We present a 19-year-old male diagnosed with ASD, intellectual disability, absent speech, and epilepsy. He was the second child delivered vaginally at term to a consanguineous 27-year-old mother and 33-year-old father. Paternal and maternal ancestry is from India, and they are 3rd cousins. He did not require cardiopulmonary resuscitation, medications, or special assistance following delivery. The neonatal period was unremarkable. His early milestones were reportedly age-appropriate. At 12 months of age, he walked independently, started running a few months later, rode a bicycle at three years old, and was skilled at climbing, roller skater, and swimming. He holds a pen in a fist and prefers to be alone, but has some interaction with individuals with whom he is comfortable. He had about 150 words at 2.5 years old, after which language regression was reported with less eye contact and loss of speech and echolalia by three years old, and he received a diagnosis of ASD at three years old. He is, however, interested in reading books, with repetitive behaviour of reading the same books and specific pages of the book. There are no concerns for hearing with a normal sedated hearing examination. Vision is normal. He is currently in an adult transition programme and is receiving special services for speech, OT (occupational therapy), and music therapy, along with learning to type and use a stencil board. He has no words currently. He had five episodes of seizures starting at 15 years old with vEEG showing rare midline epileptiform discharges suggesting a focal onset managed with oxcarbazepine. Other medical history includes multiple ear infections with antibiotics treatment throughout toddlerhood. His growth parameters, including height and weight, were normal. A brief examination on video showed the patient making non-verbal sounds, yelling, and gesturing like hands to mouth. The examination was limited due to poor cooperation. Although both parents and younger full sister were healthy, the family history is significant for developmental delay in the maternal and paternal first cousins. Genetic testing included a negative single-nucleotide polymorphism (SNP) microarray, with no diagnostic copy number changes observed. However, it detected homozygosity (ROH) regions greater than 5 Mb on chromosome 3. He also tested negative for Fragile X syndrome. MRI showed only a tiny focus of T2 flair hyperintensity in the left frontal white matter. A multidisciplinary team including a neurologist currently manages his epilepsy.

The exome sequencing trio was performed by a commercial laboratory (GeneDX/Sema4). Genomic DNA was extracted directly from the submitted specimen. The DNA was enriched for the complete coding regions and splice site junctions for most genes of the human genome using a proprietary capture system developed by GeneDx for next-generation sequencing with CNV calling (NGS-CNV). The enriched targets were simultaneously sequenced with paired-end reads on an Illumina platform. Bidirectional sequence reads were assembled and aligned to reference sequences based on NCBI RefSeq transcripts and human genome build GRCh37/UCSC hg19. Using a custom-developed analysis tool (XomeAnalyzer), data were filtered and analysed to identify sequence variants and most deletions and duplications involving three or more coding exons (PMID: 25,356,966). Smaller deletions or duplications may not be reliably identified. Reported variants were confirmed, if necessary, by an appropriate orthogonal method in the proband and, if submitted, in selected relatives. Sequence variants are reported according to the Human Genome Variation Society (HGVS) guidelines. Copy number variants are reported based on the probe coordinates, the coordinates of the exons involved, or precise breakpoints when known. Reportable variants include pathogenic variants and likely pathogenic variants. Variants of uncertain significance, likely benign and benign variants, if present, are not routinely reported. Available evidence for variant classification may change over time, and variant(s) may be reclassified according to the ACMG/AMP Standards and Guidelines (PMID: 25,741,868), which may lead to issuing a revised report. We searched Uniprot database (https://www.uniprot.org) for homologous proteins of USP51 in different species and used the Clustal Omega program (https://www.uniprot.org/align) to align them and assess their similarity. We used I-TASSER (https://zhanggroup.org/I-TASSER/) on the normal sequence and predicted truncated protein sequence from our patient.

