The neural crest-enriched microRNA miR-452 regulates epithelial-mesenchymal signaling in the first pharyngeal arch

The proper migration and differentiation of neural crest cells (NCCs) is essential for craniofacial, cardiac, peripheral and enteric nervous system, melanocyte and thymic development (Helms and Schneider, 2003; Jiang et al., 2000; Le Douarin et al., 2004; Lee et al., 2004a). After delaminating from the dorsal portion of the neural tube, NCCs migrate ventrolaterally along stereotypical routes and are induced to differentiate through reciprocal signaling with neighboring cells (Sauka-Spengler and Bronner-Fraser, 2008). Cranial NCCs populate the pharyngeal arches (PAs), where the neural crest-derived mesenchyme encounters a number of instructive signals from the pharyngeal epithelia (i.e. endoderm and ectoderm), resulting in differentiation into the proper cell lineages (Kameda, 2009; Le Douarin et al., 2004). Similarly, NCCs that populate the outflow tract of the heart and the developing aortic arch arteries rely on reciprocal signaling with neighboring cardiac progenitor cells derived from the second heart field (Waldo et al., 2005). Disruption of NCC development, cell- or non-cell-autonomously, results in numerous forms of human birth defects, including DiGeorge and Treacher-Collins syndromes (Epstein and Parmacek, 2005). Although many signaling pathways and transcription factors involved in NCC development are known (Meulemans and Bronner-Fraser, 2004; Trainor et al., 2002), the mechanism by which post-transcriptional regulation affects NCC development has not been established.

MicroRNAs (miRNAs) are an important class of post-transcriptional regulatory molecules. They typically bind to sequence-specific binding sites within the 3′-untranslated region (3′-UTR) of target mRNAs to repress translation, degrade the target message, or both (Bartel, 2009). The RNase III enzyme Dicer is required for the cleavage of precursor miRNAs into fully functional, mature miRNAs (Lee et al., 2004b). Studies with specific miRNA and conditional Dicer deletions have revealed that miRNAs are required for the proper development of a number of tissues, including lungs, cardiac muscle, cartilage, skin and limbs (Harris et al., 2006; van Rooij et al., 2007; Zhao et al., 2007; Kobayashi et al., 2008; Yi et al., 2009; Harfe et al., 2005). Deletion of Dicer in NCCs disrupts proper cranial NCC development (Zehir et al., 2010); however, the individual miRNAs that contribute to neural crest development and the mechanism by which they do so remain unknown.

Here, we show that disruption of miRNA biogenesis in NCCs not only affects cranial and cardiac neural crest development, but also specifically affects the expression of Dlx2 in the mandibular component of the first pharyngeal arch (PA1). We profiled miRNAs enriched in NCCs and found that one NCC-enriched miRNA, miR-452, was sufficient to rescue proper expression of Dlx2, a known PA patterning gene, in the mandibular component of PA1. Additionally, we found that miR-452 regulated reciprocal epithelial-mesenchymal signaling in PA1 involving Wnt5a, Shh and Fgf8 converging on Dlx2 expression. Thus, our study reveals a novel miRNA-regulated signaling cascade within NCCs and the pharyngeal apparatus.

Dicer^flox/flox mice (Harfe et al., 2005) and Wnt1-cre mice (Danielian et al., 1998) were intercrossed to generate Dicer^flox/flox Wnt1-cre mice. Genotyping was performed by PCR with primers: Cre1, 5′-AGGTCCGTTCACTCATGGA-3′; Cre2, 5′-TCGACCAGTTTAGTTACCC-3′; Dicer-For, 5′-ATTGTTACCAGCGCTTAGAATTCC-3′; and Dicer-Rev, 5′-GTACGTCTACAATTGTCTATG-3′. ROSA26 reporter (R26R)-YFP mice (Jackson Laboratory, Bar Harbor, ME, USA) were bred with Dicer^flox/flox mice to generate Dicer^floxP/+R26R-YFP mice.

