MicroRNA-17-92, a Direct Ap-2α Transcriptional Target, Modulates T-Box Factor Activity in Orofacial Clefting

Among the most common human congenital anomalies, cleft lip and palate (CL/P) affects up to 1 in 700 live births. MicroRNA (miR)s are small, non-coding RNAs that repress gene expression post-transcriptionally. The miR-17-92 cluster encodes six miRs that have been implicated in human cancers and heart development. We discovered that miR-17-92 mutant embryos had severe craniofacial phenotypes, including incompletely penetrant CL/P and mandibular hypoplasia. Embryos that were compound mutant for miR-17-92 and the related miR-106b-25 cluster had completely penetrant CL/P. Expression of Tbx1 and Tbx3, the DiGeorge/velo-cardio-facial (DGS) and Ulnar-mammary syndrome (UMS) disease genes, was expanded in miR-17-92 mutant craniofacial structures. Both Tbx1 and Tbx3 had functional miR seed sequences that mediated gene repression. Analysis of miR-17-92 regulatory regions uncovered conserved and functional AP-2α recognition elements that directed miR-17-92 expression. Together, our data indicate that miR-17-92 modulates expression of critical T-box transcriptional regulators during midface development and is itself a target of Bmp-signaling and the craniofacial pioneer factor AP-2α. Our data are the first genetic evidence that an individual miR or miR cluster is functionally important in mammalian CL/P.


Introduction
The evidence that there is a genetic component underlying CL/ P is compelling. Analysis of a Danish cohort of CL/P cases revealed that relatives of patients with CL/P have a higher relative risk for CL/P compared to background risk levels. This notion of CL/P heritability is also supported by twin studies [1], [2]. Genome wide association studies (GWAS) and mouse genetics studies have also pointed to genes and genomic regions that are associated with CL/P [3], [4].
MiRs repress gene expression post-transcriptionally by Watson-Crick base pairing to the seed sites in the 39UTR of target genes. The miR-17-92 cluster, encoding miR-17, miR-18a, miR-19a, miR-20a, miR-19b-1, and miR-92a-1, is within a region on chromosome 13q that when deletion is associated with CL/P, lung hypoplasia, microphthalmia, microcephaly, and small stature in human patients and has phenotypic similarities to Feingold syndrome [5], [6]. Moreover, miR-17-92 is found in an amplified region associated with small cell lung cancer, as well as in B-cell lymphomas, and is over-expressed in several solid tumor types, including breast, colon, lung, pancreas, and prostate cancers [7].
Our previous findings indicated that miR-17-92 is directly regulated by Bmp-signaling in heart development [9]. Bmpsignaling deficiency in mice and humans has been shown to cause CL/P and other craniofacial anomalies [11], [12]. Interestingly, miR-17-92 has also been shown to be directly regulated by Myc family transcription factors [7]. Here, we show that miR-17-92 deficiency results in orofacial clefting and that the human disease genes Tbx1 and Tbx3 are direct targets for miR-17-92. Our findings also reveal that miR-17-92 is a direct target for the master regulator of cranial neural crest development AP-2a.

miR-17-92 mutant embryos have orofacial clefting
We found that miR-17-92 (miR17-92 null/null ) mutant embryos had severe craniofacial defects including CL/P and mandibular hypoplasia with notching, revealing that miR-17-92 is a critical regulator of craniofacial development ( Figure 1A-H, Figure S1). Moreover, by genetically reducing miR-106b-25 dose on the miR-17-92 null background, the clefting phenotype was both more severe and completely penetrant, indicating that there is genetic redundancy between these two miR complexes ( Figure 1C-F, Figure S1E-H and Table S1).
In addition to cleft lip and mandible defects, both miR-17-92 mutants and miR-17-92 null ; miR-106b-25 null compound mutants had cleft secondary palate (Figure 1 B, D, H). Expression of mitotic cell marker phospho-Histone H3 (pHH3) was greatly reduced in miR-17-92 mutants, indicating that miR-17-92 is required for normal progenitor cell proliferation during orofacial development ( Figure 1I-L, Figure S2). Taken together, these data provide the first genetic evidence that miRs are important regulators of mammalian orofacial development and are involved in CL/P.

