The Wnt Co-Receptor Lrp5 Is Required for Cranial Neural Crest Cell Migration in Zebrafish

During vertebrate neurulation, cranial neural crest cells (CNCCs) undergo epithelial to mesenchymal transition (EMT), delaminate from the neural plate border, and migrate as separate streams into different cranial regions. There, they differentiate into distinct parts of the craniofacial skeleton. Canonical Wnt signaling has been shown to be essential for this process at different levels but the involved receptors remained unclear. Here we show that the frizzled co-receptor low-density-lipoprotein (LDL) receptor-related protein 5 (Lrp5) plays a crucial role in CNCC migration and morphogenesis of the cranial skeleton. Early during induction and migration of CNCCs, lrp5 is expressed ubiquitously but later gets restricted to CNCC derivatives in the ventral head region besides different regions in the CNS. A knock-down of lrp5 does not interfere with induction of CNCCs but leads to reduced proliferation of premigratory CNCCs. In addition, cell migration is disrupted as CNCCs are found in clusters at ectopic positions in the dorsomedial neuroepithelium after lrp5 knock-down and transient CRISPR/Cas9 gene editing. These migratory defects consequently result in malformations of the craniofacial skeleton. To date, Lrp5 has mainly been associated with bone homeostasis in mammals. Here we show that in zebrafish, lrp5 also controls cell migration during early morphogenetic processes and contributes to shaping the craniofacial skeleton.


Introduction
Neural crest cells (NCCs) are multipotent precursor cells that are specified at the boundary between neural plate and epidermis upon induction by growth factors such as Wnts, BMPs, FGFs and retinoic acid (RA) [1]. Committed NCCs undergo an epithelial to mesenchymal transition (EMT) before delaminating from the neural plate and migrating ventrally along distinct routes. Depending mostly on extrinsic cues derived from targeting tissues, they differentiate into various cell types and tissues such as neurons of the enteric and peripheral nervous system, endocrine and para-endocrine derivatives and pigment cells [2]. In the amniote head, cranial neural crest cells (CNCCs) migrate ventrally from hindbrain rhombomeric regions into the pharyngeal arches and the frontonasal process where they give rise to facial cartilage, bone and connective tissue. Three characteristic major streams can be distinguished: The mandibular, hyoid and branchial stream [3]. The branchial stream originates from the neuroepithelium of rhombomeres 6 to 8 and invades the 3 rd to 7 th pharyngeal arches. In actinopterygians, each of these five arches will give rise to one of the five ceratobranchials, in addition to other splanchnocranial elements. While several factors that control arch formation have been uncovered, particularly in zebrafish [4,5], the detailed mechanisms linking CNCC proliferation and migration to differentiation remain unclear.
A number of studies revealed that canonical Wnt signaling is one of the crucial signal transduction pathways involved in all NCC related processes that take place in the course of development [6]. In Xenopus laevis, it was shown that activation of Wnt signaling induces ectopic neural crest [7]. In contrast, blocking Wnt signaling by misexpression of GSK3β [8], dominant-negative Wnt8 [9], truncated Tcf3 [10] or Nkd [11] resulted in the disruption of neural crest formation. Thus, Wnt signaling is important for induction of NCCs. In zebrafish, a knock-down of Wnt8 by antisense Morpholinos blocked early NCC induction and a critical phase for NCC induction has been observed by expression of truncated Tcf under control of a heatshock-inducible promoter [10]. Wnts furthermore regulate proliferation and subsequent delamination of NCCs from the dorsal neuroepithelium in chicken [12]. A role in migration has also been suggested since LiCl-mediated GSK3β inhibition prevents cell migration and blocks cell-matrix adhesion in cultured neural crest cells [13]. In Xenopus, a role for the frizzled co-receptor low-density-lipoprotein (LDL) receptor-related protein 6 (Lrp6) has been suggested for NCC induction since its misexpression expands the neural crest. In contrast, overexpression of a truncated dominant-negative form of Lrp6 seemed to reduce the number of neural crest cells [14].