Skeletons from embryos were stained with Alcian Blue as described (McLeod, 1980). Yellow latex cast dye (Connecticut Valley Biological Supply, South Hampton, MA, USA) was injected into the beating left ventricle of wild-type or mutant hearts with a 30 1/2 gauge needle. The hearts were dehydrated and cleared in benzyl benzoate:benzyl alcohol (2:1) to visualize the yellow latex in the vasculature. Pregnant mothers were dissected to obtain E13-14.5 wild-type and mutant embryos, which were fixed in 10% formalin and paraffin embedded. Transverse sections through the heart and brain were stained with Hematoxylin and Eosin (H&E) to analyze morphology, 1:500 rabbit anti-NF-M antibody (Abcam, Cambridge, MA, USA) to visualize neuronal tissue, 1:100 rabbit anti-GFP antibody (Sigma-Aldrich, St Louis, MO, USA) to visualize NCC progeny, and 1:500 Cy5-conjugated mouse anti-smooth muscle actin (SMA) (Sigma) to visualize smooth muscle cells. Apoptosis assays were performed using the TUNEL Assay Kit (Roche, Indianapolis, IN, USA). Proliferation studies were performed using the Phospho-Histone H3^(ser10) Assay (Millipore, Billerica, MA, USA).

mRNA in situ hybridization of whole-mount embryos was carried out as described (Riddle et al., 1993) with digoxigenin-labeled probes, which were synthesized with Digoxigenin Labeling Mix (Roche) and T7 or T3 RNA polymerase (Roche). The Msx1, Dlx2, Gli1 and Fgf8 riboprobes have been described (Thomas et al., 1998; Lee et al., 1997; Meyers and Martin, 1999). Briefly, embryos were collected at E10.5, fixed in 4% paraformaldehyde and dehydrated in 100% methanol. Before hybridization, embryos were rehydrated, treated with 10 μg/ml proteinase K (Sigma-Aldrich) for 15 minutes, and placed in prehybridization buffer for 2 hours at 70°C. Probes (0.5 μg/ml) were added and embryos were hybridized overnight at 70°C. After a series of washing steps, digoxigenin was detected with an anti-digoxigenin antibody conjugated with alkaline phosphatase (Roche). Color development was visualized with BM Purple substrate (Roche), and images were obtained with a Leica microscope.

Embryos from Wnt1^Cre or Dicer^flox/+ Wnt1^Cre mice intercrossed with R26R-YFP mice were collected at E10.5 and E11.5, dissected and trypsinized. The cells were spun at 2000 rpm (425 g) and the pellet was resuspended in PBS and filtered through a 40-μm membrane (Millipore). Selection by FACS was based on YFP expression. YFP⁺ (10,000-30,000 cells per embryo) and YFP^– (50,000-100,000 cells per embryo) cells were collected, total RNA was isolated using Trizol (Invitrogen, Carlsbad, CA, USA), and 1 μg of RNA for each cell population was used for miRNA microarray hybridizations (Exiqon, Denmark). Array data were analyzed using R/Bioconductor Bioinformatic software (Gentleman et al., 2004) and the marray package. GenePix (Axon) flagged spots were removed and only unflagged mmu-miR probes were used for normalization and subsequent analysis. Cy5/Cy3 signals were calculated for each array and normalized by loess normalization (Yang et al., 2002). Microarray data are available at Gene Expression Omnibus under accession number GSE25256.

PAs were cultured according to the modified Sanger method (Trowell, 1954). Briefly, PAs from mutant and control embryos were dissected at E9.5 and incubated with 2 U/ml dispase to dissociate the epithelial layer from the mesenchymal cells. The epithelial layer was further permeabilized with fine tungsten needles, and the PAs were placed on a nitrocellulose membrane (Millipore) supported by a wire frame. The apparatus was placed in BGJb medium (Gibco, Carlsbad, CA, USA) supplemented with 50 U/ml penicillin/streptomycin and kept at 37°C for 48 hours.

Dlx2 rescue experiments were carried out in cultured PAs co-transfected with 66 pmol 2′O-methyl oligoribonucleotide mimics (miR-452 or miR-513; Dharmacon, LaFayette, CO, USA) and 33 pmol Block-iT Alexa Fluor-red fluorescent control oligo (Invitrogen) with Lipofectamine 2000 (Invitrogen). After 48 hours, PAs were fixed in 4% paraformaldehyde and dehydrated in 100% methanol for in situ hybridization or in Trizol for quantitative (q) RT-PCR.