miR-17-92 is expressed in craniofacial structures
We generated a miR-17-92 bacterial artificial chromosome (BAC) transgenic LacZ reporter line to follow the expression of primary (pri)-miR-17-92 ( Figure S3A). Three individual transgenic lines showed similar LacZ expression pattern, revealing that pri-miR-17-92 was expressed in branchial arches and frontonasal process ( Figure 2A). LacZ was also detected in the nasal structures, calvarial bones, auricle, periocular mesenchyme, and limb mesenchyme ( Figure 2K). Sagittal sections on E11.5 embryos revealed LacZ activity in epithelium and mesenchyme of first branchial arch and frontonasal process ( Figure 2I). Coronal sections through E12.5 and E13.5 embryos demonstrated LacZ staining in distal tips of the palatal shelves, the mandibular mesenchyme and mesenchyme of forming frontal bones ( Figure 2J, L).
In situ with a pri-miR-17-92 probe revealed similar expression pattern as the transgenic LacZ data ( Figure 2E-F). Furthermore, in situ analysis with locked nuclei acid (LNA) probes to detect mature miR-17 and miR-92a showed that miR-17 and miR-92a were highly expressed in branchial arch and frontonasal process (Figure 2

Direct regulation of craniofacial development genes by miR-17-92
Target genes that are repressed by miR-17-92 have a mixture of miR-17/20a/106b and miR-92a/25 family seed sites in their 39UTRs ( Figure S6). Bioinformatics analysis revealed conserved miR-17/20a/ 106b family seed sequence in the 39 UTR of Fg f10, Shox2 and Osr1 ( Figure S6A-C, F). The Tbx3 39 UTR contained both a miR-17/20a/ 106b family seed site and a miR-92a/25 family seed site ( Figure S6D-E). We cloned the 39 UTRs of Fg f10, Shox2, Tbx3 and Osr1 into luc reporter plasmids to test miR seed sequence function in vitro. Transfections with miR mimics of miR-17-92 resulted in drastic reduction in luciferase activity for all of the reporter plasmids ( Figure 3G, Figure S7). Mutation of the respective miR seed sequences within 39 UTRs of those genes ablated the inhibition by the corresponding miR ( Figure 3G, Figure S7). These data suggest that miR-17-92 directly inhibits Fg f10, Shox2, Tbx3 and Osr1.
Bmps regulate miR-17-92 complex in craniofacial development and miR-17-92 overexpression suppresses orofacial clefting in Bmp mutant mice Previous work showed that conditional inactivation of Bmpr1a, Bmp4, and Bmp2;Bmp4 in developing facial processes using the Nestin cre transgenic driver result in orofacial clefting ( [11] and Figure S8). This cre driver directs cre activity in facial prominences [11]. Moreover, miR-17-92 is a direct target for Bmp-signaling in cardiac progenitors [9]. We crossed the miR-17-92 OE line into Nestin cre , Bmp4, Bmp7 conditional mutant background to test whether miR-17-92 gain-of-function could genetically rescue the defects in Bmp mutants. All Nestin Cre , Bmp4 flox/+ , Bmp7 flox/+ embryos (23 out of 23) and embryos without Nestin Cre (29 out of 29) had normal morphology ( Figure 4A, Figure   Author Summary CL/P are very common birth defects in humans. The genetic mechanism underlying CL/P pathogenesis is poorly understood. MiRs, small non-coding RNAs that function to post-transcriptionally regulate gene expression, have been identified as pivotal modulators of various developmental events and diseases. To date, there is no individual miR or miR cluster that has been identified as functionally essential in mammalian CL/P. Here, we have discovered that deletion of miR-17-92 cluster in mice results in craniofacial malformations including CL/P. Importantly, MIR-17-92 is located on a critical human chromosome region associated with 13q deletion syndrome, a chromosomal disorder that presents with defects including CL/P, suggesting the advantages of our animal model to study human disease. Moreover, our work demonstrated that miR-17-92 cluster directly repressed Tbox factors, which have critical functions during craniofacial development. We further showed that miR-17-92 was directly activated by Bmp-signaling and transcription factor AP-2a. Together, our work identified a novel miRmediated transcriptional network underlying CL/P, providing new insights into craniofacial developmental biology. S9A, D, table S2), while all Nestin Cre , Bmp4 flox/flox , Bmp7 flox/+ mutant embryos (6 out of 6) had bi-lateral cleft lip and heart defects with incompletely penetrant embryonic lethality at E12.0 likely due to heart defects ( Figure 4B, Figure S9B, E, table S2). Most (5 out of 6) Nestin Cre , Bmp4 flox/flox , Bmp7 flox/+ , miR-17-92 OE embryos were rescued by miR-17-92 overexpression (significant different compared to Nestin Cre , Bmp4 flox/flox , Bmp7 flox/+ mutants, CHI-TEST, p,0.01), with full suppression of cleft lip and heart defect caused by Bmp loss-of-function, but not eye defect ( Figure 4C, Figure S9C, F, table S2).

