Introduction

Human evolution studies identified that an adaptive variant of ectodysplasin receptor, EDARV370A (rs3827760), had undergone strong positive selection in East Asia (Grossman et al. 2010, 2013; Sabeti et al. 2007). The derived 370A allele has extremely high prevalence in East Asians and Native American populations, but it is almost absent in Europeans and Africans (Bryk et al. 2008; Grossman et al. 2010, 2013; Sabeti et al. 2007). This gain-of-function allele alters a highly conserved amino acid in the death domain of EDAR, and it influences the binding efficiency of ectodysplasin to EDAR thereupon affects the function of downstream pathways (Cluzeau et al. 2012). It is well established that mutations on EDAR, along with other genes in EDA pathway, are causing hypohidrotic ectodermal dysplasia (HED), a rare genodermatosis characterized by sparse hair (hypotrichosis), abnormal or missing teeth (hypodontia) and inability to sweat (anhidrosis) (Cluzeau et al. 2011). This led the researchers to hypothesize that the EDARV370A allele might be selected for functionally affecting certain ectodermal-related phenotypes. Indeed EDARV370A was found to be significantly associated with hair thickness (Fujimoto et al. 2008a, b), hair straightness (Tan et al. 2013), incisor shoveling (Kimura et al. 2009; Park et al. 2012; Tan et al. 2014) and active sweat gland density (Kamberov et al. 2013). Transgenic mice studies further confirmed that EDARV370A is the responsible causal functional variant (Kamberov et al. 2013). The strong selection and the pleiotropic effect of EDARV370A on ectodermal-derived phenotypes make it a very special case in human evolution.

Intriguingly, HED is known to affect facial appearance in the patients (e.g. prominent forehead, thick lips) (Cluzeau et al. 2011). Facial characteristics are largely shaped by skull bones, which are also ectodermal derived. Therefore, it is conceivable but remains to be tested whether EDARV370A could also affect facial morphology to some extent. In this study, we aim to explore the relation between EDARV370A and facial morphology by conducting an association study in the Uyghur population.

All of the previous association studies on EDARV370A were carried out in East Asian populations (Fujimoto et al. 2008a, b; Kamberov et al. 2013; Kimura et al. 2009; Park et al. 2012; Tan et al. 2013, 2014). One conspicuous limitation is that the ancestral allele, 370 V, is at a very low frequency in most East Asian populations (Sabeti et al. 2007). It is, therefore, hard to distinguish the genetic model of EDARV370A on its affected phenotypes. The low diversity of EDARV370A also leads to low power of the association studies in East Asian populations, making it hard to explore a large spectrum of potentially affected ectodermal-related phenotypes (with lesser effects than the reported ones). In order to overcome the above limitations, we choose to conduct a study in the Uyghur population, a typical admixed population with genetic contribution from both East Asian and European ancestry populations—close to a 50:50 proportion. Population genetic studies showed that the admixture event in Uyghurs was likely to happen more than 100 generations ago. Therefore, the whole Uyghur population is thoroughly admixed, and the ancestry proportion of most Uyghur individuals are close to a 50:50 ratio (Xu et al. 2008, 2009; Xu and Jin 2008). Such an admixture property makes Uyghur an extremely ideal population to study the phenotypic effects of EDARV370A, as (1) both the 370A and 370 V alleles have a decent frequency in the population; and (2) the phenotypic diversity is also considerably high in the population.

By collecting the EDARV370A genotype and a number of ectodermal-derived phenotypes in 1027 Uyghur individuals, we aim to detect the potential facial characteristics which may be influenced by EDARV370A, and further expand our understanding of the pleiotropic nature of EDARV370A.

Materials and methods

Samples

This study recruited 1027 Uyghur samples, who were undergraduate students from Xinjiang Medicine University, including 393 males and 634 females. The age of the samples ranged from 17 to 25 years. The study only recruited individuals whose self-reported Uyghur origins traced back to all four grandparents. The research was conducted with the official approval from the Ethics Committee of Fudan University, Shanghai, China. All the participants had provided written consents.

For replication purpose, genotype and phenotype data from two other populations (323 samples in Mongolian and 973 samples in Han Chinese) were also included in this study. More details of those samples can be found in a previous study (Tan et al. 2013).

