Ontogeny of the facial phenotypic variability in Mexican patients with 22q11.2 deletion syndrome

The patient group included individuals without palatal abnormalities or cleft lip and palate surgery to avoid increasing facial variability due to the presence of this disorder. This group included 37 children (eighteen girls and nineteen boys) between the ages of 2 and 17 years of both genders and from Mexico City and the surrounding states (State of Mexico, Hidalgo, and Queretaro) with clinical diagnoses of VCFS and 22q11.2DS, as established by fluorescent in situ hybridization (FISH) using the Tuple1 probe (Vysis®). Patients with CHD were recorded, as well as their type of CHD. The patients were compared with a control group that included 200 age-matched individuals mainly from Mexico City and the State of Mexico. In order to explore the morphological differences through ontogeny, both groups were subdivided into different age groups that roughly reflect the major stages of development (i.e. early childhood, middle childhood, and adolescence), and, as far as possible, with balanced sample size (Table 1). This study was reviewed and approved by the local ethics committees. Informed consent was obtained from both groups.

Frontal facial photographs were taken of the individuals with a standardized protocol that included the placement of the faces according to the Frankfort plane and image capture using a Sony® Alpha 33 digital camera with a Sony® 50 mm lens placed at a constant distance of 1.5 m from the face, with a scale placed at ear height.

Using the program TPSDIG v.2.16, 28 landmarks were placed, including six over the sagittal line and 22 bilateral landmarks [37] (Table 2, Fig. 1). The resulting coordinates were analyzed using geometric morphometric (GM) and multivariate analyses. GM analysis is a technique used for the capture, analysis, and visualization of shape change, with advantages over traditional morphometric techniques [38]. Shape information is extracted by Procrustes superimposition, which removes variations in location, scale and rotational effects from coordinates in 2D or 3D but retains size information (i.e., centroid size) for use in subsequent analyses. Centroid size is defined as the square root of the summed squared distances of all landmarks from their centroid; it is used as the scale of the landmark configuration [39]. Subsequent to this adjustment, the standardized coordinates are analyzed with multivariate statistical methods. During the analysis, the geometric relationships among variables are preserved, thereby facilitating the visualization of shape change. Together, these features make GM analysis a powerful technique for the assessment of morphological variation.

The analysis was performed by taking into consideration the bilateral symmetry exhibited by the face, i.e. Procrustes superimposition was performed with the original and mirrored configurations of each set of landmarks, allowing the morphological variation related with asymmetry to be ruled out, as it is of no interest in this study. In this manner, a totally symmetrical morphology was analyzed [40].

First, we carried out a principal component analysis (PCA) to visualize the main axis of facial variability within our sample. The analysis generates a new set of variables called principal components (PCs) that are functions of the covariances among the original variables and that account for most of the sample variation. We considered only those PCs that explained above 10% of the total variance.

In order to explore the contribution of development (shape change with age) to the facial variation within our sample, we used a two-way multivariate analysis of variance (two-way MANOVA) to compare facial shape differences using the patient-control status and the age group (i.e. G1, G2, G3) as factors. Moreover, the developmental trajectory was analyzed in detail as multivariate trajectory in phenotype space. In this analysis, the vector that connects the phenotypic means between age groups is determined for patients and controls, and its attributes (i.e. magnitude of phenotypic vector, vector direction, vector shape) are statistically examined pairwise by permutation test with 1000 random permutations [41]. Furthermore, we explored the relative change differences through ontogeny, by computing the Euclidean distance between the mean shapes of each age group obtained in the latter analysis. Finally, we used Goodall’s F-test to evaluate mean shape differences between patients and controls at each age group, using a resampling procedure with 400 random permutations to assess its significance.

To assess the contribution of the growth process (size change with age), we performed a two-way analysis of covariance (two-way ANCOVA) with facial size (i.e. centroid size in cm) as the dependent variable, and the patient-control status and age as factors. The significance of the null model was assessed using 400 random permutations. In addition, the growth trajectory between groups was analyzed by linear regression of the log centroid size (log CS) against individual age (as continuous variable). Finally, we compared the relative scaling in facial size between groups as percentage, with the estimate of the ratio of patients’ size at a given age relative to controls in the same stage.

Allometry (shape change with size) was assessed using multivariate analysis of covariance (MANCOVA) of the facial shape, using log CS and patient-control status as factors. The statistical significance was evaluated using 400 random permutations. Finally, the allometric trajectory was analyzed by a multivariate regression of shape variables against log CS and of predicted values against log CS. The regression score can be interpreted as the shape variable that is most strongly associated with log CS [42], while the predicted shape represents the first principal component of the predicted values calculated from the multivariate regression model [43]. All morphometric and statistical analyses were performed using the R package [44], ‘Geomorph’ version 3.0.3 [45] .

