Characterization of a novel variant in the HR1 domain of MFN2 in a patient with ataxia, optic atrophy and sensorineural hearing loss

Abbreviations

BDLP: bacterial dynamin like protein

CADD: combined annotation dependent depletion

CAPOS: cerebellar ataxia, areflexia, pes cavus, optic atrophy, and sensorineural hearing loss

CMT2A: Charcot-Marie-Tooth type 2A

CO₂: carbon dioxide

DAPI: 4′,6-diamidino-2-phenylindole

DSHB: Developmental Studies Hybridoma Bank

EDTA: ethylenediaminetetraacetic acid

ER: endoplasmic reticulum

FBS: fetal bovine serum

GAD: glutamic acid decarboxylase

HR1: heptad repeat 1

HR2: heptad repeat 2

IMS: inter membrane space

MEM: minimum essential media

MERCs: mitochondria–endoplasmic reticulum contact sites

MFN: mitofusin

MRI: magnetic resonance imaging

mtDNA: mitochondrial deoxyribonucleic acid

NGS: next generation sequencing

OCR: oxygen consumption rate

OMIM: Online Mendelian Inheritance in Man

OPA1: optic atrophy 1

PBS: phosphate buffered saline

PFA: paraformaldehyde

PLA: proximity ligation assay

qPCR: quantitative polymerase chain reaction

SD: standard deviation

VCFs: Velocardiofacial syndrome

VEP: Variant Effect Predictor

Introduction

Mitochondria are highly dynamic double membrane-bound organelles that undergo continuous remodelling via fusion and fission events. These dynamic processes determine mitochondrial structure and regulate mitochondrial function.¹ Mitochondrial fusion is a multistep process mediated by several essential proteins and regulators.² Tethering of adjacent mitochondria and fusion of outer mitochondrial membrane (OMM) is performed by mitofusin1 and mitofusin2 (MFN1/2), two homologous proteins with partially redundant functions that are integral to the OMM. Meanwhile, fusion of the inner mitochondrial membrane (IMM) is carried out by optic atrophy 1 (OPA1).^3,4 Highlighting the importance of mitochondrial fusion is the fact that knockout of MFN1, MFN2 or OPA1 genes is embryonic lethal.^5,6 In addition, pathogenic variants in these genes cause human disease, with OPA1 linked to optic atrophy, and MFN2 linked to the peripheral neuropathy Charcot Marie Tooth type 2A (CMT2A).^7,8

MFNs have an N-terminal GTPase domain that is exposed to the cytosol, and two heptad repeat domains (HR1 and HR2, also referred to as coiled-coil domains), separated by a transmembrane domain. While several structural models have been generated for MFNs,^9–12 these are from artificial constructs that are based on the notion that N-terminal and C-terminal domains of the protein both face the cytosol and can interact. However, a recent reappraisal of the topology of MFNs showed that the C-terminus of the protein, including the HR2 domain, is exposed to the inner membrane space (IMS), not the cytosol.¹³ This revision raises questions about the validity of the structural models. Thus, there remain many questions as to the exact structure of MFNs and how they mediate mitochondrial fusion.

Mitochondrial fusion is also important for the maintenance of the mitochondrial genome (mtDNA), which is present in hundreds to thousands of copies per cell. The mtDNA is packaged into nucleoid structures, each of which is approximately 100 nm in size and contains a single copy of the mitochondrial genome.^14–16 Impairments to mitochondrial fusion can lead to reduced copy number and increased nucleoid size.^17,18 While the reduced copy number is thought to be a result of decreased distribution of the replication machinery,¹⁹ it is unclear exactly why nucleoid size might change. Meanwhile, mitochondrial respiration can also be compromised upon loss of fusion.^20,21

Notably, MFN2 has several functions in addition to its role in mitochondrial fusion. For example, MFN2 mediates mitochondrial autophagy (mitophagy)^22,23 and transport of mitochondria.^24,25 Meanwhile, MFN2 also localizes to the endoplasmic reticulum, where it mediates interactions between the ER and mitochondria.^26–28 These mitochondria-ER contact sites (MERCs) are specialized sites for both lipid biogenesis and exchange, are important for regulating calcium signalling, and can also mark sites of mitochondrial fission,²⁹ as well as mtDNA replication.³⁰ Notably, MFN2 has been implicated in directly binding and transferring phosphatidylserine from ER to mitochondria.³¹ While there is some debate whether MFN2 promotes or inhibits MERCs,^27,28,32,33 it is clear that it plays an important role in these organelle contacts. Finally, in addition to MERCs, MFN2 also mediates interactions between mitochondria and lipid droplets.³⁴

Pathogenic variants in genes regulating the opposing forces of mitochondrial fusion and fission can be associated with peripheral neuropathy,³⁵ suggesting that impairment in the balance of these processes contributes to the disease phenotype. To date, more than 100 pathogenic variants in MFN2 have been associated with peripheral neuropathy.³⁶ However, despite the general assumption that impaired mitochondrial fusion causes the peripheral neuropathy phenotype, only a few pathogenic MFN2 variants have been investigated for their effects on MFN2 functions. While some pathogenic MFN2 variants do impair fusion,^37,38 unexpectedly, other pathogenic variants seem to increase fusion,^39,40 while several pathogenic variants do not appear to affect fusion at all.^38,41,42 These findings raise the possibility that impaired fusion does not lead to peripheral neuropathy per se.

