Malocclusion affects individuals across all racial and ethnic backgrounds and involves disruptions to both hard and soft tissues of the craniofacial complex. Despite increasing recognition of its genetic component, the mechanisms by which common variation influences craniofacial morphology remain poorly defined (Nikopensius, et al., 2013; Guan, et al., 2015; Perillo, et al., 2015; Chen, et al., 2015; Kantaputra, et al., 2019). Unlike rare Mendelian disorders, most skeletal malocclusions arise from polygenic contributions that are incompletely understood.

In our prior study, we conducted a candidate gene analysis of lateral cephalometric radiographs in individuals with malocclusion (da Fontoura, et al., 2015). Among 273 tested variants, we identified a significant association between a noncoding single nucleotide polymorphism (SNP) located 4.2 kb upstream of the TWIST1 transcription start site (rs2189000 A>G) and a pattern of craniofacial variation characterized by shorter mandibular ramus height, increased mandibular body length, and a steeper anterior cranial base angle (P = 0.000076). Although the phenotypic variation associated with this component explained approximately 8.2% of the total variance in the sample, the biological plausibility of the association was strengthened by overlap with known features of syndromic craniofacial disorders involving TWIST1. This finding served as the rationale for the current investigation.

TWIST1 encodes a highly conserved basic helix-loop-helix (bHLH) transcription factor that plays essential roles in craniofacial development. During murine embryogenesis, Twist1 is first expressed at E7.5 in the cranial mesoderm and later in neural crest-derived mesenchyme (Stoetzel, et al., 1995). Homozygous Twist1-null embryos die mid-gestation due to failure of cranial neural tube fold fusion (Chen and Behringer, 1995). While these embryonic phenotypes are severe, postnatal studies of Twist1 haploinsufficiency and tissue-specific knockouts have revealed persistent roles in skeletal morphogenesis. In particular, Twist1 is required for the development and patterning of both the cranial base and mandible, with conditional loss in the cranial mesoderm (CM) or neural crest (NC) lineages leading to cranial asymmetry, mandibular hypoplasia, altered cranial base angle, and abnormal suture and synchondrosis development (Bildsoe, et al., 2013; Funato, 2020; Bourgeois, et al., 1998; Hermann, et al., 2012; Bildsoe, et al., 2009).

Twist1 also functions as a repressor of premature osteogenic differentiation by antagonizing Runx2, a key osteoblast regulator (Bialek, et al., 2004). Through this mechanism, Twist1 coordinates the timing of intramembranous and endochondral ossification in the cranial vault and base, thereby maintaining suture patency and regulating calvarial bone development (Rice, et al., 2000). In Twist1-haploinsufficient mice, early ossification of the spheno-occipital and intersphenoid synchondroses has been observed, leading to cranial base shortening and facial asymmetry (Behr, et al., 2011; Funato, 2020).

In humans, heterozygous coding mutations in TWIST1 cause Saethre–Chotzen syndrome (SCS), a craniosynostosis disorder that includes coronal, lambdoid, or metopic suture fusion, maxillary hypoplasia, steep mandibular plane angles, and midfacial asymmetry (Johnson, et al., 1998; Oram and Gridley, 2005; El Ghouzzi, et al., 1999). Cephalometric studies of SCS patients consistently demonstrate reduced ramus height, a steep anterior cranial base, and altered mandibular plane angulation, phenotypes that strongly resemble those observed in individuals carrying the rs2189000 variant in our non-syndromic cohort (Alawneh, et al., 2023).

Although TWIST1 is highly expressed in cranial mesenchymal tissues, its transcriptional regulation may depend on inputs from more broadly expressed developmental factors. One such factor is PITX2, a paired-like homeodomain transcription factor essential for early craniofacial and dental morphogenesis. PITX2 influences epithelial-mesenchymal signaling interactions, regulates downstream effectors of osteogenesis, and is required for neural crest cell survival and migration (Evans and Gage, 2005; Chawla, et al., 2016; Liu, et al., 2003; Li, et al., 2025). While classically viewed as an epithelial gene, Pitx2 is also expressed in periocular and pharyngeal mesenchyme, where it participates in cranial base formation (Iwata, et al., 2012). In humans, PITX2 mutations cause Axenfeld-Rieger syndrome, characterized by maxillary hypoplasia and dental anomalies, as well as prominent craniofacial features (Semina, et al., 1996).