An exome trio with both biological parents revealed a maternally inherited variant c.95del, p.Glu32GlyfsTer25 in exon 2 of the USP51 gene (RefSeq: NM_201286.3, Ensembl: ENST00000500968.3) (Fig. 1-A). For this gene, the sequencing covered 97.29% of the coding region at a minimum of 10 × sequencing. There is no indication of a multi-exon deletion/duplication involving this gene in the sequencing data. This gene is located on chromosome X, meaning the patient is hemizygous. His mother was heterozygous for the variant, and it was absent in the father. This variant is absent in large population cohorts (gnomAD) [1]. Because there is only one coding exon, this frameshift variant is predicted to result in protein truncation with the last 680 amino acids replaced with 24 different amino acids and a premature termination codon (PTC) that should not be detected by nonsense-mediated decay (NMD) nor be degraded (Fig. 1-B). Only 7 loss-of-function variants have been identified in gnomAD, 2 upstream (stop gained: p.Trp17Ter encoded by 2 different nucleotides changes) and 5 downstream (2 stop gained: p.Gln161Ter and p.Gln341Ter, and 3 frameshift: p.Lys232ValfsTer8, p.Ser251LeufsTer9 and p.Leu588IlefsTer13), all in females but one, in a male carrying the less disruptive frameshift (the most downstream, truncation of 124 amino acids replaced by 12 different and a PTC) (Fig. 1-B). Only one nonsense variant (p.Arg5Ter) was reported in ClinVar [2] (Fig. 1-B) and it was classified as pathogenic, but unfortunately, the only phenotype specified for the patient was short stature and the sex was not reported. A hemizygous missense variant (p.Thr497Ile) was identified in a patient with West syndrome (syndrome with neurologic features overlapping with our patient phenotype) (Fig. 1-B). This individual had variants in other genes (2 variants in RYR2 and 1 variant in RPGR) that may have also contributed to the phenotype [3].

A: Pedigree of the family. Proband is indicated by the arrow. Proband is the only family member affected with autism spectrum disorder and epilepsy (black filled). Exome trio performed on the proband and both his parents. B: Top panel: USP51 protein structure with known domains. On top of the protein are the loss-of-function variants reported in gnomAD. Below the protein are the variant reported in our proband and his mother (red), one variant reported as pathogenic in ClinVar, and one variant reported in a patient with West syndrome (blue). #: this variant has been reported in 2 independent individuals Bottom panel: Predicted truncated protein in our patient. In red is the new amino acids sequence created by the variant. All the important known domains would be missing. C: Alignment of truncated (mut) and normal human USP51 proteins with USP51 homologous proteins from 42 different species. Top panel: complete alignment, bottom panel: highlight of N terminus of the proteins alignment where the variant is located as well as the 24 different amino acids added before the premature termination codon. A darker blue indicates a higher similarity between the sequences. D: I-TASSER results of the top 5 final structural conformation models. Top panel: normal protein. Model 1: estimated TM-score = 0.41 ± 0.14 and estimated RMSD = 14.6 ± 3.7 Å. Bottom panel: predicted truncated mutant protein. Model 1: estimated TM-score = 0.45 ± 0.15 and estimated RMSD = 7.4 ± 4.2 Å.

In accordance with ACMG guidelines [4], because the frameshift variant (1) was not observed at a significant frequency in large population cohorts (gnomAD) (PM2 criteria), (2) although is predicted to result in protein truncation, loss-of-function variants have not been reported as mechanism of disease (not PVS1 criteria), (3) has not been previously published as pathogenic or benign to our knowledge and (4) is in a “candidate” gene for neurodevelopmental disorders, we interpreted it as a variant of uncertain significance (VUS).

There was a secondary finding of a paternally inherited heterozygous c.67458 T > G, p.Tyr22486Ter variant in the TTN gene interpreted as Likely Pathogenic to predispose to autosomal dominant familial hypertrophic cardiomyopathy (Fig. 1-A). Cardiology evaluation for the patient and his father was recommended as well as consideration of genetic testing for his sister.

When DNA damage takes place, a DNA damage response (DDR) including a rapid chromatin response is elicited, with the promotion of a variety of histone post-transcriptional modifications. Histone ubiquitination is at the heart of the DDR and corresponds to a three-step enzyme process: E1 activating, E2 conjugating, and E3 ligase that covalently bind a ubiquitin to a lysine (Lys, K) or an already bound ubiquitin residue to the protein. When the reparation is complete, a deubiquitinating enzyme (DUB) will cleave the ubiquitin marks from the ubiquitin substrates [5]. This mechanism regulates various cellular responses by controlling substrate abundance and activity.

