Characterization of normal facial features and their association with genes

ABSTRACT Background: Craniofacial morphology has been reported to be highly heritable, but little is known about which genetic variants influence normal facial variation in the general population. Aim: To identify facial variation and explore phenotype-genotype associations in a 15-year-old population (2514 females and 2233 males). Subjects and Methods: The subjects involved in this study were recruited from the Avon Longitudinal Study of Parents and Children (ALSPAC). Three-dimensional (3D) facial images were obtained for each subject using two high-resolution Konica Minolta laser scanners. Twenty-one reproducible facial soft tissue landmarks and one constructed mid-endocanthion point (men) were identified and their coordinates were recorded. The 3D facial images were registered using Procrustes analysis (with and without scaling). Principal Component Analysis (PCA) was then employed to identify independent groups ‘principal components, PCs’ of correlated landmark coordinates that represent key facial features contributing to normal facial variation. A novel surface-based method of facial averaging was employed to visualize facial variation. Facial parameters (distances, angles, and ratios) were also generated using facial landmarks. Sex prediction based on facial parameters was explored using discriminant function analysis. A discovery-phase genome-wide association analysis (GWAS) was carried out for 2,185 ALSPAC subjects and replication was undertaken in a further 1,622 ALSPAC individuals. Results: 14 (unscaled) and 17 (scaled) PCs were identified explaining 82% of the total variance in facial form and shape. 250 facial parameters were derived (90 distances, 118 angles, 42 ratios). 24 facial parameters were found to provide sex prediction efficiency of over 70%, 23 of these parameters are distances that describe variation in face height, nose width, and prominence of various facial structures. 54 distances associated with previous reported high heritability and the 14 (unscaled) PCs were included in the discovery-phase GWAS. Four genetic associations with the distances were identified in the discovery analysis, and one of these, the association between the common ‘intronic’ SNP (rs7559271) in PAX3 gene on chromosome (2) and the nasion to mid-endocanthion 3D distance (n-men) was replicated strongly (p = 4 x 10-7). PAX3 gene encodes a transcription factor that plays crucial role in fetal development including craniofacial bones. PAX3 contains two DNA-binding domains, a paired-box domain and a homeodomain. The protein made from PAX3 gene directs the activity of other genes that signal neural crest cells to form specialized tissues such as craniofacial bones. PAX3 different mutations may lead to non-functional PAX3 polypeptides and destroy the ability of the PAX3 proteins to bind to DNA and regulate the activity of other genes to form bones and other specific tissues. Conclusions: The variation in facial form and shape can be accurately quantified and visualized as a multidimensional statistical continuum with respect to the principal components. The derived PCs may be useful to identify and classify faces according to a scale of normality. A strong genetic association was identified between the common SNP (rs7559271) in PAX3 gene on chromosome (2) and the nasion to mid-endocanthion 3D distance (n-men). Variation in this distance leads to nasal bridge prominence