Phenotype-Genotype Correlations in Mouse Models of Amelogenesis Imperfecta Caused by Amelx and Enam Mutations

Mutations in human and in mouse orthologous genes Amelx and Enam result in a diverse range of enamel defects. In this study we aimed to investigate the phenotype-genotype correlation between the mutants and the wild-type controls in mouse models of amelogenesis imperfecta using novel measurement approaches. Ten hemi-mandibles and incisors were dissected from each group of AmelxWT, AmelxX/Y64H, AmelxY/Y64H, AmelxY64H/Y64H, and EnamWT, EnamRgsc395 heterozygous and EnamRgsc395 homozygous mice. Their macro-morphology, colour and micro-topography were assessed using bespoke 2D and 3D image analysis systems and customized colour and whiteness algorithms. The novel methods identified significant differences (p ≤ 0.05) between the Amelx groups for mandible and incisor size and enamel colour and between the Enam groups for incisor size and enamel colour. The AmelxWT mice had the largest mandibles and incisors, followed in descending order of size by the AmelxX/Y64H, AmelxY/Y64H and AmelxY64H/Y64H mice. Within the Enam groups the EnamWT incisors were largest and the EnamRgsc395 heterozygous mice were smallest. The effect on tooth morphology was also reflected by the severity of the enamel defects in the colour and whiteness assessment. Amelogenin affected mandible morphology and incisor enamel formation, while enamelin only affected incisors, supporting the multifunctional role of amelogenin. The enamelin mutation was associated with earlier forming enamel defects. The study supported the critical involvement of amelogenin and enamelin in enamel mineralization.


Introduction
Many genes underlying normal and abnormal dental development have been identified using studies on mouse models [Thesleff, 2006;Fleischmannova et al., 2008]. However, the underlying pathogenesis of the clinically and genetically heterogeneous group of enamel defects, amelogenesis imperfecta (AI), requires further investigation [Wright et al., 2009]. Correlating accurately defined phenotypes with different genotypes will contribute to understanding the effect of specific mutations responsible for AI.

Morphometry
In humans quantitative methods for clinical phenotyping of the dentition have been developed, from hand calliper measurements to two-dimensional (2D) [Brook et al., 1986[Brook et al., , 2005 and now three-dimensional (3D) imaging [Smith et al., 2009b]. Each new advance has enabled additional parameters to be determined and phenotyping to be enhanced.
Murine mandibles and incisors represent excellent models of complex morphological structures [Atchley and Hall, 1991], including a permanent record of all developmental stages of enamel formation, for which quantification of the effects of specific variables is valuable [Cooper and Albertson, 2008].

Enamel Colour and Surface Roughness Assessment
The physico-optical properties of dental hard tissues reflect the normal and abnormal process of mineralization . Tooth morphology and the enamel crystal surface [Risnes, 1979;Warshawsky et al., 1987] affect light reflectance and influence colour.
Colour distribution in human incisors has been used to assess enamel in three anatomical regions -cervical, middle and incisal [Brook et al., 2007]. In mice enamel development has been separated into various different stages, including secretory, pre-eruptive and mature [Robinson et al., 1983;Smith and Nanci, 1989;Wong et al., 2000;Smith et al., 2009a].
International recommendations for the objective measurement of colour [Commission Internationale de l'Eclairage, 1986] and surface texture [International Organization for Standardization, 1996] have been calibrated for human clinical trials [Smith et al., 2008] and used to interrogate enamel surface mineralization [Zhang et al., 2000;Higham et al., 2009]. These novel methods will provide a complementary approach to reflect enamel development, structure and function.
Amelogenesis Imperfecta AI is a clinically and genetically heterogeneous group of inherited dental enamel defects [Witkop, 1967]. Amelogenesis is orchestrated by genetic regulation of the secretion, organization and processing of the developing enamel extracellular matrix (ECM) [Wright, 2006]. A number of mutations in the amelogenin gene (AMELX) [Hart et al., 2002;Kim et al., 2004] and the enamelin gene ( ENAM ) [Rajpar et al., 2001;Mardh et al., 2002] are implicated in the aetiology of types of AI in humans and mice [Wright et al., 2009].