Genotyping

A qualified physician collected peripheral blood samples from each of the participants. DNA samples were extracted from the blood using the GENEray™ DNA extraction kit. The EDARV370A SNP (rs3827760) was genotyped using the SNaPshot Multiplex System. 24 Ancestral Informative Markers (AIMs) were also genotyped to calculate the ancestral contribution for each of the Uyghur samples. The detailed selection method of the AIMs is described in the “Statistical analysis” section. The 24 AIMs were genotyped using the LDR Multiplex System. Genotype calling was performed by GeneMapper v2.0.

Phenotyping

After signing a content form, each of the participants first completed a questionnaire including the information of age, gender, ethnic background, etc. We then collected phenotypic information of the participant by observing or measuring the following ectodermal related phenotypes (Table 1).

Table 1 List of assessed phenotypes by observation
Full size table

Facial characteristics We assessed 16 facial characteristics according to the human body assessment methods (Wu et al. 2010). The assessed characteristics include upper eyelid fold (right and left), epicanthic fold (right and left), type of forehead, type of cheekbone, type of chin, nasal root height, nasion depression, nasal bone, nasal-tip direction, nasal-tip shape, nasal base direction, nostril form, upper lip thickness and shape of earlobe. The rating scales and the frequencies of the facial characteristics are shown in Table 1.

Incisors shoveling and double shoveling Dental plaster models were applied to assess incisors shoveling and double incisors shoveling of the subjects. The rating criteria of incisors shoveling and double incisors shoveling follow the Arizona State University Dental Anthropology System, which includes seven degrees (0–2 indicates no incisors shoveling or double incisors shoveling, while 3–6 indicates incisors shoveling or double incisors shoveling) (Scott and Turner 1997; Turner et al. 1991).

Hair straightness The degree of hair straightness was rated on a three-point scale: straight, wavy, or curly, following the method used in previous studies (Medland et al. 2009; Tan et al. 2013). However, curly hair is scarce in the Uyghur sample, so we merged curly and wavy into one scale—“non-straight” in data analysis.

Statistical analysis

Ancestry analysis To ensure that the association analysis is not confounded by the ancestry of the samples, we inferred the ancestral proportion of each study sample and corrected the potential effect. First, we selected 24 Ancestral Informative Markers (AIMs) based on International HapMap Project Phase III (CHB as the East Asian ancestry and CEU as the European ancestry), under two criteria: (1) SNPs highly differentiate between East Asia and Europe (Fst >0.8); and (2) inter-marker distance is great enough (>0.01 cM) to avoid any substantial linkage disequilibrium. We then use STRUCTURE program to estimate the ancestral proportion for each Uyghur sample based on the 24 AIMs, using the CHB and CEU from HapMap Project Phase III as ancestral populations (Falush et al. 2003, 2007; Hubisz et al. 2009; Pritchard et al. 2000). An admixture-model with K = 2 was assumed and the program run with 5000 iterations and 10,000 burn-ins to convergence.

Association analysis General linear model was applied to estimate the association between EDARV370A and collected phenotypes. Gender and the ancestral proportion were taken as covariates in our study. We found that age was not a significant factor in any models which might be due to that all the samples were from a young group (17–25 years old), so it was not considered as a covariate in the present models. Frequency bar charts along with results of Fisher exact test (for qualitative traits), or average histograms (for quantitative traits) were also applied to demonstrate the simple association between EDARV370A and significantly associated phenotypes obtained through general linear model.

Partial least square path model analysis Partial least square path model was applied to present the pleiotropic effect of EDARV370A on detected ectodermal-derived phenotypes (Tenenhaus et al. 2005). A full path model was constructed among EDARV370A and the phenotypes while controlling for gender and ancestral proportion. The bootstrap confidence interval test (bootstrap resampling times = 1000, significance level = 0.05) was applied to test the significance of each path coefficient.

Results

For EDARV370A, the frequencies of three genotypes in the Uyghur population are 0.38 (370 V/370 V), 0.47 (370 V/370A) and 0.15 (370A/370A). The minor allele (370A) frequency is 0.39, showing a much higher diversity than other East Asian populations. All the genotyped SNPs including EDARV370A and 24 ancestral informative markers (AIMs) have passed Hardy–Weinberg equilibrium test (Table S1).