Gene expression analysis in Xenopus showed that also Lrp5, another co-receptor in canonical Wnt signaling [15], is expressed in the neural crest and its derivatives [16]. In mammals, Lrp5 plays a major role in bone homeostasis, and mutations in LRP5 are associated with reduced bone mass leading to the osteoporosis-pseudoglioma syndrome in humans [17]. Conversely, gain of function mutations in LRP5 at the N-terminus lead to a high bone mass phenotype as binding of its endogenous inhibitor Sost is prevented [18][19][20]. Mutations in Lrp5 in mice lead to reduced proliferation of osteoblast precursors [21]. On the other hand, patients with loss of function mutations in SOST suffer from sclerosteosis, a progressive sclerosing bone dysplasia, comparable to gain of function mutations of LRP5 in humans [22,23]. So far, no direct links between mutations in LRP5 and early developmental defects of the craniofacial skeleton have been made in mammals. Importantly, however, there are reports about cranial bone dysmorphologies in human patients with LRP5 gain of function mutations, such as craniosynostosis at an early age [24] or a large lobulated torus palatinus and an abnormally thick mandibular ramus [25]. This opens the possibility that a potential role of lrp5 in craniofacial morphogenesis is conserved between different groups of vertebrates.
In the present study, we analyzed lrp5 expression during zebrafish embryogenesis and used gene knock-down as well as transient CRISPR/Cas9 mediated gene editing to study its function during head skeleton formation. We show that zebrafish lrp5 is ubiquitously expressed at early stages but gets restricted to CNCC derivatives during later embryonic and larval development. Interfering with lrp5 function results in reduced proliferation of neuroepithelial cells in the hindbrain causing migration defects of CNCCs, which ultimately result in severe developmental defects of the cranial skeleton. Our data thus are consistent with a model that proposes a link between cell cycle progression and NCC delamination [12] and assigns an important role for Lrp5 in this process.

Fish keeping and husbandry
The authors confirm that the Institutional Animal Care and Use Committee (IACUC) of the National University of Singapore (NUS) has specifically approved this study (approved protocol numbers 020/08, BR19/10, 014/11). Adult zebrafish of DBS inbred wild-type strain as well as sox10:GFP [26], fli1:EGFP [27] and TOPdGFP transgenic zebrafish [28] were used to obtain embryos that were raised in 30% Danieau's solution at 28°C and staged as described previously [29]. At desired stages, embryos and larvae were fixed in 4% paraformaldehyde (PFA).

Morpholino and CRISPR/Cas9 injection
For gene knock-down of lrp5, the following splice blocking Morpholinos (MOs; Gene Tools, Corvalis, USA) were used: lrp5MoUp (AGCTGCTCTTACAGTTTGTAGAGAG) targets the Exo-n2-Intron2 splice junction; lrp5MoDown (CCTCCTTCATAGCTGCAAAAACAAG) targets the Intron2-Exon3 splice junction (see S1 Fig). A mismatch morpholino (mmlrp5) containing five nucleotide substitutions (AGgTGCTgTTAgAGTTTcTAGAcAG) was designed as control. The used p53-Morpholino was previously described [33]. For injection, 3 mM stock solutions were diluted with H 2 O and 0.1% phenol red was added. Working solutions were injected into the yolk of one or two-cell stage embryos as described previously [34]. To assess knock-down efficiency by semi-quantitative RT-PCR, total RNA was isolated from wild-type and MO injected embryos at the 25 somite stage (ss). After DNase treatment and reverse transcription, spliced transcripts were PCR amplified with primers Lrp5MoChkup (CAGTGGACTTTCTCTTCTCG) and Lrp5MoChkdown (GTCTCCGAGTCAGTCCAGTA). To amplify transcripts with retained introns, primers Lrp5MoChkup and Lrp5MointronChkdown (CTAAGATTGTGGGTCACAGG) were used.

Immunostaining and skeletal staining
Immunohistochemistry was carried out as previously described [34]. To stain for cells in Mphase, rabbit derived monoclonal anti-phosho-histone3 (pH3) antibody (Upstate Biotechnology, NY) was used in combination with anti-rabbit Alexa568 coupled secondary antibody (Invitrogen). To detect cells in S-phase, embryos were incubated in 10 mM BrdU for 30 minutes, washed several times and kept another 30 minutes before fixation in 4% PFA overnight. After this, embryos were washed in PBST and kept overnight in methanol. Then, embryos were rehydrated, followed by incubation in 2N HCl for 1 hour at 37°C. BrdU-positive nuclei were stained by using mouse anti-BrdU antibody (BSHB, Iowa City, IA; diluted 1:500 in PBDT) in combination with anti-mouse Alexa 488 coupled secondary antibody (diluted 1:1000; Invitrogen). Combined staining for cartilage (Alcian Blue) and bone (Alizarin Red) was performed as previously described [36].