Phenotyping Mouse Models
Similar enamel phenotypes have been reported for some mutant mice and humans [Gibson et al., 2001;Hart et al., 2002;Seedorf et al., 2007]. Anatomical, histological and ultrastructural assessment of mice models of AI have revealed key factors underpinning the molecular pathogenesis of AI, e.g. the disrupted secretion of enamelin interferes with initial enamel crystal formation [Masuya et al., 2005] and ameloblast cell binding and the intracellular proetin trafficking may have a mechanistic role in the failed secretion of amelogenin into the ECM [Barron et al., 2010].
However, objective morphometric measurement of mandibles and incisors, together with colour and whiteness and surface assessment of incisors, has not been described in any previous study. Quantitatively assessing the phenotype of mandibles and incisors associated with known mutations Enam S55I [Masuya et al., 2005] and Amelx Y64H [Barron et al., 2010] will aid in studying the outcome of the mutations and relating the phenotype to the genotype of these two pertinent mouse models of AI.
The aim of this study was to examine and interpret correlations between genotype and phenotype in three Amelx mouse groups, two Enam groups and their respective wild-type controls. Such correlations should provide further insight into the functions of amelogenin and enamelin during normal dental development and strengthen our understanding of aberrant enamel mineralization.

Materials and Methods
The mice were bred and reared under identical standard conditions in accordance with the Animals (Scientific Procedures) Act, UK, 1986. All mice were euthanized at the same age (90 days) and weight (25 8 5 g). Mice were gender matched within and between groups, and they were fed a soft diet. Mice were examined using the protocols of European Mouse Phenotyping Resource of Standardized Screens (www.empress.har.mrc.ac.uk/). Ethics approval was granted according to Wellcome programme GRO75945MA ethics reference number 06/Q0104/38. DNA was extracted from ear skin samples of each mouse used in the study. Genotyping, using these DNA samples, was performed for each mouse using oligonucleotide primers to PCR amplify the Amelx and Enam genes followed by DNA sequencing using dye primer chemistry.

Mandible and Incisor Extraction
Specimens were preserved in 10% neutral buffered formalin and washed in phosphate-buffered saline and distilled water before examination. Micro-dissection of hemi-mandibles and incisors was carried out under a dissection microscope (Bresser, Meade Instruments Corp., Irvine, Calif., USA). Incisors were removed after hemi-mandible imaging with care to avoid mechanical damage to surface enamel. Any incisors that were seen to be damaged at the microscopic level were discarded from the study, according to strict visual and tactile criteria, e.g. scalpel marks, and so did not interfere with roughness measurements. Specimens were kept on ice to minimize any temperature effects or dehydration during imaging.

Imaging
Standardized 2D images were taken with a 13.5-megapixel Kodak DCS Pro SLR/n (Eastman Kodak Company, Geneva, Switzerland) digital camera using an established image analysis system (IAS) [Brook et al., 2005]. An MP-E 65-mm F2.8 1-5 ! Macro Photo Lens (Sigma Corp., Kanagawa, Japan) was used. Each image contained an 11.0-mm scale and was calibrated individually using Image Pro Plus version 5.1 software (Media Cyberenetics, Inc., Bethesda, Md., USA).
The 3D IAS consisted of a customized non-contact surface profilometer (Scantron ProScan 2000; ScanTron Industrial Products Ltd., Taunton, UK) with a high systematic resolution (1.0 m). The device was modified to rotate incisors through 360° and 3D models were constructed using SolidWorks Premium 2008 software (Dassault SolidWorks, Waltham, Mass., USA).
The colour and whiteness images of the incisor labial surface were captured to contain the whole enamel surface. Images were automatically calibrated against a spectrophotometrically assessed standardized white tile (British Ceramic Research Association). Polarized images avoided interference from surface reflections.

Morphometric Assessment
After 2D image acquisition the specimens were automatically outlined by the Image Pro Plus software and morphological measurements obtained from the hemi-mandibles ( fig. 1 a) and incisors ( fig. 1 b). All 3D morphological measurements were obtained using Cloud 3D surface viewer software (Dr. Robin Richards, Westcott Road, London) that enabled both projected and actual linear measurements ( fig. 2 ).

