The patients received medical care at the Union Hospital of Tongji Medical College of Huazhong University of Science and Technology. Diagnostic assessments included pure-tone audiometry, blood tests, ultrasonography, magnetic resonance imaging, and so on. This study received approval from the Ethics Committees of Tongji Medical College of Huazhong University of Science and Technology, and BGI.

Genomic DNA was isolated from peripheral blood samples using the Magen DNA extraction kit (Magen Biotech, Guangzhou, China). Qualified genomic DNA samples were processed for library construction. Exome capture was performed using the KAPA HyperExome Probes (Roche, CA, United States), followed by high throughput sequencing at BGI. The sequencing average depth was at least 100x for the target region. Initial processing of the raw sequencing data involved rigorous quality control, followed by the mapping of high-quality reads to the human reference assembly (GRCh37/hg19). To detect single nucleotide variants (SNVs) and insertions/deletions (Indels), we utilized the Genome Analysis Toolkit, while ExomeDepth was integrated into the pipeline specifically for the identification of copy number variations (CNVs) (McKenna et al., 2010; Plagnol et al., 2012).

The pathogenicity evaluation of SNVs and Indels was conducted in strict accordance with the guideline for the interpretation of sequence variants in genetic hearing loss (Oza et al., 2018). Sanger sequencing was performed to validate the candidate variants identified by WES.

To validate the molecular impact of LARS2 c.235-2A>G variant, minigene assay was conducted. A wild-type genomic fragment of LARS2, encompassing exon 3 (234 bp) - partial intron 3 (781 bp) - exon 4 (129 bp) - partial intron 4 (448 bp), was amplified from the control’s genomic DNA. The splice variant c.235-2A>G was The wild-type and mutant-type PCR products were cloned into the pcMINI-N vector (Bioeagle Biotech Company, China) at the KpnI and EcoRI restriction sites (Figure 3A). The integrity of recombinant constructs was verified by Sanger Sequencing. HeLa and 293T cells were selected due to their high transfection efficiency and well-characterized splicing profiles. Transient delivery of the recombinant constructs (1 ug per well) was performed using a Rapid Plasmid Mini Kit (1005250, SIMGEN, China) according to the manufacturer’s instructions. Following a 48 h incubation period, total RNA was isolated using the Trizol RNAiso PLUS reagent (9109, TaKaRa, Japan) and subsequently converted into cDNA through reverse transcription. To prevent interference from endogenous LARS2 transcripts, PCR amplification was carried out using vector-specific flanking primers (pcMINI-N-F: CTA​GAG​AAC​CCA​CTG​CTT​AC and pcMINI-N-R: TAG​AAG​GCA​CAG​TCG​AGG), which specifically target the exogenous transcripts produced by the minigene construct. The amplification products were resolved on agarose gel. Distinct bands corresponding to different splicing isoforms were excised from the gel and subjected to Sanger sequencing to confirm the exact splicing events.

Peripheral blood samples from family 1 were utilized for total RNA isolation via a commercial extraction kit (TR121-50, Genstone Biotech, China). Subsequently, total RNA was treated with DNase I to remove genomic DNA contamination and reverse-transcribed into cDNA using the Hifair™ first Strand cDNA Synthesis SuperMix (11123ES70, Yeasen, China). To examine the molecular consequences of the LARS2 c.235-2A>G variant, a pair of primer sets were designed to flank the variant. The sequences of the oligonucleotides used for nested PCR were: LARS2-F1: CCT​GTG​AGC​AGA​TCC​AGA​CC; LARS2-R1: TTC​CAC​CTT​TGC​TCC​AGA​AC; LARS2-F2: ATT​TGA​GGG​CCT​TCT​CAC​CT; LARS2-R2: CCC​AGA​CGA​TCA​AGC​TGT​TT. PCR amplification was then carried out using PrimerSTAR MAX DNA Polymerase (R045A, TaKaRa, Japan). The RT-PCR reaction protocol is shown in Supplementary Table S1. The RT-PCR products were verified by Sanger sequencing to assess splicing outcomes.

The UniProt database was utilized to acquire the structure of the wild type LARS2 protein. The tertiary structures of the mutant LARS2 protein were predicted using Missense 3D (https://missense3d.bc.ic.ac.uk/) and then processed and visualized with Pymol. Additionally, evolutionary conservation analysis was performed using Constraint-based Multiple Alignment Tool (https://www.ncbi.nlm.nih.gov/tools/cobalt/cobalt.cgi).

