Postnatal Development of the Craniofacial Skeleton in Male C57BL/6J Mice

Abstract

C57BL/6J is one of the most commonly used inbred mouse strains in biomedical research, including studies of craniofacial development and teratogenic studies of craniofacial malformation. The current study quantitatively assessed the development of the skull in male C57BL/6J mice by using high-resolution 3D imaging of 55 landmarks from 48 male mice over 10 developmental time points from postnatal day 0 to 90. The growth of the skull plateaued at approximately postnatal day 60, and the shape of the skull did not change markedly thereafter. The amount of asymmetry in the craniofacial skeleton seemed to peak at birth, but considerable variation persisted in all age groups. For C57BL/6J male mice, postnatal day 60 is the earliest time point at which the skull achieves its adult shape and proportions. In addition, C57BL/6J male mice appear to have an inherent susceptibility to craniofacial malformation.

C57BL/6J is one of the most commonly used inbred mouse strains for biomedical research and is the first mouse strain that has its genome fully sequenced. In addition to its broad use in developmental biology and disease modeling, C57BL/6J mice are used in modeling craniofacial development and understanding the effects of teratogens,^3,6,16,17 particularly of ethanol, given the strain's high tolerance to this chemical.^{7-9,11,15,20-22} Most of these studies used a morphometric technique to assess the differences in the shapes of the craniofacial structures. The selection of the age at which to evaluate the outcome is typically guided by factors of experimental design and cost. However, the age at which evaluation is conducted might influence the results or their interpretation. The current study used a cross-sectional study design to determine the postnatal age at which a C57BL/6J male mouse becomes ‘fully adult’ in terms of the craniofacial shape and proportions.

Materials and Methods

Animal breeding.

Fifteen C57BL/6J female mice were mated overnight with C57BL/6J males in breeding duos and trios. All mice were purchased from Jackson Laboratory (Bar Harbor, MI). Once the presence of a vaginal plug was verified, dams were housed individually in standard mouse cages. Water and standardized mouse chow were available free choice. All mice were housed in OptiMice racks (Animal Care Systems, Centennial, CO) under a 12:12-h light:dark cycle (lights on, 1900 to 0700) at Seattle Children's Research Institute's vivarium. Sentinels were screened periodically (Quarterly Clinical Profile and annual Basic Profile assays, IDEXX, Columbia, MO) to assess the SPF status of the vivarium. All sentinels tested negative for all agents in these assays.

All litters in the current study were weaned at postnatal day 21. Only one (random) litter was used from each dam. At least one and no more than 2 male pups randomly selected from each litter were euthanized by using CO₂ asphyxiation on postnatal day 0, 3, 7, 14, 21, 30, 45, 60, 75, or 90. Each postnatal day was represented by at least 3 mice from 3 different litters. A total of 52 samples were collected, but 3 were excluded from analysis due to excessive skull malformation and asymmetry. An additional sample with a scanning artifact due to poor scanning calibration was excluded as well, leaving 48 samples. Final sample size was: postnatal day 0, n = 5; day 3, n = 5; day 7, n = 3; day 14, n = 3; day 21, n = 5; day 30, n = 6; day 45, n = 4; day 60, n = 6; day 75, n = 7; and day 90, n = 4. All procedures used in the study were approved by the Seattle Children's Research Institute Institutional Animal Care and Use Committee.

Imaging and morphometric analyses.

Fresh heads from the cadavers underwent 3D high-resolution X-ray microCT (imaging parameters: 0.5-mm Al filter; 55 kV, 180 μA, 420-ms exposure; Skyscan 1076, Bruker, Kontich, Belgium). Three images were acquired at each 0.7 degree of rotation and averaged. Images were reconstructed at 17.2-μm resolution by using the manufacturer's standard postprocessing procedures. Resultant image ‘stacks’ were loaded in the 3D Slicer (http://www.slicer.org) and rendered in 3D. Skulls were aligned according to principal anatomic directions, and 55 landmarks from each skull were collected twice by a single observer (Figure 1); the landmarks included were typical of studies focusing on the murine craniofacial shape.^8,18,22 Digitizing each landmark twice in two separate attempts captured potential digitization errors. In this study, the threshold for digitization error was set to 10 voxels (or 0.172 mm) between 2 attempts. The positions of the landmarks that exceeded this cut-off were reassessed by the observer and redigitized. After the correction step, the 2 sets of landmarks were averaged and used for the subsequent traditional and geometric morphometric analyses.

