Abstract

Objective

Osteogenesis imperfecta (OI) consists of a group of phenotypically and genetically heterogeneous connective tissue disorders that share similar skeletal anomalies causing bone fragility and deformation. The aim was to investigate the molecular genetic etiology and determine the relationship between genotype and phenotype in OI patients using targeted next-generation sequencing (NGS).

Methods

A targeted NGS analysis panel (Illumina TruSight One) containing genes involved in collagen/bone synthesis was performed on the Illumina Nextseq550 platform in patients with a confirmed diagnosis of OI.

Results

Fifty-six patients (female/male: 25/31) from 46 different families were included. Consanguinity was noted in 15 (32.6%) families. Based on Sillence classification 18 (33.1%) were type 1 OI, 1 (1.7%) type 2, 26 (46.4%) type 3 and 11 (19.6%) type 4. Median body weight was -1.1 (-6.8, - 2.5) standard deviation scores (SDS), and height was -2.3 (-7.6, - 1.2) SDS. Bone deformity affected 30 (53.5%), while 31 (55.4%) were evaluated as mobile. Thirty-six (60.7%) had blue sclera, 13 (23.2%) had scoliosis, 12 (21.4%) had dentinogenesis imperfecta (DI), and 2 (3.6%) had hearing loss. Disease-causing variants in COL1A1 and COL1A2 were found in 24 (52.1%) and 6 (13%) families, respectively. In 8 (17.3%) of the remaining 16 (34.7%) families, the NGS panel revealed disease-causing variants in three different genes (FKBP10, SERPINF1, and P3H1). Nine (23.6%) of the variants detected by NGS panel had not previously been reported and were also classified as pathogenic based on American College of Medical Genetics guidelines pathogenity scores. In ten (21.7%) families, a disease-related variant was not found in any of the 13 OI genes on the panel.

Conclusion

Genetic etiology was found in 38 (82.6%) of 46 families by targeted NGS analysis. Furthermore, nine new variants were identified in known OI genes which were classified as pathogenic by standard guidelines.

Introduction

Osteogenesis imperfecta (OI) is a hereditary disease of connective tissue characterized by increased bone fragility and multiple fractures (1, 2, 3). OI is a rare disorder with a frequency of 1/15,000-20,000. This generalized connective tissue disorder has an important effect on bone structure, leading to skeletal fragility and developmental delay. Clinical severity varies between the types of OI and additional features such as dentinogenesis imperfecta (DI), blue sclerae, short stature, hearing loss, and cardiac malformations may be present (4). Variants in the collagen genes are responsible for approximately 85-90% of OI. Patients with OI have heterogeneous clinical and genetic features (5). Thus, clinical diagnosis is insufficient for optimal diagnosis, management, prognosis, and genetic counseling. In recent years, new OI types have been discovered with the development of improved genetic analysis techniques. The disorder can also be caused by variants of genes related to collagen structure and function (6, 7). The most recently identified genes today are characterized by primary defects in osteoblast differentiation (2, 3, 6, 8).

Advances in next-generation sequencing (NGS) technology have enabled the discovery of novel genes and pathogenic variants related to OI (2, 9, 10). Molecular diagnosis is useful for early diagnosis, prognosis, identifying other individuals in the family with the same variants, and deciding on optimal treatment based on the published evidence (2, 9, 10, 11).

In this study, the aim was to investigate the molecular genetic etiology of OI using a targeted NGS panel and to determine the genotype-phenotype relationship in OI patients, and the effectiveness of this genetic panel for diagnosis.

Methods

Study Group

A cohort of clinically and/or radiologically diagnosed OI patients followed in Ege University Faculty of Medicine Pediatric Endocrinology and Diabetes Department were included in the study. Inclusion criteria were patients between 0-18 years of age with unknown molecular genetic etiology. Patients having any genetic disease other than OI that could cause bone fragility and other chronic diseases or patients with fragile bone syndrome due to medication, such as steroids or chemotherapy, were excluded.

Demographic data (age, gender, consanguinity, family history), clinical features (OI subgroup, frequency of annual bone fractures, treatment procedure and response), physical examination findings (bone deformities), and bone radiography findings were obtained from hospital records. Patients’ weight and height and their standard deviation (SD) scores (SDS) were calculated based on Turkish standards (12, 13).

