Identification of Miro1 and Miro2 as mitochondrial receptors for myosin XIX

Mitochondria are dynamic organelles that can fuse with each other, divide, move along cytoskeletal tracks and make functional contacts with other membranous compartments (Friedman and Nunnari, 2014). Interfering with this dynamic behaviour impairs mitochondrial function and may lead to a multitude of mostly degenerative diseases (Nunnari and Suomalainen, 2012; Mishra and Chan, 2014). In cells, mitochondria are constantly transported to sites where they are needed. This transport occurs either along actin filaments or microtubules depending on the organism. In animal cells, mitochondria are predominantly hauled bidirectionally along microtubules. The microtubule-based motor kinesin-1 (KIF5) transports mitochondria towards the plus-end of microtubules, which usually points towards the cell periphery (Tanaka et al., 1998). KIF5 is recruited to mitochondria with the help of adaptor proteins (Stowers et al., 2002), which mediate the binding to the outer mitochondrial membrane protein Miro (Miro1 and Miro2 in vertebrates; denoted Miro1/2). Miro encompasses a C-terminal transmembrane domain and two EF-hand or ELM domains that are in turn flanked by two GTPase domains (Fransson et al., 2003; Klosowiak et al., 2013; see also Fig. S1). The N-terminal GTPase domain shares similarity with the Rho-subfamily of monomeric GTPases. An adaptor protein that links KIF5 with Miro was first identified in Drosophila (Glater et al., 2006). Subsequently, two mammalian homologues, TRAK1 (OIP106) and TRAK2 (GRIF-1), were identified as adaptors and shown to mediate bidirectional microtubule-dependent transport of mitochondria (Brickley et al., 2005; Fransson et al., 2006; MacAskill et al., 2009a; Koutsopoulos et al., 2010; Brickley and Stephenson, 2011; van Spronsen et al., 2013). Transport of mitochondria towards the microtubule minus-ends is mediated by the dynein-dynactin complex. This complex similarly interacts with the adaptors TRAK1 and TRAK2 (van Spronsen et al., 2013; Gama et al., 2017). The microtubule-dependent transport of mitochondria can be arrested by the binding of Ca²⁺ to the EF-hands of Miro (Saotome et al., 2008; MacAskill et al., 2009b; Cai and Sheng, 2009; Wang and Schwarz, 2009). A point mutation in the N-terminal GTPase domain of Miro that is predicted to favour the GDP-bound state, revealed that this domain is essential for mitochondrial transport and morphology (Fransson et al., 2003; Babic et al., 2015). However, it is currently not known whether this GTPase domain cycles between different nucleotide-bound states under physiological conditions. The protein Alex3 of the Armcx gene family also regulates mitochondrial movements, in neurons, by interacting with the Miro1/2-TRAK2 complex and this interaction is abrogated by Ca²⁺ binding to the EF-hands of Miro1/2 (Lopéz-Doménech et al., 2012). In addition, Miro1/2 can recruit the protein CenpF, and thereby contribute to the mitotic redistribution of mitochondria (Kanfer et al., 2015). The proteins Mitofusin-1/2, DISC1 and APC are yet more proteins that are reported to interact with the Miro-TRAK complex (Misko et al., 2010; Norkett et al., 2016; Mills et al., 2016). The protein hypoxia upregulated mitochondrial movement regulator (HUMMR) was shown to interact directly with Miro1/2 (Li et al., 2009). Importantly, Miro1/2 are also substrates for the E3 ubiquitin ligase Parkin that targets them for proteasomal degradation (Sarraf et al., 2013). Mutations in Parkin have been associated with Parkinson's disease (Kitada et al., 1998).

Actin filaments in mammalian cells critically contribute to mitochondrial fission. The ER-associated formin INF2 in combination with a mitochondria-associated splice form of the actin nucleator Spire and the filament-forming motor myosin II are involved in constricting mitochondria to initiate their fission (Korobova et al., 2013, 2014; Manor et al., 2015; Ji et al., 2015). The myosin MYO6 is recruited to damaged, ubiquitylated mitochondria and promotes F-actin cage assembly to prevent refusion of dysfunctional mitochondria (Kruppa et al., 2018). Additionally, actin filaments serve as tracks for mitochondrial movement. The actin-based motor myosin XIX (Myo19) is associated with mitochondria and when overexpressed induces the formation of motile tadpole-shaped mitochondria (Quintero et al., 2009). Downregulation of Myo19 interferes with the proper partitioning of mitochondria during mitosis, and leads to a stochastic failure in cell division (Rohn et al., 2014). Myo19 consists of a head, light chain binding and tail domain (see Fig. S1). In vitro studies revealed that the head domain is a plus-end directed motor that is firmly attached to F-actin for most of the time of its chemo-mechanical cycle (Lu et al., 2014). The tail domain, on the other hand, specifies its recruitment to mitochondria (Quintero et al., 2009). A short motif in the tail domain (amino acids 860-890) is able to mediate binding to the lipids of the outer mitochondrial membrane (Shneyer et al., 2016; Hawthorne et al., 2016). However, it is not known how targeting specificity to the outer mitochondrial membrane is achieved and whether there is any coordination between actin- and microtubule-based movements of mitochondria.

In the present study, we attempted to gain further insight into the functions of Myo19 by investigating the molecular mechanism(s) of its recruitment to mitochondria. Interestingly, Myo19 shares its mitochondrial receptor with kinesin (KIF5) and dynein that potentially allows for a coordination of actin- and microtubule-based mitochondrial movements.