Figure 1.

Anatomic locations of the landmarks (LM) used in the study. Reported measurements are based on the 3D linear distance between the following landmarks: nasal, midpoint of LM1 and LM2 to LM5; frontal, LM5 to LM14; parietal, LM14 to LM15; interparietal, LM15 to LM16; premaxilla, LM29 to midpoint of LM34 and LM35; maxilla, midpoint of LM34 and LM35 to LM43; palatine, LM43 to LM47; basisphenoid, LM47 to LM52; basioccipital, LM52 to LM53; skull height, LM14 to LM52; skull width, LM10 to LM13; anterior cranial base length, LM53 to LM55; and cranial base angle, wide angle of triangle formed by LM53, LM54, and LM55.

Selected landmarks were used to calculate the length of bones forming the dorsal (nasal, frontal, parietal, and interparietal landmarks) and ventral (premaxilla, maxilla, palatine, basisphenoid, and basioccipital) aspects of the skull, the skull height and width, as well as the anterior basicranial length and basicranial angle.

Generalized Procrustes analysis (GPA) was used to assess the geometric changes in the 3D skull shape in the postnatal ontogeny. In this context, biologic shape typically is defined as the geometry that remains after the location, size, orientation,¹⁰ and any departure from perfect bilateral symmetry is removed from the landmark data.¹³ Asymmetry can arise from developmental perturbations due to nongenetic factors and therefore is of great interest for the assessment of the outcomes of perturbations such as fetal alcohol syndrome.¹⁴

A full GPA with object symmetry^13,19 was performed on these 3D landmarks by using the integrated geometric morphometrics software MorphoJ,¹² which can be downloaded free of charge from http://www.flywings.org.uk/morphoj_page.htm. In GPA, all landmark configurations are superimposed, scaled to unit size, and then rotated until the squared distance between the corresponding landmarks is minimized through a least-square optimization. Once the Procrustes alignment is finished, the new set of coordinates is used for downstream analysis. MorphoJ also was used to calculate the fluctuating asymmetry (FA) score, defined as the sum of distance between the asymmetric mean shape and the asymmetric component of the sample shape.¹² Furthermore, both principal component analysis (PCA) and canonical variates analysis (CVA) were used to reduce the dimensionality of the data and find the major axes of variation both for subjects (PCA) and groups (CVA). In GPA, size is typically expressed in terms of ‘centroid size,’ which is the square root of the summed squared distances of each landmark to its centroid.⁵ The full centroid size was calculated as a proxy for the overall skull size, by using all 55 landmarks. All GPA superimposition, PCA, and CVA were conducted using MorphoJ (version 1.0.6d);¹² additional statistical testing and 3D visualization of PCA results were done in R (version 3.1.2)²³ by using the shapes and Morpho packages.^4,24

Results

Traditional morphometrics.

Changes in cranial bone lengths and skull dimensions, size, and asymmetry in postnatal ontogeny are given in Figure 2. In addition, 2-sample t tests were used to compare the mean bone lengths between consecutive postnatal developmental days (Table 1). In most cases, the growth and elongation of the skull bones stabilized after P45. The lack of significance in the differences in parietal length and cranial base angle between age groups is noteworthy. Both of these measurements are associated with marked variation, as shown by large 95% confidence intervals around the mean. The parietal length was defined as the distance between bregma (midpoint on the intersection of sagittal suture with coronal suture) and the midpoint of the intersection of parietal bones with anterior border of interparietal bone. Because landmarks can be identified only on bone, gaps in the still-closing suture areas, in early-age specimens, resulted in greater measurement variation. Therefore, observation of increased variation in parietal length for early-age samples should be interpreted cautiously, because it might be artificially inflated.

