The role of Rock isoforms in migrating NCC, and their derivatives, was investigated using mice expressing a RockDN construct ^[12] under the control of the Wnt1-cre driver ^[13], Figure 1A. The RockDN protein RB/PH (TT) ^[12], ^[14] is unable to bind to Rho and specifically blocks Rock function ^[15]. Female mice heterozygous for the RockDN construct and male mice heterozygous for the Wnt1-cre construct were mated (Figure 1A), with resulting litters containing embryos of the genotypes: RockDN⁺;Wnt1-cre⁺ (hereafter referred to as RockDN), RockDN⁺;Wnt1-cre⁻, RockDN⁻;Wnt1-cre⁺ and RockDN⁻;Wnt1-cre⁻; the latter three genotypes were phenotypically indistinguishable and acted as controls. To assess the efficiency of cre recombination we evaluated the levels of the CAT gene cassette transcripts in pharyngeal arches 1 and 2 of embryonic heads taken from RockDN⁺;Wnt1-cre⁺ mutants and RockDN⁺;Wnt1-cre⁻ controls at E11.5, as this tissue was rich in NCC (Figure S1 C,D). Quantitative real time PCR revealed a significant 13 fold reduction in levels of the CAT cassette in the mutant heads (P = 0.006), compared with transgenic embryos where cre was not present; a reduction to almost undetectable levels (Figure 1B). Efficient removal of the CAT cassette is consistent with expression of dominant-negative Rock in the majority of NCC. Notably, expression of RockDN in all cells of the embryo, by intercrossing RockDN mice with PGK-cre mice ^[16] suggested that ubiquitously blocking Rock function results in embryonic death before E9.5 (data not shown).

We reviewed craniofacial development in the RockDN;Wnt1-cre embryos. By late gestation, RockDN embryos showed craniofacial malformations affecting the fronto-medial aspect of the head, although these varied in severity. The predominant feature was mid-facial clefting in combination with hypoplasia of the frontonasal region (Figure 2). Those with the severest phenotype (47/54, from E11.5 to E18.5) had marked reductions of the frontonasal region, maxilla and mandible (Figure 2 A,B). Facial clefting was apparent in 50/54 embryos (Figure 2G and Figure S1 C–H), although in three cases this was apparent externally only as a bifid nose (Figure S1J). At E15.5, the distance between the eyes and the width of the frontonasal processes was larger in mutant embryos compared to control littermates (P = 4.202×10⁻⁶, Figure 2I and data not shown), consistent with hypertelorism and broader skulls in the mutants. There was no significant difference between the heights of the heads (data not shown). The snout was also upturned in the RockDN embryos at E15.5, shown by a decreased angle between the snout and the forehead when compared to control littermates (P = 3.208×10⁻⁸; Figure 2 E,H,J). There was no significant difference between the two groups for any measurement at E11.5, despite mid-facial clefting being apparent at that stage (data not shown).

At E17.5–E18.5, two out of four RockDN embryos had failed to form eyelids (Figure 2B), as seen in mice null for either Rock1 or Rock2^[7]–^[9]. This was not unexpected, as the eyelids form from NCC-derived mesenchyme ^[17]. Some abnormalities may have occurred as secondary events. The hypertelorism observed throughout gestation is likely to be secondary to the mid-facial clefting as this produces a broader skull. In addition, two out of four RockDN embryos examined at E17.5–E18.5 had exencephaly (Figure 2B). As all embryos collected earlier in gestation had closed cranial neural tubes, this does not suggest a primary involvement of Rock in neural tube closure, but rather suggests secondary opening, likely as a consequence of a defect in the NCC-derived calvarias bones. Cartilage staining of E14.5 embryos revealed that the nasal capsular cartilage is missing and the Meckel's cartilage appears shortened in the RockDN embryos (Figure 3 A,B). Bone and cartilage staining of more mildly affected E18.5 embryos demonstrated that the majority of the cranium was well formed although there were subtle abnormalities in some NCC-derived bones (Figure 3 C–F); specifically there was mis-shaping and hypoplasia of the basisphenoid bone (Figure 3 E,F). The bones of the nasal septum and the maxillary bones failed to meet in the midline (double arrow in Figure 3F) and the hyoid bone was hypoplastic (red arrow in Figure 3 C,D). Thus, expression of the RockDN construct in NCC resulted in defects in the craniofacial structures derived from the frontonasal prominence and the first pharyngeal arches. Severe abnormalities of the outflow region of the heart were also observed in all RockDN;Wnt1-cre embryos by E11.5; these are described elsewhere (Phillips et al., manuscript in preparation).

