The GenoDENT panel has been evolving through times from its first published version (Prasad et al., 2016a). The last updated version 6.0 explores 567 genes (Supplementary Table S1).

The GenoDENT panel interrogates two categories of genes: a diagnostic panel (248 genes known as responsible for rare diseases with orodental expression in human) and a discovery panel (319 candidate genes reported as being involved in tooth development or orodental anomalies in animal models for example).

Probe design was performed on the Agilent SureDesign portal (https://erray.chem.agilent.com/suredesign, Agilent, United States) in order to capture, by complementarity, the exonic sequence as well as 25 bases of their flanking intronic sequences. Libraries were prepared with the Agilent SureSelect QXT protocol and sequenced on a NextSeq 550 (Illumina, San Diego, United States). GenoDENT is implemented in a diagnostic setting and its results are directly available for the individual medical file and genetic counselling. Variants are classified according to the American College of Medical Genetics (ACMG) classification (Richards et al., 2015; Harrison et al., 2019). Upon identification of variants of class 4 (probably pathogenic) or 5 (pathogenic), extended familial segregation is performed via Sanger sequencing; a detailed report is written and sent to the geneticist. A variant of uncertain significance (VUS) or class 3 should not be used in clinical decision-making. Class 2 are likely benign polymorphisms.

Trio whole Exome Sequencing (WES) was performed on trio for individuals 7.10, 7.11, 9.1, 9.2, 9.3, 17.1, 17.2, 17.3, 17.4, 18.3, 18.8, 18.10, and 24.2 by Integragen (Evry, France, 2014). Exons of DNA samples were captured using in-solution enrichment methodology (SureSelect Human All Exon Kits, Agilent, Massy, France) with the company’s biotinylated oligonucleotide probe library (Agilent Human All Exon v5+UTR 75 Mb Kit) and sequenced with an Illumina HISEQ 2000 (Illumina, San Diego, United States) as paired-end 75 bp reads, resulting in an average coverage of 80X.

STARK (Stellar Tools from raw sequencing data Analysis to variant RanKing) is a bioinformatics pipeline based on the GATK recommendations used to process the NGS data (DePristo et al., 2011). Annotation and ranking of SNV/indel were performed by VaRank (Geoffroy et al., 2015) in combination with the Alamut Batch software (Interactive Biosoftware, Rouen, France). Variant effect on the nearest splice site was predicted using MaxEntScan (Yeo and Burge, 2004), NNSplice (Reese et al., 1997) and Splice Site Finder (Shapiro and Senapathy, 1987).Very stringent criteria were applied to filter out non-pathogenic variants: 1) variants represented with an allele frequency of more than 1% in public variation databases including the 1,000 Genomes (The 1000 Genomes Project Consortium et al., 2015), the GnomAD database (Exome Aggregation Consortium et al., 2016) or our internal exome database, 2) variants in 5′ or 3′ UTR, 3) variants with intronic locations and no prediction of local splice effect, and 4) synonymous variants without pathogenic prediction of local splice effect. Annotations of structural variations (SV) were performed by AnnotSV (Geoffroy et al., 2018).

Primers’ design was done using the Amplifix v1.5.4 software. Primers were then ordered from Eurofin MWG (Supplementary Table S2). The amplicons’ sizes were checked by electrophoresis on the Caliper LabChip GX (Life science). After enzymatic purification with the Illustra™ Exoprostar™ kit (Sigma Aldrich) to remove dNTPs and salts, the PCR product was used to perform a sequence reaction with the BigDye™ Terminator v1.1 Cycle Sequencing Kit (Applied Biosystems, Thermofisher Scientific). The BigDye Xterminator™ Purification Kit (Applied Biosystems by Thermofisher Scientific) was used to purified product. This purified product was then loaded on the 3,500 Series Genetic Analyzers (Applied Biosystems, Thermofisher Scientific) sequencer. Final sequence data were analyzed with SEQUENCE Pilot (JSI medical systems).

In our current study, we found six individuals (1.1–1.6) presenting with, hypoplastic pitted AI. The enamel appears pitted, rough, of normal hardness and presents yellow-brown extrinsic stains. Radiographs show normal enamel opacity (Figure 4; Supplementary Figure S1A).

We identified heterozygous loss of function variants in COL17A1 gene in each of them (Supplementary Table S3; Supplementary Figure S2A). In individuals 1.2, 1.5, and 1.6 the heterozygous variants we identified were originally described as pathogenic in EB individuals. The heterozygous variant found in individual 1.2 was also found in her affected sister and daughters (Supplementary Figure S3.1). The heterozygous variant found in individual 1.6 was also detected in his mother but the parents’ phenotype was not known (Supplementary Figure S3.2). Individuals 1.1, 1.3 and 1.4 and their variants were already described in an autosomal dominant mode by our team in (Prasad et al., 2016b). Heterozygous variants in COL17A1 gene, historically discovered in EB families, were reported in the literature as responsible for enamel defects and an AI phenotype in heterozygous carriers, including parents non-affected with EB (McGrath et al., 1996; Prasad et al., 2016a; b).

We report three individuals (2.1, 2.2 and 2.3) presenting with an hypoplastic pitted AI phenotype with thin enamel and yellow discoloration (Figure 4; Supplementary Figure S1B). Panoramic X-rays show thinner enamel. We found in these 3 individuals COL7A1 heterozygous variants not previously described (Supplementary Table S3; Supplementary Figure S1B, S2B).

Individual 2.1 has an intronic heterozygous variant (NM_000094.4:c.2440 + 3A>C) in intron 19. This variant is predicted to have an impact on the splicing site possibly leading to an, in phase, exon 18 skipping. Exon 18 codes for a Fibronectin type III domain involved in interactions with integrins. A variant implicating a similar splice site defect has been described in an individual presenting epidermolysis bullosa with enamel defects (c.2440 + 1G>T) (Vahidnezhad et al., 2017).

Individual 2.2 and 2.3 both carry a heterozygous missense variant NM_000094.4:c.3605G>A; p.(Arg1202His) and NM_000094.4:c.3785T>C; p.(Met1262Thr) located respectively in the Von Willebrand factor type A and Collagen triple helix repeat protein domains. No other missense variant was described in the Von Willebrand domain, but other missense variants have been described as pathogenic in the Collagen triple helix domain (Yenamandra et al., 2018). Individual 2.2 has an affected sibling carrying the same variant. Individual 2.3 inherited his variant from his affected mother (Supplementary Figures S3.3, S3.4).

All these variants were, so far, classified as VUS and further investigations would be needed to determine their impacts on the phenotype.

Variants in this gene were only previously reported in individuals with epidermolysis bullosa in autosomal dominant or recessive conditions.

We report one individual (3.1) presenting severe hypoplastic AI carrying a mutation in LAMA3. A digenic inheritance with variants in both in COL17A1, and LAMA3 has been previously described (Prasad et al., 2016a) (Supplementary Table S3; Supplementary Figure S2C). The heterozygous variant, transmitted by her affected mother, in COL17A1 NM_000494.4:c.1141 + 1G>A is a pathogenic variant altering the splicing site in exon 14. The individual also carries an heterozygous variant in LAMA3 not inherited from her mother NM_000227.6:c.1650_1659del; p.(Ile550Metfs*46). We hypothesize this additional mutation could explain the phenotype severity gradient between 3.1 and her mother. Indeed, both are presenting hypoplastic AI but in 3.1 the phenotype is more severe, and the pits are numerous and clearly visible (Figure 4; Supplementary Figure S1C).

