Family elders (parents) provided written informed consent for molecular analysis and publication of the clinical data for research and diagnosis purposes. The families reported in the present study originated from Pakistan and Italy. Family A was diagnosed at Khalifa Gul Nawaz Teaching Hospital, and the district headquarter hospital in Bannu Khyber Pakhtunkhwa, Pakistan. Families (B–D) were diagnosed at the Medical Genetics department of the University of Siena, Italy (Figure 1A).

Genetic counseling was performed to evaluate each patient's personal and familial history.

Genomic DNA of both the affected and unaffected family members was isolated from peripheral blood using standard methods and quantified using the Nanodrop-2000 spectrophotometer (ThermoFisher Scientific, Waltham, MA, USA).

Sample preparation was performed following the Illumina DNA prep with enrichment manufacturer protocol. A bead-based transposome complex was used to perform segmentation, which fragments the genomic DNA and then tags it with adapter sequences in one step. After saturation with input DNA, the bead-based transposome complex fragments a set number of DNA molecules. This fragmentation allows a wide DNA input range to generate normalized libraries with a consistent tight fragment size distribution. Then a limited-cycle PCR adds adapter sequences to the ends of a DNA fragment. A subsequent target enrichment workflow is then applied. Following pooling, the double-stranded DNA libraries are denatured and biotinylated. Next, Illumina Exome Panel v1.2 (CEX) probes are hybridized into the denatured library fragments. Then Streptavidin Magnetic Beads capture the targeted library fragments within the regions of interest. Then the indexed libraries are eluted from beads and further amplified before sequencing. The exome sequencing analysis was performed on the Illumina NovaSeq6000 System (Illumina San Diego, CA, USA) according to the NovaSeq6000 System Guide. Reads were mapped against the hg19 reference genome using the Burrow-Wheeler aligner BWA (Li and Durbin, 2010). Variant calling was obtained using an in-house pipeline which takes advantage of the GATK Best Practices workflow.

Trio whole-exome sequencing (WES) was performed, including the proband and both parents. The caring clinicians gathered clinical and mutation details of the patients harboring the SPTBN5 variants by getting in touch through Gene Matcher (http://www.genematcher.org). Exome sequencing and variant interpretation and analysis were as previously described (Monies et al., 2019; Shao et al., 2021). All the variants were screened according to the location, frequency, and type of variation. Variants were filtered with a minor allele frequency (MAF) cutoff of 1% in the Exome Variant Server (http://evs.gs.washington.edu/EVS/), GnomAD (https://gnomad.broadinstitute.org), and 1000 Genomes (http://www.1000genomes.org/). Sanger sequencing was performed using stranded methods (Khan et al., 2019; Umair et al., 2019).

The bioinformatics analysis focused on non-synonymous SNVs (missense, non-sense, splice-site, and frameshift) and was submitted to Polyphen-2 (http://genetics.bwh.harvard.edu/pph2/), Sorting Intolerant from Tolerant (SIFT, http://sift.jcvi.org/), Protein Variation Effect Analyzer (PROVEAN, http://provean.jcvi.org), Mutation Taster (http://www.mutationtaster.org/), Varsome (https://varsome.com/), Mutation assessor (http://mutationassessor.org), and Combined Annotation Dependent Depletion (CADD, https://cadd.gs.washington.edu/) for functional effect prediction.

AlphaFold (Jumper et al., 2021) was used to produce high-quality structural models of three 1,000 residue segments of SPTBN5, containing the residues of interest. The average pLDDT score of the three models was higher than 80, and the residue pLDDT of His89, Asn2937, and Glu3262 and their local region was higher than 90, indicating a very confident 3D configuration. A configuration of four V100 GPUs and 16 CPU cores (provided by the KAUST IBEX cluster) was used for modeling. Models were manually inspected, and mutations were evaluated using the Pymol program (pymol.org). The American College of Medical Genetics and Genomics (ACMG) 2015 guidelines were used for the interpretation of variants (Li et al., 2020).

Gene Runner (version 5.0.69 Beta, Hastings, NY, USA) software was used for Primers design. Sanger sequencing was performed using an ABI3730 Automated Sequencer to verify co-segregation of the identified variants with the disease phenotype (Thermo Fisher Scientific, Waltham, MA, USA). The Sanger sequencing results were examined and compared with the help of visual software such as Chromas Lite (http://technelysium.com.au/wp/) and Codon Code Aligner (https://www.codoncode.com/aligner/).

This study recruited four families affected by ID from Pakistan and Italy. A clinical neurologist examined all the affected individuals. Additional clinical information on the affected individuals is summarized in Table 1.

