Identification of a Face Enhancer Reveals Direct Regulation of LIM Homeobox 8 (Lhx8) by Wingless-Int (WNT)/β-Catenin Signaling^,

Background: Lhx8 is an important gene for craniofacial development, but its regulation had been poorly understood.

Results: We identified a face enhancer of Lhx8 and discovered that it was regulated by WNT/β-catenin signaling.

Conclusion: WNT/β-catenin pathway directly regulates Lhx8 in the face via a distal enhancer.

Significance: We uncovered a molecular mechanism for the regulation of Lhx8 in the face.

Abstract

Development of the mammalian face requires a large number of genes that are expressed with spatio-temporal specificity, and transcriptional regulation mediated by enhancers plays a key role in the precise control of gene expression. Using chromatin immunoprecipitation for a histone marker of active enhancers, we generated a genome-wide map of candidate enhancers from the maxillary arch (primordium for the upper jaw) of mouse embryos. Furthermore, we confirmed multiple novel craniofacial enhancers near the genes implicated in human palate defects through functional assays. We characterized in detail one of the enhancers (Lhx8_enh1) located upstream of Lhx8, a key regulatory gene for craniofacial development. Lhx8_enh1 contained an evolutionarily conserved binding site for lymphoid enhancer factor/T-cell factor family proteins, which mediate the transcriptional regulation by the WNT/β-catenin signaling pathway. We demonstrated in vitro that WNT/β-catenin signaling was indeed essential for the expression of Lhx8 in the maxillary arch cells and that Lhx8_enh1 was a direct target of the WNT/β-catenin pathway. Together, we uncovered a molecular mechanism for the regulation of Lhx8, and we provided valuable resources for further investigation into the gene regulatory network of craniofacial development.

Introduction

Development of the face begins when cranial neural crest cells, which are a group of multipotent migratory cells, arise at the dorsal margin of the brain shortly after gastrulation (around embryonic day (E)² 8 in mice; mouse gestation is 19 days) (1–3). Around E8.5 to E10.5, cranial neural crest cells migrate to the ventral side of the head to make most of the mesenchyme in the embryonic facial primordia, i.e. the frontonasal prominence and the first pharyngeal arch (PA1) (4–7). Subsequently, through growth and morphogenesis of the tissue and cellular differentiation, frontonasal prominence develops into the middle and upper face, whereas PA1 gives rise to the jaw, lateral skull, and the middle ear (8–10). PA1 is further divided into the maxillary arch, the prospective upper jaw in the proximal half, and the mandibular arch, the prospective lower jaw, in the distal half (11).

A large number of genes are involved in patterning of the facial primordia to make various parts of the face (8–10, 12). The epithelium of frontonasal prominence and PA1 expresses genes encoding secreted signaling molecules such as the FGF, HH, WNT, and BMP families (13–16). These molecules signal to the mesenchyme, which in response expresses transcription factor genes such as Msx, Pax, Dlx, Prrx, Osr, Shox, and Lhx family members (17–23). Different combinations of these transcription factors are thought to provide region-specific developmental programs in the facial primordia.

Among the transcription factors crucial for craniofacial development is a LIM domain homeodomain transcription factor LHX8 (previously known as LHX7, L3) (17, 24). Lhx8 and its close homolog Lhx6 are expressed in the oral mesenchyme of PA1 from ∼E9.5 at least until birth (17, 25). In addition, a targeted mutation of Lhx8 in mice caused cleft palate in ∼60% of the animals (25). Although Lhx6 mutant mice had only moderate defects in the palate and teeth (26, 27), simultaneous inactivation of Lhx6 and Lhx8 resulted in profound craniofacial phenotypes, i.e. fully penetrant cleft palate, loss of molars, and duplication of upper incisors (28). These findings from mice indicated that the two Lhx genes are key regulators of craniofacial development with significantly overlapping functions. Furthermore, LHX8 has been implicated in human cleft palate and cleft lip conditions (29–31).

Given the functional importance of Lhx8, understanding what regulates its expression in PA1 can provide significant insights into the gene regulatory network of craniofacial development. To elucidate the molecular mechanism of tissue-specific expression of a gene, it is crucial to identify enhancer(s) that activate the transcription of the gene in that tissue. Enhancers can be difficult to find in the genome because they do not have distinctive signatures in the nucleotide sequence, and they can be hundreds of kilobases away from the promoter that they regulate (32–34). However, enhancers tend to be more conserved during evolution than other noncoding regions of the genome, and thus comparing the sequences of multiple species has helped to find enhancers (35, 36). In addition, active enhancers are associated with several protein markers, such as a histone acetyltransferase P300 and certain post-translational modifications of histone H3 (acetylation at lysine 27, H3K27ac; mono-methylation at lysine 4, H3K4me1) (32, 34, 37, 38). With antibodies against these markers, chromatin immunoprecipitation followed by microarray analysis (ChIP-chip) or high throughput sequencing (ChIP-seq) has been highly successful in identifying enhancers from various cells, tissues, and organs (39–46).

In this study, we aimed to elucidate upstream factors that regulate the expression of Lhx8 in the developing face. To this end, we identified a face enhancer of Lhx8 through combined methods of ChIP-seq, comparative genomics, and in vitro and in vivo functional assays. Further analysis of this enhancer revealed that Lhx8 is a direct target of the WNT/β-catenin pathway, uncovering a novel relationship between the two important regulators of craniofacial development.