Figure 2.

Lengths of selected cranial bones and skull dimensions, size, and asymmetry during postnatal development. Black circles indicate the mean for each age groups, ticks are the actual observations, and the vertical line is the 95% confidence interval around the mean.

Table 1.

Results of 2-sample t tests for difference in mean bone length, skull dimensions, and angles in consecutive age groups

Groups compared	Nasal	Frontal	Interparietal	Premaxilla	Maxilla	Palatine	Basisphnoid	Basioccipital	Skull height	Skull width	Dorsal skull length	Ventral skull length	Anterior cranial base length	Skull size
P0 & P3	0.002	0.000	0.004	0.001	0.000	0.008	0.020	0.028	0.000	0.000	0.000	0.000	0.008	0.000
P3 & P7	0.000	0.021	0.000	0.050	0.028	0.009	0.095	0.087	0.001	0.004	0.001	0.045	0.019	0.012
P7 & P14	0.004	0.006	0.004	0.002	0.002	0.006	0.003	0.030	0.002	0.006	0.000	0.003	0.003	0.001
P14 & P21	0.002	0.003		0.003	0.006		0.035	0.050		0.048	0.008	0.016		0.022
P21 & P30	0.004	0.012		0.018	0.004		0.004	0.008		0.020		0.001		0.016
P30 & P45	0.020							0.021			0.017
P45 & P60													0.026

P, postnatal day

Only P values less than 0.05 are reported (no significant differences for parietal and cranial base angle measures or between groups P60 and P75 or between P75 and P90).

Similarly the anterior landmark (crista galli) defining the cranial base angle varied considerably in its position. This greater variation coupled with the small sample sizes resulted in large confidence intervals associated with these measurements. Despite pronounced variability among neonates, the FA scores were relatively stable throughout development.

The relative contribution of each bone to the total skull length is given in Figure 3. The bones contributing to the dorsal aspect of the skull grew at different rates until postnatal day 60, as shown by the changing proportions. In contrast, growth was more uniform for the bones contributing the ventral aspect, for which the ratios remained unchanged from postnatal day 30 onward.

Figure 3.

Contributions of selected cranial bones to the composition of the (A) dorsal and (B) ventral skull lengths.

Geometric morphometric and postnatal skull shape changes in male C57BL/6J mice.

Figure 4 shows the PCA plot of the symmetric component of the skull shape. The first principal component (PC1) accounted for approximately 88% of all the variation in the dataset. Because GPA removes the isometric (or uniform) component of the size variation in the dataset but retains the allometric component, a linear model tested the extent to which PC1 scores influenced by the allometric growth. In this case, the PC1 scores of samples were highly correlated (R² = 0.98, α < 0.0001) with the overall skull size. Therefore, PC1 describes the postnatal allometric growth of the skull. In terms of shape changes, PC1 revealed the elongation of the skull, reduction in the basicranial flexion, rounding of the neurocranium, and increases in the length of the rostrum. Other PCs (not shown) accounted for the remaining 12% of variation and were related to age-independent changes in skull shape, mostly in cranial width, height, and doming.

Figure 4.

Univariate plot of PC1 of the symmetric component of the skull shape in the study sample. Black dots indicate individual values, and ticks are the mean PC1 value for each age group. The shape gradient associated with PC1 is rendered by using the mean PC1 values of each age group and Thin Plate Spline deformation.