Affi-Gel beads (Bio-Rad) were incubated in 200 ng/μl recombinant Wnt5a protein (R&D Systems, Minneapolis, MN, USA) or bovine serum albumin (BSA) (Invitrogen) for 2 hours at 37°C. During incubation, the PA1s from E10.5 Ptch1-lacZ (Jackson Laboratory) mouse embryos were dissected and prepared for explant culture as described, excluding the dispase treatment. Beads were then transplanted into the mesenchyme with fine tungsten needles and cultured for 36 hours. The PAs were then fixed in 4% paraformaldehyde and stained for β-galactosidase (β-Gal) activity. Similar experiments were performed in wild-type embryos assaying Dlx2 mRNA expression in response to Wnt5a-soaked beads.

RNA was prepared from PAs transfected with miR-452 mimic and Block-iT Alexa Fluor-red (Invitrogen) or Block-iT alone, using Trizol. For miRNA qRT-PCR, cDNA was reverse transcribed from 10 ng total RNA using the TaqMan microRNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA). miR-16 and U6 small nuclear RNA served as endogenous controls.

Wild-type FVB mice (Jackson Laboratory) were mated and plugs were identified; noon of the day of plug discovery was considered E0.5. At E8.5, pregnant mothers were anesthetized with isofluorane. A small incision was made in the abdomen and the uterus was carefully pulled through the incision to reveal the ovaries. A NanoFil syringe (World Precision Instruments) fitted with a 35-gauge beveled needle was filled with PBS (control) or 3 mg/ml miR-452 antagomir (Dharmacon) diluted in PBS. Using a microsyringe pump (World Precision Instruments), 3 μl was injected into the embryonic space through the intact deciduum of each embryo. After all embryos had been injected, the uterus was carefully placed back into the pregnant mother and the incision was closed with sutures. The embryos were allowed to develop to E11.5 or E16.5 and harvested for in situ hybridization, qRT-PCR or histological analysis.

Joma1.3 cells were cultured as described (Maurer et al., 2007). Briefly, cells were plated on cell culture dishes coated with 1 mg/ml fibronectin (Roche) in a 1:1 mixture of Dulbecco's modified Eagle's medium and F12 medium (Gibco) containing 1% N2 supplement (Invitrogen), 2% B27 supplement (Invitrogen), 10 ng/ml epidermal growth factor (Invitrogen), 1 ng/ml fibroblast growth factor (Invitrogen), 100 U/ml penicillin/streptomycin (Invitrogen) and 10% chick embryo extract (Stemple and Anderson, 1992). To maintain cells in an undifferentiated state, 200 nM 4-OHT (Sigma-Aldrich) was added fresh to the medium.

Luciferase assays to determine putative targets were performed with undifferentiated Joma1.3 cells. A ∼500 bp fragment containing the predicted miR-452 binding site for each predicted mRNA target was cloned into pMIR-Report luciferase reporter vectors (Applied Biosystems). For the luciferase assays for the mutant Wnt5a binding site, oligos for 50 bp surrounding the wild-type Wnt5a binding site (mouse chromosome 14: 29338052-74) or mutant binding site (corresponding to bases 3-6 in the seed sequence of miR-452, mutated from CAA to TCG) were annealed together to form concatamers consisting of five consecutive binding sites. These were then cloned into pMIR-Report luciferase reporter vectors. All assays were performed in triplicate in 12-well plates with Joma1.3 cells transfected with Lipofectamine 2000. After 24 hours in culture, cells were harvested and luciferase intensity was measured with the Luciferase Dual-Reporter Assay (Promega) and normalized to Renilla luciferase.

For Shh responsiveness assays, hsa-Wnt5a cDNA (Open Biosystems, Hunstville, AL, USA) was cloned into pEF6-V5 expression vectors (Invitrogen). Joma1.3 cells were transfected with Lipofectamine 2000. After 24 hours in culture, 0.5 μg/ml recombinant Shh-N protein (R&D Systems) was added to the medium and the cells were cultured for 24 hours. Cells were then harvested, RNA isolated and qRT-PCR performed to determine relative gene expression levels.

Western blots were performed using lysate from Joma1.3 cells or PA explant culture with or without transfection of 66 pmol miR-452 mimic and immunoblotted with goat anti-Wnt5a antibody (R&D Systems) and quantified using the LI-COR Odyssey Imaging System (LI-COR Biosciences, Lincoln, NE, USA).

