Signaling among fibroblast growth factor receptors (FGFRs) and their ligands (FGFs) is one of the few cell–cell signaling pathways conserved among all metazoans (Itoh and Ornitz, 2004). A growing body of evidence suggests that FGF/FGFR signaling is fundamental in the development of modern vertebrates and has played a critical role in the advent and evolution of many vertebrate characteristics including those that contribute to the vertebrate head (Bertrand et al., 2011). Mutations in FGFR2 are commonly associated with human craniosynostosis syndromes (Ornitz and Marie, 2002), common genetic disorders (∼1:60,000) typified by premature closure of cranial vault sutures (especially the coronal suture), but in fact comprising a wide spectrum of anomalies (Wilkie, 1997; Cohen and MacLean, 2000; Ornitz and Marie, 2002). The role of FGF/FGFR signaling in bone development has been well established (Ornitz and Marie, 2002) and dysmorphology of the skull is well studied in craniosynostosis syndromes, due at least in part to the invariant occurrence of premature closure of cranial vault sutures and ensuing reconstructive surgery in human patients. However, FGF/FGFR signaling is critical to basic cellular processes of development (i.e., patterning, proliferation, differentiation, and migration) in tissues other than those with an osteogenic fate (Thisse and Thisse, 2005; Hebert, 2011). Using high-resolution multimodal imaging, we explore additional potential targets of FGF/FGFR signaling in head development by assessing the phenotypic effects of altered signaling in various skeletal and non-skeletal tissues of the head, including the skull, the brain, the nasopharynx, the eye, and the inner ear of mutant and unaffected littermates of a Crouzon syndrome mouse model.
The Fgfr2cC342Y/+ Crouzon syndrome mouse model carries a cysteine to tyrosine replacement at amino acid position 342 (Cys342Tyr; C342Y) in Fgfr2 (Eswarakumar et al., 2004) equivalent to the FGFR2 mutation most commonly associated with Crouzon syndrome [OMIM no. 123500], but also causative for Pfeiffer syndrome [OMIM no. 101600]. The presence of digital and other organ abnormalities normally distinguish Crouzon from Pfeiffer syndrome patients, as differences in craniofacial phenotypes are subtle and there is marked overlap in the craniofacial features of the two syndromes (Rutland et al., 1995). In humans, Crouzon syndrome craniofacial phenotypes are variable but typically include premature closure of the coronal suture (either unilateral or bilateral) associated with a turribrachycephalic or brachycephalic cranial vault, generalized neurocranial dysmorphology, midfacial deficiency, shallow orbits, and ocular proptosis (Cohen and MacLean, 2000). Children with Crouzon syndrome can show additional symptoms affecting other head structures, such as cleft palate (Peterson and Pruzansky, 1974; Riley et al., 2007), upper airway obstruction (Moore, 1993; Sirotnak et al., 1995; Perkins et al., 1997; Scheid et al., 2002; Mitsukawa et al., 2004; Mitsukawa and Satoh, 2010; Randhawa et al., 2011), dysmorphology of the nasopharynx and contiguous structures that impair velopharyngeal function and nasal respiratory physiology (Peterson-Falzone et al., 1981; Johnson and Wilkie, 2011), and mild or moderate hearing loss (conductive, sensorineural, or mixed) often caused by recurrent otitis media effusion, ossicular chain fixation, and external auditory canal atresia (Cremers, 1981; Vallino-Napoli, 1996; Orvidas et al., 1999; Cohen and MacLean, 2000; de Jong et al., 2011; Huh et al., 2012). Brain anomalies do not occur frequently in Crouzon syndrome but ventriculomegaly is fairly common (Proudman et al., 1995). Correspondence between the Crouzon mouse model and human patients with Crouzon syndrome has been demonstrated at the morphological, histological, and molecular levels (Eswarakumar et al., 2004; Perlyn et al., 2006; Snyder-Warwick et al., 2010). Still, a significant knowledge gap exists between identification of causative genetic mutations and the development of strategies to prevent or treat these associated abnormalities. Consequently, most therapies are symptomatic and reconstructive-based and do not address the etiological origins of craniosynostosis phenotypes.
In Crouzon syndrome, the FGFR2 C342Y mutation is a gain-of-function mutation in the mesenchymal Fgfr2c variant. The distinct ligand-binding specificity and tissue-specific expression properties of alternatively spliced mRNA variants of FGFRs (b variants are epithelial-specific, c variants are mesenchymal-specific) may underlie the diverse phenotypic effects of craniosynostosis syndromes. The IIIc isoform of FGFR2 is preferentially expressed in mesenchymal tissues (Orr-Urtreger et al., 1993) and is required by the osteoblast lineage for normal skeletogenesis. However, as FGFR isoform production enables regulatory interplay between epithelial and mesenchymal layers during development in response to FGFs (Eswarakumar et al., 2005; Degnin et al., 2010), and mutations such as FGFR2 C342Y in the IIIc isoform result in loss of ligand specificity and constitutive activation of FGFR2 triggering abnormal signaling without the presence of ligand (Neilson and Friesel, 1995), the effects of a mutation in the IIIc isoform could affect cell lineages other than the those destined to become bone.