In this context, it is notable that other MFN2-mediated mitochondrial functions can also be impacted by pathogenic MFN2 variants. For example, disruptions to MERCs,^26,38 lipid metabolism,⁴³ mitochondrial respiration,⁴⁴ mtDNA copy number,^45–47 and mitochondrial transport,²⁵ have all been observed in association with various pathogenic MFN2 variants. However, it should be noted these phenotypes have not all been widely investigated across a variety of MFN2 variants.

Further complicating our mechanistic understanding of how MFN2 dysfunction causes disease is the fact that additional pathogenic phenotypes can also be linked to MFN2 variants. Although not common, some MFN2 variants are linked to other disease phenotypes such as optic atrophy.^48,49 Central nervous system involvement is also rarely described,⁵⁰ with periventricular and subcortical white matter lesions in 8 of 21 patients from one series. A small minority of cases had transient neurological deficits such as dysarthria or paraesthesiae. Other complex phenotypes have been observed in the presence of homozygous or compound heterozygous mutations. For example, the R707W MFN2 variant, which causes CMT2A when heterozygous, is also associated with lipodystrophic syndromes when homozygous. Specifically, R707W causes proliferation of adipocytes leading to adipose hyperplasia and lipomatosis.^43,51,52 Additional atypical features including severe neuropathy with hearing loss has been described with biallelic mutations.⁴⁸ We are also aware of a report of a large pedigree including optic atrophy and sensory ataxia associated with a heterozygous D210V substitution in MFN2.⁴⁹

In summary, MFN2 performs a number of functions, many of which can be impaired by pathogenic variants in MFN2. Meanwhile, MFN2 variants can lead to a number of different patient phenotypes. However, there is no clear understanding of the molecular mechanisms causing disease and whether impairment of specific MFN2 functions leads to specific phenotypes. Here, in a patient presenting with ataxia, sensorineural hearing loss, and optic atrophy leading to vision loss, we report the presence of a homozygous novel candidate pathogenic variant in MFN2, p.(D414V), located in the middle of the HR1 domain. While ataxia has not been associated previously with MFN2 specifically, it is common in mitochondrial disease. Our characterization of mitochondrial functions in patient fibroblasts is consistent with impairment of MFN2 functions, expanding the clinical spectrum of phenotypes associated with pathogenic variants in MFN2.

Methods

Results

Case description and clinical diagnosis

The patient initially came to medical attention at eight years of age, having previously had a normal birth history and development. He began to experience visual blurring affecting both eyes, and bilateral optic atrophy was identified on ophthalmologic evaluation. His vision continued to slowly deteriorate over time. In his 20s he developed gait disturbance which was diagnosed as ataxia (though in retrospect teachers had informed him of changes to his gait as early as 12 years of age). The ataxia progressed to affect limb coordination functions in his 30s, in association with dysarthria. Sensorineural hearing loss was diagnosed in his 20s and progressed gradually to deafness. In his 40s he developed hypertension as well as type 2 diabetes mellitus. He was coincidentally diagnosed with a painful small-fibre sensory neuropathy, which was thought to be attributable to his diabetes.

His family history did not include any other individuals with neurologic disease. His parents originated from India and no consanguinity was reported. His two siblings are healthy and do not have a similar condition. The patient does not have any children.

By the time of his assessment by one of the authors (GP) at age 54 years, his family described severe visual and hearing loss. He required ongoing use of a wheelchair and required assistance for feeding and all daily living tasks due to severe balance and coordination deficits. On neurologic examination, he was alert and attentive. His cognition was difficult to assess due to the above-mentioned visual/hearing loss and communication difficulties, but his overall cognition seemed grossly appropriate. Cranial nerve examination demonstrated bilateral optic atrophy, light perception only for visual acuity, full extraocular movements, no ptosis, and symmetric facial movements. His hearing was poor, but he was able to comprehend speech when spoken in a loud voice directly into his ear. He also had cerebellar dysarthria, but speech was still intelligible. He otherwise had normal lower cranial nerve functions. He had diffuse and symmetric loss of muscle bulk which was attributed to deconditioning, with normal motor tone. Deep tendon reflexes were unobtainable in all extremities. Plantar responses were flexor. Power examination revealed that muscle strength was well preserved, with some minor hip girdle weakness which was again attributed to disuse. Additionally, he exhibited a postural tremor in his upper limbs and bilateral dysmetria and intention tremor on finger-to-nose testing. He required two-person assistance to stand. Sensory examination was fairly unremarkable, with vibration sensory loss below the ankles bilaterally, but was otherwise normal.