EXPERIMENTAL PROCEDURES

Animals

The Bat-gal mice (47) were purchased from The Jackson Laboratory, and CD-1 wild type mice were purchased from Charles River Laboratories. All the experiments using mice were performed following institutional, local, and national animal welfare laws, guidelines, and policies.

ChIP-Seq

Maxillary arches were dissected from 47 of E11.5 CD-1 wild type embryos in phosphate-buffered saline (PBS) and snap-frozen on dry ice. The tissue was sent to Active Motif, Inc. (Carlsbad, CA), for HistonePath^TM service, which included chromatin preparation, ChIP-seq for H3K27ac, and bioinformatics analysis of the sequencing results to identify H3K27ac-enriched genomic regions (peaks). 20 μg of the maxillary arch chromatin was used for ChIP with a rabbit polyclonal anti-H3K27ac antibody (Active Motif, catalog no. AM 39133). Libraries for Illumina sequencing (Hi-Seq) were prepared from ChIP DNA and input chromatin as described (48), and the sequence reads from each sample were aligned to the mouse genome (NCBI Build 37, mm9) using BWA algorithm (49). The aligned sequence tags were extended in silico at the 3′-ends into 150-bp fragments, and the histograms of the fragment density along the genome were stored in binary analysis results files and visualized using the Integrated Genome Browser (50). H3K27ac peaks were identified using SICER peak-finding algorithm (51) with the following parameters: window size = 200 bp, fragment size = 150 bp, gap size = 600 bp, E value = 1000, false discovery rate = 1E-10. The coordinates of the all the peaks are listed in supplemental Table S1. The primary data have been deposited in the Gene Expression Omnibus database as binary analysis results files.

Statistical Analyses of ChIP-seq Peaks

Genomic Regions Enrichment of Annotations Tool (GREAT) Version 2.0.2 (52) was used to determine the distance between the H3K27ac peaks and the nearest TSS and to analyze the gene expression terms associated with the peaks (association rule = single nearest gene, maximum extension 1 Mb).

To compare the distribution of the H3K27ac peaks around the maxillary arch genes and around random genes, the list of 474 mouse maxillary arch genes (supplemental Table S3) was generated by searching gene expression database (53) of Mouse Genome Informatics for the genes expressed in the “maxillary arch” at Theiler stages 16–20 (equivalent to E9.5 to E13.0). As a control, 10 sets of randomly selected 474 genes were generated from UCSC RefSeq database, using Java random number generator with a linear congruential generator algorithm. For each set of 474 genes, the total number of the H3K27ac peaks within a given distance from TSS was determined using a custom program, and it was converted into a noncumulative count by subtracting the count at the preceding data point. This number was further divided by the number of the genes (474) and the length of the interval, to obtain the average number of the H3K27ac peaks per gene/kb (= peak density) along various distances from TSS. The peak densities from the 10 control gene sets were averaged, and the 95% confidence intervals were calculated using the formula in Microsoft Excel. The Z score was calculated by Z = (x − μ)/σ, in which x is the peak density from the maxillary arch gene set; μ is the average peak density from the control gene sets, and σ is the standard deviation of the peak densities from the control gene sets. Fold enrichment of the peak density was calculated by dividing the value from the maxillary arch gene set with the average value from the control sets.

To compare the degree of H3K27ac enrichment between the maxillary arch-positive and the maxillary arch-negative enhancers in supplemental Table S4, we used the fragment densities recorded in the binary analysis results files. The ChIP-seq signal for each enhancer was defined as the maximum fragment density within the region from the ChIP DNA, minus the maximum fragment density within the same region from the input DNA.

Culture of Primary Maxillary Arch Cells (PMAC) and C3H10T1/2 Cells

PMAC was prepared following a published protocol (54) with minor modifications. The maxillary arches were dissected from E10.5 or E11.5 CD-1 wild type embryos in ice-cold PBS, incubated in 0.05% trypsin/EDTA (Invitrogen) at 37 °C for 20–45 min/ and then further disrupted by pipetting into single-cell suspension. The dissociated cells were plated and cultured in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 20% newborn calf serum (HyClone), nonessential amino acids (Invitrogen), β-mercaptoethanol, and antibiotic-antimycotic (Invitrogen). For PMAC cultures without the epithelium, the dissected maxillary arches were treated with Dispase II (1 unit/ml) at 37 °C for 30 min, and the epithelium was separated from the mesenchyme using a tungsten needle before the trypsinization and dissociation into single cells. C3H10T1/2 cells were purchased from American Type Culture Collection (CCL-226) and cultured in the same medium as PMAC.

Luciferase Reporter Assays in PMAC and C3H10T1/2

All the enhancers and the 1.5-kb Lhx8 promoter were obtained by PCR from the tail genomic DNA of CD-1 wild type mice (see supplemental Table S6 for the sequences of all the PCR primers used in this study) and cloned into pGL4.23 vector (Promega), which contains the coding sequence of firefly luciferase after a minimal promotor. The LEF/TCF-binding site in Lhx8_enh1 was disrupted by PCR using primers with nucleotide substitutions. All the inserts were verified by sequencing. PMAC from E11.5 CD-1 wild type embryos or C3H10T1/2 cells were transfected with an enhancer-pGL4.23 construct and pGL4.73 (Promega) using FuGENE 6 (Promega). pGL4.73 expresses Renilla luciferase and serves as an internal control for transfection efficiency. After 48 h, the cells were lysed and analyzed by the Dual-Luciferase^TM reporter assay system (Promega). The ratio between the firefly luciferase and Renilla luciferase was calculated (= relative luciferase activity) and presented as a fold change from the relative luciferase activity from the minimal promoter control (pGL4.23).

