GJB4 and GJC3 variants in non-syndromic hearing impairment in Ghana

Abstract

The contribution of GJB4 and GJC3 gene variants to hearing impairment in Africa has not yet been studied. Here, we investigated the contribution of these genes to autosomal recessive non-syndromic hearing impairment in Ghanaian children. Hearing-impaired children from 141 simplex and 59 multiplex families were enrolled from 11 schools for the deaf in Ghana. The coding regions of GJB4 and GJC3 were amplified, sequenced, and analyzed for the study participants previously found to be negative for GJB2 and GJB6 variants. Seven GJB4 and one GJC3 variants were identified. One out of the seven GJB4 variants was classified as likely pathogenic, while the others were either benign or synonymous. The likely pathogenic variant (p.Asn119Thr/rs190460237) was predicted to be likely associated with hearing impairment. We modeled the wild-type and mutant proteins of this variant (p.Asn119Thr) to evaluate the effect of the mutation on protein structure and ligand-binding properties. The mutant and not the wild type had the potential to bind N-Ethyl-5ʹ-Carboxamido Adenosine (DB03719) which was due to a slight structural change that was observed. No clinically relevant variant was identified in the GJC3 gene. We report for the first time a likely pathogenic GJB4 variant that may be associated with non-syndromic hearing impairment in Ghana; the finding will add to the body of evidence of the contribution of GJB4 to hearing impairment cases around the world.

Impact statement

Although connexins are known to be the major genetic factors associated with HI, only a few studies have investigated GJB4 and GJC3 variants among hearing-impaired patients. This study is the first to report GJB4 and GJC3 variants from an African HI cohort. We have demonstrated that GJB4 and GJC3 genes may not contribute significantly to HI in Ghana, hence these genes should not be considered for routine clinical screening in Ghana. However, it is important to study a larger population to determine the association of GJB4 and GJC3 variants with HI.

Introduction

Hearing impairment (HI), a disabling congenital disease, is globally known as one of the major age-standardized disabilities of life.^1,2 According to a World Health Organization (WHO) report in 2019, about 466 million people are estimated to be living with HI.³ A higher prevalence is recorded in sub-Saharan Africa (about 6 out of 1000 live births) compared to the developed countries (about 1 out of 1000 live births).⁴ Reports from different populations have shown that about 50% of congenital HI cases are of genetic origin^4,5 and about 80% of the genetic cases are non-syndromic.^6,7 The majority of all non-syndromic HI cases (nearly 80%) are inherited in the autosomal recessive fashion.^8,9 HI is genetically highly heterogeneous with over 140 genes identified to date¹⁰ but the contribution of gene variants to HI has not been equally investigated across global populations, with limited studies from Africa. Hence, there is a great scarcity in the representation of known pathogenic genetic variants of African ancestry. As a result, a recent study of pathogenic and likely pathogenic (PLP) autosomal recessive non-syndromic hearing impairment (ARNSHI) variants (selected from the ClinVar and Deafness Variation Databases with their frequencies from gnomAD database) estimated the prevalence of HI due to PLP as 5.2 per 100,000 individuals for Africans/African Americans, compared to a higher prevalence of 96.9 per 100,000 individuals for Ashkenazi Jews.¹¹ The knowledge deficit is likely hindering progress in understanding the mechanism of HI and ultimately affecting the development of therapeutic strategies, genetic diagnoses, prognosis and genetic counselling.¹¹

Connexin genes are the most frequently reported known HI genes associated with HI cases, particularly in populations of European and Asian ancestries.^12–14 Connexins are a family of gap junction proteins expressed in almost all human tissues and are involved in intercellular communication,^15,16 and mutations in connexin genes have been implicated in about 28 genetic diseases,¹⁷ with deafness and skin diseases as the most frequently associated condition.^16,18 Variations in the gene GJB2 are most frequently associated with non-syndromic hearing impairment (NSHI).^18,19 Similar to GJB2, GJB4 and GJC3 gene variants are associated with skin disorders;¹⁷ however, they are seldom associated with ARNSHI. Associations have been established previously between NSHI and GJB4 in Iran^20,21 and Taiwan,²² and between NSHI and GJC3 in Taiwan²² and India.²³ However, multiple evidence from independent populations is needed for the clinical validity of hearing impairment gene-disease pairs.¹⁰ Earlier studies investigating GJB4 mutations among hearing-impaired patients found missense variants such as p.R227W (c.679C>T), p.C169W (c.507C>T), and p.R151S (c.451C>A),^20–22 though the molecular mechanisms of the cause of deafness with respect to these variants were not well elucidated. However, it was suggested that these that these variants may be pathogenic since they were identified among patients and not control participants.^20–22 Interestingly, ClinVar and the Rat Genome Database contain GJB4 variants associated with autosomal non-syndromic deafness,^24,25 further supporting the pathogenicity of the gene. Moreover, GJB4 protein was found to be expressed in the cochlea of rats.²⁶ Similar to GJB4, some GJC3 variants (e.g. p.I90A/c.569T>A and c.781 + 62 G > A) were reported only among hearing-impaired individuals without any extensive molecular study on their pathogenicity.^22,23 There is therefore the need to interrogate GJB4 and GJC3 variants from other populations across the world and to study the molecular mechanisms of pathogenicity of these gene variants.