To determine whether miRNA biogenesis is required for the proper development of NCCs, we disrupted the miRNA-processing enzyme Dicer (Dicer1 – Mouse Genome informatics) with a Dicer allele in which the exon encoding the second RNase III domains was flanked by loxP sites (Harfe et al., 2005). We crossed Wnt1^Cre transgenic mice, in which Cre recombinase is expressed in pre-migratory NCC progenitors and progeny at embryonic day (E) 8.5 (Danielian et al., 1998), with Dicer^flox/flox mice. Embryos lacking Dicer in NCCs had severe defects in the development of NCC-derived tissues and died soon after E16.5. In mutants, severe craniofacial defects were apparent by E14.5 (Fig. 1A,F). The NCC-derived maxillary and mandibular regions of the face and the frontonasal process lacked cartilaginous tissue (Fig. 1B,G). However, the presence of mesodermally derived cartilage near the base of the skull and non-NCC-derived tissue in the head (Fig. 1G, dashed lines) suggested that the defects were mainly restricted to NCC-derived cartilages at E14.5. We also observed loss of NCC-derived neuronal tissue from the dorsal root ganglia (DRG) and thoracic sympathetic ganglia (TSG) in Dicer^flox/flox Wnt1^Cre mutants by staining with H&E or the pan-neuronal marker neurofilament-M (NF-M; Nefm – Mouse Genome informatics) (Fig. 1C,H and see Fig. S1A,B in the supplementary material).

In addition to defective craniofacial and neuronal development, we also observed defects in the development of other NCC-derived tissues. In some cases, the outflow tract did not fully septate into a pulmonary artery and aorta, resulting in the persistence of a common outflow vessel (truncus arteriosus) (Fig. 1D,I). In humans, persistent truncus arteriosus is a consequence of NCC defects and is often accompanied by a ventricular septal defect and patterning defects of the aortic arch (Srivastava, 2006). Dicer^flox/flox Wnt1^Cre mice also had a ventricular septal defect (Fig. 1E,J-L, arrowhead) and a discontinuance of the ascending aortic arch with the descending aorta, reflecting improper patterning of the aortic arch due to a left fourth aortic arch defect (Fig. 1O,X, asterisk). Furthermore, thymus development, which is also dependent on NCCs, was absent in Dicer^flox/flox Wnt1^Cre mutants (Fig. 1M,O) and delayed in Dicer heterozygous embryos (Dicer^flox/+ Wnt1^Cre) at E14.5 (Fig. 1M,N). The haploinsufficiency of Dicer was transient, as the thymus was morphologically indistinguishable in wild-type and Dicer heterozygous embryos by E16.5. These diverse anomalies suggest that proper miRNA-mediated post-transcriptional regulation of gene expression is required for the development of NCC-derived tissues.

The severe defects in Dicer mutants might be caused by a failure of NCCs to delaminate from the neural tube and/or migrate properly. To address this, we crossed a Cre-dependent reporter mouse line, R26R-YFP (Soriano, 1999), which marks all progeny of Cre-expressing cells with yellow fluorescent protein (YFP), into the Dicer^flox/flox Wnt1^Cre background. We first investigated the early migratory behavior of NCCs at E8.5 and found no gross defect in the ability of mutant NCCs to delaminate and migrate (Fig. 1P,U). Additionally, NCC migration was grossly unaffected at E9.5 (Fig. 1Q,V and see Fig. S2A,C in the supplementary material) and at E10.5 in mutants, with streams of YFP⁺ NCCs migrating towards the PAs (see Fig. S2B,D in the supplementary material, arrowheads) and normal occupancy of YFP⁺ NCCs in the PAs. YFP was expressed in the craniofacial region of E14.5 mutants, albeit at lower intensity than in wild-type littermates, suggesting that NCC progeny condensed in the proper locations (Fig. 1R,W). YFP⁺ cells also lined the aorta and common carotid arteries of mutants (Fig. 1S,X), although the transverse aorta was absent, consistent with loss of the left fourth aortic arch artery. Although we detected YFP⁺ cells in the outflow tract, these cells did not differentiate into smooth muscle in the wall of the ductus arteriosus (Fig. 1T,Y, arrows), which provides a necessary vascular connection between the aorta and pulmonary artery in the fetus. These results suggest that abnormal NCC patterning, differentiation or maintenance, rather than delamination or migration, causes the defects seen in Dicer mutant NCC-derived tissues.