Each of the individuals reported in this publication (4.1, 4.2, 4.3, 4.4 and 4.5) carrying mutations in LAMB3 present with hypoplastic AI with an irregular pitted and thinner enamel and no sign of epidermolysis bullosa (Figure 4; Supplementary Figure S1C).

Among the pathogenic variants reported in this gene, bi-allelic loss of function variants was described in patients with severe EB and AI. Dominant heterozygous frameshift variations were reported in patients with isolated AI: all those variations were located in the last two exons of LAMB3 gene. For these published patients, a dominant negative effect was proposed by Smith (Smith et al., 2019). In our cohort, two patients have such 3′end variations: patients 4.4 and 4.5, NM_000228.3: c.2926del; p.(Val976Trpfs*54) and c.3305del; p.(Gly1102Valfs*7) (Supplementary Figure S3.5; Supplementary Figure S2D).

We also report in our cohort, three patients with isolated AI and a non-sense mutation located before the 3′end of the gene (4.1, 4.2 and 4.3) 4.1 and 4.3 were reported in (Prasad et al., 2016a) and reanalyzed in the context of this publication. Revisiting these data showed that 4.1 also presented a non-sense mutation in position 42 (c.124C>T; p.(Arg42*)); we also detected a new missense variant c.3490C>T; p.(Arg1164Cys) which is predicted to be deleterious by SIFT (v6.2.0) and Polyphen-2. Familial segregation showed that those two variations affected both alleles in this patient (Supplementary Figure S3.6).

Patient 4.2 has one premature stop codon in position 635 and on the second allele, a splice variation was detected, c.1288 + 1G>T (Supplementary Figure S3.7). This variation was previously reported by Kiritsi et al. (2015) and the authors confirmed the impact of the splice variant by mRNA study: the variation induced an in-frame skipping of exon 11 and was predicted to produce an incomplete protein p.(Ser378_Arg430delinsCys). According to (Kiritsi et al., 2015), the patient presented with AI and a mild form of EB with favorable evolution.

The clinical synopsis of LAMB3-related AR epidermolysis bullosa, junctional 1A intermediate includes enamel hypoplasia, enamel pitting and corneal erosion, corneal scarring besides skin, hair and nails defects. Notably, individual 4.2 has a history of recurrent corneal ulcers and might therefore present a mild form of EB.

Patient 4.3 has a stop mutation in position 635. A splice variation was also detected in this patient c.944-14C>G but mRNA analysis could not be performed and familial segregation was not possible.

The first two cases (4.1 and 4.2) could thus be compatible with a recessive form of AI: those patients combined one null allele and a possible hypomorphic second allele. An AR clinical continuum may go from severe EB/AI in patients with biallelic null variations to mild EB/AI or isolated AI in patients with one null mutation occurring with a hypomorphic allele.

A 4-year-old individual (5.1) displayed a hypoplastic/hypomature AI phenotype. The primary dentition showed thin white opaque enamel (Supplementary Figure S1C).

We found a heterozygous LAMC2 variant NM_005562.3: c.493C>T; p.(Arg165Cys) with an allele frequency of 0.2% in GnomAD, predicted deleterious by SIFT (v4.0.3) and PolyPhen-2 and located in the Laminin EGF domain (Supplementary Table S3; Supplementary Figure S2E). Heterozygous variants in this gene have not previously been associated with AI in human but this gene is known to be involved in enamel formation defects in mice (Wazen et al., 2016) and the patient’s phenotype is similar to the one observed in mice. The allele is inherited from her mother but her phenotype was not available (Supplementary Figure S3.8).

Individuals with AR variants in integrin-β6 (ITGB6), have been described as affected with hypoplastic pitted and hypomineralized AI. Enamel is less dense, presents disorganized prisms, and severe pitting in the coronal side of tooth with pigmentations (Poulter et al., 2014a; Wang S.-K. et al., 2014; Seymen et al., 2015a). Though this established phenotype infers a Type 1A classification, no ITGB6 pathogenic variants were found in our cohort.

ENAM encodes enamelin, the largest enamel matrix protein, composed of 1,142 amino acid. It is a secretory protein with a 39 amino acid signal peptide. During amelogenesis, the protein is found among the developing crystallites in the enamel rods and interrods (Daubert et al., 2016). Its presence is necessary for correct prism morphogenesis, contributing to thicker enamel (Shore et al., 2010). Variants in ENAM cause hypoplastic AI (Mårdh et al., 2002; Kim et al., 2005a), in which defective enamel is thin or absent, with horizontal row of pits, linear depressions, or one large hypoplastic area. These defects appear most prominent on the buccal surfaces of the teeth involving the middle third of the enamel.

Non-sense variants disrupting the enamelin domain, splice variants, small deletions and insertions and one big insertion have been described to be causative of Amelogenesis imperfecta (Hu and Yamakoshi, 2003).

We identified in our cohort eight individuals with hypoplastic AI manifest as thin, rough enamel in permanent teeth, columns of pits, and linear depressions in enamel. Hypoplastic enamel is clearly identified in panoramic X-Rays especially visible on unerupted teeth (Figure 4; Supplementary Figure S1D).

We detected ENAM heterozygous variants in each of these individuals (Supplementary Table S3; Supplementary Figure S2F). Three new, never reported, variants are described in this paper. Individual 6.1 presented a missense heterozygous variant in exon 3 NM_031889.3:c.101T>C; p.(Leu34Pro) (Supplementary Table S3). Missense variants in the N-terminal domain have previously been described, reportedly negatively impacting ameloblast secretory pathway leading to endothelial reticulum (ER) stress and an activated unfolded protein response (Brookes et al., 2017). With this proposed mechanism, a functional analysis could improve identification of the VUS towards pathogenic.

For individual 6.3, we detected a heterozygous 1bp duplication on a splice site NM_031889.3:c.588 + 1 dup; p.(Asn197Glufs*25). This variant was also detected in her affected sister and father (Supplementary Figure S3.9). The variant occurred at the same position than the already described deletion c.588+1delG; p.(Asn197Ilefs*81) found in individuals 6.4, 6.5, 6.6 (Supplementary Figure S3.10) and 6.7 (Supplementary Figure S3.11). Interestingly in 6.3 the phenotype seemed slightly different to the one observed in the four individuals carrying the deletion with an hypoplastic but also hypomature whitish enamel. Individual 6.3 also presented agenesis of 16, 26, 36, 46 but no further variant explaining the missing teeth was identified.

Individual 6.8, his affected sister, and their mother were all heterozygous for a non-sense mutation in exon 9 NM_031889.3:c.664C>T; p.(Gln222*) (Supplementary Figure S3.12).

Truncating variants earlier and later in the protein were already described in this gene and found to be causative of hypoplastic amelogenesis imperfecta (Ozdemir et al., 2005a; Seymen et al., 2014a).

