Dominant mutations in MIEF1 affect mitochondrial dynamics and cause a singular late onset optic neuropathy

Inherited optic neuropathies are the most common mitochondrial diseases, leading to neurodegeneration involving the irreversible loss of retinal ganglion cells, optic nerve degeneration and central visual loss. Importantly, properly regulated mitochondrial dynamics are critical for maintaining cellular homeostasis, and are further regulated by MIEF1 (mitochondrial elongation factor 1) which encodes for MID51 (mitochondrial dynamics protein 51), an outer mitochondrial membrane protein that acts as an adaptor protein to regulate mitochondrial fission. However, dominant mutations in MIEF1 have not been previously linked to any human disease. Using targeted sequencing of genes involved in mitochondrial dynamics, we report the first heterozygous variants in MIEF1 linked to disease, which cause an unusual form of late-onset progressive optic neuropathy characterized by the initial loss of peripheral visual fields. Pathogenic MIEF1 variants linked to optic neuropathy do not disrupt MID51’s localization to the outer mitochondrial membrane or its oligomerization, but rather, significantly disrupt mitochondrial network dynamics compared to wild-type MID51 in high spatial and temporal resolution confocal microscopy live imaging studies. Together, our study identifies dominant MIEF1 mutations as a cause for optic neuropathy and further highlights the important role of properly regulated mitochondrial dynamics in neurodegeneration.

