Profound hearing loss in a porcine model
Albino pigs (Fig. 1a) spontaneously arising from a native breed of swine in Southwest China (Rongchang pigs [17]) are well-studied for their observed phenotypes of deafness and depigmentation, similar to the phenotype of WS2 [11, 12]. Results from auditory brainstem response (ABR) tests show that the albino pigs produced no recognizable waveforms up to 100 dB sound pressure level (SPL) stimuli in the range from 4–32 kHz, whereas normal pigs produced ABR thresholds at 5–10 dB SPL (Fig. 1b). Loss of hair cells and stereocilia bundles were observed in the cochleae of the albino pigs by scanning electron microscopy (SEM; Fig. 1c). Because the hearing loss observed in human cases of WS2 is attributed to abnormal cochlear stria vascularis (SV) [18], the morphology of SV was examined in our study’s albino pigs using light microscopy and transmission electron microscopy (TEM). As shown in Fig. 1d and e, the albino pigs lacked intermediate cells and had thinner SVs, consisting of two layers of cells only. Because the major functions of the SV are secretion of potassium ions and production of endolymphatic potential (EP), we recorded EPs and measured the [K+] in the scala media of the cochlea. The EP and [K+] were significantly lower than those of normal pigs (P < 0.001, Student’s t-test; Fig. 1f, g). Since EP and high [K+] in the endolymph are reportedly the driving force for mechanotransduction in cochlear hair cells [19], a reduction in EP can lead to profound hearing loss. All these phenotypes were tested at postnatal day 13. Thus, these results confirmed the phenotype of profound hearing loss related to cochlear morphology defects in our study’s albino pigs. Considering that eye defects have also been observed in some WS2 patients, we also assessed the morphology of porcine eyes. The irises of the albino pigs presented with pale coloration due to lack of pigmentation (Additional file 1: Figure S1A). The paraffin-embedded sections of retinae showed hypopigmentation in the choroid, but the retinal pigment epithelium was normal in both the normal and albino pigs (Additional file 1: Figure S1B).
Gene mapping and mutation screening
To investigate whether the hearing loss in our study’s albino pigs was caused by genetic factors, a genetic analysis was performed with the aim of identifying the hereditary pattern of the deafness. The deafness incidence rate in offspring of albino × albino mating was 100 %, whereas almost all offspring of normal × albino mating yielded normal hearing offspring. Moreover, there was no significant difference in the prevalence of deafness between males and females. These results implied an autosomal recessive inheritance pattern; to confirm this, 11 pairs of putative heterozygous boars and sows were selected from the normal herd according to the selection criteria of having produced at least one albino offspring. In total, 74 piglets from 11 litters of heterozygous × heterozygous matings were phenotyped to investigate the Mendelian segregation ratio; the results were 25 piglets with hearing loss and 49 piglets with normal hearing. Results of χ2 goodness-of-fit test indicated that the segregating ratio of the hearing loss trait was 3:1 (P < 0.01; Additional file 2: Table S1), confirming that the trait’s inheritance mode was autosomal recessive. This result led us to hypothesize that all of the albino pigs had inherited a mutant allele (r allele) from a common ancestor, instead of the wild-type allele (R allele) of normal pigs. We then applied a whole genome association approach with a phenotype-segregated population (Additional file 1: Figure S2). The strongest association signals were detected at two markers (ASGA0057578 and ALGA0070138, P
genome = 0.00242; Fig. 2a) on Sus scrofa chromosome 13 (SSC 13). Then, by using haplotype association analysis we detected strong concordance of a haplotype block that was composed of five markers, with the hearing loss phenotype in the mapping population (Fig. 2b). All homozygotes of the “GGGGA” haplotype were hearing impaired and the homozygotes of the “AAAAG” haplotype had normal hearing (P
raw = 3.72 × 10–12; chromosome-wide significance, P
chr = 1.74 × 10–5, 25,000 permutations; Fig. 2b). Among the heterozygotes, a small percentage of the pigs (4/51) were hearing impaired, indicating an autosomal semi-recessive transmission mode for hearing loss. Based on these results, we mapped the causative mutant gene of hearing loss to a 763 kb interval (SSC 13: 56,170,062 to 56,933,573), which was defined by the haplotype block and the proximal recombinant markers (Fig. 2c). The melanogenesis- and hearing-related gene Mitf was the only annotated gene located in this sequence interval (Fig. 2c). A search of the literature determined that previous studies had associated mutations in Mitf with auditory-pigmentary syndrome in humans, mice, cattle, horses and dogs [20–24], all of which have reported phenotypes similar to those of the albino pigs used in our study. Based on these results we speculated that Mitf was likely the causative mutant gene of hearing loss in these albino pigs.