To date, only GJB2 and GJB6 contributions to NSHI have been systematically investigated in Ghana^27–30 and other parts of Africa.^31,32 There is no study from Africa on the role of GJB4 and GJC3 variants in HI. In this study, we investigated the contribution of GJB4 and GJC3 to NSHI in Ghana. We report for the first-time, variants in GJB4 and GJC3 genes in a Ghanaian HI cohort, and we have used in silico protein modeling approaches to explore the possible molecular mechanisms through which a likely pathogenic variant found in GJB4 could cause deafness.

Materials and methods

Ethics consideration

The set of ethical principles of the Declaration of Helsinki were adhered to in this study. Ethical approvals were sought and obtained from two ethics review boards: the Noguchi Memorial Institute for Medical Research Institutional Review Board (NMIMR-IRB CPN 006/16–17) and the University of Cape Town’s Faculty of Health Sciences’ Human Research Ethics Committee (HREC 104/2018). Prior to patient enrolment, the study was explained to each study participant in their native language and informed consent was confirmed by signature.

Study participants

The participants in this study were grouped into three categories: (1) isolated/non-familial simplex cases (n = 141) living with severe to profound HI with putative genetic cause of deafness; (2) multiplex/familial cases consisting of 59 individuals, each one selected from 59 families who had at least two affected family members with HI (Figure 1 and Figure S1); and (3) control participants (n = 47) randomly selected from a general Ghanaian population, with no personal and family history of HI. The medical records of the hearing-impaired students were evaluated to identify families with congenital HI. Both families and isolated cases were compatible with autosomal recessive inheritance, and each hearing-impaired participant was carefully examined and interviewed with a structured questionnaire to eliminate syndromic and environmental causes of HI as described previously.²⁷ All the study participants including the controls had been previously screened and were found to be negative for GJB2 and GJB6 gene variants (Figure 1).^27,30

Figure 1.

Flow chart of genetic screening of patients with GJC3 and GJB4 variants, and in silico analysis of GJB4 c.356A>C (p.Asn119Thr) variant.

Genetic analyses

DNA extraction

At the West African Centre for Cell Biology of Infectious Pathogens (WACCBIP), University of Ghana, Accra, Ghana, DNA were extracted from the blood samples collected from each participant using QIAamp DNA Blood Maxi Kit ® (Qiagen, USA).

Polymerase chain reaction and Sanger sequencing

The molecular analyses were conducted at the Division of Human Genetics, University of Cape Town. Previously published primers²⁰ specific for GJB4 exon 2 (F1B4: 5′-TCAATCGCACCAGCATTAAG-3′ and R1B4: 5′-GGGGGACCTGTTGATCTTATC-3′) and GJC3 exon 1 (F1C3: 5′-GCTCCCTCTGAAGGACAGTG-3′ and R1C3: 5′-GGGAGGAGATCATCAGGACA-3′) and GJC3 exon 2 (F2C3: 5′-TGGGTACGCACTGTGAAAAA-3′ and R2C3: 5′-AGCTCCTCCTTGGACAGGAT-3′) were used to amplify the coding regions of GJB4 and GJC3. The PCR amplicons were Sanger sequenced as described by Bosch et al. in 2014 using ABI 3130XL Genetic Analyzer® (Applied Biosystems, Foster City, CA).

