ENAM mutations and digenic inheritance

Abstract Background ENAM mutations cause autosomal dominant or recessive amelogenesis imperfecta (AI) and show a dose effect: enamel malformations are more severe or only penetrant when both ENAM alleles are defective. Methods Whole exome sequences of recruited AI probands were initially screened for mutations in known AI candidate genes. Sanger sequencing was used to confirm sequence variations and their segregation with the disease phenotype. The co‐occurrence of ENAM and LAMA3 mutations in one family raised the possibility of digenic inheritance. Enamel formed in Enam+/+Ambn +/+, Enam+/−, Ambn+/−, and Enam+/−Ambn+/− mice was characterized by dissection and backscattered scanning electron microscopy (bSEM). Results ENAM mutations segregating with AI in five families were identified. Two novel ENAM frameshift mutations were identified. A single‐nucleotide duplication (c.395dupA/p.Pro133Alafs*13) replaced amino acids 133‐1142 with a 12 amino acid (ATTKAAFEAAIT*) sequence, and a single‐nucleotide deletion (c.2763delT/p.Asp921Glufs*32) replaced amino acids 921‐1142 with 31 amino acids (ESSPQQASYQAKETAQRRGKAKTLLEMMCPR*). Three families were heterozygous for a previously reported single‐nucleotide ENAM deletion (c.588+1delG/p.Asn197Ilefs*81). One of these families also harbored a heterozygous LAMA3 mutation (c.1559G>A/p.Cys520Tyr) that cosegregated with both the AI phenotype and the ENAM mutation. In mice, Ambn+/− maxillary incisors were normal. Ambn+/− molars were also normal, except for minor surface roughness. Ambn+/− mandibular incisors were sometimes chalky and showed minor chipping. Enam+/− incisor enamel was thinner than normal with ectopic mineral deposited laterally. Enam+/− molars were sometimes chalky and rough surfaced. Enam+/−Ambn+/− enamel was thin and rough, in part due to ectopic mineralization, but also underwent accelerated attrition. Conclusion Novel ENAM mutations causing AI were identified, raising to 22 the number of ENAM variations known to cause AI. The severity of the enamel phenotype in Enam+/−Ambn+/− double heterozygous mice is caused by composite digenic effects. Digenic inheritance should be explored as a cause of AI in humans.

Enamelin is particularly critical for dental enamel formation (Hu & Yamakoshi, 2003). When both Enam alleles are defective or absent in mice, no true enamel layer forms . As the dentin mineral continues to thicken, no enamel ribbons are deposited on the dentin surface, and ameloblasts become increasingly pathological, and some epithelial cells undergo apoptosis (Hu, Lertlam, et al., 2011). Instead of enamel, ameloblasts in Enam -/mice deposit a thin, disorganized mineralized crust (rather than true enamel) on the dentin surface as well as ectopic calcifications within the enamel organ Smith et al., 2009). Enamelin function is dependent upon phosphorylation by Golgi Casein Kinase (Cui et al., 2015). The absence of a single phosphorylation causes severe enamel defects in humans (Chan et al., 2010) and in mice (Yan et al., 2017).
The enamel phenotypes caused by ENAM mutations in humans show a clear dose effect. When a single ENAM allele is defective, the enamel malformations vary in severity and can be nonpenetrant (Seymen et al., 2014). The mildest observable enamel defects are well-circumscribed enamel pits, often arranged in horizontal lines , or horizontal grooves, usually in the cervical third of the crown (Kang et al., 2009;Mardh et al., 2002), with the most severe heterozygous phenotype being profound generalized thinning of the enamel layer (Hart, Michalec, Seow, Hart, & Wright, 2003;Kida, Ariga, Shirakawa, Oguchi, & Sakiyama, 2002). When both ENAM alleles are defective, the enamel layer is either completely absent or appears as a very thin mineral layer that only partially covers the crown Ozdemir et al., 2005;Simmer, Estrella, Milkovich, & Hu, 2013). The severity of enamel defects caused by a single defective Enam allele in mice varies in different teeth. Enam +/− maxillary incisors deposit virtually normal enamel, while the enamel mineral formed on Enam +/− mandibular incisors is reduced by as much as 50% (Smith et al., 2009).
