This study included all individuals with a typical OI phenotype who were evaluated in the Metabolic and Genetic Diseases Unit of the Hospital “Giovanni XXIII”, General Hospital Policlinico, Bari, Italy, between 2013 and 2023. This study was approved by the Institutional Review Board of the General Hospital Policlinico. Written informed consent was obtained from all patients aged 16 years or older whereas written parental consent was obtained for individuals under 16 years of age.

Clinical information, including medical history, age at diagnosis, gender, family history, and non-skeletal features such as hearing loss, was collected.

Blood samples were isolated from 10 patients affected by OI. DNA was isolated from 200 μL of peripheral blood using InviMag Blood DNA mini kit/IG kit and the automated platform for nucleic acid purification InviGenius Plus according to the manufacturer’s instructions (Stratec Molecular GmbH, Berlin, Germany). The concentration of the isolated DNA was determined by Qubit dsDNA HS Assay Kit (ThermoFisher Scientific, Waltham, MA, USA).

To identify PVs in genes involved in OI, we used a custom panel (Ampliseq custom DNA Panel) designed using the Designstudio V5.0 software (Illumina), which covers 99.93% of the COL1A1, COL1A2, FKBP10, IFITM5, PPIB, SERPINF1, WNT1, BMP1, CRTAP, P3H1, and SERPINH1. The custom panel consists of 266 total amplicons (1st pool of 133 and 2nd pool of 133 amplicons) having a complete coverage (99.93%) of all genes listed above. For each gene, all exons and adjacent intronic regions ± 25 nucleotides were sequenced (exon padding: 25). The only bases that are not covered by the in silico design were 22 bases of the padding (intron) of exon 8 of the FKBP10 gene, due to the presence of a highly repeated region (homopolymer region).

Briefly, DNA samples were diluted to 10 ng/μL, and subsequently, the target regions were amplified by PCR using the Ampliseq kit (Illumina) and the custom panel containing the specific primers of the genes listed above. The generated amplicons were partially digested with the FuPa reagent and then the adapter (i5/i7) indexes (Ampliseq UD indexes, Illumina) necessary for sequencing were added by PCR. A clean-up of the libraries was then carried out with the AMPure XP magnetic beads with a subsequent amplification of the libraries to ensure an adequate quantity of sequences for the Illumina system. In the next phase, a second clean-up has been performed and then all libraries were quantified by the QUBIT fluorometer (Invitrogen). All the generated libraries were checked on the bioanalyzer 2100 (Agilent) to evaluate their quality before NGS sequencing. As expected, each library presents a correct profile of the amplicons with an average size of approximately 390 bp.

In the final step, each library was diluted to 10 nM in Low TE, and then all libraries were pooled at 10 nM and diluted to 4 nM. Finally, the pooled libraries were denatured and diluted at 12 pM and loaded on a MiSeq Reagent Nano Kit v2 (300-cycles) cartridge with 12.5 pM PhiX (1.5%) sequencing control V3 (Illumina). Libraries were then sequenced on Illumina MiseqDX (Illumina) instrumentation in RUO mode. The whole process also required the use of a Veriti 96 thermal cycler (Applied Biosystem) for the amplification of the libraries and the Dynamag-96 side (Invitrogen) for the clean-up of the libraries.

NGS performance was evaluated using the DNA Amplicon module (Illumina). The qualitative analysis of the NGS run shows optimal metrics that have passed all the quality controls of the primary analysis including %Q30>84.02, a percentage (%) of PF (Passing Filter) equal to 94.07% with a yield of approximately 442 Mb. The cluster density is also in specification with 1.091 K. Furthermore, the percentage of PhiX loaded turns out to be 1.36% with an error rate of 0.47%. The secondary analysis of the NGS run for each sample in terms of coverage and uniformity results in optimal values with an average Amplicon coverage greater than 1,000 and a uniformity greater than 96%. Typically, the average coverage of 100–300 with a coverage (uniformity of coverage) of 90% of the amplicons is sufficient for an NGS analysis of germ samples. In our analyses, we obtained an average coverage (Amplicon mean coverage) greater than 1,000 and uniformity of coverage greater than 96% for each sample; these are optimal values to proceed with a correct tertiary analysis (variants of biological significance/germline variants).