The phenotype is more severe than in local hypoplastic autosomal dominant AI and has also been linked to ENAM variants but in a recessive mode of inheritance (2 alleles affected). This leads to the conclusion that ENAM variants associated phenotype is dosage dependent. Indeed, Ozdemir et al., 2005a, Hart et al., 2003 described families for which carriers of only one ENAM variant were less affected presenting only localized enamel pitting, whereas family members with compound heterozygous or homozygous ENAM variants presented severe hypoplastic AI. Lindemeyer et al., 2010 also described the case of a nine-year-old boy with homozygous variants in ENAM and severe hypoplastic AI and smooth teeth.

We did not detect any cases AR mode of inheritance associated with ENAM in our cohort.

SP6 encodes a transcription factor expressed during amelogenesis and involved in regulating proliferation and differentiation of ameloblasts (Nakamura et al., 2004; 2008; Utami et al., 2011; Muto et al., 2012; Ruspita et al., 2020). Autosomal dominant pathogenic variants have been published in two unrelated families in association with severe hypoplastic AI (Smith et al., 2020; Kim et al., 2021b). Interestingly in the two families, the same amino acid (Ala273) was modified. No variant in this gene was evident in our cohort.

AMELX, Amelogenin X-linked, has been identified as the gene involved in this AI. The gene resides in intron 1 of ARHGAP6 in the opposite orientation (Iwase et al., 2007). Amelogenins are highly conserved proteins secreted by ameloblasts that constitute 90% of the enamel organic matrix. As the proteins are digested and removed in maturation phase, mineral crystals grow in well-organized prism patterns (Gibson et al., 2001).

Genotype-phenotype correlation revealed that variants in the N-terminal (Lagerström et al., 1991; Kim et al., 2004) and C-terminal parts (Greene et al., 2002) of AMELX cause hypoplastic AI.

In our cohort, we identified five females (7.1, 7.3, 7.4, 7.8 and 7.11) presenting the typical hypoplastic AI banding pattern (Figure 4; Supplementary Figure S1E).

We identified heterozygous variation in AMELX in each of these individuals (Supplementary Table S3; Supplementary Figure S2G). Female individuals 7.1 to 7.3 were carrying a non-sense AMELX variant NM_182680.1:c.11G>A; p.(Trp4*) firstly reported by (Hart et al., 2002b). Individual 7.4 carried a newly reported variant NM_182680.1(AMELX):c.47C>A; p.(Ala16Asp) predicted damaging by SIFT (v6.2.0) and Polyphen-2 and inherited by her affected father (Supplementary Figure S3.13). Individuals 7.5 to 7.7 were presenting already reported missense variant. Individuals 7.8 to 7.11 were carrying frameshift variant originally reported by (Lench and Winter, 1995).

Male individual 7.9 is the cousin of 7.8 female and carried the same variant (Supplementary Figure S3.14). He presented the characteristic female lyonisation banding pattern thus requiring further understanding and clarification. His karyotype revealed XXY aneuploidy, the most common disorder of sex chromosomes in humans, with a prevalence of one in 500 males. This finding explained the lyonisation pattern in a male and was consistent with his associated behavioral disorder.

Individual 7.11 is the mother of 7.10 male. The mother presented the typical female hypoplastic banding pattern. The boy showed a severe hypoplastic AI with almost no enamel associated with delayed dental eruption and behavioral issues. As the boy’s phenotype presented as syndromic, this family was directly sequenced by whole exome sequencing. We identified a pathogenic 1bp deletion leading to a premature stop codon c.541del; p.(Leu181Cysfs*8), a mutation previously reported (Kindelan et al., 2000) in both the mother and the boy (Supplementary Figure S3.15). This variation clearly explained the AI phenotype but no other variation was discovered as possibly explaining behavioral issues presented by this boy.

Male individuals 7.5, 7.6 and 7.7 presented a different hypomature amelogenesis imperfecta phenotype with a smooth yellow opaque enamel (Supplementary Figure S1E). Individuals 7.6 and 7.7 carry a c.208C>A; p.(Pro70Thr) variant first described by Collier et al., 1997 (Supplementary Figures S3.16, S3.17). Functional analyses showed that this variant is reducing the interactions between amelogenin and the MMP20, metalloproteinase degrading amelogenin in the maturation phase (Tanimoto et al., 2008). Therefore, depending on the patho-physiological mechanism, AMELX might also be the gene involved in the hypomature form of AI type IIC (snow-capped teeth X-linked) or eventually IIB (hypomaturation X-linked). This would be discussed further in these mentioned categories.

Individual 7.7 also had a frequently reported heterozygous variant in WNT10 gene (c.682T>A; p.(Phe228Ile), explaining the 12, 15, 22, 28 agenesis.

AMBN gene (4q13.3), containing 13 exons was discovered as the causative gene for non-syndromic autosomal recessive amelogenesis imperfecta (Toyosawa et al., 2000). It encodes ameloblastin (AMBN, also named “amelin” or “sheathlin”), the second most abundant enamel matrix protein, critical to amelogenesis (Lu et al., 2018).

Four cases have been reported so far in the literature with AMBN variants associated with non-syndromic AI (Poulter et al., 2014b; Lu et al., 2018; Liang et al., 2019).

In our cohort we identified three individuals with a hypoplastic amelogenesis imperfecta combined with recessive variants in AMBN (Supplementary Table S3; Figure 4; Supplementary Figures S1F, S2H). Individuals 8.1, and 8.2 presented compounds heterozygous variants never described before. 8.1 displayed a splice variant in intron 1 NM_016519.6:c.15 + 1G>A and a non-sense variant in exon 8 c.577G>T; p.(Gly193*) (Supplementary Figure S3.18). 8.2 was carrying a non-sense variant in exon 5 NM_016519.6:c.209C>G; p.(Ser70*) and a splice variant in intron 7 c.571-1G>C. The non-sense mutation hadn’t previously been described, however Poulter et al. (2014b) described a large deletion resulting in the loss of exon 6. Individual 8.3 was already described in (Prasad et al., 2016a). We detected a homozygous splice variant in intron 6 never described before NM_016519.6:c.532-1G>C. This is the first splicing variant described in this gene.

This phenotype is associated with recessive homozygous or compound heterozygous variants in ACP4 gene encoding testicular acid phosphatase (Seymen et al., 2016; Smith C. E. et al., 2017). All variants reported so far were localized in the histidine phosphatase domain. This domain contains conserved Histidine residues that are phosphorylated during the catalytic activity of the protein and other residues that are forming a phosphate pocket and interact with the phospho group of substrates before, during and after its transfer to the Histidine residue. Variants were reported mostly in exon 4 and 7 and only one was reported in exon 3 (Seymen et al., 2016; Smith C. E. et al., 2017).

We found three unrelated families (individuals 9.1, 9.2 and 9.3) with hypoplastic amelogenesis imperfecta, with thin enamel with irregular surface, and a yellowish color. X-rays showed a thinner enamel of normal density, long teeth, and slim roots (Figure 4; Supplementary Figure S1G).