Family A, of Pakistani origin, has a single affected individual (II:3), a 11-year-old male with ID had a developmental delay born from non-consanguineous parents. The pregnancy and delivery events were unremarkable. The parents of the affected individuals are normal and have no neurological symptoms related to cognition. Patient II-1 was examined at the age of 10 and 11 years, respectively. Clinical examination revealed early-onset epilepsy and seizures at 2–3 years of age, anxiety, severe chronic constipation, aggressive behavior, feeding difficulties, and harming himself by beating his head against the wall. Physical examination revealed a prominent metopic ridge, strabismus, epicanthus, flat philtrum, stuttering, and relatively thin upper lips with mild wide spacing of teeth. He had a normal 46 XY karyotype. At the last examination (11 years of age), his weight, height, and Occipital Frontal Circumference (OFC) were 45.9 kg, 52.1 cm, and 49 cm, respectively. No cardiac, respiratory, skeletal, or skin anomalies were observed. His vision and hearing were seen as normal (Figure 1A).

We reported a 15-year-old girl with ID born from non-consanguineous parents regarding the Italian family B. She was born after pregnancy with threats of miscarriage; a growth deficit was signaled during the last month. She was diagnosed with drug-induced autoimmunity at birth and interatrial and interventricular defects (DIA and DIV). She came to our attention when she was 3-year-old, and since her first months of life, she presented recurrent infections in the upper respiratory vias and feeding difficulties with gastroesophageal reflux. The anamnestic collection reported a paternal cousin who acquired severe ID after meningitis. Physical examination at the age of 15 years showed a height of 144.5 cm, a weight of 35 kg, and a head circumference of 50.5 cm. She suffered from ID, hyperactivity, language/speech delay and aggressiveness, and a disturbing wake-sleep cycle. Arachnodactyly of the hands and feet was noticed. She showed triangular facies, prominent ears, thin upper lip, absent eyebrows, broad nasal bridge, bulbous nasal tip, thin and sparse hair, fusiform fingers of the hands and the feet, everted lower lip, M-shaped upper lip, hairline anteriorly advanced. Her array-CGH exams gave negative results (Figure 1A).

Family C includes a 14-year-old Italian boy with a history of physical dysmorphisms and behavioral and nonverbal learning disorders. He was born at 38 weeks with a cesarian section. Her anamnestic collection presented a brother with language and behavior delay, a maternal cousin with language delay, and a paternal cousin with seizures at pediatric age. His weight and height were 34 kg and 150.5 cm, and OFC was 46.5 cm, respectively. He showed psychomotor delay, cognitive delay, and manifested aggressiveness through himself and the others. His physical examination reveals thin upper lips, bilateral clinodactyly of the 3rd−4th and 5th fingers, depressed nasal bridge, turricephaly, high anterior hairline, hypertelorism, nasal voice, prominent forehead suture. His karyotype, array-CGH, and X-fragile exams were normal (Figure 1A).

Family D of Italian origin had an affected 6-year-old female child diagnosed with autism and ID. Her weight was 15 kg, height 97 cm, and chest circumference (CC) 50 cm. Her anamnestic collection highlighted language delay in the maternal line and autistic disorders in the paternal line. She was born at term and suffered from gastroesophageal reflux and feeding difficulties. Her weaning and her wake-sleep cycle were normal. However, she manifested aggressiveness, language delay, and expression disorder. Her ADOS score was 16 (moderate ASD), and her auditory evoked potentials were negative. In addition, her array-CGH exam was negative (Figure 1A).

By applying an iterative filtering strategy based on variant frequency, functional consequences on coding sequence, and heredity, de novo variants in the SPTBN5 gene [NM_016642] were considered the best candidate. These include c.266A>C; [p.(His89Pro)] in Family A, c.9784G>A; [p.(Gly3262Lys)] in Family B, c.933C>G; [p.(Tyr311^*)] family C, and c.8809A>T; [p.(Asn2937Tyr)] in family D. These candidate variants were confirmed by Sanger sequencing in all the available family members (Figure 1B).

The SPTBN5 encodes a 3674 long amino acids protein. It consists of two N-terminal calponin homology (CH) domains, actin-binding domains that cross-link actin filaments into bundles and networks (Richards et al., 2015). The α-helical CH domains are common in actin-binding proteins and play important regulatory roles in cytoskeletal dynamics and signaling. Most of the region C-terminal to the CH domains, almost 90% of the protein, is formed by spectrin repeats. The spectrin repeats form a tandem arrangement of helical coiled coils, where each repeat forms a three-helix bundle. Through tension-induced unfurling of the three-helix repeats, these domains allow spectrin to expand and contract, conferring flexibility to the protein (Borrego-Diaz et al., 2006). The spectrin repeats are also self-association and tetramer formation with alpha spectrins (Nicolas et al., 1998; Stradal et al., 1998). Finally, there is a pleckstrin homology domain (PH) at the C-terminal end. The PH domains bind to phosphatidylinositol lipids, allowing the recruitment of proteins to cellular membranes (Mayer et al., 1993).