Here we describe a family in which a novel heterozygous mutation in AFG3L2 co-segregates with ADOA. Interestingly, the identified p.G337E mutation localizes close to the AAA domain of AFG3L2, in contrast with those causing SCA28 and SPAX5, which clusterize in the proteolytic domain of the protein. Functional studies of the p.G337E mutation demonstrate its pathogenicity, as it strongly affects the processing of L-OPA1 leading to aberrant mitochondrial fragmentation. The defective mitochondrial dynamics is in line with most optic neuropathies exhibiting mitochondrial dysfunction in RGCs as underlying mechanism [12].
Other heterozygous/compound heterozygous mutations in AFG3L2 have been described in isolated cases with nonsyndromic optic atrophy (although not supported by functional studies) [13, 14], and in familial syndromic and non-syndromic optic atrophy [15, 16]. The present study further demonstrates that heterozygous mutations in AFG3L2 shall be considered as a genetic cause for ADOA in OPA1 and OPA3- negative cases.
The family described here has features of ADOA with no clinical evidence of spinocerebellar ataxia in any of the affected members but the proband, who experienced an acute episode of cerebellar ataxia. As proband’s clinical features, MRI and CSF investigations were consistent with relapsing remitting MS and the demyelinating lesions (including the cerebellar lesions) settled on treatment with Natalizumab, his cerebellar symptoms were most likely due to MS. However, it remains to be determined if these demyelinating neurological symptoms could be directly related to the AFG3L2 mutation. Although mouse studies have shown that deletion of AFG3L2 (either constitutive or in mature mouse oligodendrocytes) can cause myelin abnormalities [17, 18], no such phenotype has been described in humans before. Comorbidity cannot therefore be excluded in the proband.
The functional studies we conducted on the p.G337E mutation clearly prove its pathogenicity. Exogenous expression of p.G337E AFG3L2 in an Afg3l2 null background indicates that the mutant protein has no residual function. Indeed, the complete loss of L-OPA1 and mitochondrial fragmentation in this condition were comparable to those previously observed in Afg3l2 null MEFs [10]. In agreement, we found swollen mitochondria with altered cristae in the optic nerve of Afg3l2 null mice (VB and FM unpublished observation). Investigations in patient fibroblasts also revealed faster abnormal processing of OPA1 despite the heterozygous state of the mutation, with strong reduction of L-OPA1, accumulation of S-OPA1 and altered mitochondrial fusion. Interestingly, the decrease in L-OPA1 is significantly more pronounced in the proband compared to his affected mother, in line with the more severe phenotype, indicating that OPA1 processing might be considered as an outcome of disease severity in this form of ADOA. We also demonstrated that OPA1 processing is caused by strong OMA1 hyper-activation, which is reduced in amount in patients versus controls because of its faster autocatalysis. On the contrary, YME1L1 levels were not altered in patients, thus excluding a compensatory upregulation of YME1L1 on the final outcome on mitochondrial dynamics.
OPA1 processing is more severely compromised in this family compared to what we previously described in SCA28 and SPAX5-patient fibroblasts, where the levels of L-OPA1 were moderately reduced compared to controls, the accumulation of S-OPA1 was not appreciated, and the mitochondrial network presented shorter tubules, but not evident fragmentation [7]. Interestingly, the mutation we identified localizes close to the AAA domain, while most of those associated with SCA28 or SPAX5 affect the proteolytic domain, suggesting that mutations in different domains of this protein could differently affect its molecular function. Mutations localizing in the AAA domain of AFG3L2 can abolish ATP binding/hydrolysis and impact more severely on proteolytic activity, in agreement with a recent work in which ATPase and proteolytic activity of AFG3L2 carrying different mutations were assessed in vitro [19].
SCA28 and SPAX5 predominantly affect the cerebellum, while this novel AFG3L2 mutation predominantly affects the optic nerve. The aberrant OPA1 processing and severe mitochondrial fragmentation we observed, together with the fact that most ADOA patients carry OPA1 mutations, indicates that a fine control of mitochondrial dynamics is crucial for RGC survival. We may speculate that AFG3L2 mutations that predominantly and severely impact on OPA1 processing affect specifically RGCs, while those mostly impinging on other AFG3L2-related functions (oxidative phosphorylation and mitochondrial calcium homeostasis) affect Purkinje neurons in the cerebellum [10, 20, 21]. Purkinje neurons are selectively vulnerable to these defects, since they are characterized by a high oxidative metabolism and experience elevated calcium fluxes due to massive glutamatergic stimulations [22, 23].
In conclusion, our study broadens the spectrum of neurodegenerative diseases associated with AFG3L2 mutations and expands the genetic causes leading to ADOA, enforcing aberrant OPA1 processing as common mechanism for this disease.