The differential diagnosis at this time was considered to most likely include genetic or metabolic disorders, such as mitochondrial disorders, CAPOS syndrome (cerebellar ataxia, areflexia, pes cavus, optic atrophy, and sensorineural hearing loss, due to mutations in ATP1A3), or other multisystem genetic disorders such as Wolfram syndrome (due to WFS1 mutations). Investigations included vitamin B12, folate, and vitamin E levels, which were normal. Very long chain fatty acids and hexosaminidase A and B were also normal. Testing for anti-GAD antibodies was negative. A muscle biopsy demonstrated some denervation atrophy but did not reveal any changes suggesting a mitochondrial cytopathy. MRI of the brain revealed diffuse cerebral volume loss affecting the cerebrum, cerebellum and brainstem structures, and volume loss was also observable in the optic nerves bilaterally and optic chiasm (Figure 1A). Clinical genetic testing included a first-line genetic screening for spinocerebellar ataxia types 1, 2, 3, 6, 7, 8, and Friedreich ataxia, which was negative. He also had an NGS-based sequencing panel for ataxic syndromes including 277 genes. This found a homozygous variant of unknown significance in MFN2, c.1241A>T p.(D414V).

Figure 1. Pathogenic variants in MFN2.

Representative images of MRI head scans of the patient at 55 years of age. Top-left: T1-weighted sagittal image through the midline demonstrates volume loss in the frontal lobe, brainstem, and midline cerebellar structures. Top-right: T1-weighted transverse axial image at the level of the lateral ventricles and basal ganglia demonstrate volume loss predominantly affecting the bilateral frontal lobes. Bottom panels: T2-weighted transverse axial images at the levels of the medulla and pons again demonstrate brainstem atrophy as well as accentuated cerebellar foliae indicative of diffuse volume loss in the posterior fossa. (B) Sequencing chromatograms confirms the 1241A>T variant, resulting in a missense mutation of Aspartic acid (D) to Valine (V) at position 414 in the MFN2 protein. Sequencing data are shown from patient derived (upper panel) or normal control (lower panel) derived fibroblasts. (C) Alignment of the region of the HR1 domain of MFN2, showing D414V is conserved throughout vertebrate species and with MFN1. Residues highlighted in yellow are conserved residues determined by Clustal Omega analysis. (D) Diagram showing the topology and domains of the 757 amino acid MFN2 protein, which contains a GTPase domain, two HR domains and a transmembrane (TM) domain. The number of reported pathogenic missense mutations in the indicated regions of the protein are indicated on the right. (E) Western blot analysis of control and MFN2-D414V patient fibroblasts for expression of MFN2, MFN1, OPA1, β-Actin, VDAC and HSP60 proteins. The graphs indicate ratio of MFN2, MFN1, or OPA1 normalized to cellular protein loading control β-Actin as ratio of percentage of control fibroblasts. n = 3 biological replicates. Error bars indicate mean ± SD, unpaired t-test.

Exome sequencing was performed and variants were filtered as follows: maximum allele frequency in population databases of <0.0001, predicted to cause protein-coding changes (simple substitutions, frameshifts, splicing alterations, or early termination), and present in genes associated with neurological phenotypes. We considered variants to be reasonable candidates if they met these criteria and if they fit the known mode of inheritance for these conditions (e.g.: recessive disease genes would require two heterozygous variants or a homozygous variant), and were not classified as “benign” or “likely benign” in ClinVar. Using these criteria, we again identified the above-mentioned homozygous variant in MFN2 (c.1241A>T, p.(D414V)). Using these criteria, we did not identify any other monogenic disease candidates. Coverage of the mitochondrial genome was highly limited (6% coverage at ≥20X, which is considered the minimum read depth to accurately call mtDNA variants from off-target reads).⁵⁹ Within these limitations we did not identify any pathogenic mtDNA variants.

Impact of the MFN2-D414V variant on mitochondrial morphology

We compared mitochondrial morphology in patient and control fibroblasts to examine the well-defined role for MFN2 in mediating mitochondrial fusion. While the mitochondria were long and often reticular in the control fibroblasts, those in the patient fibroblasts were noticeably shorter (Figure 2A). We quantified this observation by measuring the mitochondrial length and found shorter mitochondria in patient fibroblasts, indicative of more fragmented mitochondrial networks compared to those of control fibroblasts (Figure 2B). These results are consistent with a reduction in mitochondrial fusion.

Figure 2. Mitochondrial dysfunction in MFN2-D414V patient-derived fibroblasts.