To provide a broad characterization of the potential concurrent phenotypic effects of mutations in the FGFR2c isoform on skeletal and non-skeletal phenotypes of the murine head, we conduct an extensive quantitative analysis of heads of newborn Fgfr2cC342Y/+ Crouzon syndrome mice and their unaffected littermates using multimodal imaging. We test the hypothesis that a mutation in the Fgfr2 IIIc splice variant causes numerous and varied changes in quantitative effects on several skeletal and non-skeletal head tissues of varied embryonic origin. Head phenotypes of newborn Fgfr2cC342Y/+ Crouzon mice and unaffected littermates are quantified using landmark and volumetric data collected from high-resolution micro-computed tomography (μCT) and magnetic resonance microscopy (MRM) images. Under the null hypothesis, we expect that effects of a mutation of the Fgfr2 IIIc variant will primarily affect bone tissue and thus significant differences between mutant and unaffected mice will be limited to the skull. The null hypothesis is rejected if we determine statistically significant contrasts between mutant and unaffected littermates in non-skeletal tissues (i.e., brain, nasopharynx, eye, inner ear), indicating that the Fgfr2 IIIc variant affects morphogenesis of several tissues in addition to those formed from mesenchyme. The alternative hypothesis proposes multiple tissue effects of FGF/FGFR signaling in head development, and, if supported, provides novel data pertaining to the broad effects of FGFR mutations on the development of head tissues and evidence for the importance of this signaling pathway in the integration and coordination of changes that occurred with the evolution of the vertebrate head (Bertrand et al., 2011; Martínez-Abadías et al., 2011).
We compared the phenotypic variation of skeletal and non-skeletal structures of P0 heads of Fgfr2cC342Y/+ Crouzon mice and their unaffected littermates (Table 1) using an array of morphometric and statistical methods (see Experimental Procedures section for more details) that allow us to characterize overall variation and pinpoint highly localized shape and size differences among Fgfr2cC342Y/+ and unaffected littermates.
Table 1. Sample Sizes (N) of Fgfr2cC342Y/+Mutant Mice and Unaffected Littermates Used in Analyses
Analysis of skull morphology was accomplished using data from micro-computed tomography (μCT).
Analysis of soft tissue structures was accomplished using data from magnetic resonance microscopy (MRM). MRM samples represent subsets of the same individuals that underwent μCT imaging. These samples vary in size due to quality of images for segmentation of various structures.
Skull Phenotypes of Fgfr2cC342Y/+ Newborn Mice Show Global and Localized Differences Relative to Unaffected Littermates
Skull shape was captured by collecting the 3D coordinates of 39 anatomical landmarks on μCT images of P0 heads (Table 2). Generalized Procrustes analysis was used to superimpose the coordinate data and to extract shape information (Rohlf and Slice, 1990a; Dryden and Mardia, 1998). This procedure minimizes the influence of size and adopts a single orientation for all specimens by shifting the landmark configurations to a common position, scaling them to a standard size, and rotating them until a best fit of corresponding landmarks is achieved. For the configuration of landmarks representing each animal, a measure of size is computed as the centroid size, the square root of the summed distances between each landmark coordinate and the centroid of the landmark configuration. Shape variation as represented by Procrustes coordinates was analyzed using Principal Component Analysis (PCA), a data exploration technique that performs an orthogonal decomposition of the data and transforms the resulting Procrustes coordinates into a smaller number of uncorrelated variables called principal components (PCs).
Table 2. Anatomical Definitions of Skull Landmarks Collected From μCT Images and Shown in Figure 1a
Intersection of frontal process of maxilla with frontal and lacrimal bones
Intersection of malar with zygomatic process of maxilla, taken on maxilla
Intersection of malar with zygomatic process of maxilla, taken on malar
Intersection of zygoma with zygomatic process of temporal, taken on zygoma
Most anterior point of the anterior palatine foramen
Most posterior point of the anterior palatine foramen
Most infero-lateral point of the premaxillary-maxillary suture, taken on premaxilla
Midline point on the premaxilla between the incisor and the nasal cavity just anterior of the incisive foramen, right side only
Most antero-superior point of the vomer
Most posterior point of the vomer
Most anterior point of the indentation in the center of the presphenoid
Most anterior point on the anterior projection on the presphenoid
Postero-medial point of the inferior portion of the left alisphenoid
Most antero-lateral indentation at the posterior edge of the horizontal plate of the palatine bone
Most posterior tip of the medial pterygoid process
Most antero-lateral point on the corner of the basioccipital at the basioccipital synchondrosis
Most infero-lateral point on the squamous occipital
Mid-point on the posterior margin of the foramen magnum, taken on basioccipital
Mid-point on the anterior margin of the foramen magnum, taken on squamosal occipital
Superior posterior point on the ectocranial surface of occipital lateralis on the foramen magnum
Most posterior point on the posterior extension of the forming squamosal
Most lateral intersection of the frontal and parietal bones, taken on the frontal
Most lateral intersection of the frontal and parietal bones, taken on the parietal
Most postero-medial point on the parietal
Most antero-medial point on the body of the sphenoid
Most anterolateral point on the posterior palatine plate
Most anteriomedial point on the posterior palatine plate
Most posteromedial point on the posterior palatine plate
Most posterolateral point on the posterior palatine plate
The PCA based on the Procrustes coordinates of the 39 cranial landmarks that define the global skull configuration (Table 2; Fig. 1A, 1B) showed clear separation between Crouzon syndrome Fgfr2cC342Y/+ mutant mice and unaffected littermates along PC1, which explained 24.36% of total morphological variation (Fig. 1A). Shape changes associated with positive values of PC1 and corresponding to Fgfr2cC342Y/+ mutant mice are located on the parietal and temporal regions of the cranial vault, as well as the cranial base and the face. PC2 accounted for 12.49% of total shape variation but did not separate mutant and unaffected littermates, indicating skull shape variation along this axis is shared by the two groups (Fig. 1A). The multivariate regression of global skull shape on size showed that allometry (size-related differences in shape) was not a significant factor affecting newborn skull shape variation (i.e., Fgfr2cC342Y/+ mutant and unaffected littermates were similar in size), with size predicting only 2.18% of shape variation (P = 0.21).