Nearly 20 genes have been found to cause isolated (nonsyndromic) forms of AI, and more AI-causing genes have yet to be identified (Smith et al., 2017). Global efforts are underway to identify the genes and mutations that cause inherited enamel defects. By identifying the critical participants in amelogenesis, genetic studies are improving our understanding of the basic mechanisms of dental enamel formation while providing the means for genetic testing to diagnose the cause of inherited enamel malformations in patients. Many genes are also known to cause AI in syndromes (Wright, Carrion, & Morris, 2015), where the enamel malformations often precede the onset of other phenotypes, including kidney calcifications (Jaureguiberry et al., 2012;Wang, Aref, et al., 2013) renal tubular acidosis (Demirci, Chang, Mah, Romero, & Gorin, 2006), immunodeficiency (Lacruz & Feske, 2015), and blindness (Parry et al., 2009). A diagnosis that identifies the genetic cause gives the affected family the relief of knowing their condition is isolated, or awareness that future health risks go along with their enamel malformations and might be mitigated by therapeutic interventions.
Candidate gene approaches for genetic testing to identify the etiology in AI cases are impractical due to the increasing number of AI candidate genes and unnecessary because of advances in high-throughput DNA sequencing, which can identify sequence variations throughout the exome in a single experiment . Whole exome sequence (WES) analyses can also identify potential digenic inheritance, the simplest form of genetically complex diseases (Schaffer, 2013). Although human enamel malformations are not yet associated with digenic inheritance, Mmp20 +/− /Klk4 +/− double heterozygous mice show an enamel phenotype, whereas the single heterozygous (Mmp20 +/− or Klk4 +/− ) mice do not .
In this study we identify five families with hypoplastic AI caused by dominant heterozygous or recessive compound heterozygous ENAM mutations. All six affected members of one family also expressed a potentially deleterious heterozygous LAMA3 defect, raising the possibility of digenic effects. The enamel phenotypes of Enam +/− and Ambn +/− mice were compared with Enam +/− /Ambn +/− double heterozygous mice, demonstrating composite digenic effects in mice harboring a single allele Enam defect.

| Recruitment and ethical compliance
The study protocol and subject consent forms were reviewed and approved by the Ethics Committee at the University of Istanbul, and the Institution Review Boards at the University of Michigan, the University of North Carolina, and National Taiwan University Hospital. Two unrelated Turkish families with generalized hypoplastic enamel were characterized and recruited by Dr. Seymen and her team. Two unrelated Caucasian families with hypoplastic AI were recruited by Dr. Tim Wright at the University of North Carolina and Dr. Jan Hu at the University of Michigan School of Dentistry. A Taiwanese family was recruited by Dr. Shih-Kai Wang at National Taiwan University. Study explanation, pedigree construction, subject enrollment, clinical examinations, and collection of blood or saliva samples were completed under the proper consenting procedure specified in the study protocols and according to the Declaration of Helsinki.
This study complied with US National Research Council's Guide for the Care and Use of Laboratory Animals, the US Public Health Service's Policy on Humane Care and Use of Laboratory Animals, and Guide for the Care and Use of Laboratory Animals.