In Vivo Transgenic Tests of Enhancer Activity

Lhx8_enh1, Lhx8_enh2, and Ror2_enh were amplified from mouse tail genomic DNA and cloned into Hsp68-LacZ plasmid (55). All the subsequent steps, including purification of the plasmid, pronuclear injection, and β-galactosidase staining of the embryos, were performed by Cyagen Biosciences, Inc. (Santa Clara, CA) following standard protocols.

Whole Mount and Section RNA in Situ Hybridization, β-Galactosidase Staining of Bat-Gal Embryos

The embryos were processed as described previously (26, 56).

RNA Extraction, Reverse Transcription, and Quantitative Real Time PCR (RT-qPCR)

Total RNA was extracted from PMAC or C3H10T1/2 using the RNeasy mini kit (Qiagen) and subjected to reverse transcription by transcriptor first strand cDNA synthesis kit (Roche Applied Science). qPCR was performed using PerfeCTa SYBR Green Supermix (Quanta Biosciences), and cyclophilin B (Ppib) was used as an internal standard (57). The ΔΔCt method was used to compare gene expression levels between control and test samples. PMAC cultures were treated with 1 μm XAV939 (Sigma) in dimethyl sulfoxide (DMSO) or DMSO alone or with 10 ng/ml mouse recombinant WNT3A (R&D Systems) or bovine serum albumin (BSA) as a control. The cells were treated for 16 h and then harvested for RNA extraction. C3H10T1/2 cells were transfected with 200 nm small interfering RNA (siRNA) against β-catenin (Santa Cruz Biotechnology, sc-29210) or Silencer Select Negative Control no. 1 siRNA (Invitrogen, 4390844) using FuGENE 6, and the cells were harvested after 48 h.

ChIP-qPCR

Maxillary arches were dissected from 35 to 40 CD-1 wild type embryos at E11.5. The chromatin was cross-linked with formaldehyde and sonicated with Bioruptor (Diagenode) into 0.5–1-kb fragments. The rest of the chromatin preparation and immunoprecipitation were performed with ChIP-IT high sensitivity kit (Active Motif). 4 μg of anti-LEF1 antibody (rabbit, Cell Signaling, 2230) and normal rabbit IgG (Millipore, 12-370) were used for immunoprecipitation. DNA was purified from the immunoprecipitated chromatin and input chromatin using the ChIP-IT high sensitivity kit, and subjected to qPCR. Primers amplifying the part of Lhx8_enh1 inclusive of the LEF/TCF-binding motif were used to detect LEF1 binding to Lhx8_enh1. The negative control primers were from Ref. 58, amplifying a part of a p63 intron that was previously shown to lack LEF1 binding in the embryonic face (58). Enrichment of each region in LEF1 ChIP DNA was calculated as a fold difference from the IgG control ChIP DNA. The ChIP experiment was repeated three times using a new preparation of chromatin each time, and the results were combined for Fig. 8.

FIGURE 8.

Regulation of Lhx8_enh1 by WNT/β-catenin signaling. A, position weight matrix for LEF1 from JASPAR database (102), and the LEF/TCF-binding motif in Lhx8_enh1 of chicken, mouse, and human. mLEF, mutated LEF/TCF-binding motif. B and C, result of the luciferase reporter assay for Lhx8_enh1 and Lhx8_enh1^mLEF with indicated treatments. D, anti-LEF1 ChIP-qPCR result from E11.5 maxillary arches. Fold enrichment over IgG mock ChIP is shown for the region of Lhx8_enh1 encompassing the LEF/TCF motif and a negative control region in the genome (see “Experimental Procedures” for details). The error bars indicate standard deviation from three experiments.

RESULTS

Genome-wide Mapping of the Maxillary Arch Enhancers

To facilitate our search for a face enhancer of Lhx8, we sought to build a comprehensive catalog of enhancers that are active in PA1 when Lhx8 is strongly expressed (E10.5–12.5) (17, 25). A couple of recent studies reported identifying large numbers of craniofacial enhancers; one study combined ChIP-seq for P300, H3K27ac, and H3K4me1 from neural crest cells derived from human embryonic stem cells (43), and the other study used ChIP-seq for P300 from the whole face of E11.5 mouse embryos followed by in vivo functional tests in transgenic embryos (39). Both studies contributed profoundly to our understanding of the transcriptional regulation during craniofacial development; however, neither found a face enhancer near Lhx8 (see under “Discussion”).

To generate a data set that is tailored to our goal, we performed ChIP-seq for H3K27ac from the maxillary arches of E11.5 mouse embryos (Fig. 1A). H3K27ac was shown to be more specific to active enhancers than H3K4me1 or P300 (37, 44). In addition, by using only the maxillary arches instead of the whole face, we aimed to reduce the heterogeneity of the cell population and consequently to increase the sensitivity of the ChIP-seq detection of the genomic regions with moderate enrichment of H3K27ac. We chose the maxillary arch over the mandibular arch because the former contributes to the structures that are more frequently affected in human birth defects, i.e. the upper lip and the palate (59, 60).