Inherited Optic Neuropathies (ION) are the most common mitochondrial diseases, leading to the irreversible loss of retinal ganglion cells, optic nerve degeneration and central visual loss [1,2]. They can be maternally transmitted by the mitochondrial genome in the case of Leber hereditary optic neuropathy (LHON, MIM#535000), or by mono or bi-allelic autosomal mutations. In dominant optic atrophy (DOA, MIM165500), it is now well established that mutations in genes involved in mitochondrial dynamics such as OPA1 (MIM#605290) are the main cause of the disease. More than 400 distinct pathogenic variants have been described in OPA1 [3][4][5][6], affecting the pro-fusion activity of this large dynamin-related GTPase resulting in defective mitochondrial fusion [7,8]. The implication of mitochondrial dynamics in DOA was further supported by the identification of dominant mutations in MFN2 (MIM#608507) [9] and OPA3 (MIM#606580) [10], two additional genes acting on mitochondrial dynamics.
More recently, we and others described cohorts of individuals affected by DOA caused either by mutations in SPG7 (MIM#602783) or in AFG3L2 (MIM#604581) [11][12][13], two genes encoding mAAA-proteases acting on mitochondrial fusion through the indirect control of OPA1 processing, and also involved in recessive hereditary spastic paraplegia type 7 [14] and dominant spinocerebellar ataxia 28 [15], respectively. In addition, homozygous mutations in YME1L1 (MIM#617302), which also disrupt OPA1 processing have been identified in individuals with optic atrophy and mitochondrial disorders [16]. Interestingly, DNM1L mutations resulting in excessive mitochondrial fusion have also been identified in patients with a similar ophthalmological presentation [17] (MIM603850), thus providing evidence that alterations of both fusion and fission compromise retinal ganglion cell survival. DNM1L encodes DRP1, another dynamin-related GTPase which requires a set of adaptor proteins including MID51, MID49, and MFF to be recruited on the mitochondrial outer membrane to exert and regulate its pro-fission activity [18][19][20]. So far, only biallelic nonsense MFF mutations (MIM#614785) have been identified in individuals with a severe encephalopathy associated with optic atrophy and peripheral neuropathy [21], while no mutation has been reported for optic atrophy in MIEF1 or MIEF2, which encode for MID51 and MID49, respectively.
Together, these observations prompted us to screen all known genes involved in mitochondrial dynamics in molecularly undiagnosed ION cases. Here, we identify the first heterozygous pathogenic disease-causing mutations in MIEF1, found in individuals with non-syndromic lateonset ION characterized by initial loss of peripheral visual fields. Importantly, ION disease-linked MIEF1 variants affect mitochondrial network dynamics, and ultimately highlight the crucial role of properly regulating this pathway in optic nerve physiology.
Two hundred individuals in France affected by an ION were included in this study. None of the individuals included had a molecular diagnosis after screening for pathogenic variants in OPA1, OPA3 and WFS1 exonic sequences and LHON-associated mtDNA mutations. Individuals were analyzed using a targeted resequencing panel of 22 genes, including those already published as causing ION, and genes involved in mitochondrial dynamics (Supplementary Table 1). After eliminating Fig. 1 Characterization of inherited optic neuropathy patients with dominant MIEF1 mutations. a Eye fundus of the right (RE) and left (LE) eye exhibiting severe optic disk pallor in MIEF1 individuals with inherited optic neuropathy (Patient 1 (left 2 panels); Patient 2 (right 2 panels)). b Evaluation of the visual fields revealed peripheral visual field defects in MIEF1 individuals with inherited optic neuropathy (Patient 1 (left 2 panels); Patient 2 (right 2 panels)). c Optical coherence tomography (OCT) recordings of the retinal nerve fiber layer (RNFL) at the papilla of the right (RE, left panel) and left (LE, right panel) eye from Patient 2 with inherited optic neuropathy frequent (allele frequency > 1/10.000) and nonpathogenic variants, according to prediction tools (Sift, Polyphen, and Mutation-Taster), we identified two individuals ( Fig. 1) harboring a MIEF1 heterozygous variant, which were confirmed by Sanger sequencing. The first individual harbored a c.718 T > A variant, not referenced in any database, and the second individual harbored a c.436C > T variant, referred to as rs778124994, with a frequency of 1.99e-5 in GnomAD database. Both variants were predicted to be damaging by the Sift and Polyphen programs and disease causing by Mutation Taster (Fig. 2a). These variants lead to the p.Y240N and the p.R146W amino-acid changes in MID51, respectively, with p.Y240N located in a DRP1 binding domain, and p.R146W located in another domain conserved within MID49, but without known function (Fig. 2b, c).
The two middle-aged women, 55 and 47 years old, carrying MIEF1 mutations were initially referred to ophthalmology departments for a sudden painless visual acuity impairment (see Methods for detailed patient descriptions). At initial assessments, neither of the two individuals had symptoms or immunological profiles compatible with Glaucoma, Neuro Myelitis Optica or myelin oligodendrocyte glycoprotein optic neuritis. At fundus examination, both individuals had normal retina, but presented pale and moderately excavated optic disks (Fig. 1a). Importantly, both had an alteration of peripheral visual fields, with a clear-cut bilateral superior field defect in the second individual, together with variable central visual alterations (Fig. 1b). In both individuals, the disease progressed to poor vision (Patient 1: RE: 1/ 20, LE: 18/20; Patient 2: RE: hand moving, LE: 4/20), asymmetrically in the first one and symmetrically in the second one. Accordingly, optical coherence tomography (OCT) scanning disclosed severe retinal nerve fiber layer loss in all quadrants in the second individual (Fig. 1c). No additional neurological or systemic symptoms were observed in either patient.