| Whole-exome sequencing and bioinformatics analysis
Either the nonstimulated saliva sample of 2 ml or peripheral blood sample of 5 ml was collected from each participant. Each sample was inspected, coded, then a small aliquot was removed for genomic DNA isolation following the manufacturer's protocol. Genomic DNA quality was assessed by 1.5% agarose gel electrophoresis and quantity was determined using a Qubit TM Fluorometer (ThermoFisher Scientific). Samples from the parents and proband of each family were selected for wholeexome sequencing (WES), while DNA samples from the other family members were used for segregation analyses. DNA samples, following the initial quality control, were submitted to Johns Hopkins Center for Inherited Disease Research (CIDR) for WES. Each DNA sample, at the concentration of 50 ng/µl, volume of 50 µl, and total amount of 2.5 µg, was plated onto a 96-well plate. A manifest file with coded sample information and the plated samples were shipped to the CIDR overnight on dry ice. Each sample was genotyped using an Illumina QC Array. Once sample aliquoting errors were ruled out, and the performance potential and genotypes were determined to be appropriate, then samples were subjected to the WES procedure. Exome capture was completed using the SureSelect Human All Exon Enrichment System (Agilent Technologies). Pairedend sequencing was generated using the Illumina HiSeq 2500 (CIDR). Sequencing reads were aligned to the 1000 genomes phase 2 (GRCh37) human genome reference using BWA version 0.7.8 (Li & Durbin, 2010). Duplicate reads were flagged with Picard version 1.109. Local realignment around indels and base call quality score recalibration was performed using the Genome Analysis Toolkit (GATK) (McKenna et al., 2010) version v3.3-0. GATK's reference confidence model workflow was used to perform joint sample genotyping to generate a multi-sample VCF file. Variant filtering was done using the Variant Quality Score Recalibration (VQSR) method (DePristo et al., 2011). Multi-sample VCF files from each family containing variants that were polymorphic among the family members were extracted from the multi-sample VCF file derived from the specific cohort with similar phenotypes. All variants in individual VCF files were annotated using VarSeq (Golden Helix) against a variety of data sources including gene annotation, function prediction, and frequency information (a cutoff value of 0.01 for the minor allele frequency). Following the comparisons between the affected and unaffected individuals, a list of prioritized variants was then subjected to segregation analysis. ENAM mutations are reported with respect to the first nucleotide of the ENAM genomic reference sequence (NG_013024.1) and the first nucleotide of the ENAM translation initiation site in the ENAM cDNA reference (NM_031889.2). The LAMA3 sequence variation associated with the disease phenotype in Family 5 is reported with respect to the first nucleotide of the translation initiation site in the LAMA3 cDNA reference sequence (NM_198129.2).

| Segregation analyses using sanger sequencing
The prioritized DNA sequence variations and their segregation within each family was confirmed by Sanger sequencing. The PCR primers were designed to bracket the candidate variant, and the reactions were conducted following the established protocols (Kim et al., 2005). PCR primers and amplification conditions are provided in Table S1.

| Scanning electron microscopy
Tissue processing: 7-week-old mice were anaesthetized with isoflurane and fixed by cardiac perfusion. Blood was cleared from the vasculature using lactated Ringer's solution (30-45 s) followed by 4% paraformaldehyde in PBS (135 mM NaCl, 2.7 mM KCl, 4.3 mM Na 2 HPO 4 , 1.4 mM Na 2 H 2 PO 4 ; pH 7.3) for 20 min. The mouse mandibles were removed and stripped of soft tissue, fixed by immersion in 4% paraformaldehyde overnight, and washed in PBS 3x (every 0.5-1 hr) with one overnight wash at 4°C. The embedding, cross-sectioning, polishing and imaging using a Cameca SX-100 Electron Microprobe Analyzer on backscatter mode were described previously (Hu, Hu, Smith, Bartlett, & Simmer, 2011).

| RESULTS
Whole exome sequence analyses of genomic DNA obtained from families with amelogenesis imperfecta (AI) identified five probands with potential disease-causing mutations in ENAM. Sanger sequencing confirmed the presence of these ENAM mutations, determined their distribution among the recruited members in each family, and established segregation of the ENAM defects with the AI phenotype.

| Family 1
The enamel malformations in Family 1 (Figure 1) followed an autosomal dominant pattern of inheritance with no consanguinity. Five members of this Turkish family were recruited, with three being affected (Figure 1a). The three affected members harbored one defective ENAM allele that had a frameshift mutation in Exon 7 (XM_006714056.4: c.395dupA/p.Pro133Alafs*13) (Figure 1b), which has not been reported previously (Table S2). This single-nucleotide duplication truncates the enamelin protein (normally 1,142 amino acids) after Lys132 and replaces amino acids 133-1142 with the 12 amino acid sequence ATTKAAFEAAIT. The aberrant transcript from the mutant allele may not be translated into protein as its mRNA is likely to undergo nonsense mediated decay (Kurosaki & Maquat, 2016). The three affected individuals (III:1, III:2, and III:5) were all heterozygous for this frameshift mutation, while the two unaffected individuals (II:6 and III:4) did not have this mutation on either ENAM allele and had normal enamel ( Figure S1). The proband's father (II:5) and oldest brother were not enrolled and their affection statuses and genotypes are unknown, although it is presumed that the three affected siblings characterized in this study inherited the mutant ENAM allele from their father. The proband (III:2) presented with generalized hypoplastic enamel with brown discoloration, pitting, and marked attrition of the occlusal and incisal surfaces ( Figure  1c). The anterior teeth exhibited horizontal grooves and banding, predominantly on their buccal surfaces. The incisal edges of these teeth (at age 24) showed signs of wear and chipping on their incisal edges. The posterior dentition was affected by dental caries, with large cavitated carious lesions mainly on occlusal surfaces. The bicuspids had rough surfaces where the enamel had apparently crumbled near the cusp tips. The proband's affected younger sister at age 13 (III:5; Figure 1d) displayed similar hypoplastic enamel with generalized interdental spacing, anterior incisal wear, chipping and horizontal banding, and posterior chipping of cusps, occlusal wear, and occlusal caries. The younger sister's (III:5) teeth were not as brown as those of the proband. The oldest brother (III:1) was affected, but the enamel phenotype was relatively mild ( Figure S2). His teeth were not significantly stained ( Figure  S3a). The panorex detected generalized enamel hypoplasia ( Figure S2c), although interdental spacing was only observed between the anterior teeth. Many well-circumscribed enamel pits were observed most notably on buccal surfaces. The more severe enamel phenotypes of the younger siblings (III:2 and III:5) relative to their mildly affected older brother (III:1) with the same heterozygous ENAM defect suggests sequence variations in other genes might contribute to the severity of the enamel defects.

