Reduced penetrance of COL1A1/2 pathogenic variants linked with osteogenesis imperfecta: analysis of a large population cohort

Osteogenesis imperfecta (OI) is under consideration for inclusion in several genomic newborn screening initiatives, but its penetrance in clinically-unselected populations is currently unknown. It is an exemplar condition for evaluating penetrance in adult cohorts due to its relatively low mortality, variable expressivity and link to several large genes. Using genome sequencing data from ~500,000 adults in UK Biobank, we curated a set of rare pathogenic/likely pathogenic (P/LP) variants in COL1A1, COL1A2 and IFITM5 using annotations from gnomAD, ClinVar and SpliceAI. Analysis of summed read-count data from genome and exome sequencing led to exclusion of 16 mosaic variants with consistently low allelic balance of 3.4-35.9%. We identified 61 likely constitutive heterozygous P/LP variants in COL1A1 and COL1A2 (29 loss-of-function or splice variants and 32 missense) in 115 participants, with a mean age at recruitment of 55.1 years; no P/LP variants were identified in IFITM5. Phenotypes were assessed using ICD-10 codes, self-reports and heel bone mineral density (BMD). Overall disease penetrance was lower than anticipated: 40.7% for COL1A1 and 21.3% for COL1A2, potentially due to depletion of severe early-onset disease. When considering only truncating variants in COL1A1, disease penetrance increased to 73.1%, and 90% of individuals had reduced levels of circulating COL1A1 protein. For COL1A2, BMD data supported the low penetrance, whereas for COL1A1, data suggested the possibility of a subclinical phenotype. Overall, for P/LP missense variants (including those altering Gly-Xaa-Yaa repeats), the low penetrance observed suggests reliance on current ClinVar assertions to support pathogenicity may overstate OI risk in population screening.