Introduction

Mutations in MARVELD2 cause autosomal recessive non-syndromic deafness at the DFNB49 locus (Riazuddin et al. 2006). Seven different pathogenic variants of human MARVELD2 have been identified in families segregating moderate to profound hearing loss, with no obvious vestibular phenotype (Chishti et al. 2008; Riazuddin et al. 2006; Safka Brozkova et al. 2012; Yang et al. 2013). MARVELD2 encodes tricellulin, a tricellular tight junction protein (Ikenouchi et al. 2005; Riazuddin et al. 2006). Tight junctions are a type of apical junctional complex which occur between epithelial cells. Bicellular tight junctions are formed when two cells meet and tricellular tight junctions are formed where three cells converge (Tsukita et al. 2008; Yap et al. 1998). Tight junctions regulate ionic and other molecular traffic via the paracellular route or through the intercellular spaces between epithelial cells (Ikenouchi et al. 2005; Schneeberger and Lynch 2004; Steed et al. 2010; Tsukita et al. 2001). In the inner ear, epithelial cells delimit the fluid-filled compartments and the tight junctions between adjacent cells prevent intermixing of the extracellular fluids (Gulley and Reese 1976; Jahnke 1975). Several tight junction proteins are found in the inner ear epithelia and variants in many of these have been associated with hearing loss in humans and mouse mutants (Ben-Yosef et al. 2003; Borck et al. 2011; Citi et al. 1988; Gow et al. 2004; Higashi et al. 2013; Kitajiri et al. 2004a, b; Lee et al. 2012; Morozko et al. 2014; Nakano et al. 2009; Nayak et al. 2013; Raphael and Altschuler 1991; Riazuddin et al. 2006; Wilcox et al. 2001).

Mice with a Marveld2 p.Arg497* knock-in variant display hearing loss by postnatal day 16 (P16) and are profoundly deaf by P30 (Nayak et al. 2013). These mice lose their cochlear hair cells in a temporal pattern that coincides with the observed hearing loss. Similar to the affected human subjects, these mice have no vestibular phenotype and have an apparently normal stria vascularis. Intriguingly, the tricellulin mutant mice also showed phenotypic changes in the heart, thyroid, mandibular salivary glands and olfactory epithelium (Nayak et al. 2013). Besides hearing loss, the overall body weight and the weights of several other organs were higher in the knock-in mice, along with altered histology of the granular ducts of the mandibular salivary glands, rostral olfactory epithelium and cardiac myocytes (Nayak et al. 2013).

Given the wide distribution of tricellulin in human epithelial cells, one would have predicted that the hearing loss would have been accompanied by multiple other disorders and deafness would be part of a syndrome rather than a non-syndromic hearing loss (Chishti et al. 2008; Riazuddin et al. 2006; Safka Brozkova et al. 2012) as we initially reported. The goal of this study was to determine the spectrum of mutant alleles, and revisit families and obtain a more comprehensive clinical evaluation of DFNB49 families from Pakistan and Slovakia to determine if individuals homozygous for MARVELD2 mutation also have other clinical symptoms in addition to the hearing loss, similar to that observed in the tricellulin knock-in mice.

Materials and methods

Family participation

This study was approved by IRBs at the National Centre of Excellence in Molecular Biology (NCEMB), Lahore, Pakistan, at the National Institutes of Health, USA (Combined Neuroscience IRB; OH-93-N-016), at the University of Maryland, USA and at the University Hospital in Bratislava, Slovakia. Written informed consent was obtained from adult subjects and parents of minor subjects. The study participants comprised of twelve DFNB49 families, which were identified through linkage studies of 800 families segregating moderate to profound, prelingual- or congenital-onset, recessive deafness ascertained through the NCEMB including the original 550 families screened by Riazuddin et al. (2006).