| Family 2
We performed whole-exome sequence analyses of genomic DNA from two affected and two unaffected members of a nonconsanguineous Caucasian family, referred to us with a diagnosis of autosomal recessive amelogenesis imperfecta. Two different ENAM mutations were identified ( Figure S3). The first was a frameshift mutation in Exon 10 (NM_031889.2: c.1259_1260insAG/p.Pro422Valfs*27) that had previously been reported to cause AI (Chan et al., 2010;Kang et al., 2009;Lindemeyer, Gibson, & Wright, 2010;Ozdemir et al., 2005;Pavlic, Petelin, & Battelino, 2007;Wright et al., 2011). This mutation (rs587776588) causes a frameshift after Gly421 that deletes amino acids 422-1142 and replaces them with 26 extraneous amino acids VPNLALLFAMKKSKIQRRSPWVQKNK*. The defect was observed in one ENAM allele (heterozygous) in the unaffected father (II:9) and in his two affected daughters (III:1 and III:2). The second ENAM defect (NM_031889.2: c.2763delT/p.Asp921Glufs*32) was also a frameshift mutation in Exon 10. This novel frameshift truncated the enamelin protein after Arg920 and replaced amino acids 921-1142 with 31 extraneous amino acids ESSPQQASYQAKETAQRRGKAKTLLEMMCPR*. This frameshift was identified in the unaffected mother (II:8) and in her two affected daughters (III:1 and III:2). Therefore, the unaffected parents were simple heterozygotes for one of the frameshifts, while both of the affected children were complex heterozygotes with frameshifts in both ENAM alleles ( Figure  S3). The novel frameshift mutation (p.Asp921Glufs*32) has not been previously reported to cause amelogenesis imperfecta, but had been submitted to the dbSNP database without any clinical information (rs529979202). These two frameshift mutations in Family 2 are in Exon 10, the most 3′ ENAM exon, making them more likely to escape nonsense mediated decay than premature stop codons in upstream exons (Kurosaki & Maquat, 2016).

