Patient-specific mutation of contact site protein Tomm70 causes neurodegeneration Open Access

Translocase of outer mitochondrial membrane 70 (Tomm70) is a 70 kDa mitochondrial import receptor protein composed of 26 α helices, the majority of which contribute to the formation of tetra-trico-peptide repeat (TPR) motifs. These TPR motifs mediate the interaction of Tomm70 with other proteins (Kreimendahl and Rassow, 2020). Although the role of Tomm70 in facilitating the uptake of proteins into the mitochondria is well established, recent studies have shown that Tomm70 is also a key component of the endoplasmic reticulum (ER)-mitochondria contact site (MCS) (Filadi et al., 2018), where it binds with an ER-localised sterol transporter protein Lam6 (also known as Ltc1) (Murley et al., 2015). ER-MCSs are known to regulate a number of key cellular processes, including lipid metabolism, calcium homeostasis, mitochondrial functions, ER stress and autophagy, and the defects in contact site proteins are associated with neurodegenerative disorders (Wilson and Metzakopian, 2021). TOMM70 has been reported to be a potential candidate gene for hereditary spastic paraplegia (HSP) (Novarino et al., 2014). HSPs are a complex group of neurodegenerative disorders, characterised by progressive lower-limb spasticity and weakness, that result from the degeneration of corticospinal motor neurons (Blackstone, 2012, 2018; Fink, 2006, 2014; Finsterer et al., 2012; Harding, 1984, 1993; Klebe et al., 2015; Tesson et al., 2015)_.

Although the pathophysiology of HSP is well described, (Blackstone, 2012, 2018; Elsayed et al., 2021; Fink, 2006, 2014; Finsterer et al., 2012; Giudice et al., 2014; Harding, 1993, 1984; Klebe et al., 2015; Parodi et al., 2018; Tesson et al., 2015), little is known about its aetiology. A maternal effect mutant screen in zebrafish (Dosch et al., 2004) generated several lines characterised by opaque eggs. Here, we show that one of these, ruehrei^p25ca, possess a single base change that changes the conserved isoleucine at position 525 of Tomm70 to a threonine (Tomm70^Ile525Thr). Mutation of the corresponding isoleucine to phenylalanine (TOMM70^Ile554Phe) was reported to lead to dystonia, hypotonia, hyper-reflexia, ataxia, dysarthria, tremor, ptosis and white-matter abnormality in a human patient (Dutta et al., 2020). Here, we used this zebrafish model to examine the effects of this missense mutation at the molecular, neuronal, physiological and behavioural levels, providing insights into its potential role in neurodegenerative processes.

Using bimolecular fluorescence complementation (BiFC) assay, we found that, in zebrafish, the mutation leads to a partial impairment in the interaction between Tomm70 and Lam6 at the ER-MCS. Reduced interaction with this sterol transporter was associated with a lowered level of cholesterol in the opaque eggs produced by the tomm70 (also known as tomm70a) mutant zebrafish. Primary brain neuronal culture studies revealed that the non-synonymous mutation also affected the transport of mitochondria into axons and dendrites. Using high-pressure freezing and electron microscopy, we demonstrated that the mutation affects the integrity of the myelin sheath and that tomm70 mutants show signs of demyelination of the large calibre axons in the spinal cord. These physiological changes were accompanied by locomotion defects. The tomm70 mutants could not maintain continuous swimming for extended periods, owing to their reduced endurance and diminished propulsion efficiency. The tomm70 mutants also showed defective axial bending during the C-start escape response. This indicated that Mauthner neurons, giant rapid-firing neurons that drive this rapid escape response, were also affected by the mutation in the tomm70 gene.

Hence, in our zebrafish model, we show how a missense mutation alters the functioning of an ER-MCS protein, leading to demyelination of the spinal cord neuron and subsequent functional degeneration. This provides evidence that mutations affecting the function of tomm70 causes neurodegenerative symptoms.

The zebrafish mutant ruehrei^p25ca, identified in a maternal-effect mutant screen, shows defective stage IV oocyte maturation (Dosch et al., 2004). The mutant produces opaque eggs that fail to undergo cell division, suggesting a critical role for the affected gene in oogenesis. We used RNA sequencing and analysis to show that the ruehrei^p25ca mutants carry a single point mutation in the tomm70 gene. This mutation results in a base change from thymine to cytosine, leading to an amino acid substitution from isoleucine to threonine at position 525 in the C-terminus of the protein Tomm70 (Fig. S1A,B). The amino acid isoleucine affected by the Tomm70^Ile525Thr mutation is highly conserved across vertebrate orthologues of Danio rerio Tomm70, including human TOMM70 (Fig. S1C). In human TOMM70, the equivalent isoleucine resides at amino acid position 554, and its substitution by phenylalanine, TOMM70^Ile554Phe, causes neurological defects in a patient, including involuntary muscle contraction, decreased muscle tone, abnormal posture, spastic tendencies, white-matter abnormalities and ataxia (Dutta et al., 2020). Using western blot analysis, we found that the Danio rerio Tomm70^Ile525Thr mutation reduced the abundance of the Tomm70 protein in the brain of adult female and male tomm70 heterozygous mutants. The reduction in protein abundance intensified in homozygous mutants to ∼50% (Fig. S1D-G).