To explore regional shape variation, we summarized four distinct configurations of 3D landmark coordinates representing the facial skeleton, the cranial base, the cranial vault and the palate (Table 2, Fig. 1) and analyzed them separately using PCA of Procrustes-based coordinates. To avoid problems associated with mathematical invariance that make the Procrustes parameters used for inference non-identifiable (Lele and McCulloch, 2002; Richtsmeier et al., 2002), we also used Euclidean Distance Matrix analysis (EDMA) (Lele and Richtsmeier, 2001), a coordinate system invariant approach to statistically test for localized differences in shape between Fgfr2cC342Y/+ mutant mice and unaffected littermates (see Experimental Procedures section). Separation of mutant and unaffected mice is apparent in the results of the PCA of the facial skeleton (Fig. 1C) and cranial base (Fig. 1D) but is less clear for cranial vault (Fig. 1E) and especially palatal shapes (Fig. 1F). The hypothesis of similarity in facial skeletal shape tested with EDMA was rejected (P = 0.03) and revealed statistically significant differences in facial shape of the two groups of mice. The facial skeleton of Fgfr2cC342Y/+ mutant mice is statistically reduced relative to unaffected littermates for measures on the nasal bones and premaxillae (derivatives of the medial nasal prominence) and the rostrocaudal length of the maxillary palatal shelves (formed in the maxillary prominences) (Figs. 1C, 2A,B; see Supp. Video S1, which is available online). Statistical results indicate a reduction in midline elements of the facial skeleton and expansion in height and width of the more caudal aspect of the face in mice carrying the Fgfr2cC342Y/+ mutation. Cranial base shapes of mutant and unaffected mice are statistically different (P = 0.001) and a distinct demarcation between the effects of the C342Y mutation on rostral and caudal cranial base elements is apparent (Figs. 1D, 2C,D; Supp. Video S2) with the line of demarcation of these opposite effects localized to the basisphenoid bone. Dimensions measured from the midsphenoidal body to rostral aspects of the cranial base are relatively reduced in mutant mice while dimensions of the caudal aspects of the cranial base are relatively enlarged in mutant mice (Fig. 2D). EDMA demonstrated that the anterior cranial vault of Fgfr2cC342Y/+ mutant mice is reduced rostrocaudally local to the parietal and squamous temporal bones, but posteriorly it is increased along rostrocaudal and mediolateral axes (interparietal, squamous occipital) relative to unaffected littermates, with an overall statistically different shape (P = 0.001) (Figs. 1E,2C,D). Finally, EDMA also confirmed statistically different palatal shapes in the two groups of mice (P = 0.001) (Fig. 1F). Analysis of palatal landmarks revealed increased rostrocaudal dimensions of the horizontal plates of the palatine bones but reduction along the mediolateral axis with increased distance between these plates at the inter-palatine suture in Fgfr2cC342Y/+ P0 mice relative to unaffected littermates, though no mice in our sample had overt cleft palate.
Suture patency was visualized and qualitatively scored in Fgfr2cC342Y/+ Crouzon mice and unaffected littermates from the 3D μCT isosurfaces (Fig. 3). Over 95% of Fgfr2cC342Y/+ mutant mice showed partial to complete coronal suture fusion bilaterally at P0 (Table 3, Fig. 3C). Coronal sutures were completely patent in most unaffected littermates. Fgfr2cC342Y/+ mutant mice show a higher tendency towards patency of several facial sutures relative to unaffected littermates (inter-premaxillary, inter-palatine, maxillary-palatine), with the exception of fusion of zygomatic-maxillary suture. Sutures connecting the anterior neurocranium to the facial skeleton (fronto-maxillary, fronto-premaxillary, and fronto-nasal sutures) are patent in both mutant and unaffected littermates (Table 3, Fig. 3).
Table 3. Counts (and Corresponding Percentages) for Suture Patency in Fgfr2cC342Y/+ Crouzon Syndrome Mice and Unaffected Littermates at P0
For each group, qualitative scores on suture patency (open, partial, fused) as visualized on μCT images are provided. Open: No closure or overlap visualized at either side of the suture (both sides scored as open). Partial: Partial fusion or “bridging” apparent. Fused: Almost no open areas visualized along sutural margins. Suture patency data are presented as a histogram in Fig. 3C. For a more complete definition of patency states, see Experimental Procedures section.
The inter-premaxillary suture was divided into thirds: the most anterior third, medial third, and most posterior third. Each section was scored on the degree of fusion: a score of 1 indicated the suture was completely patent; 2 indicated partial fusion; 3 indicated complete fusion. The sutures were scored from anterior to posterior. A score of 123 indicates the anterior third is open, the medial third is partially fused, and the posterior third is fused; 233 refers to inter-premaxillary sutures that remained patent anteriorly but were otherwise not patent; 333 refers to inter-premaxillary sutures that showed no visible patency by inspection. See Supp. Experimental Procedures section.
This Fgfr2c+/+ mouse displayed complete fusion of the left coronal suture and partial fusion of the right coronal suture.
Totals vary for suture counts in cases where a mouse was damaged during processing and specific suture(s) could not be scored for patency reliably.
Brain Phenotypes of Fgfr2cC342Y/+ Mutant Mice at Birth Show Localized Shape Differences Relative to Unaffected Littermates
Global brain shape was defined by the 3D coordinates of 15 anatomical landmarks (Table 4, Fig. 4D,E,F). The PCA analysis based on the Procrustes coordinates of brain landmarks did not show a distinct separation between Crouzon syndrome Fgfr2cC342Y/+ mutant mice and unaffected littermates along PC1 (27.47% of total morphological variation) or PC2 (17.30% of total shape variation), indicating that the majority of brain shape variation is shared by the two groups (Fig. 4A). Brain volume, estimated as the geometric mean of all linear distances measured among a set of 15 neural landmarks (Darroch and Mosimann, 1985) (Table 4) was not significantly different in Fgfr2cC342Y/+ mutant mice and unaffected littermates (P = 0.862; Mann-Whitney U test; IBM SPSS 20.0; IBM, New York). Multivariate regression of global brain shape on size showed that allometry was not a significant factor affecting newborn brain shape variation, with size predicting only 4.65% of shape variation (P = 0.36).