Next, we investigated whether patterning or maintenance of the bilaterally symmetric PAs was defective in Dicer^flox/flox Wnt1^Cre mutants. Morphologically, the emergence, size and shape of the PAs patterned along the anterior-posterior axis were indistinguishable in mutant and wild-type embryos at E10.5. However, expression of the distal-less homeobox gene (Dlx2), a regulator of NCC patterning in PA1 (Qiu et al., 1995), was downregulated in the mandibular and, less severely, in the maxillary portions of PA1 in mutants (Fig. 2A,D). The mandibular component of PA1 contributes to the lower jaw, whereas the maxillary region of PA1 populates the palatal region and other craniofacial bones. Other markers of NCC-derived PA mesenchyme, such as Msx1 (Satokata and Maas, 1994), were not diminished (Fig. 2B,E). By contrast, expression of fibroblast growth factor 8 (Fgf8), which is required for survival of NCC-derived PA mesenchyme (Macatee et al., 2003; Trumpp et al., 1999), was also greatly downregulated in the adjacent mandibular ectoderm of PA1 in mutants (Fig. 2C,F).

To determine whether loss of Dlx2 and Fgf8 expression in PA1 affected NCC maintenance, we assessed cell proliferation with an anti-phospho-histone H3 antibody and apoptosis using the TUNEL assay. At E10.5, no statistically significant difference between mutant and wild-type embryos was detected by either assay (Fig. 2G,K and 2H,L; quantified in Fig. 2O,P). Concordantly, at E11.5 proliferation remained unchanged in wild-type and mutant embryos (Fig. 2I,M; quantified in Fig. 2Q); however, apoptosis increased dramatically in the PA region of the mutants (Fig. 2J,N; quantified in Fig. 2R). Thus, loss of Dicer activity in NCCs disrupted normal PA1 gene expression and the survival of NCC-derived mesenchyme.

We sought to identify specific miRNAs that when disrupted could contribute to the gene expression changes in Dicer^flox/flox Wnt1^Cre mutants. We searched for miRNAs enriched in normal NCC-derived PA mesenchyme. Cells from the frontonasal process and PAs of Wnt1^Cre R26R^YFP embryos were sorted into YFP⁺ (NCC) and YFP^– (non-NCC) populations at E10.5 and E11.5 and analyzed by miRNA microarrays. The NCC population at E10.5 was enriched for a single miRNA, miR-452, and for eight additional miRNAs at E11.5 (Fig. 3A). qRT-PCR revealed a 7-fold enrichment of miR-452 in E10.5 YFP⁺ (NCC) compared with YFP^– (non-NCC) populations, confirming our microarray data (Fig. 3B). Interestingly, a few miRNAs were downregulated in E10.5 NCCs heterozygous for Dicer (Dicer^flox/+ Wnt1^Cre R26R^YFP), of which miR-452 was the most downregulated (Fig. 3C). Thus, miR-452 was abundant in the NCC population and was most sensitive to the protein dosage of Dicer.

To determine the relative levels of miR-452 within PA1 compared with other regions of the embryo, we harvested E10.5 wild-type or Dicer^flox/flox Wnt1^Cre mutant embryos and dissected them into PA1, midbrain region of the head, heart, limb buds, and tail to measure miR-452 levels by qRT-PCR. The qRT-PCR data revealed that miR-452 was most highly expressed in PA1 and the tail, with lower levels within the head and the limb buds, and was nearly undetectable in the heart (Fig. 3D). We observed a 60% decrease in the levels of miR-452 in PA1 of Dicer^flox/flox Wnt1^Cre embryos, but not other regions of the embryo, suggesting significant expression within the NCC-derived mesenchyme of PA1 (Fig. 3D).