To explore whether the Danio rerio Tomm70^Ile525Thr mutation compromises Tomm70 protein structure, we compared the predicted structures of wild-type and mutant protein using the PyMol application and the template modelling (TM)-align server of Iterative Threading ASSEmbly Refinement (I-TASSER) (Zhang and Skolnick, 2005). A value of 0.233 for the root-mean-square deviation of atomic positions and 0.83586 for TM score, a measure of similarity between two protein structures, indicated that the mutation has no effect on the structural conformation of the Tomm70 protein (Fig. S2A-C).

Tomm70 is an outer mitochondrial membrane protein, known to interact with Tomm40, another mitochondrial protein, and with Lam6, an ER protein responsible for sterol transport (Murley et al., 2015; Tang et al., 2019). To assess whether the mutation might affect these interactions, we used a BiFC assay, in which two proteins are tagged with unfolded complementary N- and C-terminal fragments of a fluorescent protein reporter (Kerppola, 2006, 2008). We used N-terminal and C-terminal split fragments of the yellow fluorescent protein Venus as the reporter, termed ‘VN’ and ‘VC’, respectively (Fig. S3A). Danio rerio Tomm70 was tagged with VN and yeast Lam6 with VC. We cloned the LAM6 gene from yeast genomic DNA because, for Danio rerio, there are eight different Lam6 homologues with different transcripts. Four different combinations were tested: Tomm70-VN+VC-Lam6, Tomm70-VN+Lam6-VC, VN-Tomm70+VC-Lam6 and VN-Tomm70+Lam6-VC. Of these four different combinations, Tomm70-VN+VC-Lam6 was the only combination that generated fluorescence (Fig. S3C and Fig. S4A-F). We repeated the experiment using all four possible combinations with mutant Tomm70^Ile525Thr, and, in this case as well, fluorescence was observed only for Tomm70^Ile525Thr-VN+VC-Lam6 (Fig. S3C and Fig. S4A-F). Quantification revealed a significant decrease in the percentage of fluorescent embryos for the mutant Tomm70 combination compared to the wild-type combination (Fig. S3D). The reduced fluorescence indicates that the Tomm70^Ile525Thr mutation likely increases the spatial separation between Lam6 and Tomm70.

Next, we investigated whether the Tomm70^Ile525Thr mutation affects the interaction with Lam6 only or also that with Tomm40. We tagged Tomm40 with VC and tested all four possible combinations. In this control experiment, strong fluorescence signals were observed when VN was attached to the C-terminus of Tomm70 with both Tomm40-VC and VC-Tomm40 (Fig. S3E and Fig. S5A). On quantification, we found no change in the percentage of fluorescent-positive embryos between wild type and mutant for the two combinations (Fig. S3F and Fig. S5B). The attachment of VN to the N-terminus of Tomm70 yielded only minimal fluorescence, rendering any fluorescence differences between mutant and wild type negligible, given the considerably lower absolute fluorescence levels observed (Fig. S5C-F).

Yeast Lam6 reportedly acts as a transporter of sterols at the ER-MCS (Murley et al., 2015). Because the Tomm70^Ile525Thr mutation impairs the interaction with Lam6, we wondered whether it might affect sterol levels, such as those of cholesterol. Using gas chromatography-mass spectrometry (GC-MS), we determined cholesterol levels in the eggs of wild-type fish and tomm70 mutants. Cholesterol levels were significantly lowered in the mutant opaque eggs (Fig. S3H).

As our study revealed a partial loss in the interaction of Tomm70 with Lam6 and a potential consequence on sterol levels, we investigated the effects of this mutation on the central nervous system. We used primary neuronal culture and immunocytochemistry to examine the presence of Tomm70 in the brain neurons of the mutant fish. Fig. 1A illustrates the different types of neuronal staining with the anti-Tomm70 antibody. In order to quantify the staining, we devised four categories: category 1 represents the most severe phenotype, with antibody signals for Tomm70 only present in the soma and absent from the axon; category 2 includes neuronal staining in which the Tomm70 signal was observed in the soma and the initial part of the axon; in category 3, Tomm70 signal was present in the soma and halfway along the axon; and category 4 represents the neuronal staining with signal for Tomm70 in the soma and throughout the axon. Fig. 1B represents typical neuronal staining found in female wild-type fish, which predominantly belonged to category 4. Fig. 1C-E display representative images of neuronal staining predominantly observed in mutant female fish brains, corresponding with categories 1, 2 and 3, respectively. More than 80% of the neuronal staining in the brain of wild-type female fish belonged to category 4, whereas there was a significant decrease in the percentage of staining in this category in the mutant female fish. Conversely, there was a significant increase in the percentage of staining in categories 1, 2 and 3 in mutant female fish compared to in the wild-type female fish (Fig. 1F). Compared to that in wild types, there was still a significant increase in the percentage of neuronal staining in mutants when the data for categories 1, 2 and 3 were pooled (Fig. 1G). Although male fish showed no significant changes in categories 2-4, the most severe phenotype, category 1, was significantly increased in male mutants compared to male wild types (Fig. S7F). All trends were similar to those of the female data, and pooled categories were significantly different (Fig. S7G). These findings revealed absence of Tomm70 from the axons in a length-dependent manner.

Mitochondria with mutated Tomm70^Ile525Thr protein could not reach axon terminals (Fig. 2A,B), showing that the mutation blocks transport into axons. Mitochondria without Tomm70 protein moved freely (Fig. 2C). Similar disruptions were observed in dendrites (Fig. 3A-D) of multipolar neurons. The male mutants exhibited identical effects to those seen in female mutants (Fig. S8A-C and Fig. S9A-D).