Most superolateral point of intersection of olfactory bulb with anterior frontal lobe surface
Most caudolateral point on occipital lobe surface
Most caudal point on cerebellar surface
Most lateral point on cerebellar surface
Centroid of head of caudate nucleus
Origin of middle cerebral artery from Circle of Willis on ventral cerebral surface
Genu of corpus callosum
Splenium of corpus callosum
Intersection of pons with most caudal aspect of the ventral cerebral surface
To explore regional shape variation, we divided the 3D landmark coordinates into two distinct configurations, landmarks located on the brain surface (K=10) and those located deep to the surface (K=5) (Table 4; Fig. 4D–F), and analyzed them separately using PCA of Procrustes-based coordinates and EDMA as described for skull analyses. PCA demonstrates no distinct separation between mutant and unaffected brain shapes for the surface shapes (Fig. 4B), but the subsurface landmarks reveal a tendency towards separation suggesting that the Fgfr2c mutation affects surface and subsurface regions of the brain differently (Fig. 4C,D). Size-related differences in shape (allometry) did not have a significant effect on shape difference of the whole brain or surface landmarks, but there was a significant allometric effect of size (P ≤ 0.05) in the analysis of the subsurface landmarks, indicating a size difference in the distribution of subsurface features in the brains of mutant and unaffected mice.
EDMA non-parametric hypothesis testing for overall shape difference between mutant and unaffected littermates for brain surface and subsurface landmark subsets did not show statistical differences between groups (P = 0.297, P = 0.565, respectively), but non-parametric confidence interval testing revealed significant localized differences in the neonatal brains of Fgfr2cC342Y/+ mice and unaffected littermates (Fig. 5). Specifically, dimensions describing the height of the cerebrum of Fgfr2cC342Y/+ mutant mice are significantly increased (Fig. 5B,C), while the mediolateral breadth of the rostral cerebrum is significantly decreased in Fgfr2cC342Y/+ mice relative to unaffected littermates (Fig. 5A). The caudolateral surface of the brain is located farther from the midline in mutant mice as compared to unaffected mice (Fig. 5A).
Nasopharyngeal and Vitreous Humor Volume, But Not Inner Ear Volume Differs in Fgfr2cC342Y/+ Mutant Mice Relative to Unaffected Littermates
To analyze and compare the effects of the Fgfr2 IIIc mutation on other non-skeletal head tissues, we obtained volumetric quantitative measures of the nasopharynx, eye ball, and inner ear from MRM images. Volume of the nasopharynx was estimated for each mouse by combining standard segmentation techniques with brain landmarks (Aldridge et al., 2010) that were used to define a plane determining the caudal end of the nasopharynx (see Experimental Procedures section). Fgfr2cC342Y/+ mutant mice had significantly restricted nasopharyngeal volumes (2.81 ± 0.17 mm3) compared to unaffected littermates (3.28 ± 0.13 mm3; P = 0.012, Mann-Whitney U test) (Fig. 6). Neither the length of the rostral aspect of the nasopharynx (estimated as the distance from most rostro-superior point of the nasopharynx to the most caudo-superior point), nor the length of the caudal aspect of the nasopharynx (estimated as the distance between the most caudo-superior point to the most infero-caudal point on the nasopharynx) is significantly different in mutant as compared to unaffected littermates. 3D visualizations of the virtual endocasts of the nasopharynges (Fig. 6) suggest differences in the complex shape of these negative spaces in mutant and unaffected littermates and reduction in the cross-section of the nasal passages.
Vitreous humor, an avascular mass of transparent, gelled, intercellular substance occupying a chamber posterior to the lens and making up a large proportion of the eye volume, was segmented from MRM images and used as a volumetric measure for globe of the eye (Fig. 6). There was no significant difference in vitreous humor volume estimated for the left and right globes of the eyes of Fgfr2cC342Y/+ mutant mice or unaffected littermates, indicating no asymmetry for eye volume in either sample. Therefore, left and right eye volumes were averaged for each animal and used in statistical analysis. The mean vitreous humor volume of the eyes of Fgfr2cC342Y/+ mutant mice (0.60 ± 0.04 mm3) is significantly larger as compared to that of unaffected littermates (0.36 ± 0.03 mm3) (P < 0.001, Mann-Whitney U test).
Virtual endocasts of the fluid filled spaces of the cochlea and vestibular canals were segmented using MRM images of each mouse using standard segmentation techniques (Fig. 6; see Experimental Procedures section). Total inner ear volumes were estimated (cochlea and vestibular canals) from these segmentations. No difference in inner ear volume was found between newborn mutant (0.97 ± 0.04 mm3) and unaffected (1.02 ± 0.03 mm3) littermates (P = 0.248, Mann-Whitney U test).