Data analysis

The Sanger sequence data were cleaned and analyzed using FinchTV chromatogram viewer and Unipro UGENE Integrated Bioinformatics Tools.^33,34 Odd ratios were calculated to examine how strongly the identified variables are associated with the HI phenotype. We used Fisher’s exact test to determine if there is an association between the number of alleles obtained for each variant in different populations. P-values less than 0.05 were considered significant. We used the following online databases, genome browser, and predictive programs to predict the clinical significance of the identified gene variants: VarSome,³⁵ ClinVar,²⁵ Align GVGD (Align Grantham Variation/Grantham Deviation),^36,37 FATHMM (Functional Analysis Through Hidden Markov Models),^38–40 MutationAssessor,^41,42 MutationTaster,⁴³ MutPred2 (Mutation Prediction 2), PROVEAN (Protein Variation Effect Analyzer),^44–46 PolyPhen-2 (Polymorphism Phenotyping V-2),⁴⁷ SIFT (Sorting Intolerant From Tolerant),^48–51 EIGEN,⁵² MPV (pathogenicity of missense variants), PrimateAI⁵³ and InterVar⁵⁴ (Table S1).

In silico analysis of c.356A>C (p.Asn119Thr) variant

The “ab1” file (obtained from the ABI 3130XL Genetic Analyzer^®) of the sample with the GJB4 c.356A>C (p.Asn119Thr) variant was trimmed and edited using the SnapGene Viewer v5.0.6 (https://www.snapgene.com/). The resulting sequence was then saved as a FASTA file which was used to perform a BLASTx search in the non-redundant protein data bank (nrPDB) accessed via the NCBI BLAST web interface. Six hits were obtained from the BLASTx search, of which four were human proteins which were retrieved as .pdb files. Only the “A” chain of the PDB hits showed homology with GJB4 protein; hence, they were the only chains considered for further analysis. The “A” chains were retrieved as PDB files using the “indicate chain” command of PyMOL v1.8.4.0.⁵⁵ For each of the four templates (the retrieved “A” chains), the wild type and mutant proteins of the GJB4 c.356A>C (p.Asn119Thr) variant were modeled.

Modeler v9.0.3⁵⁶ was used to perform a template-based (TM) modeling of both the wild type and mutant proteins of the GJB4 c.356A>C (p.Asn119Thr) variant using two strategies; (i) single template-based modeling and (ii) multiple-template-based modeling (Figure 1). All the scripts used for the modeling were obtained from the Modeler web tutorial and changes were made where necessary.

Single template-based modeling

To identify the best template, the four templates were compared against each other using multiple sequence alignment and phylogenetic tree reconstruction. Pairwise alignment of the best template was performed with both the wild type and mutant proteins of the GJB4 c.356A>C (p.Asn119Thr) variant from which 50 models were built. The best model was selected based on the lowest DOPE (Discrete Optimized Protein Energy) score.⁵⁷

Multiple template-based modeling

A multiple sequence alignment (MSA) was performed for all the four templates followed by pairwise sequence alignment with the wild type and mutant proteins of the GJB4 c.356A>C (p.Asn119Thr) variant. Similar to the single template-based modeling, 50 models each of the wild type and mutant proteins were built and the top 10 models were selected based on the lowest DOPE score.⁵⁷ The best model was selected from the top 10 models based on the Ramachandran plot evaluation (RAMPAGE) and Z-score (ProsA-Web) (Figure 2).

Figure 2.

Chromatograms and multiple sequence alignment of GJB4 p.Asn119Thr variant. Chromatogram of Sanger sequence of (a) wild type and (b) mutant of GJB4 c.356A>C (p.Asn119Thr) variant. The position of the nucleotide change is highlighted in blue (c) Multiple sequence alignment of GJB4 protein in different species. Position 119 for the c.356A>C (p.Asn119Thr) variant is boxed.

Model refinement

The best models were run through the Galaxy server’s^58,59 refinement (Refine2) pipeline, which iteratively optimizes the initial structure using global and local operators as loop modeling and hybridization. The top-ranked model based on Galaxy energy, in combination with other parameters, was selected for virtual screening of possible ligands.

Virtual screening

To assess the possible effect of the GJB4 c.356A>C (Asn119Thr) variant on the binding property of the protein, virtual screening for ligands was performed using the Galaxy server’s Site algorithm. The algorithm predicts binding by comparing the distance between an amino acid residue and a ligand atom with the sum of their van de Waals radii + 0.5angstrom. Binding site residues are considered as those with a smaller difference in distance.

Results

Molecular analysis of GJB4 and GJC3

To identify GJB4 and GJC3 variants that may be associated with HI in Ghana, we investigated hearing-impaired patients identified to be negative for GJB2 and GJB6 gene variants, from both multiplex (n = 59/127 affected individuals from 59 unrelated families) and simplex (n = 141) unrelated families segregating ARNSHI (Figures 1 and S1). These patients were found to have severe to profound congenital HI and their clinical and demographic data were previously reported.²⁷ The GJB4 and GJC3 gene variants were identified in the hearing-impaired patients and were further examined among the control individuals not affected by HI (Table 1).