The study was approved by the Ethics Committee of Ege University Faculty of Medicine (ethic committee number: 18-3.1/55, date: 20.03.2018), and samples from the patients were obtained in accordance with the Helsinki Declarations. Written informed consent for molecular analysis was obtained from all cases or their parents/guardians.

Molecular Analysis

Genomic DNA samples were extracted from leukocytes from 1 mL of peripheral blood obtained from all patients using the QIAamp DNA Blood Mini Kit (QIAGEN, Hilden, Germany) in accordance with the manufacturer’s instructions. DNA quality and quantity were assessed using a NanoDrop 2000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA). For sequence analysis, a targeted NGS panel (TruSight One Panel by Illumina^®) including 13 genes (COL1A1, COL1A2, IFITM5, SERPINF1, CRTAP, P3H1, PPIB, SERPINH1, FKBP10, SP7, BMP1, MBTPS2, PLOD2) associated with OI was used.

Data Analysis

Sequencing data was analyzed using Illumina VariantStudio software and IGV (Integrative Genomics Viewer). Firstly, 13 genes known to be responsible for OI were analyzed. Variants in these genes with a frequency of less than 0.5% in public databases including NCBI dbSNP build155 (http://www.ncbi.nlm.nih. gov/SNP/), 1000 Genomes Project (http://www.1000genomes.org/), gnomAD (https://gnomad.broadinstitute.org/) and NHLBI Exome Sequencing Project Exome Variant Server (http://evs.gs.washington.edu/EVS/) were selected. The impact of the variants on the protein structure was identified using several in silico prediction tools, such as MutationTaster, PolyPhen-2, and SIFT. The conservation of residues across species was evaluated by the PhyloP algorithm and GERP (14, 15, 16, 17). The pathogenicity of all variants identified was classified according to the American College of Medical Genetics (ACMG) recommendations. The pathogenity scores were obtained from the https://www.acmg.net/ website. The ACMG Guidelines were established by clinicians and clinical lab directors who are experts in clinical genetics and members of the ACMG, the Association for Molecular Pathology (AMP), and/or the College of American Pathologists. Franklin Genoox software and database was used for ACMG Classification. There are 28 criteria in the ACMG guidelines. During variant interpretation, variants are classified into five tiers: pathogenic (P), likely pathogenic (LP), uncertain significance (VUS), likely benign, and benign (B), depending on the applicable criteria. These criteria can be classified by the weight and type of evidence indicated by each criterion. The 28 criteria can be classified into eight types: population data, computational data, functional data, segregation data, de novo data, allelic data, other databases, and other data, depending on the source of evidence (18).

Confirmation

The most likely disease-causing variants, identified by data analysis, were confirmed using direct Sanger sequencing on ABI PRISM 3130 DNA analyzer (Applied Biosystems) and Big Dye Terminator Cycle Sequencing V3.1 Ready Reaction Kit (Life Technologies) (Fisher Scientific - Göteborg - Sweden).

Statistical Analysis

Analysis was conducted using Statistical Package for the Social Sciences for Windows, version 25.0 (IBM Inc., Armonk, NY, USA). Descriptive statistics are reported as mean±SD for normally distributed variables and median (range) for skewed data. Groups were compared using independent samples t-test for normally distributed variables and the Mann-Whitney U test for skewed data. Trends across time were analyzed using linear polynomial contrasts (ANOVA). A p<0.05 was considered statistically significant. No adjustment was made for a multiplicity of statistical tests.

Results

Clinical Manifestations

Fifty-six patients (female/male: 25/31) from 46 different families were included in the study. In 15 (32.6%) families, consanguineous marriage was noted. The mean age of the cohort on admission was 4.5±3.7 years, median body weight was -1.1 (-6.8, - 2.5) SDS, and height was -2.3 (-7.6, - 1.2) SDS. Based on the actualized Sillence classification (Table 1), 18 (33.1%) patients were considered to be type 1, 1 (1.7%) type 2, 26 (46.4%) type 3, and 11 (19.6%) type 4 (19). Bone deformity was detected in 30 (53.5%) of the patients, while 31 (55.4%) were evaluated as mobile. Thirty-six (60.7%) patients had blue sclera, 13 (23.2%) had scoliosis, 12 (21.4%) had DI, and 2 (3.6%) had hearing loss.

