Novel neuro-audiological findings and further evidence for TWNK involvement in Perrault syndrome

Two affected sisters from a Polish non-consanguineous family and their unaffected mother were available for the study. The proband was born at term by Cesarean section after an uneventful pregnancy with 3600 g of body weight and an Apgar score of 1/2/3, indicating severe birth depression. Her psychomotor development was normal until the third year of age, when the motor performance begun to deteriorate progressively and the electromyography (EMG) examination was abnormal (she is unable to run since the age of 15 years). Speech development was age appropriate. At 5 years of age SNHL was diagnosed and pathologic auditory brainstem responses (ABRs) were observed. She has been fitted with hearing aids at the age of 13 but they were of limited benefit. Testing for primary amenorrhea and delayed pubertal development at the age of 15 revealed hypergonadotropic hypogonadism, streak gonads, rudimentary uterus and a normal female karyotype 46,XX. Hormone replacement therapy was introduced. At the age of 16 chronic thyroiditis with elevated levels of anti-thyroid peroxidase antibodies was diagnosed. Nerve conduction studies at the age of 21 showed axonal sensorimotor polyneuropathy. Ophthalmological examination was unremarkable except for impaired eye movements (Table 1).

The proband’s sister was born at term with 2850 g of body weight and an Apgar score of 5/6/7. Horizontal nystagmus and imbalance began at the age of 11, walking was gradually deteriorating and sensorimotor polyneuropathy was identified. At the age of 12 SNHL and ovarian dysgenesis with normal female karyotype were diagnosed. From the age of 16 years she has a Hashimoto’s disease. Biochemical studies showed mildly elevated levels of serum lactate and creatine kinase. Metabolic disease screening for organic acids with gas chromatography-mass spectrometry (GS/MS) in urine, amino acids and acylcarnitines with liquid chromatography-tandem mass spectrometry (LC–MS/MS) in dried blood spot and congenital disorders of glycosylation gave normal results. Wilson and Refsum diseases were excluded based on normal blood concentrations of ceruloplasmin and cooper and phytanic acid, respectively.

Written informed consent was obtained from each participant. The study was approved by the ethics committee at the Institute of Physiology and Pathology of Hearing and performed according to the Declaration of Helsinki. The patients underwent thorough clinical evaluation. Functional impairment was assessed according to spinocerebellar degeneration functional score (SDFS) [20]. In brain and cervical spine MRI (3T Siemens Magnetom Trio, 12-channel Head Matrix Coil) T1-weighted, T2-weighted and diffusion-weighted images were acquired. In addition, high-resolution 3D structural T1-weighted volumes were acquired using an MPRAGE sequence with 208 sagittal slices and an isotropic resolution 0.9 × 0.9 × 0.9 mm. Sequence parameters were: TR = 1900 ms, TE = 2.21 ms, TI = 900 ms, FA = 9, FOV = 26 × 28.8 cm, matrix = 320 × 290, Pixel bandwidth = 200 Hz/pix, iPAT = 2, TA = 5 min. Brain structures were segmented by Freesurfer version 5.3 (http://​surfer.​nmr.​mgh.​harvard.​edu/​). The diameter of the vestibulocochlear nerve was evaluated directly and compared with the neighboring facial nerve used as an internal reference. Assessment of the cochlear and vestibular components was based on their visual comparison and with reference to the facial nerve. For comparisons of the cerebrum and cerebellum volumes and their white and gray matters, age- and sex-matched controls [21‐23] from the Internet Brain Volume Database funded by The Human Brain Project (http://​ibvd.​virtualbrain.​org/​) were used. A difference above 2.3 of the standard deviation was considered as statistically significant.

Assessment of auditory function consisted of pure-tone and speech audiometry, impedance audiometry, otoacoustic emissions (OAE) and ABRs. Hearing thresholds for air and bone conduction were determined at frequencies 125–8000 and 250–4000 Hz, respectively, using the AC40 clinical audiometer (Interacoustics, Middelfart, Denmark) and the 10/5 dB descending-ascending threshold estimation procedure [24]. Speech comprehension was tested using monosyllabic Polish words and the AC40 audiometer (Interacoustics) [25]. Acoustic impedance measurements (tympanograms and stapedius reflex) were performed with the Zodiac 901 instrument (Madsen Electronics, Copenhagen, Denmark). Stapedius reflexes were analyzed for the frequencies 500, 1000, 2000 and 4000 Hz in the ipsi and contralateral modes [26]. OAE were evoked by standard-click stimuli and 500 Hz tone bursts by using the ILO-292 system (Otodynamics Ltd, Hatfield, United Kingdom) [27‐29]. ABRs were recorded using the Integrity V500 system (Vivosonic Inc., Toronto, Canada). The stimuli were 0.1 ms clicks with alternating polarity presented with 90 dB normal hearing level (nHL) intensity at a repetition rate of 11/s. The amplifier bandwidth was 30–1500 Hz and analysis time 12 ms. The number of sweeps required for an averaged response was 1024 [30].