Table 1.

GJC3 and GJB4 variants found in hearing-impaired patients and control subjects from Ghana.

					Number of participants, n (%)
Gene	Variants	Protein change	Clinical significance	Geno-types	Multiplex family N = 59	Simplex family N = 141	Total affected (N = 200)	Controls N = 47	Odds ratio	P value
GJC3	c.490C>T(rs73405465)	p.Pro164Ser	Benign	GG	50 (84.75%)	124 (87.94%)	174 (87.00%)	41 (87.23%)
				GA	7 (11.86%)	17 (12.06%)	24 (12.00%)	6 (12.77%)	0.94	0.45
				AA	2 (3.39%)	0	2 (1.00%)	0	–	–
GJB4	c.611A>C (rs3738346)	p.Glu204Ala	Benign	AA	49 (83.05%)	106 (75.18%)	155 (77.50%)	38 (80.85%)
				AC	4(6.78%)	16 (11.35)	20 (10.00%)	6 (12.77%)	0.81	0.34
				CC	6(10.17%)	19 (13.47%)	25 (12.50%)	3 (6.38%)	2.04	0.13
GJB4	c.451C>A (rs78499418)	p.Arg151Ser	Benign	CC	47 (79.66%)	111 (78.72%)	158 (79.00%)	40 (85.11%)
				CA	8 (13.56%)	18 (12.77%)	26 (13.00%)	3 (6.38%)	2.19	0.11
				AA	4 (6.78%)	12 (8.51)	16 (8.00%)	4 (8.51%)	1.01	0.49
GJB4	c.516T>C (rs111693060)	p.Thr172=	Variant of uncertain significance	TT	56 (94.92%)	136 (96.45%)	192 (96.00%)	46 (97.87%)
				TC	0	3 (2.13%)	3 (1.50%)	1 (2.13%)	0.72	0.39
				CC	3 (5.08)	2 (1.42%)	5 (2.50%)	0	–	–
GJB4	c.369G>A (rs142843509)	p.Lys123=	Benign	GG	59 (100%)	139 (98.58%)	198 (99.00%)	46 (97.87%)
				GA	0	2 (1.42)	2 (1.00%)	1 (2.13%)	0.46	0.26
				AA	0	0	0	0	–	–
GJB4	c.356A>C(rs190460237)	p.Asn119Thr	Likely pathogenic	AA	59 (100%)	140 (99.29%)	199 (99.50%)	47 (100.00%)
				AC	0	0	0	0	–	–
				CC	0	1 (0.71%)	1 (0.50%)	0	–	–
GJB4	c.303C>G(rs138184343)	p.Arg101=	Synonymous	CC	55 (93.22%)	135 (95.74%)	190 (95.00%)	46 (97.87%)
				CG	1 (1.69%)	4 (2.84%)	5 (2.50%)	1 (2.13%)	1.21	0.43
				GG	3 (5.09%)	2 (1.42)	5 (2.50%)	0	–	–
GJB4	c.238C>T(rs114429815)	p.Gln80Ter	Benign	CC	59 (100%)	139 (98.58%)	198 (99.00)	45 (95.74%)
				CT	0	1 (0.71%)	1 (0.50%)	2 (4.26%)	0.11	0.039
				TT	0	1 (0.71%)	1 (0.50%)	0	–	–

The clinical significance and pathogenicity of the identified variants were predicted using 2 online databases and 12 predictive bioinformatic tools (Table S1). The sensitivity, accuracy, and specificity of these predictive tools vary based on the algorithms used.⁶⁰ It was, therefore, important to use a combination of predictive tools.⁶¹

Variants in GJC3

The molecular, clinical, and pathogenic evaluation of the variants identified in heterozygous state a GJC3 variant predicted as benign (p.Pro164Ser). Two familial cases were found to be homozygous for the same mutation (Table 1).

Variants in GJB4

Three GJB4 synonymous variants (p.Lys123=, p.Arg101=, and p.Thr172=) were identified in all three groups of samples. Of the three synonymous variants, p.Lys123= was classified as benign and p.Thr172= as a variant of uncertain significance. The GJB4 sequence analysis also identified one nonsense and two non-synonymous variants classified as benign (p.Gln80Ter, p.Arg151Ser, and p.Glu204Ala). An additional variant (p.Asn119Thr) was classified as likely pathogenic (Table 1). Although some predictive tools suggested that the GJB4 p.Glu204Ala variant was likely pathogenic, this was not supported by clinical gene variants-disease correlations with HI since the homozygous form of the variant was identified in both affected hearing-impaired (n = 25) and unaffected hearing control individuals (n = 3) with an odds ratio of 0.81. In addition, the minor allele frequency of GJB4 p.Glu204Ala within the global and African populations exceeds the threshold of 0.05, suggesting that it is not disease-causing (Table 2).