| Family 3
In Family 3, four members of a nonconsanguineous Caucasian family were recruited, three of whom were affected ( Figure 2). All affected family members (II:2, III:2, and III:3), but not the unaffected father (II:1), had a F I G U R E 1 Family 1. (a) Pedigree of Family 1. Asterisks mark the five recruited members, three of which were affected. The proband (III:2) is indicated by an arrow. (b) DNA sequence chromatograms confirmed the WES results showing that the proband (III:2) was heterozygous for the ENAM mutation NM_031889.2: c.395dupA/p. (Pro133Alafs*13), as were his affected older brother (III:1) and younger sister (III:5). The DNA sequence chromatogram for the younger brother (III:4) and mother (II:6) matched the reference sequence, and these individuals had normal enamel ( Figure S2). heterozygous ENAM c.588+1delG mutation. The deleted guanine (G) follows a run of six Gs at the end of ENAM Exon 9 (Figure 2d). Functionally, the sixth G (the seventh being deleted) acts as the first nucleotide of the splicing donor site (GT), so splicing occurs normally, but Exon 9 ends up with one less nucleotide (G), causing a frameshift (p.Asn197Ilefs*81) that truncates the enamelin protein after 197 amino acids (which normally has 1,142 amino acids), and adds 80 extraneous amino acids to the C-terminus of the truncated protein (Table S2). Deletion of any one of the seven consecutive Gs results in this frameshift. The ENAM c.588+1delG mutation causes a severe, readily detected ADAI enamel phenotype and is the most frequently reported ENAM defect Kida et al., 2002;Kim et al., 2005;Pavlic, Battelino, Trebusak Podkrajsek, & Ovsenik, 2011;Pavlic et al., 2007;Wright et al., 2011). Although reported to be a mutational "hot spot," the mutation is rare, and is not listed in the 1000 Genomes database (Ensembl GRCh38 of the 1000 Genomes Project data) (1000Genomes Project Consortium et al., 2015 or National Center for Biotechnology Information (NCBI) Single Nucleotide Polymorphism Database (dbSNP) (Kitts & Sherry, 2012), which includes short deletion and insertion polymorphisms (indels/DIPs) as well as single-nucleotide polymorphisms (SNPs). A similar ENAM defect in which a G in the same run of seven Gs was duplicated rather than deleted (rs1085307975; NG_013024.1: c.588+1dupG; p.Asn197Glufs*25) has been identified by clinical testing and submitted to the dbSNP, but the enamel phenotype was not described or published. Although only one ENAM allele is defective, the enamel is markedly reduced in thickness, and on many crowns the enamel appears to be nonexistent. Generalized, severely hypoplastic enamel has been observed in all reported cases of AI caused by heterozygous ENAM c.588+1delG mutations. In previous cases where the primary teeth were available for scanning electron microscopic (SEM) analysis, the enamel was thin and lacked prism organization Pavlic et al., 2007). This same ENAM mutation was identified in Families 4 and 5.

| Family 4
Family 4 was a Taiwanese nuclear family of presumed East Asian descent (Figure 3). The proband was a 25-year-old male presenting with generalized severe enamel hypoplasia but was otherwise healthy. Clinically, there was almost no enamel on all the permanent teeth, except some thin patches F I G U R E 2 Family 3. (a) Clinical photos of the proband (III:2) at age 12 in the mixed dentition stage of dental development and exhibiting severely hypoplastic enamel, interdental spacing, premature attrition, and chipping of the more recently erupted permanent incisors. (b) Pedigree showing autosomal dominant inheritance. (c) Clinical photos of the affected younger sister (III:3) at age 10 also showing severely hypoplastic enamel, interdental spacing, premature attrition, and chipping of the incisors. (d) The three recruited affected members of the family were heterozygous for the ENAM c.588+1delG/(p.Asn197Ilefs*81) mutation in Exon 9/Intron 9 causing a frameshift in the mRNA. The three affected members were heterozygous for this ENAM frameshift while the unaffected father (II:1) lacked the mutation. (e) Panoramic radiograph of the proband (III:2) showing mixed dentition, with little or no enamel evident found at the proximal area of upper incisors and lower anterior teeth, corresponding to the findings on panoramic radiographs. On posterior teeth, there was also a thin bandlike enamel structure around the middle of the tooth crown. Narrowed dental arch forms with a high palatal vault and anterior open bite were also evident. Exome analysis of proband's DNA identified the recurrent ENAM c.588+1delG mutation without any other potential disease-causing mutation in known AI candidate genes. The ENAM mutation was confirmed by Sanger sequencing.

| Family 5 (Digenic Effects?)
