Results of genome-wide SNP genotyping showed that none of the known deafness genes co-segregated with the phenotype. In the family, five autozygous segments longer than 2 Mb were present on chromosomes 8, 15, 16, 19 and 21 (). These five autozygous regions include 382 annotated genes.
Five autozygous regions detected with Affymetrix 6.0 arrays in the family.
The exome sequencing experiment of one affected individual (IV-7) achieved the expected number of reads (87,586,240) and target coverage plus average read depth. Eighty seven million reads were generated which constitutes 8.6 gigabases of raw sequence. More than 95% of the reads mapped to the reference genome. Comparable with other labs 
, when measured at a minimum depth of 8×, 82% of the target region was covered with an average depth of 68× (). Likewise, when measured at 1× and 20× coverage, nearly 95% and 69% of the intended target was covered with an average depth of 68× and 66× respectively. In terms of variant calls, the MAQ predicted 99,374 SNPs and 5,420 indels.
The five autozygous regions from genome wide genotyping data on chromosomes 8, 15, 16, 19 and 21 were investigated using the results of the exome sequencing. For the five autozygous regions much variation on the read depth was observed. Using a minimum depth of 8 as a filter, the chromosome 8 region has high average read depth of 101× with 92% coverage compared to the chromosome 19 region which has low average read depth of 15× with 38% coverage (). There was a strong negative correlation (r
−0.96) between the percentage of gunanine cytosine (GC) bases and the read depth or coverage of autozygous regions (). We focused on exonic and flanking intronic variants within these five autozygous regions. shows the four novel homozygous missense, nonsense, splice site and frame shift variants (not reported in dbSNP132) in the five autozygous regions when we used a filter of minimum 8× read depth. Sanger sequencing confirmed one novel missense variant in the second longest autozygous region on chr8 (76476256A>T) in HNF4G
(MIM 605966) and one novel missense variant in the third longest autozygous region on chr19 (2917947C>T) in ZNF57
(no MIM number available) (nucelotide numbers are according to Hg19). We then recruited additional family members who were not typed with Affymetrix 6.0 chips to evaluate co-segregation of these variants. The novel variant c.1263A>T (p.Q421H) in HNF4G
did not co-segregate with the phenotype in the entire family but variant c.1328C>T (p.T443M) in ZNF57
did (). For the variant in ZNF57
PolyPhen2 classification was possibly damaging with a score of 0.938, MutPred predicted that T443M amino acid substitution caused a gain of catalytic residue at V439 (p
0.0472) and predicted the g score (probability of deleterious mutation) of 0.497 
. Four coding exons and intron-exon boundaries of ZNF57
were Sanger sequenced and no other nucleotide change was found. The indentified nucleotide change was not found in 335 Turkish controls via Sanger sequencing.
Novel missense, nonsense, splice site, and frameshift variants in top five autozygous regions.
ZNF57 is a recently discovered human zinc finger gene which has not been implicated in hearing loss. The ZNF57 protein product comprises 555 amino-acids with a KRAB-A domain at the amino-terminus and 13 tandemly arranged C2H2 zinc fingers at the carboxyl-terminus. Over expression of ZNF57 was shown to inhibit the transcriptional activities of NFAT and p21 demonstrating that ZNF57 is likely to function as a negative transcriptional regulator in NFAT-p21 signaling pathway 
. The variant p.T443M is located in the linker between zinc fingers 10 and 11. The wild type linker in ZNF57 (has the sequence TQEQL) and the canonical zinc finger linker sequence is TGEKP. Both linkers comprise five residues and have a conserved threonine at the first position. Threonine at this position is highly conserved and attains a ConSeq conservation score of 8 in a scale of 1 to 9 (where 9 is most conserved) 
. This conserved threonine is changed to methionine in the variant form of ZNF57 p.T443M. The functional unit for the zinc finger protein ZNF57 is unknown. Whilst zinc fingers are known to bind DNA, zinc fingers also interact directly with proteins 
and RNA 
and many have more than one role and form both protein-DNA and protein-protein interactions 
The full-length ZNF57 sequence was BLASTed against the sequences of the PDB. A designed zinc finger peptide with six zinc fingers known as Aart 