FIGURE 1.

Genome-wide mapping of the enhancers that are active in the maxillary arch. A, overview of the ChIP-seq experiment. The maxillary arches were dissected following the dotted line. A representative image of the ChIP-seq result is shown (visualized by integrated genome browser). The y axis corresponds to the fragment density derived from the Illumina sequencing result (see “Experimental Procedures” for details). The orange bars indicate the H3K27ac peaks determined by SICER peak-finding algorithm. B, distribution of the H3K27ac peaks relative to the nearest TSS.

The ChIP-seq identified 25,772 regions that were enriched with H3K27ac (peaks) (Fig. 1A, supplemental Table S1). Approximately one-third of the H3K27ac peaks were within 5 kb of the nearest TSS (Fig. 1B), consistent with an earlier finding that promoters tend to be enriched with H3K27ac (61). However, another one-third of the peaks was over 50 kb away from the nearest TSS, suggesting that they were distant-acting enhancers (Fig. 1B).

H3K27ac Peaks Are Associated with Known Maxillary Arch Genes and Maxillary Arch Enhancers

To assess the biological relevance of our ChIP-seq data, first we performed functional annotation using GREAT (52). We found that the H3K27ac peaks were significantly associated with the genes expressed in the craniofacial tissue of an embryo (Fig. 2A; supplemental Table S2).

FIGURE 2.

Association of the H3K27ac peaks with known maxillary arch genes and enhancers. A, gene ontology analysis by GREAT. PA1-related terms are selected from the 20 most-enriched “expression” terms associated with the H3K27ac peaks. The binomial p value for each term is shown. All 20 terms are listed in supplemental Table S2. B, comparison of the H3K27ac peak densities around the maxillary arch genes and around random genes at various distances from TSS. Error bars for the random genes indicate 95% confidence interval. C, Z-scores for the data in B. D, fold differences in the H3K27ac peak densities between the maxillary arch genes and random genes, calculated from B. E, left panel, percentages of the enhancers marked by H3K27ac peaks, among 20 maxillary arch enhancers and 20 control enhancers. Right panel, comparison of the H3K27ac ChIP-seq signals between the two groups of enhancers.

Next, we compiled a list of 474 genes that are expressed in the maxillary arch between E9.5 and E13.0 (the maxillary arch genes, supplemental Table S3), and we compared the distribution of the H3K27ac peaks around TSS of the maxillary arch genes versus random genes. We examined up to 200 kb away from TSS, and along the entire distance, the H3K27ac peaks were present at higher densities around the maxillary arch genes than around the random genes (Fig. 2, B and C; Z score >2.58, or p < 0.01, at all positions). The Z score and the fold difference in the peak density were greatest between 2.5 and 30 kb from TSS (Fig. 2, C and D).

Finally, we used information of previously identified enhancers to validate our data. From VISTA Enhancer Browser (35), we selected 20 mouse enhancers that gave lacZ reporter expression in the maxillary arch and 20 control enhancers that were not active in the maxillary arch (supplemental Table S4). 18 of the maxillary arch-positive enhancers (90%) co-localized with an H3K27ac peak, whereas only six of the maxillary arch-negative enhancers (30%) did (Fig. 2E). Similarly, the H3K27ac ChIP-seq signals of the individual enhancers were significantly higher in the maxillary arch-positive group than in the negative group (p < 0.0001 from Student's t test) (Fig. 2E).

Functional Test of the H3K27ac Peaks Identified Novel Enhancers Around Human Palate Defect Genes

Following the in silico validation of our data, we tested whether they could accurately predict active enhancers. The current standard for confirming the activity of an enhancer is an in vivo reporter assay using transgenic mice. However, generating transgenic animals is costly and slow, and thus it is difficult to test more than a couple of candidate enhancers in most cases.

For a rapid and cost-effective test of candidate maxillary arch enhancers, we adopted PMAC from E11.5 embryos for an in vitro luciferase reporter assay (Fig. 3A). We prepared PMAC following the protocol developed by other researchers for the culture of mouse embryonic palatal mesenchyme cells from older embryos. Mouse embryonic palatal mesenchyme cells have been widely used as a biologically relevant in vitro surrogate for the developing palate (54, 62–70). In our experiment, an actual maxillary arch enhancer would be activated by transcription factors that are endogenously expressed in PMAC, leading to up-regulation of the reporter expression. We confirmed the accuracy of this method using three genomic regions that had been tested by the in vivo transgenic assay (Fig. 3B).

FIGURE 3.

Luciferase reporter assay in PMAC for rapid functional tests of candidate maxillary arch enhancers. A, maxillary arches were dissected following the dotted line. A representative image of the cells after they were dissociated and plated is shown. B, to evaluate the accuracy of the reporter assay in PMAC, we used three genomic regions that had been tested by the in vivo transgenic assay (mm622, mm75, and mm70 from VISTA enhancer browser). mm622 induced robust expression of lacZ reporter in the maxillary arch of E11.5 embryos (39), whereas mm75 and mm70 were heart enhancers that gave no reporter expression in the maxillary arch (40). Consistent with the in vivo results, mm622, but not mm75 or mm70, activated luciferase reporter expression in PMAC. minP, minimal promoter control. **, p < 0.01 from Student's t test.