For evaluation of the vestibular endorgan function Fitzgerald and Hallpike bithermal caloric test with video eye movement recordings (Visual Eyes Micromedical Technologies, Chatham, USA) were used. Sinusoidal harmonic acceleration testing at frequencies 0.01–0.32 Hz was conducted with a Rotational Vestibular Chair System 2000 (Micromedical Technologies, Chatham, USA). Otolith function was measured using air-conducted sound stimulation cervical and ocular vestibular evoked myogenic potentials (cVEMP, oVEMP) at 500 Hz, 95 dBnHL (EclipsVemp, Interacoustics, Assens, Denmark).

DNA was isolated from blood sample by a standard procedure. WES was performed using SureSelect Target Enrichment (Agilent Technologies, Palo Alto, CA, USA) according to the manufacturer’s protocol. The sample was run on 16% of a lane on HiSeq 1500 using 2 × 100 bp paired-end reads. All bioinformatics analysis was done as described previously [31]. After primarily CASAVA processing, all reads were aligned to the hg19 reference genome with the Burrows-Wheeler Alignment Tool and analyzed with Genome Analysis Toolkit [32]. Indel realignment, base quality score recalibration, duplicates elimination as well as SNP/INDEL calling were performed [33]. The retrieved variants were annotated with ANNOVAR and converted to MS Access format for subsequent manual analyses. Total exon coverage by 20 reads or more was 89% and by 10 reads or more 95.9%. Alignments were inspected with Integrative Genomics Viewer [34] and analyzed with a pipeline combining protein coding changes, splice site prediction, prevalence in populations, evolutionary conservation and scores from PolyPhen-2 [35], SIFT [36] and MutationTaster2 [37] prediction algorithms of non-synonymous single-nucleotide variants. Sanger sequencing with 3500xL Genetic Analyzer (Applied Biosystems, Foster City, CA, USA) and BigDye Terminator cycle sequencing kit v. 3.1 (Applied Biosystems) were used to confirm the presence of variants identified by WES.

Homologs of the human Twinkle protein were identified with a PSI-Blast [38] search (E-value threshold of 0.005) performed against the NCBI non-redundant protein sequence database. The collected 15,000 sequences were initially clustered at 60% sequence identity with cd-hit [39] and any sequences shorter than 300, longer than 1000 amino acids or described as “hypothetical protein” were removed. The multiple sequence alignment of the twinkle family was derived using MAFFT program [40].

For preparation of the Twinkle homology model, the crystal structure of the gp4 protein from bacteriophage T7 (Protein Data Bank code1e0j) was selected as a template after analyzing the GeneSilico Metaserver results [41]. The sequence-to-structure alignment between the Twinkle protein and the template (Additional file 1: Figure S1) was built using the consensus alignment approach and 3D assessment [42] based on the results of FFAS [43], HHSearch [44] and the alignment proposed by Fernandez et al. [19]. Multiple sequence alignment of the family was also taken into consideration. The 3D model of the protein was built with MODELLER [45]. A model quality assessment was carried out using ProSA-web server [46]. Secondary structure elements were predicted with PSI-PRED [47]. Structure visualization was carried out with PyMOL (http://​www.​pymol.​org).

The proband and her sister came to our observation at the age of 27 and 19, respectively. They both had progressive neurologic symptoms, including postlingual progressive SNHL, cerebellar syndrome and flaccid paresis (Table 1) with muscle atrophy being more prominent in the proband.

Brain MRI in both sisters revealed considerable bilateral thinning of the vestibulocochlear nerve and its cochlear and vestibular components, which was defined as partial atrophy. It was more pronounced in the cochlear than in the vestibular nerves. Morphological signs of a subtle cerebellar atrophy (widening of the sulci more pronounced in the vermis than in hemispheres) were found in the proband but not in the sister. In cervical spine MRI, a diminished cervical enlargement was observed in both patients (Fig. 1). In the proband volumetric measurements of the brain showed significantly increased cerebellum white and decreased cerebellum gray matter. The results of other volumetric measurements were unchanged (Table 2).

In the proband and her sister, a different degree of sensorineural hearing loss mainly affecting high frequencies has been diagnosed in pure-tone audiometry (Fig. 2a). In contrast to pure-tone thresholds, speech discrimination was unproportionally poor and the test has not been continued. In both patients, the tympanograms revealed normal middle ear function. Ipsi- and contralateral acoustic reflexes were absent for all tested frequencies. The analysis of OAE recordings from both patients showed the presence of otoacoustic emission signals in the frequency range up to 2 kHz in the right ear and up to 4 kHz in the left ear. OAE in both patients were largely consistent with the results of pure-tone audiometry, demonstrating a partially impaired function of the outer hair cells. In ABR recordings, no responses at the maximum level of 90 dB nHL were obtained biaurally in the proband (Fig. 2b) and her sister. Comprehensive audiological evaluation revealed auditory neuropathy that was accompanied by a certain degree of cochlear dysfunction.