Molecular Analysis Findings

Sequence analysis of the COL1A1 and COL1A2 genes revealed heterozygous variants in 24 (52.1%) and 6 (13%) families, respectively. The (NM_000088.4(COL1A1):c.3677A>G/p.(Asp1226Gly)/rs1319157667) variant was detected together with the c.2296G>C variant in COL1A2 in one patient. The variant detected in the COL1A1 gene was also found in the asymptomatic father of the patient in segregation analysis. Based on this family history, she was excluded from the COL1A1 group. The remaining 16 families were molecularly analyzed using the NGS panel, and in 8 (17.3%) families, a disease-causing variant in three different genes (FKBP10, P3H1, and SERPINF) was identified. Nine (23.6%) of the detected variants in all genes have not been previously reported and were considered to be deleterious based on prediction tools. Following a two-step NGS-based molecular analysis, a molecular diagnosis was achieved in 38 (82.6%) families in the study group.

Genotype-phenotype Relations

Fifteen of the COL1A1 variants were boys, and 14 were girls. The mean age at admission was 4.69±3.66 years, and weight and height SDS were -0.73±1.39 and - 2.41±4.45, respectively. The distribution of clinical diagnosis was as follows: 13 (44.8%) type 1; 1 (3.4%) type 2; 10 (34.4%) type 3; and 5 (17.2%) type 4. Recurrent pathological fractures were detected in 25 (86.2%) of the patients, and deformity of extremities in 7 (24.1%) patients. Six (20.6%) patients were mobile with help or had the ability to sit. The rest of the patients were completely mobile. Twenty-five patients (86.2%) had blue sclera, and 8 (27.5%) had DI.

The mean age of admission of those with variants in the COL1A2 gene (n=6, 4 girls) was 5.17±3.45 years, weight SDS was -2.86±2.33, and height SDS was -3.07±1.01. The distribution of clinical types was four (66.6%) type 3 and 2 (33.3%) type 4. In all patients, 2 or more recurrent pathological bone fractures and deformities were detected. In 5 (50%) patients, blue sclera and DI was found in one (16.6%) patient.

Biallelic variants in the SERPINF1 gene were detected in four patients (patients 36, 37, 38, and 39) from three families. In one case, compound heterozygous variant c.80dupA/c.907C>T was present, and this patient’s OI phenotype was compatible with type 3 with severe deformities, recurrent fractures, and short stature. However, family segregation was not performed in this patient.

A homozygous novel c.446T>G p.(Leu149Arg) variant was detected in the P3H1 gene in a 0.2-year-old patient with a history of consanguineous marriage. The patient’s weight SDS was -0.36, and height SDS was -2.12.

A homozygous c. 15dupC variant was identified in the FKBP10 gene in a male patient with Bruck syndrome and clinically type 4 OI. Her parents were heterozygous for the same variant. The same variant was demonstrated in the sibling of the index case with a similar phenotype. These cases had severe OI and congenital contractures of large joints, short stature, and scoliosis.

Genotype and phenotype characteristics of patients with OI-related variants are given in Tables 2 and 3.

Discussion

The present study investigated the molecular etiology of 56 clinically diagnosed OI patients from 46 different families using an NGS panel, including a total of 4800 genes, including 13 genes related to OI. Genetic etiology was found in 38 (82.6%) of 46 families by targeted NGS analysis with this TruSight One Panel. Such targeted gene panels are extremely reliable and validated and can be used in a wide range of indications for genetic diseases. Panels containing the genes of most diseases inherited as Mendelian in humans, such as the Illumina TruSight One Panel, are also referred to as “clinical exomes”. This expression should not be confused with whole exome sequencing (WES) because approximately 20,000 genes detected in mankind are analyzed by WES analysis, while the clinical exome only contains genes associated with disease in humans. The advantage of the TruSight One panel compared to WES is that it is easier to analyze the results, and the cost is lower (20).

It is generally accepted that in-frame, partial deletions in the COL1A1 or COL1A2 genes can result in a lethal or severe OI phenotype when the resulting abnormal protein is not rapidly degraded but instead is incorporated in the triple helix exerting a dominant negative effect. In a study by (21), multiplex ligation-dependent probe amplification analysis was performed in the analysis of the COL1A1 gene in a group of 106 index patients. These authors found seven patients with deletion of the complete COL1A1 gene on one allele. In the present study, we did not evaluate the gross deletions and duplications in exons because this was outside the scope of the analysis methods. Although these types of variants are generally identified at low rates, they may be more relevant for the subgroup without any variants, in terms of disease severity through haploinsufficiency of COL1A1 and COL1A2 genes.