We selected eight of the H3K27ac peaks for the in vitro reporter assay (Fig. 4, A–H; Table 1). The candidate enhancers were chosen from within or near a gene implicated in congenital palate defects in humans, so any novel enhancer that we identify would be of medical significance. We compiled a comprehensive list of the genes that had been connected to palate abnormalities in humans (supplemental Table S5) and then selected candidate enhancers from various positions relative to the palate defect gene (Fig. 4, A–H). Specifically, ABCA4, CHD7, MAFB, MYH9, PAX9, ROR2, and TGFBR2 were found to be associated with human palatal defects in multiple studies for each gene (see supplemental Table S5). In addition, there were conspicuous H3K27ac peaks around these genes that were high and yet narrow and thus amenable to cloning. Therefore, we cloned and tested candidate enhancers located in the first intron of Abca4 (Fig. 4A), 43 kb upstream of Chd7 (Fig. 4B), 84 kb upstream of Mafb (Fig. 4C), 63 kb upstream of Myh9 (Fig. 4D), 25 kb upstream of Pax9 (Fig. 4E), 22 kb upstream of Ror2 (Fig. 4F), in the second intron of Tgfbr2 (Fig. 4G), and 208 kb upstream of Tgfbr2 (Fig. 4H).

FIGURE 4.

Functional test of candidate enhancers near human palate defect genes. A–H, H3K27ac ChIP-seq result for the mouse genomic regions near human palate defect genes. The red boxes indicate the H3K27ac peaks selected for functional tests. The black arrows next to gene names indicate the direction of transcription. I, result of the luciferase reporter assay for the candidate enhancers from A to H, performed in E11.5 PMAC. The relative luciferase activity of each enhancer is shown as the fold change from the minimal promoter control. The error bars indicate the standard deviation from triplicates. ***, p < 0.001 from Student's t test. n.s., not significantly different (p > 0.05). J, lateral view of the head of an Ror2-enh_lacZ transgenic embryo stained for β-galactosidase activity (blue/green). Abbreviations used are as follows: mxPA1, maxillary arch; mdPA1, mandibular arch.

TABLE 1.

List of candidate maxillary arch enhancers selected for functional tests

Enhancer name	Nearest gene (TSS)	Enrichment found in ChIP-seq				Result of functional test
		Human ESC-derived NCCa		Mouse E11.5 whole faceb, P300	Mouse E11.5 maxillary arch, H3K27Ac	VISTA Enhancer Browser, in vivo transgenic	This study
		P300	H3K27Ac				Primary culture	In vivo transgenic
Abca4_enh	Abca4	Yes	Yes	Yes	Yes	+	+
Chd7_enh	Chd7	No	Yes	No	Yes		−
Mafb_enh	Mafb	No	No	Yes	Yes	−	−
Myh9_enh	Txn2	No	No	No	Yes		+
Pax9_enh	Pax9	No	No	No	Yes		+
Ror2_enh	Ror2	Yes	Yes	Yes	Yes		+	+
Tgfbr2_enh1	Tgfbr2	No	No	Yes	Yes		+
Tgfbr2_enh2	Tgfbr2	No	No	No	Yes		+
Lhx8_enh1	Lhx8	No	No	No	Yes		+	+
Lhx8_enh2	Lhx8	No	No	Yes	Yes		−	−

^a Data are from Rada-Iglesias et al. (43).

^b Data are from Attanasio et al. (39).

Of the eight candidate enhancers, six up-regulated luciferase reporter expression in PMAC (Fig. 4I). The confirmation rate (75%) was comparable with what had been reported for candidate enhancers from P300 ChIP-seq tested by the in vivo transgenic assay (60%) (39). Importantly, three of the six enhancers that we confirmed in PMAC (Myh9_enh, Pax9_enh, and Tgfbr2_enh2) had not been identified as candidates from the two previous ChIP-seq searches (Table 1) (39, 43), demonstrating the unique value of our data set in uncovering the maxillary arch enhancers.

Among the enhancers that were active in PMAC, Ror2_enh displayed the strongest activity (Fig. 4I). Therefore, we tested it in vivo by generating transgenic embryos with Ror2_enh-lacZ construct. Four of eight transgenic embryos showed a consistent pattern of β-galactosidase stain in PA1 (Fig. 4J; supplemental Fig. S1), overlapping with the broad expression of Ror2 in the face (71). Although Ror2_enh had been identified as a candidate enhancer in previous studies (Table 1), our data provided the first evidence for its activity from functional tests.

Enhancer Located 33 kb Upstream of Lhx8 Drives Gene Expression in PA1

Using the ChIP-seq data and the methodology established above, we set out to find a face enhancer of Lhx8. There was a large H3K27ac peak centered at the promoter of Lhx8 and several smaller peaks upstream and downstream (Fig. 5A). In addition, Lhx8 is expressed in very similar patterns in PA1 of mouse and chick embryos (17, 72), which suggested that the DNA sequence of the enhancers driving this expression was likely to be conserved.

FIGURE 5.

Identification of an Lhx8 face enhancer. A, mouse genomic region around Lhx8 showing the result of H3K27ac ChIP-seq and evolutionary conservation of the sequence from Multiz Alignment of UCSC Genome Browser (100). prom, 1.5 kb Lhx8 promoter used in C. B, enlargement of the boxed area in A. enh1 and enh2 are the two candidate Lhx8 enhancers tested in C. C, result of the luciferase reporter assay in E11.5 PMAC. Relative luciferase activity is shown as the fold change from the minimal promoter control (minP). The error bars indicate standard deviation from triplicates. D, lateral view of the head of an E10.5 wild type embryo processed by whole mount in situ hybridization for Lhx8 mRNA. E and F, transgenic embryos stained for β-galactosidase activity. G, alignment of Lhx8_enh1 sequences from mouse, human, and chicken generated by Clustal Omega (101). The boxes indicate the conserved binding motifs for transcription factors.