(A) Representative confocal images of healthy control (left panel) and patient derived (right panel) fibroblasts labelled with TOMM20 immunostaining (purple) to visualize mitochondrial networks. Scale bar, 20 μm. (B) Quantification of average mitochondrial length in fibroblasts from control (n = 24 cells) or MFN2-D414V patient (n = 27 cells). Error bars indicate mean ± SD. p < 0.0001, unpaired t-test (C) Oxygen consumption rate (OCR) traces in fibroblasts from control or MFN2-D414V patient measured using the Seahorse XF24 extracellular flux analyzer (n = 12 replicates). (D) Basal (upper) and maximum (lower) OCR in control and MFN2-D414V fibroblasts calculated from C. Error bars indicate mean ± SD. p < 0.0001, unpaired t-test.

Mitochondrial nucleoids

In order to understand how the D414V variant might be impacting mitochondrial respiration, we employed several approaches to examine the mtDNA genome, which is essential for respiration, and which can be impacted by impairments to mitochondrial fusion.¹⁹ First, using confocal microscopy, we observed a significant decrease in the average number of nucleoids in patient fibroblasts compared to control (Figure 3A, B). Given previous findings that impaired mitochondrial fusion can lead to larger mitochondrial nucleoids, we also imaged and quantified the size of mtDNA nucleoids. Unexpectedly, we observed a reduction in the average size of nucleoids in patient fibroblasts when compared to those from control (Figure 3C). Consistent with the reduced number of nucleoids evident by confocal microscopy, quantitative PCR analysis also showed a reduction in mtDNA copy number in patient fibroblasts (Figure 3D). Meanwhile, D414V patient fibroblasts did not exhibit any evidence for mtDNA deletions (Figure 3E), as has been reported previously for some pathogenic MFN2 variants.⁴⁹

Figure 3. Altered mtDNA in MFN2-D414V fibroblasts.

(A) Representative confocal images of mitochondria and mitochondrial nucleoids labelled by immunostaining with anti-TOMM20 (purple) and anti-dsDNA (cyan) antibodies in control and MFN2-D414V fibroblasts. The area indicated by white box in the upper panel is shown in lower panel at higher magnification. Scale bar, 20 μm. (B) Quantification of the number of mtDNA nucleoids. Each data point indicates the total number of nucleoids in a control or MFN2-D414V fibroblast. Error bars indicate mean ± SD. p = 0.0012, unpaired t-test. (C) Quantification of the average mtDNA nucleoid size in fibroblast cells from control (n = 26) or MFN2-D414V patient (n = 40). Error bars indicate mean ± SD. p < 0.0001, unpaired t-test. (D) Relative mtDNA copy number analyzed using quantitative PCR with 18S DNA or Beta-2-Microglobulin as reference genes. Different colours indicate biological replicates, and the same colour indicates technical replicates. Error bars indicate mean ± SD. p < 0.0001, unpaired t-test. (E) Integrity of the mtDNA genome was assessed on 0.6% agarose gel following long-range PCR of DNA isolated from control or MFN2-D414V fibroblasts. No mtDNA deletions were detected.

Mitochondria-ER contacts (MERCs)

Next, we tested whether the D414V variant might impair MFN2 protein function in mediating MERCs. To estimate the size and number of MERCs in patient and control fibroblasts, we employed a proximity ligation assay (PLA), which indicates when two target proteins (mitochondrial-TOMM20 and ER-calnexin) are within ~40 nm.^60,67 We observed a significant reduction in both the number and size of MERCs in the patient cells (Figure 4A-D), consistent with the notion that the D414V variant might be causing reduced interactions between mitochondria and ER.

Figure 4. Mitochondria-ER contacts (MERCs) are reduced in MFN2-D414V fibroblasts.

(A and B) Representative images showing MERCs in the control (A) or MFN2-D414V (B) fibroblasts, visualized using a proximity-ligation assay (cyan). Mitochondrial networks and ER are visualized via immunostaining with TOMM20 (pink) and calnexin (purple), respectively. The area indicated by white box in the upper panel is shown in lower panel zoomed in. Scale bar, 20 μm. (C) Quantification of the number of MERCs. Each data point indicates the total number of MERCs in fibroblast cells from control (n = 35) or MFN2-D414V patient (n = 49 cells). Error bars indicate SD. p = 0.0002, unpaired t-test. (D) Quantification of the average size of MERC’s in control or MFN2-D414V fibroblasts. Each data point indicates average size of MERCs from an individual cell. Error bars indicate SD. p < 0.0001, unpaired t-test.

Lipid droplet regulation

Our observations of reduced MERCs, and the probable compromise in tethering function of MFN2 in the patient fibroblasts, led us to hypothesize that lipid droplet metabolism could also be affected by the MFN2-D414V variant. To this end, lipid droplets were imaged by staining with a neutral lipid dye and imaged by confocal microscopy. Notably, the number of lipid droplets was reduced in the patient fibroblasts (Figure 5A, B and C). In addition, the total lipid content, as measured by the total intensity of the lipid dye, was also reduced (Figure 5D). Finally, we also noted an unusual perinuclear arrangement of lipid droplets in patient fibroblasts compared to control fibroblasts, in which lipid droplets were spread throughout the cell (Figure 5A, B). To quantify the differences in lipid droplet distribution, we measured the distance of individual lipid droplets from the center of the nucleus. We found that the average distances of lipid droplets from the nucleus was significantly lower in the patient fibroblasts (Figure 5E, F). Altogether, these observations are consistent with the notion that the D414V variant in MFN2 affects lipid droplets.