Image acquisition
For microscopy, stained embryos were mounted in 100% glycerol. For flat-mount preparations, the yolk was manually removed. Images were taken with a Nikon SMZ1000 stereomicroscope, a Nikon T1-SM inverted microscope with GFP filter set and a Nikon Eclipse 90i upright microscope using the NIS-element BR software (Nikon). For confocal microscopy, a LSM 510 Meta laser scanning confocal microscope (Zeiss) was used. Alexa488 was detected by excitation with an argon multi-line gas laser at 488 nm and detection through the BP 505-530 nm filter. Alexa568 was detected by excitation with a Helium Neon gas laser at 543 nm and detection through the LP 560nm filter. LSM software (Zeiss) was employed for confocal image processing. For histological analysis, specimen were fixed in a mixture of 1.5% glutaraldehyde and 1.5% paraformaldehyde in 0.1 M cacodylate buffer and processed for embedding in Epon, according to standard procedures. Transverse 1 μm semi-thin cross sections of the head were stained with toluidine blue for 1-2 min (0.2% toluidine blue, 2% Na2CO3), rinsed with water, air-dried and mounted with DPX (Fluka, Buchs, Switzerland). A Zeiss Axio Imager-Z1 compound microscope equipped with a Zeiss Axiocam MRc camera was used for imaging.

Quantification and statistical analysis
For analysis of cell proliferation, cell numbers were determined in images of multiple individuals (n). A region of interest (ROI) was defined to cover the hindbrain region between rhombomeres 4 to 8 (r4-8). Within this area, all positively stained nuclei were counted. Average, standard deviation and P-values (Student's T-test) were determined using Microsoft Excel 2011.

Characterization of the zebrafish lrp5 sequence
Sequence information for zebrafish lrp5 is available in Ensembl (ENSDARG00000006921). Lrp5 is located on chromosome 25 and spans a region of more than 120 kb (10,854,981,755). It contains an open reading frame (orf) of 4845 base pairs (bp) in 24 exons. The deduced amino acid (aa) sequence is 1615 aa. Sequence comparisons revealed identities of 76% (to human); 75% (mouse); 79% (chicken); 76% (Xenopus); 67% (Fugu); 80% (Medaka) and 45% (Drosophila), respectively. Thus, the degree of conservation of the Lrp5 amino acid (aa) sequence is high. Fig 1A shows an aa sequence alignment for the first YWTD propeller motif (β1). This motif is an important regulatory domain and is the binding site for SOST [37]. A glycine residue at position 171, which is mutated to valine in human patients with high bone mass phenotypes [25], is conserved in all analyzed species including zebrafish (highlighted in grey in Fig 1A).

lrp5 expression during zebrafish embryogenesis
In zebrafish, lrp5 is maternally expressed and transcripts are found throughout the embryo at the eight-cell stage (data not shown). lrp5 remains ubiquitously expressed through gastrulation until the 10 somite stage (ss) (Fig 1B). Around 25 ss, its expression becomes restricted with elevated levels found in the developing brain and tail (arrow in Fig 1C). In the brain, regionally restricted expression is found in the fore-, mid-and hindbrain with the anterior hindbrain region being devoid of any expression (arrowheads in Fig 1C and 1D). At 48 hpf, strong lrp5 expression persists in the head region ( Fig 1E) and weaker expression is observed in the trunk (not shown). Broad expression is found throughout the head with elevated levels in the dorsal hindbrain (white arrowhead in Fig 1E) and the diencephalon (arrow in Fig 1E). In contrast, expression is weak or absent in tectum and telencephalon. lrp5 is also strongly expressed in the ventral head region (black arrowhead in Fig 1E). At 72 hpf, lrp5 expression becomes further restricted to distinct domains ( Fig 1F). It is strongly expressed in the cranial region as well as in the pectoral fin anlagen (not shown). In the head, expression is found predominantly in the dorsal hindbrain (white arrowhead in Fig 1F) including the rhombic lip, as well as epiphysis (arrow, Fig 1F). No expression is detectable in tectum and telencephalon but strong expression is found throughout the forming pharyngeal skeleton in the ventral head region (black arrowhead, Fig 1F), which contains CNCC derivatives.