In agreement with what has been previously described for adult Fgfr2cC342Y/+ mice (Eswarakumar et al., 2004; Snyder-Warwick et al., 2010), the dysmorphology of skulls of newborn Fgfr2cC342Y/+ mice parallels skull phenotypes of human infants with Crouzon syndrome, even though most of the bones of the murine neonatal neurocranium and cranial base are only partially formed at P0. Skulls of Fgfr2cC342Y/+ mice and unaffected littermates are similar in size at P0, but different in shape (Figs. 1, 2). These differences are accompanied by premature closure of the coronal sutures and variation in patency of facial sutures (Fig. 3, Table 3). Similar to previous studies (Eswarakumar et al., 2004; Snyder-Warwick et al., 2010), the incidence of cleft palate is extremely low in mice heterozygous for this mutation. Our analysis is the first to estimate additional effects of the Fgfr2 IIIc mutation on shape and size of several non-skeletal tissues of the head in newborn Fgfr2cC342Y/+ mice. We catalogue additional targets of the Fgfr2 C342Y mutation and reveal specific significant changes in brain shape (but not brain size) (Figs. 4, 5), significant reductions of nasopharyngeal volume, and significantly increased eye volume (Fig. 6), supporting the alternative hypothesis that a mutation in the Fgfr2 IIIc variant affects development of various tissues of the embryonic head.
Direct and Indirect Effects of the Fgfr2 IIIc Mutation
Genetic variants causative for the FGFR-related craniosynostosis syndromes are many, and variation in midfacial deficiency and neurocranial dysmorphology are characteristic of these syndromes (Wilkie, 2005). These traits are usually attributed to dysmorphogenesis of the skull, a composite and complex osseous structure, and our Fgfr2cC342Y/+ mice show many, and some of the most severe malformations seen in the skulls of humans carrying this mutation. Though it is true that mineralized skeletal tissue provides a structural framework for overall head morphology (Abzhanov et al., 2007), mineralization of intramembranous and endochondral bones of the skull occurs late in development relative to the appearance of other soft tissue structures and spaces (e.g., brain, naso- and oropharynx, eyes). Moreover, the skull is not a product of the activities of isolated osteogenic precursor cells, but takes its form from the interaction of multiple cell types that send and respond to signals from many sources (Noden and Trainor, 2005), all combining in a highly coordinated 4-dimensional space-time environment to differentiate into the varying tissues that together form the head.
Historically, the functional matrix (FM) hypothesis proposed that the presence, size and shape, growth, and position in space of all skeletal tissues are secondary, compensatory, and mechanically obligatory responses to the temporally prior demands of related soft tissue functional matrices (Moss, 1962). The proposed mechanism (e.g., biomechanical, biophysical) was put forward prior to current knowledge of the genetic basis of development and failed to provide an explanation for cause and effect in development, though the generation of signaling via mechanical forces through cell–cell contact between tissues continues to be a productive area of research (e.g., Yu et al., 2001). Specific to craniosynostosis, the FM hypothesis predicted that changes in the orientation of fiber tracts of the dura mater changed tensile forces that affected the position of cranial bone anlagen and suture development resulting in premature suture fusion (Moss, 1959). More recently, data pertaining to localized expression of FGF/FGFR pathways, ligand specificity, the interaction of gene products, and cellular events happening local to the suture have been gathered with the knowledge that FGFs are integral to the growth of normal bone, and FGFR mutations are perhaps the direct and critical players in the process of premature suture fusion. Work presented here and previously (Martínez-Abadías et al., 2010) demonstrates that coronal suture closure is neither the primary nor sole locus of skull dysmorphology in mouse models for craniosynostosis syndromes.
Determining to what extent size and shape changes quantified in various tissues result from a direct influence of the mutation on each tissue, and/or are the result of indirect effects of an alternate malforming tissue (those we analyzed or alternatives) on the morphogenesis of another tissue requires further analysis. We provide a preliminary analysis of the interdependence of development among traits of the head by estimating the correlation of size of skeletal and non-skeletal structures of the head (Table 5). The correlation analysis reveals similar patterns of relationships among these tissues in Fgfr2cC342Y/+ mice and their unaffected littermates. In mutant and unaffected mice, the correlation among skull regions is strongest and the correlation among non-skeletal traits is low to moderate. In both groups, the correlation between sizes of developing skeletal and non-skeletal cranial traits is very low or absent, a finding that is surprising given the apparently precise coordination among multiple developing tissues during head morphogenesis. Our preliminary analyses suggest that this detailed coordination occurs with little or no discernible statistical association among skeletal and non-skeletal trait size at P0. However, change in size is simply one aspect of development. Study of early postnatal growth (from P0 to P2) in the Fgfr2+/P253R mouse model for Apert syndrome indicates that size and shape change differently in the skull and brain, suggesting some level of independence in the pattern of growth for the two tissues (Hill et al., 2012). Future correlation analyses with larger sample sizes and additional data (i.e., quantitative measures of shape) will provide further insights into the interdependence of these developing tissues. We propose that variation in localized shape change of complex head traits contributes to the obvious differences in relative position and overall spatial association of traits in Fgfr2cC342Y/+ mice relative to unaffected littermates (Fig. 7).
Table 5. Correlations of Size of Skeletal and Non-Skeletal Structures of the Head in Fgfr2cC342Y/+ Crouzon Syndrome Mice (Below the Diagonal) and Unaffected Littermates (Above the Diagonal)
Complete bicoronal craniosynostosis was observed in 75% of Fgfr2cC342Y/+ newborn skulls (Table 3), in agreement with previous observations of bilateral coronal suture closure in 80% of adult mice (Perlyn et al., 2006), and confirming that coronal suture fusion occurs prenatally when this mutation is present. Fusion of the presphenoid-basisphenoid synchondrosis observed in adult Fgfr2cC342Y/+ mice (Perlyn et al., 2006) was not apparent in our newborn sample where mineralized bones of the cranial base are small and clearly separated from each other (Fig. 2 and Supp. Videos S1, S2). Consequently, premature fusion of mineralized cranial base elements cannot be the cause of the significant changes in cranial base, facial or cranial vault morphology quantified in Fgfr2cC342Y/+ skulls at P0. Our shape analysis reveals a boundary of cranial base effects at the midpoint of the cranial base where rostral cranial base elements and tissues that occupy the non-mineralized spaces between them are reduced in size while those caudal to the basisphenoid are elongated (Figs. 1 and 2). Whether or not this boundary of effects corresponds with the division between neural crest-derived (presphenoid, basisphenoid) and mesoderm-derived (basioccipital) elements of the cranial base (McBratney-Owen et al., 2008) cannot be determined directly by image data. Variation in cranial vault suture patency and changes in bone and brain shape contribute to neurocranial dysmorphology in Fgfr2cC342Y/+ mice.