Clinical evaluation

We obtained a detailed clinical history for each affected individual of the DFNB49 families to determine the neonatal course, any developmental delays, and signs of polyuria, polydipsia, goiter and hypotonia. For all affected individuals, we also performed a physical examination to evaluate motor skills, body weight, stature, and morphology. Hearing was evaluated on few participating individuals in audiology clinics by pure tone audiometry at octave frequencies with intensities up to 110 dBHL. Vestibular function was evaluated by tandem gait and Romberg testing (Khasnis and Gokula 2003). In seven subjects from family PKDF1069 and five from the Slovak family, we determined serum and urinary chemistry values to define aspects of metabolic and renal function. We also performed complete blood count to determine the hematopoietic abnormalities. Electrocardiogram and echocardiograms were obtained to determine the structure and function of the heart. Plain and contrast computed tomography (CT) scans of the brain were obtained for assessment of brain lesion and bony abnormalities including temporal bone, pituitary gland and peripheral nervous system.

Genotype and mutational analysis

Genomic DNA was extracted from 10 ml of peripheral venous blood as described (Grimberg et al. 1989). Three fluorescently labeled microsatellite markers (D5S2036, D5S637 and D5S1982) linked to chromosome 5q13.2 were PCR genotyped as described (Ahmed et al. 2001). Primers for polymerase chain reaction (PCR) amplification and MARVELD2 sequencing were described previously (Riazuddin et al. 2006). For sequencing, 15 μl of PCR product was treated with 0.3 U of shrimp alkaline phosphatase and 3 U of exonuclease I at 37 °C for 1 h, followed by incubation at 80 °C for 15 min. DNA sequencing reactions contained 3.2 pmol of primer, 0.6 ml of Big Dye Terminator Ready Reaction Mix (Applied Biosystems, CA, USA), and 2 μl of 5× dilution buffer (400 mM Tris–HCl pH 9 and 10 mM MgCl2) in a 10 μl reaction. Cycling conditions were 96 °C for 2 min and 45 cycles of 96 °C for 10 s, 55 °C for 10 s, and 60 °C for 4 min. Sequencing reaction products were ethanol precipitated and the pellets were resuspended in 10 μl of deionized formamide and sequence obtained using an ABI 3730 XL (Applied Biosystems).

To map the deletion breakpoints encompassing MARVELD2 exons 4 and 5 in family PKDF1069, we performed long-range PCR analysis using the DNA samples from the participating individuals of PKDF1069 family. For long-range PCR, the analysis reaction contained the kit recommended concentrations of primers, dNTPs, PCR buffer and LA-Taq polymerase (Clontech Laboratories Inc., CA, USA). For genotyping of deletion segregating in family PKDF1069, we used the forward primer (5′-CTGGCCTCAACCTCACTTCTT-3′) and reverse primer (5′-ACTATAGCAATGCCCTCGTTCA-3′). Cycling conditions were 96 °C for 2 min and 30 cycles of 96 °C for 30 s, 60 °C for 30 s, and 68 °C for 9 min. For sequencing of the long-range PCR products, we used the same parameters as described above.

Co-segregation of the mutations with hearing loss in each family was demonstrated for all subjects participating in this study. Control DNA samples from ethnically matched Pakistanis were sequenced for novel variant of MARVELD2. We also estimated the contribution of MARVELD2 mutations to recessive, severe to profound, congenital or prelingual-onset deafness using a sample size of 800 Pakistani families and 95 % confidence interval (CI).

Fluorescently tagged tricellulin expression constructs

A wild-type tricellulin construct (TRICWT) in pEGFP C2 vectors (Invitrogen) was a gift from Alan Fanning, PhD, Department of Cell Biology and Physiology, University of North Carolina, Chapel Hill, North Carolina, USA. The TRICC395−Q501del construct was made using the In-Fusion® HD cloning kit and the TRICWT cDNA as template.

Primer pairs were designed to amplify the N-terminal region encoding the transmembrane domain [Primer pair Inf-exon2F (5′-GGACTCAGATCTCGAGGATGTCAAATGATGGAAGATCCAG-3′) and Exon3R (5′-AATTCTCTCATGTTCCATTTTCCTTTTCGATGACAAT-3′)] fused to that encoding the C-terminal end [(Primer pair Exon6F (5′- TCGAAAAGGAAAATGGAACATGAGAGAATTTC AAGAA-3′) and Inf-exon7R (5′-GAAGCTTGAGCTCGAGTCAAGAATAACCTTGTACA TCCCAAT-3′)], such that the resulting construct lacked the region encoded by exons 4 and 5. The sequences of primers Exon3R and Exon6F are partially complimentary and are a fusion of exon 3 and exon 6, which results in the deletion of exons 4 and 5.