The downregulation of Dlx2 in PA1 of Dicer^flox/flox Wnt1^Cre mutant embryos provided an important assay to determine the contribution of specific miRNAs to Dlx2 expression in the PA. We developed a Lipofectamine-based transfection approach to efficiently introduce individual mature miRNA locked nucleic acid (LNA) mimics into cultured PAs of E9.5 Dicer^flox/flox Wnt1^Cre mutant embryos to test for rescue of Dlx2 expression (Fig. 4A-C). In situ hybridization on cultured PA1 from wild type and Dicer^flox/flox Wnt1^Cre mutants revealed that Dicer^flox/flox Wnt1^Cre mutant PA1 exhibited Dlx2 downregulation in culture, similar to that observed in vivo (Fig. 4D,G). We introduced mimics for five of the nine NCC-enriched miRNAs. Only miR-452 consistently rescued Dlx2 expression in Dicer^flox/flox Wnt1^Cre mutant PAs (Fig. 4J and data not shown; representative PA1s are shown in Fig. 4D-I).

We used antagomirs (Krutzfeldt et al., 2005) targeted against miR-452 to determine whether loss of miR-452 activity affects Dlx2 expression in vivo. Antagomirs are cholesterol-modified antisense oligonucleotides that bind to and inhibit the function of endogenous miRNAs by blocking their incorporation into the RNA-induced silencing complex. Because antagomirs do not cross the placental barrier, we injected miR-452 antagomirs directly into the embryonic sac of wild-type embryos in utero. To validate this delivery method, we injected a fluorophore-modified antagomir into the embryonic sac of E8.5 embryos and observed near ubiquitous uptake of the antagomir throughout the embryos at E10.5 (Fig. 4K,L). After confirming the delivery method, we next injected miR-452 antagomir in utero at E8.5. At E11.5, miR-452 knockdown was variable and bimodal as observed by qRT-PCR (Fig. 4M). We found that this method of in vivo knockdown resulted in ∼35% of embryos with efficient (>70%; n=9; mean=86%; median=94%) miR-452 knockdown and these embryos were used for subsequent experiments; the remainder were considered failed knockdowns.

In embryos with efficient miR-452 knockdown, Dlx2 expression in the distal mandibular portion of PA1 in E11.5 embryos was significantly reduced, similar to that in Dicer^flox/flox Wnt1^Cre mutants (Fig. 4N,O). Maxillary expression of Dlx2 was only slightly affected, consistent with what was observed in Dicer^flox/flox Wnt1^Cre mutants. At E16.5, some craniofacial structures of miR-452 antagomir-injected embryos were hypoplastic (Fig. 4P-S), particularly the ala orbitalis, palatal process of palatine and anterolateral process of ala temporalis (Fig. 4Q,S). Interestingly, some mesoderm-derived tissues, such as the parietal bone, were also affected in the antagomir-injected embryos. This finding implies that a non-NCC function of miR-452 or a non-cell-autonomous effect in the NCCs results in the disruption of mesodermal bone development. Not surprisingly, the defects resulting from miR-452 knockdown were less severe than those upon complete loss of miRNA biogenesis, reflecting the function of other miRNAs in craniofacial development or the incomplete loss of miR-452. Nevertheless, these results indicate that miR-452 is involved in proper Dlx2 expression in NCC-derived mesenchyme and also for a subset of subsequent craniofacial development.

To understand the mechanism by which miR-452 regulates Dlx2 expression, we sought to identify the direct target(s) of miR-452 in NCCs and their derivatives. Putative targets were identified with an in-house miRNA:mRNA algorithm that considers sequence specificity, binding site accessibility and evolutionary conservation of the binding site, and potential sites were then tested experimentally (Cordes et al., 2009). The neural crest stem cell (NCSC) line Joma1.3 (Maurer et al., 2007) was transfected with the miR-452 mimic along with constitutively active luciferase reporter genes containing the 3′-UTR of putative targets. Repression of luciferase activity was only observed in the presence of the Wnt5a 3′-UTR (Fig. 5A). Wnt5a, a non-canonical member of the Wnt family of signaling molecules, is expressed in NCC-derived mesenchyme of the PAs (Gavin et al., 1990), and its 3′-UTR contains a miR-452 binding site (Fig. 5B). Mutating the seed region of the miR-452 binding site within the Wnt5a 3′-UTR led to loss of the miR-452-dependent decrease in luciferase activity, suggesting that the activity of the predicted binding site depends on miR-452 binding (Fig. 5B). In addition to changes in luciferase activity, western blotting and qRT-PCR revealed reduced Wnt5a protein, but not mRNA, levels in NCSCs transfected with miR-452 mimics (Fig. 5C,D). We also observed a decrease in Wnt5a protein levels in our PA culture system upon transfection of miR-452 mimic (Fig. 5C). Furthermore, co-transfection of miR-452 inhibitors into NCSCs containing a luciferase vector with the Wnt5a 3′-UTR resulted in a dose-dependent increase in luciferase activity (Fig. 5E). These results suggest that endogenous Wnt5a protein levels are sensitive to miR-452 overexpression in vitro and in vivo.