Midfacial deficiency in Fgfr2cC342Y/+ mice was evidenced by a reduction in bones of the face and palate (Figs. 1, 2), variation in facial suture patency (Fig. 3), and significantly reduced nasopharyngeal volume (Fig. 6). Though midfacial hypoplasia, a characteristic feature of craniosynostosis phenotypes, is often associated with maldevelopment of facial bones, we have no evidence to limit the primary target of Fgfr2 mutations to osteoprogenitor cells. The primitive pharynx is composed of endoderm internally, which is Fgf-dependent for morphogenesis and it has been shown in zebrafish that pharyngeal endoderm plays a role in promoting the survival of local skeletogenic neural crest cells and patterning of bones and cartilages of the pharyngeal arches (Crump et al., 2004). Consequently, pharyngeal endoderm may also be a primary target of aberrant FGF/FGFR signaling that directly affects nasopharyngeal shape and indirectly affects the size and shape of facial bones developing within the pharyngeal arches. There is no doubt of the role of this Fgfr2 mutation in osteogenesis and maldevelopment of skeletal components, as this has been well studied. Our analyses reveal additional possible primary effects of the Fgfr2 IIIc variant in non-skeletal tissues of the head that might contribute to dysmorphogenesis of skull bones.
FGF/FGFR Signaling and the Role of Head Development in Vertebrate Evolution
Our analysis establishes parallels in human and mouse head phenotypes resulting from the C342Y mutation on FGFR2, revealing strong conservation of the evolved developmental genetic program that underlies mammalian head development and further validates the use of this mouse model in the analysis of the development of Crouzon syndrome phenotypes. These data provide new factors for future molecular and cellular investigations targeting the effects of this mutation on tissue-specific cell signaling that contributes to craniofacial dysmorphogenesis and the complex etiology of Crouzon syndrome head phenotypes. At the same time, our data provide a basis for thinking more broadly about organismal response to mutations as developmental reactions that are constrained to follow established evolutionary patterns.
FGFs are small proteins characterized by a conserved functional domain that act by binding to FGFRs causing them to homodimerize, and eventually activate and phosphorylate multiple cytoplasmic signaling cascades (Eswarakumar et al., 2005). FGFRs are key regulators of numerous cellular processes (Turner and Grose, 2010). The large number of genes coding for FGFs (at least 22 in vertebrates), along with four coding for their FGFRs appear to have been generated from two rounds of whole genome duplication with many functions added over evolutionary time through the accumulation of genes and of splice variants with different ligand-binding specifications (Itoh and Ornitz, 2004; Popovici et al., 2005). The expression of FGFs in neural crest cell derivatives, cranial placodes, and specific regions of the vertebrate brain, and the conservation of distinct ligand-binding properties of alternatively spliced mRNA variants of FGFRs increases their functional diversity and renders FGF/FGFR signaling pervasive in development of the head (Yeh et al., 2003; Itoh and Ornitz, 2004; Eswarakumar et al., 2005).
A growing body of evidence suggests that FGF/FGFR signaling is fundamental to diverse processes in normal vertebrate development (Ornitz and Itoh, 2001; Thisse and Thisse, 2005; Wilkie, 2005; Szabo-Rogers et al., 2008; Hatch, 2010; Hebert, 2011). Evidence presented here of the multiple phenotypic effects of a mutation in FGFR2 demonstrates the influence of FGF/FGFR signaling on the development of multiple cell and tissue types. Importantly, this evidence also suggests a role for FGF/FGFR signaling in the accommodation of changing tissue associations that occur during head morphogenesis, a role that would require compliant cell-to-cell signaling mechanisms in which cells of various tissues are cooperative partners in the development of the overall assembly. This type of system would not arise de novo with the introduction of a mutation in a modern vertebrate, but instead is a product of the evolution of developmental systems.
In a previous study of morphological integration of various parts of the skull in two Fgfr2 mutant mouse models for Apert syndrome (Martínez-Abadías et al., 2011), we suggested that FGF/FGFR signaling is a covariance-generating mechanism established early in vertebrate evolution that acts as a global factor modulating coordinated development of various skull components. We further suggested that this mechanism could constrain clinically relevant shape changes of the skull induced by Fgfr2 mutations (Martínez-Abadías et al., 2011). While skull morphological integration was assessed by Martínez-Abadías et al. (2011) by statistical analysis of covariation among the shapes of skull components, we show here that this covariance-generating system is not restricted to skeletal tissues.
Covariation generated by FGF/FGFR signaling occurs at the level of the cell. FGF/FGFR signaling may both induce specific patterns of gene expression and confer specific responses to naive cells providing dynamic accommodation of shape change as the head assembles from simple homogenous cell populations into precisely coordinated traits of mixed cell origin. Cells respond to the FGF/FGFR signal by changing size or shape, by differentiating, by dying, by dividing, by sending an appropriate signal, by enabling a cell to receive and respond appropriately to a signal, or by changing the genes the receiving cell uses to respond. Importantly, the tissue-specific and global responses of head tissues to FGFR mutations are constrained by cell-to-cell signaling patterns produced over millions of years of evolution. Knowledge of the evolution of these systems may ultimately bring us closer to harnessing their dynamics for therapeutic purposes.