In a general linear model, we find significant associations between EDARV370A and a number of facial characteristics, including type of chin, shape of earlobe, epicanthic fold, type of forehead and upper lip thickness (Table 2). After Bonferroni correction, the associations between EDARV370A and type of chin and shape of earlobe are still significant. Additionally, we perform ancestral proportion correction for these associations and obtain similar results (Table S2).

Table 2 The association between EDARV370A and 16 facial characteristics
Full size table

Significant association between 370A allele and triangular shape of earlobe

In the Uyghur population, individuals carrying more 370A alleles are less likely to have circular shape of earlobe and more likely to have triangular shape of earlobe (P = 3.91 × 10 −4, Fisher exact test; Fig. 1b). In a general linear model corrected for gender, EDARV370A is significantly associated with shape of earlobe and fitted well in an additive model (P = 3.64 × 10 −6; Table 2). We confirm this association in two other East Asian populations (P = 4.97 × 10 −5 in Han Chinese, P = 9.60 × 10 −3 in Mongolian, Fisher Exact test; Fig. 1c, d). Furthermore, we have collected the frequency data of the EDARV370A allele as well as the shape of earlobe in 19 East Asian populations (Table S3). A linear regression analysis shows that the frequency of triangular earlobe is positively correlated with the frequency of 370A at the population level (R 2 = 0.27, P = 0.02; Fig. 1e).

Fig. 1
figure 1

EDARV370A is significantly associated with the shape of earlobe. a Different shapes of earlobe (circular, square and triangular); b the association between EDARV370A and shape of earlobe in the Uyghur population (P = 3.91 × 10 −4); c The association between EDARV370A and shape of earlobe in a Han population (P = 4.97 × 10 −5); d The association between EDARV370A and shape of earlobe in a Mongolian population (9.60 × 10 −3); e The linear regression on frequencies of triangular earlobe and 370A frequencies in 19 Asian populations

Full size image

Significant association between 370A allele and retruded chin

In the Uyghur population, we also find that individuals carrying 370A alleles are more likely to have retruded chin and less likely to have protruded chin (P = 4.75 × 10 −5, Fisher Exact test; Fig. 2b). In a general linear model corrected for gender, the association between type of chin and EDARV370A remains significant (P = 9.23 × 10 −5, Table 2). Compared with individuals carrying 370A homozygotes, the individuals carrying 370 V alleles are more inclined to have protruded chins. We further confirm the association between EDARV370A and type of chin in a Han Chinese population (P = 0.03; Fig. 2c).

Fig. 2
figure 2

EDARV370A is significantly associated with the type of chin. a Different types of chin (protrude, normal and retrude); b The association between EDARV370A and type of chin in the Uyghur population (P = 4.75 × 10 −5); c The association between EDARV370A and type of chin in a Han population (general linear model: P = 0.03)

Full size image

Significant associations of 370A allele on hair straightness and incisor shoveling best fitting in an additive genetic model

Previous studies found that EDARV370A is significantly associated with hair straightness and incisors shoveling in East Asian populations (Kimura et al. 2009; Tan et al. 2013, 2014). Our data on the Uyghur population confirm the trend that individuals with the 370A allele tend to have straight hair (P = 4.07 × 10 −16, Fisher Exact test; Fig. 3b), shovel-shaped incisors (P = 1.76 × 10 −6, Fisher exact test; Fig. 3c) and double shovel-shaped incisors (P = 5.15 × 10 −12, Fisher exact test; Fig. 3d). In a general linear model corrected for gender, the associations remain significant between 370A and straight hair (P = 3.99 × 10 −16), incisors shoveling (P = 1. 20 × 10 −7) and double incisors shoveling (P = 1.86 × 10 −12). Moreover, our results clearly show that the associations fit best in an additive model (Table S4).