Table 2.

Differential allele frequencies of GJB4 and GJC3 variants in the global population.

				Our data			Allele frequency (Ensembl)
Gene	Variants	rs number	Allele	Cases	Control	P-value (cases vs. control)	Global	P-value (our cases vs. global)	Africa	P-value (our cases vs. Africa)	America	P-value (our cases vs. America)	East Asia	P-value (our cases vs. East Asia)	Europe	P-value (our cases vs. Europe)
GJC3	c.490C>T (p.Pro164Ser)	rs73405465	G	0.93	0.94	1.0000	0.98	0.0001	0.94	0.6441	0.99	0.0001	1.00	0.0001	1.00	0.0001
			A	0.07	0.06		0.02		0.06		0.01		0.00		0.00
GJB4	c.611A>C (p.Glu204Ala)	rs3738346	A	0.83	0.87	0.3548	0.89	0.0001	0.75	0.0023	0.89	0.0022	0.88	0.0075	0.99	0.0001
			C	0.17	0.13		0.11		0.25		0.11		0.12		0.01
GJB4	c.451C>A (p.Arg151Ser)	rs78499418	C	0.86	0.88	0.6197	0.96	0.0001	0.90	0.0134	0.92	0.0010	0.97	0.0001	1.00	0.0001
			A	0.14	0.12		0.04		0.10		0.08		0.03		0.00
GJB4	c.516T>C (p.Thr172=)	rs111693060	T	0.97	0.99	0.4863	0.99	0.0001	0.98	0.1931	1.00	0.0001	1.00	0.0001	1.00	0.0001
			C	0.03	0.01		0.01		0.02		0.00		0.00		0.00
GJB4	c.369G>A (p.Lys123=)	rs142843509	G	0.99	0.99	0.4699	1.00	0.0296	1.00	0.0532	1.00	0.1340	1.00	0.0806	1.00	0.0808
			A	0.01	0.01		<0.01		0.00		<0.01		0.00		0.00
GJB4	c.356A>C (p.Asn119Thr)	rs190460237	A	0.99	1.00	1.0000	1.00	0.0156	1.00	0.1366	1.00	0.1335	1.00	0.0806	1.00	0.0808
			C	0.01	0.00		<0.01		<0.01		0.00		0.00		0.00
	c.303C>G (p.Arg101=)	rs138184343	C	0.96	0.99	0.3282	1.00	0.0001	0.98	0.0302	1.00	0.0001	1.00	0.0001	1.00	0.0001
			G	0.04	0.01		<0.01		0.02		0.00		0.00		0.00
GJB4	c.238C>T (p.Gln80Ter)	rs114429815	C	0.99	0.98	0.2425	0.99	0.7962	0.96	0.0013	1.00	0.6744	1.00	0.0228	1.00	0.0229
			T	0.01	0.02		0.01		0.04		<0.01		0.00		0.00

The GJB4 p.Asn119Thr variant predicted as likely pathogenic was identified in only one hearing-impaired individual from a simplex family (Figure 2(a) and (b)). The sample from the participant in whom this variant was found was independently sequenced three times, with each run from a new PCR product. Furthermore, the GJB4 p.Asn119Thr variant had less than 0.01 allele frequency in the global and African populations, indicating that it is a rare variant (Table 2). Since the GJB4 p.Asn119Thr variant was predicted to be likely pathogenic, we examined it further using protein modeling approaches (Figures 3 to 5).

Figure 3.

Evaluation and validation of GJB4 protein models: ProsA web evaluation of (a) wildtype (p.Asn119=) and (b) mutant (p.Asn119Thr) proteins. Ramachandran plot of (c) wildtype (p.Asn119=) and (d) mutant (p.Asn119Thr) proteins. (e) Discrete optimized protein energy (DOPE) profile for wildtype (p.Asn119=) and mutant (p.Asn119Thr) proteins.

Figure 4.

Refinement of GJB4 protein models. Galaxy refinement of (a) wild type and (b) mutant c.356A>C (p.Asn119Thr) GJB4 protein models. The best-ranked models are highlighted with red rectangles. Refined and unrefined models of (c) wild type and (d) mutant GJB4 c.356A>C (p.Asn119Thr) GJB4 protein models.