Thus, the optic atrophy that we report here is an original ophthalmological presentation based on three discriminating criteria. First, the visual loss was noticed during adulthood, rather than during the first two decades, as commonly observed in other DOA, although a latent alteration of the peripheral visual fields might have preceded the loss of visual acuity by several years, as observed in one individual. Secondly, both patients complained of a rather sudden painless loss of visual acuity, similar to those observed in maternally inherited LHON, although without optic disk elevation or edema, which is distinct from what is generally observed for other autosomal optic neuropathies, except for a recent report presenting a LHON-like patient with bi-allelic NDUFS2 mutations [22]. Thirdly, both cases examined here showed that the disease progressed from the peripheral to the central visual field, still preserving some visual acuity in the left eye of the first patient. This is in contrast with all the reports of ION cases, in which the alteration of the visual field occurs and evolves from the center to the periphery.
To assess the pathogenicity of the MIEF1 variants, we expressed wild-type and both mutants (p.Y240N and p.R146 W) of MID51 in HeLa cells. MID51 is embedded in the outer mitochondrial membrane, where it homooligomerizes to regulate mitochondrial fission [18,19]. Thus, we first examined whether p.R146W and p.Y240N mutants displayed a mitochondrial localization, using high spatial and temporal confocal microscopy in live cells. Expression of wild-type MID51 robustly localized to mitochondria (mEmerald-Mito) (Fig. 3a, b), consistent with its reported localization within the outer mitochondrial membrane [18,19]. This was also true for mutant MID51 p.Y240N and p.R146W proteins, which also localized to the mitochondrial network in live cells (Fig.  3a, b), demonstrating that these MIEF1 missense variants linked to optic neuropathy do not alter MID51's mitochondrial localization.
We next investigated whether MIEF1 variants disrupted the ability of MID51 to self-oligomerize. To test this, we immunoprecipitated myc-tagged wild-type and mutant MID51 proteins, and analyzed their oligomerization patterns. Wild-type MID51 formed monomers as well as oligomers, which consisted of dimers, tetramers, and very few high molecular weight (HMW) species (Fig. 3c). Similarly, we found that both mutant proteins MID51 p.Y240N (Fig. 3d-f) and MID51 p.R146W (Fig. 3g-i) also showed similar oligomerization patterns to wild-type MID51 (Fig. 3c), with similar ratios of each oligomeric species to monomer levels. Thus, these MIEF1 missense variants linked to optic neuropathy do not significantly disrupt the oligomerization of MID51.
Finally, as MID51 is a known regulator of mitochondrial fission dynamics [18,19], we studied whether mitochondrial fission/fusion dynamics were perturbed by the MIEF1 variants. Expression of wild-type MID51 has previously been shown to inhibit mitochondrial fission, resulting in increased mitochondrial fusion events [19]. We analyzed the rate of mitochondrial fusion events in live cells, using a photo-activatable mitochondrial matrix probe (mito-PAGFP) in high spatial and temporal resolution confocal microscopy live imaging studies. A subpopulation of mitochondria was selectively photoactivated using a 405 nm laser, and the rate of fusion events with the surrounding mitochondria was analyzed by measuring the increase in fluorescence intensity over time at a distal region of 10 μm from the site of photoactivation (Fig. 3j). Consistent with previous reports, we found that wild-type MID51 expression increased the rate of fusion events (Fig. 3j-m) compared to control cells. In contrast, both MID51 p.Y240N and p.R146W mutants resulted in significantly decreased mitochondrial fusion events and disrupted mitochondrial network dynamics, as compared to wild-type MID51 (Fig. 3j-m). Together, these results show that MIEF1 mutations linked to optic neuropathy preferentially disrupt the ability of MID51 to regulate mitochondrial fission/fusion dynamics.
In summary, combining our clinical, genetic and pathophysiological data, we identify the first known dominant MIEF1 mutations linked to a human disease, resulting in a specific ophthalmological neurodegenerative disease. MID51 is a key regulator of mitochondrial dynamics [18,19,[23][24][25][26][27]. Our findings that Mid51 optic neuropathy-linked variants disrupt mitochondrial fission/fusion dynamics but not its localization or oligomerization are consistent with the fact that p.Y240 is a residue located in the loop region (residues 238-242) critical for DRP1 binding [28,29] which regulates mitochondrial fission [30]. In contrast, neither diseaselinked mutations (p.Y240N or p.R146W) are located in MID51's transmembrane domain which mediates its outer mitochondrial membrane localization (residues 1 to 48) [19], or in previously characterized residues mediating Mid51's oligomerization [29]. Together, our work should prompt the molecular screening of MIEF1 in ION individuals with severe alterations of the peripheral visual field, and further emphasizes the crucial role for properly regulated mitochondrial dynamics in neurodegenerative diseases.