To identify potential mitochondrial receptors for Myo19, we used tail proximity-dependent biotinylation (BioID) (Roux et al., 2012), since the tail region contains the mitochondrial targeting information and localizes to mitochondria (Quintero et al., 2009). We generated a stably transfected HeLa cell clone that expressed low levels of the fusion protein BirA*-Myo19 tail when incubated with sodium butyrate (Fig. 1A,B). Biotinylated proteins were affinity purified by magnetic streptavidin-beads and analysed by mass spectrometry (Table S1). Thereby, we identified peptides from Miro1 and Miro2 (Miro1/2), the mitochondrial receptors for microtubule-based transport. Immunoblotting experiments further demonstrated that Miro1/2, but not the outer mitochondrial membrane protein VDAC-1, were biotinylated specifically by BirA*-Myo19 tail. Miro1/2 were eluted after affinity purification from streptavidin exclusively upon expression of the BirA*-Myo19 tail construct (Fig. 1C,D). In accordance with our results, Miro-2 has recently been identified as a binding partner of Myo19 in two independent large-scale protein-protein interaction screens (Huttlin et al., 2015; Hein et al., 2015). Two additional outer mitochondrial membrane proteins were discovered in the Myo19-tail BioID screen, namely metaxin-3 and MAVS (Table S1). Subsequently, we focused on Miro1/2 proteins and tested whether Myo19 interacts with Miro1 directly (Fig. 2). We expressed and purified C-terminally His-tagged recombinant human Miro1 lacking its transmembrane domain (Miro1-ABC, aa 1-592; Klosowiak et al., 2016) (Fig. 2A). As a potential binding partner, we expressed and purified a FLAG-tagged C-terminal tail fragment (aa 898-970-FLAG) of human Myo19 that lacks the more N-terminally located lipid-binding motif (Fig. 2A). The purified Myo19 C-tail fragment still contained a major contaminant protein of ∼70 kDa from E. coli. Pull-down experiments were performed with Miro1 adsorbed to Talon beads. The Myo19 C-tail fragment was eluted specifically together with Miro1, but not from control beads lacking Miro1-ABC (Fig. 2B), indicating that this Myo19 tail fragment interacts directly with Miro1.

To test for potential involvement of Miro1/2 in the localization of Myo19 to mitochondria, we downregulated Miro1/2 by siRNA in HeLa cells. Notably, downregulation of Miro1/2 resulted in a concomitant downregulation of Myo19 as revealed by immunoblotting (Fig. 3A-D). This drop in Myo19 protein level was not due to a general reduction in mitochondrial mass, as the signal for the outer mitochondrial membrane protein VDAC-1 was not affected by downregulation of Miro1/2 (Fig. 3A,B). Transfection of cells with either pools or individual siRNAs directed against Miro1/2 led to time-dependent downregulation of Miro1/2, with clear reduction after 48 h and 72 h (∼85% reduction) (Fig. 3A,B). Notably, downregulation of Miro1/2 was accompanied by comparable time-dependent downregulation of Myo19 (∼70% after 72 h). The concomitant downregulation of Myo19 with Miro1/2 could be attenuated by transient overexpression of Miro2 resistant to the siRNA. The ability of Miro2 to rescue Myo19 expression depended on the nucleotide-state of the N-terminal GTPase domain. Introduction of a point mutation (T18N) that in analogy to other GTPases is predicted to induce a constitutively GDP-bound or nucleotide-free state prevented the rescue of Myo19 upon transient transfection (Fig. 3B,D). Conversely, downregulation of Myo19 with siRNA did not alter the expression of Miro1/2 (Fig. 3A,C). These results strongly argue that Miro1/2 stabilize Myo19, but that conversely, Myo19 does not influence the stability of Miro1/2. Indirect immunofluorescence labelling of Myo19 in cells treated with siRNA against Miro1/2 confirmed that the amounts of Myo19 protein associated with mitochondria depended on the expression of Miro1/2 (Fig. 3E). The mitochondrial signal of Myo19 was greatly reduced in cells with downregulated Miro1/2. To verify that Miro1/2 indeed regulate Myo19 protein stability, pulse-chase experiments were performed with Halo-tagged Myo19. Stable cell lines expressing Halo-tagged Myo19 were incubated with siRNA directed against Miro1/2 and the Halo-tag of Myo19 protein was pulse-labelled with the fluorophore TMR. Labelled Myo19 decayed faster when Miro1/2 were simultaneously downregulated (Fig. 3F). Interestingly, Myo19 protein turnover was slower when the Halo-tag was fused to the C-terminus (Myo19-Halo; Fig. 3G). This was the case irrespective of whether the cells were treated with non-targeting control siRNA or Miro1/2 siRNA. The stable expression of the Halo constructs reduced the levels of endogenous Myo19 so that the sum of exogenous and endogenous Myo19 amounted to the level of endogenous Myo19 in wild-type cells. The ratio of N-terminally-tagged Halo-Myo19 to endogenous Myo19 was 1:1, whereas the ratio of Myo19-Halo to endogenous Myo19 was 1:3. These data reveal that the positioning of the tag influences the life-time of Myo19 protein and that Miro1/2 increase the life-time of Myo19 protein.