Figure 5.

GJB4 mutant protein in complex with NEC. A LigPlus plot shows the interacting residues in detail.

Evolutional evaluation of amino acid at position 119 of GJB4 protein

Since the molecular analysis suggested GJB4 c.356A>C (p.Asn119Thr) as a likely pathogenic variant, a multiple sequence alignment was performed with GJB4 protein sequences from different species to investigate the evolutional conservation of the amino acid residue at position 119 of the protein (Figure 2). Asparagine (Asn) at position 119 was conserved among all the different species investigated suggesting that the residue is important for the protein’s function. It is worth mentioning that some of the amino acid residues around the asparagine 119 were not conserved among some of the species studied.

Modeling of wild type and mutant (c.356A>C (p.Asn119Thr)) GJB4 protein

We examined the possible molecular effect of the change in the conserved amino acid at position 119 of the protein by modeling and comparing the wild type and GJB4 c.356A>C (p.Asn119Thr) mutant proteins. Good quality models with DOPE scores of ∼ −26,500 were obtained from the modeling experiment from which the best models were selected. Multiple-template modeling performed better than the single-template modeling (Figure 1(e)). Both models were evaluated and found to be within the range of expected values for X-ray crystallography-determined and nuclear magnetic resonance (NMR)-determined proteins. Z-scores of −4.56 and −4.28 were obtained for wild type and mutant (c.356A>C (p.Asn119Thr)) GJB4 proteins (Figure 3(a) and (b)), respectively. In addition, more than 98% of the residues were observed to fall within favorable and allowed regions on the Ramachandran plot with highly favorable ProsA Z-scores for both models (Figure 3(c) and (d)).

The Galaxy refinement of the wild type and mutant GJB4 proteins produced 10 models, from which we selected the best-refined (Figure 4(a) and (b)). The model labeled “MODEL 1” appeared to be the overall best for the wild type, while the model “MODEL 7” appeared as the best-refined for the mutant (Figure 4(a) and (b)). Figure 3(a) and (b) shows the quality improvement of the selected refined models compared to the unrefined models.

The GJB4 c.356A>C (p.Asn119Thr) mutation slightly modifies the protein structure, which we can observe when the mutant protein is compared with the wild type protein. On the wild type protein, asparagine at position 119 forms part of a random coil, however the same position in the mutant model harboring a threonine residue forms a helix (Figure 4). There was, generally, a high degree of conservation of the extracellular E1 and E2 loops, as expected. Refinement further saw the modeling of two short helixes in the C-terminus, in regions of random coil expected for gap junction proteins (Figure 5).

Virtual screening

Connexins are characterized by four transmembrane helices that form the transmembrane pore and extracellular domains, which form two loops (E1 and E2) that help in cell–cell recognition and docking. These loops are mostly involved in protein–protein interactions, while residues on the alpha-helix transmembrane domains are involved in the process of small molecule shuttling. To the best of our knowledge, the GJB4 c.356A>C (p.Asn119Thr) mutation (rs190460237) has not been previously reported, hence we modeled the 3D structures of GJB4 wild type and mutant proteins which revealed subtle but fundamental differences that may have significant implications on the protein function. To assess the possible effect of these differences, we performed virtual screening for ligands using the Galaxy server’s Site algorithm. The virtual screening predicted four ligands and their corresponding binding sites for the wild type GJB4 (1KS, SNT, A8T, and SG8) and five ligands for the mutant GJB4 c.356A>C (p.Asn119Thr) (NEC, 1KS, SNT, A8T, and SG8) proteins. Although none of the ligands interacts with the position 119 residues of both the wild type and the mutant models, it appears that the residue change caused a perturbation in the protein structure that is significant enough to alter ligand binding (Figure 5).

Discussion

Mutations in connexin genes have been implicated in about 28 genetic diseases, with HI and skin disorders as the predominant cases.¹⁷ Although the GJC3 gene has been associated with NSHI with specific pathological alterations in the cochlea,^62,63 there are limited studies globally and especially from Africa. Unlike other epidermal disease-associated connexins, the role of GJB4 variants in NSHI is not well elucidated.⁶⁴ To the best of our knowledge, this is the first report on GJB4 and GJC3 variants in African hearing-impaired patients and will add to the current knowledge, as well as help refine gene-disease pairs and clinical validity curation.