Data (VCF files) were obtained using the DNA Amplicon (Illumina) module and analyzed by Evai (Engenome, https://evai.engenome.com/#app/analysis/11919), Variant Interpreter (Illumina, https://variantinterpreter.informatics.illumina.com/registry/cases), and Varsome Clinical (https://varsome.com). Briefly, the DNA Amplicon module was used to align the reads to the reference genome (hg19) and then to run to search for germline variants in the targeted regions. Variant classification is based on the current ACMG standards and guidelines. All variants were reviewed using VARSOME (https://varsome.com), the Osteogenesis Imperfecta Variant Database (oi.gene.le.ac.uk), dbSNP (http://www.ncbi.nlm.nih.gov/projects/SNP/), and Clinvar (https://www.ncbi.nlm.nih.gov/clinvar/).

The software VarSeq 2.3.0 (Golden Helix) was also used to confirm PVs detected in this study.

In this study, we enrolled 10 patients from seven different families, comprising 7 male and 3 female patients. The age range was between 4 and 57 years, and the cohort included six pediatric (<18 years) and four adult patients. Clinical information, including age, gender, family history, OI type, and non-skeletal features like blue sclera and hearing loss, and other clinical details were collected ( Table 1 ). In addition, X-ray radiography of the upper and lower extremities was examined for patients 22/21 and 02/22 ( Supplementary Image 1, 2 ).

The proband 21/21 (III.1) was an 18-year-old Italian man, the only son of non-consanguineous parents, a normal pregnancy lasting 9 months, and a Cesarean section due to severe joint limitation, weighing 3.16 kg and artificial feeding. The mother has OI, diagnosed at the age of 6. The patient has suffered from numerous fractures with functional outcomes. X-ray radiography of the upper and lower extremities was examined for patient 22/21 ( Supplementary Image 1 ).

The proband 22-21 (III.2) was a 6-year-old Italian boy. The historical medical background is as follows: the proband was the second-born child, with a father (patient 23/21, II.2) who had OI and a history of limb fractures. The proband’s uncle (II.4) and grandmother (I.2) were also affected by OI, while the proband’s mother and older brother were healthy.

Proband 24/21 (II.1) was a 15-year-old boy, the only son of non-consanguineous parents ( Figure 1 ). The mother (patient 25/21: I.2) was affected by OI. His biological father died at 29 years of age due to cardiac ischemia.

The proband 04-22 (II.2) was a 8-year-old male subject. He was born as the third child to a mother aged 28, who unfortunately passed away due to uterine cancer. The family history includes a younger sister (II.1) aged 4, exhibiting apparent good health; an older half-sister (II.4) aged 10, also in apparent good health; and an older half-brother aged 14 (II.3), conceived with a previous husband, and similarly in apparent good health. The father, aged 38, showed no notable health concerns.

The proband 05/22 (II.1) was a 15-year-old Italian boy. The patient was admitted due to a right olecranon fracture associated with OI. Additionally, the patient’s 57-year-old father is also affected by OI.

Proband 01/22 (II.1) was a 14-year-old girl who was the firstborn in her family. She was diagnosed with OI and exhibited blue sclerae since the age of 2. The proband experienced her first fracture in the left tibia at 1 year of age, subsequently followed by fractures in the right ulna and tibia at 2 years, and the right metatarsal at 3 years old. There were no family members affected with OI.