Further evidence suggesting that release of Myo19 from mitochondria promotes its degradation was provided by the observation that stable expression of BirA*-Myo19 tail displaced endogenous Myo19 from mitochondria without a corresponding increase of Myo19 protein in the cytosol fraction (Fig. 4A). Analysis of cell homogenates confirmed that Myo19 is degraded upon overexpression of Myo19 tail (Fig. S2). Additionally, overexpression of BirA*-Myo19 tail led to a slight, although not significant, reduction of the amount of TRAK1 that was associated with purified mitochondria (Fig. 4). Next, we wondered whether adaptor proteins TRAK1 and TRAK2 could compete with Myo19 for binding to Miro1/2 and hence their overexpression would induce a downregulation of Myo19. Indeed, transient transfection of either TRAK1 or TRAK2 led to a downregulation of Myo19 according to the rate of transfection (Fig. 4C,D). Comparable results were obtained when C-terminal fragments of TRAK1 (aa 396-953) or TRAK2 (aa 324-913), which harbour the Miro-binding region, but not the KIF5- or dynactin-binding regions, were transiently overexpressed (Fig. 4C,D). In contrast, Miro1/2 levels were unaltered after transient overexpression of the TRAK constructs. These results indicate that Myo19 competes with TRAK proteins for binding to Miro1/2.

The simultaneous downregulation of Myo19 with Miro1/2 precluded a meaningful analysis of potential redistribution of Myo19 in the absence of Miro1/2. The proportion of background signal significantly increased in relation to the specific signal, so that no meaningful quantification of the remaining mitochondrial localization of Myo19 was possible (see Fig. 3E). Therefore, we devised an alternative strategy to test for in vivo recruitment of Myo19 by Miro1/2. We retargeted Miro1 to early endosomes by replacing its C-terminal transmembrane domain with 2xFYVE domains that mediate binding to phosphatidylinositol-3-phosphate (see schematic representation in Fig. S1). This Miro1/2 construct was indeed targeted to early endosomes as it colocalized with fluorescently labelled transferrin that was taken up by receptor-mediated endocytosis (Fig. S3A). To monitor retargeting of the Myo19 tail that directs Myo19 to mitochondria (Quintero et al., 2009; see also Fig. 4) and to delineate the mitochondrial and Miro1/2 binding site(s) in the Myo19 tail region more precisely, we analysed various fragments of the Myo19 tail region for Miro-dependent retargeting to early endosomes. We found that the tail region of Myo19 localized both to mitochondria with endogenous Miro1/2 and to early endosomes with retargeted Miro1 (Fig. 5A-C), indicating that Miro1/2 are sufficient for the specific recruitment of Myo19 tail in vivo. An N-terminal tail fragment (aa 824-897) encompassing a reported lipid-binding domain (Shneyer et al., 2016; Hawthorne et al., 2016) localized exclusively to mitochondria and was not redirected to early endosomes (Fig. 5A-C). In contrast, the C-terminal tail fragment (aa 898-970) that bound to Miro1 in vitro, colocalized with Miro1/2 that was retargeted to early endosomes (Fig. 5A-C). Part of this C-terminal tail fragment was also found in the cytosol, but it was not obviously colocalized with mitochondria (Fig. 5A-C). These results suggest that the tail region of Myo19 can be targeted to mitochondria by an N-terminal lipid-binding motif and a C-terminal Miro-binding domain. For recruitment of the Myo19 tail region, a fragment of Miro1/2 containing the N-terminal GTPase domain and the two ELM domains (aa 1-408) was sufficient (Fig. 5D-F); the C-terminal GTPase domain was dispensable. Next, we examined potential regulation of Myo19 recruitment by the nucleotide state of the N-terminal GTPase domain of Miro. Introduction of the point mutation T18N in the N-terminal GTPase domain abrogated recruitment of the Myo19 tail and C-terminal tail fragment to mitochondria (Fig. 5G-I). The correlation between localization of retargeted Miro1/2 and TOMM20 was barely affected by the different Myo19 tail constructs (Fig. S2B).

In mammalian cells, the subcellular distribution of mitochondria is regulated by microtubule-based motor proteins and their mitochondrial receptor Miro1/2 (Maeder et al., 2014; Mishra and Chan, 2014). Silencing of Miro1/2 or Kif5 protein expression has been shown to induce perinuclear accumulation of mitochondria (Tanaka et al., 1998; Liu et al., 2012). To analyse whether Myo19 contributes to mitochondrial distribution in cells, we either silenced or overexpressed Myo19 in HeLa cells. We specifically measured the perinuclear accumulation factor of mitochondria in cells, which we defined as the percentage of mitochondria within the perinuclear region, divided by the fraction of the cell area occupied by this region. This quantification confirmed that depletion of Miro1/2 or Kif5B with siRNAs causes significant perinuclear accumulation of mitochondria (Fig. 6). Interestingly, siRNA-mediated depletion of Myo19 expression also significantly shifted the localization of mitochondria to the perinuclear region (Fig. 6). Next, we addressed whether overexpression of Myo19 or different tail fragments would affect mitochondrial distribution. We found that overexpression of GFP-Myo19 phenocopied the effect of Myo19 downregulation (Fig. 7). This effect was independent of the actin-based motor capacity of Myo19, since the GFP-labelled tail region of Myo19, when overexpressed, induced an even more pronounced perinuclear accumulation of mitochondria (Fig. 7). Overexpression of the Myo19 C-tail fragment, which has been shown to interact with Miro, also induced perinuclear accumulation of mitochondria (Fig. 7). The N-tail lipid-binding fragment might still compete to some extent for binding of endogenous Myo19 to mitochondria, as its overexpression caused a less-pronounced, but still significant, perinuclear accumulation of mitochondria (Fig. 7). To exclude the possibility that perinuclear accumulation of mitochondria was due to oligomerization of the fused EGFP label, the experiments were repeated with full-length Myo19 and Myo19 tail, respectively fused with a Halo tag and labelled with TMR. These experiments confirmed that the perinuclear accumulation of mitochondria occurred independently of EGFP, and again showed that the tail region of Myo19 has a stronger effect on mitochondrial distribution than full-length Myo19 (Fig. 8). We also assessed whether overexpression of Myo19 or tail fragments thereof led to a change in cell size or shape. Such changes might impact the quantification of perinuclear clustering. We found that cells overexpressing full-length Myo19 covered a smaller area, had a shorter perimeter and increased circularity (Fig. S4). However, this effect seemed to be dependent on the motor domain of Myo19, since none of the Myo19 tail fragments caused similar changes when overexpressed (Fig. S4). In cells overexpressing EGFP-tail, for which the most pronounced perinuclear accumulation of mitochondria had been observed, the perinuclear accumulation factors were only weakly correlated to the cell size and shape as assessed by Pearson's correlation coefficient. The strongest correlation was found between a cell's perinuclear accumulation factor and its covered area or perimeter (R=0.46 and 0.43, respectively). This positive correlation confirmed rather than invalidated the observed increased perinuclear accumulation of mitochondria in the smaller Myo19-overexpressing cells, since increased perinuclear clustering in smaller cells would actually be detected less readily. In summary, both the downregulation of Myo19 and the overexpression of full-length Myo19 or Myo19 tail fragments that contain the Miro-binding site induced perinuclear clustering of mitochondria.