For the PCR reaction, the annealing temperature was 60 °C, with 90 s extension at 72 °C for primers Inf-exon2F and Exon3R and 30 s extension for primers Exon6F and Inf-exon7R. Meanwhile the pEGFP C2 vector was linearized by restriction enzyme XhoI (Fermentas) and dephosphorylated. Both PCR products and the linearized vector were gel-purified and ligated together in the recommended molar ratio, using the In-Fusion® HD enzyme as per the manufacturers’ protocol. The ligation reaction was used to transform Stellar competent cells (Clontech Laboratories) and resulting clones were sequence verified before using for transfection of cells.

Cell culture conditions and transfection

Madin-Darby Canine Kidney Epithelial (MDCK) cells were transiently transfected using Fugene-6 (Roche Applied Science, Indianapolis, IN, USA) with 1.5 μg of the desired construct per well in a 6-well dish. After transfection, cells were incubated for 48 h at 37 °C followed by fixation with 4 % paraformaldehyde. For visualization of the bicellular tight junctions, anti-ZO1 rabbit polyclonal (Life Technologies) antibody was used at 1:100 dilution. A Zeiss LSM700 confocal microscope was used for imaging and the digital images were adjusted in Photoshop 6.0 (Adobe) for optimal contrast and brightness. For quantification, the GFP fluorescence intensities at three tricellular and bicellular tight junctions per image were measured using ImageJ. The anti-ZO-1 signal intensity peak was used as the reference point to locate the tight junctions. The ratio of the average GFP fluorescence intensity at tricellular tight junctions to bicellular tight junctions was calculated for each image to determine if the protein was enriched specifically at either type of tight junctions. Statistical analyses were performed using the Student’s t test. All average data on the graph are shown as mean ± standard error of mean (SEM). P values less than 0.05 were considered significant.

Results

Mutation analysis of MARVELD2

We identified five families co-segregating deafness (Figs. 1, 2) with microsatellite markers linked to the DFNB49 locus (Riazuddin et al. 2006). The Slovak Roma family has two affected members who are homozygous for a pathogenic allele of MARVELD2 (Fig. 1). Sequence analysis of MARVELD2 revealed three probable homozygous pathogenic variants co-segregating with prelingual, severe to profound, sensorineural hearing loss in all five of these families (Figs. 1, 2). The previously reported splice junction variants (c.1331+2T>C and c.1183−1G>A) were found in four families (Fig. 1). Both these alleles are known to cause aberrant splicing of the MARVELD2 mRNA, resulting in premature stop codons (Riazuddin et al. 2006). In the affected individuals of another DFNB49 family (PKDF1069), several attempts at PCR amplification of exons 4 and 5 of MARVELD2 did not produce an amplimer, while we were able to amplify the neighboring exons 3 and 6. These observations prompted us to search for an intragenic deletion of MARVELD2 exons segregating in family PKDF1069. We performed long-range PCR to amplify 5–15 kb products using several combinations of primers from upstream of MARVELD2 exon 4 and downstream of exon 5 using DNA samples from the affected individuals and their normal hearing parents. Eventually, we identified an 8046 bp deletion encompassing both exons 4 and 5 of MARVELD2 segregating with hearing loss in family PKDF1069 (Fig. 2). The proximal breakpoint of this deletion is in intron 4 at chr5:68,726,959 (Human genome build GRCh37/hg19), while the distal breakpoint occurs at position chr5:68,735,005 located in intron 6 of MARVELD2 (Fig. 2c).

Fig. 1
figure 1

Four families from Pakistan and Slovakia with known DFNB49 mutations. Hearing loss in these families co-segregates with known DFNB49 mutations, c.1331+2T>C and c.1183−1G>A, except for the gray individual in family PKDF895, who also has prelingual profound hearing but is heterozygous for the MARVELD2 mutation, c.1331+2T>C, and does not share the disease haplotype with the other affected members of the family. Filled symbols indicate affected individuals, circles represent females and squares represent males