To investigate whether the absence of Tomm70 from the mitochondria affects the structure of myelin in the large calibre axons of the spinal cord, we used electron microscopy. Two characteristic pathological features in myelin, splitting and vesiculation, were observed in the large calibre axons in the cranial portion of the spinal cord of mutants. Representative pictures of intact myelin in the cranial region of the spinal cord of wild-type fish are shown in Fig. 4A,D. Fig. 4B,E illustrate splitting and vesiculation of the myelin sheath in the cranial region of the spinal cord in mutant zebrafish. Although zebrafish myelin is inherently less compact than mammalian myelin (Siems et al., 2021), and artefacts can affect the quality of electron micrographs, quantification showed a significant increase in splitting events (Fig. 4C) and a slight increase in vesiculation of the myelin sheath in mutants compared to wild types (Fig. 4F). Although breaks in the myelin sheath can occur due to artefacts, we observed a significant increase in break events in the caudal region of the spinal cord in mutants compared to wild types (Fig. S10A-C).

To examine whether mutant fish exhibit reduced endurance, we replicated the natural environmental conditions of zebrafish, whereby they swim against the water currents (Arunachalam et al., 2013; Parichy, 2015; Spence, 2011). Our experimental setup enabled fish to swim against an incoming water stream, strongest in the centre and weaker along the sides. In the two-dimensional heat map with the marginal histogram in Fig. 5A, blue represents the lowest probability density, and yellow represents the highest probability density, for the animal's presence at a given position within the tank. The results revealed that wild-type fish preferred the centre of the current. In contrast, the heterozygous mutant fish resided at an intermediate distance from the centre of the stream. Homozygous fish positioned themselves even further away from the centre of the stream, confirming that they possess less endurance and cannot constantly swim against the water current (Fig. 5A,B; Fig. S11A-C). This endurance reduction was further corroborated by observations during motivated swimming trials: tomm70 mutant fish, particularly females, were unable to maintain constant swimming for 30 s, with most ceasing activity after ∼15-20 s (Fig. 5C).

Upon further analysis, a notable decrease in the propulsion efficiency emerged when comparing heterozygous and homozygous mutants to their wild-type counterparts. The primary decrease in propulsion efficiency was observed in heterozygous fish, with only a minimal additional decrease in the homozygous fish. This decline in propulsion efficiency was consistently observed across different test conditions, encompassing free and motivated swimming trials, as depicted in Fig. 5D,E. The disparity in propulsion efficiency was more pronounced in male specimens, a finding that aligns with the known sexual dimorphism in swimming velocities, as detailed in Garg et al. (2022). Propulsion efficiency, in this context, is quantified as the forward movement achieved per degree of body bending. Forward movement of the fish's body is intrinsically linked to the extent of its body bending (Gray, 1933; Lauder and Tytell, 2005; Muller et al., 2000; Müller and Van Leeuwen, 2006; Smits, 2019; Wardle et al., 1995). The observed decrease in propulsion efficiency in mutant fish thus suggests a fundamental alteration in their locomotive mechanics compared to that of wild types. Mutant fish have to bend their body more in order to move forward at the same speed as wild types, resulting in reduced propulsion efficiency and endurance.

In order to investigate potential differences in locomotion behaviour between mutant tomm70 fish and wild-type controls, we further compared the well-known C-start escape response between the two. The giant Mauthner cells facilitate this escape response, allowing the fish to rapidly flee from potential harm (Eaton et al., 1977; Faber and Korn, 1978; Korn and Faber, 2005; Zottoli and Faber, 2000). Notably, these large-calibre neurons and other related cells produce high-amplitude action potentials, the field potentials of which can be detected by electrodes placed in the fish tank (Issa et al., 2011).

Wild-type fish exhibited an immediate pivot and surge in speed (Fig. 6A; Fig. S12A). This instantaneous reaction was visually captured as a widely spaced silhouette trajectory, denoting swift movement away from the perceived threat (Fig. 6A). In contrast, the heterozygous mutants displayed a more subdued response, and the silhouettes appeared more closely clustered (Fig. 6B; Fig. S12B). The trajectory of homozygous mutants lacks the distinct sharp angulation associated with the C-start, suggesting a compromised or entirely absent reflexive response to the stimuli (Fig. 6C; Fig. S12C).

The discernible delay in orientation alteration and the less abrupt acceleration profile of mutants (Fig. 6B,C, bottom) are accompanied by an increased neuronal stimulus latency (Fig. 6D) and a reduction in the number of field potentials (Fig. 6E). Heterozygous fish exhibited a subtle phenotype in spike count and latency compared to that of homozygous fish. The slower and less pronounced neuronal activity of mutants correlated with a significant decrease in their bending ability (Fig. 6F), which is crucial for the C-start response. These findings suggest that the altered physiology observed in mutant tomm70 fish, as evidenced by the BiFC assay (Fig. S3) and immunocytochemistry (Fig. 1), not only impacts the neurons themselves (Fig. 4) but also directly influences locomotion behaviour (Fig. 5) and its neuronal control, as demonstrated by the changes in the flight response (Fig. 6).