These families were previously negative on the GenoDENT panel first version (Prasad et al., 2016a), and were further investigated by whole exome sequencing analysis. We found compound heterozygous variants in the ACP4 gene (Supplementary Table S3; Supplementary Figure S2I). Indeed ACP4, discovered in 2016 (Seymen et al., 2016), was not on the first version of GenoDENT but was added at version 0.4. Individual 9.1 had a previously reported missense variant in exon 4 NM_033068.3:c.331C>T; p.(Arg111Cys) and a splice variant in intron 6 c.645 + 1G>A; p.? which was predicted to lead to an in frame exon 6 skipping. This is the first case reported incidence of a splice variant of ACP4. Individual 9.2 had an already described missense variant in exon 4 NM_033068.3:c.428C>T; p.(Thr143Met) and a second variant not yet reported in exon 7, in the histidine phosphatase domain c.736G>A, p.(Val246Met). This variant is predicted damaging by SIFT (v6.2.0) and Polyphen-2. Individual 9.3 had 2 new missense variants, the first one is in exon 6, NM_033068.3:c.626T>C; p.(Leu209Pro) is in the domain but the second one is in the exon 11, in cytoplasmic domain c.1199C>A; p.(Ala400Asp). No variants were reported in the cytoplasmic part of the protein. Our results are generally consistent with previously described variants, but also suggest that the cytoplasmic region of the protein, as well as the extracellular region could have a role in amelogenesis.

Individual 9.3 also had agenesis of teeth 18, and 28, consistent with one of the most frequent variant found in WNT10A gene NM_025216.3:c.682T>A; p.(Phe228Ile).

KLK4encodes an enamel matrix protease requiring cleavage for its activation after secretion by ameloblasts. KLK4 encodes a serine protease that, once activated during maturation stage, will degrade enamel proteins to allow the growth in width and thickness of enamel crystallites (Simmer et al., 2009).

Truncating variants in this gene have been previously described (Wang et al., 2013; Seymen et al., 2015b; Smith C. E. L. et al., 2017). In our cohort, individual 10.1 displayed pigmented hypomature, hypomineralized enamel. As a compound heterozygous, he is carrying a previously reported non-sense variant (Hart et al., 2004) associated with a new missense mutation in KLK4 NM_004917.4: c.443G>T; p.(Cys148Phe) and c.458G>A; p.(Trp153*) (Supplementary Table S3; Supplementary Figure S2J and Supplementary Figure S3.19). The missense variant is affecting a highly conserved amino-acid and is predicted to be deleterious by SIFT and Polyphen-2. This is the first report of the putative pathogenic missense variants c.443G>T; p.(Cys148Phe) in this gene.

MMP20 encodes a zinc-dependent endopeptidase activated during the secretion stage and continuing to be expressed by ameloblasts during the maturation stage. It supports enamel maturation by degrading the enamel protein matrix. As KLK4 and MMP20 seem to work in a collaborative manner, it is not surprising to note a similar dental phenotype in individuals with KLK4 (Hart et al., 2004) or MMP20 (Kim et al., 2005b) variants. We found 16 individuals (Supplementary Table S3; Supplementary Figure S2K) presenting with hypomature AI both in their primary and permanent dentitions; their phenotypes are similar to the ones described in the literature and associated with MMP20 recessive inheritance (Kim et al., 2005b; Ozdemir et al., 2005b; Papagerakis et al., 2008; Lee et al., 2010; Gasse et al., 2013; 2017; Kim et al., 2017; 2020; Wang et al., 2020; Nikolopoulos et al., 2021).

Individuals 11.1, 11.2, 11.3 and 11.4 have compound heterozygous variants in MMP20 gene (Supplementary Figures S3.20, 3.21, 3.22). All have in common the first variant NM_004771.4: c.103A>C; p.(Arg35 = ). This pathogenic variant was firstly reported in Prasad et al., 2016a in individual V2.13. Its splicing impact was then functionally demonstrated by Kim et al., 2020. For individual 11.4 the second variant c.1362C>G; p.(Tyr454*) is novel and has been discovered thanks to GenoDENT panel. Other variants truncating the protein were already described in this gene (Papagerakis et al., 2008; Kim et al., 2017) but never so late in the protein. Indeed c.1362C>G; p.(Tyr454*) occurred in the last exon, inducing the truncation of the hemopoxin domain involved in binding inhibitors of metallopeptidases. It is also plausible that the domain may restrict cleavage site selection (Lee et al., 2010). We have tentatively classified this variant as probably damaging. We also report another new heterozygous pathogenic mutation found in individual 11.8 c.359dup; p.(Asn120Lysfs*9), along with a second variant on the other allele (Supplementary Figure S3.23) already reported c.954-2A>T (Kim et al., 2005b). This new variant was classified as probably damaging in light of later truncations being so described (Kim et al., 2017). The GenoDENT panel highlighted a compound heterozygous VUS in individual 11.11 c.530G>A; p.(Gly177Glu) associated to a known pathogenic variant (c.954-2A>T) (Kim et al., 2005b) (Supplementary Figure S3.24). This missense VUS creates an amino-acid substitution of the conserved Gly177 in the Matrixin protein domain, involved in the peptidase activity, and is likely predicted deleterious by SIFT (v4.0.3), VariantTaster (v2021) and PolyPhen-2. This VUS was not previously reported in the GnomAD database.

Individuals 11.12 and 11.13 both have a heterozygous variant NM_004771.4:c.566T>C; p.(Leu189Pro) already reported by the team (Gasse et al., 2017) along with respectively c.910G>A; p.(Ala304Thr) described in (Lee et al., 2010) and a novel variant in individual 11.3 reported c.1126C>T; p.(Gln376*). Due to the lack of parental DNA, we could not confirm the relative position of the variant in individual 11.13. It would however lead to the loss of almost all the hemopoxin domain, so, we classified this variant as probably damaging. Individuals 11.14 to 11.17 are all homozygous for the pathogenic variant c.954-2A>T (Supplementary Figure S3.25).

WDR72 is a transporter, it encodes a protein predicted to be an intracellular vesicle coat protein that is mostly expressed during maturation stage but also at secretory stage. The stronger expression has nevertheless been detected during maturation initiation (El-Sayed et al., 2009; Katsura et al., 2014). WDR72 has recently been confirmed as regulating vesicle trafficking in ameloblasts (Katsura et al., 2022) and being involved in distal renal tubular acidosis (Khandelwal et al., 2021).

We identified four individuals with hypomature AI (Figure 4).

New variants in WDR72 were identified by GenoDENT (Supplementary Table S3; Supplementary Figure S2L) in individuals 12.1, 12.2 and 12.4. Individual 12.1 presented a homozygous deletion including a part of intron 1 and the beginning of exon 2 NM_182758.4:c.-13 + 989_7del. In individual 12.2, we detected, a non-sense mutation in exon 2 and a 10 bp deletion at the intron 9 splice site. Truncating variation are known to be pathogenic but no splice variants were previously described in this gene. Intriguingly, the 12.2 phenotype is more complex than in other reported individuals as this patient also presented with small teeth, tooth agenesis, mild deafness and nephrocalcinosis. The recent description of the role of WDR72 in kidney-associated diseases and the syndromic nature of the diseases presented by 12.2 would demand possible future consideration of WDR72 in a class of syndromic AI. Individual 12.3 is compound heterozygous for two known pathogenic variants reported in (Prasad et al., 2016a). In individual 12.4 we detected two novel compound heterozygous 1 bp deletion in exon 15 c.2388del; p.(Lys796Asnfs*16) and c.2146del; p.(Ala716Profs*10), resulting in a frameshift and a premature stop codon a known mechanism in this disease (Supplementary Figure S3.26).

