The Rosa26 ^{CAG-LSL-Dlx2−3xFlag} mouse model was developed by Shanghai Model Organisms Center, Inc(Shanghai, China). This model was generated using the CRISPR/Cas9 system in a C57BL/6J mouse background (Figure 1A). The sequence information used in the construction of the mouse model is NM_010,054.2→NP_034,184.1 homeobox protein DLX-2. The knockin site was Gt (ROSA)26Sor (MGI number: 104,735). The chromosome position of the Rosa26 gene is in Chromosome 6: 113,067,428–113,077,333 reverse strand.

Briefly, a targeting vector containing the following components was constructed: CAG-LSL-Dlx2-3xFlag-woodchuck hepatitis virus post-transcriptional regulatory element (WPRE)-bGH poly(A). Cas9 mRNA was in vitro transcribed with the mMESSAGE mMACHINET7 Ultra Kit (Ambion, TX, United States) according to the manufacturer’s instructions, and subsequently purified using the MEGAclear™ Kit (ThermoFisher, United States of America). 5′-GGG​GAC​ACA​CTA​AGG​GAG​CT-3′ was chosen as Cas9 targeted guide RNA (sgRNA) and in vitro transcribed using the MEGAshortscript Kit (ThermoFisher, United States) and subsequently purified using MEGAclear™ Kit. The donor vector with sgRNA and Cas9 mRNA was microinjected into C57BL/6J fertilized eggs. F0 generation mice positive for homologous recombination were identified by long PCR. The primers (P1-P4) used for genotyping the correct homology recombination were P1: 5′-GCC​GGG​CCT​CGT​CGT​CT-3′ and P2: 5′-TTT​TTG​GGG​GTG​ATG​GTG​GTC-3′ for the correct 5′ homology arm recombination, and P3: 5′-CTG​CGC​GGG​ACG​TCC​TTC​TGC​TAC-3′ and P4: 5′-GGG​ATC​CAT​TGC​CAC​CTT​TCA​CTT-3′ for the correct 3′ homology arm recombination. The PCR products were further confirmed by sequencing. F0 mice were crossed with C57BL/6J mice to obtain Rosa26 ^{CAG-LSL-Dlx2−3xFlag} heterozygous mice.

After the Rosa26 ^{CAG-LSL-Dlx2−3xFlag} mouse model was generated, the primer sequences used for genotype identification were P1: 5′-TCA​GAT​TCT​TTT​ATA​GGG​GAC​ACA-3′, P2: 5′-TAA​AGG​CCA​CTC​AAT​GCT​CAC​TAA-3′, P3: 5′-GGT​GTT​GTC​GGG​GAA​ATC​ATC​GTC-3′ and P4: 5′-AGG​AGC​CTG​CCA​AGT​AAC-3’. The primer pairs (P1, P2) and (P3, P4) were amplified separately. In WT mice, only (P1, P2) amplified a PCR product of 994bp, whereas (P3, P4) should not give rise to any PCR product. In heterozygous mice, (P1, P2) amplified a PCR product of 994bp, while (P3, P4) amplified a smaller PCR product of 617bp (Figure 1B).

WT C57BL/6J mice were purchased from Shanghai Jihui Laboratory Animal Care Co.Ltd.(Shanghai, China). All mice were maintained under SPF conditions at the Animal Center of the Ninth People’s Hospital affiliated to Shanghai Jiao Tong University School of Medicine. All animal experiments were performed by following protocols approved by the Animal Care and Usage Committee of the Ninth People’s Hospital affiliated to Shanghai Jiao Tong University School of medicine.

After removing the skin, internal organs, and back fat on the day of birth, the mice were treated with 95% ethanol overnight and stained with Alcian Blue at 37°C for 42 h. Following shaking and washing twice with 95% ethanol for 1 h and 2% KOH digestion for 2.5 h, the mice were stained with Alizarin Red for 1 h, digested with 1% KOH for 1 day, soaked in gradient glycerol until there were no bubbles, and photographed.