Consanguineous marriage was present in 28.5% of patients, and 39.3% had a family history of OI. Consanguineous marriage may lead to a high rate of autosomal recessive variants being found. In a study in India with seven patients of consanguineous marriage, SERPINF1, PPIB, and CRTAB mutations were detected (22). In a study evaluating COL1A1 and COL1A2 gene variants in 364 patients of Italian origin, the rate of positive family history rate was 57.7% (23). However, this increased rate of more than half may have been present because, with the exception of the two autosomal dominant gene variants in COL1A1 and COL1A2, other types of OI were not studied. Family history rates were reported to be 53% in the Korean population (24), and 32-33% in different societies (25, 26, 27). The reason for these differences may be due to genetic differences in societies, variations in the genetic analysis methods used, and the changing frequency of de novo variants. Furthermore, the frequency of consanguineous marriage and/or founder variants in a society will lead to variability in the distribution of genes responsible for OI.

The clinical finding of blue sclera is one of the distinctive clinical features of OI and is frequently observed in patients with type 1 OI. Patients with type 3 and 4 OI may have blue sclera at birth, but the bluish color disappears with increasing age (2, 8, 28). In our study group, 34 (60.7%) of the patients had blue sclera.

Bone fractures and deformities in OI usually occur at an early age and are often caused by repeated bone remodeling in long bones (2, 29). This affects patients’ growth, functional status, and mobility. In our study, 20.6% of patients with the COL1A1 variant and all of those with the COL1A2 variant had difficulty walking. In 24.1% of patients with the COL1A1 variants, and all of those with the COL1A2 variants, deformities were detected in the extremities. Mohd Nawawi et al., (30) showed that 63.6% of all OI patients had bone deformities at the age of nine years and needed help to walk. Studies have shown that bone deformities are more common in patients with qualitative variants than quantitative variants (2, 29). Hald et al. (31, 32) showed that OI patients with quantitative defects had normal protein structure in bone, despite collagen deficiency. This allows bone mineralization and thus leads to fewer breakages than qualitative defects (11, 31).

DI has been reported in more frequently type 3 and rarely type 1 OI (33, 34). DI was detected in 12 (21.4%) of the patients in the present study. The clinical diagnosis of the patients with DI was as follows: 5 patients (41.6%) were type 1, 4 (33.3%) type 3, and 3 (25%) type 4. In another study, DI was reported to be more frequent in patients with more severe clinical types (type 3 and 4) than in moderately affected groups (type 1) (34). Those with a qualitative variant, that is a problem in collagen structure, are more at risk of developing DI. Structurally abnormal collagen affects the development of dental germ cells in the predentin during the mineralization process (35).

In the present study, 63.1% had a variant in the COL1A1 gene, 13.1% in the COL1A2 gene, and 2.6% in both genes; in total, 78.8% of patients had variants in these two genes. In 11 (19.6%) out of 21 patients without variants in these genes, by NGS analysis, three other gene (SERPINF1, FKBP10, and P3H1) variants were detected. Three (7.8%) families had FKBP10, 3 (7.8%) families had SERPINF1, and 2 (5.2%) families had the P3H1 variant. Abali et al. (36) studied 89 patients with OI. Similarly to our study, these authors reported a majority (61.4%) having variants in COL1A1 and COL1A2 genes, while much lower proportions had varinats in other OI-associated genes: 5 (5.6%) with FKBP10; 2 (2.2%) LRP5; 1 (1.1%) P3H1; 1 (1.1%) CRTAP; 1 (1.1%) BMP1; and 1 (1.1%) SPARC variant.

COL1A1 and COL1A2 Gene Variants

Variants in COL1A1 and COL1A2 encoding type 1 collagen are responsible for most of the etiology in OI. In the present study, 62.4% of patients had variants in these two genes, similarly to previous repots where variants in these two genes were responsible for 51-73% of the disease (25, 30, 37, 38). Moreover, variants in the COL1A1 gene were detected more frequently in both the present study and earlier studies than the COL1A2 gene (4, 25, 27, 30).