To further substantiate that the locomotion defects in the tomm70 mutants stem from neuronal origins rather than skeletal deformations, we utilised micro-computed tomography (CT) scanning to conduct a comprehensive examination of the vertebral columns in wild-type and tomm70 mutant fish. Our analysis revealed no discernible differences in skeletal structure between wild types and heterozygous or homozygous mutants, regardless of sex (Fig. S13A,B). Additionally, we investigated the muscle area in the caudal region of the fish body. However, no difference was detected between wild-type, heterozygous and homozygous fish, suggesting absence of muscular atrophy in the caudal region of the mutant fish body (Fig. S14A,B). Taken together, our findings support the notion that the locomotion defects in tomm70 mutants arise from neuronal dysfunctions.

Tomm70 is an outer mitochondrial membrane transport protein principally involved in mediating the uptake of proteins into the mitochondria (Kreimendahl and Rassow, 2020) and interacts with a sterol-transporter ER protein called Lam6 at the ER-MCS (Murley et al., 2015). A mutation in the C-terminus of Tomm70, which results in a conserved amino acid change from isoleucine to threonine in zebrafish (Tomm70^Ile525Thr; Fig. S1A-C), has been identified in a human patient (TOMM70^Ile554Phe) exhibiting similar symptoms as those documented here, such as movement disorders and spasticity (Dutta et al., 2020). The absence of any impact on the structural conformation of the Tomm70 protein (Fig. S2A-C), but instead impact on its abundance (Fig. S1D-G), suggests that the mutation's consequences are functional rather than structural.

Using the BiFC assay, we could confirm that the mutation affects the interaction of Tomm70 with Lam6 specifically, as the interaction of Tomm70 with Tomm40 is unaffected (Fig. S3C-F). As Lam6 functions as a sterol transporter (Murley et al., 2015), we hypothesise that loss of the interaction of Tomm70 with Lam6 impacts cholesterol abundance in cholesterol-rich organs of the mutant fish. A significant reduction in cholesterol abundance was observed in the opaque eggs produced by the tomm70 mutants (Fig. S3H). The observed anomalies suggest that a dysfunction in yolk cholesterol processing at stage IV of oogenesis interrupts the maturation process of oocytes, impeding the typical progression to a transparent state and leading to sustained opacity (Dosch et al., 2004).

Lipid metabolism disturbances, along with impaired transport of critical components such as RNA, proteins and organelles to the axon and axon terminal, are known to contribute to neurodegeneration (Guo et al., 2020; Liu et al., 2012). Among the organelles, mitochondria play a crucial role in ATP and metabolite production, thereby regulating cellular energy levels and metabolic pathways. Impaired mitochondrial transport machinery can lead to progressive neuronal degeneration (Guo et al., 2020; Liu et al., 2012; Maday et al., 2014; Saxton and Hollenbeck, 2012).

In wild-type neurons, Tomm70 is incorporated into the outer mitochondrial membrane (Eaglesfield and Tokatlidis, 2021), and these mitochondria can freely travel into the axon and its terminals. In fish carrying the mutant Tomm70^Ile525Thr, mitochondria that incorporate the mutant protein appear to be retained near the soma and are unable to reach the axon terminals. However, in mutant neurons, mitochondria that do not incorporate any Tomm70 protein can still be transported freely (Fig. 2A-C). This is likely because the overall abundance of Tomm70 is significantly reduced in mutants, resulting in many mitochondria lacking Tomm70 altogether.

The absence of Tomm70 in mitochondria transported to axons could profoundly impact their function, as Tomm70 is crucial for importing numerous essential proteins into mitochondria (Kreimendahl and Rassow, 2020). Our data suggest a possible sexual dimorphism in this phenotype, warranting further investigation. The phenotype in females (Fig. 1) seems to be more pronounced than that in males (Fig. S7). Additionally, Tomm70 is part of the ER-MCS, facilitating the transport of sterols and Ca2+ into mitochondria (Filadi et al., 2018; Murley et al., 2015), and its absence could disrupt these processes. Furthermore, dendritic mitochondria are essential for maintaining synapses (Li et al., 2004), and the absence of Tomm70 could contribute to synaptic instability. As the mutation does not appear to affect the protein's structural conformation, further investigation is needed to determine how the mutation influences the transport of mitochondria to axons and dendrites and identify the processes most affected by this alteration.

Absence of Tomm70 protein from the distal parts of long axons can affect the downstream cellular processes for which the supply of Tomm70 or mitochondria is essential. One such process is the formation and maintenance of myelin sheath surrounding the large calibre axons. Myelin pathology and demyelination are common features of the degeneration of neurons in many neurodegenerative disorders (Weil et al., 2016). Although myelin composition in zebrafish differs from that in other vertebrates (Siems et al., 2021), our findings reveal that prominent signs of pathology, such as splitting and vesiculation of myelin sheath, were more frequently observed in large-calibre spinal cord axons of tomm70 mutants than in those of wild types (Fig. 4B-F). This suggests that the integrity of the myelin sheath is affected in the absence of Tomm70. Given the essential role of mitochondria in oligodendrocyte function and myelin sheath development (Meyer and Rinholm, 2021), our observations suggest that the myelin pathology and absence of Tomm70 in axons are functionally related.