To achieve more fine mapping of the causative mutation, we carried out gene screening for 12 mutant (Mitf
r/r) and 12 wild-type (Mitf
R/R) pigs. The porcine Mitf gene was not completely annotated in the reference genome (Sus scrofa genome 10.2) due to the poor assembly quality in this region. Therefore, we performed a homology annotation using the human Mitf mRNA from the Reference Sequence (RefSeq) database and the scaffold (TP_scaffold_24421) from our recently reported genome of Tibetan wild boars [25]. We found that 15 exons, spanning 243 kb of consecutive sequence, had a high syntenic relationship with the Mitf genes in both human and mouse (Additional file 2: Table S2). We next amplified and sequenced all exons, exon-intron boundaries and proximal promoters of the full Mitf gene, out to 10 kb upstream of the transcription start site of the encoded MITF-M isoform. In total, 21 co-segregated variants were identified, including 16 SNPs, four insertions, and one deletion (Additional file 2: Table S3). A C > A non-synonymous variant, which induced a N106K transition, was detected in the protein-coding region (exon 3). However, because only the Mitf
R/R pigs carried this variant, it seemed unlikely to be correlated with the hearing loss phenotype. The other 20 variants were located in regulatory regions.
As the associated region we discovered was too large for complete screening, the causative mutation could have been missed due to incomplete coverage by Sanger sequencing. To help rule out this possibility, we used whole-genome re-sequencing data for three of the Mitf
r/r pigs and three of the Mitf
R/R pigs (Additional file 2: Table S4). A total of 1711 SNPs were detected in the associated region (SSC13: 56,170,062 to 56,933,573; Additional file 3: Table S5) and 961 of these co-segregated with the hearing loss phenotype, including 362 ambiguous SNPs with missing data (Additional file 3: Table S6). Furthermore, 103 open-access porcine whole genome re-sequencing data sets (Additional file 2: Table S4) were included in the analysis to filter out common variants, which are not expected to be related to the hearing loss phenotype. Because this phenotype has not been reported in any other pig breed beyond the Rongchang breed, the 103 pigs from the open-access database were regarded as wild-type pigs (MITF
R/R). Next, 946 of the abovementioned 961 co-segregated SNPs were sought in the 103 MITF
R/R pigs (Additional file 3: Table S6); only 15 of the co-segregated SNPs were identified as carried exclusively by the Mitf
r/r pigs (Additional file 3: Table S7). Additionally, nine of those 21 co-segregated variants detected in the mutation screening noted above were excluded due to their existence in any Mitf
R/R pigs (marked with bold text in Additional file 2: Table S3). Combining the data from our mutation screening and re-sequencing analysis provided a total of 26 co-segregated variants that were deemed as candidate mutations for further study (Additional file 4: Table S8).
Mitf expression analysis
Because an association assay was incapable of further identifying the causative mutation, we next investigated the differential expression of Mitf between the mutant and wild-type pigs. In humans, at least seven transcript variants (encoding seven isoforms) with the same number of promoters and first exons have been identified [15]; in pigs, the Mitf gene has not yet been fully characterized and only one transcript has been identified [26]. Thus, the techniques of reverse transcription-PCR and 5’-rapid amplification of cDNA ends (commonly known as 5’-RACE) were used to investigate the transcript variants in the porcine Mitf gene. The transcript variants of Mitf-m, Mitf-a and Mitf-h were identified in the cDNA from the porcine inner ear (Fig. 3a).