Enamel Colour and Whiteness, and Surface Roughness Assessment
For colour and whiteness assessment, incisors were held in a customized holder with the proximal end fixed in black modelling clay. The buccal surface enamel was imaged from the labial view ( fig. 3 ). At the proximal end of each incisor a distinctive colour and surface texture change was used as a readily identifiable anatomical landmark feature. This was consistently observed in the 2D morphometric images, the colour and whiteness images and in the 3D images. This white opaque boundary was previously reported [Robinson et al., 1983;Smith and Nanci, 1989].
Images were opened in customized Adobe Photoshop software CS2 version 9 (Adobe Systems, Inc., San Jose, Calif., USA) and the 'Magnetic Lasso Tool' feature was used to objectively trace the observable incisor perimeter that encompassed the whole la-  (2) angle of curvature (degrees); (3) width at midpoint (mm); (4) labial length (mm); (5) incisor perimeter (mm), and (6) incisor area (mm 2 ). Left hemi-mandible and mandibular incisor shown from the buccal view. Scale bar = 11.0 mm. bial enamel surface. Using customized hotkeys, red, green and blue colour channel outputs were automatically derived either from the whole enamel surface area [(i) whole] or from one of the three regions that were separated equidistantly into (ii) cervical, (iii) middle and (iv) incisal. Simultaneously, a Microsoft Excel (Microsoft Corp., Albuquerque, N. Mex., USA) algorithm used the red, green and blue outputs to calculate calibrated CIE L, A and B (L = lightness, A = green/red, B = yellow/blue) colour space and WI (WI = whiteness) values, where L = 0 yielded black and L = 100 yielded white [Smith et al., 2008]. These additional three regions were chosen as they provided a greater degree of analysis than just using the entire surface as a whole. The authors appreciate that the actual underlying histological regions differ; however, they are not visible at the tooth surface to use as a method guide. The software and algorithm minimized human subjective input and error, were highly reproducible, objective and practical and expedited data collection efficiently.
A surface roughness measurement was taken in each of the three specific enamel surface regions in a 3D image from each experimental group (sample size n = 1) using ProScan 2000 software (ScanTron Industrial Products Ltd.) ( fig. 4 ). A 200 ! 500 m area was selected equidistantly along the longitudinal axis of the incisor to minimize subjectivity. The surface roughness measurement quantified enamel surface texture, as distinct from form or waviness components [International Organization for Standardization, 1996].

Reliability and Validation
Measurements were taken by two independent operators. The initial and repeat imaging and measurements were carried out on different days. No repeats were obtained for surface roughness. Using SPSS (SPSS, Inc., Chicago, Ill., USA), intra-class correlation coefficient (ICC), Pearson's correlation coefficient (PCC) and repeated measures t tests were used to determine method reliability and agreement [Fleiss, 1986]. Using MedCalc (MedCalc bvba, Mariakerke, Belgium) Bland-Altman plots visualized limits of agreement and bias.

Phenotypic Comparison
One operator implemented the phenotypic comparisons between the four Amelx genotype groups and the three Enam genotype groups. Percentage values were obtained by dividing the number of significantly different variables by the total number of variables to indicate the relative number of significant differences found using each variable. Descriptive statistics provided the mean, mean difference, standard error and 95% confidence intervals. Bonferroni's corrected one-way ANOVA multiple comparisons (p = 0.002) and post hoc Tukey's honestly significant difference tests (p = 0.05) were used to identify significant phenotype variation between groups. Significant differences (p = 0.05) observed before the robust Bonferroni correction are also detailed because of the varying degrees of independence of the variables.

Reliability and Validation
Intra-operator repeatability (ICC 6 0.75) and interoperator reproducibility (ICC 6 0.77) were predominantly substantial to excellent for all 2D and 3D morphometric variables according to the classification of Donner and Eliasziw [1987]. A similarly excellent (ICC 6 0.96) intra-operator repeatability was demonstrated for colour and whiteness assessment across all of the enamel surface regions.
The 2D and 3D methods showed significant (p ^ 0.01) method agreement (PCC 0.710-0.999) [Rodgers and Nicewander, 1988]. Repeated measures t tests showed no significant differences (p 6 0.01) between measurements, except for the width-at-midpoint variable. The Bland-Altman plots used to assess limits of agreement and bias gave satisfactory results Altman, 1986, 1999].