Notably, in silico motif analysis identified rs2189000 within a highly conserved PITX2 binding motif upstream of TWIST1. This raises the possibility that allelic variation at this site could disrupt transcriptional regulation of TWIST1 by PITX2, providing a functional mechanism for the observed association with skeletal morphology. While PITX2 and TWIST1 have not been extensively studied in a common regulatory context, their spatial and temporal co-expression during craniofacial development suggests potential convergence in gene networks governing postnatal morphogenesis (Iwata, et al., 2012; Minoux, et al., 2017).

This study builds on our earlier genetic association work to interrogate the regulatory mechanisms underlying craniofacial growth variation. Based on this rationale, we hypothesize that rs2189000 is a functional noncoding variant that affects TWIST1 expression by disrupting PITX2-mediated transcriptional activation, and Twist1 disruption contributes to altered craniofacial phenotype during postnatal development, leading to malocclusion-related phenotypes. Additionally, by integrating deep sequencing of the TWIST1 locus, functional characterization of enhancer activity, and phenotypic analysis of Twist1 conditional knockout mice, we aim to delineate a developmental pathway linking common genetic variation to postnatal craniofacial dysmorphology.

Craniofacial morphology results from the integrated actions of developmental patterning, transcriptional regulation, and tissue growth across embryonic and postnatal stages (Tseng and Crump, 2023). TWIST1, a basic helix–loop–helix transcription factor, plays an essential role in mesenchymal lineage specification and the maintenance of cranial suture patency during skull development (Bourgeois, et al., 1998; Johnson, et al., 1998; Chen and Behringer, 1995). While loss-of-function mutations in TWIST1 are known to cause SCS, a craniosynostosis disorder with cranial abnormalities (Wilkie, et al., 1995; Alawneh, et al., 2023), the mechanisms by which common noncoding variants at this locus contribute to craniofacial variation in non-syndromic populations remain largely unexplored.

This study builds upon our prior human genetic analysis, which first identified the common SNP rs2189000 located upstream of TWIST1 as significantly associated with a principal component of craniofacial shape variation, capturing mandibular ramus height, mandibular body length, and anterior cranial base inclination in our subjects. These traits are central to sagittal jaw discrepancies and are commonly implicated in malocclusion (Sanggarnjanavanich, et al., 2014; Gong, et al., 2016). Notably, the shape axis defined by PC3 parallels several cephalometric features characteristic of SCS, including mandibular deficiency and cranial base steepness, suggesting that common noncoding variants may phenocopy aspects of syndromic phenotypes in a dosage-sensitive manner.

In our extended cohort of 277 non-syndromic individuals with malocclusion, we identified rs2189000 as the top-associated SNP within the TWIST1 interval (da Fontoura, et al., 2015). This common variant demonstrated a strong association with craniofacial shape variation, particularly affecting mandibular ramus height, mandibular body length, and anterior cranial base inclination. Sequencing of the full TWIST1 interval did not identify any coding mutations, reinforcing the hypothesis that the phenotypic effects of rs2189000 arise from regulatory rather than structural gene changes. In addition to rs2189000, four other SNPs in strong linkage disequilibrium (D' = 1) also showed significant associations with PC3 (P < 0.001). Importantly, non-coding mutations were detected in TWIST1 across the sequenced cohort. Annotation of the non-coding variants (SNV) of TWIST1 was reported in patients with SCS, further supporting a regulatory role in craniofacial development and emphasizing the importance of screening the 5′ UTR in clinically diagnosed patients (Zhou, et al., 2018).