In this work, we present a detailed analysis of the global and local phenotypic effects of the C342Y mutation on Fgfr2c on various skeletal and non-skeletal tissues of the head. The approach that we present is directly applicable to the study of other animal models of human disease. Our findings contribute to the growing evidence that FGF/FGFR signaling is part of a multifaceted set of interactions among genes and regulatory networks that drive communication among cells in the development of the head. The nature of the signals vary by cell type, location, and developmental timing so that the challenge becomes one of understanding the complexity of cell–cell signaling and phenotypic assembly at many levels across time (evolutionary and developmental) and space. In addition to the highly successful foundation of developmental biology built largely on molecular analyses of cells and tissues at specific developmental time points, the knowledge that complex craniofacial traits vary together by exploiting patterns of signaling, association, and covariation established over hundreds of millions of years of evolution should be used in the design of new tools for decoding the complex effects of specific mutations on head morphogenesis.
The generation and analysis of Crouzon syndrome Fgfr2cC342Y/+ mouse model are described elsewhere (Eswarakumar et al., 2004; Snyder-Warwick et al., 2010). Fgfr2cC342Y/+ mice were maintained on a CD1 genetic background. Our sample (Table 1) was generated and imaged after death in compliance with animal welfare guidelines approved by the Mount Sinai School of Medicine and the Pennsylvania State University Animal Care and Use Committees. Newborn mice (P0, gestation time was 19.0 ± 0.5 days) were euthanized by inhalation anesthetics and fixed in 4% paraformaldehyde. The genotypes were determined by PCR analysis of tail DNA.
μCT and MRM Imaging Protocols at P0
High-resolution micro-computed tomography (μCT) images with pixel size and slice thickness ranging from 0.014 to 0.016 mm were acquired by the Center for Quantitative X-Ray Imaging at the Pennsylvania State University (www.cqi.psu.edu) using the HD-600 OMNI-X high-resolution X-ray-computed tomography system (Bio-Imaging Research Inc, Lincolnshire, IL). Image data were reconstructed on a 1,024 × 1,024 pixel grid as 16 bit TIFFs but reduced to 8 bit for image analysis. High-resolution micro-magnetic resonance microscopy (MRM) images were acquired of the same specimens for which we have μCT images by the High Field MRI Facility at the Pennsylvania State University (www.imaging.psu.edu/facilities/high-field). The fixed animals were immersed in 2% Magnevist (Bayer Health Care, Wayne, NJ) phosphor-buffered solution for 10 days to reduce the T1 and T2 relaxation time. All experiments were conducted on a vertical 14.1 Tesla Varian (Varian Inc., Palo Alto, CA) imaging system with direct drive technology. To prevent drying and to minimize magnetic susceptibility artifacts during scanning, specimens were immersed in a flourinert liquid, FC-43 (3M, St. Paul, MN). A standard imaging experiment with an isotropic resolution of 80 μm comprised a field of view of 15.4 × 14 × 11 mm3 and a matrix size of 192 × 132 (75% partial Fourier: 176) × 137. With eight averages and a repetition time of 75 ms (echo time 25 ms), the total scan time was three hours. Matlab (The MathWorks, Inc., Natick, MA) was used for image postprocessing. By zero-filling all directions by a factor of two, the pixel resolution of a standard imaging experiment was 40 (μm)3.
Landmark Data Collection
Three-dimensional (3D) coordinates of 59 skull landmarks (Table 2, Fig. 1) were recorded on the 3D μCT isosurfaces reconstructed using AVIZO 6.2 (Visualization Sciences Group, VSG). Landmarks are locations on biological forms that hold biological significance and that can be precisely defined and located with an acceptable degree of measurement error (Richtsmeier et al., 2002). 3D coordinates of landmark locations were digitized twice on each specimen by the same observer and measurement error was minimized by averaging the coordinates of the two trials (Richtsmeier et al., 1995; Aldridge et al., 2005). To ascertain the accuracy and reproducibility of landmark placement, intraobserver error (i.e., absolute difference between the two trials) was checked for every landmark. If landmark placement differed by more than 0.05 mm, the position of the landmark was remeasured. In addition to skull shape defined by a selection of 39 landmarks that cover the whole skull, subsets of 3D landmarks were defined to represent the shape of the three major regions of the mammalian skull: facial skeleton (N=22), cranial vault (N=10), cranial base (N=15), and palate (N=8) (Table 2, Fig. 1).
For analysis of the brain, non-neural tissue was removed from the MRM images using Amira 5.2 (Visualization Sciences Group, VSG). Three-dimensional coordinates of 15 brain landmarks (Table 4, Fig. 4) were collected from the 3D reconstructions of the brain using etdips (http://www.cc.nih.gov/cip/software/etdips/). 3D coordinates of landmark locations were digitized twice by the same observer and averaged to minimize measurement error. Coordinate data were checked for gross errors, and the absolute difference between the two trials was compared; if landmark placement differed by more than 0.08 mm (i.e., two pixels), the position of the landmark was remeasured. Brain shape defined by all landmarks was analyzed, as were two subsets of landmarks defined to represent the shape of the surface of the brain and the shape of the brain deep to the surface (Table 4, Fig. 4). As the dominant feature of Crouzon syndrome is craniosynostosis and skull dysmorphology, we hypothesized that the effects of the Fgfr2c C342Y mutation would have differential effects on brain structures located adjacent to the cranium (i.e., the surface) as compared to structures located farther from the cranium (i.e., deeper, internal structures).