C4ORF26, named also ODAPH, odontogenesis associated phosphoprotein, function is not yet well defined. ODAPH encodes a proline rich protein expressed during the secretory and maturation stage, suggesting a possible role in hydroxyapatite crystallization (Parry et al., 2012). ODAPH is important to maintain the integrity of the atypical basal lamina at maturation stage (Ji et al., 2021). Our cohort included individual 13.1 with hypomature AI (Figure 4). This individual was previously described in (Prasad et al., 2016b). He carries a small homozygous deletion NM_178497.5:c.39_46del; p.(Cys14Glyfs*18) identified using the GenoDENT panel (Supplementary Table S3; Supplementary Figure S2M). The phenotype was consistent with previous individuals described with truncating variants in this gene (Parry et al., 2012).

SLC24A4 is a potassium-dependent sodium/calcium exchanger expressed by maturation stage ameloblasts (Hu P. et al., 2012). It likely performs a role in calcium provision to the enamel during maturation (Wang S. et al., 2014). One large deletion encompassing intron 14 to exon 17 has been described (Seymen et al., 2014b) whereas Prasad et al., 2016a described a homozygous deletion including the last three exons (15,16 and 17) and Parry et al., 2013 a non-sense homozygous variant.

Two individuals (14.1 and 14.2) with hypomature AI (Figure 4) showed novel homozygous SLC24A4 variants using our GenoDENT panel (Supplementary Table S3; Supplementary Figures S1I, S2N). Individual 14.1 was already described in (Prasad et al., 2016a) with a homozygous deletion including the 3 last exon (15,16 and 17) of the SLC24A4 gene NM_153646.4:c.(1,537 + 1_1538-1)_*67. Individual 14.2 and her affected sister carried a homozygous variant affecting a splice site with strong exon skipping predictions c.1716 + 5G>A (Supplementary Figure S3.27). No splice variant has previously been reported so we have classified this variant as a VUS pending functional analysis.

GPR68 is a proton-sensing protein present during all stages of amelogenesis. It has been suggested that the protein acts as a pH sensor directing ameloblasts to switch between the ruffle ended and smooth ended conformations during the maturation stage (Parry et al., 2016b).

pH-Sensing G-protein-Coupled Receptor (GPR68) has been showed to be implied in ameloblast (Ludwig et al., 2003; 2003; Tomura et al., 2008; Frick et al., 2009) and odontoblast (Yang et al., 2006; Pereverzev et al., 2008) function. First human pathogenic variants were reported in (Parry et al., 2016b) associated with hypomaturation AI in three families (#617217). The clinical phenotype showed an enamel of apparent normal thickness but poorly mineralized, with brown discoloration. Weaker enamel tended to be prematurely lost secondary to attrition or masticatory stress, especially in the posterior teeth. Only few other cases, were additionally published, associated with the same type of AI (Seymen et al., 2021; Spedicati et al., 2021). No variants in this gene were found in our cohort.

Complete deletion of AMELX has been associated to this phenotype (Hu J. C.-C. et al., 2012).

We did not find this phenotype among our cohort.

Family with sequence similarity 83, member H (FAM83H), is an intracellular protein with ubiquitous expression (Lee et al., 2011). It reaches maximum expression in ameloblasts during the secretory stages. FAM83H regulates the organization of the keratin cytoskeleton and is involved in desmosome formation (Kuga et al., 2016). Variants identified in FAM83H cause autosomal dominant hypocalcified AI (Mendoza et al., 2007; Kim et al., 2008).

We report 15 individuals (15.1-15.15) presenting with hypocalcified amelogenesis imperfecta (Figure 4), affecting both primary and permanent dentitions with secondary loss of enamel, and colored teeth. Occlusal and incisal wear gave a conical form to the canines. There was no difference in radio-opacity between enamel and dentin.

The 15 individuals carried autosomal dominant variants in exon 5 (Supplementary Table S3; Supplementary Figures S1J, S2O, S3.28–3.37). This is consistent with previously reported variants as each of the variants identified to date have mapped to this largest and final exon of FAM83H. All of them except the one encountered in 15.13 are frameshift or non-sense variants, a mechanism already described as disease causing. Only three of those individuals had a novel variant: 15.1 NM_198488.5:c.930_939dup; p.(Val314Argfs*14), 15.6 c.1309_1311delinsTAG; p.(His437*) and 15.9 c.1375C>T; p.(Gln459*). Individual 15.4 c.1282C>T; p.(Gln428*) carries a variant previously described in (Prasad et al., 2016a). Individual 15.13 is the only one presenting a missense variant NM_198488.5:c.1498C>G; p.(Leu500Val). He also displays a variant in WDR72 NM_182758.4:c.1283T>G; p.(Ile428Ser) (Supplementary Figure S3.36). The phenotype in this individual appears more severe possibly due to the compound effect of the two variants or to his more advanced age and accentuated tooth wear.

Amelotin, encoded by AMTN, is a proline, leucine, threonine and glutamine rich protein binding to ODAM (odontogenic, ameloblast associated) and SCPPPQ1 (secretory calcium-binding phosphoproteins proline-glutamine rich 1) to form aggregates able to maintain the attachment between ameloblasts and the produced enamel during maturation stage (Holcroft and Ganss, 2011; Fouillen et al., 2017).

A large deletion in AMTN spanning exon 3–6 has been reported yet to cause hypomineralized AI (Smith et al., 2016). No variant in this gene was found among our cohort.

The protein encoded by this gene is a member of the TNF-receptor superfamily able to activate the NF-kappaB pathway and selectively bind TNF receptor-associated factor 1 (TRAF1).

In mice, Relt is expressed in the odontoblast and ameloblast layers, specifically in secretory stage ameloblasts where there is ∼20-fold higher expression than in maturation stage ameloblasts. It has been published that Relt ^−/− incisor enamel was of normal thickness but rough-surfaced and generally hypomineralized correlating with the phenotype found in humans. RELT was originally classified as causative of a new type of AI called type IIIC (#618386) but following Witkop’s classification, as recessive hypocalcified hypoplastic AI, it could be classified in type IIIB.

We did not detect any individuals with RELT variants in our cohort.

DLX3 is an important transcription factor involved in osteogenic differentiation (Sun et al., 2019). It is located on chromosome 17q21.3-q22, and contains 3 coding exons (Scherer et al., 1995; Price et al., 1998). DLX3 plays a role in craniofacial development, and in the development of the ventral forebrain. DLX3 has three main domains: The N- and C-terminus transactivation domains, and a central sequence-specific DNA-binding distal-less-like homeodomain, encoded by exons 2 and 3. The homeodomain can interact directly with DNA in a sequence-specific way and regulates the expression of target genes throughout numerous developmental processes (Whitehouse et al., 2019). Only one missense variant had been described in DLX3 gene for AI with taurodontism and attenuated tricho-dento-osseous syndrome in exon 2 by (Whitehouse et al., 2019).