In one (3.3%) case (patient 7), a heterozygous variant was detected in both COL1A1 c.3677A>G p.(Asp1226Gly) and COL1A2 c.2296G>C p.(Gly77Arg) genes. The variant detected in the COL1A1 gene was also found in the asymptomatic father of the patient in segregation analysis and we made the decision to discount the VUS variant in the COL1A1 gene as causing the phenotype. The variant in the COL1A2 gene was previously reported as pathogenic. Therefore, this variant may be responsible for the clinical findings in this case, who had severe clinical type with recurrent fractures and severe deformities. No cases carrying variants in both these two genes simultaneously have been reported previously. However, Ji et al. (39) reported a case with a severe clinic variant in COL1A1 and SERPINF1. Oligogenic inheritance should also be considered in cases with severe clinical features.

In patient 5, the variant c.1699C>T p.(Pro567Ser) in COL1A1 gene. The variant c.1699C>T in COL1A1 was detected once in the GME Variome database and once in the Turkish Variome database. The frequency of the variant in GMA Variome was 0.05% and it was 0,02% in the Turkish Variome. Both frequencies are less than 0.06%, which is the cut-off level for ACMG-PM2 criteria for this gene, and this is supporting evidence for the possibly pathogenic nature of the variant. This variant has not been reported in association with OI or any disease.

P3H1 Gene Variants

A homozygous c.446T> G p.(Leu149Arg) variant was detected in the P3H1 gene in two unrelated female patients (Patients 40 and 41) who were admitted with recurrent fractures before two months of age. This variant was thought to be disease-causing in in silico analysis and was not previously reported. Recurrent fractures continued with severe clinical phenotypes. In the literature, clinical OI type VIII due to P3H1 gene variants have moderate/severe phenotypic features (31). In the West African community, in the P3H1 gene, a relatively high c.1080+1G>T carriage was detected: 1/240. The homozygous form was associated with the perinatally lethal form of OI. This variant was thought to be a founding variant (40, 41, 42) but was not observed in any of the 56 patients in the present study.

SERPINF1 Gene Variants

In some populations, SERPINF1 and CRTAP variants have been reported to be responsible for recessive OI types. Some variants have been suggested as causing a “founder” effect (43).

In the present study, four different variants were detected in three patients from different families, and another variant was detected in two siblings. Patients with SERPINF1 variants had a more severe clinical picture and early admission. Compound heterozygous c.80dupA p.(Glu28Glyfs*37) / c.907C>T p.(Arg303Ter) variant was found in an infant (patient 36) with fractures from birth, widespread deformities, and severe short stature. In in silico analyses, both variants were disease-causing. No blue sclera, DI, or hearing loss was detected. In the follow-up, despite treatment, his fractures recurred, his deformities increased, and independent mobilization never developed. Another patient (patient 39) who presented with a severe clinical picture at the age of 6 months had a homozygous c.988C>T p.(Gln330Ter) variant. Vertebral and lower extremity fractures were present. Similar to these patients, most SERPINF1 gene variants previously reported have been due to frameshift and nonsense variants (44, 45). A missense homozygous c.317G>C p.(Arg106Pro) variant was found in a patient with a milder clinical picture who presented at 8.6 years with recurrent fractures without deformities. This variant was predicted to be a VUS by Franklin Genoox and Varsome programmes, by the ACMG 2015 criteria. Most of the predictions tool predicted that this variant will be pathogenic or VUS. This variant has low population frequency. Given this evidence, it was thought that this homozygous c.317G>C p.(Arg106Pro) variant detected in SERPINF1 might be responsible for the clinical picture in the patient. These variants impair production of circulating pigment epithelium-derived factor (PEDF) as well as loss of PEDF protein function (2). Rauch et al. (46) reported that measuring PEDF concentration in serum may be a potential marker in assessing patients’ clinical severity.

