The present study was approved by Ethics Committee of the First Hospital of Jilin University (approval number: 2021-485). All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Written informed consent was obtained from the parents.

The proband was an 8-year-old male child delivered at full term by normal delivery. He was brought to the First Hospital of Jilin University for delayed growth, delayed mental development and speech disorder. At the time of presentation, the child underwent a detailed medical examination with the following findings: Height, 120 cm; weight, 19.3 Kg; head circumference 43 cm; teeth, dysplasia and disorder of arrangement. Since his parents wanted to have another child, they wanted to find the genetic cause behind his presentation. A karyotype test with 320-400 bands performed previously had found no abnormalities. Furthermore, the brain parenchyma in the head MRI scan, the amino acid and acylcarnitine profiles, the urine organic acid analysis and the EEG test were all normal. Upon inquiry, the parents reported no family history of any genetic disease. For a genetic examination, 2 ml of peripheral blood sample was taken from the child as well as the parents after obtaining their written informed consent. The blood was anticoagulated with EDTA and stored at 4˚C.

The genomic DNA from the peripheral blood of the patient and his parents was extracted using the QIAamp Blood DNA Mini kit (Qiagen GmbH). The concentration and purity of the sample DNA were detected using a UV-240 UV spectrophotometer. The fragmented DNA was then screened by magnetic beads and the screened fragments measured 280-320 bp, with the ends filled and the base ‘A’ added at the 3' end to make the fragments compatible with the 3' end of the fragment. After amplification and purification by ligation-mediated PCR, the library of a single subject sample was constructed. A customized gene capture probe (MGIV4 59M; MGI) was used to hybridize 10-20 labeled patient DNA libraries at 47˚C for 16-24 h. Probe washing and elution reactions were performed after hybridization, followed by post-PCR reactions of the captured samples. The libraries were tested for fragment size and concentration using Agilent 2100 Bioanalyzer (Agilent Technologies, Inc.) and a BMG microplate reader; after passing the test, libraries requiring different data amounts were used for pooling and quantification. Then, the pooling libraries were subjected to single-stranded cyclization. After making a DNA nanoball (DNB), the libraries were sequenced using a high-throughput sequencer, MGISEQ-2000 (MGI) with PE100+10 type and the raw sequencing data were obtained after sequencing.

After the data were offloaded, they were analyzed. First, the sequencing quality of raw data (raw reads) was evaluated and the low-quality and junction-contaminated reads were eliminated. Then, sequence alignment was performed with the Burrows-Wheeler Aligner (BWA) software (version 0.7.17; https://bio-bwa.sourceforge.net/) and HG19/HG20 human genome reference sequences and the sequence capture performance was simultaneously evaluated. Single nucleotide variant and InDel (insertion and deletion) search was performed with the Genome Analysis Toolkit (GATK) software (version 4.1.9.0; https://gatk.broadinstitute.org/hc/en-us) to generate base polymorphism results in the target region and subsequently databases (NCBI dbSNP (build 151; http://www.ncbi.nlm.nih.gov/SNP), 1000 human genome dataset (phase 3; https://www.internationalgenome.org/) and database of 100 Chinese healthy adults (phase I; http://cmdb.bgi.com/) were compared and the suspected mutations were annotated and screened for.

The suspected pathogenic mutations identified by high-throughput sequencing were verified by Sanger sequencing of the patient and his parents and primers were designed upstream and downstream of exon 8 of the SATB2 gene: forward primer sequence, GTCAGTGGTTTCTGTCTTGGC and reverse primer sequence, GTCAGTGGTTTCTGTCTTGGC. PCR amplification was performed. PCR reaction system was as follows: Total volume was 25 µl, including 2 X PCR MasterMix (Promega Corporation; 12.5 µl), dNTPs (3 µl), ddH2O (5.2 µl), upstream and downstream primers (1 µl each), DNA template (2 µl) and LA taqase (0.3 µl). Reaction conditions were as follows: Pre-denaturation at 95˚C for 5 min, denaturation at 95˚C for 30 sec, annealing at 68˚C for 30 sec and extension at 72˚C for 30 sec for 12 cycles. This was followed by the amplification step, wherein the sample was subjected to 94˚C for 30 sec, 58˚C for 30 sec, 72˚C for 30 sec for 35 cycles, which was followed by 72˚C extension for 5 min. Next, 1.5% agarose gel was prepared and then 2 µl of loading buffer was mixed with 4 µl of DNA amplification product. After that, Sanger sequencing was performed using the ABI3730XL (Applied Biosystems; Thermo Fisher Scientific, Inc.) sequencer and the results were compared with the hg19 reference sequence in University of California Santa Cruz (UCSC) Genome Browser (https://genome.ucsc.edu/) to verify the results of gene chip capture and high-throughput sequencing.

The acquired sequenced segments were compared by BWA and UCSC hg19 human reference genome to remove duplicates. InDel and genotyping. ExomeDepth was used to detect CNV at the exon level.