had the highest percent sequence identity with the query. The query sequence ZNF57 (with only the four zinc fingers 9, 10, 11 and 12) was aligned with the structural template Aart (PDB Code: 2I13 – chain B) using the Accelrys global alignment program. The alignment comprises 144 topologically equivalent position sharing 47.2% sequence identity (ZNF57:362–505 and 2I13:152–295). Indels were absent in the alignment. Protein structural models for ZNF57 (the wild type and p.T443M) were built (). Threonine 443 of the wild type is the last residue at the C-terminal end of the α-helix and likely contributes to the α-helix cap (). The methionine side chain is longer, more flexible and is unbranched compared with threonine (). Mutation of threonine to methionine is likely to affect DNA binding capability indirectly in several different ways 
. Threonine is capable of being phosphorylated whereas methionine is not. Phosphorylation and dephosphorylation is implicated in the regulation of zinc finger protein binding and function and the conserved threonine in the linker region is a prime candidate for this type of regulation. Threonine is on the surface and accessible to possible phosphorylation events 
. Others suggest that the DNA-induced helix capping in the conserved linker sequence is a determinant of binding affinity in C2H2 zinc fingers 
. In evolution, threonine is one of the most frequently observed amino acids at this position in the zinc finger domain topology 
. In mutational studies of the linkers between two contiguous zinc fingers, mutation of threonine to alanine had deleterious effects on DNA binding 
. Similarly, mutating threonine to leucine in the linker was shown to reduce DNA binding 
Molecular modeling of p.T443M in ZNF57.
While this work was ongoing a missense mutation in Gipc3
was reported to be associated with age-related sensorineural hearing loss in the mouse, and two missense variants in GIPC3
(MIM 608792) in two small families with sensorineural hearing loss 
. The autozygous region on chromosome 19 in our family also includes GIPC3
, in which no novel variant had passed our filters. We then re-analyzed the exome sequencing data reducing the filter for read depth to ≥4× instead of ≥8×; two additional variants in autozygous regions were detected () and only the variant c.508C>A (p.H170N) in exon 3 of GIPC3
was confirmed by Sanger sequencing (). Read depth for this variant was 5 and the exon containing this variant was poorly covered (). PolyPhen2 classification for this variant was probably damaging (score 1.0) and MutPred predicted the g score (probability of deleterious mutation) as 0.850. A ConSeq conservation score of 9 is obtained for H170 showing that this residue is highly conserved. The mutation was absent in 335 healthy ethnicity-matched controls.
The GIPC3 sequences (wild type and p.H170N) were each aligned with the PDZ domain in GIPC2 (PDB code: 3GGE - chain B). The alignment (GIPC3:108–199 and 3GGE:3–95) comprises 92 positions sharing 26.6% sequence identity with no indels. The structural template (3GGE) and the models each have two α-helices and six β-strands. Protein structural models for GIPC3 (the wild type and p.H170N) were built (). The mutated form of GIPC3 was compared with the wild type, structural differences were observed. In the mutated form of the model, the substrate molecular recognition pocket was larger and the associated charge distribution was reduced () compared with the wild type. In the wild type H170 side chain is pointing away from the core and the resulting side chain is solvent accessible whilst the asparagine side chain 170 in the mutated form of GIPC3 points inwards towards the hydrophobic core and forms a tight network of H-bonds. The asparagine side chain forms two side chain H-bonds with two main chain atoms (ASP 128 NH: ASN 170 OD1 and ASN 170 OD2-HD22:THR 127 O) which renders the side chain solvent inaccessible. In addition, N170 forms two main chain to main chain H bonds (ALA 174: ASN 170 O and VAL 173 N:ASN 170 O). Residue 170 is the first residue of α-helix 2. In the wild type the side chain solvent accessibility for residue H170 is greater than 10%. H170 does not form side chain to side chain H-bonds within the PDZ domain (). Whilst the accepted amino acid residue substitution profile is variable at position 170 across the PDZ superfamily 
, histidine for the GIPC family members at this position is invariant. This position 170 coincides with a key ligand binding pocket 
and the mutation from histidine to asparagine is predicted to alter the ligand binding site with a change of substrate specificity resulting in an adverse alteration in the protein function.
Diagrams of structural models for GIPC3.