Individuals 16.1 and 16.2 showed hypoplastic amelogenesis imperfecta with taurodontism (Figure 4), and enamel surface with striae (16.1), or thinner enamel (16.2). Taurodontism was really apparent on 16.2. 16.3 presented an hypomature/hypoplastic enamel and smaller 12, 22, as well as probable third molar agenesis. Dental agenesis could be linked to the additional variant discovered in WNT10A gene NM_025216.3:c.637G>A; p.(Gly213Ser). Dental radiographs confirmed the decreased thickness of enamel, and various degrees of molar taurodontism (Supplementary Figure S1K). In 16.1 we detected a heterozygous missense variant in exon 1 NM_005220.3:c.92C>G; p.(Thr31Ser) predicted possibly damaging by Polyphen-2 and localized in the distal-less-like homeobox protein domain. In individual 16.2 the heterozygous variant was located in exon 3 c.537C>A; p.(Asn179Lys) concerning a well conserved amino acid and predicted damaging by SIFT (v6.2.0) and Polyphen-2. Individual 16.3 carries the variant c.710A>G; p.(Tyr237Cys) (Supplementary Table S3; Supplementary Figure S2P). 16.1-16.3 did not present any bone nor hair additional phenotypes.

As this gene has also been linked to tricho-dento-osseus syndrome, we will describe further syndromic individuals linked to DLX3, in the coming syndromic section of this publication, illustrating therefore the tight and thin border between isolated and syndromic AI.

Hypoplastic AI is the hallmark of numerous syndromes. To date 22 genes have been associated to syndromes including hypoplastic AI in their clinical synopsis. Thirteen genes (FAM20A, GALNS, TSC1, TSC2, TP63, MSX2, FAM20C, ARHGAP6, RAI1, PEX1, PEX2, PEX26, TGFBR2 and ATP6V1A) have been associated with clinical signs within the head and neck area, eleven with skin, nail and hair defects (TSC1, TSC2, TP63, ARHGAP6, PORCN, TGFBR2, PEX1, PEX2, PEX26, ORAI1 and STIM1), three with immune deficit (AIRE, ORAI1 and STIM1), seven with skeletal defects (LTBP3, GALNS, TP63, MSX2, PORCN, and FAM20C), five with neurological issues (ATP6V1A, SLC13A5, PEX1, PEX2 and PEX26), two with cardiovascular defects (LTBP3, TGFBR2), and eight with genitourinary defects (FAM20A and FAM20C, CLDN19, CLDN16, WDR72, TSC1, TSC2, RAI1). It is therefore of importance to explore these potential associated phenotypes while taking medical history and examination.

DASS is an autosomal recessive disorder caused by homozygous or compound heterozygous variants in LTBP3 gene (11q12) (Li et al., 1995; Huckert et al., 2015). LTBP3 encodes latent transforming growth factor-beta-binding protein 3, modulating TGFbeta bioavailability in the extracellular matrix. Deleterious variants in LTBP3 have also been associated with autosomal dominant Geleophysic dysplasia 3 (#617809). Ultrastructural enamel defects showed an absence of initial aprismatic enamel layer and an abnormal secretion of non-prismatic bulk enamel, suggesting LTBP3 plays a role in the life cycle of ameloblasts especially at the secretory stages with Tomes process formation (Huckert et al., 2015).

We identified LTBP3 as the gene behind short stature and absent enamel (individuals 17.1–17.4) using exome sequencing (Huckert et al., 2015). GenoDENT panel was subsequently exanded with addition of this gene and we identified in 2 unrelated consanguineous families, additional individuals 17.5 and 17.6 with a similar phenotype and two new homozygous loss of function variants (Table 2; Supplementary Figures S1L, S2Q, S3.38). Teeth were small, spaced, and had a yellowish color. No enamel, impacted teeth, and irregular alveolar bone level were visible on panoramic radiographs. The first pathogenic variant (Individual 17.5) NM_001130144.3:c.3087del; p.(Asn1030Thrfs*47) was a one nucleotide deletion leading to a frameshift and a premature stop codon. The second variant (17.6) was an intronic mutation c.3629-2A>G which led to an aberrant exon 27 splice site. This would probably induce exon 27 skipping resulting in a 132 bp in phase deletion equivalent to a 44 amino acid deletion (position 1,211–1,254).

Mucopolysaccharidosis type IVA (MPS IVA; Morquio syndrome A) is an autosomal recessive lysosomal storage disease caused by variants in the galactosamine-6-sulfate sulfatase gene (GALNS), located on chromosome 16q24.3. GALNS contains 14 exons and 13 introns (Sawamoto et al., 2020).

We report two individuals with compound heterozygous variations in the GALNS gene. Individual 19.1 was previously described by our team (Prasad et al., 2016a).

In individual 19.2, we found a missense variant NM_000512.5:c.1156C>T; p.(Arg386Cys) firstly described by (Ogawa et al., 1995), the second variant c.1558T>C; p.(Trp520Arg) was a class 2 variant so we cannot conclude that this variant is indeed involved in the individual phenotype as it is inherited by her homozygous unaffected mother (Table 2). Nevertheless, individual 19.2 presents the typical enzymatic deficiency and was given a confirmed MPS clinical diagnosis. So, either this class 2 variant contributes to the phenotype, or the individual is carrying another variant in this gene not detected by the panel such as a deep intronic variation. Such deep intronic variants have already been described in GALNS (Caciotti et al., 2018).

The disease has been associated to variations in the autoimmune regulator gene (AIRE) which consists of 14 exons (Nagamine et al., 1997) and is located on chromosome 21q22 (Aaltonen et al., 1997). AIRE protein is localized in the cell nucleus, where it forms distinct speckles (Björses et al., 1999). Analysis of its multidomain structure reveals that human AIRE belongs to the group of proteins able to bind to chromatin and regulate the process of gene transcription (Perniola, 2018). No variant in this gene were found in our cohort.

Individuals with TSC present variants of the TSC1 and TSC2 genes, which intervene in cell cycle regulation. The TSC1 gene (9q34) encodes hamartin, a protein that interacts with tuberin (TSC2) to form a protein complex that inhibits signal transduction to the downstream effectors of the mammalian target of rapamycin (MTOR) (Inoki et al., 2002). The TSC1 gene consists of 23 exons, of which the last 21 contain coding sequence and the second is alternatively spliced (Slegtenhorst et al., 1997).

The TSC gene on chromosome 16 was named TSC2. TSC2 has 41 small exons spanning 45 kb of genomic DNA and encodes a 5.5-kb mRNA (van Bakel et al., 1997). No variants of TSC1 or TSC2 were identified in our cohort.

TP63 plays an important role allowing cells to undergo apoptosis in response to DNA damage (Flores et al., 2002) and is involved in tumor and metastasis suppression (Su et al., 2010). Given its broad expression pattern, we suggest that TP63 mutation affect the cell differentiation or fate of ameloblasts in development, though we await experimental evidence. Kantaputra et al., 2003 identified a heterozygous missense variant (S545P) in the TP63 gene in a Thaï teenager presenting all the characteristics of the syndrome. No variant in this gene is reported in this publication.