To explore the mechanism by which rs2189000 might alter TWIST1 expression, we performed in silico transcription factor binding analysis. We identified that rs2189000 disrupts a highly conserved consensus DNA binding motif for PITX2, a paired-like homeodomain transcription factor essential for craniofacial morphogenesis, asymmetry, and odontogenesis (Liu et al., 2003; Evans and Gage, 2005; Cao(Liu, et al., 2003; Evans and Gage, 2005; Cao, et al., 2013). Luciferase assays in five independent cell lines—including human embryonic palatal mesenchymal (HEPM) cells, demonstrated that PITX2 robustly activates transcription through the ancestral allele, but not the risk (G) allele of the enhancer. Chromatin immunoprecipitation was performed in HEPM cells homozygous for the rs2189000-A allele to assess PITX2 occupancy at the TWIST1 upstream enhancer and confirmed that PITX2 binds the rs2189000-containing enhancer sequence in HEPM cells, with approximately six-fold enrichment over IgG controls, directly validating the enhancer’s putative function in vitro. Although we did not perform allele-specific ChIP, the loss of PITX2-mediated enhancer activation was functionally validated through luciferase assays using the risk (G) allele.

While our in vitro assays support PITX2 as a transcriptional activator of TWIST1 through the rs2189000 enhancer, we acknowledge the well-established spatial separation between PITX2 and TWIST1 expression domains, specifically, PITX2 being enriched in craniofacial epithelia and TWIST1 in cranial mesenchyme. This has raised valid concerns about the physiological relevance of a direct regulatory interaction in vivo. However, transcriptomic and gene expression atlas data suggest that Pitx2 is transiently expressed in cranial neural crest and mesoderm-derived mesenchyme during early craniofacial development, particularly during palate and jaw formation, during stages when mesenchymal–epithelial crosstalk is prominent (Chawla, et al., 2016; Minoux, et al., 2017; Franco and Aranega, 2018; Evans and Gage, 2005). We interpret this not as evidence of a stable PITX2–TWIST1 axis, but as supporting a temporally restricted regulatory event that could plausibly influence enhancer activation during critical windows of craniofacial development. Thus, the PITX2 experiments presented here serve as a functional test of the rs2189000 variant’s putative regulatory potential in vitro. Future studies will be needed to identify whether PITX2 or other transcription factors engage this enhancer in vivo within the mesenchymal lineage.

To evaluate the physiological relevance of TWIST1 dysregulation, we examined the craniofacial phenotypic variations of a Mesp1 ^Cre conditional knockout model targeting cranial mesoderm. While the mandibular skeleton is derived primarily from cranial neural crest cells, the cranial base originates from cranial mesoderm and plays a critical role in establishing mandibular position through its influence on occlusal plane angulation, condylar orientation, and vertical facial proportions. Prior studies confirm that Mesp1^Cre efficiently targets early cranial mesoderm without affecting epithelial or neural crest-derived domains (Bildsoe, et al., 2013; Chen, et al., 2007). Our decision to employ Mesp1 ^Cre mediated deletion of Twist1 was therefore based on the hypothesis that early perturbations in cranial base development could drive secondary alterations in mandibular shape and spatial orientation. This rationale is supported by our human genotype–phenotype analyses, in which the rs2189000 risk allele was associated with reduced ramus height, increased mandibular body length, and a steep anterior cranial base, features consistent with cranial base–driven Class III skeletal relationships. Although Wnt1-Cre–mediated deletion of Twist1 has been shown to cause profound disruptions in mandibular development, the Mesp1 ^Cre model enabled assessment of more subtle, postnatal morphologic changes spanning both cranial base and mandibular structures. This approach thus captures a distinct axis of craniofacial variation that is directly relevant to the pathogenesis of malocclusion.

Postnatal morphometric analyses of high-resolution microCT scans of mice heads at postnatal days 14 and 21 revealed significant genotype-dependent craniofacial differences. PCA of three-dimensional landmark data demonstrated separation from wild-type mice along PCs 1 and 2, reflecting differences in cranial base angle, cranial vault doming, and mandibular projection. Wireframe comparisons further highlighted biologically meaningful variation in these axes. While scatterplots showed some group overlap, wireframes revealed consistent shape trends distinguishing genotypes. Notably, homozygous mutants exhibited a domed cranial vault, shortened cranial base, and posteriorly rotated mandible. Specifically, wild-type mice exhibited flatter cranial vaults, longer cranial bases, and more prominent mandibular extension, whereas homozygous mutants presented with domed calvaria, shortened cranial bases, and posteriorly rotated mandibles. These differences are visualized in wireframe overlays that contrast the extremes of PC1 between genotypes, despite partial overlap in scatterplots. Notably, the craniofacial variations observed in mutant mice closely resemble those observed in the human cohort, underscoring the conserved craniofacial phenotypic effects of TWIST1 dysregulation across species.