Inter-crossing of ROSA-EYFP mice with Wnt1-cre enabled the co-localisation of Rock1 protein with NCC. At E9.5 Rock1 was expressed in the neural tube and the dorsal root ganglia (Figure 4 A,D). Expression localised to the perimeter of the cells was found in the mass of the NCC within the pharyngeal arches (Figure 4 B,E and Figure S2 A–D) and in the frontonasal processes (Figure 4 C,F). Rock1 immunoreactivity was most abundant, however, at the boundaries of the NCC-derived mesenchyme with the epithelia of the surface ectoderm and the neural ectoderm in the frontonasal processes (arrows and arrowheads in Figure 4 C,F). Rock2 was expressed in similar NCC-rich regions (Figure S2 E–G) confirming that both isoforms were found in regions rich in migrating and post-migratory NCC. By E11.5, both Rock1 and Rock2 proteins were down-regulated in the craniofacial region (data not shown).

We used the Rosa 26R (β-galactosidase; β-gal) reporter to investigate the extent of migration of NCC and the distribution of their derivatives with regard to the observed craniofacial phenotypes. At E8.5, control and RockDN embryos were indistinguishable based on their external appearance. Staining for β-gal showed normal migration of NCC towards pharyngeal arches 1 and 2 (Figure 5 A,C), suggesting that NCC induction and delamination had occurred normally. However, there was a reduction in β-gal expression anterior to the stream of NCC migrating to the 1^st pharyngeal arch (Figure 5 A,C). At E9.5–E10.5, the mutant embryos were still not reliably distinguishable from control littermates on the basis of their craniofacial phenotype. However, the β-gal expression was generally more patchy in the craniofacial region of RockDN embryos examined at E9.5–E10.5 (Figure S1 A,B and Figure S3) and there appeared to be fewer β-gal-stained NCC in all embryos examined at this stage (9/9). Although the frontonasal structures appeared hypoplastic in 87% of RockDN mutant embryos by E11.5–E18.5 (Figure S1 C–H), β-gal expression was confluent in the developing facial prominences. It was deficient, however, in the posterior parts of the medial calvarias bones in embryos with the most severe craniofacial phenotype (data not shown) and below the eyes (Figure 5 K,L). In addition to the structural abnormalities already described, the thymus was ectopically located in the cervical region (data not shown) and the cranial ganglia were misshapen and reduced in size (arrowheads in Figure 5 E,G,I,J and data not shown).

The β-gal staining pattern also revealed abnormalities in NCC distribution during the development of the pharyngeal arches. At E8.5, the streams of NCC migrating towards the pharyngeal arches were similar in both wild type and RockDN mutant embryos (arrows in Figure 5 A,C). However, at E9.5, and persisting at E10.5, the first and second pharyngeal arches were hypoplastic and the β-gal expression in the arches was less intense and patchy (Figure 5 E–J). These data suggest that although the formation, delamination and early migration of NCC appeared to occur normally at E8.5 in the RockDN mutants, NCC numbers declined later in development.

Reduced Rock function is associated with increased apoptosis in developing motor neurones ^[12]. We therefore examined the expression of activated caspase 3 in the developing RockDN embryos, hypothesising that increased levels of apoptosis in the affected tissues might explain the emergent phenotype. NCC migration begins at E8.0 in the cranial region of mouse embryos and is complete by the end of E9.5 ^[18]. At E8.5, after the first NCC had delaminated from the neural tube, but before the appearance of any craniofacial phenotype, there was little if any activated caspase 3 labelling anywhere in the control embryos. However, a few activated caspase 3-expressing, dying, cells were observed in the frontonasal region and pharyngeal arches of RockDN embryos at this stage (data not shown). By E9.5, and continuing at E10.5, activated caspase 3-positive cells were abundant in the dorsal part of the neural tube in RockDN embryos, localising to the Wnt1-cre domain (Figure 6 A–D). This correlates with the latter stages of NCC delamination from this region but also continues after migration from the cranial neural tube is complete. The mesenchymal cells within pharyngeal arches 1–3 were compact in control embryos by E9.5. The surface ectoderm formed a smooth layer surrounding the inner ectomesenchyme (Figures 5F and 6 E,F). This outer layer appeared uneven in the RockDN mutants and the mesenchymal cells appeared loosely arranged and disorganised (Figures 5H and 6 G,H). Similar observations were made in the frontonasal processes of control and stage matched mutant embryos (arrowheads in Figure 6 I–L). Examination of the craniofacial region of RockDN embryos revealed that more than 25% of NCC were caspase 3-positive in pharyngeal arches 1–3 and the dorsal frontonasal prominences at E9.5 and E10.5 (Figure 6 E–N and data not shown), compared with only occasional caspase 3-expressing cells in comparable sections from control littermates (P = 0.019; Figure 6 E–N and data not shown). Thus, this high incidence of cell death likely explains the less compact appearance of the ectomesenchyme of the pharyngeal arches and frontonasal processes. In contrast to the elevated cell death, analysis of cell proliferation in NCC-rich regions of the developing craniofacial region revealed no significant differences between RockDN and control littermates at any stage examined (P = 0.433; Figure 6O and data not shown). These data suggest that the progressive reductions in NCC numbers observed in the craniofacial and pharyngeal regions at E9.5–E10.5, and the hypoplasia of NCC-derived structures observed in these regions later in development, were caused by loss of NCC by apoptosis.