The physiological changes induced by the mutation in tomm70 mutants had notable effects on their motor behaviour. The tomm70 mutants required substantially greater bending to generate thrust velocities comparable to wild-type fish (Fig. 5D,E), suggesting a more energetically demanding locomotion that contributes to their diminished endurance and premature cessation of flight behaviour (Fig. 5B,C). The mutation also affects the Mauthner cell-governed C-start escape response. Electrophysiological analysis of the neuronal activity during C-starts revealed that there is an increase in the latency of the field potentials and a reduction in the number of spikes (Fig. 6D,E). Activating the escape response revealed that tomm70 mutant fish are less able to bend their bodies than their wild-type counterparts (Fig. 6F). Heterozygous mutants display intermediate locomotion phenotypes compared to homozygous fish, but the latency during C-starts is ten times higher in homozygous fish, suggesting a greater impact of the mutation on Mauthner cells than on other motor neurons. As Mauthner cells are considered functional homologues of the mammalian corticospinal tract (Davis and Farel, 1990; Will, 1986), these findings suggest a potential evolutionary link between the two systems. Furthermore, our results indicate that the mutation not only affects the motor neurons themselves but also directly influences the locomotion activity governed by them, highlighting the mutation's impact on both neuronal structure and function.

Our analysis revealed no significant differences in the vertebral column anatomy between wild-type, heterozygous and homozygous tomm70 mutant fish (Fig. S13A,B), confirming that the alterations in locomotion behaviour stem from motor neuron degeneration rather than skeletal deformities. Additionally, the absence of significant caudal muscle area reduction in tomm70 mutants (Fig. S14A,B) further implies that the behavioural phenotypes are likely rooted in neuronal control factors.

By examining the effects of a point mutation in the ER-MCS protein Tomm70 across multiple biological layers in zebrafish, our study provides insights into potential mechanisms of motor neuron dysfunction. In light of these findings, the report of neurodegenerative symptoms in a human patient with an identical variant (TOMM70^Ile554Phe; Dutta et al., 2020) and the previous prediction of tomm70 as a potential causal gene for HSP (Novarino et al., 2014), we propose its inclusion in genetic screenings to improve the early detection and diagnosis of neurodegenerative disorders.

Zebrafish were maintained according to the guidelines provided by the Westerfield zebrafish book (Westerfield, 2000) and EuFishBioMed/Federation of European Laboratory Animal Science Associations (FELASA) (Aleström et al., 2020), in compliance with the regulations of Georg-August-University Göttingen and Bielefeld University, Bielefeld, Germany (540.4/16 25/Engelmann). The zebrafish experiments were approved by the Lower Saxony State Office for Consumer Protection and Food Safety (AZ14/1681/Dosch) and conducted according to European Union directive 2010/63/EU.

The ruehrei^p25ca mutant we analysed in this study was originally identified through a maternal-effect mutant screen in zebrafish (Dosch et al., 2004). We identified this mutant as Tomm70^Ile525Thr by RNA sequencing. Another mutant from this screen, souffle, exhibiting an identical egg phenotype, was found to carry a mutation in the spastizin (also known as zfyve26) gene (Kanagaraj et al., 2014). Prior research in our laboratory showed that neurodegenerative symptoms in spastizin mutants only manifest in late-stage adults (Garg et al., 2024). Consequently, we conducted our study on Tomm70^Ile525Thr adult fish aged 12 to 18 months to capture a more accurate assessment of these late-onset symptoms.

We extracted the total RNA from the ovary of wild-type and ruehrei^p25ca mutant fish. The mRNA was enriched by selecting poly(A) tails using oligo(dT) beads, thereby removing ribosomal RNA and mitochondrial RNA. Once mRNA was isolated, it was converted into sequencing libraries using a TruSeq RNA Library Prep Kit v2 (Illumina, Germany) and subjected to sequencing using an Illumina HiSeq 2000. Sequence images were transformed with Illumina software BaseCaller to base call (BCL) files, which were demultiplexed to fastq files with consensus assessment of sequence and variation (CASAVA) (v1.8.2). Subsequently, reads were aligned by spliced transcripts alignment to a reference (STAR) (2.3.9e) (Dobin et al., 2013) to the Ensembl Danio rerio genome (version Zv9). Duplicates were eliminated and single-nucleotide polymorphism (SNP) piled up by samtools (0.1.19). VarScan (v2.3.6) (Koboldt et al., 2012) was applied for SNP calling and limiting SNP areas. Data underwent pre-processing and analysis within the R/Bioconductor environment (3.0.2/2.14). Mutation effects were identified via the Ensembl variant effect predictor (McLaren et al., 2010).

Multiple sequence alignment was performed using Clustal Omega program (Sievers et al., 2011). The I-TASSER server was used to predict the structure of wild-type and mutant Tomm70 protein (Yang and Zhang, 2015; Zhang et al., 2017; Zhou et al., 2022). The TM-align server of I-TASSER (Zhang and Skolnick, 2005) and PyMOL application were used to align the predicted model for mutant and wild-type Tomm70.