In this study, we identified Miro1/2 as mitochondrial receptors for Myo19 and showed that this interaction stabilizes Myo19, but not Miro1/2. We further found that both downregulation of Myo19 and overexpression of Myo19 or Miro-binding Myo19 fragments phenocopy the loss of Miro1/2 and the microtubule-based motor KIF5. KIF5-dependent transport of mitochondria towards the plus-end of microtubules and dynein-dependent transport to the opposite end are mediated by the protein TRAK1/2, which serves as an adaptor for Miro1/2 and both KIF5 and dynactin/dynein (Glater et al., 2006; van Spronsen et al., 2013). We further demonstrate that Myo19 interacts directly with Miro1/2 and competes with TRAK1/2 for binding. These results indicate that Miro1/2 coordinate actin- and microtubule-based mitochondrial movements. The perinuclear collapse of mitochondria observed upon both deletion and overexpression of Myo19 could be induced by an imbalance of Miro-mediated microtubule-dependent transport. Deletion of Myo19 might affect the equilibrium of kinesin- and dynein-dependent mitochondrial movements, whereas its overexpression might compete differentially with these two motors. Similarly to the deletion of Myo19, deletion of Miro1/2 caused perinuclear accumulation of mitochondria. The search for nearest neighbours of the Myo19 tail region using BioID did not reveal any known binding partners of Miro1/2. However, additional proteins were identified that are associated with the mitochondrial outer membrane, such as MAVS (mitochondrial anti-viral signalling protein) and metaxin-3. The connection of these proteins to Myo19 should be investigated further.

Intriguingly, downregulation of the Myo19 receptors Miro1/2, or overexpression of either TRAK1 or TRAK2 caused concomitant reduction of Myo19 protein. The binding of Myo19 to Miro1/2 on mitochondria stabilizes the protein and the loss of this interaction leads to Myo19 destabilization and degradation. Miro1/2 can be downregulated by the two Parkinson's disease genes PINK1 and Parkin, which participate in mitochondrial quality control (Chan et al., 2011; Wang et al., 2011). Additionally, the degradation of Miro2 is regulated by the PGAM5-KEAP1-Nrf2 complex. Miro2, but not Miro1, is a substrate for KEAP1–cullin-3 E3 ubiquitin ligase and the proteasome (O'Mealey et al., 2017). Following Miro1/2 degradation by either pathway, microtubule-based mitochondrial movement is halted (Wang et al., 2011; O'Mealey et al., 2017). Similarly, actin-based motility via Myo19 will be abrogated by loss of Miro1/2. In agreement with our work, López-Doménech et al. (2018) reported in an independent study that Miro1/2 stabilize Myo19 on mitochondria. A remarkable analogy represents vacuole inheritance in S. cerevisiae that relies on actin-based transport by the class V myosin Myo2. Unloading of the vacuole at its proper destination from Myo2 is controlled by the degradation of its vacuole adapter Vac17 (Yau et al., 2017).

Recently, it has been reported that a peptide encompassing residues 860-890 in the tail region of Myo19 serves as an outer mitochondrial membrane-binding motif (Shneyer et al., 2016; Hawthorne et al., 2016). This finding agrees well with our observation that an N-terminal tail fragment (aa 824-897) of Myo19 is able to localize to mitochondria. However, this lipid-binding motif alone appears not to provide sufficient specificity and functionality under physiological conditions. To establish specificity and fully functional association with mitochondria, Myo19 has to interact additionally with Miro. Although the overexpressed Myo19 Miro-binding motif appeared to be localized mainly in the cytosol, it induced perinuclear collapse of the mitochondria, indicating that it successfully competed with other proteins for binding to Miro. An increase of the Miro1/2 concentration resulted in the distinct recruitment of the Miro1/2 binding C-terminal tail fragment (aa 898-970) of Myo19 to mitochondria (data not shown). Miro1/2 were not only necessary, but also sufficient to recruit the Myo19 tail. When Miro1/2 were redirected to early endosomes, the Myo19 tail was also localized on early endosomes. Furthermore, this result demonstrates the predominance of the interaction with Miro1/2 over the lipid binding by Myo19. In a recent report it was suggested that the mitochondrial localization of Myo19 depends exclusively on the presence of Miro1/2 (López-Doménech et al., 2018).