Almost 20 years ago, Quesada and colleagues identified and cloned 22 human cDNAs encoding novel members of the ubiquitin-specific protease (USP) family, including USP51 [6]. USPs constitute the largest deubiquitinating enzyme family and exhibit several specific highly conserved structural features characteristic of this family of cysteine proteases including (1) the Cys box, which contains a conserved cysteine residue essential for the catalytic properties of these enzymes (Cys372 in USP51), (2) the “QQD” box, which contains an aspartic acid residue absolutely conserved in all USPs (Gln447Gln448Asp449 in USP51), and (3) the His box that contains a histidine that has also been proposed to participate in the catalytic mechanism of USPs and conserved in all USPs but one (His665 in USP51). USP51 also contains a Zn finger ZnF-UBP domain (amino acids 193–311) that allows it to bind to ubiquitin. All these domains, crucial for the function of USP51, are localized downstream to the PTC identified in our patient and will be missing which is in favour of a loss-of-function variant (Fig. 1-B). We searched for USP51 homologous in other species and identified 42. Alignment shows a high similarity demonstrating a conservation throughout evolution, even upstream of the location of the variant identified in our patient (Fig. 1-C). We used I-TASSER to predict the top 5 models of the structural conformation of the normal and predicted truncated mutant USP51 protein (Fig. 1-D) [7]. Model 1 of the normal protein had an estimated TM-score of 0.41 ± 0.14 and an estimated RMSD of 14.6 ± 3.7 Å, whereas the predicted truncated mutant protein model 1 had an estimated TM-score of 0.45 ± 0.15 and an estimated RMSD of 7.4 ± 4.2 Å.

USP51 gene starts to preferentially be expressed in the brain (mostly forebrain but also hindbrain) during the second part of the prenatal period and then continues through childhood and adulthood (GXD [8], Supplementary Table S3 from Peng et al. (2018) [3], GTEx [9], and BrainEXP [10]). The phenotyping of female homozygous and male hemizygous knockout mice for USP51 by the International Mouse Phenotyping Consortium (IMPC) shows significantly decreased bone mineral density in early adulthood [11]. Additionally, 2 females showed abnormal digit morphology and 1 male had a hindlimb defect (MGI: 3,588,217) [12]. Regarding a possible link between variants in the USP51 gene and human disease, the literature suggests the USP51 gene plays a role in the mouse-developing brain [13]. Wang and colleagues characterized USP51 as a DUB specifically involved in the DDR by removing H2AK13,15ub, following the DNA repair. Depletion of USP51 results in increased H2AK13,15 levels on chromatin and delayed disassembly of proteins at DNA damage foci [14]. By using a machine learning approach to predict autism risk genes, Lin and colleagues were able to validate known genes and discover new candidate genes. Among the top ten, 5 are related to protein ubiquitination (ZYG11B, HERC1, HECTD1, CAND1, and MYCBP2, ranked second, third, fourth, seventh, and tenth respectively) which is consistent with the significant enrichment of protein ubiquitination process in their Gene Ontology (GO) enrichment analysis [15]. Thus, by its function of protein deubiquitination, USP51 could be causing overlapping phenotypes. Clinicians should consider including USP51 in genetic testing panels for neurodevelopmental disorders, which could lead to more precise diagnoses and tailored interventions. It will help with genetic counselling to help families understand the potential implications of the variant, including risks for current and future generations. Expanding genetic screenings across diverse populations could help assess the prevalence and impact of this variant on a broader scale. Additionally, longitudinal studies tracking the developmental trajectories of individuals carrying this variant could offer valuable insights into its phenotypic consequences and interaction with environmental factors. Further research is still needed to explore a possible link between variants in the USP51 gene and human disease and potentially identify targeted therapies, for example, by utilizing the USP51 knockout mouse model available at the International Mouse Phenotyping Consortium (IMPC) to study behaviour that could resemble ASD in humans [11].

The USP51 loss-of-function variant is maternally inherited, but because his mother is heterozygous, she is only carrying one mutated copy and the normal copy of the gene seems enough to maintain a sufficient function of the encoded protein and thus explains why she does not express any phenotype (no haploinsufficiency).

Due to the parents being 3rd cousins, SNP microarray detected 2 ROH on chromosome 3 (Chr3:11,893,606–28270426; 16.38 Mb and Chr3:45,170,642–55189959; 10.02 Mb) that contain 333 genes including 217 OMIM genes and 75 OMIM diseases (26 with a dominant inheritance and 51 with a recessive inheritance) [16]. There are no known imprinted genes on chromosome 3 and uniparental isodisomy of chromosome 3 (paternal or maternal) has not been associated with disease [17,18,19]. Please note that Genescout also calculated a coefficient of inbreeding of 1/117 and a coefficient of relationship of 1/58 which is closer to a 6th degree of genetic relationship and consanguinity than the 7th degree of expected 3rd cousins (great grandfathers were brothers) [16]. Although the identification of multiple ROH throughout the genome itself is not associated with an abnormal phenotype, having identical copies of the genes in these regions increased the possibility of inheriting homozygous pathogenic variants in an autosomal recessive gene [20].

We also acknowledge that non-coding regions are not covered by this test and that deep intronic variants or changes in regulatory regions could have been missed and be responsible for the phenotype of the patient.