Whole-brain samples from each genotype were prepared for western blotting by homogenising the tissue in a mix of protease inhibitor (Thermo Fisher Scientific, Germany) and radio-immunoprecipitation assay (RIPA) buffer [0.1% sodium dodecyl sulphate (SDS; Serva Electrophoresis, Germany), 1% Nonidet P-40 (NP40; Merck/Sigma-Aldrich, Germany), 1% sodium deoxycholate (Merck/Sigma-Aldrich), 150 mM sodium chloride (NaCl; Carl Roth, Germany), 50 mM Tris (hydroxymethyl) aminomethane (THAM) hydrochloride (Tris-HCl; pH 7.2-7.5; Carl Roth)] in a ratio of 1:100. The lysates were centrifuged at 10,000 g for 10 min at 4°C. The supernatant was collected, and samples were subjected to Bradford assay for estimating the protein concentration (Bradford, 1976). Then, 30 µg lysate from each sample type was separated on a 8% SDS-PAGE gel at 30 mA for 1-1.5 h and blotted on nitrocellulose membrane (Bio-Rad, Germany) at 180 mA for 1.5 h. The membrane was blocked for 1 h at room temperature in 4% powdered milk (Carl Roth) in phosphate buffered saline (PBS; ChemSolute, Germany) with 0.1% Tween (AppliChem, Germany) (PBST). For this study, anti-β-Tubulin antibody was used as the loading control because Tomm70 and β-Tubulin are nearly in the same size range, and the available antibodies were both raised in mouse. We incubated the blot with anti-Tomm70 primary antibody (1:250 or 1:300; Proteintech, Germany) in 4% milk in PBST overnight at 4°C, and then horseradish peroxidase (HRP) anti-mouse secondary antibody (1:2500; Invitrogen, Germany) in 4% milk in PBST for 30 min at room temperature. After capturing the images for the Tomm70 signal, the membrane was washed with PBST for 5 min and then with stripping buffer [0.5 M NaOH (AppliChem, Germany) in water] for 5 min. It was followed by two more washes with PBST for 5 min each. The membrane was again blocked for 1 h in 4% milk in PBST, followed by incubation with anti-β-Tubulin (1:1000; Developmental Studies Hybridoma Bank, USA) and HRP anti-mouse secondary antibody (1:2500; Invitrogen) for 30 min each. Images were captured and analysed using iBright CL1000 (Invitrogen) and ImageJ/FIJI (Schindelin et al., 2012), respectively.

The investigation of Tomm70 interactions with Lam6 and Tomm40 was conducted using the BiFC assay (Kerppola, 2006, 2008). In this study, Venus served as the fluorescent reporter protein. Its N-terminal half (VN) was fused to both the N- and C-termini of Tomm70, whereas its C-terminal half (VC) was fused to the N- and C-termini of Tomm40 and Lam6. All fusion constructs were generated using either the Gateway or In-Fusion cloning method, as previously described (Perera and Dosch, 2021). For Tomm70 and Tomm40 fusion constructs, templates were amplified from complementary DNA, derived from the reverse transcription of ovarian RNA. The template for Lam6 was amplified from the genomic DNA of yeast.

A total of 2 nl capped sense RNA, diluted with 0.1 M KCl and 0.05% Phenol Red (Merck/Sigma-Aldrich), was injected into one-cell-stage zebrafish embryos using PV820 WPI injecting apparatus (Sarasota, USA). Injected embryos were raised at 28°C in E3 medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl₂ and 0.33 mM MgSO₄). The injected embryos were sorted and manually counted and imaged 7-8 h post fertilisation for fluorescence using the green fluorescent protein (GFP) channel of a fluorescence stereomicroscope (Lumar V12, Zeiss, Germany). Images were processed using ImageJ/FIJI (Schindelin et al., 2012).

The transparent and opaque eggs produced by wild-type and mutant zebrafish were flash frozen in liquid nitrogen in pre-weighed Eppendorf cups and stored at −80°C until further processing. The weight of samples was determined after lyophilisation. Lyophilised samples were ground in a ball mill to make a fine powder. For the two-phase extraction process, 400 µl methyl-tert-butyl ether [methanol (3:1; v/v; all solvents were high-performance liquid chromatography grade (Thermo Fisher Scientific)] was added, followed by vortexing and addition of 200 µl ultra-pure water. As an internal standard, 5 µg 17:0 free fatty acid (Merck, Germany) was added, and samples were extracted for 30 min. The samples were then centrifuged at maximum speed (21,130 g) for 1 min, and the resulting upper phase was transferred to a new tube and stored at −20°C until further processing. For GC-MS measurements, 20-50 µl of the upper phase was evaporated under a stream of nitrogen, redissolved in 10-15 µl anhydrous pyridine (Merck) and derivatised with twice the volume of N-methyl-N-trimethylsilyltrifluoroacetamid (Merck) to yield trimethylsilylated analytes. Samples were analysed on an Agilent 7890B gas chromatograph connected to an Agilent 5977N mass-selective detector, as described (Berghoff et al., 2021). Cholesterol was identified in comparison to an external standard. For quantification, GC/MSD Mass Hunter software with MSD ChemStation Data Analysis (Agilent Technologies) was used. For the internal standard and cholesterol, the mass-to-charge ratios 327 Da/e and 458 Da/e, respectively, were quantified as target ions. Values for cholesterol were normalised to the internal standard and the mass of the sample.

The protocol for primary neuronal cultures for zebrafish brain was adapted from studies with insect neurons (Knorr et al., 2020, 2022). One whole brain per culture from wild-type and tomm70 mutants was dissected and collected in Leibovitz 15 (L-15) medium (Gibco, Life Technologies, Germany) supplemented with 1% penicillin/streptomycin (Merck/Sigma-Aldrich). Then, the samples were enzymatically digested with collagenase/dispase (2 mg/ml; Merck/Sigma-Aldrich) for 30 min at 27°C. The enzymatic reaction was stopped by replacing the solution with Hank's balanced salt solution (Thermo Fisher Scientific). The brain stem was triturated in the Hank's solution and centrifuged for 3 min (2191 g), after which a pellet was formed. After removal of supernatant, the pellet was resuspended in L-15 medium, and the cell suspension was seeded on a concanavalin A (Merck/Sigma-Aldrich)-coated coverslip and incubated for 1.5 h. Subsequently, culture dishes were filled with L-15 medium and supplemented with 4% foetal bovine serum gold (PAA Laboratories GmbH, Austria). Medium was changed every second day for ∼2 weeks until the neurons developed long axons. The cultures were maintained at 27°C without CO₂ buffering.