Entire heads of E15.5 and E13.5 mice embryos were surgically dissected. Following fixation in 4% paraformaldehyde and gradient dehydration in ethanol, the tissues were embedded in paraffin and sections at a thickness of 8 µm were cut and stained with Hematoxylin and Eosin (HE). Immunofluorescence staining was performed with anti-Dlx2 polyclonal antibody (ab272902, Abcam, United States, 1:250), anti-DDDDK tag antibody (ab205606, Abcam, United States, 1:200), goat secondary antibody to rabbit IgG (ab150077, Abcam, 1:500) following a previously described protocol (Ha, Sun, Bian, Wu, & Wang, 2022). Images were captured using an Olympus IX83 inverted microscope.

The head sections of E13.5 WT and wnt1 ^cre ; Rosa26 ^Dlx2/- mice were dewaxed to water and a HA staining kit (g3710, Solarbio, China) was used for HA staining. Following hyaluronidase treatment for 3 h, the treated sections were soaked with Alcian Blue for 30 min, counterstained with nuclear solid red for 30min, dehydrated and transparent, air dried, and sealed with neutral gum.

In situ hybridization was performed on head tissue sections of mouse embryos to detect the Dlx2 transcript levels using the RNAscope^® 2.5HD Reagent Kit- RED (322,350, ACDbio, California, US) and the RNAscope^® Probe-Mm-DLX2 (555,951, ACDbio, California, US) according to the manufactor’s instructions. Briefly, the slides were incubated twice at room temperature in xylene and then also 100% ethanol, incubated with the RNAscope^® hydrogen peroxide, and washed with distilled water. The slides were then transferred to 99°C RNAscope^® 1× Target Retrieval Reagent for 15 min. Following rinsing and ethanol, the slides were incubated in RNAscope^® protease Plus at 40°C for 30 min. The RNAscope^® Probe-Mm-DLX2 probe was applied and the slides were incubated at 40°Cfor 2 h. The slides were incubated in six amplifiers reagents in turn at toom temperature and washed with 1× wash buffer during each interval. Each section was incubated with sufficient RED working solution prepared using a 1:60 ratio of RNAScope Fast RED-B to RNAScope Fast RED-A. The slides were stained with hematoxylin and mounted.

Total RNA was extracted using Trizol from freshly dissected E12.5 maxillary prominence tissues from six littermates. For each genotype (WT and wnt1 ^cre ; Rosa26 ^Dlx2/- ), three independent RNA samples were prepared. RNA purity was tested using the kaiaoK5500^® Spectrophotometer (Kaiao, Beijing, China). RNA integrity and concentration were assessed using the RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies, CA, United States of America). For each sample, 2 μg total RNA was used as input material for the library preparations. Sequencing libraries were generated using the NEBNext^® UltraTM RNA Library Prep Kit for Illumina^® (#E7530L, NEB, United States of America) following the manufacturer’s recommendations and index codes were added to attribute sequences to each sample. The ligated products were retrieved and PCR amplification was performed to generate sequencing libraries. The RNA concentration of the library was determined using the Qubit^® RNA Assay Kit in Qubit^® 3.0 to preliminary quantify and then dilute to 1 ng/μL. The insert size was assessed using the Agilent Bioanalyzer 2100 system (Agilent Technologies), and the qualified insert size was accurately quantified using a StepOnePlus™ Real-Time polymerase Chain Reaction (PCR) System. The clustering of the index-coded samples was performed on a cBot cluster generation system using HiSeq PE Cluster Kit v4-cBot-HS (Illumina) according to the manufacturer’s instructions. Following cluster generation, the libraries were sequenced on an Illumina HiSeq X ten platform and 150 bp paired-end reads were generated. The raw reads of RNA-seq libraries were filtered by removing adaptor sequences, contamination, and low-quality (Phred quality <20) reads. Reads quality was assessed using FastQC. Sequenced reads were mapped to the mm10 genome using STAR aligner version 2.7.3a. Reads were counted using htseq-count version 0.12.4 at the union mode. Comparison between the RNA-seq datasets was performed using the DESeq2 package in R. Differentially expressed genes were defined as genes with a Benjamini–Hochberg adjusted p-value (padj) of <0.05 and log2 expression fold change (log2FC) > 1 or < −1. Enrichment analysis and visualization of functional profiles (GO and KEGG) of differentially expressed genes were performed using the clusterProfiler package in R.