Moreover, we found that the Mitf-m transcript was normally expressed in Mitf
R/R cochlea, but not in Mitf
r/r cochlea at any developmental stage (Fig. 3b). No obvious differences were found in the expression levels of Mitf-a or Mitf-h between the Mitf
R/R and Mitf
r/r cochlea (Fig. 3b). RNA-seq assay was used to detect differences between the transcriptome profiles of the Mitf
R/r and Mitf
r/r SVs at embryonic day 85; at this time in the development, melanocytes remained in the Mitf
r/r SVs (Additional file 1: Figure S3). A total of 28 genes showed two-fold differential expression, including some genes with known functions related to pigmentation and melanogenesis, but the Mitf gene was not among them (Additional file 4: Table S9). Because the algorithm used for calculating the gene RPKM values (reads per kilobase per million mapped reads, which serve to estimate gene expression) cannot separate the expression of Mitf-m from other equally expressed transcript variants, we obtained the normalized read count for each exon individually, and found that expression of the M-exon in wild-type SVs was approximately 11.5-fold higher than in mutant SVs (Fig. 3d). In addition, RNA-seq data indicated that expression of some marker genes of melanocytes or the SV intermediate cells remained in the MITF
r/r SVs; these genes included S100, KIT, MUM1, and Kir1.2 (Additional file 4: Table S10). Most of the genes expressed in MITF
r/r SVs, however, did not show significantly lower expression than the genes in the MITF
R/r SVs. When we considered these results along with those from TEM analysis of prenatal SVs (Additional file 1: Figure S3), we determined that the intermediate cells remained in the MITF
r/r SVs at the embryo stage and disappeared around birth. Thus, the expression differences between MITF
R/r and MITF
r/r SVs that we observed were indeed caused by Mitf mutation, rather than a lack of intermediate cells.
Subsequent immunoblotting assay showed an undetectable level of polypeptides of 55–70 kDa (the reported size range of the MITF-M isoform in melanoma [27]) in the Mitf
r/r cochleae (Fig. 3c). Consistent with the previous data, the levels of MITF-A and MITF-H detected by immunoblot (Fig. 3c) were similar between the Mitf
R/R and Mitf
r/r cochleae. These results indicated that the expression of MITF-M was eliminated in the Mitf
r/r pigs at both the transcript and protein levels, and this differential expression pattern itself indicated the existence of regulatory mutations in M isoform-specific regions. Consistently, eight of the 26 co-segregated variants were located in the M isoform-specific promoter (M-promoter; Additional file 4: Table S8).
Transcriptional activity analysis of the MITF M-promoter
To test whether the variants in the M-promoter were capable of altering transcriptional activity, a transient transfection assay was performed using mouse B16 melanoma cells. The luciferase reporter constructs contained varying lengths (7.8, 6.4, 5.2, 3.7 or 1.2 kb) of truncated M-promoter from the R and r alleles, as shown in Fig. 4a. The construct pGL3-r-7.8 k, which contained a 7852-bp promoter region of the r allele, exhibited significantly lower luciferase activity than the R allele construct (pGL3-R-7.8 k; Fig. 4a). There was no significant difference in activity between the pGL3-r-7.8 k construct and the null construct (pGL3-Basic vector), the pGL3-R-6.4 k and pGL3-r-6.4 k constructs, or the pGL3-R-1.2 k and pGL3-r-1.2 k constructs (Fig. 4a). Together, these results suggested that the sequence variations involving the sequences between –7852 and –6416 bp, relative to the transcription start site of Mitf-m, were responsible for the elimination of MITF-M expression. Four co-segregated variants were located within that region, including two insertions (9 and 14 bp, respectively) and two continuous SNPs (Fig. 4b, c, red box). The mutations were densely clustered within a 96 bp region (Fig. 4a, b, black box). To further validate their effects on transcription, we knocked out the 96-bp fragment from the pGL3-r-7.8 k construct (named pGL3-r-7.8D) and detected restoration of the transcriptional activity (Fig. 4a). Using TFSEARCH [28], a transcription factor binding site searching tool, we predicted that the 9- and 14-bp insertions would create two putative binding sites for SOX family proteins (Fig. 4c, red underlined in red). As SOX proteins can act as suppressors of gene expression [29], we speculated that these insertions and the SNPs could be functional mutations.
To address whether these candidate mutations were able to alter the protein binding landscape in the involved sequence region, we used electrophoretic mobility shift assay (EMSA) to compare the binding capacity of the wild-type and mutant sequences. Two sets of oligonucleotides, R1 and r1 (Fig. 4c, red and yellow highlight, respectively), which differed only in the 9-bp insertion and the GC>TT replacement, and R2 and r2, which differed only in the 14-bp insertion (Fig. 4c, Additional file 4: Table S11), were incubated with nuclear extracts from mouse B16 melanoma cells. Only one differential complex (C1 in Fig. 4d) was formed, namely that with the r2 probe and lacking the R2 probe. The specificity of the complex was confirmed by competition EMSA, wherein a 50- and 100-fold molar excess of unlabeled r2 probe showed effective binding competition but a 50- and 2100-fold excess of unlabeled R2 probe did not. Another complex (C2 in Fig. 4d) formed was a non-specific complex, because no competition was observed with even 100-fold molar excess r2 cold probe. No difference in protein binding was observed between the R1 and r1 probes (Additional file 1: Figure S4). Collectively, these results show that only the 14-bp insertion can induce specific transcription factor binding events, while the 9-bp insertion and the two continuous SNPs do not. Thus, the 14-bp insertion is the only variation detected in our study that was considered as potentially responsible for the observed down-regulation of Mitf-m transcription activity.