Human embryonic kidney (HEK) cells were grown in Dulbecco's modified Eagle medium (DMEM; Thermo Fisher Scientific) supplemented with 10% foetal calf serum (DMEM+; Merck/Sigma-Aldrich) and split when the confluence reached ∼95%. Old medium was removed by aspiration, and cells were washed with 1 ml 1× PBS. To detach the cells, 750 µl 1× trypsin (Thermo Fisher Scientific) was added. Detached cells were washed with 5 ml DMEM+ and pelleted down at 300 g for 5 min at 24°C. Pelleted cells were resuspended in 1 ml fresh DMEM+, and 200-250 µl was seeded in a new flask. For transfection, 70-100 µl of the cells was seeded on concanavalin A (Merck/Sigma-Aldrich)-coated coverslips and filled with 1.4 ml DMEM+. An Effectene transfection kit (Qiagen, Germany) was used to transfect the cells with mCherry-Mito-7 (Addgene #55102) plasmid. The transfection mix was prepared by adding 94 µl EC-buffer, 500 ng plasmid and 4 µl enhancer for one well of a six-well plate. The compounds were mixed by vortexing for 5 s and incubated for 5 min at room temperature. Then, 5 µl Effectene was added and mixed by flipping the vial manually, followed by incubation for 8-10 min at room temperature. This mixture was then added to the cells before incubating at 37°C with 5% CO₂. HEK cells were seen to express the protein within 24-48 h.

As neuronal development progressed in the primary culture, cells were fixed with 400 µl 4% paraformaldehyde (PFA; Carl Roth) for 30 min. Unless otherwise stated, all washes were for 5 min each. Coverslips were first washed with PBS three times, then five times with PBST [PBS containing 0.1% Triton X-100 (Merck/Sigma-Aldrich)]. Cells were blocked with 300 µl blocking solution [5% normal goat serum (NGS; Jackson ImmunoResearch, UK), 0.25% bovine serum albumin (BSA; Carl Roth) in PBS containing 0.3% Triton-X 100] for 1 h at room temperature. Subsequently, the cells were incubated with anti-Cytochrome c antibody (1:450; Proteintech) in blocking solution overnight at 4°C. After washing coverslips five times with PBST and three times with PBS, the cells were incubated with goat anti-mouse Alexa Fluor 633 (1:1500; Invitrogen) and 4′,6-diamidino-2-phenylindole (DAPI; 1:1000; Merck/Sigma-Aldrich) in PBST for 1 h at room temperature. After five washes with PBS, the cells were stained with CoraLite Plus 488-conjugated anti-β-Tubulin (1:450 or 1:500; Proteintech) and CoraLite Plus 594-conjugated anti-Tomm70 (1:400; Proteintech) primary antibodies for 3-4 h at 4°C. Then, again after washing, coverslips were flipped upside down and mounted with 1,4-diazabicyclo[2.2.2]octane (DABCO; Carl Roth), and slides were stored at 4°C. As anti-β-Tubulin and anti-Tomm70 antibodies are directly conjugated to fluorophore, for staining neurons with only these two antibodies, the protocol remained the same, except for adding no secondary antibodies but only DAPI after the primary antibody.

To confirm the specificity of the anti-Cytochrome c antibody, coverslips with HEK cells were transferred to four-well plates and fixed with 300 µl 4% PFA for 10 min. After washing cells three times with PBS and then twice with PBST, they were blocked with blocking solution (5% NGS, 0.25% BSA in PBS containing 0.3% Triton-X 100) for 1 h at room temperature. Following this, cells were incubated with anti-Cytochrome c antibody (1:300; Proteintech) in blocking solution for 30 min at 4°C and washed twice with PBST and twice with PBS. Next, cells were incubated with secondary goat anti-mouse Alexa Fluor 488 (1:1500; Life Technologies, Germany) and DAPI (1:1000) in PBST for 30 min at room temperature. After three washes with PBS, coverslips were flipped upside down and mounted with DABCO, and slides were stored at 4°C. In the case of the anti-Cytochrome c and mCherry-Mito-7 co-staining, we could directly observe the signal overlap and thereby confirm the specificity of our anti-Cytochrome c antibody (Fig. S6A). CoraLite Plus 594-conjugated anti-Tomm70 antibody has a similar emission range to mCherry, so to make sure that we saw a true signal for this antibody, we quenched the original mCherry signal by fixing cells in 100 µl more 4% PFA and for 5-10 more min. The rest of the procedure was the same as for anti-Cytochrome c antibody, until the blocking step. After blocking, cells were incubated with CoraLite Plus 594-conjugated anti-Tomm70 (1:400; Proteintech) and Fluo Tag X4 anti-red fluorescent protein (RFP), Alexa Fluor 647 (1:150; NanoTag, Germany) (to recognise the mCherry) antibodies. This process led to an overlay of the anti-mCherry signal with the signal from anti-Tomm70 antibody (Fig. S6B). The distinct overlay of the different fluorescent signals confirmed the accuracy of our antibody targeting, thus validating our experimental approach.