Phenotypic Comparison Amelx Groups
Twenty-five percent of mandible and 82% of incisor 2D variables showed significant differences (p ^ 0.05) between one or more of the Amelx groups ( table 1 ). The Amelx WT group had the largest mandibles (e.g. ascending height and mandible angle) and incisors (e.g. overall length, perimeter and area), followed by the Amelx X/Y64H , then the Amelx Y/Y64H and finally the Amelx Y64H/Y64H groups (online suppl. fig. 1; for all online suppl. material, see www.karger.com/doi/10.1159/000336440).
Eighty-four percent of colour and whiteness variables showed significant differences ( ^ 0.05) between the Amelx groups, notably between all of the groups in the incisal and whole regions ( table 1 ). The major colour and whiteness differences between the Amelx WT and Amelx Y/Y64H groups and between the Amelx WT and Amelx Y64H/Y64H groups occurred in the lightness, yellow/blue and whiteness colour components (online suppl. fig. 1). The Amelx WT and Amelx X/Y64H groups' yellow/blue values were similar and significantly higher than those of the Amelx Y/Y64H and Amelx Y64H/Y64H groups.
Seventy-three percent of incisor 3D variables showed significant differences (p ^ 0.05) between one or more of the Amelx groups ( table 1 ). The Amelx WT incisors were the largest (e.g. surface area and volume), and the Amelx Y64H/Y64H incisors were the smallest; the Amelx X/Y64H and Amelx Y/Y64H incisors were of an intermediate size (online suppl. fig. 2).

Enam Groups
There were no significant differences in mandible morphometry between the Enam groups. Only 14% of incisor 2D variables showed significant differences (p ^ 0.05) between the Enam WT and Enam Rgsc395 heterozygous groups ( table 2 ).
Forty-one percent of the colour and whiteness variables showed a statistically significant difference ( ^ 0.05) between the Enam groups ( table 2 ). Significant colour and whiteness differences occurred between the Enam WT and Enam Rgsc395 heterozygous groups and between the Enam WT and Enam Rgsc395 homozygous groups (online suppl. fig. 3). The average yellow/blue component of Enam WT was higher than the Enam Rgsc395 heterozygous and the Enam Rgsc395 homozygous groups in the middle, incisal and whole regions in that order.
Eighty percent of incisor 3D variables showed significant differences (p ^ 0.05) between the Enam WT and the Enam Rgsc395 heterozygous mice ( table 2 ). The Enam WT incisors were the largest (e.g. projected overall length), followed by the Enam Rgsc395 homozygous group and then the Enam Rgsc395 heterozygous group (online suppl. fig. 3).
Regarding surface roughness assessment, Amelx WT and Enam WT had similar marginally higher values compared to their respective mutant groups ( tables 1 , 2 ). In all groups the enamel surface roughness increased through the cervical, middle and incisal surface regions (online suppl. fig. 2, 4).