Total RNA was extracted using the RNeasy Mini Kit (Qiagen, Valencia, CA, United States of America) from freshly dissected E12.5 maxillary prominence tissues from six littermates, following the manufacturer’s instructions. For each genotype (WT and wnt1 ^cre ; Rosa26 ^Dlx2/- ), three independent RNA samples different from the Bulk RNA-seq were prepared. The reverse transcriptase reaction was performed using the Hifair III first Strand cDNA Synthesis SuperMix for qPCR (11137ES60, Yeasen, China). Fluorescence-based real-time PCR was performed with Hieff qPCR SYBR Green Master Mix (11201ES08, Yeasen, China). The primer sequences were derived from gene sequences available through the PrimerBank or NCBI (Table 1). The relative expression was calculated for each gene by the 2−∆∆Ct method, normalized against β-actin expression, and presented as fold changes relative to the control. All analyses were performed in triplicate. Statistical analysis was performed using an appropriate Student’s t-test when two values were being compared, with p < 0.05 considered as significant.

Conditionally overexpressing Dlx2 in NCCs using a high copy-number transgenic mouse model (Wnt1 ^Cre :iZEG-Dlx2) led to a wide range of craniofacial abnormalities, from cleft lip to neural tube defects and exencephaly. These developmental defects with different severity may be attributed to the variable, and possibly leaky Dlx2 expression from the high copy-number transgenes. To avoid this technical problem and more explicitly address the effect of Dlx2 overexpression on craniofacial development, we constructed a new transgenic mouse model in which a CAG-LSL-Dlx2 expression cassette was inserted into the endogenous Rosa26 locus (Rosa26 ^{CAG-LSL-Dlx2−3xFlag} ) (Figure 1). The Rosa26 ^{CAG-LSL-Dlx2−3xFlag} mice were bred with wnt1 ^cre mice to obtain wnt1 ^cre ; Rosa26 ^Dlx2/- mice, which specifically overexpressed Dlx2 in neural crest-derived cells.

To further clarify the Dlx2 overexpression, we examined the temporal Dlx2 expression pattern during the early-stage development of maxillary processes from E9.5 to E13.5 (Figure 2).

In the whole in situ hybridization of E9.5 mouse head, it can be seen that the expression amount and expression range of Dlx2 in wnt1 ^cre ; Rosa26 ^Dlx2/- mice were increased (Figure 2A). In situ hybridization on head tissue sections of E10.5, E11.5, E12.5, and E13.5 WT mouse embryos revealed that Dlx2 was expressed in the epithelium and mesenchyme of the upper and lower jaws of the first branchial arch. At E10.5, the Dlx2 expression was readily detected in the distal region of both the maxilla and mandible of the first branchial arch (Figure 2B). At E11.5, concurrent with the rapid growth of the maxillary prominence, the Dlx2 expression level exhibited a significant elevation compared to that at E10.5. At the same time, the region exhibiting Dlx2 expression extended to the proximal end of the maxilla, occupying a large fraction of the maxillary prominences (Figure 2C). Interestingly, at E12.5, the Dlx2 expression level and area both exhibited a marked reduction (Figure 2D). Dlx2 expression was largely diminished at E13.5 (Figure 2E). Together, these observations revealed that Dlx2 expression in cranial neural crest-derived mesenchyme is largely restricted to E10.5 and E11.5, suggesting that Dlx2 may play a role in promoting the early-stage development of maxillary prominences. These findings also suggest that the ectopic Dlx2 overexpression in wnt1 ^cre ; Rosa26 ^Dlx2/- mice may influence the developmental trajectory of the palate from E12.5 on, when endogenous Dlx2 has been largely downregulated. Compared with E13.5 WT mice, the Dlx2 transcript level in E13.5 wnt1 ^cre ; Rosa26 ^Dlx2/- mice was increased (Figure 2FG). We also performed immunofluorescence staining of Dlx2 and Flag in the maxillary process of E13.5 mice (Figure 2H,I). Dlx2 was also overexpressed at the protein level, which was consistent with the results of in situ hybridization.