A comprehensive literature review was conducted to systematically summarize all reported LARS2 variants associated with Perrault syndrome. The literature search was conducted using PubMed, Google Scholar, and CNKI databases with keywords including “LARS2,” “Perrault syndrome,” and “variant/mutation”. After excluding duplicate reports of the same variant or family, 30 LARS2 variants were finally included for functional domain mapping and genotype-phenotype correlation analysis.

Proband 1 (F1 II-1): She is a 12-year-old daughter of unrelated Chinese parents. She did not pass the newborn hearing screening and was diagnosed with congenital profound SNHL. The patient exhibited delayed development, being able to walk independently at 2 years old. At the age of 5, she underwent right cochlear implantation; and the left ear did not receive a cochlear implant or a hearing aid. Due to the presence of a magnet in her cochlear implant, cranial magnetic resonance imaging is not feasible. Recent pure-tone audiometry revealed bilateral, symmetric, profound hearing loss, with air-conduction thresholds of 100 dB HL and bone-conduction thresholds of 60 dB HL (F1 II-1, Figure 1B). Now she could hear some sounds from her surroundings, but is unable to speak. Now she uses sign language as her primary mode of daily communication. She has poor attention span and limited comprehension ability. Additionally, menarche occurred at 11 years old, with only three menstrual episodes in total, followed by complete amenorrhea to date. Hormonal assays revealed significantly elevated FSH (66.43 IU/L) and LH (29.36 IU/L), along with markedly decreased AMH (0.01 ng/mL) (Table 1). Ultrasonography showed a left ovary measuring 11 mm × 8 mm × 6 mm with no visible follicles, and a right ovary measuring 15 mm × 9 mm × 7 mm with only one follicle larger than 4 mm (Figure 1C). These findings indicate POI. Furthermore, complete blood count, blood glucose, lipid profile, and lactic acid levels were all within normal limits (Table 1).

Proband 2 (F2 II-1): He is the 7-year-old son of a non-consanguineous Chinese couple. Newborn hearing screening was not performed. The patient has not received cochlear implantation or hearing aid assistance in either ear. He presented with bilateral, severe SNHL. Recent pure-tone audiometry revealed average air conduction thresholds of approximately 82 dB HL and bone conduction thresholds of 73 dB HL. The audiogram suggested more pronounced hearing loss in 500 Hz and 1,000 Hz (F2 II-1, Figure 1B). Currently, the patient demonstrates poor academic performance and limited comprehension ability. Developmental delay, seizures, or epilepsy were not observed. Cranial MRI showed symmetric brain architecture, preserved gray-white matter differentiation, and no abnormal signal intensities in the parenchyma. Laboratory investigations, including complete blood count, sex hormone profiles, blood glucose, lipid panel, and lactic acid levels, were all within normal limits (Table 1).

Through whole exome sequencing, two variants in the LARS2 gene were identified in each of the two unrelated probands. Proband 1 carries compound heterozygous variants in the LARS2 gene: c.880G>A (p.Glu294Lys), which was previously reported by Zazo Seco (Zazo Seco et al., 2016), and a novel variant c.235-2A>G. Proband 2 harbors two additional LARS2 missense variants: c.1886C>T (p.Thr629Met), which has been documented by Pierce (Pierce et al., 2013), and a second novel variant c.1661T>C (p.Val554Ala). All variants were confirmed by Sanger Sequencing in the respective family members (Figure 2). The parents of each proband are only heterozygous carriers and exhibit normal hearing (Figure 1A). No pathogenic variant was identified in other hearing loss-associated genes.

Computational analysis via the RDDC^SC and Varseak platforms suggested that the LARS2 c.235-2A>G variant is likely to disrupt a conserved splice acceptor site of exon 4 of LARS2, potentially leading to exon skipping. Additionally, SpliceAI indicated that the LARS2 c.235-2A>G variant leads to the loss of the canonical acceptor site with a high confidence score of 0.99. All three algorithms consistently suggested that this variant is likely to disrupt normal splicing. To functionally assess the impact of this variant on splicing, we performed in vitro and in vivo experiments. Expression plasmids containing the variant-bearing exon 4 of LARS2 and its flanking regions were constructed (Figure 3A); these minigene systems were then utilized to assess splicing patterns in 293T and HeLa cellular environments. The variant construct produced aberrant splicing patterns, including a 9-bp deletion within exon 4, a 54-bp deletion spanning part of exon 4, and complete exon 4 skipping (Figures 3B–D). To validate the splicing effect in vivo, we performed RT-PCR using peripheral blood from family 1. RT-PCR products from both the proband and her mother who carries the variant only showed exon 4 skipping, whereas her father who does not carry LARS2 c.235-2A>G exhibits normal inclusion of exon 4 (Figure 4). Both the minigene assay and RT-PCR analysis confirmed that the LARS2 c.235-2A>G variant led to aberrant splicing.