Fig. 3
figure 3

EDARV370A is associated with hair straightness, incisors shoveling and double incisors shoveling, fitting well in an additive genetic model. a The rating scales of hair straightness, incisors shoveling and double incisors shoveling. For incisors shoveling and double shoveling, 0, 1 defined as normal while 2–6 defined as incisors shoveling or double shoveling. b The association between EDARV370A and hair straightness (P = 4.07 × 10 −16). c The association between EDARV370A and incisors shoveling (P = 1.76 × 10 −6). d The association between EDARV370A and double shoveling (P = 5.15 × 10 −12)

Full size image

Systematical associations of EDARV370A and weakly related ectodermal-derived characteristics

Pairwise correlation analyses show that the eight ectodermal-derived characteristics detected in our study are weakly correlated with each other (Fig. 4a). Furthermore, we apply a full partial least square path model to systematically learn the relation among EDARV370A and eight ectodermal-derived characteristics. The result shows that EDARV370A is significantly associated with hair straightness (path coefficient = −0.27), incisors shoveling (path coefficient = 0.12), double incisors shoveling (path coefficient = 0.20), shape of earlobe (path coefficient = 0.10), type of chin (path coefficient = 0.11), upper lip thickness (path coefficient = 0.07) and type of forehead (path coefficient = 0.08) in the path model (Fig. 4b; Table S5). Each path coefficient represents the effect of EDARV370A on each studied characteristic. The path model also indicates that there are weak relations among eight ectodermal-derived characteristics, where only 4 out of 28 path coefficients among these characteristics are significant (Table S5). Those non-significant path coefficients are not presented in Fig. 4b, while more information about the path model is included in supplement Table S5. These results confirm that EDARV370A systematically influences a series of weakly related ectodermal-derived characteristics.

Fig. 4
figure 4

The correlation among eight ectodermal-derived phenotypes and the partial least square path model among them and EDARV370A. a The correlation among eight ectodermal-derived phenotypes which are associated with EDARV370A. a Upper panel are the correlation coefficients among eight phenotypes; b lower panel shows the P values of significance test for the corresponding correlation coefficients. There is weak correlation among the eight ectodermal-derived phenotypes. b The partial least square path model between EDARV370A and eight ectodermal-derived phenotypes. Only path coefficients that are significant by bootstrap confidence interval testing (bootstrap resampling times = 1000, significance level = 0.05) are presented in this panel

Full size image

Discussion

EDARV370A and facial morphology

Bones are mostly mesodermal derived, with the exception of skull bones, which are mostly ectodermal derived. Since EDARV370A is already known to play a crucial role in the ectodermal-related phenotypes, we hypothesize that facial characteristics—largely affected by skull bones—could be associated with EDARV370A as well. In this Uyghur study, we did find at least two facial characteristics, the type of chin and the shape of earlobe, are significantly associated with EDARV370A. Both associations have been independently replicated in other East Asian populations and, therefore, should not be artifacts. For the type of chin, 370A carriers tend to have retruded chins, while 370 V carriers are more likely to have protruded chins. The worldwide distribution does indicate that protruded chin is prominent in Europe, matching well with the distribution of EDARV370A. The type of chin might be a combined result from the shape of mandible and teeth (Aiello and Dean 1990; Fuerle 2008; Schwartz and Tattersall 2000).

The shape of earlobe is also largely affected by skull bone. The external ear derives from tissue in the area of the contiguous edges of the first pharyngel pouch where the mandibular and hyoid arches join (Hunter and Yotsuyanagi 2005; Moldenhauer 1877). Specifically, the earlobe is believed to derive from one or two hillocks in this region (Park 1999). In our results, we find that for the shape of earlobe, 370A carriers tend to have more triangle earlobes. Kaustubh and colleagues also independently found this association in Latin American populations (Adhikari et al. 2015). Looking at the worldwide distribution of shape of earlobe, it may not be a coincidence that triangular earlobes mostly present in East Asia, where 370A is predominant; while circular earlobes mostly present in Africa and Europe, where 370 V prevails (Bhowmik 1971; Nakamura et al. 1995; Overfield and Call 1983). It is also worth noting that the shape of earlobe is long thought to be a polygenetic trait with more than one gene involved. Indeed our results suggest that EDARV370A only explains a small portion of the variance of the phenotype. Future studies at the genome-wide level shall be able to discover other genes also responsible for the shape of earlobe phenotype.