In Family 5, the proband at age 14 (III:3) presented with generalized hypoplastic AI, interdental spacing, and an anterior open bite (Figure 4). The maxillary right premolars exhibited localized pitting and brown staining on the buccal surfaces that was not observed on other teeth (Figure 4a). There was evidence of enamel attrition, such as large wear facets and chipped enamel. The incisal edges of the mandibular anteriors and premolars had fractured or chipped, plausibly due to the interplay of malformed enamel and misaligned occlusion (Figure 4a). Nine members of this nonconsanguineous Turkish family were recruited, six of which were affected (Figure 4b). The enamel malformations followed an autosomal dominant pattern of inheritance ( Figure  4b). There was little or no evidence of enamel on the panoramic radiograph. If enamel was present, it did not contrast well with dentin and could not be distinguished from it (Figure 4c). All six affected family members, and none of the three unaffected persons, had a heterozygous ENAM c.588 +1delG/p. Asn197Ilefs*81 mutation as well as a heterozygous LAMA3 c.1559G>A/p.Cys520Tyr mutation. The enamel phenotypes of the unaffected mother (II:3) and affected father (II:4) are provided in Figure S4. Those of the two affected cousins, age 11 and 9 (III:1 and III:2) are shown in Figure S5, and the enamel phenotypes of the unaffected younger sister, age 11 (III:4) and affected younger brother, age 8 (III:5), are displayed in Figure  S6. Despite sharing the same ENAM mutation, Family 3 and Family 4 are of different ethnicity, so the ENAM mutation almost certainly occurred independently in the two families. The LAMA3 missense mutation is rare, is not listed in dbSNP, and is predicted to interfere with protein function (Sorting Intolerant from Tolerant (SIFT) score = 0 (Sim et al., 2012); Polyphen2 score = 1.0 (Adzhubei, Jordan, & Sunyaev, 2013). Unfortunately, these two unlinked genes (ENAM Chr. 4q13.3; LAMA3 Chr. 18q11.2) cosegregated in all of the six recruited affected individuals and were both absent from all three of the recruited, unaffected members of the family, so the phenotypic consequence of the heterozygous LAMA3 defects alone could not be observed. Other heterozygous LAMA3 mutations are reported to cause localized enamel hypoplasia and pitting (Gostynska et al., 2016;Yuen, Pasmooij, Stellingsma, & Jonkman, 2012), but the effects on enamel of this particular heterozygous LAMA3 mutation (NM_198129.2: c.1559G>A/p. Cys520Tyr) have not been reported. In Family 4, the ENAM c.588+1delG mutation caused a severe generalized enamel hypoplasia and localized pitting, so the potential contribution of the LAMA3 defect to the enamel phenotype in this family is unclear. The low frequency of the defect and the prediction software results indicate that the heterozygous LAMA3 c.1559G>A/p.Cys520Tyr mutation would likely cause AI, but such evidence is not decisive by itself (Miosge et al., 2015), so the enamel pathogenicity of this LAMA3 defect remains uncertain.

| Mouse Enam-Ambn digenic effects
To help assess the potential for digenic effects to contribute to the enamel phenotype in ENAM heterozygotes, Enam -/- (Hu et al., 2008) and Ambn -/-(Liang et al., 2019) mice were crossed to generate double heterozygous offspring. The erupted dental phenotypes of 7-week-old wild-type, single heterozygous, and double heterozygous mice were first evaluated by dissection microscopy. The Ambn +/− maxillary incisors and molars both resembled the wild type (Figure 5a). The Ambn +/− mandibular incisors were chalky white and sometimes showed minor chipping (Figure 5b). The Enam +/− molars closely resembled the wild type in overall shape, but the enamel surface was rough textured (Figure 5b). The Enam +/− Ambn +/− molars showed more severe surface roughness, chalkiness, and greater attrition than either of single heterozygous mice. The Enam +/− mandibular incisor enamel was chalky white and consistently chipped. The Enam +/− Ambn +/− double heterozygous maxillary and mandibular incisors were both chalky white and much of the mandibular incisor enamel had been lost to attrition. Even at low magnification under a dissecting microscope digenic effects on surface roughness, chalkiness, and rate of attrition were evident on both molar and incisors.
Mandibular incisors at 7 weeks were cross-sectioned at the level of the labial alveolar crest (prior to this portion of the incisor erupting into the oral cavity) and characterized by bSEM (Figure 6 and Figure S7). The enamel layer looked nearly normal in the Ambn +/− mandibular incisors, with characteristic rows of enamel rods (Smith, Hu, Hu, & Simmer, 2019). The Enam +/− mandibular incisor enamel showed normal rod patterns, but the enamel was thinner and part of the outer enamel layer did not form properly and its surface was rough. Enam +/− Ambn +/− mandibular incisor was thinner still, had a rough, crusty surface, and ectopic concretions in the soft tissue.