Discussion
Reliability was substantial to excellent for almost all variables. The 2D and 3D method agreement validated the new 3D IAS [Rodgers and Nicewander, 1988]. The complementary methods provide objective approaches to quantitative phenotypic analysis of mice mandibles and incisors that are of comparable reliability to those used for human teeth [Smith et al., 2009b]. These novel experimental approaches have considerable potential for future applications, for example small mammalian dentition and other similar murine model organisms.
The significant mandible morphological differences found between the Amelx groups, but not between the Enam groups, support a role for amelogenin in mandible development. Amelogenin is expressed in various developing structures including dental supporting tissues and during alveolar bone formation and remodelling [Gruenbaum-Cohen et al., 2008]. Our findings are consistent with its involvement in formation and growth of the  Mean values are shown. Values in parentheses represent the standard error. One-way ANOVA multiple comparisons and post hoc tests determined * Bonferroni corrected significant differences (p ≤ 0.002) and ** significant differences (p ≤ 0.05). Right side not displayed for brevity. mandible ramus. This supports amelogenin's role as a multifunctional protein in the craniofacial complex.
There was evidence of significant phenotypic differences between the controls and the mutant mice. These enamel mineralization defects were associated with the absence of the full length amelogenin and enamelin proteins in the developing enamel ECM [Seedorf et al., 2007;Smith et al., 2009a]. The significant macroscopic differences between the incisors, for the wild-type controls and the mutant groups, concur with the important contribution of amelogenin and enamelin in structural organization and enamel mineralization, detectable at the phenotype level.
The 2D and 3D morphological data suggested the enamel in the Enam Rgsc395 homozygous incisors was less affected than that in the Enam Rgsc395 heterozygous mutants. This contrasted with the report by Smith et al. [2009a] that indicated a more severe phenotype for the Enam Rgsc395 homozygous null mutants. However, only 14% of the 2D morphological variables were significantly different between the Enam wild types and the Enam heterozygous mice. Also, there were no significant differences between the 3D incisor morphology of the Enam wild types and the Enam homozygous, or between the Enam heterozygous and Enam homozygous mice.
The wild-type mouse incisors showed typical rodent enamel and dentine colouration, i.e. opaque white with yellow/orange/brown colouration, due to the deposition of iron pigment in the superficial layer of enamel [Halse, 1972]. However, this layer may have been disrupted during preeruptive enamel maturation leading to the observed chalky white enamel indicative of porous hypoplastic enamel of AI. The significant differences in colour and whiteness between the Amelx X/Y64H and Amelx Y/Y64H groups reflect a mosaic genotype in the Amelx X/Y64H females concordant with the expression of the mutant Amelx allele according to lyonization [Lyon, 1961] or X-chromosomal inactivation [Huynh and Lee, 2005]. The Amelx X/Y64H females showed hypomineralized enamel and the Amelx Y/Y64H males and Amelx Y64H/Y64H females displayed thin severely hypoplastic enamel characteristic of AI [Wright, 2006].
The Amelx WT and Enam WT control groups both had high yellow/blue and low whiteness and lightness values in the incisal region, correlating with the contribution of both amelogenin and enamelin to the intact and normally mineralized enamel phenotype. In contrast, the mutant Amelx and Enam mice had significantly lower yellow/blue values and higher whiteness and lightness values. Smith et al. [2009a] reported flaky enamel in the Enam homozygous null incisors compared to the Enam heterozygous incisors; however, both the Enam Rgsc395 heterozygous and homozygous mice incisors displayed flaky enamel.
The site of these significant differences varied between the Amelx groups and the Enam groups: in the Amelx mice the differences between the groups were found in the incisal region, while in the Enam mice the differences between the groups occurred in both the middle and incisal regions. This suggests th at Enam may have an earlier, more generalized effect on colour and whiteness than Amelx . The study differentiates between the overlapping enamel phenotypes of hypomineralized Amelx X/Y64H females and severely hypoplastic Amelx Y/Y64H males and Amelx Y64H/Y64H females, and local hypoplastic Enam Rgsc395 heterozygous and homozygous mice, according to the two mutations, in an enamel surface regionspecific manner that correlates to the distinct stages of enamel formation [Gibson et al., 2007].
The enamel surface roughness increased through the cervical, middle and incisal surface regions that represented the progressive developmental stages of enamel mineralization. However, the sample size did not allow for statistical significance to be tested and was an experimental limitation. Also, these findings contrasted with the diminishing surface roughness expected from a loss of organic matrix and an increasingly smooth crystal surface morphology as revealed by atomic force microscopy [Kirkham et al., 1998]. The incisal surface region was the only enamel surface region to have erupted into the oral cavity and be exposed to attrition or abrasion. However, all mice were maintained under identical standard conditions and fed on soft diets, which minimized the potential impact of any external environmental influences.
The increased surface roughness observed was consistent with the presence of pathological enamel, as mutations that disrupt ECM processing impair enamel mineral formation and disrupt crystal morphology [Robinson et al., 2003]. This supports the recently proposed hypothesis that intracellular protein-protein interactions involved in the secretion of amelogenin are a key mechanistic factor underpinning AIH1 [Barron et al., 2010].
Comparing these phenotypic observations in mice with the enamel defects due to AMELX and ENAM mutations in humans [Gibson et al., 2001;Hart et al., 2002] must be undertaken with care [Masuya et al., 2005;Gibson et al., 2007], noting the variation in the splicing of amelogenin, the cleavage products of enamelin, in protein function and epigenetic effects Hu et al., 2007;Wright et al., 2009].
In conclusion, significant mandibular and incisor morphometric as well as colour and whiteness differences between the wild-type controls and specific mutant phenotypes were related to aberrant enamel mineralization caused by the presence of amelogenin and enamelin proteins during amelogenesis. The multifunctional role of amelogenin in mandibular development was also supported.