The gene is located on chromosome 5q34-q35 (Jabs et al., 1993). (Hassan et al., 2004) showed that Msx2 regulates the expression of osteocalcin and therefore is implicated in the control of bone formation. This gene is reported for craniosynostosis in human. Two previous publications (Aïoub et al., 2007; Molla et al., 2010) reported that, in the targeted deletion mouse model Msx2 ^−/−, Msx2 was implicated in both isolated enamel dysplasia (regulating amelogenin, enamelin) and syndromic enamel dysplasia (through alterations in cell-cell junctions). To date one duplication of the entire gene MSX2 has been reported in a syndromic (craniofacial, eye and limb anomalies) individual associated with hypoplastic AI (Plaisancié et al., 2015). No variant in this gene was identified in our cohort.

(Simpson et al., 2007) identified homozygous and compound heterozygous variants in FAM20C (7p22) in individuals with Raine syndrome. FAM20C is a Golgi associated secreted protein kinase, partnering with FAM20A, phosphorylating small integrin-binding ligand N-linked glycoproteins SIBLINGS, among other proteins, and playing a substancial role in osteogenesis and amelogenesis. No variant in this gene was identified in our cohort.

Most of the individuals with FDH are female, with heterozygous or mosaic variants in the PORCN gene. Males (mosaic variants) account for 10% of affected individuals; heterozygous non-mosaic variants are lethal in males (Deidrick et al., 2016).

PORCN gene is located in chromosome Xp11.23, and contains 15 exons and spans about 12 kb. The first exon is non-coding (Caricasole et al., 2002). PORCN encodes and endoplasmic reticulum protein: the protein-serine O-palmitoleoyl transferase porcupine or porcupine O-acyltransferase. Although the exact function of the PORCN protein is uncertain, proteins in the porcupine (PORC) family are involved in WNT (wingless and int homologue) signaling pathway which is extremely important for embryonic development (Durmaz et al., 2018) including odontogenesis. No variant in this gene was identified in our cohort.

Because the Amelogenin gene (AMELX) is nested within intron 1 of ARHGAP6, partial deletions in ARHGAP6 can completely remove AMELX giving a distinctive enamel phenotype resembling “snow-capped” teeth (Hu J. C.-C. et al., 2012). This phenotype was already described in isolated hypomature AI IIC section.

Ras homologue GTPase activation protein 6 (ARHGAP6), belongs to the Rho GTPase-activating protein (RhoGAP) family. ARHGAP6 does not appear essential for normal enamel formation. Failed ARHGAP6 expression did not appreciably alter the severity of enamel defects when AMELX was absent (Hu J. C.-C. et al., 2012).

We report one individual (20.1) carrying a 12 Mb deletion of the Xp22.2 region (Table 2). This region is including not only the AMELX gene but also the full ARHGAP6 gene.

This female individual suffers from microphthalmia with linear skin defects (MLS) syndrome and associated hypoplastic enamel with a banding pattern characteristic of the female lyonisation effect (bands of normal enamel next to defective enamel; AI hypoplastic IE) (Supplementary Figure S1N).

The syndrome is in 90% of the cases due to a 3.7 Mb deletion in chromosome 17p11.2 encompassing the RAI1 gene (Vieira et al., 2012).

We identified with the GenoDENT panel, in individual 21.1, a deletion in the 17p11.2 region leading to a possible diagnosis of Smith-Magenis syndrome with hypoplastic AI (Table 2). The exact size of the deletion was further characterized by array analysis. Individual 21.1 presented the classical 3.7 Mb deletion associated with the syndrome as well as classical associated phenotype including the hypoplastic enamel described by (Wright et al., 2015).

Variants in TGFBR2 have been associated with Loeys Dietz syndrome type 2. TGFBR2 codes for a transmembrane protein with a serine/threonine-kinase domain forming a heterodimeric complex with TGF-beta type I receptor, TGFBR1, binding TGF-beta, TGF-beta1, TGF-beta2 and TGF-beta3, ligands. This receptor/ligand complex phosphorylates proteins, which then enter the nucleus and regulate the transcription of genes related to cell proliferation, cell cycle arrest, wound healing, immunosuppression, and tumorigenesis.

Here we report an individual 22.1 with a heterozygous missense variant in the TGFBR2 NM_003242.6:c.1561T>C; p.(Trp521Arg) (Table 2; Supplementary Figures S1O, S3.43) and diagnosed with Loeys-Dietz syndrome including hypoplastic amelogenesis imperfecta. This variation was already reported by (Mátyás et al., 2006). The variant is most probably transmitted by the affected mother though we were not able to access the mother’s DNA to confirm.

This disorder has been associated with ATP6V1A (ATPase, H + Transporting, Lysosomal, 70-KD, V1 Subunit A). It codes for a component of vacuolar ATPase, a multimeric enzyme, an ATP-dependent protein pump function, which mediates acidification of eukaryotic intracellular organelles and is necessary to activate mTORC1 (Zoncu et al., 2011). No variant in this gene was found among our cohort.

(Simon et al., 1999) identified homozygous and compound heterozygous variants in the CLDN16 gene in ten individuals presenting renal hypomagnesemia. CLDN16 is localized on chromosome 3q28, and consists of 5 exons and encodes a protein of 305 AA with 4 transmembrane domains and intracellular N and C termini (Simon et al., 1999). CLDN16 is required for cell division (Kittler et al., 2004), is selectively expressed at tight junctions of renal epithelial cells and plays a central role in the reabsorption of divalent cations (Kausalya et al., 2006). No variant in this gene was identified among our cohort.

Konrad et al., 2006 found two different homozygous missense variants in the CLDN19 gene in families with renal magnesium wasting, renal failure, and severe ocular involvement. Indeed, the syndrome is caused by homozygous or compound heterozygous variants in the claudin-19 gene on chromosome 3q28. Claudins, such as CLDN19, are transmembrane proteins found in tight junctions. Tight junctions form barriers that control the passage of ions and molecules across an epithelial sheet and the movement of proteins and lipids between apical and basolateral domains of epithelial cells (Lee et al., 2006). No variant in this gene was identified among our cohort.

Ratbi et al. (2015) identified homozygous and compound heterozygous variants in the peroxisome biogenesis factor 1 gene (PEX1) in 4 families, including the family reported by (Heimler et al., 1991). Human PEX1 gene is located on chromosome 7q21, and is composed of 24 exons and encodes a 147-kD protein member of the AAA protein family (ATPases associated with diverse cellular activities). Those proteins participate in a broad range of cellular processes, as indicated by the designation AAA which comes from ATPases associated with diverse cellular activities and are specially required for peroxisomal matrix protein import (Portsteffen et al., 1997; Reuber et al., 1997). No variant in this gene was identified among our cohort.

Another form of Heimler syndrome (HMLR2; #616617) is caused by a variant in the PEX6 gene (601,498) located on chromosome 6p21, consisting of 17 exons and 16 introns, spanning about 14 kb (Zhang et al., 1999). No variant in this gene was identified among our cohort.

Developmental and epileptic encephalopathy 25 with amelogenesis imperfecta is a rare disease caused by variants in SLC13A5 genes. SLC13A5 is located in the chromosome 17p13-p12, and contains at least 12 exons (Inoue et al., 2002). SLC13A5 encodes a high affinity sodium-dependent citrate transporter, which is mainly expressed in liver and brain. Neurons are considered incapable of de novo synthesis of tricarboxylic acid cycle intermediates; therefore, they rely on the uptake of intermediates, such as citrate, to maintain their energy status and neurotransmitter production (Hardies et al., 2015).