Subject 02/22 (III.1) was a 4-year-old Italian girl with OI, and her mother, aged 27 (II.2), and her grandmother (I.2) also exhibited the same condition. Multiple fractures were reported during her infancy. At the age of 2, she was admitted to the Orthopedics Department at the hospital due to an unprovoked fracture in the right tibia and fibula. Additionally, during the same month, a metaphyseal fracture in the right femur was treated by immobilizing it with a cast. X-ray radiography of both the upper and lower extremities was conducted for patient 02/22 ( Supplementary Image 2 ).

Ten patients from seven different OI families were analyzed by a targeted NGS panel for the common OI-related genes including COL1A1, COL1A2, FKBP10, IFITM5, PPIB, SERPINF1, WNT1, BMP1, CRTAP, P3H1, and SERPINH1. Among the 10 OI-positive patients, 7/10 PVs were found in the COL1A1 whereas 3/10 were in the COL1A2 gene. The PVs identified in COL1A1 were c.1767 + 1G>A, c.658C>T, c.2890_2893del, c.517G>T, and c.3887del, whereas in COL1A2, they were c.1009G>A and c.596G>T. Therefore, the majority of PVs were found in COL1A1 (n = 5), followed by COL1A2 (n = 2) ( Table 2 ). No likely pathogenic (LPVs) or PVs were detected in FKBP10, IFITM5, PPIB, SERPINF1, WNT1, BMP1, CRTAP, P3H1, and SERPINH1 genes.

Proband 21/21 (Family 1, III.1), an 18-year-old male patient affected by OI, was found to have the heterozygous PVs c.1767 + 1G>A in the COL1A1 gene ( Figure 1 ; Table 2 ). The proband’s mother (II.2) also affected by OI did not receive any genetic testing. The splice-site COL1A1 variant near exon 25 of 51 (c.1767 + 1G>A) detected in patient number 21/21 has been reported to ClinVar as pathogenic (VCV001380204.2). This splice-site variant disrupts the triple-helix domain of COL1A1 and has been associated with OI in several studies (18–21, 23).

The proband 22/21 (Family 2, III.2) is a 6-year-old boy affected by OI, with his father 23/21 (II.2), uncle (II.4), and grandmother (I.2) also affected by OI. Patients 22/21 and 23/21 were found to have the heterozygous PVs c.658C>T (p.Arg220ter) in the COL1A1 gene ( Figure 1 ). The COL1A1 nonsense variant c.658C>T (p.Arg220ter) in exon 9 of 51 found in patients 22/21 and 23/21 was classified as pathogenic in Clinvar (VCV000425637.13) and described to be associated with OI in several studies (24, 32–35).

Patient 24/21 (Family 3, II.1) and his mother 25/21 (I.2) both affected by OI were found to have the heterozygous variant c.2890_2893del (p.Pro964Valfs*143) in the COL1A1 gene. To our knowledge, this COL1A1 frameshift deletion in exon 40 of 51 has not been reported previously as a PV or as a benign variant, and it is predicted to cause loss of normal protein function through protein truncation caused by a frameshift mutation.

Patient 01/22 (Family 6, III.1) affected by OI was found to have the pathogenic heterozygous variant c.517G>T (p.Gly173Ter) in the COL1A1 gene. This sequence change in exon 6 of 51, which creates a premature translational stop signal (p.Gly173*) in the COL1A1 gene, has been observed in individuals with OI (36) and ClinVar contains an entry for this variant (Variation ID: 35924).

Patient 02/22 (Family 7, III.1) affected by OI who had clinical manifestations typical of OI, was a heterozygous carrier of the c.3887del (p.Gly1296ValfsTer35) in the COL1A1 gene. The proband’s mother (II.2) and grandmother (I.2) also affected by OI did not receive any genetic testing. Currently, this COL1A1 frameshift variant in exon 49 of 51 has not been previously described in publications and has not been registered in any databases.