lrp5 knock-down leads to defects in hindbrain and CNCCs
To analyze the role of lrp5 during zebrafish embryogenesis, we first used a Morpholino (MO) based knock-down approach. Two separate splice blocking MOs (lrp5MoUp; lrp5MoDown, Figure  Consistent with an earlier report [38], knock-down of lrp5 resulted in severe hindbrain defects in embryos, which were morphologically most obvious at 48 hpf. Compared to wildtype controls (Fig 2A), lrp5 morphants had widely inflated hindbrain ventricles ( Fig 2B). CNCCs originate from the dorsal hindbrain, migrate ventrally and form large parts of the cranial skeleton (Fig 2C). We checked for the morphology of the ventral cranial skeleton in lrp5 morphants at larval stages by bone and cartilage staining. Compared to wild-type ( Fig 2D) and MoMM injected embryos (Fig 2E), lrp5 morphants exhibited severe malformations of the cranial skeleton (Fig 2F and 2G). In lrp5 mmMO injected embryos, general development was  slightly delayed as evident by delayed mineralization of the ceratohyals (compare Alizarin red staining in Fig 2D and 2E). Importantly however, CNCC derived cartilage structures formed normally in lrp5 mmMO injected embryos (Fig 2E). In lrp5 morphants, cranial phenotypes were grouped into two classes of severity. ClassI morphants were characterized by a complete loss of ceratobranchials 1-4 (arrowhead in Fig 2F) and reverse oriented ceratohyals. The 5 th ceratobranchial with attached pharyngeal teeth appeared normal (arrow in Fig 2F). In more severely affected classII morphants, only rudiments of the ventral craniofacial skeleton such as Meckel's cartilage or ceratohyal remained while the 1 st to 5 th ceratobranchials were completely absent (arrowhead in Fig 2G). Importantly, these defects could not be rescued by co-injection of a p53 MO (Fig 2H), excluding the possibility of unspecific apoptosis caused by MO off-target effects.
Analysis of toluidin blue stained histological sections of classI morphant larvae at 5 dpf revealed that pharyngeal cartilages were present and well differentiated in some areas, however their number, shape and position was affected (Fig 2J). This made their identification difficult with exception of the hyosymplectic. Wild-type larvae on the other hand showed well differentiated cartilaginous arches at this stage ( Fig 2I). Interestingly, dentition progressed normally in lrp5 morphants despite severe skeletal malformations. According to [39], wild-type larvae at 5 dpf ( Fig 2K) possess three teeth on each side, labeled 3V 1 , 4V 1 and 5V 1 , with 4V 1 attached and possessing a replacement tooth, 4V 2 , in early cytodifferentiation. Teeth 3V 1 and 5V 1 are in a similar stage at late cytodifferentiation and do not have a replacement tooth yet. In lrp5 morphants (Fig 1L), the same three teeth are present: 3V 1 , 4V 1 and 5V 1 . Individuals with tooth 4V 1 in late cytodifferentiation have teeth 3V 1 and 5V 1 in the morphogenesis stage. Individuals with tooth 4V 1 in an early cytodifferentiation stage have teeth 3V 1 and 5V 1 in an initiation stage only. Tooth 4V 1 displays no replacement tooth, but this is to be expected given that a replacement tooth develops only once its predecessor is attached, which is not the case. Taken together, analyzing tooth organization in lrp5 morphant larvae revealed that not all ventral head structures are generally affected. Rather, tooth development as well as development of other dermal skeletal elements, such as cleitra and operculae, appeared normal (or at most slightly delayed), while CNCC derived cartilage elements of the head skeleton were strongly affected. This opens the possibility that lrp5 at this stage is required for morphogenesis of specific CNCC derived craniofacial cartilage structures but not for head development in general.

Knock-down of lrp5 reduces canonical Wnt signaling
We next tested whether a lrp5 knock-down leads to reduced Wnt activity, in particular in the hindbrain [38]. For this, lrp5MOs were injected into TOPdGFP transgenic zebrafish embryos expressing destabilized GFP under control of a Lef1/β-catenin responsive element [40]. At 20 ss, all analyzed control embryos (n = 33) showed strong reporter activity in the mid-hindbrain boundary (MHB), as well as the hindbrain and tailbud as analyzed by in situ hybridization with a gfp riboprobe (Fig 3A and 3B) [40]. In lrp5 morphants, however, 80% (n = 51) of embryos showed an overall decrease of GFP expression including in the MHB and hindbrain (Fig 3C and 3D). This suggests a drastic down regulation of Wnt signaling after injection of lrp5 MOs.
To test whether Wnt signaling is reduced specifically in neural crest cells, we examined expression of lymphoid enhancer-binding factor 1 (lef1), a key downstream factor in Wnt signal transduction [41]. lef1 is expressed in migratory CNCCs at 20 ss as two bilateral stripes adjacent to the hindbrain (Fig 3E and 3F) [42]. 73% (n = 80) of lrp5 morphants showed a clear reduction in lef1 expression in all expression domains including the CNCCs (asterisk in Fig  3G, arrows in Fig 3H).