In various developmental roles, Wnt5a acts as a signaling mediator of epithelial-mesenchymal interactions. For example, Wnt5a expression in lung epithelium negatively regulates Shh responsiveness of the underlying mesenchyme (Li et al., 2005). In PAs, Wnt5a is primarily expressed in NCC-derived mesenchyme (Gavin et al., 1990), and whole-mount in situ hybridization revealed higher levels in the mandibular component of PA1 (Fig. 6A). By contrast, Shh is produced in the pharyngeal endoderm, where it diffuses into the NCC-derived mesenchyme and activates downstream signaling cascades (Yamaguchi et al., 1999; Washington Smoak et al., 2005). Shh is required for proper development of PAs and, by positively regulating Fgf8 expression in the pharyngeal ectoderm, supports Dlx2 expression in the NCC-derived mesenchyme (Haworth et al., 2007; Yamagishi et al., 2006; Thomas et al., 1998).

To determine whether Wnt5a inhibits the responsiveness of NCSCs to the Shh stimulus, we transfected NCSCs with a human WNT5A expression plasmid, then 24 hours later treated cells with recombinant Shh protein and measured the expression levels of two Shh-responsive genes, Gli1 and Ptch1. Overexpression of Wnt5a blunted the increase in Gli1 and Ptch1 mRNA expression in response to Shh treatment (Fig. 6B). To test this in vivo, we implanted beads soaked in recombinant Wnt5a or BSA (control) into PA1 of mice harboring a lacZ gene in the Ptch1 locus (Goodrich et al., 1997) as a reporter of Shh signaling. β-Gal expression was decreased circumferentially around the Wnt5a bead, but not around the control bead placed in the contralateral PA (Fig. 6C). Next, we asked whether increased Wnt5a protein levels were sufficient to result in decreased Dlx2 expression, as observed in the Dicer mutants and miR-452 knockdown experiments. Again, Wnt5a- and BSA-soaked beads were implanted into the mandibular component of cultured PA1 of wild-type E10.5 embryos. In-situ hybridization to assay Dlx2 expression revealed a marked decrease in Dlx2 expression specifically surrounding the Wnt5a-soaked beads, whereas Dlx2 expression was unaffected surrounding the contralateral BSA-soaked beads (Fig. 6D). Thus, increased Wnt5a protein levels in the mandibular component of PA1 led to a decrease in Shh signaling and Dlx2 expression.

We investigated whether miR-452 promotes Dlx2 expression in part by repressing the translation of Wnt5a, allowing efficient Shh signaling to NCCs. Shh positively regulates expression of Fgf8 in the pharyngeal ectoderm, and Fgf8 expression is required for Dlx2 expression in the NCC-derived mesenchyme (Haworth et al., 2007; Thomas et al., 2000). Thus, we hypothesized that loss of miR-452 would decrease endogenous Shh signaling and Fgf8 expression within the endoderm and ectoderm, respectively, of PA1. Indeed, expression of Gli1 was decreased in the mandibular component of PA1 in miR-452 antagomir-injected embryos at E11.5 and expression of Fgf8 was markedly reduced in the mandibular ectoderm, with less of a change in the maxillary region (Fig. 6E-H). The effects of miR-452 knockdown were consistent with this miRNA participating in the reduction in mandibular Fgf8 and Dlx2 expression in the Dicer^flox/flox Wnt1^Cre mouse and the enrichment of mandibular Wnt5a expression. These data suggest that loss of miR-452, resulting in an increase in Wnt5a protein, leads to decreases in Shh signaling and Fgf8 expression, which in turn result in the decrease in Dlx2 expression seen in both Dicer mutant and miR-452 antagomir-injected embryos.