The proband 04/22 (Family 4, II.2), an 8-year-old boy who had clinical manifestations typical of OI, was a heterozygous carrier of the c.1009G>A (p.Gly337Ser) in the COL1A2 gene ( Figure 1 ). The proband’s mother (Family 4, I.2), who died due to ovarian cancer, and two more proband’s sister (Family 5, II.1 and II.4) and one proband’s half-brother (Family 4, II.4), who were 10, 4, and 14 years old, respectively, and who were referred to be healthy, did not receive any genetic testing. This COL1A2 missense variant in exon 19 of 52 has been observed in individuals with OI types I, III, and IV (24, 29, 30, 33, 37–43).

The proband 05/22 (Family 5, II.1) is an 8-year-old boy affected by OI, and his father 06/22 (Family 5, I.1) also had clinical manifestations typical of OI ( Figure 1 ; Table 2 ). After NGS genetic analysis, they were both found to be heterozygous carriers of the c.596G>T (p.Gly199Val) missense variant in exon 13 of the COL1A2 gene ( Figure 1 ). Notably, the PV was present only in affected patients. Currently, this COL1A1 missense variant has not been previously registered in any databases and has not been described in publications.

To our knowledge, the COL1A1 frameshift variant c.2890_2893del (p.Pro964ValfsTer143) identified in patient 24/21 and his mother 25/21 has not been reported previously as a PV. The p.Pro964Valfs*143 variant is novel in gnomAD and 1kG database, and it is predicted to cause loss of normal protein function through protein truncation caused by a frameshift mutation. The COL1A1 frameshift variant identified here was classified as likely pathogenic based on the American College of Medical Genetics (ACMG) criteria (PVS1 and PM2) (44) ( Table 3 ).

To our knowledge, the frameshift deletion c.3887delG (p.Gly1296Valfs*35) in COL1A1 found in patient 02/22 has not been reported previously as a PV or as a benign variant as well. The p.Gly1296Valfs*35 variant is novel in gnomAD and 1kG database, and it is predicted to cause loss of normal protein function through protein truncation caused by a frameshift mutation. This variant is a frameshift variant that occurs in an exon of COL1A1 upstream where nonsense-mediated decay is predicted to occur. The COL1A1 frameshift variant identified here was classified as likely pathogenic based on the ACMG criteria (PVS1, and PM2) (44) ( Table 3 ).

The missense variant c.596G>T(p.G199V) in the COL1A2 gene detected in patients 05/22 and his father 06/22 was not observed in the large population cohorts of gnomAD or 1kG database (Genome Aggregation Database and Genomes Consortium) ( Figure 2 ). In addition, the computational tools used to evaluate the possible pathogenicity of this variant showed that the c.596G>T COL1A2 variant is predicted to be damaging by both SIFT and PolyPhen2. The COL1A2 missense variant identified here, not previously described in any database including HGMD, LOVD, and ClinVar, was classified as likely pathogenic based on the ACMG criteria (PM1, PM2, PP2, PP3, and PM5) (44) ( Table 3 ).

Importantly, the c.596G>T (p.Gly199Val) variant on exon 13 of COL1A2 is not predicted to disrupt the existing acceptor splice site 2 bp upstream by any splice site algorithm such as GeneSplicer, MaxEntScan, NNSplice, PWM, and VarSeak ( Supplementary Image 3 ), whereas HSF PRO showed a significant alteration of ESE/ESS motifs ratio, suggesting that this mutation could have a potential impact on splicing ( Supplementary Table 1 ). A PV c.596G>A (Gly199Asp) at the same amino acid residue of COL1A2 was associated with OI, confirming that Glycine 199 (G199) is an essential residue for COL1A2 function ( Figure 2 , Clinvar VCV000496615.30, dbSNP 1554395833). Importantly, the glycine residue at codon 199 of COL1A2 is conserved in all mammalian species ( Supplementary Image 4 ). Snapshots of the Integrative Genomics Viewer (IGV) showing all the germline variants identified in this study can be found in the Supplementary Material ( Supplementary Image 5 ).