Recruitment and phenotypic analysis

The probands were research patients at the Institute of Dentistry, University of Helsinki, and patients referred to the Helsinki University Central Hospital (HUCH) because of need of dental treatment. They were informed of the purpose and procedures of the research with discussion and with information leaflets and they, or in case of children under 15 years of age, their parents, signed a form of an informed consent. The relatives were also asked to participate in the study. All clinical and molecular genetic studies were conducted according to the principles expressed in the Declaration of Helsinki and under permission from the Ethics Committee, Department of Surgery, HUCH. Diagnosis of tooth agenesis was confirmed by clinical and radiographic examination or from previous documents. The general health, including ectodermal features and other syndromic indications as well as history of cancer, were assessed by clinical examination and interviews. The patients with diagnosis or suspected syndrome were excluded from this study. The genomic DNA was isolated from blood samples in the DNA isolation core unit of the National Institute for Health and Welfare, Biomedicum Helsinki, or from buccal swabs in the laboratory of the research group by using Qiaamp DNA mini kit (Qiagen).

Mutational analysis

For a mutational analysis by sequencing of exons of MSX1, PAX9, AXIN2, EDA, EDAR (ENSG00000135960), EDARADD (ENSG00000186197), WNT6 (ENSG00000155196), WNT10A and WNT10B (ENSG00000169884) primers were planned to span exons and at least 30 bp of the flanking intronic sequences with Primer 3 software (http://frodo.wi.mit.edu/). The target was amplified by PCR and the products purified with ExoSAP-method (USB) and used as templates in the sequencing reaction by ABI BDRR reagent v 3.1 (Applied Biosystems). The sequencing products were subjected to capillary electrophoresis in the Biomedicum Helsinki Molecular medicine sequencing laboratory. The results were compared with known genomic sequences for each gene (BLAST, http://www.ncbi.nlm.nih.gov/blast; MSX1: NT_006051.15; PAX9: NT_026437.10; AXIN2: NT_010783.15; EDA: NT_011669.16; EDAR: NT_022171.14; EDARADD: NT_004836.17; WNT10A and WNT6: NT_005403.17; WNT10B: NT_029419.12) and the observed deviation from reference stored in a proprietary database. For the primer sequences and PCR and sequencing conditions, see Table 1. Observed missense variants were subjected to bioinformatic and conservation analysis by SIFT (http://sift.jcvi,org) and PolyPhen2 (http://genetics.bwh.harvard.edu/pph2/index.shtml). Eventual effect on splicing sites were studied by NetGene2 (http://www.cbs.dtu.dk/services/NetGene2/) and Berkeley splicing site prediction tool (http://www.fruitfly.org/seq_tools/splice.html) and effects on functional sites by ELM (Eukaryotic linear motif resource (http://elm.eu.org/).

Nature of the identified mutations

Frameshift and nonsense mutations are highly likely all causative because they involve profound alteration of the protein primary structure. The two mutations in MSX1 are the first mutations that affect the C-terminal part of the protein downstream of the homeodomain the function of which is poorly understood. The in-frame deletion in the collagen-like domain observed in EDA is similar albeit not identical to several previously reported mutations in HED families [33–35].