We showed here that the N-terminal Rho-like GTPase domain of Miro1/2 is critically involved in the recruitment of Myo19. Currently, there is no conclusive evidence available as to whether this domain cycles between a GDP- and a GTP-bound state in vivo. Based on sequence comparisons, it appears likely that this domain in Miro1/2 has GTP bound constitutively, and is not cycling, since some residues important for hydrolysis are not conserved (Longenecker et al., 2003; Fransson et al., 2003; Li et al., 2002; Foster et al., 1996; Nobes et al., 1998). Marginal GTPase activity has been reported for this domain in Gem1p, the yeast analogue, with one GTP hydrolysed in 5-10 min (Koshiba et al., 2011). Phosphorylation of a serine residue was recently suggested to increase GTPase activity (Lee et al., 2016). Mutation in Miro1/2 of a threonine residue at the end of the P-loop that is conserved in nucleotide binding proteins to asparagine (T18N) abrogated the recruitment and stability of Myo19. Introduction of the analogous mutation (T19N) in RhoA produces a dominant-negative form. This RhoA mutant has an accelerated rate of GTP hydrolysis, but is mostly in a nucleotide-free state exhibiting a high affinity for guanine nucleotide exchange factors (Strassheim et al., 2000; Miyazaki et al., 2002). Genetic experiments in Drosophila with this Miro mutant suggested that it is essentially a recessive loss-of-function mutation (Babic et al., 2015). Residues mutated in dominant active constructs of RhoA are not conserved in Miro1. Therefore, it is not certain whether the corresponding mutations in Miro1 will induce similar alterations. Neuronal phenotypes observed for one such point mutation in Drosophila (dMiroA20V) were not compatible with a constitutively active mutation (Babic et al., 2015). It remains to be determined whether cycling between different nucleotide states of the N-terminal GTPase domain of Miro1/2 represents a physiological mechanism to regulate Myo19 recruitment. The protein VopE from the bacterial pathogen Vibrio cholerae has been proposed to act as a GTPase-activating protein (GAP) for the N-terminal GTPase domain of Miro1/2 and the mammalian protein RAP1GDS1 as a guanine-nucleotide exchange factor (GEF) (Suzuki et al., 2014; Ding et al., 2016).

Myo19 was shown to regulate the subcellular distribution of mitochondria. Both downregulation and overexpression of Myo19 induced a perinuclear accumulation of mitochondria. A similar perinuclear accumulation of mitochondria was observed in the absence of Miro1 which negatively affected cytoplasmic energy distribution, cell migration and cell adhesion (Schuler et al., 2017). During cell division, Myo19 regulates the allocation of mitochondria to daughter cells. Loss of Myo19 causes asymmetric partitioning of the mitochondria to one or both spindle poles in mitosis and stochastic failure of cytokinesis (Rohn et al., 2014). A similar phenotype in mitochondria segregation during mitosis was recently reported for the loss of Miro1/2 (López-Doménech et al., 2018). In the presence of Myo19, mitochondria do not localize to spindle microtubules, but instead are localized in the periphery (Lee et al., 2007; Chung et al., 2016). Forced attachment of dynein or kinesin motors to the mitochondrial surface during mitosis generates a phenotype comparable to that observed after downregulation of Myo19 (Chung et al., 2016; Rohn et al., 2014). These findings further hint at a role for Myo19 in the regulation of microtubule-based transport of mitochondria. Therefore, it will be interesting to elucidate further the interplay between Miro1/2 and Myo19 in the control of microtubule- and actin-based transport of mitochondria.

The antibodies used were as follows: Myo19 (EPR12551-13/ab174286, Abcam, WB: 1:1000; HPA021415, Sigma-Aldrich, WB: 0.18 μg/ml); Miro-1/2 (NBP1-59021, Novus Biologicals, WB: 0.25 μg/ml); KIF5B (ab167429; Abcam, WB: 0.14 μg/ml); TRAK1 (PA5-44180, Thermo Fisher, WB: 0.5 μg/ml); VDAC1 (20B12AF2/ab14734, Abcam, WB: 1 μg/ml); myc (ab9106, Abcam, WB: 0.04 μg/ml); GAPDH (TA802519; OriGene, WB: 0.5 μg/ml); β-actin (AC-15/A1978, Sigma-Aldrich, WB: 1 μg/ml); mCherry 1C51 (ab125096 Abcam, WB: 0.5 μg/ml); Goat Anti-Rabbit IgG (H+L)-HRP (111-035-003, Jackson ImmunoResearch, WB: 0.16 μg/ml); Goat Anti-Mouse IgG (H+L)-HRP (115-035-003, Jackson ImmunoResearch, WB: 0.16 μg/ml); Streptavidin-HRP (016-030-084, Dianova, WB: 0.2 μg/ml);

Transfection reagents used in this study were as follows: Lipofectamine LTX & PLUS Reagent (15338100, Thermo Fisher Scientific); Lipofectamine RNAiMAX Reagent (13778150, Thermo Fisher Scientific); PolyFect Transfection Reagent (301105, Qiagen); Nanofectin Kit (Q051-005, PAA Laboratories); polyethylenimine (PEI 25K, 23966-1, Polysciences); and ScreenFect A (299-73203, Wako Laboratory Chemicals).