Figure 5. The intracellular distribution and number of lipid droplets (LD) are reduced in MFN2-D414V fibroblasts.

(A and B) Representative confocal images of control (A) and MFN2-D414V (B) fibroblasts, with lipid droplets stained with LipidTox Green (green) and mitochondria visualized following immunofluorescence with TOMM20 (purple). The area indicated by white box in the upper panel is shown in lower panel zoomed in. Scale bar, 20 μm. (C) Quantification of the number of LDs. Each data point indicates the total number of LDs in fibroblast cells from control (n = 35) or MFN2-D414V patient (n = 49). Error bars indicate SD. p = 0.0002, unpaired t-test. (D) Quantification of total lipid content. Each data point indicates the total intensity given by the LipidTox Green dye for fibroblast cells from control (n = 35) or MFN2-D414V patient (n = 45). Error bar indicates SD. p = 0.0073, unpaired t-test. (E) Quantification of the intracellular distribution of LDs. Distance of each LD from centre of the nucleus of a respective cell was calculated for fibroblasts from control (n = 8303 LD from 24 cells) and MFN2-D414V (n = 5259 LD from 25 cells). The median and inter-quartile ranges are indicated in the violin plot. p<0.0001, unpaired t-test. (F) The average distance of LDs from nucleus. Each data point indicates the average distance of all the LDs from the center of respective cell’s nucleus for fibroblast cells from control (n = 24) or MFN2-D414V patient (n = 25). Error bar indicates SD. p = 0.0006, unpaired t-test.

Discussion

We identified a novel homozygous MFN2 variant, D414V, in a patient with ataxia, deafness and blindness. While hearing and vision loss have been observed in a few cases linked to MFN2 variants,^48,49 cerebellar ataxia has not. There is one described report of an MFN2 patient having neuropathy with sensory ataxia and optic atrophy.⁴⁷ The index case was actually described as having “sensory and cerebellar ataxia”, however the presence of cerebellar dysfunction in this case is uncertain because physical findings were not described, the MRI did not have cerebellar abnormalities, and none of the other 11 cases in the pedigree were considered to have cerebellar findings. In contrast, our case is the first to demonstrate clear-cut cerebellar ataxia, with cerebellar examination findings, absence of major sensory findings, and presence of cerebellar atrophy on neuroimaging. As pathogenic variants in MFN2 are typically associated with the peripheral neuropathy CMT2A, it was not clear whether the D414V variant is responsible for these patient phenotypes. Notably, though the patient did have a small-fibre sensory neuropathy, this was attributed to diabetes given the relatively late onset compared to the other phenotypes. A second question arising from the D414V variant is whether different mechanisms underlying MFN2 dysfunction might explain the distinct patient phenotypes. Unfortunately, the D414V variant has not been described to date in any other patients. As such, we only have a single patient cell line in which to study this variant. Here, we discuss our characterization of the D414V variant, and the novel insight provided into the function of MFN2 and molecular mechanisms underlying pathology associated with dysfunction of this protein.

Despite the limitations of existing structural models, homology modelling of MFN2 with the bacterial homolog BDLP may still be informative.^68,69 These reports suggest that the HR1 domain, which is predicted to form part of a larger helical bundle, comprises two alpha helical subdomains (~350-383, and ~390-419) that are separated by a flexible hinge domain. Notably, of the 167 reported MFN2 missense/nonsense variants in the Human Gene Mutation Database,⁷⁰ only a handful are found in three positions in the second helical domain of HR1 (aa390-419), with two of the affected residues at the extremities of this subdomain (i.e. aa390 and aa 418). These variants have been published in the context of classic CMT2A phenotypes, including: C390R,⁴⁸ C390F,⁸ R400X,⁷¹ R400P,⁷² R418Q⁷³ and R418X.⁷⁴ However, none of these variants have been characterized functionally in human cells. Moreover, in contrast to the homozygous D414V reported here, the other published variants in the second helical domain of HR1 are all heterozygous. Notably, two variants were compound heterozygous, with the R400X variant found in combination with another pathogenic variant (R250W, also linked to CMT2A), and the C390R variant in combination with N214D. As such, some functional rescue by the other allele is possible for these other variants. These observations suggest that the region comprising the second helical domain within HR1 (aa390-419), where the D414V variant is present, is of critical importance for proper MFN2 function.