To visualize the ontogenetic shape change expressed by PC1, thin-plate spline deformation involving all 55 landmarks was used to warp the surface mesh of the sample closest to the grand mean shape of the population to the sample mean.² The mean PC1 scores for each age group were used to scale the PC1 eigenvector; the resulting vector was added to (or subtracted from) the mean shape. No additional scaling of the PC values was necessary, because the changes already were sufficiently large for visualization, especially at early ages. The resultant set of coordinates was used to create a representative 3D image for the age group, again by using thin-plate spline deformation. All deformations and 3D renderings were conducted in R by using the Morpho and rgl packages.

The CVA procedure of MorphoJ performed a Procrustes distance-based permutation test of 10,000 replicates to assess the statistical significance of differences in skull shape between age groups (Table 2). Although younger age groups were highly significantly different from each other as well as from older groups, the 60-, 75-, and 90-d groups did not differ from one another in terms of skull shape. In addition, the 45-day group differed from those for 60 and 75 d but not from the 90-d mice. This test, however, only considers the univariate Procrustes distance as a measure of difference between groups and is not multivariate in nature. Therefore, a modification of the Hotelling T² test¹ was used to test differences in mean skull shape through permutation-based resampling as implemented in the shapes package of R statistical software.⁴ Because this test is extremely computationally intensive, only differences between consecutive developmental days were tested. Results were similar to those for the Procrustes distance-based method, except that the differences between the day 3 and day 7 groups and between day 7 compared with day 14 were not significant. This outcome is not surprising given that, in the current study, the Hotelling T² test is impeded by small sample size and numerous variables. Specifically, each landmark in our 3D landmark dataset had 3 variables (x, y, and z coordinates), meaning that the total number of variables for a single observation was 165. In addition, the day 7 and 14 age groups had the smallest sample size (n = 3) in our study population. For all other age-group pairs, the Hotelling T² result was identical to that for the Procrustes distance-based test.

Table 2.

P values from permutation resampling (10000 replicates) for Procrustes distances among groups (upper triangle) and from the mean shape tests of 2 consecutive ages by using a modified Hotelling T-squared statistic¹based on permutation resampling (5000 replicates; lower triangle). In both cases, groups that are notsignificantly different from each other at P < 0.05 are bolded. The non significance of the P3–P7 and P7–P14 mean shape differences are most likely due to the low power of the T-squared statistic, given the small sample sizes (8 and 6, respectively) and the variation among the samples at these ages.

Postnatal day	0	3	7	14	21	30	45	60	75	90
0		0.0088	0.0165	0.0171	0.0080	0.0016	0.0073	0.0005	0.0009	0.0055
3	0.0026		0.0069	0.0149	0.0071	0.0012	0.0089	0.0013	0.0013	0.0068
7		0.6625		0.0335	0.0175	0.0053	0.0201	0.0057	0.0056	0.0058
14			0.0646		0.0175	0.0009	0.0048	0.0021	0.0047	0.0006
21				0.0004		<0.0001	0.0006	0.0001	<0.0001	0.0001
30					0.0002		0.0049	0.0004	0.0004	0.0018
45						0.0012		0.0281	0.0189	0.1481
60							0.0128		0.5813	0.3563
75								0.3179		0.4281
90									0.8442

The variation in the asymmetric component of the skull shape was much less structured than that for the symmetric component. The first PC accounts only about 27% of the total variation, and 13 PCs were required to explain the same amount of variation as that explained by the first PC of the symmetric component. Although some of the neonates appeared to cluster around a negative PC1 score, there was no clear segregation by postnatal age. The plot of the mean asymmetry in the skull shape (figure not shown) indicated that most of the asymmetry arose due to landmarks located on the neurocranium (landmarks 14, 15, 43, and 47) and those in the nasal area (landmarks 1 and 2). These same regions showed pronounced asymmetry and malformation in some of our samples (Figure 5).

Figure 5.

Three male samples excluded from the morphometric analyses due to severe craniofacial malformations and asymmetry. (A) Rightward deviation of the presphenoid bone; foramen rotundum and foramen ovale differ in size on left and right sides. (B and C) Severe midfacial asymmetry.