This study is the first to report two novel LARS2 variants (c.235-2A>G and c.1661T>C) associated with PS in the literature. We identified the LARS2 c.235-2A>G variant in 2023. Although this variant was registered in ClinVar in 2024, it has not been reported in the literature, and no clinical phenotypic information is available for this variant in ClinVar. Additionally, the application of the PVS1 criterion for LARS2 c.235-2A>G in ClinVar is incorrect.

This study reports a new proband (F1 II-1) who carries LARS2 c.235-2A>G in trans with the likely pathogenic variant LARS2 c.880G>A, providing additional evidence for the PM3 criterion. Furthermore, the patient presents with POI, which contributes evidence for the PP4 criterion. LARS2 c.235-2A>G is observed at an extremely low frequency (popmax filtering allele frequency is 0.000019 in East Asian) in the gnomAD dataset, meeting the PM2_Supporting criterion. Notably, in silico predictions, minigene assays as well as RT-PCR experiments indicate that LARS2 c.235-2A>G leads to exon skipping, with the missing protein regions constituting less than 10% of the full-length LARS2 protein. According to the latest PVS1 decision tree, this variant can only be assigned PVS1_Moderate, rather than PVS1 as annotated in ClinVar. Based solely on the information provided in the ClinVar database, the evidence is insufficient to classify the variant as LP. However, when integrated with the proband’s detailed clinical phenotypes and functional experimental validation conducted in this study, the variant meets the criteria for classification as LP (Table 2).

LARS2 c.1661T>C does not exist in gnomAD, meeting the PM2_Supporting criterion. Computational prediction tools support a deleterious effect of the LARS2 c.1661T>C variant on the gene, providing the PP3 criterion. Moreover, this study reports two unrelated hearing loss patients who carry LARS2 c.1661T>C in trans with LARS2 c.1886C>T (another pedigree is shown in Supplementary Figure S1), providing PM3_Strong evidence. In addition to SNHL, the patient exhibits academic challenges and impaired comprehension, which aligns with the PP4 evidence criterion. According to the ACMG/AMP guideline, the LARS2 c.1661T>C variant was classified as LP (Table 2).

Three-dimensional structural modeling was performed to assess the pathogenic potential of LARS2 c.1661T>C (p.Val554Ala). The p.Val554Ala variant is located within the catalytic domain of LARS2. This substitution replaces valine with alanine, converting the side chain from an isopropyl group to a smaller methyl group. Although both the wild-type (valine) and mutant (alanine) residues are non-polar and form backbone hydrogen bonds with HIS591 and PRO589 at distances of 2.8 Å each, the Val-to-Ala change is predicted to alter side chain interactions with surrounding residues (Figures 5A,B). This disruption is likely to decrease the protein stability, and potentially impair aminoacylation efficiency.

Evolutionary conservation analysis reveals that the valine residue at position 554 is highly conserved across diverse species, including mammals and non-mammals (Figure 5C). This finding suggests a critical role of Val554 in the LARS2 protein’s structure and function.

Through a comprehensive literature review, we systematically mapped the distribution of 30 LARS2 variants onto the functional domains of LARS2 protein (Figure 6). 15 variants were predominantly localized to the catalytic domain, which is responsible for amino acid activation and tRNA aminoacylation. Eight variants were localized to the editing domain, which ensures translational fidelity by proofreading tRNAs.

Moreover, genotype-phenotype analysis revealed that all individuals harboring P/LP LARS2 variants presented with deafness and/or POI, with a substantial subset exhibiting additional neurological symptoms. Notably, specific variants (p.N153H, p.T522N, p.Y557C) were associated with Marfanoid features. The novel LARS2 c.235-2A>G, a splice-site variant in intron 3 preceding the catalytic domain, was identified in a 12-year-old girl presenting with both profound SNHL and POI. The novel LARS2 p.Val554Ala, located within the catalytic domain, was found in a 7-year-old boy with hearing loss accompanied by learning and comprehension difficulties.