Discussion
We report the first miR underlying mammalian CL/P. miR-17-92 is located on human chromosome 13q31.3 in a critical region for CL/P associated with 13q deletion syndrome highlighting the importance of our findings to human disease. Our data indicated that miR-17-92 promotes proliferation in developing midface by regulating a group of progenitor genes including Tbx1 and Tbx3 that are known human disease genes ( Figure S13). Our findings reveal that timely down-regulation of progenitor genes in developing midface by miR-17-92 is critical for normal midface development.
Mir-17-92 regulates Tbx1 and Tbx3 genes in craniofacial development Tbx1 loss-and gain-of-function result in cleft palate in human DGS patients and mouse models [13], [17]- [19]. Consistent with our findings, Tbx1 gain-of-function results in cell cycle arrest [17]. DGS is characterized by highly variable phenotypes indicating that there are strong modifiers in the human genome [18], [20]. Our findings suggest miR-17-92 as a candidate genetic modifier for Tbx1 since it fine-tunes Tbx1 expression levels.
Mouse mutants for Tbx3 and the related Tbx2 have cleft palate [21]. Furthermore, human patients with UMS have abnormal and distinct facial appearance indicating a requirement for Tbx3 in human craniofacial development [22]. While our findings suggest that elevated Tbx3 inhibits proliferation, there is other evidence suggesting that Tbx3 promotes proliferation [23]. However, an in vivo study reveals that Tbx3 overexpression results in reduced cardiomyocyte proliferation in the zebrafish heart [24]. More work will be required to evaluate Tbx3 function and target genes in vivo in the context of the miR-17-92 mutant midface to better understand contextual Tbx3 function.

MiR-17-92 regulates Fgf signaling
Both Fg f10 and Fg fr null mice have cleft secondary palate [25], [26]. Mutations in Fg f10 and Fgf receptors cause lacrimoauricular-dento-digital (LADD) syndrome in human patients indicating a requirement for Fgf-signaling in human craniofacial development [27]. Fg f10 mRNA is enriched in anterior and middle regions of the secondary palate. Moreover, Fg f10 deficiency results in abnormal fusion of the palatal to oral cavity epithelium, suggesting that Fg f10 is required for maturation of palate epithelium. Importantly, elevated Fgf signaling is pathologic in human patients as shown by the extensive investigations into Fgf receptor mediated craniosynostosis [28]. Homozygosity for the Fg fr2 gain-of-function Crouzon mutation in mice results in cleft palate, as well as, craniosynostosis [29] indicating that elevated Fgf signaling also causes cleft palate. Our data demonstrate that miR-17-92 directly represses Fg f10 as a mechanism to maintain correct levels of Fg f10 during palate closure.

Micro RNAs in human orofacial development and disease
Currently, there are no other genetic loss-of-function data indicating that single miRs or miR clusters are important in mammalian orofacial clefting. Data from zebrafish indicate that miR-140 targets pdg fra to regulate primary palate development [30]. GWAS in human patients reveal important genome regions that are associated with CL/P, including 8q24 [3], [31]. Within the 8q24 region is the c-myc gene, a known miR-17-92 regulator [32], [33]. Chromosomal deletions that include miR-17-92 cause a variant of Feingold syndrome in human patients with small stature and skeletal abnormalities [6]. Human patients with hemizygous miR-17-92 deletion do not have CL/P likely reflecting phenotypic heterogeneity in miR-17-92 loss of function families. These data are consistent with our findings indicating that there is incomplete penetrance of the CL/P phenotype in miR-17-92 mutant mouse embryos (table S1).