Therefore, we intersected the H3K27ac peaks with the sequence conservation between the mammals and the chick to refine the selection of the candidate enhancers. Notably, although all the exons of Lhx8 were strongly conserved, the promoter region was not (Fig. 5A). However, the conservation was found in multiple short fragments in the intergenic region around Lhx8 (Fig. 5A). In particular, two ∼300-bp conserved fragments (red bars in Fig. 5B) were located within the H3K27ac peaks upstream of Lhx8, one (enh1) at ∼33 kb from TSS and the other (enh2) at ∼37 kb from TSS. These fragments, along with 50–100-bp flanking sequences, were cloned for functional tests of the enhancer activity. For comparison, 1.5 kb of the Lhx8 promoter was also cloned (−1 to −1501 bp; prom in Fig. 5A).

In the luciferase reporter assay in PMAC, Lhx8_prom caused a significant up-regulation (∼6-fold) of reporter expression over the minimal promoter (Fig. 5C). Strikingly, Lhx8_enh1 induced over 16-fold up-regulation of the reporter (Fig. 5C), showing a much greater effect than Lhx8_prom, even though Lhx8_enh1 was less than one-third the size of Lhx8_prom. However, Lhx8_enh2 did not activate reporter expression in PMAC (Fig. 5C). This result suggested that the crucial regulatory information for Lhx8 expression in the face resided in the distal enhancer Lhx8_enh1.

To test the activity in vivo, we generated transgenic embryos carrying the Lhx8_enh1-lacZ reporter. Among the nine transgenic embryos, six embryos stained positive for β-galactosidase (supplemental Fig. S2). Importantly, four of the six embryos showed the reporter activity in PA1, closely resembling the normal expression pattern of Lhx8 (Fig. 5, D and E; supplemental Fig. S2). Contrary to Lhx8_enh1, Lhx8_enh2 failed to induce reporter expression in PA1 (none of the six transgenic embryos) (Fig. 5F and supplemental Fig. S3). Together, the in vivo transgenic test confirmed the findings from the luciferase assay in PMAC (Fig. 5C) that Lhx8_enh1, but not Lhx8_enh2, is a PA1 enhancer.

Lhx8_Enh1 Contains an Evolutionarily Conserved Binding Motif for LEF/TCF Transcription Factors

To identify the regulators of Lhx8_enh1, we examined its sequence for binding motifs of transcription factors. We used rVISTA 2.0 (73) to detect the binding motifs of 467 transcription factor families in the TRANSFAC library (74). The sequence conservation between the mouse and chick was included as a criterion. This analysis found four transcription factor-binding sites within Lhx8_enh1, which were predicted to bind to STAT, ETS, cAMP-response element-binding protein, and LEF/TCF family members (Fig. 5G). Discovery of the LEF/TCF-binding site was particularly intriguing because the LEF/TCF family mediates the transcriptional outcome of the WNT/β-catenin signaling pathway, which plays a crucial role in craniofacial development (75–77); however, there had been little evidence to suggest that Lhx8 was directly regulated by WNT/β-catenin signaling (see “Discussion”). Therefore, in this study, we focused on further investigation into the connection between Lhx8 and WNT/β-catenin pathway; the significance of the other transcription factor-binding sites in Lhx8_enh1 will be addressed in future studies.

Lhx8 Expression Is Regulated by WNT/β-Catenin Signaling within the Maxillary Mesenchyme

To determine whether WNT/β-catenin signaling regulated the expression of Lhx8 in PA1, first we compared the patterns of Lhx8 expression and WNT/β-catenin pathway activation during normal craniofacial development. We used two indicators of WNT/β-catenin signaling activity, the Bat-gal transgene (47), which has a β-galactosidase coding sequence following LEF/TCF-binding motifs, and Axin2, which is a direct transcriptional target of WNT/β-catenin pathway (78). At E10.5, we found extensive overlap in the expression of Lhx8 and Bat-gal in the face, especially in the anterior half of the maxillary arch (Fig. 6, A and B). In the sections of the E10.5 maxillary arch, Axin2 expression was detected in most of the Lhx8 expression domain, although the expression of Axin2 was weak medially and strong laterally unlike Lhx8 showing the strongest expression around the center (Fig. 6, C and D). However, at E11.5, a common pattern appeared in the distribution of Lhx8 and Axin2 transcripts in the oral mesenchyme (Fig. 6, E–H); in the medial-to-lateral progression, both genes showed the sequence of moderate (black arrowheads), lower (white arrowheads), and the highest (black arrows) levels of expression when the comparison was made among the three positions. However, Axin2, but not Lhx8, was expressed in the aboral mesenchyme below the eye (Fig. 6, C–H). This observation suggested that WNT/β-catenin signaling positively regulated Lhx8 expression in PA1, but only the oral mesenchyme was competent to respond.

FIGURE 6.