As Rock is a key regulator of the actin cytoskeleton, and disruption of the actin cytoskeleton has been linked to induction of cell death in a variety of contexts ^[19] we carried out phalloidin staining for filamentous actin in sections taken from E9.5 control and RockDN embryos, as this was the stage when cell death was at its peak. In control embryos, filamentous actin was found in a marked cortical distribution, lining the perimeter of the cell, in the NCC-derived ectomesenchyme of the pharyngeal arches (Figure 7 A,E) and frontonasal region (Figure 7 C,G), as well as the non-NCC-derived neural ectoderm of these regions. Cortical phalloidin staining was markedly reduced in the mesenchyme of the pharyngeal arches (Figure 7 B,F) and frontonasal processes (Figure 7 D,H) of stage-matched RockDN embryos, but was found in the same pattern as the controls in the neural ectoderm, which is not derived from NCC (Figure 7 G,H). Moreover, there were foci of intense phalloidin staining scattered throughout the ectomesenchyme of the pharyngeal arches and frontonasal process of mutant embryos, which were rarely seen in control embryos (blue arrow in Figure 7 F,H). These foci of rounded up, intensely phalloidin-stained cells were interspersed with activated caspase 3-expressing cells (white arrow in Figure 7 F,H), with occasional phalloidin-intense cells also expressing activated caspase 3 (arrowheads in Figure 7 F,H). Thus, disruption of the actin cytoskeleton is associated with cell death in the craniofacial region.

In cell culture, vinculin localises to integrin-mediated focal adhesion complexes, linking the internal cytoskeleton of the cell to the substrate on which it sits, and also, at lower levels, in cadherin-mediated cell-cell junctions (reviewed in ^[20]). As detachment of cells from the matrix has been observed in RockDN-expressing motor neurones ^[12], we looked for evidence of disruption of cell-matrix adhesion, as a possible cause of the cell death we observe in the craniofacial region of RockDN mutants. Whereas in cultured NCC, vinculin localised to distinct foci at the periphery of the cells as expected (data not shown), this was not the case in the three dimensional environment of the developing embryo. In control embryos, vinculin was expressed strongly in discrete regions of the tissue, associated with boundaries between different cell types, such as the boundary between the NCC-derived ectomesenchyme and the neural ectoderm (Figure 7 K,O). Lower levels were also found at the periphery of individual cells, likely reflecting its association with cadherin-based cell-cell junctions. In matched RockDN embryos, vinculin was lost at the ectomesenchyme : neural ectoderm boundary but was expressed more broadly in the NCC-derived ectomesenchyme, in both the first pharyngeal arch (Figure 7 I,J,M,N) and in the frontonasal region (Figure 7 K,L,O,P). Ectopic vinculin expression was also found in the surface ectoderm of the frontonasal region and pharyngeal arches (Figure 7 J,L). Co-staining for vinculin and activated-caspase 3 showed some localisation within the same areas, suggesting a possible link between disruption of the actin cytoskeleton, cell-substrate adhesion and cell death of the RockDN-expressing NCC. Staining for paxillin, another component of focal adhesions, confirmed these observations (Figure 8 A–H). Paxillin was localised to the ectomesenchyme : neural ectoderm boundary and individual cell membranes in control embryos (Figure 8 A,C). However, staining was lost from the ectomesenchyme : neural ectoderm boundary in RockDN embryos and was observed in punctate foci in the ectomesenchyme (Figure 8 F,H) in a similar distribution to activated-caspase 3 immunoreactivity, supporting the idea that cell death was occurring at sites of abnormal focal adhesion formation. Immuno-staining for laminin confirmed that the interactions with the extracellular matrix were abnormal, as basal lamina localisation was lost from the ectomesenchyme : neural ectoderm boundary in the frontonasal processes and was abnormally distributed throughout the surface ectoderm in both the frontonasal processes and the pharyngeal arches (Figure 8 I–L and data not shown).