We excluded missense variants that were present with high frequencies in our patient and unaffected control cohorts or in the dbSNP or 1000genomes databases. The remaining missense variants were almost all suggested to be harmful by the bioinformatic tools and most of them affected known functional domains. The two missense changes in PAX9 affected the highly conserved paired domain and are expected to affect the DNA-binding properties of this domain. The EDA p.R357W and p.T378M affect the TNF homology domain (InterPro IPR006052) of this extracellular signaling protein, and similar mutations in this domain have previously been shown to cause tooth agenesis in males who carry a single copy of this X-chromosomal gene and in part of heterozygous female carriers [14,16,36]. The EDAR missense mutations reside in the intracellular death and death-like domains (InterPro IPR000488 and IPR011029) important for receptor multimerization and the interaction with the adapter protein EDARADD essential for signal transduction. The EDARADD p.R170W resides in the death domain interacting with EDAR. Previously, both homozygous and heterozygous mutations affecting these domains have been described in families with obvious recessive and dominant inheritance of HED, respectively (OMIM 129490 and 614941) [34,37–40]. We suggest that p.M391K which was detected in a father and daughter with a severe tooth agenesis and perhaps also the EDARADD p.R170W have some extent of dominant negative effect. The other mutations in EDAR were associated with more variable and usually less severe phenotype and some relatives were unaffected. We suggest that these amino acid changes essentially cause loss of function that causes tooth agenesis with incomplete penetrance and that the patients express carrier phenotypes of autosomal recessive HED (OMIM 224900). Indeed, the p.E379K mutation is homologous to the original mutation described in downless (dl), the mouse model of autosomal recessive HED [41]. EDARADD p. S103F was reported in heterozygous state earlier in a single patient who lacked six permanent teeth [42]. We detected this variant in heterozygous state in 11 probands but also in some unaffected controls, suggesting low penetrance if present isolated in heterozygous state.

Based on alignment with XWnt8, the Xenopus Wnt protein whose three-dimensional structure and interactions with the Frizzled receptor has been elucidated by X-ray crystallography [43], all missense variants of WNT10A affect the N-terminal domain which consists of an alpha-helical bundle. p.F228I homozygous and heterozygous genotypes were observed in this and earlier studies on ectodermal dysplasia syndrome and isolated tooth agenesis [17,18,22,38] in striking contrast with the prevalences among the controls. p.R113C was detected in several probands but not in controls and it was previously found in a compound heterozygote patient [19]. We observed p. R70W, p. V145M, p.V145L and p.L154M only in single families and only in heterozygous state but they were also predicted as harmful by both SIFT and PolyPhen2. p.V145M has been described earlier in patients [18,38]. A previously known variant p.G165R was predicted as benign, and was previously observed in trans together with p.F228I in unaffected persons [17]. However, the presence in six patients with severe tooth agenesis in our cohort but not in controls, and the consequences of the amino acid change on the level of the protein primary structure, suggest that association of this variant with abnormal tooth development should not be excluded.

Several of the probands who had heterozygous genotypes of WNT10A had equally severe phenotypes as those with biallelic genotypes. Heterozygous genotypes of WNT10A were described also in earlier reports in association of syndromic phenotypes or isolated severe tooth agenesis [18,19,22,38]. Phenotypic variation is inherent in tooth agenesis but we consider it plausible that the extreme variation observed in these cases reflects existence of another mutation. In two of our families all affected relatives did not have the mutation detected in the proband, and these included sons of the proband in family 12 and mother of the proband in family 261 (Table 2). In these and several other families we could not identify any alterations in the promoter and excluded whole exon deletions in WNT10A as well as coding region mutations in the co-expressed WNT10B and WNT6. However, we observed the EDARADD mutation, p. S103F, in several probands with heterozygous WNT10A genotypes, as well as in two with heterozygous genotypes of EDAR. Most of the probands with these allele combinations had a severe phenotype as compared to the other probands or relatives who had only the EDAR or WNT10A allele. One of the probands (family 266) was actually heterozygous for three variants, EDAR p.R325W, EDARADD p. S103F and WNT10A p.G165R, consistent with the most severe tooth agenesis observed among the EDAR heterozygotes. Combinations of two mutations affecting the EDA pathway, i.e. the EDA receptor and the adapter protein EDARADD, may be expected to have a stronger effect than a single mutation even though the combination shall not affect the pathway to a similar extent as biallelic mutations in one gene. It would neither be surprising if combinations of harmful albeit heterozygous mutations in EDA and WNT pathways would cause additive effects. It is known that both pathways have an instrumental role in the dental epithelium from the early stages of tooth development and it is known that they interact during dental and hair development [44,45]. This is the first molecular observation of combined effects of mutations in distinct genes on tooth agenesis, implying that monogenic inheritance does not explain all cases of tooth agenesis.