Comparison of the patterns of Lhx8 expression and WNT/β-catenin signaling during normal development of PA1. A, wild type embryo processed by whole mount RNA in situ hybridization for Lhx8. The dotted line demarcates PA1. B, Bat-gal embryo stained for β-galactosidase activity. C–H, coronal sections of the head of wild type embryos processed by section RNA in situ hybridization. The right half of the face is shown. The black arrowheads, white arrowheads, and black arrows in E–H point to the areas with moderate, low, and highest levels of gene expression, respectively. Abbreviations used are as follows: ab, aboral; E, eye; la, lateral; me, medial; or, oral; PS, palatal shelf; UM, upper molar. Bar, 0.25 mm.

To directly test the effect of altering WNT/β-catenin signaling on Lhx8 expression, we treated PMAC from E10.5 and E11.5 embryos with XAV939, a chemical inhibitor of the WNT/β-catenin pathway (79). XAV939 greatly reduced Axin2 expression, confirming the inhibition of WNT/β-catenin signaling (Fig. 7, A and B). At the same time, Lhx8 expression was almost completely abolished, indicating that WNT/β-catenin signaling was essential for Lhx8 expression (Fig. 7, A and B). In contrast, Lhx6 expression was not affected by XAV939.

FIGURE 7.

Regulation of Lhx8 expression by WNT/β-catenin signaling. A–G, result of RT-qPCR showing the changes in gene expression levels upon inhibition (A–E) or stimulation (F and G) of WNT/β-catenin signaling. The error bars indicate standard deviation from triplicates. *, p < 0.05 from Student's t test; β-cat, β-catenin; w/o epi, without epithelium; **, p < 0.01; ***, p < 0.001.

PMAC cultures in the above experiment contained both the epithelium and mesenchyme cells of the maxillary arches. As such, we could not distinguish whether WNT/β-catenin signaling was required within the mesenchyme to regulate Lhx8 or whether it was required in the epithelium only to produce a relay signal, which then acted on the mesenchyme to regulate Lhx8 (see “Discussion” for details). Therefore, we prepared PMAC cultures without the epithelium and repeated the above experiment. Although the epithelium is a source of multiple WNT proteins (15, 80), PMAC without the epithelium expressed detectable levels of Axin2. This might reflect the activities of WNT proteins that had been transferred to the mesenchyme before the epithelium was removed, or the mesenchymal expression of Wnt(s) that had yet to be identified. Axin2 expression in PMAC without the epithelium was significantly inhibited by XAV939, and Lhx8 expression was reduced accordingly (Fig. 7, C and D). This result indicated that WNT/β-catenin signaling was necessary within the maxillary arch mesenchyme for the expression of Lhx8.

To corroborate the above result, we performed the same test using another method of blocking the WNT/β-catenin pathway, i.e. siRNA-mediated knockdown of β-catenin. Because of the modest transfection efficiency of PMAC, we turned to C3H10T1/2, an immortalized cell line of undifferentiated mesenchyme cells. C3H10T1/2 endogenously expresses many transcription factors that are found in craniofacial mesenchyme, including LHX8 (81). Transfection with β-catenin siRNA efficiently inhibited both WNT/β-catenin signaling and Lhx8 expression in C3H10T1/2 (Fig. 7E), and thus it confirmed the result from XAV939 treatment.

Having established that WNT/β-catenin signaling was necessary for Lhx8 expression, we asked whether an increase in WNT/β-catenin signaling was sufficient to up-regulate Lhx8. Addition of WNT3A protein moderately enhanced WNT/β-catenin pathway activity in PMAC without the epithelium, according to the changes in Axin2 expression (Fig. 7, F and G). Interestingly, WNT3A was able to up-regulate Lhx8 in PMAC from E11.5 but not E10.5 embryos (Fig. 7, F and G). This result suggested that at E10.5, WNT/β-catenin signaling in the mesenchyme provided only a permissive condition to express Lhx8, whereas at E11.5, WNT/β-catenin signaling played a more instructive role to determine the level of Lhx8 expression within the responsive domain.

Lhx8_enh1 Is Directly Regulated by WNT/β-Catenin Pathway

The presence of a LEF/TCF-binding motif in Lhx8_enh1 suggested that it might be regulated by WNT/β-catenin signaling, and therefore, we tested this possibility using the luciferase reporter assay.

To determine whether the LEF/TCF-binding site was important for the function of Lhx8_enh1, we disrupted it by replacing three nucleotides (mLEF; Fig. 8A). This mutation (Lhx8_enh1^mLEF) completely abolished the activity of Lhx8_enh1 in PMAC, and it also diminished, although it did not abolish, the enhancer activity in C3H10T1/2 (Fig. 8, B and C; compare the 1st 3 columns). Furthermore, XAV939 and β-catenin siRNA significantly lowered the reporter up-regulation by Lhx8_enh1 (Fig. 8, B and C; compare columns 2 and 5), whereas these treatments did not have any effect on Lhx8_enh1^mLEF (Fig. 8, B and C; compare the columns 3 and 6). Finally, we performed ChIP from the maxillary arches with anti-LEF1 antibody to detect direct binding of LEF1 to Lhx8_enh1. LEF1 was found enriched at the LEF/TCF motif of Lhx8_enh1 but not at a negative control region of the genome (Fig. 8D).

Together, we conclude that WNT/β-catenin pathway directly regulates Lhx8_enh1 through the conserved LEF/TCF-binding site, and this input is essential for the robust enhancer activity of Lhx8_enh1.