Here we show that miRNA biogenesis is essential for the proper development of NCC-derived tissues, including those that contribute to craniofacial, cardiovascular, thymic and nervous system structures. More importantly, we found that miR-452 was the most enriched miRNA in early NCCs and regulated epithelial-mesenchymal interactions in the mandibular region of PA1 by directly targeting Wnt5a in the NCC-derived mesenchyme. Our findings suggest that miR-452 negatively regulates secretion of Wnt5a from the NCC-derived mesenchyme of PA1, and that Wnt5a normally negatively regulates Shh signaling to the ectoderm (Fig. 7). Since Shh produced in the endoderm activates ectodermal expression of Fgf8 (Haworth et al., 2007) and secreted Fgf8 functions to promote Dlx2 expression in the neighboring NCC-derived mesenchyme (Thomas et al., 2000), miR-452 might regulate Dlx2 positively in part through this complex epithelial-mesenchymal interaction. Of course, miR-452 and other NCC-enriched miRNAs are likely to target other mRNAs in the NCC-derived mesenchyme of PA1 that further regulate the epithelial-mesenchymal interactions required for proper PA development.

In the present work, we did not address which specific NCC-enriched miRNAs function in the other arches that give rise to the thymus (PA3) or the cardiac outflow tract and aortic arch arteries (PA3, PA4 and PA6), all of which were affected in the Dicer mutants. One or more of the other eight NCC-enriched miRNAs is likely to contribute to the development of these structures in a stage-specific manner. Notably, the thymus and cardiac outflow tract also require cross-tissue interactions for their development and maturation (Gordon et al., 2010; Olson, 2006). For example, reciprocal signaling between the second heart field and cardiac NCCs is required for proper patterning of the aortic arch arteries (Waldo et al., 2005). Deletion of Notch signaling specifically in the second heart field leads to altered NCC behavior, resulting in a decrease of NCC-derived tissue in the outflow tract and improper patterning of the arch arteries (High et al., 2009). Other NCC-enriched miRNAs probably regulate similar pathways to coordinate the complex tissue-tissue interactions required to pattern the outflow tract and aortic arch arteries.

Interestingly, the phenotype of the Dicer^flox/flox Wnt1^Cre mutant mice has similarities to the human disease DiGeorge syndrome (DGS). DGS patients typically present with craniofacial, cardiovascular and thymic anomalies arising from a 3-Mb deletion of chromosome 22q11 (Lindsay, 2001). In that region, Tbx1 has been strongly implicated in the etiology of DGS. However, Tbx1 is not expressed in NCCs; rather, it is expressed in the pharyngeal ectoderm, endoderm and early second heart field progenitors (Merscher et al., 2001; Yamagishi et al., 2003). Tbx1 is downstream of Shh in the pharyngeal endoderm and regulates the expression of Fgf8, resulting in a secondary effect on the adjacent neural crest-derived mesenchyme that expresses Fgf receptors (Arnold et al., 2006; Garg et al., 2001; Hu et al., 2004; Zhang et al., 2005). Another intriguing gene in the DGS-deleted region is DGCR8, which is required for miRNA biogenesis (Gregory et al., 2004), and thus suggests a possible contribution of miRNA dysregulation in DGS. Our finding that Dicer^flox/+ Wnt1^Cre heterozygous embryos have hypoplastic thymus primordia indicates that a hemizygous deletion of members of the miRNA biogenesis pathway in NCCs might result in predisposition to some DGS-like phenotypes.

Disruption of cross-tissue signaling cascades has been found in a number of human neural crest disorders (Lindsay et al., 2001; Ornitz and Marie, 2002). The results presented here suggest a crucial role for miRNAs in titrating such reciprocal signaling during NCC development and might have implications for the potential contribution of miRNA dysregulation in human neural crest disorders.

We thank B. Harfe and M. McManus for Dicer^flox/flox mice; J. Maurer for the JoMa neural crest stem cell line; J. Reiter for Gli1 in situ hybridization probes; Y. Huang for performing antagomir injections; J. Fish and C. Miller for histopathology support; G. Howard and S. Ordway for scientific editing; B. Taylor for manuscript preparation; and members of the D.S. lab for helpful discussion. D.S. was supported by grants from the NHLBI/NIH and the California Institute for Regenerative Medicine (C.I.R.M.). This work was also supported by NIH/NCRR grant C06 RR018928 to the Gladstone Institutes. Deposited in PMC for release after 12 months.