Rodent models of craniofacial abnormalities represent a valuable tool for studying, understanding and ultimately preventing human malformations. For example, mutations in ALX3, whose mouse counterpart has been implicated in the development of the midface ^[27] also cause frontorhiny, a form of midfacial clefting in humans ^[32]. Our study suggests an important role for Rock and other regulators of the cytoskeleton in maintaining normal craniofacial morphogenesis. In humans, frontonasal dysplasia (OMIM: 136760) is considered to be a sporadic event, supporting the idea that craniofacial malformations have a complex aetiology. Although there are no reports of mutations in ROCK1 or ROCK2 in patients with craniofacial malformations, loss of both chromosome 18q and 2p (where ROCK1 and ROCK2 reside, respectively) result in a spectrum of abnormalities that include hypertelorism, palatal defects and micrognathia and cardiac (OMIM 218340 and 601808). Thus, haploinsufficiency for these genes might contribute to the phenotype of these patients. From our data, and a review of the current literature relating to frontonasal dysplasia and midface clefting resulting from mutations in a range of different genes, we suggest that the common link is a reduction in NCC in the developing frontonasal prominences and first pharyngeal arch, in many cases related to increased NCC apoptosis. The key roles that Rho kinase plays in regulating the cytoskeleton and maintaining cell-matrix interactions suggests that it may be a common downstream effector in this cell death process.

Ethical approval of animal work carried out in this project has been authorised by the Newcastle University Ethics Committee and was covered by Project Licence PPL 60/3876 approved by the UK Home Office.

Rock dominant-negative (RockDN) mice ^[12] (BRC no. 01294) from RIKEN BioResource Center (Tsukuba, Japan) were inter-crossed with the Rosa 26R reporter line ^[33], ROSA-EYFP line ^[34], Wnt1-cre line ^[13] or the PGK-cre line ^[16]. Mice were maintained according to the Animals (Scientific Procedures) Act 1986, United Kingdom. Genotyping for RockDN positive mice was performed as described ^[12]. CD1 mice were obtained from Charles River. Stage matched mutant and control embryos were used for all experiments.

The relative quantities of the CAT gene cassette was measured by quantitative real time PCR. RNA was extracted from four E11.5 control and four mutant pharyngeal arches 1 and 2 using the Trizol reagent kit (Invitrogen) in duplicate. cDNA was produced using 1 µg of RNA with an initial DNase treatment step (Invitrogen DNase kit) followed by reverse transcription (Invitrogen SuperScript II RTase kit). Random 15-mers were used (Sigma).

Quantitative real-time PCR was performed using a 7900ht fast real-time PCR system (Applied Biosystems) and the SYBR green JumpStart Taq ReadyMix kit (Sigma). All reactions were performed in triplicate. Two housekeeping genes were used for normalisation, β-actin and GAPDH, and were found to be stably expressed in this experimental setting. Relative quantities of gene expression were calculated using the ΔΔCt method. Primer sequences were for β-actin, F: [gene: 5′-GCTGGTCGTCGACAACGGCTC-3′], R: [gene: 5′-CAAACATGATCTGGGTCATCTTTTC-3′], for GAPDH F:[gene: 5′-GCTGGTCGTCGACAACGGCTC-3′], R:[gene: 5′-CAAACATGATCTGGGTCATCTTTTC-3′] and for CAT primers see ^[12].

Heads were collected from E15.5 embryos, digital protographs were taken, and measurements were made using ImageJ software ^[35].

β-gal staining of RockDN;Rosa 26R embryos was performed as described previously ^[36]. For skeletal staining, the superficial muscle layers, eyes and internal organs were removed before fixing in Bouin's solution. E18.5 embryos were dehydrated in 95% ethanol and washed in acetone before staining for cartilage in alcian blue for eight days. Bone was stained with alizarin red. E14.5 embryos were washed with 70% ethanol/0.1% NH₄OH and cartilage stained in 0.05% alcian blue/5% acetic acid. Embryos were imaged in either glycerol or benzyl alcohol/benzylbenzoate.

Immunolabelling was performed on PFA-fixed, paraffin-embedded sections using the Rock2 antibody (Santa Cruz Biotechnology). Cryosections were labelled using antibodies raised against activated Rock1 (abcam), caspase 3 (Cell Signalling), vinculin (Sigma), paxillin (abcam), laminin (Sigma), GFP-FITC (Molecular Probes) and filamentous actin was stained with rhodamine-phalloidin (Sigma). Each experiment was repeated a minimum of three times and included appropriate controls.

The mean percentage cell death and cell proliferation was calculated from three matched sections of the first pharyngeal arch from four control and four mutant E9.5 embryos.

Statistical analysis was carried out using SPSS (IBM). The one way ANOVA test and one sample t-test were used as appropriate.