In both patients bi-thermal caloric irrigation results were within the normal limit. Rotational testing showed abnormally reduced gain and increased phase at 0.01, 0.02 and 0.04 Hz and was unaffected at the remaining frequencies. In the proband cVEMP and oVEMP were not recorded on either side. In her sister only oVEMP on the left side was registered with a prolonged latency of N1 and P1 peaks (data not shown). The results are suggestive of a neuropathy of both the superior and the inferior vestibular nerves in addition to the involvement of the auditory branch of the vestibulocochlear nerve.

DNA sample of the proband was analyzed by whole exome sequencing. After exclusion of variants found with a prevalence of 1% or more in the databases of the Exome Aggregation Consortium (ExAC, http://​exac.​broadinstitute.​org/​), 1000 Genomes Project (http://​www.​1000genomes.​org) and the NHLBI GO Exome Sequencing Project (ESP, http://​evs.​gs.​washington.​edu/​EVS/​; all accessed 05/2016) and in a set of 816 exomes of Polish patients (ZGM, R. Płoski, unpublished results) in the first line we searched for variants reported in the Human Gene Mutation Database (www.​hgmd.​cf.​ac.​uk/​ac/​index.​php) and variants predicted to be pathogenic by bioinformatic tools. We found a rare TWNK heterozygous missense variant NM_021830.4:c.1196A>G (rs863223921), causing the missense change NP_068602.2:p.Asn399Ser (Fig. 3a, b) that has been identified for the first time in 2016 in a Norwegian female with PRLTS [10] (Table 3). The variant is predicted to be damaging by PolyPhen-2 (score 0.993), SIFT (score 0.02) and MutationTaster2 (score 0.998).

The second TWNK variant was a very rare heterozygous missense change NM_021830.4:c.1802G>A (rs141315771), causing the amino acid substitution NP_068602.2:p.Arg601Gln (Fig. 3a, b). The p.Arg601Gln variant was reported only in the ExAC and ESP databases (accessed 07/2016) with an allele frequency of 3.29e−5 (4/121412 alleles) and 1.54e-4 (2/13006 alleles), respectively but heretofore it has not been identified in a homozygous state or associated with any disease. The G>A transition is predicted to be damaging by PolyPhen-2 (score 0.895), SIFT (score 0.02) and MutationTaster2 (score 0.997).

Presence of the two heterozygous TWNK mutations was confirmed in the proband’s sister. The mother was a carrier of the p.Asn399Ser mutation, showing that the two TWNK mutations are biallelic and the patients are compound heterozygous for the mutations (Fig. 3a, b). DNA sample from the deceased father was not available for the study.

Multiple sequence alignment demonstrated that Asn399 and Arg601 are conserved amino acid residues among vertebrates (Fig. 4a). In the crystal structure of the bacteriophage T7 gp4 protein (human Twinkle homolog), Asn289 (human Asn399) forms two hydrogen bonds with the backbone of Phe296 (human Trp392). A similar scenario is observed for the human protein. The side chain of Asn399, which is located on the short helix, forms two hydrogen bonds with the backbone atoms of Trp392. Both amino acids are located in a region between the linker and helicase domains. The p.Asn399Ser mutation disrupts these interactions as Ser399 is located too far away to form hydrogen bonds with the main chain of Trp392 (Fig. 4b, c). As a result, the entire region may adopt a different conformation than in the wild type protein. This may impair the correct orientation of the linker region and hinder the hexamer/heptamer formation [19].

In the T7 gp4 protein the adenine ring of the ATP analog is stacked between Arg504 (human Asp591) and Tyr535 (human Phe621) [48]. Val514 (human Arg601) is positioned next to the adenine ring and placed on the same plane. In the human Twinkle protein the key Arg504 is replaced by an aspartic acid, which is unable to stabilize the adenine ring. Simultaneously, an arginine (Arg601) is introduced at the position occupied by a short-chained amino acid (Val514) in the T7 phage. According to our 3D model, human Arg601 could adopt a conformation, which allows the formation of a hydrogen bond with the ATP N7 atom. This conformation can be stabilized by hydrogen bond interactions with Asp591. In contrast to the T7 phage Arg504, human Arg601 seems not to form a cation-π interaction with the adenine ring. However, we cannot exclude the possibility that the loop, on which both Arg601 and Asp591 are located, assumes a different conformation than in our model. This may allow Arg601 to adopt a conformation where the adenine ring is stabilized via stacking rather than hydrogen bonding. The interaction between ATP and Phe621 remains unaffected. Substitution of Arg601 to Gln most likely weakens the binding of ATP as Gln is unable to form a hydrogen bond with the adenine N7 atom (Fig. 4d, e).