After obtaining informed consent, peripheral blood samples were drawn from the parents of the patient. Then, peripheral blood lymphocytes were routinely cultured, produced and G-banded and 50 mid-phase divisions were counted microscopically and analyzed for five karyotypes.

AlphaFold is a deep neural network developed by DeepMind that predicts the 3D structure of a protein from its amino acid sequence (9). AlphaFold can predict the distance distribution between each pair of amino acids in the protein and the angle between the chemical bonds that connect them, then aggregates the measurement results of all pairs of amino acids into 2D distance histograms. The convolutional neural network can learn from these pictures and subsequently construct the 3D structure. The Fasta file format of the acquired protein sequence wase first downloaded, then submitted to the supercomputer center for AlphaFold modeling. The structured model needs to be assessed by the ERRAT program (version saves6.0; https://saves.mbi.ucla.edu/), which refers to the calculation of the number of non-bond interactions between different atomic types within the range of 0.35 nm, the higher the overall quality factor value indicates a greater resolution of the crystal structure.

The phenotypic progression of the child was as follows. He was born at 39-week gestation without obvious abnormality. He could sit at 6 months and walk without support at 26 months. However, at the time of writing, the child was 8 years old; his walk was not stable and he remained non-verbal. His height, weight and head circumference were 120 cm, 19.3 kg and 43 cm, respectively, with a body mass index (BMI) of 13.4. His limbs were long and thin and both knee joints were large and had cavities. He showed a significant developmental delay in terms of intelligence, particularly when compared with other children of his age. Further, he showed severe adaptability retardation, poor cognition and understanding abilities and could only understand simple commands. He showed severely impaired language development, which manifested as absent language function. Moderate lag was noted in grand motor and fine motor function. His walk was not steady with his forward-set posture and he could not run or jump. In terms of cephalic and facial abnormalities, both his eyebrows were thin and curved, bilateral downward slanting eye fissures were present and he had hypoplastic, crowded teeth, a high jaw arch, a small, pointed lower jaw with backward tilt (Fig. 1) and a flat occipital bone. He was not epileptic but showed excitable behavior, with frequent episodes of crying, hyperactivity, timidity and restlessness, as well as poor bowel awareness.

Genetic testing of blood samples from the child and his parents revealed that the exon 8 of the SATB2 gene had a c.771dupT (p.Met258Tyrfs*46) de novo repeat insertional shift mutation, which resulted in a mutation from methionine to tyrosine at the amino acid site 258 (Project number: PRJNA916183, database link: https://www.ncbi.nlm.nih.gov/bioproject/PRJNA916183). Consequently, this shift mutation altered all amino acids after site 258 and led to the production of a truncated protein with 46 amino acids missing. The parents of the child harbored no mutation at this locus (Fig. 2). The family pedigree of this child was shown in Fig. 3. The mutation identified in SATB2 gene (NM_015265.3), which is transcript variant 2, has not been reported in the search from the OMIM database (https://omim.org/) or PubMed database (https://www.ncbi.nlm.nih.gov/pmc/), NCBI-Nucleotide (https://www.ncbi.nlm.nih.gov/nucleotide/) showed that SATB2 NM_015265.3 mRNA transcript variant 2 contains 12 exons.

To the best of the authors' knowledge, the STAB2 gene c.771dupT (p.Met258Tyrfs*46) variant detected in this case is novel. According to the American College of Medical Genetics and Genomics guidelines (10), this variant was consistent with PVS1 and PM2 in the guidelines and was suspected to be a pathogenic variant.

The modeling scores of STAB2 and mutant protein were 91.348 and 96.257, respectively. For the STAB2 protein, the constructed protein structure has high reliability and can be used as a template for subsequent studies according to Ramachandran plot and VERFY-3D score (11) (Fig. 4A). For the mutant protein, the evaluation score was greater than 95%, which indicated high reliability and can be used as a template for subsequent studies (Fig. 4B). The three-dimensional structure of STAB2 protein consisted of 22 α-helical structures, 4 β-folds and 10 LOOP-regions. However, the mutation of p.Met258Tyrfs*46 in STAB2 leads to the premature termination of protein translation. The mutant protein contained only 6 α-helical structures, 4 β-folds and 4 LOOP-regions, thus causing a large amount of missing structure (Fig. 4C and D). The wild-type protein sited in 258 is a nonpolar uncharged methionine, while the p.Met258Tyrfs*46 mutation results in the replacement of methionine at position 258 by tyrosine, which was large, polar and uncharged, thus increasing the aromaticity and polarity of the protein. Property changes of these amino acids may lead to alternation in the structure and function of the mutant protein (Fig. 4E and F).

CNV analysis at the exon level did not detect pathogenic chromosomal CNV variants of more than 1M associated with the phenotype of the preexisting patient.

No abnormal karyotypes (320-400 bands) were found in the child and the parents' chromosome Giemsa staining bands.