We carefully examined the craniomaxillofacial abnormalities caused by Dlx2 overexpression using this new Rosa26 knock-in model. On the day of birth (P0), wnt1 ^cre ; Rosa26 ^Dlx2/- mice exhibited markedly shorter maxilla compared to their WT littermates, rendering the lips unable to close (Figure 3A). Wnt1 ^cre ; Rosa26 ^Dlx2/- mice also displayed cleft palate (Figure 3B, black arrow), which is consistent with the previously reported phenotypes of Wnt1 ^Cre :iZEG-Dlx2 mice. Interestingly, the midfacial cleft phenotype associated with the Wnt1Cre:iZEG-Dlx2 mice was not observed in any P0 wnt1 ^cre ; Rosa26 ^Dlx2/- mice. Instead, the wnt1 ^cre ; Rosa26 ^Dlx2/- mice consistently exhibited a tiny cleft in the upper lip (Figure 3A, white arrow), considering the phenotype of previous mouse models, the manifestation of the midfacial cleft may dependent on the dose of the Dlx2 expression.

In addition to the short maxilla and cleft palate, the wnt1 ^cre ; Rosa26 ^Dlx2/- mice displayed abnormal eye development and soft tissue bulges on the head (Figure 3A). Skeletal staining of P0 wnt1 ^cre ; Rosa26 ^Dlx2/- mice revealed a partial defect of the frontal bones (Figure 3B, black star), suggesting that the swelling of the head tissues and the bulging out of the brain could be attributed to defects in skeletal development. The mandible of wnt1 ^cre ; Rosa26 ^Dlx2/- mice was also shortened to some extent (Figure 3C,D). Moreover, the temporomandibular joints of the wnt1 ^cre ; Rosa26 ^Dlx2/- mice were abnormally developed, with smaller joint heads and reduced articular cartilage (Figure 3D). These observations on P0 wnt1 ^cre ; Rosa26 ^Dlx2/- mice are correspond with the previous reports that the three-month-old Wnt1 ^Cre :iZEG-Dlx2 mice exhibited postnatal condyle malformation, subchondral bone degradation, and an irregular histological structure of the condylar cartilage (Dai et al., 2016), suggesting that Dlx2 plays critical roles in regulating condyle development during embryogenesis. Together, these results revealed that Dlx2 overexpression exerts a profound effect on a diverse array of neural crest-derived craniofacial skeletal tissues.

Of particular note, we found that all six P0 wnt1 ^cre ; Rosa26 ^Dlx2/- mice from three litters of newborn mice (22 in total) across three generations exhibited highly consistent craniofacial abnormalities, including short maxilla and cleft palate. Such a nearly homogeneous phenotypic manifestation demonstrates the stability of our mouse model. The wnt1 ^cre ; Rosa26 ^Dlx2/- mouse model enables further elucidation of Dlx2 function in the regulation of craniofacial development with greater spatiotemporal precision.