FKBP10 Gene Variants

A homozygous c.21dupC p.(Ser8Glnfs * 67) variant in the FKBP10 gene was detected in an 8.3-year-old girl (patient 45) who presented with congenital joint contractures, recurrent fractures, and chest deformity resembling Bruck syndrome. No consanguineous marriage was reported. FKBP10 gene variants have been associated with severe OI and Bruck syndrome (2, 11). In two brothers with Bruck syndrome, Shaheen et al. (47) reported a homozygous 8-bp insertion variant in the FKBP10 gene. Alanay and Krakow (48) reported that the patients in Shaheen et al.’s (47) study may have Bruck syndrome and that the clinical picture may be milder because they received bisphosphonate treatment. Researchers have reported that different variants in the FKBP10 gene can explain the variable severity of phenotypes.

Genotype-phenotype correlations in OI have been extensively studied over the years, with certain investigations revealing significant associations (49, 50). Notably, more severe phenotypes have been observed in patients harboring pathogenic variants in COL1A1 compared to those in COL1A2 (51). Mrosk et al. (52) suggested a robust correlation between genotype and the severity of phenotypes. They proposed a ranking based on phenotype severity as follows: P3H1, COL1A1, and COL1A2, respectively. In the present study, 60% of the patients with variants detected in the COL1A1 and COL1A2 genes, 50% of the patients with variants detected in the SERPINF1 and P3H1 genes, and 60% of the patients with variants detected in the FKBP10 gene had a severe phenotype. In addition, we identified a variant in a total of 15 affected individuals across seven families. Remarkably, the clinical types and features of cases with the same variant within these families were similar. On the other hand, c.1299+1G>A variant was detected in the COL1A1 gene in cases 11 and 12 from different families, and c.446T>G variant was detected in the P3H1 gene in cases 40 and 41. The clinical features of pateints 11 and 12 were similar. However, while case 41 was type 4 OI, case 41 was type 3 OI and case 41 showed more severe type features. The literature indicates that while varying phenotypes can exist within the same family, similarities can also exist among individuals from different families who share the same genetic variant (2, 53). The genotype-phenotype relationship in OI remains variable, as carriers of the same variant may develop diverse phenotypes. Furthermore, the factors influencing additional phenotypic differences have yet to be fully elucidated.

In the present study, no variants were detected in any of the genes covered by TruSight One used in the targeted NGS analysis in 10 (17.8%) patients. There may be more OI-associated genes or perhpas new candidate genes that were not covered by this panel. Targeted gene panels are highly efficient in the diagnosis of genetic disorders, which have genetic heterogeneity. This panel gave an 82% diagnostic yield. For most of the cases this high diagnostic yield was clinically useful but current technology allows us to perform WES-CNV at a comparable price. However, in most centers the capacity of the genetic laboratories is the main determinant of the genetic approach.

In one study, it was noted that an unusually high percentage of AR forms due to mutations in genes such as BMP1, FKBP10 were reported in their cohort of 50 patients. This highlights the utility of gene panel testing in a setting where specific mutations are known to be more common. WES can be particularly useful in cases where patients present with atypical features or where the targeted gene panel does not provide a definitive result. For example, in one patient, WES revealed no significant mutations, suggesting the presence of non-coding or complex inter-related genetic contributions to the disease that may have been missed by targeted panels. Targeted sequence analysis is often more practical and cost-effective when a patient’s clinical presentation strongly points to mutations in known OI-associated genes. WES is more comprehensive and can reveal unexpected mutations, but is also more resource intensive. The choice between these techniques may depend on clinical indications, resources and the possibility of atypical genetic contribution to the disease (53).

Study Limitations

There are some limitations to the present study due to the small size of the study population, single-center nature, and possible selection bias due to being only one tertiary center. Furthermore, the cases’ pedigrees were not considered because the study was conducted a long time ago. The exclusion of certain genes associated with osteogenesis imperfecta (CCDC134, CREB3L1, KDELR2, MESD, SPARC, TENT5A, TNEN38B, WNT1) also contributes to the study’s limitations.

Conclusion

This fairly comprehensive study demonstrated the clinical and molecular features of OI in a clinically diagnosed Turkish cohort. Genetic etiology was identified in 82.6% of 46 families with targeted NGS analysis. In addition, nine novel variants in OI-associated genes were identified. However, no new candidate gene related to OI could be detected by NGS analysis and eight genes known to be associated with OI were not analyzed (see limitations above). In patients where variants cannot be detected by targeted NGS, advanced genetic analysis, such as WES analysis may be planned. Finally, targetted panel studies in genetically heterogeneous diseases like OI are helpful for increasing the rate of variant detection.