Mir-17-92 regulation in midface development by Bmp and AP-2a
Consistent with previously finding that Bmp-deficiency results in CL/P in mice and humans [11], [12], our data indicate that Bmp signaling activates miR-17-92 in craniofacial development. Moreover, we show that AP-2a also regulates miR-17-92 expression although our transfection assays failed to uncover synergistic miR-17-92 activation by AP-2a and Bmp-signaling (not shown). One possibility is that Bmp-signaling and AP-2a activate miR-17-92 sequentially during craniofacial progenitor cell development. The assays we employed here cannot easily distinguish molecular events that occur in neighboring or closely apposed cells rather than in the same cell. We also failed to detect up-regulated miR-17-92 target genes in AP-2a mutants perhaps due to functional redundancy with other AP-2 family members [34]- [36]. Nonetheless our findings have important implications since AP-2a has been shown to regulate Irf6, a common genetic defect in syndromic and non-syndromic CL/P in human patients [37]. AP-2a regulation potentially connects miR-17-92 to a gene regulatory network that may be involved in a large portion of human CL/P. In summary, we identified a miR-mediated genetic pathway that plays critical roles during orofacial development ( Figure S13).

Ethics statement
All animal experiments detailed within the manuscript were approved by the Baylor College of Medicine review board.

Immunofluorescence
Embryos were fixed in 4% PFA, embedded in paraffin and cut to 5 mm sections mounted on Superfrost/Plus slides (Fisher Scientific). The antigens were retrieved by incubating in the citrate buffer (10 mM) for 2 minutes in microwave oven. The primary antibody was anti-Phospho-Histone H3 with 1:200 dilution (Cell Signaling). Broad HRP conjugated secondary antibody (Invitrogen) was used and visualized using TSA Plus Fluorescence Systems from PerkinElmer on a Zeiss LSM 510 Confocal Microscope. Nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI).

In situ hybridization
Tissue preparation and in situ hybridization were as previously described [39], [40]. The gene probes were synthesized using DIG RNA Labeling Kit (Roche) following manufacturer's guidelines. The enzymes used for digestion and transcription of in situ constructs are SacII and T7 for Fg f10, XhoI and T7 for Shox2 (gift from Dr. Yiping Chen's lab), EcoRI and T7 for Osr1(gift from Dr. Rulang Jiang's lab), EcoRI and T3 for Tbx1(gift from Dr. Antonio Baldini's lab), PstI and T3 for Tbx3 (gift from Dr. Robert Kelly's lab). miRCURY LNA probes were purchased from Exiqon and used per manufacturer's guidelines.

Generation of constructs
To generate 39 UTR luciferase reporter plasmids, 39 UTR genomic sequence of genes including Fg f10, Osr1, Shox2 and Tbx3 were amplified using a high-fidelity PCR system (Roche) with designed oligonucleotides and subcloned into the pMIR-RE-PORT Luciferase miRNA Expression Reporter Vector (Ambion).

Chromatin immunoprecipitation
Wild type mouse embryonic orofaces were dissected at E12.5 (for Smad1/5/8 ChIP) or E10.5 (for AP-2a ChIP) and followed by ChIP analysis as previously described [9]. As control, normal rabbit immunoglobulin G was used as a replacement for the anti-Smad1/5/8 (sc-6031 X, Santa Cruz) and 3B5 mouse monoclonal AP-2a antibody [41] to reveal nonspecific immunoprecipitation of the chromatin. The PCR products were evaluated for appropriate size on a 2% agarose gel and were confirmed by sequencing.

Real time RT-PCR
Total RNA was isolated using RNeasy Micro Kit (QIAGEN) and real-time thermal cycling was performed using StepOne Real-Time PCR Systems (Applied Biosystems). Super Script II Reverse Transcriptase (Invitrogen) was used for RT-PCR and SYBR Green JumpStart Taq ReadyMix (SIGMA) was used for real-time thermal cycling. All error bars represent SEM.

Luciferase activity assay
Plasmids used for transfection were generated as described above or previously reported [9]. LS8 cells were transfected using Lipofectamine 2000 (Invitrogen). Luciferase activity assays were measured using the luciferase Assay System (Promega).

ChIP-seq analysis
hNCC AP-2a ChIP-seq and histone modification markers ChIP-seq datasets were accessed from GEO under accession number GSE28876 [16], [42]. Raw fastq reads were mapped to hg19 genome using Bowtie2 [43]. The total number of tags of each ChIP-seq run was normalized to 10 million. ChIP-seq tracks were visualized and compared in UCSC Genome Browser. Table S1 Summary of phenotypes of miR-17-92 single and miR-17-92;miR-106b-25 compound mutant embryos. For cleft lip and palate, the penetrance and severity of the phenotype was more severe in compound mutants for both miR clusters. A ''wide mouth'' phenotype that was observed in some miR17-92 mutants represents an increased distance between the two frontonasal processes and we believe this is an intermediate phenotype between normal and cleft lip. (DOCX)