Full size image
Fig. 2
figure 2

Non-syndromic profound deafness co-segregates with a novel MARVELD2 deletion in family PKDF1069. a Genotypes of the participating family members are shown below each symbol. b Long-range PCR used to determine the presence of the allele with the deletion versus the wild-type allele. The wild-type allele gives a 8.7 kb band, while the allele with the in-frame deletion gives a 681 bp band. The labeling of the lanes corresponds to the participating individuals shown in a. c The genomic position of the deletion breakpoints. Underlined is the 18 bp sequence with 100 % nucleotide identity that flanks the deletion on both sides and is oriented in the same direction. d–f Confocal images of MDCK cells transfected with tricellulin wild-type and mutant constructs. d TRICWT is enriched at tricellular junctions in transfected MDCK cells. e TRICC395−A501del is found at both tricellular and bicellular junction regions and is no longer enriched at the former. It is also found in the basolateral domain of transfected MDCK cells. The intracellular GFP signal may represent misfolded TRICC395−Q501del protein that may be retained in the endoplasmic reticulum. f TRICR500X is also mislocalized and is no longer enriched at tricellular junctions. Scale bar 5 μm. g The fluorescence intensities of the GFP signal at bicellular and tricellular junctions were measured and the ratio of TJ/BJ was calculated for each construct. The wild-type protein is enriched at tricellular junction and this enrichment is lost or significantly reduced in the mutant proteins. ***p < 0.001

Full size image

To investigate the mechanisms involved in the deletion detected in family PKDF1069, the interval surrounding the breakpoints was analyzed through RepeatMasker (http://www.repeatmasker.org/). In addition, significantly over-represented motifs within ±15 bp of Gross Rearrangement Breakpoint Database (GRaBD) translocations breakpoints were also sought (Abeysinghe et al. 2003). Bioinformatic analysis revealed that the 5′ breakpoint lies within an AluSg repeat located in intron 4, whereas the 3′ breakpoint lies within an AluSx1 repeat located in intron 6. Alignment of 308 bp AluSg repeat and the 284 bp AluSx1 repeat showed their sequence identity to be 86 % (243 of 283 nucleotides). Notably, an 18 bp sequence with 100 % nucleotide identity and oriented in the same direction flanks the deletion on both sides (Fig. 2c), suggesting that this short repetitive element within highly similar Alu short interspersed nuclear elements (SINEs) may have served as a substrate for non-allelic homologous recombination. The loss of exons 4 and 5 would lead to an in-frame deletion of 107 amino acids at the protein level (p.Cys395-Gln501del; GenBank accession # NP_001033692), which comprises part of the occludin–ELL domain at the carboxy-end of tricellulin (Riazuddin et al. 2006).

Heterologous expression and targeting of mutant tricellulin with exon 4–5 deletion

We concluded from previous studies of over-expression of tricellulin in stably transfected MDCKII cells that the carboxy terminus of the protein is required for efficient transport and targeting to the basolateral membrane (Westphal et al. 2010). To test the effect of the MARVELD2 allele segregating in family PKDF1069, which is predicted to delete 107 amino acid in-frame from the cytoplasmic tail of tricellulin, MDCK cells were transfected with the plasmid constructs encoding the full-length wild-type (TRICWT), p.Arg500* (TRICR500X; as positive control) and p.Cys395-Gln501del (TRICC395−Q501del) mutant tricellulin, all fused to GFP at the N-terminus (Fig. 2d–g). In cells with higher expression levels of TRICWT, the protein was found at the basolateral plasma membrane and weakly at bicellular tight junctions but was enriched at tricellular tight junctions (Fig. 2d). TRICR500X localized to the basolateral membrane and infrequently to the bicellular tight junctions as discrete puncti (Nayak et al. 2013), but it was not observed at tricellular junctions (Fig. 2f). Similarly, TRICC395−Q501del localized to the basolateral membrane and was almost uniformly distributed along with ZO-1, but was not enriched at the tricellular junctions (Fig. 2e). The mislocalization of the tricellulin mutants is apparent when the ratio of the fluorescence intensities of GFP at the tricellular to bicellular junctions is calculated (Fig. 2g). TRICWT shows tricellular to bicellular junction localization ratio of over 2.5, indicating that wild-type full-length tricellulin preferentially, but not exclusively, localizes to tricellular junctions in comparison to bicellular junctions. The mutant proteins had near-equal fluorescence intensities at both types of tight junctions resulting in significantly lower ratios compared to TRICWT protein (p < 0.001), which suggests that the mutant proteins had lost their tricellular tight junction localization signal. Thus, our in vitro studies suggest that tricellulin with an in-frame deletion of 107 amino acids from the cytoplasmic tail is unlikely to target to tricellular tight junctions. Future work will focus on identifying the domain of tricellulin involved with tricellular tight junction targeting and the proteins, such as ILDR1 (Morozko et al. 2014), with which it interacts at this location in the cell.