DISCUSSION

In this study, we used the approach of enhancer discovery to investigate the molecular mechanisms for the regulation of Lhx8, one of the key genes for craniofacial development. We generated a comprehensive map of putative maxillary arch enhancers that are active during early craniofacial morphogenesis. We validated our data by in silico statistical analyses and functional assays of select candidates, which uncovered multiple novel enhancers near the genes implicated in human palate defects. Finally, by identifying and characterizing a face enhancer of Lhx8, we provided compelling in vitro evidence that Lhx8 is a direct target of WNT/β-catenin signaling within the facial mesenchyme.

New Resource and Method for Identifying and Characterizing Craniofacial Enhancers

It has now become evident that enhancers play crucial roles in development, evolution, and human diseases. Consequently, significant efforts have been made to find enhancers in various contexts (32–34, 82, 83).

Prior to our study, two reports have used ChIP-seq to successfully identify craniofacial enhancers in large scales, from human embryonic stem cell-derived neural crest cells using multiple markers of enhancers (43) or from the whole face of mouse embryos using P300 as a marker (39). However, several of the maxillary arch enhancers found in this study, including Lhx8_enh1, were not detected in either report (Table 1), which indicates that our ChIP-seq result contains unique information that can supplement the data from the two previous reports. A potential explanation for the differences in the ChIP-seq results is that, given the dynamic changes in gene expression during development, the list of active enhancers is likely to vary depending on the precise developmental status of the cells. Neural crest cells have broad developmental potential beyond contributing to the maxillary arch, making the two samples significantly different. In addition, P300 is associated with only a subset of active enhancers, although this appears to be true for H3K27ac also (38, 41). Therefore, combining our ChIP-seq data with the two other sets will provide a more comprehensive and powerful resource for finding craniofacial enhancers. Our data will be particularly useful for the studies focusing on the early stages of maxillary arch/upper jaw development.

Functional tests found that a significant portion of the candidate enhancers from ChIP-seq failed to activate reporter expression in the expected tissue (39–42, 46, 84), highlighting the importance of confirming the candidate enhancers through functional assays. In this study, we used the in vitro luciferase assay in PMAC as an efficient method to test maxillary arch enhancers. We performed the in vitro assay for 14 genomic regions (those listed in Table 1 and Fig. 3, and Lhx8_prom), eight of which were also tested by the in vivo transgenic assay (three in this study and five from VISTA Enhancer Browser) (39, 40). Remarkably, the results from the two assays were consistent for all eight genomic regions, validating our in vitro approach. Because PMAC can be subject to transient transfection and/or treatment with various modulators of signaling pathways, it can facilitate investigation into the molecular mechanism of enhancer function, as we have demonstrated with Lhx8_enh1.

Regulation of Lhx8 Expression in PA1

Earlier studies found that Lhx8 expression in PA1 mesenchyme was dependent on the signals from the overlying ectoderm, and specifically, FGF was identified as a key regulator (17, 85). Genetic deletion of Fgf8 in PA1 ectoderm severely reduced the expression of Lhx8 in the mesenchyme (16), and a consistent result was obtained from pharmacological inhibition of FGF signaling in mandibular arch explants (86). Conversely, FGF8 protein was sufficient to induce Lhx8 expression from the mandibular arch mesenchyme that was cultured without the epithelium (17, 85); however, it is unknown whether FGF8 is also sufficient in the maxillary arch mesenchyme for Lhx8 expression. The fact that Lhx8_enh1 contains two binding motifs for ETS family transcription factors, which are effectors of FGF/MAPK pathway (87), suggests that FGF signaling may directly regulate Lhx8. Further investigation is necessary to define the exact relationship between FGF and Lhx8.

Compared with FGF signaling, little was known about the potential role of WNT/β-catenin signaling in the regulation of Lhx8. Prior to our work, the only report that suggested this connection was from a study in chicks (88). Here, they inhibited the WNT/β-catenin pathway by implanting a bead soaked in an antagonist DKK-1 in the intact maxillary arch and found that Lhx8 expression was reduced. However, because WNT/β-catenin signaling within the oral ectoderm is necessary for Fgf8 expression (89–91), it remained unknown whether WNT/β-catenin pathway had any direct role in regulating Lhx8 in PA1 mesenchyme. By separating the epithelium and mesenchyme of PA1, we showed that WNT/β-catenin signaling within the mesenchyme regulated Lhx8 expression. More importantly, we elucidated the molecular mechanism underlying this regulatory relationship, i.e. the direct activation of Lhx8_enh1 by WNT/β-catenin pathway through the conserved LEF/TCF-binding site.

Downstream Targets of WNT/β-Catenin Signaling in the Developing Face

The WNT/β-catenin pathway is one of the major regulatory pathways in development, and thus many researchers have investigated its role in craniofacial development (76, 92). In the orofacial epithelium, WNT/β-catenin signaling regulates the fusion of the lip and palate and the number of the teeth (58, 89–91, 93–96). In the mesenchyme, this pathway is important for the growth of the facial primordia and morphogenesis of the teeth (75, 77, 97–99). However, only a few genes have been identified as direct targets of WNT/β-catenin pathway in the developing face, which are p63 and Fgf8 in the epithelium (58, 91) and Bmp4, Msx1, and Msx2 in the mesenchyme (77, 98). We have now added Lhx8 to the list, and it is likely that Lhx8 mediates some of the functions of WNT/β-catenin pathway in craniofacial development.