The following siRNAs were used (5′-3′): siRNA Myo19 (GCAAAUGACUGGAGCCGCA, UAACAACAGCAGUCGCUUU, GGGAGGUCCUGCUGUACAA, CGCCCGAGCUAAUGAGAGA (L-017137-01-0005, Dharmacon, GE Healthcare); siRNA non-targeting (UGGUUUACAUGUCGACUAA, UGGUUUACAUGUUGUGUGA, UGGUUUACAUGUUUUCUGA, UGGUUUACAUGUUUUCCUA (D-001810-10, Dharmacon, GE Healthcare); siRNA Miro-1 (#-09 UGUGGAGUGUUCAGCGAAA, #-10 GCAAUUAGCAGAGGCGUUA, #-11 CCAGAGAGGGAGACACGAA, #-12 GCUUAAUCGUAGCUGCAAA (J-010365-XX, Dharmacon, GE Healthcare); siRNA Miro-2 (#-09 GCGUGGAGUGUUCGGCCAA, #-10 CCUCAAGUUUGGAGCCGUU, #-11 GAGGUUGGGUUCCUGAUUA, #-12 AGGAGAUCCACAAGGCAAA, J-008340-XX, Dharmacon, GE Healthcare); SMARTpool Kif5B siRNA (Dharmacon: ON-TARGETplus KIF5B siRNA L-008867-00-0005).

Human Myo19 cDNA (AK304073/FLJ61052/TRACH 3006685) was obtained from the National Institute of Technology and Evaluation (NITE) Japan. It was subcloned into pEGFP-C1 by PCR amplification of a 5′- and a 3′-fragment, introducing a BglII and an XbaI site, respectively. The plasmid pEGFP-Myo19 tail codes for a fusion protein of EGFP with amino acids 824-970 of Myo19, pEGFP-Myo19 N-tail with amino acid residues 824-897 of Myo19 and pEGFP-Myo19 C-tail with amino acid residues 898-970 of Myo19, respectively. The myc-BirA R118G-MCS (BioID) plasmid was obtained from Addgene (plasmid #35700; Roux et al., 2012). The Myo19 tail cDNA coding for residues 824-970 was subcloned into the XhoI and HindIII restriction sites. To construct pMyo19-Halo, the Halo-tag flanked by a BshTI restriction site and a stop codon followed by a NotI restriction site was amplified by PCR using plasmid pFN21AA1042 (Kazusa DNA Research Institute) as a template. The Halo-tag cDNA was used to replace the cDNA for EGFP in pMyo19-EGFP-N1. pHalo-Myo19 was constructed by replacing EGFP in the plasmid pEGFP-C1-Myo19 with amplified Halo-tag using the restriction sites BshTI and BglII. Plasmid pMyo19tail-Halo-N1 was constructed by amplifying Myo19tail from pEGFP-C1-Myo19tail by PCR, introducing BglII and EcoRI restriction sites. This PCR product was then used to replace the Myo19 coding sequence in pMyo19-Halo-N1. Plasmid pHalo-C1-Myo19tail was constructed by amplifying Halo from plasmid pFN21AA1042 by PCR, introducing AgeI and BspEI restriction sites. Using these sites, this PCR product was then used to replace the EGFP coding sequence in pEGFP-C1-Myo19tail-avi-FLAG. To express GST-Myo19 C-tail (aa 898-970)-FLAG, the sequence coding for C-tail-FLAG was amplified by PCR and subcloned into pGEX-4T-1 using EcoRI/SalI restriction sites. pRK5-Myc-Miro-1 and pRK5-myc-Miro-2 plasmids were gifts from Pontus Aspenström, Department of Microbiology, Tumor and Cell Biology, Karolinska Institutet, Stockholm, Sweden (Fransson et al., 2003). The plasmid pRK5–myc-Miro2 T18N was obtained from Addgene. The plasmid mCherry-Miro-1 was constructed by inserting Miro-1 into the XhoI/SacII sites of pmCherry-C1. The point mutation to create plasmid mCherry-Miro-1 N18 was introduced by QuikChange mutagenesis. The GFP-2xFYVE^Hrs plasmid was a gift from Marco Falasca, Centre for Cell Biology and Cutaneous Reserch, London, UK (Maffucci et al., 2003). The plasmid mCherry-Miro-1-ΔTM-2xFYVE has the sequences coding for amino acid residues 592-618 of Miro-1 replaced by the sequences coding for 2xFYVE domains. The plasmid coding for the fusion protein pmCherry-Miro-1-aa1-408-2xFYVE was constructed by standard molecular biology techniques. The pET28a(+) plasmid encoding hMiro1-ABC(aa1-592)-6xHis was a gift from Sarah E. Rice, Department of Cell Biology and Molecular Biology, Chicago, USA (Klosowiak et al., 2016). Plasmid pFN21AA1042 coding for Halo-tagged human TRAK1 was obtained from Kazusa DNA Research Institute. The cDNA sequence coding for TRAK1 was subcloned into pmCherry-C1 using XhoI and SalI restriction sites. The plasmid coding for mCherry-TRAK1 amino acid residues 396-953 (TRAK1-C) was constructed using PCR. Plasmid pCMV-4a rat TRAK2-FLAG was a generous gift from A. Stephenson (University College London, UK). The TRAK2 cDNA encoding either full-length TRAK2 or residues 324-914 (TRAK2-C) were subcloned into pmCherry-C1 vector using XhoI and EcoRI restriction sites. The plasmid encoding the mitochondrial targeting sequence from the subunit VIII of the human cytochrome c oxidase fused to RFP was purchased from BD Pharmingen and the sequence coding for RFP was exchanged for EGFP and mCherry, respectively. Plasmid mTagBFP2-TOMM20-N-10 was from Addgene (plasmid #55328, deposited by Michael Davidson; Subach et al., 2011). All fragments amplified by PCR were sequenced.