Clinical characteristics of DFNB49 families

We performed several screening clinical tests to evaluate the physical, audiometric, biochemical, hematological and physiologic function of some of the affected individuals from DFNB49 families. The affected individuals of Pakistani family PKDF1069 had bilateral, moderate to profound sloping sensorineural hearing loss (Fig. 3). There is intra-familial variability seen in the severity of the hearing loss among the affected individuals of family PKDF1069 (Fig. 3). In addition to hearing loss, the affected individuals of the PKDF1069 family did not show signs or symptoms of other pathology that could constitute a syndromic, rather than a non-syndromic deafness. The affected members of the PKDF1069 were born from full-term pregnancies with reportedly normal birth weight and length. Serum biochemical analyses of these individuals, revealed normal liver, kidney and cardiac function (Tables 1, 2). The affected individuals also did not display problems with polydipsia, polyuria, muscle weakness and fatigue, growth or mental retardation or any evidence of blood disorder. Brain computed tomography (CT) scans or magnetic resonance imaging (MRI) of three affected individuals revealed normal brain, sinuses and orbits with no signs of bone destruction or structural abnormality, space occupying lesions or peripheral nervous system abnormalities (Table 3). Collectively, these clinical studies suggest that the deafness segregating in family PKDF1069 is non-syndromic.

Fig. 3
figure 3

Audiograms of the PKDF1069 members. a A non-carrier member (s) of the family shows completely normal hearing. b–d Affected members p, n and o that are homozygous for the deletion. These members show moderate to profound, bilateral sensorineural hearing loss

Full size image
Table 1 Clinical finding of PKDF1069 family
Full size table
Table 2 Clinical findings of PKDF1069 family
Full size table
Table 3 Brain computed tomography (CT) scan of PKDF1069 family members
Full size table

To determine if other non-Pakistani DFNB49 families also present with non-syndromic hearing loss, we performed a comprehensive clinical investigation of the Slovak Roma family (Tables 4, 5, 6, 7), segregating a splice site variant allele of MARVELD2 (Fig. 1). Similar to family PKDF1069, the affected individual of the Slovak family also did not show any other abnormality besides hearing loss and also had normal vestibular function. However, based on the non-specific abnormalities from the electrocardiogram and echocardiography results of this proband (the orientation of the electrical axis, increased trabeculation in the apical half of the left ventricle and the aortic valve anomalies noted) there could be syndromic cardio-pathology associated with the proband’s deafness (Table 6). But these specific cardiovascular changes were not seen in any of the evaluated controls or any of the other affected individuals (Table 1), suggesting that these borderline non-specific abnormal findings, observed in the proband from the Slovak family, are likely not associated with the DFNB49 allele. The other abnormalities (microtia, microdontia) found in several members of the Slovak Roma family did not co-segregate with hearing loss and the MARVELD2 genotype, so they likely have different but probably still genetic causes.

Table 4 Clinical finding of Slovak family
Full size table
Table 5 Clinical finding of Slovak family
Full size table
Table 6 Echocardiogram and electrocardiogram findings of the Slovak family
Full size table
Table 7 General physical findings of the Slovak family
Full size table

Discussion

Taken together with our previous study, mutations of MARVELD2 account for 1.5 % (12 of 800 families; 95 % CI 0.8–2.6) of recessive hearing loss in the NCEMB Pakistani study cohort (Riazuddin et al. 2006). To date, eight variants of MARVELD2 have been identified and all of them are clustered in the cytoplasmic tail of tricellulin (Chishti et al. 2008; Riazuddin et al. 2006; Safka Brozkova et al. 2012). Among the known alleles, c.1331+2T>C is reoccurring having been identified in eight Pakistani and three Czech Roma families (Chishti et al. 2008; Riazuddin et al. 2006; Safka Brozkova et al. 2012). It accounts for ~53 % of the DFNB49 hearing loss phenotype segregating in Pakistani families (Chishti et al. 2008; Riazuddin et al. 2006), and is the only allele reported so far in the Czech Roma population (Safka Brozkova et al. 2012). Thus, hierarchical mutation screening examining the c.1331+2T>C in deaf individuals would be a cost-effective approach to molecular diagnosis in both populations.