Mouse models created with alterations in the GJC3 gene indicated that about 50% of homozygous GJC3 null mice had delayed maturation of hearing thresholds, high-frequency hearing loss, and were vulnerable to noise-induced hearing loss.⁶⁵ An earlier study, however, did not describe any significant difference between the phenotypes (including auditory brainstem response) of the GJC3 deficient and the wildtype control adult mice.⁶⁶ The authors stated that the gene might be functionally associated with other connexins such as connexin 32 and connexin 47 which suggested that it may not be independently associated with the HI phenotype. Our study identified p.Pro164Ser (c.490C>T/rs73405465) variant in the GJC3 gene of both hearing-impaired and hearing individuals in Ghana with a 0.94 odds ratio. The missense GJC3-p.Pro164Ser variant had a minor allele frequency of 0.064 in the African population⁶⁷ which is greater than the 0.050 thresholds for calling uncommon variants. Considering the odds ratio, minor allele frequency, and occurrence of the variant in control hearing participants, the GJC3-p.Pro164Ser variant may not be associated with HI. The GJC3 p.Pro164Ser variant had no record/phenotypic data in ClinVar²⁵ and Ensembl⁶⁷ and was labeled as benign, non-pathogenic, neutral, or polymorphism by the majority of predictive tools used (Table S1) as well as on the VarSome database,³⁵ further supporting its non-pathogenicity.

The expression pattern and contribution of GJB4 to HI remain unclear. A GJB4 deficient mouse model generated by replacing the coding region of GJB4 with a lacZ gene did not show any auditory abnormality when assessed by brain stem evoked potentials.⁶⁸ Interestingly, these mice did not show any skin abnormality, which made it difficult to interpret the role of GJB4 in humans; however, there have been some studies that investigated and detected GJB4 gene variants in deaf individuals.^20–22 In a rat study, GJB4 was found to be expressed in rat cochlea, suggesting its role in the hearing process. The present study identified synonymous GJB4 variants (p.Lys123=, p.Arg101=, and p.Thr172=) in both affected and control samples of which p.Lys123= and p.Thr172= were classified as benign and variant of uncertain significance, respectively. But these three variants had no effect on the resultant protein; hence, they may not be responsible for HI pathogenesis. We also identified GJB4 p.Arg151Ser and p.Gln80Ter variants previously predicted to be benign. There was no published data on the GJB4 p.Gln80Ter variant in hearing HI patients. Similar to our study results, GJB4 p.Arg151Ser was found in both HI patients and controls in Iran²⁰ suggesting that it may not be associated with the HI phenotype. The variant was associated with skin disorders and found in patients without hearing loss^69,70 hence confirming the above observation.

Similar to our findings, a Spanish study also identified the GJB4- p.Glu204Ala in hearing-impaired patients.⁶⁴ We found the variant in both control and affected samples which are consistent with findings from Iran²⁰; our findings suggest that there is no likely association between the GJB4 p.Glu204Ala variant and HI. The p.Asn119Thr variant may be of clinical significance since it was reported as “likely pathogenic,” according to InterVar and the majority of the predictive tools (Table S1). GJB4 p.Asn119Thr was predicted to be a variant of uncertain significance by VarSome.³⁵ According to the automated clinical interpretation of genetic variants by ACMG/AMP 2015 guideline,⁵⁴ the variant was found to fall within the categories of PM1, PM2, PP3, and BP1. This implies that the variant is located within a mutational hot spot or a well-established functional domain without benign variation (PM1), and absent from controls in the ESP, 1000Genomes, and ExAC databases with extremely low frequency if recessive (PM2) with multiple lines of computational evidence supporting a deleterious effect of the gene product (PP3).⁵⁴ A supporting evidence for benign status of a missense variant in a gene which when truncated are known to cause disease (BP1).⁵⁴ When the variant was analyzed for “Pathogenic variants Enriched Regions (PER) for genes and gene families” in the PER viewer,⁷¹ it was observed to fall within a region of pathogenic missense burden for both gene family-wise and gene-wise analyses (Figure S2). PER sources disease-associated missense variants from ClinVar and the Human Gene Mutation Database (HGMD), retaining only “pathogenic” and/or “likely pathogenic” variants in ClinVar, and variants with “high confidence” calls in HGMD, all in the GRCh37.p13/hg19 coordinate. Interestingly, GJB4 p.Asn119Thr (N_119) variant was observed to align with a GJB2 variant (E_120) which is associated with sensorineural hearing loss.⁷¹ Our study identified the variant in one patient with allele frequency less than 0.01 and none in the control population, but there was not enough evidence to conclude on its pathogenicity.