Images of primary brain neurons and HEK cells were acquired using a TCS Sp8 confocal microscope (Leica, Germany) with 63× glycerol immersion objective and scanning resolution of 1024×1024 pixels. Axons positive for the Tomm70 and Cytochrome c signal were counted manually, and the images were processed using ImageJ/FIJI (Schindelin et al., 2012).

After sacrificing zebrafish, the spinal cord was dissected and separated into three segments using the number of vertebrae for orientation. A 3 mm-long segment of the upper and lower part of the cord was cryofixed in 20% polyvinylpyrrolidone (molecular mass 10,000 g/mol) (Sigma-Aldrich, Germany) in PBS, using a Leica HPM100 high-pressure freezer. The sections were then freeze substituted using a Leica AFS2 and Epon (Serva Electrophoresis) embedded, as previously described (Weil et al., 2019). Then, 0.5 µm semi-thin or 50 nm ultra-thin sections of the embedded tissue were cut using a Leica UC7 ultramicrotome, which were contrasted with UranylLess stain (Science Services GmbH, Germany). Electron micrographs were acquired using a Zeiss LEO EM912AB equipped with a wide-angle dual speed 2k-charged-coupled device camera (TRS, Germany). Images were marked manually for signs of demyelination using the cell counter feature of ImageJ/FIJI.

Genie HM1024 high-speed camera (Dalsa Imaging Solutions GmbH, Germany) linked with an Optem Zoom 125C 12.5:1 Micro-Inspection Lens System was used to record the movement of zebrafish from above. A light-emitting diode light plate (Lumitronix, Germany) and an aquarium light control (Elektronik-Werkstatt SSF, University of Göttingen, Göttingen, Germany) were used for illuminating the setup from below. Recordings were conducted between 10:00 and 20:00 in the diurnal rhythm.

The experiments for free and motivated swimming trials were conducted in a 24.9×11.4 cm acrylic glass aquarium with a shallow water depth of 1.6 cm. Videos for these experiments were recorded for 30 s at 200 frames/s. The recording for cruising started 30 s after transferring the fish to the setup tank. The motivated trial started directly after cruising recording ended by tossing a 474 g metal weight through a plastic tunnel, which struck the setup table and created a mechanical stimulus of 18.7 N on the surface.

To generate a constant water flow, we used a custom-built 17.2×4.4 cm Plexiglas^® aquarium fitted with an installed water pump, running at 150 ml/s. The movement of fish against the water stream was recorded for 30 s at a frame rate of 501 frames/s.

To measure the electric field potential from the neurons of freely behaving zebrafish, we adapted the experimental setup from Issa et al. (2011). The tank used for this study was 8×4×4 cm. The escape response of the fish was evoked using the water jet produced by a picospritzer pressure pump (Parker Hannifin, USA). The electric signal was amplified 2000 times and band-pass filtered with a pass window of 300-500 Hz, and 50 Hz noise was additionally filtered out using a Hum Bug (Quest Scientific, USA). Milli-Q water was used to achieve a resistance of ⁠. The filtered and amplified electric signals were recorded using a micro2 1401 (Cambridge Electronic Design, UK) data acquisition system and subsequently analysed with Spike2 software (Cambridge Electronic Design). The picospritzer pump triggered the video camera, and the movement of animal in response to the water jet was recorded for 5 s at a frame rate of 923 frames/s.

Limbless Animal traCkEr (LACE), a MATLAB R2012b script (MathWorks, USA), was used to track the animals (Garg et al., 2022).

After sacrificing and briefly rinsing in water, zebrafish were transferred to 35% and 70% ethanol for 1 h each. For staining and fixation, fish were placed for ∼10 days at room temperature under slow rotation in a 4% PFA solution (Serva Electrophoresis) in PBS, pH 7.4 (Invitrogen), containing 0.7% phosphotungstic acid solution (Merck/Sigma-Aldrich) diluted in 70% ethanol. For storage, fish were briefly rinsed in water and embedded in 1% agarose (Carl Roth) in 1.8 ml vials (Nunc CryoTube Vials, Merck/Sigma-Aldrich). The specimens were scanned with an in vivo micro-CT system Quantum FX (Perkin Elmer, USA) operated with the following settings: tube voltage, 90 kV; tube current, 200 µA; field of view (FOV), 10×10 mm (Arunachalam et al., 2013); total acquisition time, 3 min; resulting in a reconstructed pixel size of 20 µm and an image matrix of 512×512×512 voxel (Dullin et al., 2017). Two to three consecutive scans were acquired and stitched together using a custom-made Python script that, in addition to finding the perfect overlap, also corrects for drift in between the scans. Data were analysed using ImageJ/FIJI (Schindelin et al., 2012).

Independent t-test (William, 1908) and Wilcoxon rank-sum test (Mann and Whitney, 1947) were used to calculate the statistical significance for western blot and cholesterol measurements, respectively. For all other datasets, Fisher's permutation test (Crowley, 1992; Ernst, 2004; Fisher, 1970) was used, in which samples were bootstrapped 200,000 times. All P-values were corrected with Benjamini–Hochberg false discovery rate correction (Benjamini and Hochberg, 1995).

We thank Gudrun Kracht, Nicola Schwedhelm-Domeyer, Stephanie Pauls, Silvia Gubert and Tatjana Openkowski for technical assistance; Gudrun Kracht, Axel Zigan and Helene Schellenberg for animal care; and Dr Philip Hehlert for his advice and guidance in multiple experiments, including neuronal culturing and staining. We also thank Professor Dr Caroline Beck for her invaluable assistance in revising the manuscript.