Interfering with lrp5 function affects migration, but not induction of CNCCs Wnt signaling has been shown to play multiple roles in neural crest induction, migration and differentiation [10]. To examine whether lrp5 is required for neural crest induction, we analyzed expression of the early CNCC marker forkhead box d3 (foxd3 or fkd6) [43]. At 10 ss, foxd3 expressing premigratory CNCCs are found bilaterally to and overlaying the dorsal neuroepithelium in the caudal hindbrain (Fig 4A and 4B). Expression of foxd3 was normal in lrp5 morphants (100%, n = 28; Fig 4C and 4D) when compared to wild-type controls (n = 32; Fig  4A and 4B). This suggests that lrp5 is not involved in early CNCC induction.
Next, we analyzed whether CNCC migration is altered after lrp5 knock-down. At 14-15 hpf, CNCCs start to migrate from positions lateral to rhombomeres 2, 4 and 6 towards the ventral pharyngeal arch region in three distinct streams. At 20 ss, cells in these streams express distal-less homeobox 2a (dlx2a; Fig 4E-4I) [44]. No dlx2a expression is found in the medial neuroepithelium (arrow, Fig 4G; transverse section, I). Knock-down of lrp5 drastically changed the pattern of migratory CNCCs. First, all CNCC streams appeared strongly reduced and disorganized in 45% of the lrp5 morphants (n = 64; Fig 4J and 4K). Moreover, a patch of ectopic dlx2a positive cells was observed in a dorsomedial position above the neural tube between the branchial CNCC streams at r6 (asterisks and arrow, Fig 4J,4K,4L and 4N). The r6 stream was severely reduced in size (Fig 4M) when compared to wild-type controls (Fig 4H). No changes in pattern were evident in lrp5 mmMo injected embryos (n = 59; Figures G and H in S1 Fig). This suggests that CNCC migration in the branchial stream is disrupted upon reduced Lrp5 activity.
To confirm this, migratory CNCCs were also examined for expression of the pan-neural crest marker crestin [45]. In wild-type embryos, crestin is expressed in cranial and trunk neural crest cells (Fig 4O). In controls at 20 ss, no crestin positive cells are found in the region dorsal to the neuroepithelium (arrow, Fig 4Q) but in two bilateral streams of migratory NCCs ( Fig  4P,4Q and 4R; transverse section in S). In 55% (n = 64) of lrp5 morphants, however, clusters of ectopic crestin positive NCCs were found in dorsomedial positions comparable to the situation for dlx2a (arrow, Fig 4V; asterisk in transverse section X). Likewise, branchial clusters of crestin positive migratory CNCCs were drastically reduced in size (Fig 4W) when compared to wildtype (Fig 4R) or MM morphants (n = 54; Figures K and L in S1 Fig).
To obtain further evidence for migratory defects in lrp5 morphants, lrp5MO was injected into sox10:GFP transgenic embryos [26]. In this line, cells of the neural crest lineage express GFP in migratory streams around 20 ss, and no GFP-positive cells are found in the dorsal hindbrain (arrow in Fig 4Y). In contrast, 54% of sox10:GFP embryos injected with lrp5Mo (n = 74) showed ectopic clusters of GFP positive cells in dorsomedial positions (arrow in Fig 4Z) similar to the situation for dlx2a and crestin. Also in lrp5 deficient sox10:GFP embryos, caudal clusters of migratory CNCCs were smaller when compared to controls.
Taken together, this suggests that a knock-down of lrp5 results in altered migratory behavior of CNCCs, while induction is not affected. Although the observed ectopic cells have migratory CNCC character as evident by dlx2a expression, they fail to follow the migratory streams and instead are retained in dorsomedial positions [46].
To validate the MO induced phenotypes, a CRISPR/Cas9 approach was used and two guide RNAs were designed and injected separately in a transient gene targeting assay. RFLP analysis revealed that both guide RNAs were efficient in mutating the chosen target sequence (Fig 5A).  Fig 5F-5Q), thus recapitulating the situation in lrp5 MO injected embryos. lrp5 CRISPR/Cas9 injected embryos also developed the typical severe defects in the ventral craniofacial skeleton (Fig 5R-5U), recapitulating the phenotypes seen in lrp5 morphants. Together this shows that the two used CRISPR/Cas9 guide RNAs validate the phenotypes observed after lrp5 knock-down and therefore confirms that lrp5 is required for neural crest cell migration and branchial arch formation.