It is also worth briefly discussing a study in which two rare human MFN2 variants were modelled in Drosophila.⁶⁵ Tissue-specific overexpression of these variants (M393I and R400Q) led to cardiac and eye phenotypes. Notably, these variants are also present in the second helical region of the HR1 domain, suggesting a distinct role for this region of the HR1 domain in mediating MFN2 function, and hint at additional phenotypes that could be associated with human disease due to pathogenic variants in MFN2.

An initial investigation in the expression of MFN2 and other mitochondrial fusion proteins MFN1 and OPA1 provided interesting findings. The observation of reduced MFN2 protein expression in our patient fibroblasts likely reflects a total loss of mitochondrial mass as the expression of several other mitochondrial proteins is also reduced. While this reduction could be due to reduced biogenesis or increased turnover, we favour the later explanation as MFN2 has a known role in mediating mitochondrial mitophagy.⁷⁵ Notably, reduced MFN2 protein expression has also been reported for MFN2 variants linked to mtDNA depletion,⁷⁶ though total mitochondrial abundance was not investigated in that study.

The observation of altered abundance of OPA1 isoforms in the D414V fibroblasts is also intriguing in the context of the functional link between MFN2 and OPA1. While some MFN2 variants linked to mtDNA instability have reduced total OPA1 expression,⁴⁷ to our knowledge, a link between pathogenic MFN2 variants and alterations in OPA1 isoform abundance has not been reported previously. Elucidating the exact cause of the changes in the relative abundance of OPA1 protein bands in our patient fibroblasts is beyond the scope of the current work. However, based on previous work examining OPA1 isoforms⁷⁷ we hypothesize the following explanations for the observed pattern of OPA1 bands in the patient fibroblasts: 1) the loss of OPA1 bands a and c is likely due to reduced expression of OPA1 isoform 7; 2) band d may be the d’ band described recently for OPA1 isoforms containing exon 4b (i.e. isoforms 3,5,6,8), which are cleaved by YME1L; 3) the increased abundance of band b and reduced abundance of band e is likely due to reduced OMA1 processing of isoform 1.

While it is uncertain how changes in MFN2 function might impact OPA1 expression and processing, it is worth noting that MFN2 interacts SLP2 (STOML2),⁷⁸ which is an IMM protein that regulates OMA1 cleavage of OPA1,⁷⁹ and mediates stress-induced mitochondrial hyperfusion.⁸⁰ Meanwhile, MFN2 has also been implicated in additional processing of OPA1 linked to starvation.⁸¹ Additionally, while OMM and IMM fusion events are sequential and can be uncoupled,⁸² direct interaction between MFNs and OPA1 have been observed in yeast,^83,84 fly,⁸⁵ and humans.⁸⁶ Regardless of the mechanism by which MFN2 impacts OPA1, it is possible that alterations in the OPA1 isoform expression also contribute to the mitochondrial dysfunction observed in these cells.

To investigate whether the D414V variant is indeed pathogenic, we performed a comprehensive analysis of mitochondrial functions linked to MFN2, beginning with the well-recognized role of MFN2 in mediating mitochondrial fusion. Patient fibroblasts harboring the D414V variant displayed fragmented mitochondrial networks compared to control, consistent with impaired fusion in these cells, which could be due to impaired MFN2 functionality and/or altered expression of OPA1 isoforms. This finding is notable in comparison to the few pathogenic MFN2 variants causing CMT2A that have been investigated for their role in fusion. While a few of the CMT2A MFN2 variants studied appear to be fusion incompetent,^5,65,87 there are other CMT2A MFN2 variants that do not seem to affect mitochondrial morphology.^41,88 Conversely, some CMT2A MFN2 variants actually enhance fusion.^39,40 Thus, impaired mitochondrial fusion does not appear to be necessary to cause CMT2A.

As mitochondrial fusion is important for maintenance of the mitochondrial genome, we also investigated mtDNA in patient fibroblasts. The reduced mtDNA copy number and nucleoid number observed in patient fibroblasts could be a result of several abnormalities observed in these cells. First, reduced fusion could be involved, as inefficient fusion-dependent distribution of the mtDNA replication machinery is proposed to lead to reduced mtDNA copy number.¹⁹ Second, the decrease in MERCs could contribute, as the ER is thought to licence mtDNA replication.³⁰ Third, alterations to the levels of OPA1 isoforms could also play a role, as OPA1 also mediates mtDNA.⁸⁹ Finally, the reduced abundance of several mitochondrial proteins likely reflects a global loss of mitochondrial mass, which would also be consistent with decreased mtDNA copy number. In this light, it is relevant that mitophagy can be mediated by MFN2,⁹⁰ MERCs,^22,23 and OPA1.⁹¹ To our knowledge, a link between pathogenic MFN2 variants and increased mitophagy has not been described. However, increased mitophagy has been noted in the context loss of FBXL4,⁹² another mitochondrial fusion protein,²¹ mutations in which cause a mtDNA depletion syndrome.