It has been stated that the expression of facial characteristics in the context of 22q11.2DS may vary depending on the ethnic group [19]. For example, African and Afro-American patients have shown subtle facial dysmorphism [17, 18], with auricular abnormalities reported as more frequent than other features [53], whereas the phenotypes of Chinese and Chilean patients have been described as similar to that of Caucasian patients [54]. However, a study in European patients has observed some differences from the typical Caucasian phenotype, where patients exhibit downslanted palpebral fissures [27]. Likewise, a long face has not been associated with Japanese patients [55], although this feature has been frequently observed in the syndrome. Our results obtained in Mexican population show that (Fig. 4a) there are differences between the mean shape of patients and controls, differences consistent with the characteristics reported as typical for Caucasian population [56]. That is, patients have longer faces than controls, short and upslanting palpebral fissures, hypertelorism, shorter philtrum, downward labial commissures and short ears. Nevertheless, the nasal differences are not as marked as the other facial differences in this overall comparison, maybe due to the fact that the characteristic nose previously reported includes lateral variation (e.g. a bulbous nasal tip) which was not analyzed in this study; another likely factor is that the nasal trait that discriminates between groups in this study (i.e. hypoplasia of the nasal wings) is only observed in the older age group (Fig. 4d).

Also, facial size has been reported to be an important criterion included in the definitions and descriptions of different syndromes [57]. However, to the best of our knowledge, there are few studies that consider the analysis of facial size in 22q11.2DS and its impact on facial shape. For example, in a dense surface analysis performed in individuals pertaining to a similar age group to that of the present study, a slight increase in overall facial size was found in patients relative to controls [48]. Similarly, in the study by Sinderberry et al. [58], the facial size is one of the traits that statistically differentiate the two subtypes related to mental health problems in individuals with 22q11.2DS. Our results showed that patients have smaller facial size than controls when all the ontogenetic stages were taken into account, and also within each age group (see below).

In addition to explore the mean facial phenotype and size related with 22q11.2DS, we present a detailed description of the morphological variation of Mexican patients. We found that the patterns of facial covariation between patients and controls are not the main source (i.e. PC1) of variation within our sample (Fig. 2). Prasad et al. [48] reported a similar finding using 3D facial surfaces of affected and unaffected siblings; nonetheless, they found significant distinction between patients and controls in the second PC. Altogether, these results could support claims on the difficulty of recognizing the typical facial morphology of patients based on qualitative inspection only [14], and contrast with the facial variability observed for other syndromes. For example, Goodwin et al., studied the facial dysmorphology related to hypohidrotic ectodermal dysplasia using 3D imaging, and found that patients and controls differed along the first PC [59].

The present study also confirms the range of morphological variation that has been previously reported [5, 11, 14, 15, 27, 46, 56]. The facial features that showed greater morphological variation in our patients included the mouth, which was observed to be small and arc-shaped, with thinner lower lip compared to the upper lip in some patients, while in others it was found to be wide with the thinning of both lips. In addition, other variable features were observed, for instance, the palpebral fissures, which were upslanting and oblique or normal, and the implantation of the ear, which was low or normal. Finally, the facial thirds showed variability in different proportions.

These results, along with the ontogenetic-related changes would benefit the diagnosis of 22q11.2DS since it has been reported as more likely when a systematic evaluation that considers the presence of three or more of these characteristic features is performed [14]. Moreover, given that these facial traits represent the most descriptive characteristics for this group of individuals, they could be used for clinical training [32, 60]. Finally, the description of the facial phenotypic variation in patients may also be useful for the identification of greater number of individuals, thereby improving the comprehensiveness of their medical care and genetic counseling of the families.

We explored whether part of the morphological differences found between patients and controls are related to ontogenetic processes, i.e. development, growth, and allometry. According to previous studies, facial appearance in 22q11.2DS varies among ontogenetic stages. For example, some studies have mentioned that adults have more subtle manifestations [25‐27], while others have reported that newborns and patients in early childhood are more difficult to diagnose [61‐63]. Our results showed a pattern of morphological differentiation between patients and controls that begin at G1, which supports the existence of prenatal or early postnatal morphological differences in our sample. Features like reduced philtrum, smaller ears and the inverted v-shaped mouth could be useful to diagnosis in all ontogenetic stages (Fig. 4, Table 3). In addition, small mouth and eyes, nasal hypoplasia, and low-set position of the ears could be included as discriminative traits at G3 (Fig. 4d).

In the same manner, we found that facial size differences have an age-specific pattern that is more pronounced at G3, where patients are 91.4% smaller than controls at such stage. This result is similar to previous reports of growth retardation associated to 22q11.2DS [34, 35, 64] related to height, weight [35] and head circumference [34] of Caucasian patients.