We next focused on understanding how Dlx2 overexpression in neural crest-derived mesenchymal cells affects the development of maxillary processes and causes short maxilla and cleft palate. We examined the morphology of maxillary processes of wnt1 ^cre ; Rosa26 ^Dlx2/- mouse embryos at different developmental stages. At E15.5, all four wnt1 ^cre ; Rosa26 ^Dlx2/- embryos from three litters (17 in total) across three generations displayed a depression-like cleft in the palate (Figure 4A, black arrow; Figure 4B), which penetrates the anterior and posterior of the palate. Notably, the tongue of E15.5 wnt1 ^cre ; Rosa26 ^Dlx2/- mouse embryos also appeared towering (Figure 4B, white star). Such a phenotype may be due to the changes in the pressure of the palate, which caused the tongue to fill the space of the cleft.

We next investigated whether the cleft palate in E15.5 wnt1 ^cre ; Rosa26 ^Dlx2/- embryos resulted from defective palate fusion or maxillary growth. Because the palate fusion in mouse embryos occurs at E14.5, we compared the morphology of E13.5 wnt1 ^cre ; Rosa26 ^Dlx2/- mice embryos with their WT littermates. At E13.5, the maxillary processes of the wnt1 ^cre ; Rosa26 ^Dlx2/- mice were markedly smaller than those of the WT (Figure 5A, black star). These findings collectively suggested that the cleft palate phenotype in wnt1 ^cre ; Rosa26 ^Dlx2/- mice cannot be solely attributed to defective palatal adhesion and fusion. Rather, the conditional Dlx2 overexpression already led to abnormal development of the maxillary process during the vertical outgrowth stage.

We used HA staining to analyze its accumulation. At E13.5, HA expression in wnt1 ^cre ; Rosa26 ^Dlx2/- mice was decreased compared with WT mouse embryos (Figure 5B). Of note, in the anterior and middle regions of the palatal process of wnt1 ^cre ; Rosa26 ^Dlx2/- mice, HA downregulation was more marked. The palatal process sections of WT and wnt1 ^cre ; Rosa26 ^Dlx2/- mice stained with HA were locally enlarged to observe HA expression and cell density (Figure 5C,D). We found that WT mouse palatal mesenchymal extracellular matrix (ECM) was filled with HA. By constrast, in the palatal process of wnt1 ^cre ; Rosa26 ^Dlx2/- mice, HA failed to fill the ECM space. In addition, an decrease in cell density was observed in the palatal process of wnt1 ^cre ; Rosa26 ^Dlx2/- mice, indicating that the number of cells decreased and the growth of palatal process was damaged. Therefore, it is reasonable to propose that the mutation resulted in downregulation HA expression, which affects the filling of extracellular stroma of palatal process cells, leading to insufficient uplift of palatal process. In addition, the decrease in the number of mesenchymal cells leads to insufficient growth of palatal process tissue, and finally results in problems in the lifting and fusion of palatal process.

We reasoned that Dlx2 overexpression may disrupt the complex gene expression programs during craniofacial development, thereby causing craniofacial defects. The wnt1 ^cre ; Rosa26 ^Dlx2/- mice that exhibited consistent and synchronized cleft palate phenotype enable us to profile the impact of Dlx2 overexpression on the transcriptome of mouse maxillary processes.

We surgically isolated maxillary processes from E12.5 WT or wnt1 ^cre ; Rosa26 ^Dlx2/- embryos and performed bulk RNA-seq on three independent biological replicates for each genotype. The individual RNA-seq replicate exhibited a high degree of correlation with those of the same genotype, while correlating less well with the replicates of the different genotype, indicating the good quality of our RNA-seq datasets (Figure 6A). Analysis of the RNA-seq data revealed that Dlx2 was upregulated by approximately 3-fold in wnt1 ^cre ; Rosa26 ^Dlx2/- than in WT. Moreover, Msx2, a transcription factor that is known to be a transcriptional target for Dlx2 (Sun, Liu, Li, Dai, & Wang, 2015; Dai et al., 2017; Zeng, Wang, Dong, Ma, & Liu, 2020), also exhibits a 3-fold upregulation upon Dlx2 overexpression (Figure 6B). The Msx2 expression in E12.5 maxillary process was confirmed by qPCR (Figure 6H), which was consistent with the bulk RNA-seq results. Collectively, these results demonstrate the efficacy of the Dlx2 overexpression in our novel mouse model and its value in elucidating the functions of Dlx2.