Surprisingly, we also observed slightly smaller nucleoids in patient cells, an unexpected finding given that impaired mitochondrial fusion has previously been linked to enlarged mitochondrial nucleoids.^19,21,93 However, it should be noted that it is not clear why nucleoid size increases in response impaired fusion. One possible explanation for this discrepancy in nucleoid size could be that size does not correlate directly with the degree of fusion impairment. Perhaps, slight impairments to mitochondrial fusion lead to smaller nucleoids, while greater impairments lead to larger nucleoids. The fact that each nucleoid is estimated to contain ~1.4 copies of the mtDNA genome,⁹⁴ is consistent with observations that a significant subset of nucleoids is actively undergoing replication.³⁰ If one logically assumes that replicating nucleoids are larger than non-replicating nucleoids, then the smaller nucleoids we observe could be due to a reduced number of nucleoids being actively replicated.

Meanwhile, the larger nucleoids described in cells completely lacking fusion are due to clustering of multiple individual nucleoids that cannot be resolved by traditional confocal microscopy.¹⁹ Notably, cells where mitochondrial fission is inhibited also exhibit large, clustered nucleoids, demonstrating that mitochondrial dynamics are important for the distribution of nucleoids.^95,96 Perhaps in cells with more severe fusion impairment, individual mtDNA nucleoids also cluster, leading to apparently larger nucleoids. It is also notable that while nucleoid size has not previously been quantified in cells with different pathogenic MFN2 variants, a subset of pathogenic MFN2 variants are linked to mtDNA depletion and deletions.^47,97

Given the links between mitochondrial structure and function, as well as the critical role for mtDNA encoded proteins in oxidative phosphorylation, we also examined mitochondrial function in patient fibroblasts. The reduced OCR we observed in MFN2-D414V patient fibroblast demonstrated clear mitochondrial dysfunction in these cells, possibly as a result of reduced mitochondrial fusion leading to reduced levels of mtDNA, or impaired MERCs affecting Ca⁺⁺ transfer to mitochondria. While MFN2 is important for maintaining mitochondrial bioenergetics,⁹⁸ the consequences of pathogenic CMT2A MFN2 variants on mitochondrial function are conflicting, raising questions about the contribution of impaired mitochondrial bioenergetics to CMT2A. For example, some pathogenic CMT2A MFN2 variants lead to a decrease in the mitochondrial bioenergetic function,^41,98,99 others cause an increase,^100,101 while still more variants do not cause any change at all.^25,38,42

MFN2 also plays a key role in mediating contact sites between mitochondria and ER, though conflicting reports debate the exact role of MFN2 in maintaining MERCs. Thus, we also investigated MERCs in D414V patient fibroblasts, where we observed a decrease in number as well as size. These results suggest that the D414V variant reduces MERCs. While only a limited number of CMT2A MFN2 variants have been investigated for their role in MERCs,^26,38 disruption of contacts between mitochondria and ER seems to be a common theme underlying neuropathies.^87,102,103 Further, a study in Drosophila indicated that MFN2 could regulate ER stress,¹⁰⁴ and peripheral neuropathy in diabetic patients is also thought to be partly due to ER stress.^105,106 These observations suggest that deregulation of MERCs and their many functions could be a significant contributor to the peripheral neuropathy phenotype associated with CMT2A MFN2 variants.^38,87

Although a peripheral neuropathy noted in the D414V patient was initially attributed to a diabetic neuropathy, it is worth considering the contribution of MFN2 dysfunction to this phenotype. Notably, peripheral neuropathy can appear at later stages in life and with varying severity.^36,107 Phenotypic variation may occur between individuals with the same mutation suggesting a potential role for other genetic or environmental factors. In this regard, the alteration to MERCs are consistent with the fact that impaired MERCs are a feature of MFN2 CMT2A variants.^38,108 As such, it may simply be a coincidence that the peripheral neuropathy appeared around the same time as the patient’s diabetes, and that MFN2 dysfunction was responsible. Alternatively, it is possible that the development of diabetes exacerbated an already unstable situation due to the MFN2 variant.

Next, we examined lipid droplets, another organelle whose interactions with mitochondria are mediated by MFN2.³⁴ We found that the abundance of neutral lipid signal and the number of lipid droplets are reduced in patient fibroblasts. In contrast to our finding, a previous report has shown that CMT2A MFN2 variants increase the lipid droplet signal and apparent abundance.³⁸ Interestingly, we also discovered an unexpected perinuclear accumulation of lipid droplets in fibroblasts with the D414V variant when compared to healthy control. This phenotype has not been described previously in the context of MFN2 dysfunction.