To assess the form and surface textures of the single and double heterozygous mouse molars after they had fully formed but before they could be altered in the oral cavity, we removed the overlying soft tissue from nearly erupted D14 mandibular first molars and examined them using bSEM (Figure 7). The overall shape of the Ambn +/− ( Figure S8) and Enam +/− ( Figure S9) D14 mandibular first molar crowns was normal, with minor surface roughness. In contrast, the D14 mandibular first molars showed major surface roughness with surface nodules ( Figure S10). The enamel defects observed in incisors and molars were consistently most severe in the Enam +/− Ambn +/− double heterozygotes relative to the Ambn +/− or Enam +/− single heterozygous mice. Asterisks indicate family members that were recruited for this study, nine in total. The proband is marked by an arrow. (c) Panoramic image of the proband. Very little enamel is evident, either because it is absent or does not contrast with dentin (is hypomineralized). Two potential disease-causing mutations were identified: (d) ENAM: c.588+1delG/ p.Asn197Ilefs*81, which was previously reported to cause ADAI in multiple families Kang et al., 2009;Kida et al., 2002;Pavlic et al., 2007;Wright et al., 2011). (e) LAMA3 missense mutation c.1559G>A/p.Cys520Tyr. All recruited affected family members (II:2, II:4, III:1, III:2, III:3, and III:5) were heterozygous for both of these (ENAM and LAMA3) mutations. All recruited unaffected family members (II:3, II:5, and III:4) had neither mutation. No family members were heterozygous for the ENAM or LAMA3 mutation alone

| DISCUSSION
Enamelin was expressed when enamel first formed on the scales and teeth of fish more than 450 million years ago. Enamelin is specifically expressed by secretory stage ameloblasts and its primary structure has been conserved during evolution (Gasse & Sire, 2015). Analysis of 36 known mammalian Enam sequences identified 77 unchanged and 346 conservative amino acid positions distributed throughout the eight coding exons with conserved intron boundaries, demonstrating significant selection pressure to maintain ENAM structure and function (Al-Hashimi, Sire, & Delgado, 2009). Defects in ENAM also cause inherited enamel defects in dogs. A five base pair deletion in Enam (c.1991_1995delTTTCC/p.
This report increases to 22 the number of reported novel human ENAM mutations that cause AI, which are widely distributed across the coding exons and intron borders (Table S2). ENAM disease-causing mutations include small deletions and insertions that truncate the protein, shift the reading frame, and add potentially toxic C-terminal adducts. There are also splice donor and acceptor site, missense, and stop-gain mutations. The enamel phenotypes caused by ENAM mutations sometimes show incomplete penetrance and variable expressivity. ENAM heterozygotes sometimes display no detectable AI phenotype (lack of F I G U R E 5 Dissecting light microscopy comparison of 7-week Ambn +/+ Enam +/+ , Ambn +/− , Enam +/− , and Ambn +/− Enam +/− molars and incisors. (a) Buccal, occlusal and lingual views of mandibular molars. The Ambn +/− and Enam +/− molars closely resembled the wild type in crown form and level of attrition. The Ambn +/− molars showed minor surface roughness. The Enam +/− molars showed increased surface roughness and a chalky appearance. The Ambn +/− Enam +/− molars showed clear digenic effects on their enamel: more severe surface roughness, chalkiness, and rapid attrition than either single heterozygote. (b) Mesial, buccal, lingual, and lateral views of the mandibular incisors and facial view of the maxillary and mandibular incisors. Maxillary incisor chalkiness and mandibular incisor attrition both follow the pattern of progressively increasing severity going from Ambn +/− to Enam +/− to Ambn +/− Enam +/− mice, indicative of digenic effects penetrance) (Seymen et al., 2014), but might still affect the enamel in subtle ways. A combination of two ENAM amino acid substitutions (p.Ile648Thr and p.Arg763Gln) expressed from the same allele (which did not cause AI) were associated with a two-to threefold increase in risk for caries susceptibility (Chaussain et al., 2014).