Population screening
Two hundred individuals in France (112 males and 88 females) with an inherited optic neuropathy were included in this study. Their DNA samples were collected at the Department of Biochemistry and Genetics from the Angers University Hospital (France), which is a national centre for the molecular diagnosis of ION. All patients presented with an isolated dominant or recessive optic atrophy, and were diagnosed by an ophthalmologist from the French National Reference Centres for Rare Blinding Diseases.

Subjects
The first affected individual aged 55 is from Maghreb origins (Patient 1). In 2002, she noticed a faint visual problem on the right eye, with a reported best visual acuity (bva) of 10/20 at first examination. In 2015, she complained of an acute painless visual loss, affecting mainly the peripheral visual field of both eyes, and also the central visual field of the right eye (bva:1/20), but not that of the left eye (bva: 18/20) (Fig. 1b, left). Visual evoked potentials were strongly affected for both eyes, while scotopic and photopic electroretinograms (ERGs) were normal. Eye fundus examination revealed a normal retina, while the optic disks appeared pale and moderately excavated (Fig. 1a, left). Blood analyses, and in particular vitamin B9 and B12 concentrations and Lyme and Treponema serology, were normal, but she had higher sedimentation rate (44 mm, with normal Protein C Reactive concentration). She also suffered from seronegative rheumatoid polyarthritis treated by ARAVA (20 mg/day), while no peripheral neuropathy was noticed.
The second individual aged 47 was born in Egypt (Patient 2). She did not complain of any visual defect during her youth. In 2002, 1 month after an uneventful cesarean delivery, she complained of a painless visual loss first on the right eye (bva: 3/20), then on the left eye (bva: 10/ 20), while the visual field examination disclosed severely narrowed isopters, in particular on the superior quadrant, reflecting a bilateral superior altitudinal defect (Fig.  1b, right). She presented with pale and excavated optic disks (Fig. 1a, right) with normal intra-ocular pressure (IOP: RE: 9 and LE: 10 mmHg). Her bilateral optic neuropathy progressed to 2/20 and 8/20 in 2012, to moving hand and 4/20 in 2015, respectively for the right and left eye. VEP (visual evoked potential) were strongly affected, while the retina was normal, except for the presence of macular microcysts. OCT (optical coherence tomography) examination revealed collapsed RNFL (retinal nerve fiber layer) in all quadrants in both eyes (Fig. 1c). Anti-NMO and anti-MOG serology were negative. Brain MRI and ENT examinations were normal. She suffers from hypothyroidism, hypercholesterolemia and iron deficiency, but has no diabetes, epilepsy or neuro-muscular disorder. Biotinidase activity was normal.
In both individuals, Sanger sequencing did not reveal mutations in OPA1, OPA3, WFS1, or mtDNA LHON mutations. In addition, SNP array analyses revealed no chromosomal abnormality, no copy number variation and no large region of homozygosity that might have indicated a consanguinity.

Standard protocol approvals, registrations, and patient consents
Written informed consent to perform genetic analyses was obtained from each subject involved in this study, according to protocols approved by the Ethical Committees of the different Institutes involved in this study, and in agreement with the Declaration of Helsinki (Institutional Review Board Committee of the University Hospital of Angers, Authorization number: AC-2012-1507).