A knock-down of lrp5 leads to reduced proliferation of premigratory CNCCs The size reduction of migratory CNCC streams suggested that cell numbers were affected by lrp5 deficiency. Premigratory NCCs are highly proliferative and canonical Wnt signaling controls the NCC cell cycle [12]. It has also been reported that NCC delamination is synchronized with the cell cycle [47]. We therefore assessed the proliferative status of neuroepithelial cells in the region of interest (roi), i.e. in rhombomeres 4 to 8 in lrp5 morphants at 20 ss. For this, we performed immunohistochemical staining for phosphorylated Histone 3 (pH3), which labels nuclei in M-phase (Fig 6A and 6B). WT control embryos showed an average of 42 pH3-positive nuclei in the roi (n = 11), whereas the number of pH3 positive nuclei in the roi of lrp5 morphant embryos was reduced to an average of 26 cells (37%; p < 10 −6 ; n = 11; Fig 6C).
To determine the number of cells in S-phase, we analyzed Bromodeoxyuridine (BrdU) incorporation in lrp5 morphants compared to wild-type controls at the same developmental stage and in the same roi. It was reported earlier that NCCs synchronize in S-phase before delamination from the neuroepithelium [47], thus being a prerequisite for the start of migration. WT embryos on average displayed 148 BrdU positive cells in the roi (n = 9). In contrast, lrp5 morphants on average only had 95 BrdU positive cells in the roi (36%; n = 11) indicating a reduction of S-phase nuclei (p = 1.05x10 -6 ; Fig 6D-6F).
To examine which cell cycle phase is affected in lrp5 morphants, we analyzed cyclin d1 (ccnd1) transcript levels. ccnd1 is expressed in G1 and responsible for G1/S-transition. It has been shown to be under positive transcriptional control of Wnt signaling [48], also in zebrafish neural crest [46]. Unexpectedly, ccnd1 was significantly up-regulated in the hindbrain of 68% (n = 22) of lrp5 morphants (Fig 6G-6J). This accumulation of ccnd1 transcripts suggests a possible G1 cell cycle arrest in hindbrain cells of lrp5 morphants. Consequently, this could lead to the reduced number of cells in S-phase as observed in lrp5 morphants (Fig 6E).

Reduced CNCC migratory streams cause cranial skeleton defects
Postmigratory CNCCs, once they reach their final destinations, differentiate to establish the head skeleton. We tested whether the observed defects in CNCC proliferation and migration result in reduced numbers of postmigratory CNCCs in the pharyngeal arches. For this, we knocked-down lrp5 in fli1:EGFP transgenic zebrafish that express GFP in CNCC derivatives as well as vascular endothelial cells [27]. At 30 hpf, the mandibular (md), hyoid (hy) and three branchial (br) patches of postmigratory CNCCs show distinct GFP expression in wild-type embryos (Fig 7A). In contrast, 65% of the lrp5 morphants (n = 32) showed disturbed organization in this area (Fig 7B). We followed development in the affected embryos over time and analyzed morphogenesis and position of GFP positive CNCC derivatives. At 48 hpf, pharyngeal arches were well established in the caudal head region of control embryos and visible as five clearly distinguishable columns of GFP positive cells (Fig 7C and 7E). In contrast, lrp5 morphants failed to establish proper pharyngeal arch morphology. Only one group of migratory Lrp5 in Craniofacial Development ceratohyal (ch) and the five ceratobranchials (cb) are distinguishable (Fig 7G and 7I). In lrp5 morphants, on the other hand, pharyngeal arches are absent and severe malformations are observed in the cranial skeleton. Only rudiments of the caudal pharyngeal arches remain ( Fig  7H). While most parts of the anterior head skeleton are visible in ventral views (mc; Fig 7J), more posterior structures are morphologically not distinguishable (ch, cb; Fig 7J). Together, this suggests that a lrp5 knock-down initially leads to proliferation defects in premigratory CNCCs, consequently leading to reduced migratory CNCC streams and decreased numbers of postmigratory CNCCs, which ultimately results in morphological defects in the cranial skeleton. In this study, we provide the first analysis of expression and activity of the Wnt co-receptor Lrp5 during zebrafish craniofacial development. Sequence alignments showed that zebrafish lrp5 is highly conserved with significant similarities to lrp5 sequences in other vertebrates as well as its ortholog arrow in Drosophila. Elevated and regionally restricted expression of lrp5 in the early hindbrain are a first hint that lrp5 might be involved in CNCC formation and migration. As reported previously in zebrafish [46], CNCCs not only derive from an area lateral to the neuroepithelium but also from the neuroepithelium itself, where notably lrp5 is expressed at this stage. CNCC migration starts at around 14 hpf and results in three distinct streams of migrating cells on both sides of rhombomeres 2, 4 and 6. Thus, lrp5 expression is found in areas with forming and migrating CNCCs. As development progresses, its expression remains associated with CNCC derivatives as they form the cartilage elements of the ventral cranial skeleton. Interestingly, the overall spatiotemporal expression of lrp5 in brain and the developing cranial skeleton corresponds well with that of Sost [49], a Wnt antagonist known to exert its function by binding to Lrp5 [25]. This suggests that also in teleosts both proteins might interact to control Wnt signaling.