Comparison between the bulk results of wnt1 ^cre ; Rosa26 ^Dlx2/- and WT mice revealed that 2428 genes exhibited significant expression changes, of which 447 genes were significantly upregulated and 440 genes were significantly downregulated (Figure 6C). The most upregulated genes in wnt1 ^cre ; Rosa26 ^Dlx2/- mice include multiple transcription factors that have been implicated in craniofacial development, including Lef1, Irx4, and Pou3f3, demonstrating the critical functions of Dlx2 in orchestrating the transcriptional networks in neural crest-derived cells. Notably, Lef1, a transcription factor involved in regulating the Wnt pathway, has been shown to express at a high level in both epithelial and peri-epithelial mesenchyme during the early developmental stage of maxillary processes (Chen, Lan, Baek, Gao, & Jiang, 2009), consistent with the expression pattern of the WT Dlx2. Therefore, that a moderate Dlx2 upregulation is sufficient to induce marked remodeling of the transcriptional network and severe craniofacial defects, this highlights the critical functions of Dlx2 in regulating the development of craniofacial tissues.

The GO and KEGG enrichment analyses on the differentially regulated genes revealed that multiple development-related biological processes were impacted by Dlx2 overexpression. Genes associated with epithelial and neuronal development were enriched among the upregulated genes (Figure 6D), while genes associated with blood vessel development were enriched among the downregulated genes (Figure 6E). Hoxd1, an early neurodermal marker, which is part of a development regulatory system (Appukuttan et al., 2000), was significantly increased by qPCR(Figure 6H), which was consistent with the results of the bulk RNA-seq. Observing the down-regulated GO-term “exteacellular structure organization” and “extracellular matrix organization”, it can be found that their enriched genes contain Mmp9 (Dias et al., 2022), Ccn2 (Takeuchi-Igarashi, Tachibana, Murakashi, Kubota, & Numabe, 2021), Kazald1(H. Wang et al., 2013) and Col9a3(P. Zhang & Chang, 2020), which are genes related to cell proliferation. Therefore, the Dlx2 overexpression leads to the down-regulation of a group of genes related to cell proliferation, which hinders the development of palate by inhibiting cell proliferation. Notably, genes associated with the GO-term “ossification” were enriched among both the upregulated and downregulated genes upon Dlx2 overexpression, suggesting that the skeletal defects observed in wnt1 ^cre ; Rosa26 ^Dlx2/- mice result from profound dysregulation of genes related to skeletal development.

KEGG enrichment analysis suggested that the genes involved in Wnt and Hippo signaling were enriched among the upregulated genes (Figure 6F), in agreement with the upregulation of the Lef1 gene described above, suggesting that Dlx2 may indeed play an important role in regulating the Wnt pathway during craniofacial development. Axin2 (Jho et al., 2002) and Wnt3a (Samarzija, Sini, Schlange, Macdonald, & Hynes, 2009), the classical reporter genes of the Wnt signaling pathway, were detected by qPCR. Their expressions were consistent with bulk RNA-seq, both being significantly increased (Figure 6H). Ctnna2, which is closely associated with Hippo signaling pathways (Pavel et al., 2021), also displayed a consistent upregulation (Figure 6H). Moreover, genes related to the KEGG terms “ribosome” and “oxidative phosphorylation” were significantly enriched among the downregulated genes (Figure 6G), implying that Dlx2 overexpression may cause decreases in cellular biosynthesis and energy metabolism, thereby inhibiting cellular growth during craniofacial development. In summary, our systematical characterization of the transcriptome changes induced by Dlx2 overexpression revealed the critical functions of Dlx2 in the regulation of a variety of development-related pathways, providing important information for understanding the regulatory functions of Dlx2 during craniofacial development.