Contributions to tooth agenesis

Overall, we have described mutations in 58 probands of our Finnish patient cohort, including our previous reports on two families with mutations in PAX9 and two with mutations in AXIN2 [13,26,27]. These account for 44% of all our probands. However, in one third of the probands the observed genotype of WNT10A or EDARADD may not alone explain the phenotype. The significant majority of these probands had a severe phenotype, and most of them fulfilled the criterium for oligodontia (agenesis of six or more permanent teeth, excluding third molars). Several of the other probands had a specific phenotype that involved both permanent and deciduous incisors. Only a few of the probands with identified mutations conformed to the typical incisor premolar hypodontia [25] and no mutations were detected in the probands or other members of the multiplex families in which the latter phenotype is inherited in an autosomal dominant manner [25]. However, this kind of phenotypes were observed in several heterozygote relatives of the probands with oligodontia.

Considering our whole patient cohort, i.e. including the previous reports, we have detected mutations in PAX9 in six probands, and two in both AXIN2 and MSX1, all with severe phenotypes. Our results are in line with reports from the earlier reports in which population-based patient cohorts of severe tooth agenesis have been analysed for candidate gene mutations [18,42], supporting the conclusion that dominant mutations in these genes explain severe tooth agenesis in a relatively minor proportion of all families. However, the relative contribution of these three genes with dominant mutations to severe tooth agenesis becomes more significant if all patients, and not only probands, are considered. Thus, in our whole patient cohort, severe tooth agenesis is explained in altogether 13 patients by mutations in PAX9 and in 11 by mutations in AXIN2.

We detected mutations in EDA in four probands, three with severe phenotypes. While EDA mutations are revealed as a cause of isolated tooth agenesis, in female carriers or as hypomorphic mutations in males often appearing recessive or sporadic, they appear to have a minor contribution [18,42]. However, our results suggests that the whole contribution of EDA pathway is more significant, considering also the mutations in the EDAR and EDARADD genes. To our knowledge, this is the first report describing mutations in EDAR in isolated tooth agenesis. They were detected in eight probands, three of whom had a severe phenotype. Heterozygous EDARADD p. S103F variant was detected in 11 of our probands, most with a severe phenotype. Altogether this variant was present in 15 unrelated samples, 11 affected and four unaffected, which is in line with the reported allele frequency of 2% in the dbSNP database. Thus, this variant has potentially a significant contribution to tooth agenesis on the population level.

We found biallelic genotypes of WNT10A in 11 probands, corresponding to 8.4% of all probands, and 13% of those with a phenotype defined as oligodontia. Heterozygote genotypes of WNT10A – including p.G165R - were observed in 19 probands with severe tooth agenesis and two other probands, which increased involvement of this gene to 23% of all probands and to 35% of those with a severe phenotype. We also detected several affected, most with typical incisor premolar hypodontia phenotype, as well as unaffected heterozygote relatives, consistent with reduced penetrance of these alleles. Thus, WNT10A has the most significant contribution of so far identified genes to tooth agenesis, in line with previous reports [18]. Several of the mutations have been detected in multiple studies and thus the high contribution of WNT10A to tooth agenesis is most apparently based on the presence, and even with moderate frequencies, of the disease causing alleles in all studied populations. However, the contribution reported here is remarkably smaller than in recent reports from the Netherlands and Poland [18,19]. Our cohort was larger than in the previous studies and includes more probands with less severe phenotypes (mean number of missing teeth of 9.3 in oligodontia as compared to 14.6 in the Dutch cohort). In addition, it appears that there are differences in the allele frequencies between different studied populations. We have detected no cases of a common nonsense allele, p.C107X, in our cohort or controls. The Dutch study reported a frequency of 2.3% and the Polish study a frequency of 1,3% for the most common disease allele, p.F228I [18,19]. dbSNP database reports similar frequencies for this allele in two large cohorts but the German study reported a significantly lower frequency of 0.5% [17]. We did not detect this allele once among the more than 100 unaffected controls, suggesting frequency similar to the German study in the Finnish population.

