithelial-to-mesenchymal transition 10 [58][59][60][61][62] . Our data bring new light on common aetiology for a spectrum of idiopathic anomalies characterizing certain human congenital disorders.

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 the French Agriculture Ministry (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 63 . 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, controls and morphant embryos were raised at similar densities in embryo medium in a 28.5°C incubator. Embryos were treated with 0.0015% 1phenyl 2-thiourea (PTU) to inhibit melanogenesis. Embryos 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 strain Dlx5/6 +/was maintained on a hybrid genetic background resulting from the cross between a C57BL/6N female and a DBA/2JRj male (B6D2N; Janvier Labs, France).

Morpholino-mediated knock down
The morpholino-mediated knock down was performed as previously described 18 . To ensure specificity of the morpholinos, rescue of the resulting morphant phenotypes was performed as previously described 18 .

Histological analyses
In situ hybridization on whole-mount zebrafish embryos were performed as previously described 18 . Skeletal preparations on mouse foetuses were performed as previously described 64 .
Scale bar in A for A-F 100µm, for A'-D' 25 µm and for G-L 50µm.