The pedigrees are shown in . Both the families showed an apparent X-linked dominant form of tooth agenesis. Family A comprises 56 members spanning four generations, 9 members in this family show congenital tooth agenesis (male 6, female 3). Family B consists of five generations including 13 members with congenital tooth agenesis (male 11, female 2).
Pedigree structure of the two Chinese families with tooth agenesis.
In both families, the manifestation of tooth agenesis is not uniform (). In Family A, the hypodontia involved all classes of teeth. However, in Family B, all the affected members predominantly lacked incisor teeth but had all the permanent molars. The phenotype in Family A was more severe in terms of the numbers of missing teeth, i.e., with the proband lacking all permanent premolars, cans, incisors, as well as all the third molars and three second molars. Phenotypic characteristics of scalp and body hair, skin, nails and ability to sweat were examined in all individuals of both families. All the individuals had normal sweating and had no complaints about intolerance to heat, and their facial features, skin, and nails all appeared normal.
Positional cloning of the oligodontia gene
By haplotype analysis of the pedigree of Family A (), the affected locus was confined to the region between DXS1039 and DXS8064. In the two-point linkage analysis, two loci in this region, DXS1196 and DXS986, both gave the highest two-point LOD value of 3.13 (), strongly suggesting that the disease gene associated with tooth agenesis is closely linked to the two markers.
In this critical region, the candidate genes for the oligodontia locus include those encoding for transcription factors or proteins involved in signal transduction. The relative positions of these genes are shown in . We therefore proceeded to screen for mutations in these genes. By directly sequencing all exons and flanking splice junctions of the candidate genes, ITM2A, TBX22, SH3BGRL, ZNF711, KLHL4, CPXCR1 and TGIF2LX were ruled out because no mutation was detected (data not shown). Our best candidate pointed to EDA. The EDA gene structure is shown in .
Candidate genes identified in the critical region in the Xp11-Xq 21.
To identify possible mutations in the EDA gene, we first sequenced all eight exons coding for EDA in two affected males and two female carriers in Family A. We found a novel missense mutation c.947A>G in exon 9 of EDA, and found the mutation segregating with affected or carrier status in the other family members (). This mutation resulted in the non-synonymous substitution of aspartic acid for glycine at amino acid residue position 316 (p.D316G) in the TNF domain of EDA protein (). Exon 9 of EDA was also sequenced from 300 unrelated normal Chinese individuals with the same Han ethnic background, 150 females and 150 males, without detecting any guanine at the c.947 base position of EDA gene allele.
DNA sequencing chromatograms showing the mutations of affected members in Family A and Family B.
Though the phenotype of hypodontia in Family B was milder than Family A, the hypodontia in Family B was inherited as the same X-linked dominant trait as in Family A. Therefore we also selected EDA as a candidate gene in Family B and directly examined Family B for possible mutations in EDA. After mutation screening of eight exons of EDA gene in Family B, a c.1013C>T transition mutation of EDA gene was observed (). The c.1013C>T transition resulted in p.T338M substitution in the TNF domain of the EDA protein (). The c.1013C>T mutant allele was present in all affected males, and in affected and obligate carrier females, but in none of the other individuals in Family B. The c.1013C>T nucleotide substitution was also not found in any of 300 control individuals of the same ethnic background, which strongly suggests that this is the causative mutation in this family.
In both families, the male individuals bearing the mutant gene were associated with complete penetrance, however, in female heterozygotes incomplete penetrance was observed. In Family A, female carries IV:08, IV:09 and IV:10 showed hypodontia, but the other carriers displayed normal teeth (). Among female carriers in Family B, only individuals IV:27 and V:05 showed missing teeth, while the other female carriers showed normal teeth. The cause of this phenomenon is unclear; however, it may have resulted from the differential pattern of X-chromosome inactivation between the symptomatic carriers and the non-symptomatic carriers.
gene product is a type-II transmembrane protein with a small N-terminal intracellular domain followed by a larger C-terminal extracellular domain. The C-terminal extracellular domain contains a collagen-like repeat domain and a tumor necrosis factor (TNF) domain 
(). The TNF domain has been shown to form homotrimers which are believed to be required for receptor interactions 
. The HED-causing mutations in the TNF domain which affect the function of EDA have been previously analyzed: most mutations (His252Leu, Gly291Trp, Gly291Arg, Gly299Ser, Tyr320Cys, and Ala349Asp) are likely to affect the overall structure of EDA, and some mutations (Tyr343Cys, Ser374Arg, Thr378Pro, and Thr378Met) alter the receptor binding site 
. However, Asp316Gly and Thr338Met mutations have not been reported previously.
Structural analysis of EDA protein showed that D316 and T338 were located at two adjacent loops at the bottom of the TNF domain (D316 at loop CD and T338 at loop EF respectively) (). Unlike the above mentioned HED-causing mutations, both p.D316G and p.T338M mutations were a distance away from the receptor-binding site, and thus are not likely required for the receptor binding. To understand the structural effects of these two point mutations, the detailed molecular modeling analysis of the interactions between mutated residues and their surrounding residues was performed and is presented below.
Structures of wild type, D316G and T338M EDA.
Residue Asp316, which is located at loop CD, interacted with the residues including Ser335, Ile336 and Thr338 in the adjacent loop EF via van der Waals force or hydrogen bonds. In addition, the side chain of Asp316 extended and formed salt bridges or hydrogen bonds with Lys340 or Asn342 from the neighboring monomer, thus taking part in subunit-subunit interactions (). Replacing this residue by Gly abolished the contacts between the side chain of Asp316 and its surrounding residues and its neighboring monomer (). This could reduce the inter-subunit interactions in this region and further affect the stability of the trimer. The mutation to Gly also decreased the hydrophilicity and negative charges of this site and may increase the flexibility of this region. Considering its role in interactions between loops, this Gly mutation may also affect the fluctuation of the adjacent loop EF.
Thr338, located in loop EF, could interact with the residues from loop CD, forming a hydrogen bond with the carbonyl oxygen atom of Ile312. In addition, backbone and side chain atoms of Thr338 could interact with Asn313, Thr315 and Asp316 via van der Waals force or hydrogen bond (). The replacement of hydrophilic Thr with hydrophobic Met increased the hydrophobicity at this site. The absence of Thr338 in p.T338M mutant would cause a decrease in the stability of this region. Meanwhile, the large hydrophobic side chain of Met338 in the mutant might make hydrophobic and van der Waals interactions with Asn313 and Phe314 from neighbor loop CD and thus may contribute to the conformational rearrangement surrounding the residues. As a consequence, it could affect the conformation of Asp316 side chain and in turn disrupt the inter-subunit interactions (). These results by molecular modeling suggest that replacing hydrophobic residue at this site would not only affect the stability of loop EF but also affect the stability of the nearby loop CD.
In conclusion, this molecular modeling analysis suggests that although Asp316 and Thr338 are located at different sites in the primary sequence, they are spatially close and might interact with each other; the mutation in either site could affect the stability of its adjacent loop. It is possible that p.D316G and p.T338M might have similar impacts on the structure of EDA by interacting between each other. As strands E, F and C connected by CD and EF loops are directly involved in the inter-subunits interactions and Asp316 and its neighboring residues are directly involved in monomer-monomer interactions, it could be expected that significant variations occurring at these mutated sites would affect the stability of EDA trimer.