What are the implications of our results for the genetic background of the common types of tooth agenesis, especially the common hypodontia, which involves agenesis of one or a few permanent incisors and second premolars and most often follows dominant inheritance [25,46]? We or others have not identified coding region mutations in MSX1, PAX9 and AXIN2 associated with this common phenotype. However, putative associations of several silent or non-coding region variants to common hypodontia have been reported [47,48]. The results from this and earlier studies show that several variants of EDAR and WNT10A in heterozygous state contribute to incisor premolar hypodontia phenotypes with a penetrance that conforms to dominant inheritance. This kind of alleles – and also in other genes – connect the genetic background of common hypodontia and severe tooth agenesis: the alleles that in heterozygous state cause common hypodontia by dominant inheritance, cause severe tooth agenesis when present as biallelic recessive genotypes or allele combinations of different genes. However, they were not present in most of our probands with a hypodontia phenotype and especially not in the families in which we have described dominant inheritance [25], suggesting that they have a minor contribution to these phenotypes.

Gene specific effects on phenotypes

Mammalian teeth are divided into different classes according to their morphology and position on the dental arches. Despite increasing knowledge of the genetic mechanisms underlying tooth development, and even of the regulatory mechanisms important in modulating the cusp pattern and complex morphologies of multicusped teeth [49], the genetic basis providing different morphologies (heterodonty) in mammalian dentitions has not been elucidated.

As supported by our results, heterozygous loss of function mutations in MSX1 and PAX9 have a strongest effect on the posterior multicusped teeth. MSX1 mutations affect almost uniformly all second premolars and third molars, while permanent molars are the most sensitive to PAX9 mutations [8,12]. In HED, even complete anodontia may sometimes be observed as a consequence of complete inactivation of the EDA signaling pathway, but in female carriers of the X-linked mutations, the tooth agenesis is most prominent in the anterior dentition [50]. The hypomorphic EDA mutations have an even more specific effect on both deciduous and permanent incisors in both sexes [14,16]. The phenotypes of our male and female patients with mutations in EDA conform to this tendency but it is also recapitulated by the phenotypes associated with heterozygous mutations in EDAR. However, this feature was not observed in the patients with heterozygous mutations in EDARADD.

Compilation of the data on tooth specific phenotypes associated with biallelic WNT10A mutations [17–19,22] shows that these mutations affect all tooth types, in line with the expression of the mouse homolog in both incisor and molar tooth germs [23]. Most commonly affected are the third molars, and after them second premolars and maxillary lateral and mandibular incisors whereas the first permanent molars and maxillary central incisors are most stable. This order follows the prevalences of agenesis of different tooth types in general [4]. However, frequencies of missing canines, first premolars and second permanent molars are also quite high among the patients with biallelic WNT10A mutations. Furthermore, there is significant variation between patients in the severity of agenesis and in the tooth types that are affected. As a further distinction to EDA pathway, deciduous teeth are usually not affected in the non-syndromic patients who most often carry missense than stop-codon creating mutations that are more common among syndromic patients [18,21,38,51].

The differential sensitivities of specific parts of dentition for reduced doses of genes may be dependent on differential expression levels of these genes and they may reflect different roles of the genes during normal development. Posterior, multicusped teeth may be especially sensitive to the role of MSX1 and PAX9 as regulators of the recruitment of the mesenchyme to dental fate and the regulation of mesenchymal signaling necessary to the cascade of the morphogenetic epithelial signaling centers enamel knots [49,52]. On the other hand, EDA signaling regulates the size of the ectodermal signaling centers [53,54], which may be critical for the development of anterior dentition. Wnt10a expression during mouse dental development is largely superimposed to that of Edar [23], but its target cells are not known. Despite the regulatory interaction between these pathways, the differences in human phenotypes suggests that WNT10A deficiency exerts its effect through different mechanisms than EDA signaling.