HeLa and HEK 293T cells (ATCC) were cultivated in DMEM supplemented with 10% (v/v) FCS, 100 U/ml penicillin and 100 μg/ml streptomycin (complete DMEM) at 37°C, 95% humidity and 5% CO₂. For transfection of HeLa cells, 20,000 cells were seeded onto glass coverslips in 24-well plates. Cells were transfected 1 day after seeding as described in the manual of the manufacturer. Briefly, 0.5 µg of the plasmid DNA was mixed with 1 μl Lipofectamine LTX and 0.5 µl PLUS Reagent in 50 μl DMEM, vortexed, incubated at room temperature (RT) for 5 min and added drop-wise to the cells. For cotransfection of two plasmids, the amount of plasmid DNA was reduced to 0.25 µg for each plasmid. After 4 h of incubation the DNA-lipid complexes were removed, cells were washed with PBS and fresh complete DMEM was added. Cells were fixed 24 h after transfection.

For live cell imaging, ∼40,000 HeLa cells were seeded onto a 35 mm µ-dish (ibidi) in 1 ml complete DMEM 1 day before transfection. For transfection, 2 µg of the plasmid DNA in 50 μl diluent (150 mM NaCl) was mixed with 6.4 μl Nanofectin in 50 μl diluent, and vortexed. The mixture was incubated at RT for 30 min and added to the µ-dish drop wise. The transfection complexes were removed after 4 h of incubation, and fresh complete DMEM was added to the µ-dish. After 24 h of incubation, the expression of recombinant proteins was analysed.

To generate the stable HeLa myc-BirA* Myo19 tail, Halo-Myo19 and Myo19-Halo cell lines, 100,000 cells per well were seeded in a 6-well plate. The following day, 2.5 µg of the plasmid DNA with 5 μl Lipofectamine LTX and 2.5 µl PLUS Reagent in 250 μl DMEM was used for transfection. After 48 h, complete DMEM was supplemented with 600 µg/ml G418 and 60%, 30% and 10% of cells, respectively, were replated in 150-mm-diameter cell culture dishes. Cells still viable after 2 weeks were considered stably transfected. Single colonies of cells were isolated with a metal ring, trypsinized and transferred to 6-well plates. Cells were processed for western blot analysis and further expansion.

For knockdown studies of Miro-1 and Miro-2, 100,000 HeLa cells per well were seeded in 6-well plates. The following day, cells in one well were transfected with 25 pmol of each siRNA or each pool of 4 siRNAs that was mixed with 7.5 μl Lipofectamine RNAiMAX in 183 μl DMEM. Cells were harvested for western blot analysis 48 h and 72 h after transfection. Transient transfections of HEK cells were performed in 6-well plates by seeding 150,000 cells per well. The following day 4 μg of plasmid DNA in 100 μl DMEM were mixed with 12 μg PEI in 100 μl DMEM, vortexed, incubated for 30 min at RT and then added drop-wise to the cells. Cells were harvested 24 h after transfection and processed for immunoblot analysis.

BioID was carried out essentially as described in Roux et al. (2013) with some modifications. Cells were incubated for 24 h in complete DMEM supplemented with 50 µM biotin before lysis and affinity purification of biotinylated proteins. To enhance expression of the construct, 10 mM sodium butyrate was added to the medium 15 h before harvesting of the cells.

Approximately 2×10⁸ stably transfected myc-BirA* Myo19 tail HeLa cells (10 confluent 150-mm-diameter cell culture dishes) were washed three times with cold PBS and collected by scraping. They were resuspended in 20 ml lysis buffer [50 mM Tris-HCl, pH 7.4, 500 mM NaCl, 0.2% (w/v) SDS, 0.1 mg/ml Pefabloc, 0.01 mg/ml Leupeptin and 0.02 U/ml Aprotinin] before 2 ml of 20% (v/v) Triton X-100 were added. The subsequent steps were performed at 4°C. Lysates were sonicated with two 30 pulses using an amplitude of 100% and a proportional sonication period of 0.6 (UP 100H sonicator, Hielscher Ultrasound Technology). Prior to the second sonication cycle 20 ml of 50 mM Tris-HCl, pH 7.4, were added. Lysates were clarified by centrifugation at 16,600 g for 10 min. Supernatants were incubated with 8 mg of equilibrated streptavidin magnetic beads (Pierce 88817, Thermo Fisher Scientific) overnight. Beads were collected and washed with wash buffer 1 [2% (w/v) SDS in ddH₂O] for 8 min at RT. This step was repeated with wash buffer 2 [0.1% (w/v) deoxycholic acid, 1% (w/v) Triton X-100, 1 mM EDTA, 500 mM NaCl and 50 mM Hepes, pH 7.5], wash buffer 3 [0.5% (w/v) deoxycholic acid, 0.5% (w/v) NP-40, 1 mM EDTA, 250 mM LiCl and 10 mM Tris-HCl, pH 7.4] and wash buffer 4 (50 mM Tris-HCl, pH 7.4), respectively.

Biotinylated proteins were eluted from the beads at 98°C for 8 min with 200 µl buffer containing 40 mM Tris-HCl, pH 6.8, 1 mM EDTA, 8% (w/v) sucrose, 3% (w/v) SDS, 2% (w/v) 2-mercaptoethanol, 0.004% (w/v) bromophenol blue and 820 µM biotin. Aliquots of the eluates were analysed by immunoblotting (20 µl) and by Coomassie Blue staining following SDS-PAGE (40 µl). Isolated bands were subjected to mass spectrometry (University Hospital Cologne, Centre of Molecular Medicine (ZMM), Central Bioanalytics (ZBA)).