Recently, comprehensive phenotypic analysis of Marveld2 knock-in mice revealed a more widespread pathology in various body organs besides inner ear defects (Nayak et al. 2013). In contrast to mice, in the previous studies, the affected members of the DFNB49 families did not appear to have any other obvious relevant deficit besides prelingual, sensorineural hearing loss (Riazuddin et al. 2006; Chishti et al. 2008; Safka Brozkova et al. 2012). However, these human families were not assessed to the same extent as the phenotypic evaluation of the tricellulin mutant mice. We considered the possibility that the affected members of the DFNB49 families might have other, yet unreported disorders (Nayak et al. 2013). Therefore, we performed comprehensive clinical investigations of affected individuals of a new DFNB49 family, PKDF1069 and of one Slovak Roma family. However, analysis of liver, kidney, heart, brain, pancreas and hematological functional analyses did not reveal any functional abnormalities in these organs co-segregating with hearing loss among the affected individuals. We cannot rule out the possibility of late onset of other tissue abnormalities in affected individuals homozygous for pathogenic MARVELD2 alleles, as the oldest affected individual evaluated in this study is 22 years old.

There are several possibilities that might explain the discrepancy between the phenotype observed in humans and mice due to tricellulin deficiency. With the exception of the inner ear, there may be redundant tight junction complexes in humans that bypass the defects observed in Marveld2 mutant mice. Defects in tricellulin may be compensated by other tight junction proteins in humans but not rodents. Alternatively, the broader phenotypic changes in Marveld2 mutant mice may reflect signaling defects due to tricellulin deficiency, which have no functional consequence in humans.

An inter- and intra-familial variability in the severity of hearing loss due to recessive mutations of MARVELD2 has been reported in Pakistani families (Riazuddin et al. 2006) including this study. Severity of hearing thresholds in our DFNB49 families does not show any correlation with age of the affected individual (Fig. 3) (Riazuddin et al. 2006). It is possible that an environmental factor (for example, infections, drugs or loud noise) may have contributed to this inter- and intra-familial phenotypic variability in hearing thresholds. However, we cannot rule out the possibility of genetic modifiers modulating the severity of hearing loss phenotype, especially in individuals harboring the same MARVELD2 mutation (Riazuddin et al. 2006). Consistent with a genetic modifier as a possible explanation, we have observed that the Marveld2 (accession numbers DQ682659) knock-in allele (p.Arg497*) produces varying degrees of hearing loss when this allele is present in different congenic mouse strains (unpublished data).

Tricellulin is necessary to maintain fluid barriers in the inner ear and loss of this barrier in the cochlea leads to hearing loss (Nayak et al. 2013). Loss of tricellulin protein from the sensory epithelia in the inner ear resulted in ultrastructural abnormalities in the tricellular tight junctions, where defects were seen in the formation of the central element of the junctions as well as with the convergence of bicellular tight junction strands. It was proposed that tricellulin is necessary to connect the bicellular junctions with the tricellular tight junctions (Nayak et al. 2013). In this regard, it is important to note that the C-terminal region of tricellulin was shown to bind ZO-1, a bicellular tight junction protein (Riazuddin et al. 2006). It has been proposed that this interaction may allow proper incorporation of tricellulin in the tight junction assembly (Riazuddin et al. 2006). Accordingly, loss of the C-terminal region in the TRICR500X and TRICC395−Q501del mutants results in mislocalization of the proteins from the tricellular junctions and non-specific targeting of the mutant proteins to basolateral domains of transfected cells (Fig. 2e, f).

In summary, the hearing loss segregating in DFNB49 families due to recessive mutations of MARVELD2 is non-syndromic in nature, which is in contrast to the phenotype observed in Marveld2 knock-in mice (Nayak et al. 2013). This may reflect functional redundancy with other tight junction proteins in humans or differences in other genetic, stochastic and environmental factors.