In zebrafish lrp5 morphants, the most severe defects in viscero-cranial development were observed in ceratobranchials 1-4, while the 5 th ceratobranchial containing pharyngeal teeth and other dermal skeletal elements, such as cleitra and operculae, appeared unaffected at least in classI morphants, further underlining the complex role of Wnt signaling in zebrafish pharyngeal tooth formation [50]. In contrast to ceratobranchials 1-4, the 5 th ceratobranchials consist mostly of sox10:GFP-negative cells (data not shown). Pharyngeal teeth start to form from the pharyngeal epithelium lining the floor of the pharyngeal cavity opposite the ceratobranchials 5 [51] and no sox10:GFP positive cells are present in this region. Dentition was normal in lrp5 morphants and teeth formed in correct positions and with regular alignment according to the developmental stage. The stage of dentition was slightly delayed compared to wild-type, and corresponded to dentition at 56 to 72 hpf. Taken together, the absence of ceratobranchials 1-4 and presence of the 5 th ceratobranchial with normal dentition suggested that the effect of the lrp5 knock-down was restricted to rostral, sox10 positive subsets of CNCC derivatives in the ventral cranial skeleton.
We hypothesize that the induced skeletal defects resulted from events occurring earlier in development. Wnt signaling is involved in different steps of NCC development including NCC induction, also in zebrafish [10]. However, although lrp5 is expressed at the right place during NCC induction, it seems dispensable for this process, as the pattern of premigratory CNCCs was not affected in morphants. It is possible that our MO data reflect a hypomorphic condition due to incomplete knock-down by the used MOs. However, also after lrp5 CRISPR/Cas9 injection, normal foxd3 expression was observed therefore strongly suggesting that lrp5 is not required for NCC induction. This is particularly interesting since misexpression of a truncated dominant-negative Lrp6 variant in Xenopus leads to reduced NCC induction [14]. In lrp5 morphants, on the other hand, we observed aberrant localization of migratory CNCCs at more advanced stages. At 20 ss, when CNCCs have evaded from the neuroepithelium in wild-type embryos, cells of the branchial stream were retained in the dorsal regions of r6. These cells were identified as NCCs as they expressed crestin, sox10 and dlx2a [52]. Importantly, dlx2a is only expressed in migratory CNCCs. This therefore suggests that upon lrp5 knock-down early CNCCs had already completed EMT but failed to initiate migration towards the prospective location of pharyngeal arches. It remains a possibility that the observed ectopic CNCCs are a consequence of a delay rather than a complete stop in migration. In lrp5 CRISPR/Cas9 injected embryos, we found embryos where crestin positive cell clusters spread from medial to lateral positions (see Fig 5K). This could be suggestive for a delay in migration of individual cells. Alternatively, this pattern could also reflect possible defects in collective CNCC migration, where individual mutant cells affect neighboring wild-type cells in a migrating mosaic cell cluster. Based on our observations, we conclude that canonical Wnt signaling through Lrp5 is essential to initiate CNCC migration (Fig 8), as has been proposed from previous experiments in vitro [13]. In zebrafish, it had been shown that Wnt signaling affects N-cadherin localization through Ovo1 and thereby regulates NCC migration [53]. Wnts also activate snail, which is a repressor of E-cadherin [54]. Thus, the observation that CNCC migration is deficient in zebrafish lrp5 morphants and after CRISPR/Cas9 injection provides a first hint as to how cell migration is regulated by Wnt signaling in the cranial neural crest.