To extract shape information for each dataset/subset of skull and brain landmarks, we performed separate General Procrustes Analysis (GPA) to superimpose the coordinate data (Rohlf and Slice, 1990b; Dryden and Mardia, 1998) using the MorphoJ software package (Klingenberg, 2008). This procedure minimizes the influence of size and adopts a single orientation for all specimens by shifting the landmark configurations to a common position, scaling them to a standard size, and rotating them until a best fit of corresponding landmarks is achieved. GPA removes the effects of scale, translation, and rotation (Rohlf and Slice, 1990b), but does not eliminate the allometric shape variation that is related to size. In order to test how allometry affected our samples, we computed a regression of shape (represented by Procrustes coordinates) on centroid size (Drake and Klingenberg, 2008).
To statistically evaluate shape differences between Fgfr2cC342Y/+ mice and unaffected littermates, we used Euclidean Distance Matrix Analysis (EDMA; Lele and Richtsmeier, 2001). EDMA is a 3D morphometric technique that is invariant to the group of transformations consisting of translation, rotation, and reflection (Lele and McCulloch, 2002; Richtsmeier et al., 2002). The original 3D coordinates of landmark locations collected from the forms are re-written and analyzed as a matrix of all unique linear distances among landmarks called the form matrix, (FM) and are used to estimate a mean FM for each sample being considered following Lele and Richtsmeier (2001). Form difference between samples is evaluated by calculating ratios of like-linear distances using the mean FMs of the two samples. The matrix of ratios, the form difference matrix (FDM), is used to estimate a statistic against which information from boot-strapped samples are evaluated. The null hypothesis of similarity in shape is rejected if estimates from the data fall outside of the 95% bootstrapped confidence intervals. Further examination of the FDM identifies patterns of localized differences in form between the two samples. To statistically test for localized differences in form, an alternate nonparametric bootstrap procedure calculates the confidence interval for each linear distance (α = 0.10). If this interval contains the value 1.0, the null hypothesis of similarity for that linear distance is accepted. Confidence interval testing revealed the localized effects of the Fgfr2c C342Y mutation on the facial skeleton, the cranial base, the neurocranium, and the palate. EDMA analyses were performed using WinEDMA v. 1.0.1 (Cole, 2002).
Evaluation of Cranial Suture Fusion
All sutures were scored using isosurfaces segmented for bone using hydroxyapatite phantoms imaged with the specimens as described above. In the case of the appearance of fusion, we examined individual slice images for that portion of the suture to verify fusion. All sutures were scored bilaterally, excepting the inter-premaxillary and inter-palatine sutures that are midline sutures. For each specimen, sutures were scored qualitatively as open when more than 75% of the length of the suture was completely patent, partial when more than 25% but less than 75% of the length of the suture was patent, and fused when less than 25% of the length of the suture was patent. Given the discrete nature of patency of the right and left zygomatic-maxillary and premaxilla-maxillary facial sutures, we assigned a discrete score of either open or fused. Because the inter-premaxillary suture showed highly localized variation in patency, we attempted to determine whether or not this variation was patterned by dividing the suture into thirds for scoring. The divisions are: the most anterior third, medial third, and most posterior third (see Table 3).
Segmentation of Upper Airway, Vitreous Humor, and Inner Ear Volumes
The nasopharynx, eyes, and inner ear volumes of a subset of specimens were analyzed (Table 1). The irregular and relatively smooth shapes of these biological structures prevented the acquisition of repeatable landmark data. Consequently, volumetric analyses were conducted. To meet requirements for segmentation, the nasopharynx of the individual had to be visibly perfused with Magnevist solution (a contrast medium used in MRM; see “μCT and MRM Imaging Protocols at P0” above for details). Structures were segmented from MRM images of each mouse twice using label fields and the segmentation editor of Avizo version 6.3 (VSG). Quantification of volume was performed using the MaterialStatistics module of Avizo, computed as the number of voxels multiplied by the size of a single voxel. Mean volumes from the trials were used in analyses and statistical analyses were conducted in SPSS version 20 (IBM SPSS Statistics, SPSS Inc., v. 20). Threshold values (min/max range = 89–255) were set for each specimen depending upon the individual MRM scan. As the inner ears and vitreous body are closed structures, boundaries of these structures were defined by the change in density of tissues that defined the outer limits of these regions. In essence, virtual endocasts of the fluid-filled inner ear structures and the vitreous humor-filled vitreous body were segmented and their volumes were estimated. Since the nasopharynx is a structure that continues to the oropharynx and trachea, an inferior endpoint was needed in order to limit the segmentation. To accomplish this, 3D coordinates of 4 neural landmarks (denoting the most dorsolateral point of intersection of the olfactory bulb with rostral frontal lobe surface [left and right] and the origin of middle of cerebral artery from the circle of Willis on ventral cerebral surface [left and right]) (Aldridge et al., 2010) were located on the MRM of the brain using Avizo. These landmarks were used to create an oblique plane through the specimen that defined the caudal endpoint of the nasopharynx as measured in this study. The nasopharynx of each specimen was segmented and the volume estimated twice by the same observer.
We thank Timothy Ryan, Timothy Stecko, and Thomas Neuberger for excellent technical work in acquiring μCT and MRM images. Ken Weiss, Anne Buchanan, Chris Percival, and Yann Heuzé offered helpful suggestions throughout the development of this study. David Ornitz generously provided the Fgfr2cC342Y/+ Crouzon syndrome mice for this study. This research was supported in part NIDCR/NIH R01DE018500 (to J.T.R.), R01DE022988 (to J.T.R. and E.W.J.), NIDCR/NIH and ARRA 3R01DE018500-02S1 (to J.T.R. and E.W.J.), Comissionat per a Universitats i Recerca (CUR), Generalitat de Catalunya, Spain 2008 BP A 00170 (to N.M.A.).