SOX proteins (SOX2, SOX3 and SOX9) can regulate inner ear development [30–32]. Our RNA-seq data showed that Sox2 and Sox9, but not Sox3, were expressed in porcine SVs (Additional file 4: Table S12), suggesting that SOX2 and/or SOX9 is capable of binding ectopically to a new negative CRE (i.e. a silencer), which may have been generated by the 14-bp short insertion in the Mitf regulatory region and which may have caused the abrogation of Mitf-m expression (Fig. 6). Thus, the phenotypes of hearing loss and depigmentation in the albino pigs appear to be caused, at least partially, by the 14-bp insertion that is located at –7532 bp relative to the transcription start site of Mitf-m. Additionally, we performed a further genotyping analysis of the causative mutant in 311 individual Rongchang pigs (Additional file 1: Figure S5). Both the 14- and 9-bp insertions were found to be carried by all albino pigs, and to co-segregate with the albino phenotype completely (Additional file 4: Table S13). No recombination event involving the two insertions was observed.
These results provide convincing evidence that even a small insertion in a region that lacks regulatory activity in the melanocyte lineage can create a transcription factor binding site (TFBS). Therefore, we next investigated whether this region was a non-regulatory sequence in other cell/species lineages. By exploiting the available data from the human ENCODE project, which had previously identified large numbers of regulatory elements, we were able to obtain the human ortholog of the mutant region identified in pigs and investigate whether this region overlapped with the CREs reported in the ENCODE data (Additional file 1: Figure S6A). The published RNA-seq data indicated that the wild-type M-exon was indeed transcribed in the human melanocyte lineage (Additional file 1: Figure S6C).
Using the UCSC browser to search the ENCODE data, we noted the following characteristics of these putative CREs. (1) The flanking regions of the causative mutant point has a relatively low level of conservation in mammals, suggesting a low probability of conserved functional CREs (Additional file 1: Figure S6B). (2) DNAse I hypersensitivity data indicate that chromatin accessibility around the mutant point is low in melanocytes and various cell lineages (Additional file 1: Figure S6D). (3) Chromatin immunoprecipitation and DNAse I footprinting data provide no evidence of protein binding sites near the mutant point (Additional file 1: Figure S6E). (4) The H3K4Me1 and H3K27Ac histone markers show no evidence of CREs in the flanking region of the mutant point (Additional file 1: Figure S6F). Because none of these data supported the existence of CREs near the mutant point, we concluded that the 14-bp insertion found in the albino pigs resulted in the de novo genesis of a silencer in the M-promoter (Fig. 6).
Mitf-m-specific mutations result in hearing loss in a mouse model
To investigate whether a loss-of-function mutation in Mitf-m is sufficient to cause the deafness and depigmentation phenotypes, we constructed a mouse model with null Mitf-m alleles (Mitf
mi-ΔM/mi-ΔM mouse; Fig. 5a). A quantitative PCR assay confirmed that the expression of Mitf transcriptional variants in the Mitf
mi-ΔM/mi-ΔM-targeted mouse was similar to those detected in the albino pigs (Fig. 5d). The Mitf
mi-ΔM/mi-ΔM mice displayed profound hypopigmentation, with white hair and skin (Fig. 5b). ABR testing also revealed profound hearing loss in the Mitf
mi-ΔM/mi-ΔM mice (n = 10; Fig. 5c). Finally, the Mitf
mi-ΔM/mi-ΔM mice showed thinner SVs, compared to the wild-type mice, and fused or missing stereocilias of hair cells (Fig. 5e, f), similar to the cochlear morphology seen in the albino pigs. Together, these data demonstrate that an exclusive malfunction in the m transcript isoform is sufficient to cause the auditory-pigmentary phenotypes.