Delamination of NCCs is tightly intertwined with the cell cycle. Delaminating cells are synchronized in S-phase in a process mediated by Wnts [12]. In lrp5 morphants, the cell cycle of premigratory CNCCs appeared arrested with reduced numbers of rhombomere cells in M-and S-phase. This suggests a possible involvement of lrp5 in cell cycle control of premigratory CNCCs. Comparable phenotypes, i.e. reduced numbers of S-phase nuclei in premigratory CNCCs, were reported after expressing a dominant suppressor variant of TCF to inhibit Wnt signaling [46]. However, it remained unclear which Wnt components were responsible for this defect. Wnts regulate the cell cycle through transcriptional regulation of ccnd1 and thereby control G1/S-phase transition [46,48]. In our studies, we found that ccnd1 expression was significantly up-regulated in lrp5 morphants. The elevated ccnd1 levels in parallel with reduced numbers of cells in S-phase suggest an arrest during G1/S-phase transition. We speculate that there is a feedback on the ccnd1 promoter in the arrested cells resulting in accumulation of ccnd1 transcripts. Importantly, a comparable situation has been described in cell cultures where RNAi mediated inhibition of LEF1 also resulted in the up-regulation of Cyclin D1 [55]. Noteworthy, migration defects such as observed in our study are not necessarily a direct consequence of reduced proliferation of premigratory CNCCs and impaired S-phase synchronization. Knock-down of Semaphorin3D in zebrafish for example resulted in reduced proliferation of premigratory CNCCs yet importantly did not lead to a loss of migratory behavior [46]. This suggests that migratory properties of CNCCs are not directly linked to proliferation and that the migration defects observed in this study could also be caused by an independent function of Lrp5 in controlling migratory properties of CNCCs. We therefore speculate that Lrp5 mediated Wnt signaling could be essential for multiple steps in CNCC migration, including proliferation of premigratory CNCCs, S-phase synchronization during EMT as well as migration of delaminated CNCCs (Fig 8). Our finding that zebrafish lrp5 is involved in proliferation but not specification of CNCCs is consistent with findings for mouse Lrp5, which is required for proliferation of osteoblast precursors but not their differentiation [21].
The morphogenetic fates of individual migratory CNCCs streams have been described earlier [3]. We employed the fli1:EGFP transgenic line to follow the fate of postmigratory cells in the craniofacial skeleton of lrp5 morphants. In contrast to sox10:GFP, where GFP expression is restricted to NCCs, the fli1 line shows EGFP expression in NCCs and the entire pharyngeal arches, thus revealing information about the overall morphology of the cranial skeleton. This allowed us to visualize how the lack of branchial CNCCs resulted in deficient ceratobranchial morphogenesis and revealed that morphogenesis of the 5 th ceratobranchial was less affected in lrp5 morphants.
A teleost specific function for lrp5 in craniofacial development?
Present studies on Lrp5 functions are limited to bone metabolism and eye vascularization [17,22]. Loss-of-function mutations in Lrp5 cause the osteoporosis pseudoglioma (OPPG) syndrome [21]. Surprisingly, no craniofacial deficiencies or any other neural crest related abnormalities were reported in Lrp5 knock-out mice. However, observations in human patients suffering from genetically inherited LRP5 gain-of-function mutations reported mild aberrations in skull anatomy [25]. Some of these patients are characterized by abnormally thickened jaws or lobulated palates from earliest ages onwards. Patients with a specific gain-of-function mutation (A214T) also suffer from craniosynostosis [24]. The early onset of these deformations suggests that they are not a result of a progressive sclerosteosis as described in other Lrp5 gainof-function mutants. As jaw, palate and skull originate from the CNCC, it is thus tempting to speculate that the symptoms observed in human patients are caused also by neural crest aberrations that occurred during embryogenesis.
Our observations suggest a role for Lrp5 in craniofacial morphogenesis that appears more imperative in zebrafish than compared to its role in mammals. Lrp5 knock-down in fish resulted in severe craniofacial defects compared to the milder defects described in mouse models or human patients. We found that the most severely affected structures in the craniofacial skeleton were the ceratobranchials that support the gills. These structures derive from branchial streams of CNCCs that shifted their morphogenetic destination in the course of vertebrate evolution. In amphibians, they generate cells that eventually build up bones of the skull proper [56]. During human embryonic development, however, branchial NCCs play a less important role in cranium formation and contribute to the formation of squamosal, alisphenoid and hyoid bones [57]. It is therefore tempting to speculate that the different destinations of migratory CNCCs during branchial morphogenesis in various vertebrates reflects a modification or possible neo-functionalization of lrp5's role in teleosts and other non-mammalian vertebrates.