Genetic analysis
Genomic DNA was extracted from peripheral blood cells from a cohort of individuals in France with ION and screened for pathogenic variants in OPA1, OPA3 and WFS1 exonic sequences and the three primary LHON mtDNA mutations. Negative cases were analyzed by targeted resequencing panel of 22 genes (Supplementary  Table 1) including published ION genes and candidate genes involved in mitochondrial dynamics (FIS1, MFF, MFN1, MIEF1, MIEF2, OMA1 and YME1L1), but not known to be involved in ION.
The amplicon library targeting the exons from the 22 genes (Supplementary Table 1) was designed with the Ion AmpliSeq Designer (http://ampliseq.com). The Qubit dsDNA High Sensitivity Assay Kit (Thermo Fisher Scientific) was used to quantify DNA for NGS library construction. Library preparation for each sample was performed using Ion AmpliSeq technologies, sequencing was undertaken using 540 ChIPs on Ion S5™ Sequencer using barcoded samples, and sequencing data was processed using our own dedicated bioinformatics pipeline as described elsewhere [31,32]. Within the design regions, 95% of bases had coverage over 100X, indicating that sufficient coverage was achieved to enable high variant detection sensitivity. Sanger sequencing was performed in parallel for regions that were not or poorly covered by ChIP sequencing.
For each patient, we screened for causative variants using the following prioritization strategy. Firstly, we checked reported pathogenic variants in ION published genes, then we looked for novel loss-of-function (LOF) variants (stop-gain, frameshift and splicing) or novel missense variants and finally, we selected pathogenic variants in candidate unpublished genes. All dominant variants were either absent or had a minor allele frequency (MAF) threshold of < 0.0001 in public databases (GnomAD), while recessive variants were considered with a MAF threshold < 0.005. For familial cases, we specifically selected the variants that matched the inheritance pattern predicted from the pedigrees [32]. The two candidate pathogenic variants in MIEF1 were validated by Sanger sequencing.

Confocal microscopy
Confocal images were acquired on a Nikon A1R confocal microscope with GaAsp detectors using a Plan Apo λ 100 × 1.45 NA oil immersion objective (Nikon) using NIS-Elements (Nikon). Live cells were imaged in a temperature-controlled chamber (37°C) at 5% CO 2 at 1 frame every 2-3 s. For mito-PAGFP experiments, the matrix of a subpopulation of mitochondria were selectively labeled by localized photoactivation, using a 405 nm laser (100% for 4 s) in cells transfected with photoactivatable mitochondrial matrix marker mito-PAGFP and either control (mApple-Tom20) or mCherry MID51 (WT, Y240N or R146W). Fluorescence intensity at a distal region of 10 μm from the site of photoactivation was analyzed.

Immunoprecipitation
To examine MID51 oligomerization, HeLa cells were transfected for 24 h with myc-tagged MID51 (WT, Y240N or R146W) or not transfected (control). Cells were lysed in EBC buffer (Boston Bioproducts; C14-10) with cOmplete™ Protease Inhibitor Cocktail (Roche; 11, 873,580,001) and sonicated. Lysates were immunoprecipitated using Protein G-coupled Dynabeads (Invitrogen; 100003D) incubated in anti-myc antibody (ms), and washed in EBC buffer. Equal total protein levels of immunoprecipitate were analyzed by SDS-PAGE and Western blot according to standard protocols using antimyc antibody (Rb). Measured band intensities of immunoprecipitated MID51 for oligomeric species (dimer, tetramer or high molecular weight (HMW)) were normalized to monomeric MID51, and further normalized to MID51 (WT), and expressed as the ratio of MID51 oligomer/monomer for each condition.

Image analysis
MID51 localization to mitochondria was quantified as the percentage of cells that had MID51 (mCherrytagged) colocalized with mitochondria (mEmerald-mito).
Mitochondrial fusion dynamics were analyzed as the mito-PAGFP fluorescence intensity at a distal region (10 μm from the site of photoactivation) at 1 min and 5 min post-photoactivation, and subsequently normalized to the intensity at t = 0 min. Example traces of mito-PAGFP max intensity (Fig. 3k) are the average of 3 independent examples (n = 3 cells from 3 experiments/ per condition). The mito-PAGFP fluorescence intensity used was the maximum intensity in a 4 μm × 4 μm area in the distal region. Immunoblots were quantified using ImageJ (NIH).

Statistical analysis, graphing and figure assembly
Data were analyzed using unpaired two-tailed t test (for two datasets) and one-way ANOVA with Tukey's post hoc test (for multiple datasets), and statistics are shown comparing MID51 (WT) to MID51 mutants. Data presented are means ± s.e.m. with n ≥ 3 independent experiments (biological replicates) per condition. Statistics and graphing were performed using Prism 7 (GraphPad) software. All videos and images were assembled using ImageJ 1.51j8 (NIH) and Illustrator CC (Adobe).