Growth disruption in 22q11.2DS individuals have been associated with environmental perturbations such as nutritional disorders [64] or hormonal deficiencies [13, 65]. Studies using mice as model of craniofacial growth development have confirmed that there is a differential response of the craniofacial components to such perturbations, with the viscerocranium being the most affected structure compared to the neuro- or basicranium. Gonzalez et al., [66] showed that in the case of postnatal malnutrition exposure, the effect of growth reduction is related to the time and magnitude of the perturbation and to the rates and relative timing of growth of the structure being affected. They found that the face component experienced larger size reduction when the exposure was late in ontogeny or was maintained throughout development and smaller reduction when the malnutrition occurred only in early ontogeny. In contrast, Gonzalez et al., found in mice that is deficient in growth hormone a catch-up growth pattern in which facial morphology was affected independently of the timing of treatment [67].

Along with size deficiency due to environmental disturbances, the underlying genetic basis of this syndrome could also be affecting the facial size and shape variation in patients, especially, genes critical for CHD. For example, one gene identified as crucial for cardiovascular morphogenesis in 22q11.2DS is TBX1, which encodes a T-box transcription factor that may disrupt the Shh or Fgf signaling pathways [68]. In avian models, this disruption has been shown to contribute to the variation of adult mid-facial growth, shape and size (e.g. brain-face covariation). Furthermore, Tbx1 haploinsufficiency in mice produces embryos that exhibit facial dysmorphism typically associated with 22q11.2DS and that appeared smaller than the littermates [69]. In humans, a recent study of Israeli patients has indirectly supported the genetic effect on the ontogenetic disruption by finding no correlation between growth delays and palate anomalies or recurrent infections, and significant association with CHD [36].

It remains to be determined if there is a relationship between catch-up growth in the facial morphology of 22q11.2DS individuals and nutritional disturbance, hormone deficiency or genetic variability. In the first case, since the amount of exposure to poor diet could have different outcomes on the facial phenotype, and feeding difficulties are frequently observed in early ontogeny (i.e. infants and children) given velopharyngeal insufficiency and other gastrointestinal complications [4, 70]. In the second case, because, although it has been shown that there is a low prevalence of growth hormone deficiency in 22q11.2DS individuals, those patients with this endocrine disorder treated with growth hormone show growth improvement [36], even when an adequate dietary intake showed no results [71]. In the last case, given that CHD is commonly observed in patients and a recent study shows promising results that suggest that the embryonic treatment using vitamin B12 ameliorates Tbx1 gene haploinsufficiency in mice [72]. This latter result could be relevant for the population analyzed in the present study, since CHD was highly prevalent (91.9%) in patients, as it has been reported in a previous study of Mexican patients [73].

The overall growth disruption observed in the present study could be explained altogether by environmental and genetic factors associated with the syndrome. Equally important, our results showed that whatever the cause, such size variation influences differently the shape of patients and controls, and that these allometric differences may help to explain the phenotypic facial variability of the syndrome. In this regard, we found that patients and controls share similar developmental and allometric trends, but they differ in allometric scaling (Fig. 6b). Our study showed that, in controls, the facial shape changes related to the developmental and allometric processes share similarities (Fig. 7a and b). For instance, in both cases (i.e. young/old and small/large comparisons) we observed a less protrusive position of the ears and an enlargement of the middle and lower third of the face outline that includes changes in the nose and lips. On the contrary, although patients exhibited a similar change in the facial outline in both processes, there were differences in the midline structures, that is, eyes, nose, and lips (Fig. 7c and d).

Taken together, these results suggest that facial dysmorphology within our population is not further accentuated by alterations of the amount of shape change occurring over time (development) or by modifications of the shape and size relationship (allometry). Nevertheless, it seems to be affected by truncation of the ontogenetic trajectory due to the underlying growth delay experienced by patients (Fig. 5). These same mechanisms have been suggested as possible explanations of the contrasting adult morphologies between populations; for instance, Inuit and Khoisan [31], or Caucasians and African-Americans [30]. At the same time, it has been shown [19, 20] that 22q11.2DS individuals of the latter populations differ in the degree of severity of the facial phenotype. Therefore, it remains to be explored if some amount of inter-population variability exhibited by these 22q11DS patients is related to the already present developmental differences.

In developing countries, where only limited resources are available to conduct molecular biology-based tests (e.g. Multiplex Ligation-dependent Probe Amplification, Fluorescence in situ hybridization or CGH microarray) in disorders like 22q11.2DS, it is of utmost importance to establish the robustness of patients’ clinical features, like facial appearance, as valid elements to reach a diagnosis. This is especially true when canonical diagnostic criteria have been developed in populations with different ancestry.