Linear measurements confirmed these findings, demonstrating statistically significant reductions in skull length, cranial base length, palatal dimensions, and mandibular body length in mutant mice compared to controls. These phenotypes closely mirror cephalometric patterns documented in SCS patients, who present with shortened ramus height, steep mandibular planes, and anterior cranial base defects (Lonsdale, et al., 2019; Evans and Christiansen, 1976).

A particularly novel finding was the premature ossification of cranial base synchondroses in mutant mice. In wild-type animals, the intersphenoid (ISS) and spheno-occipital (SOS) synchondroses remain patent into adolescence, serving as critical growth centers for anterior cranial base elongation (Vora, et al., 2015; Hallett, et al., 2025; Funato, 2020; Wei, et al., 2017). However, Twist1-deficient mice exhibited partial or complete fusion of these synchondroses by P21, as revealed by microCT and Alizarin Red staining. These fusions were accompanied by cranial vault doming, snout asymmetry, and malocclusion, including the distinctive “elephant tusk” appearance previously described in Twist1 deletion models (Bourgeois, et al., 1998). Although incisor overgrowth is a common confounder in rodent malocclusion models, we attribute these skeletal deformities to underlying asymmetries and premature fusion events that disrupt postnatal growth trajectories.

These findings highlight the capacity of common regulatory variants to contribute to complex traits and bridge the gap between syndromic and non-syndromic presentations. While rs2189000 does not cause disease, it contributes to a spectrum of craniofacial variation that overlaps with TWIST1 loss-of-function syndromes. This concept of “intermediate phenotypes” is particularly informative for understanding gene dosage effects in craniofacial development (Sharma, et al., 2013). Moreover, our results reinforce the importance of considering noncoding variation in genetic studies of malocclusion and jaw development. Despite most clinical genetics focusing on coding mutations, regulatory elements may underlie unexplained cases of craniofacial asymmetry or growth anomalies.

Our study also illustrates the value of integrating genomic, functional, and phenotypic datasets across species. While the PITX2–TWIST1 regulatory axis was validated in vitro, future studies should test in vivo enhancer occupancy using ChIP-seq or CRISPRi strategies in embryonic mesenchymal populations. Additionally, given that the TWIST1 locus is enriched for regulatory elements based on FANTOM5 and ENCODE annotations, further chromatin conformation and enhancer–promoter interaction mapping will be necessary to define the whole regulatory architecture underlying craniofacial growth.

This study has limitations. While the knock-in model did not contain an rs2189000 directly, our integrated data, combining genotype–phenotype associations, allele-specific luciferase assays, and conditional Twist1 deletion, support the functional relevance of this site. Future studies employing precise genome editing to model this variant in vivo will be necessary to definitively test its phenotypic consequences.

Furthermore, the mouse sample size, especially in homozygous mutants, restricts statistical power. While appropriate statistical tests were applied, the qualitative nature of morphometric findings and the variability across individuals underscore the need for larger cohorts. Additionally, the PITX2–TWIST1 regulatory axis was validated in vitro; future studies should test in vivo enhancer occupancy using ChIP-seq or CRISPRi strategies in embryonic mesenchymal populations. Given that the TWIST1 locus is enriched for regulatory elements based on FANTOM5 and ENCODE annotations, further chromatin conformation and enhancer–promoter interaction mapping will be necessary to define the whole regulatory architecture underlying craniofacial growth.

In conclusion, we provide evidence that a common noncoding variant upstream of TWIST1 alters enhancer activity in vitro experiments, suggesting a possible regulatory mechanism contributing to variation in jaw relations and cranial base morphology These results extend the functional impact of TWIST1 beyond syndromic contexts and demonstrate that noncoding regulatory variation can influence craniofacial structure in non-syndromic individuals. This work offers a framework for future studies aiming to unravel the complex genetic underpinnings of malocclusion and craniofacial diversity.