dentification of ACO2 variants in individuals with ION Cohorts including about 1000 molecularly undiagnosed ION individuals from France, Germany, Italy, Spain, Belgium and the UK were screened by targeted sequencing for the presence of ACO2 variants. Among individuals with a single ACO2 variant, we selected cases with an MAF lower than 0.0001, whereas in individuals with biallelic heterozygous composite or homozygous variants, we selected the cases with an MAF lower than 0.005 for each variant. This led to the identification of 50 individuals harbouring 1 of 43 different heterozygous mutations, among which 29 were novel, as well as 11 individuals with biallelic mutations, among which 8 were novel. A compilation of these variants and the associated amino acid changes are displayed on ACO2 primary sequence and 3D structure in Fig. 1A and C and listed in the Supplementary Table 1. Notably, four dominant variants p.(Arg474Gly), p.(Trp603Ser), p.(Ser669Leu) and p.(Arg671Trp) were identified in two unrelated individuals and p.(Arg671Gln) in four unrelated individuals, while one recessive variant, p.(Leu74Val), with an MAF value of 0.00369, was found in five unrelated ROA individuals, and also formerly described in eight individuals with syndromic ACO2 presentations. This p.(Leu74Val) variant was associated once with p.(Gly661Arg) in the P56 individual, a combination already reported in an ROA family.6 Segregation analysis was performed where available. In DOA cases, all affected relatives carried the variant, whereas some relatives in the families of patients P6, P8, P9, P13 and P33 carrying the variant were asymptomatic, as occasionally observed in families with OPA1 mutations.15,41 Moreover, segregation analysis revealed de novo mutations in the P15 and P18 index cases. Analysis of the types of mutations identified in ACO2 (Fig. 1B) revealed that two-third of all variants were missense mutations, irrespective of the dominant or recessive mode of inheritance, while nonsense, frame-shift, splice-site mutations and deletions accounted together for the third part of all mutations identified, without significant difference between the distribution of the mutation types between dominant and recessive cases (P-value: 0.80). Notably, the identification of null alleles in dominant cases suggests that ACO2 haploinsufficiency contributes mainly to DOA pathophysiology.