This epilepic encephalopathy was described as Kolschütter-Tönz like syndrome (KTS like) as the phenotype is similar to the one observed in individuals with variations in the ROGDI gene except for the enamel defects: AI is hypoplastic in patient with SLC13A5 variations and hypomature in individuals with ROGDI mutations. Kolschütter-Tönz syndrome will be described later in the paper as individuals present a hypomature AI.

We report one individual 23.1 carrying a compound heterozygous missense variants affecting exons 2 and 4 NM_177550.5:c.203C>A; p.(Pro68Gln), c.434C>A; p.(Thr145Lys) showing a hypoplastic AI with thin opaque enamel (Table 2; Supplementary Figures S1O, S2T). The SLC13A5 individual in our panel, is an individual described with SLC13A5 variant discovered through GenoDENT panel after the identification of the gene by WES in another family of the Schossig et al., 2017 cohort (Schossig et al., 2017) with a phenotype described as Kohlschütter-Tonz like syndrome (KTS).

By analyzing individuals described by (Feske et al., 1996), (Feske, 2010) showed for the first time that individuals with this type of immune dysfunction were homozygous for a variant in the ORAI1 gene. Indeed, the disease is caused by homozygous or compound heterozygous variants in ORAI1 (610,277). This gene located on chromosome 12q24, encodes a subunit of the plasma membrane calcium channel CRAC essential for store-operated calcium entry (Vig et al., 2006) and the channel function. The gene is expressed in cells and organs involved in immunity (CD4⁺ and CD8⁺ T-cells, CD19⁺ B-cells, and in a subset of cells in the thymus, spleen, and tonsils) but also in sarcolemma of muscle fibers, eccrine sweat glands, skin, vascular endothelium, hepatocytes, lung, and kidney (McCarl et al., 2009) and in ameloblast (Nurbaeva et al., 2015). No variant in this gene was identified among our cohort.

Immunodeficiency 10 is due to a homozygous truncating variant in the STIM1 gene (Picard et al., 2009; Parry et al., 2016a). STIM1 gene is located to chromosome 11p15.5 and contains 12 exons (Parker et al., 1996; Sabbioni et al., 1999). It encodes a 746 AA calcium sensor that conveys the calcium load of the endoplasmic reticulum to store-operated channels at the plasma membrane (Yuan et al., 2007) and mediate the gating of CRAC channels (McNally et al., 2012). ORAI1 and STIM1 are interacting as the ORAI1 membrane calcium channel is activated by the calcium sensor STIM1 when calcium reservoirs are depleted (Lacruz and Feske, 2015). AI, related to Immunodeficiency 10, is classified as hypomineralized on OMIM but due to recent publications we transfer it to hypomature AI section (Wang S. et al., 2014; Furukawa et al., 2017). No variant in this gene was identified among our cohort.

(Schossig et al., 2012) reported for the first time the causative gene by identifying homozygous or compound heterozygous variants in ROGDI gene in 3 members of 3 unrelated families with KTS. ROGDI maps to chromosome 16p13.3., contains 11 exons and spans over 5.98 kb. It encodes a leucine-zipper protein with high expression in the human brain and spinal cord (Mory et al., 2012).

Both variants in ROGDI and SLC13A5 cause epileptic encephalopathy and AI. Whereas ROGDI mutation manifests in an hypomature-hypomineralized AI with a rough colored dental surface and with seizure onset may be as late as age 3, individuals mutated for SLC13A5 present hypoplastic AI, a smooth the dental surface sometimes with only mild discoloration and seizure onset is within the first days after birth.

Here we report three individuals with stop variants in ROGDI (24.1, 24.2, 24.3) (Table 2; Supplementary Figures S1P, S2U). Individuals 24.1 and 24.3 present compound heterozygous variants (Supplementary Figures S3.44) already described in the literature, while 27.2 has a homozygous variant found with our panel and previously published in (Huckert et al., 2014).

This autosomal recessive syndrome was firstly described by (Dubail et al., 2018) in 6 unrelated individuals. It is due to variants in the SLC10A7 gene encoding a calcium transporter.

The key phenotype to distinguish individuals with mutation in SLC10A7, within the wide spectrum of skeletal dysplasia, was the hypomineralized/hypomature enamel defects observed in all the individuals and the hypoplastic lower jaw.

Here we report 2 individuals with homozygous variants in SLC10A7. Individual 25.1 presents a novel homozygous variant in the gene. The variant identified by GenoDENT NGS panel affects exon 3 NM_001300842.3c.269T>G; p.(Leu90Arg) in which another homozygous variant was already described to be pathogenic (Ashikov et al., 2018). Individual 25.2 was already described in (Laugel-Haushalter et al., 2019, WES). She has a homozygous missense variant in the very last exon of the gene c.908C>T; p.(Pro303Leu) (Table 2; Supplementary Figure S3.45) and presents a mild phenotype of the disease but with the characteristic enamel defects (Supplementary Figures S1Q, S2V).

The disorder is caused by homozygous or compound heterozygous variants in CNNM4 gene (607,805) sitting on chromosome 2q11.2. More than 24 different variants have been identified on individuals all around the world but the molecular mechanism of the disease remains unclear (Daneshmandpour et al., 2019). CNNM4 encodes a deduced 775 amino-acids protein. (Guo et al., 2005) hypothesized that the protein may have a role in metal ion transport and homeostasis. Indeed, (Yamazaki et al., 2013) showed in mice the role of Cnnm4 in Mg²+ transport. The protein is localized in keratocytes, in the retina, and in developing teeth specifically in ameloblasts (Parry et al., 2009; Polok et al., 2009).

Here we report two individuals with homozygous missense variants in this gene. In individual 26.1 we detected a homozygous missense variant, firstly described by (Parry et al., 2009): NM_020184.4:c.1495G>A; p.(Val499Met) (Table 2; Supplementary Figure S3.46). Other missense variants, were reported as pathogenic and causative of Jalili syndrome. Individual phenotype was consistent with the literature and described clinical synopsis. The “hypoplastic” enamel appearance of both primary and permanent teeth was due to extensive post-eruptive loss of soft enamel detaching easily from dentin. The teeth displayed yellow-brownish coloration with brown spots (Supplementary Figures S1R, S2W). Individual 26.2 has already been reported in (Prasad et al., 2016a) c.1495G>A; p.(Val499Met).

TDO is caused by variants in a transcriptional regulator, Distal-less homeobox 3 gene, DLX3.

Here we describe two unrelated individuals, one girl (27.1) and one boy (27.2) presenting with tricho-dento-osseous syndrome (Supplementary Figure S1S). They are both carrying the same DLX3 heterozygous frameshift variant in exon 3 NM_005220.3:c.561_562del; p.(Tyr188Glnfs*13) (Table 2; Supplementary Figures 3.47 and 3.48). This variant firstly described by (Dong et al., 2005) is altering the two amino-acids of the DNA-binding homeodomain and truncating the protein by 88 amino-acids.