Including Families 3, 4, and 5 from this study, the ENAM c.588+1delG mutation has been identified in seven unrelated families (Table S2) with various ethnicities, suggesting that this is a mutation "hot spot." The mutation-prone nature of this spot might be attributed to its mononucleotide repeat sequence of seven Gs (Figure 3c). Tandemly repeated sequences have long been recognized as hot spots for frameshift mutations, presumably caused by "DNA polymerase slippage" (Strauss, 1999). This mechanism involves a transient dissociation of the replication DNA template and a re-annealing within the repeat tract, forming an extrahelical loop comprised of one or more repeat units, which would in turn cause an insertion or deletion. A characteristic of the repeated sequences in humans is the large contribution of runs of As in noncoding regions of the genome, which is more permissive. In contrast, repeat runs in coding regions are restricted by the amino acid code, and a run of glycines (GGG) might not be tolerated (Strauss, 1999). In E. coli, repeats of Gs and Cs were observed to frameshift more frequently than repeats of As in a mismatch repair deficient strain (Sagher, Hsu, & Strauss, 1999). The ENAM c.588+1delG mutation, removing one G out of seven, should allow normal splicing between Exon 9 and Intron 9, but cause a frameshift (p.As-n197Ilefs*81) that introduces a premature termination codon in the last exon (Exon 10). This mutant transcript would presumably escape nonsense mediated decay and produce F I G U R E 6 Level 8 cross-sections from 9-week Ambn +/− , Enam +/− , and Ambn +/− Enam +/− mouse mandibular incisors imaged by backscattered scanning electron microscopy (bSEM). The Ambn +/− incisor is similar to the wild type. The Enam +/− incisors exhibit a somewhat thinner enamel with normal-looking enamel rods. The enamel is rough, with some ectopic mineral nodules on its surface. The Ambn +/− Enam +/− incisors shows a more severe enamel phenotype, with thin enamel and large ectopic nodules that appeared to form as concretions. Severe enamel malformations were consistently found in the Ambn +/− Enam +/− mouse ( Figure S7) F I G U R E 7 bSEM images of D14 mouse molars from Ambn +/− , Enam -/-, and Ambn +/− Enam +/− mice. These images were taken after removing the soft tissue covering these erupting molars. The Ambn +/− ( Figure S8) and Enam -/-( Figure S9) molars resembled the wild type except for showing localized areas of surface roughness. The Ambn +/− Enam +/− mouse molars showed extensive areas of surface roughness and regions covered by a mineral crust. A crusty, rough surface was a consistent finding in Enam +/− Ambn +/− mice ( Figure S10) a mutant ENAM protein. This mutant protein, carrying 80 aberrant amino acids at its C-terminus, might interact with the wild-type ENAM protein or other EMPs, causing a dominant negative effect. Alternatively, the mutant protein might be toxic to ameloblasts and cause cell pathology, which leads to a more severe enamel defect when compared with phenotypes caused by simple loss-of-function mutations in a single ENAM allele. The ENAM c.588+1delG enamel defect is readily recognizable clinically and hard to overlook, which might also contribute to the frequent identification of this mutation.
ENAM mutations show a dose effect: when only one allele is defective the phenotype is nonpenetrant, or milder than when both alleles are defective. The enamelin dose effect is largely due to the fact that detectable enamel malformations are often observed when only one ENAM allele is defective (dominant pattern of inheritance). Most genes that are required for amelogenesis are expressed at sufficiently high levels that a loss of one allele (haploinsufficiency) does not cause a detectable enamel phenotype and produces a recessive pattern of inheritance. Over-expression as well as under-expression of early secreted enamel proteins interferes with enamel ribbon deposition. Expression of Ambn (Teepe et al., 2014) or Mmp20 (Shin, Chavez, Ikeda, Foster, & Bartlett, 2018) transgenes in wild-type mice already expressing these genes at normal levels causes enamel malformations. It is plausible then that natural selection limits ENAM expression levels to less than twice the amount required to avoid over-expression defects, resulting in an autosomal dominant pattern of inheritance with a dose effect. In any case, amelogenesis is a sensitive process that can be readily disturbed. Here, we establish that digenic effects can cause AI in mice.
The Online Mendelian Inheritance in Man (OMIM) lists 91 conditions having an enamel phenotype, with 71 having a known genetic etiology (Wright et al., 2015). About half of AI conditions show a recessive pattern of inheritance (Wright et al., 2011). A genetic search for the etiology of AI in a given family has a success rate of about a 60% (Wright et al., 2011). A disproportionate share of the unsolved cases is recessive. There are multiple reasons why cases go unsolved, but the most likely is that there are still many genes that cause AI that have not yet been identified. Another reason is the inherent difficulty in determining the etiology when more than one gene is involved. Inherited enamel malformations are highly heterogeneous and some are likely to be caused by variations in gene regulatory sequences and digenic inheritance.