Besides the type of chin and the shape of earlobe, there are also several facial characteristics (epicanthic fold, type of forehead and upper lip thickness) showing nominal association with EDARV370A. The association between EDARV370A and type of forehead and upper lip thickness are also confirmed in partial least square path model analysis. Interestingly, prominent forehead and thick lips are two most conspicuous facial characteristics in the HED patients (Cluzeau et al. 2011). Although we do not find significant associations of these two characteristics in our replicate populations, it would be interesting to confirm the results in further studies with increased samples size.

Pleiotropic effect of EDARV370A

EDARV370A arose in East Asia more than 30,000 years ago and since then had undergone strong positive selection in East Asia (Kamberov et al. 2013). The estimated selective coefficient of EDARV370A was found to be extremely high in various studies (Bryk et al. 2008; Fujimoto et al. 2008a; Kamberov et al. 2013; Sabeti et al. 2007). This could be due to the pleiotropic nature of EDARV370A, as many related traits could be the subjects under selection. It is even possible that at different time of human evolution, different traits could be the subjects under selection, collectively driving the 370A allele into high frequency in East Asia. Unfortunately, also because of its pleiotropic nature, the selective pressure on this unique variant remains a mystery.

While it is difficult to determine the adaptive fitness of EDARV370A during its strong positive selection in East Asia, exploring more potential traits which are influenced by EDARV370A may provide clue to its selective pressure. In previous studies, the EDARV370A allele was reported to be associated with hair straightness, hair thickness, sweat gland density, incisor shoveling and other dental morphologies (Fujimoto et al. 2008a, b; Kamberov et al. 2013; Kimura et al. 2009; Mou et al. 2008; Park et al. 2012; Tan et al. 2013, 2014). In this study, we further discover that EDARV370A is associated with a number of facial characteristics, particularly the type of chin and the shape of earlobe. This study indicates that EDARV370A may play a role in the development of facial morphology. With the new associated phenotypes reported in this paper, the pleiotropic nature of EDARV370A is further extending.

By applying a partial least square path model, we confirm that EDARV370A is systematically affecting a number of ectodermal-derived characteristics. However, these characteristics are weakly correlated. This might because that EDARV370A mainly plays roles in early embryo development, and subsequently various genes and environmental factors affected the development of these characteristics during late stage of development.

Use of the admixed Uyghur population in association studies

The Uyghur population used in this study is a typical admixed population between East Asia and Europe. Estimations of the admixture time range from 800 to more than 2000 years ago (Xu et al. 2008, 2009; Xu and Jin 2008). Therefore, the Uyghur admixture happened much earlier than the admixture events of African Americans or the Mestizos. The admixture happened quite thoroughly in the population at the individual level, making most of the Uyghur individuals ~ 50:50 admixture of their East Asian and European ancestries (Xu et al. 2008, 2009; Xu and Jin 2008). This unique feature helps to reduce substructure within the Uyghur population and makes it an ideal population for association studies. In this study, since EDARV370A has an extremely contrasting frequency between East Asia and Europe, one could raise the concern that the association between EDARV370A and a related phenotype may result from a real association between the ancestry proportion and the phenotype. We cautiously add the ancestry proportion as a co-variant in all of the association analysis to exclude this possibility. The association results before and after corrections remain largely the same, making it clear that ancestry proportion is not a confounding factor in the association results reported in this study. Our unpublished results using genome-wide SNP markers in a subset of the samples also confirm that substructure is not much a concern for the well-admixed Uyghur population. Contrarily, because of the admixture feature, the Uyghur population turns out to be an excellent subject to study the phenotypic effect of EDARV370A. Its high genetic diversity (minor allele frequency 0.39) and phenotypic diversity bring at least ten times more power than a study of same size in Han Chinese (simulation results not shown).

Interestingly, by studying the Uyghur population, one can infer that EDARV370A is most likely not to be under any selection at all for its history of 800–2000 years. The 370A allele, selected in most of the East Asian populations, is actually the minor allele in the Uyghurs, with a frequency of 39 %. Had it been under selection in the Uyghurs, one would expect the allele to also raise to a high frequency, significantly higher than 50 %, the genome-wide average contribution from East Asia. The absence of selection in Uyghurs could shed light on the intriguing case of EDARV370A. It also raises an interesting question as to whether EDARV370A is no longer under positive selection in East Asia in the recent human history, or if EDARV370A is just not under positive selection in the environment that the Uyghurs live in.