It is also worth discussing the role of MFN2 in cellular lipid homeostasis, though there is clearly more to learn, especially in the context of disease. In addition to regulating lipid droplets and MERCs, MFN2 was recently shown to have a direct role in transferring phosphatidylserine from the ER to mitochondria, a function implicated in non-alcoholic steatohepatitis.³¹ Moreover, in adipose tissue the association between mitochondria and lipid droplets is important for both lipid storage and consumption,^109,110 while adipocyte specific knockout of MFN2 leads to obesity in mice.^34,111,112 Finally, the R707W MFN2 variant, which causes CMT2A when present heterozygously, also causes lipomatosis when present homozygously.^51,52 Thus, it is clear that the roles of MFN2 in maintaining lipid homeostasis are important, though the exact molecular mechanisms remain undefined.

Our results indicate that the D414V MFN2 variant behaves differently than the few other CMT2A variants that have been investigated in the context of lipid droplets. Furthermore, in contrast to complete loss of MFN2 function, which seems to increase lipid accumulation, the reduced lipid droplet abundance we found in D414V fibroblasts could be due to reduced lipid storage or increased lipid consumption. We suggest that this reduced lipid storage is most likely due to impaired lipid droplet tethering, given the reduced oxidative phosphorylation we observed is inconsistent with an increase in lipid consumption. Furthermore, the distinct distribution pattern of lipid droplets in D414V fibroblasts is consistent with a role of MFN2 regulating interactions between mitochondria and lipid droplets via perilipin 1,³⁴ an important mediator of lipid droplet distribution.¹¹³ Though it should be noted that perilipin 1 is more abundant in adipose tissue than in fibroblasts.³⁴ In contrast, the other CMT2A MFN2 variants investigated previously are proposed to increase lipid droplet signal via alterations to MERCs.³⁸ Though we also see evidence of MERC disruption in D414V cells, the differences in lipid droplet signal between the D414V and other CMT2A variants could also be due to the severity of MERC disruption, which cannot be compared across studies, as different methods of analysis were used.

In comparison to the characteristics of previous published CMT2A variants, our findings describing the cellular characteristics of the D414V variant, begin to provide insight that may explain the different patient phenotypes and the underlying mechanisms of disease. Though mounting evidence clearly shows that dysregulation of MFN2 causes CMT2A, the exact molecular mechanism underlying this pathology is complicated by the fact that MFN2 is a multifunctional protein. As such, impairment of any or all functions performed by MFN2 could be pathogenic. Here, our description of the D414V variant suggests that impairment of different MFN2 functions may be associated with different pathological phenotypes. Notably, the patient phenotypes described here with the MFN2-D414V variant are reminiscent of variants in OPA1, where patients also have optic atrophy, hearing loss and ataxia. This similarity suggests that impaired mitochondrial fusion may be the underlying mechanism driving these specific phenotypes.

To date, only a few CMT2A MFN2 variants have been investigated for their effects on the various functions of MFN2, making it difficult to generalize which MFN2 function(s) may be causative for which phenotype. Nonetheless, the conflicting findings of pathogenic CMT2A MFN2 variants that have been studied with respect to their consequences on mitochondrial fusion, mtDNA depletion, and bioenergetics suggest that these functions may not be the primary mechanisms underlying the peripheral neuropathy phenotype. However, impairment of these other functions could certainly contribute to disease and may explain some of the additional features sometimes associated with CMT2A (e.g. hearing loss, optic atrophy or lipomatosis). It is also important to note that many of these conflicting reports are from distinct studies by different groups that do not always use the same methods, making direct comparisons difficult. Other limitations of this study include the possibility that a clinically relevant intronic variant was missed due to the use of exome sequencing. We did not perform mitochondrial genome sequencing (and coverage of mtDNA from off-target reads was very low in our exome analysis), so it is also possible a relevant mtDNA variant was missed.

In summary, our study establishes that the MFN2-D414V variant is incompetent in carrying multiple MFN2 functions, arguing for the importance of the HR1 domain. One of the key cellular differences between D414V and CMT2A variants pertains to lipid droplets abundance, as well as a previously unreported perinuclear lipid droplet distribution. Furthermore, the fact that nearby variants also cause distinct phenotypes compared to ‘classic’ CMT2A variants when modelled in flies supports the notion that alterations to the HR1 domain could have different patient phenotypes in humans. However, as only a single D414V patient has been described to date, we cannot infer definitively on genotype/phenotype correlation. At the same time, the cellular phenotypes we describe are all consistent with impaired function of MFN2. Combined with bioinformatic predictions that this variant is likely to be deleterious, we believe the homozygous D414V variant to be the cause of the patient phenotypes, thus expanding the clinical spectrum MFN2-associated mitochondrial diseases.

Author contributions

Conceptualization: G.S., R.S., G.P., T.S.; Methodology: G.S., R.S., K.M.; Formal analysis: K.M. M.J., A. dK., G.P.; Investigation: G.S., R.S. M.J.; Resources: D.M., G.P.; Data Curation: M.J.; Writing – original draft: G.S., G.P., T.S.; Writing – review & editing: G.S., R.S., D.M., G.P., T.S.; Visualization: G.S.; Supervision: A. dK., G.P., T.S.; Project Administration: G.P., T.S.; Funding acquisition: T.S.