Discussion

The current morphometric analysis of skull shape and size in postnatal C57Bl/6J male mice revealed that different regions of the skull have different paces of development. For example, the relative contributions of the bones to the ventral aspect of the skull length were virtually unchanged from postnatal day 30 onward (Figure 3) yet they clearly continued to grow (Figure 2). In contrast, the stabilization of the dorsal skull ratios occurred at day 60 and thereafter (Figure 3). From postnatal day 60 to 90, the skull shape, skull proportions, and individual bone lengths were statistically indistinguishable from each other (Table 1). According to the current analyses, postnatal day 60 was the earliest time point at which the adult cranial shape of C57BL/6J male mice was obtained in either traditional or geometric morphometric studies.

Postnatal day 45 can potentially be used for analyses focusing solely on the skull bone dimensions, with the exception of the anterior cranial base angle. However day 45 is not well supported as an endpoint in geometric morphometric analyses, given the considerable variation in the skull shape at that stage, as shown by the wide dispersion of the data points on the PCA plot (Figure 4). In particular, 2 day-45 specimens clustered with day-30 samples and another with the day-60 samples. This problem can be overcome by maximizing the sample sizes, a practice that is always advisable for geometric morphometrics in view of the large number of variables involved. In addition, increasing the sample size would help with measurements that were highly variable, namely parietal length, FA score, and cranial base angle.

Although the skull shape did not differ from postnatal day 60 onward, this finding does not imply that full adult proportions have been achieved by that point. Skulls continued to remodel throughout the postnatal ontogeny, albeit at a much slower rate. Significant differences in the skull shape might occur beyond day 90, especially with larger sample sizes, but these experiments were beyond the scope of the current study.

The most surprising outcome of this study was the high frequency of spontaneous craniofacial malformations in C57BL/6J mice. In our original study population of 52 male mice, 3 unrelated animals had severe malformations in the midface region (Figure 5), thus yielding an incidence rate of approximately 5.7%. The mean asymmetry of the study population suggested increased asymmetry regarding landmarks associated with this region. In-utero exposure to citalopram reportedly caused craniofacial malformations in litters of C57BL/6J mice.³Among the gross craniofacial malformations reported in those pups,³ the severely deviated snout is strikingly similar to the spontaneous malformations in 2 of our samples (Figure 5). Furthermore the authors of the cited study³ reported that no such malformations occurred in their control population (n = 26), and the 2 deviated snouts in their study population (n = 24) corresponded to an incident rate of 8.3%, which is higher than that for the current study's wild-type population. In addition, chronic in-utero exposure to a low concentration of ethanol led to a deviated snout in a male C57BL/6J mouse in a study population of 10 control and 7 exposed mice.⁸ In both of the previously reported cases,^3,8 the observed snout malformations likely were indeed due to the teratogenic exposure. However, the exposure to the teratogen might simply be amplifying an already existing inherent susceptibility to such malformations rather than causing them de novo.

Without argument, the current study population is small, and further study is warranted to obtain better rate estimates for inherent craniofacial malformations in C57BL/6J. In addition, lack of female mice in the present study precludes investigation of sex-specific patterns of development and frequencies of asymmetry. Statistically significant sexual dimorphism in murine skulls has been documented.¹⁸ A large back-crossed population of A/J and C57BL/6J previously revealed that sex accounts for about 1% of the variation in the skull shape, and male and female A/J × C57BL/6J mice differed most in cranial vault curvature and in the length of the basicranium.¹⁸ However, a further investigation specific to the C57BL/6J strain would be informative in this regard.

This first attempt to study the postnatal development of the C57BL/6J skull used both traditional and geometric morphometrics. Further studies with larger sample sizes and multiple inbred strains would increase our understanding of the normal craniofacial development and variation in inbred laboratory mice and guide the selection of appropriate model strains. Given the importance of C57BL/6J as a model for teratogenic studies of craniofacial development, detailed investigation of the types and frequency of inherent craniofacial malformations in this strain is warranted.