Analysis of in silico protein modeling revealed a striking difference between wildtype and mutant models of the p.Asn119Thr variant. The asparagine at position 119, which is on a cytoplasmic loop, forms random coils in the wild type model, whereas threonine in the same position forms a helix in the mutant model. It appears that the presence of Threonine at this position increases the overall propensity for a helix.

The ligand-binding property of the mutant p.Asn119Thr protein was slightly different from the wild type GJB4 protein. An extra ligand, N-Ethyl-5ʹ-Carboxamido Adenosine (NEC), was found to bind the GJB4 p.Asn119Thr mutant protein and not the wild type. NEC (DB03719) is a non-carcinogenic purine nucleoside, a cAMP/cGMP phosphodiesterase (PDE) inhibitor⁷² that doubles as a human adenosine A (2 A) receptor agonist.⁷³ PDE inhibitors are often used in the treatment of erectile dysfunction because of their adenosine A (2 A) receptor agonist role. Post-marketing and retrospective clinical trial analysis has shown that these PDE inhibitors have severe side effects such as hearing loss.⁷⁴ However, the above observation is inconclusive as there is no direct association established between hearing loss and PDE inhibitors.

Limitation of the study

The study identified a rare missense variant GJB4-p.Asn119Thr in a single hearing-impaired patient which makes it difficult to associate the variant to the hearing impairment phenotype. The pathogenicity of the variant was predicted using in silico predictive tools. Although these tools give a good prediction of the possible clinical effect of the variant which is very useful, they are not as accurate as functional assays. We therefore recommend the use of cell and animal models to confirm the pathogenicity of the GJB4-p.Asn119Thr variant.

Conclusions

In this study, only one possibly pathogenic GJB4 variant (p.Asn119Thr) was identified in a hearing-impaired patient. The protein modeling and virtual screening identified differences in the protein structure and binding properties of the mutant p.Asn119Thr GJB4 protein compared to the wild type. There is a need for functional studies and investigations from larger populations to elucidate the pathogenicity of the variant (GJB4-p.Asn119Thr) predicted as “likely pathogenic”. We did not identify any GJC3 variant of clinical significance in the study population. Hence, GJB4 and GJC3 variants were found not to be significant contributors to non-syndromic autosomal recessive hearing impairment in Ghana. We therefore recommend the used of modern genomic approaches to investigate the associated HI gene variants in the study participants.

Authors’ contributions

Conceptualization, AW, GAA, and SMA; methodology, SMA, and KKE, validation, AW, GAA, GKA, and OQ; formal analysis, SMA, AW, and KKE; resources, AW, GAA, GKA, and OQ; writing—original draft preparation, AW, SMA; writing—review and editing, SMA, KKE, OQ, GKA, GAA, and AW; supervision, AW, GAA, GKA, and OQ; funding acquisition, AW and GAA. All authors have read and agreed to the published version of the manuscript.

DECLARATION OF CONFLICTING INTERESTS

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article

Ethical approval

Ethical approvals were obtained from the Noguchi Memorial Institute for Medical Research Institutional Review Board (NMIMR-IRB CPN 006/16–17) and the University of Cape Town’s Faculty of Health Sciences’ Human Research Ethics Committee (HREC 104/2018).

FUNDING

This work was supported by funds from the World Bank African Centres of Excellence grant (ACE02-WACCBIP: Awandare) and a Developing Excellence in Leadership, Training and Science Initiative (DELTAS) Africa grant (DEL-15–007: Awandare). The DELTAS Africa Initiative is an independent funding scheme of the African Academy of Sciences (AAS)’s Alliance for Accelerating Excellence in Science in Africa (AESA) and supported by the New Partnership for Africa’s Development Planning and Coordinating Agency (NEPAD Agency) with funding from the Wellcome Trust (107755/Z/15/Z: to G.A.A. and A.W.) and the U.K. government; the National Institutes of Health (NIH), USA, grant number U01-HG-009716 to AW; and the African Academy of Science/Wellcome Trust, grant number H3A/18/001 to A.W. Samuel Mawuli Adadey is supported by WACCBIP DELTAS Ph.D. fellowship and Africa Regional International Staff/Student Exchange (ARISE) II mobility fund. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

ORCID iDs

Kevin K Esoh https://orcid.org/0000-0002-4024-5681

Osbourne Quaye https://orcid.org/0000-0002-0621-876X

Ambroise Wonkam https://orcid.org/0000-0003-1420-9051

SUPPLEMENTAL MATERIAL

Supplemental material for this article is available online.