Miro1 (aa 1-592, C-terminal 6xHis-tag) (Miro1-ABC) was expressed in E. coli Rosetta 2 (DE3) cells and purified by metal ion affinity chromatography with TALON beads (Clontech Laboratories, Inc.). Cells were cultured in TPM medium (20 g/l tryptone, 15 g/l yeast extract, 8 g/l NaCl, 2 g/l Na₂HPO₄, 1 g/l KH₂PO₄) containing 50 µg/ml kanamycin and 34 µg/ml chloramphenicol. Protein expression was induced overnight at 18°C with 125 µM IPTG when cell density had reached an OD₆₀₀ of 0.9. Cells were harvested by centrifugation (4400 g, 15 min, 4°C), washed with cold PBS (137 mM NaCl, 2.7 mM KCl, 10.2 mM Na₂HPO₄, 1.8 mM KH₂PO₄, pH 7.0) and stored at −80°C. For protein purification, cell pellets were resuspended in lysis buffer (25 mM Hepes, pH 7.4, 300 mM NaCl, 0.5 mM TCEP, 8 mM imidazole, 2 mM MgCl₂, 5% sucrose, 0.02% Tween-20, 0.1 mg/ml Pefabloc^®, 0.01 mg/ml Leupetin, 0.02 U/ml Aprotinin, 2 mM ATP) and cells were lysed by sonication for 4×2 min with 2 min intervals using an amplitude of 100% and a proportional sonication period of 0.6 (UP 100H sonicator, Hielscher Ultrasound Technology). The lysate was clarified by ultracentrifugation (139,000 g, 45 min, 4°C) and batch adsorbed to pre-equilibrated TALON beads for 1 h at 4°C. The beads were collected and washed three times with washing buffer (25 mM Hepes, pH 7.4, 300 mM NaCl, 0.5 mM TCEP, 12 mM imidazole, 2 mM MgCl₂, 5% sucrose, 0.02% Tween-20) before they were transferred to a gravity-flow column. Bound protein was eluted with elution buffer (25 mM Hepes, pH 7.4, 300 mM NaCl, 0.5 mM TCEP, 300 mM imidazole, 2 mM MgCl₂, 5% sucrose, 0.02% Tween-20), pooled and stored at −20°C.

Myo19 C-tail (aa 898-970) fused to a N-terminal GST-tag and a C-terminal FLAG-tag was expressed in E. coli Rosetta 2 (DE3) cells and purified by glutathione affinity chromatography followed by FLAG-antibody affinity chromatography. The procedure for cell growth, induction of protein expression and harvest of cells was as described above for Miro1-ABC protein purification. Cell pellets were stored overnight at −80°C. They were resuspended in lysis buffer (20 mM Hepes, pH 7.4, 100 mM NaCl, 50 mM KCl, 2 mM MgCl₂, 10% glycerol, 1 mM 2-mercaptoethanol, 0.1 mg/ml Pefabloc^®, 0.01 mg/ml Leupetin, 0.02 U/ml Aprotinin, 2 mM ATP) and lysed by sonication. Following ultracentrifugation (139,000 g, 45 min, 4°C), the supernatant was incubated with pre-equilibrated glutathione Sepharose beads. The beads were collected and washed three times with washing buffer (20 mM Hepes, pH 7.4, 100 mM NaCl, 50 mM KCl, 2 mM MgCl₂, 1 mM 2-mercaptoethanol). The protein bound to the resin was eluted by cleavage of the GST-moiety by the addition of 40 U of thrombin overnight in washing buffer. Beads were centrifuged at 700 g for 5 min at 4°C and washed twice. The supernatants were combined and passed several times over FLAG-antibody agarose (ANTI-FLAG^® M2 Affinity Gel, Sigma-Aldrich). The column resin was washed with washing buffer, followed by two alternating high salt (20 mM Hepes, pH 7.4, 450 mM NaCl, 50 mM KCl, 2 mM MgCl₂, 1 mM 2-mercaptoethanol) and low-salt (20 mM Hepes, pH 7.4, 50 mM KCl, 2 mM MgCl₂, 1 mM 2-mercaptoethanol) washes, followed by washing buffer. Finally, Myo19 C-tail was eluted with 0.25 mg/ml FLAG^®-peptide (Sigma-Aldrich) in washing buffer.

Purified hMiro1-ABC was loaded with GTP by adding 2 mM GTP and 6 mM EDTA for 10 min at RT followed by the addition of 15 mM MgCl₂. Next, the protein was diluted with assay buffer (25 mM Hepes, pH 7.4, 300 mM NaCl, 0.5 mM TCEP, 2 mM MgCl₂, 5% sucrose, 0.02% Tween-20) to a final concentration of 10 mM imidazole and adsorbed to pre-equilibrated TALON beads at 4°C. Beads were washed three times with washing buffer (25 mM Hepes, pH 7.4, 300 mM NaCl, 0.5 mM TCEP, 12 mM imidazole, 2 mM MgCl₂, 5% sucrose, 0.02% Tween-20) and transferred to a gravity-flow column. Myo 19 C-tail was passed repeatedly over the column before it was washed three times with washing buffer. Adsorbed Miro1-ABC was eluted with an excess of imidazole in the buffer (25 mM Hepes, pH 7.4, 300 mM NaCl, 0.5 mM TCEP, 400 mM imidazole, 2 mM MgCl₂, 5% sucrose, 0.02% Tween-20). Eluted proteins were analysed by SDS-PAGE and immunoblotting.

Cells grown in 6-well plates were washed three times with cold PBS and gently scraped off. They were centrifuged at 600 g at 4°C for 5 min and resuspended in 100 µl lysis buffer [50 mM Tris-HCl pH 7.4, 2 mM MgCl₂, 100 mM NaCl, 1% (v/v) NP-40, 10% (v/v) glycerol, 1 mM DTT, 0.1 mg/ml Pefabloc, 0.01 mg/ml Leupeptin and 0.02 U/ml Aprotinin]. Protein concentrations were determined by Bradford assay using BSA as a standard. Equal amounts of protein were separated by SDS-PAGE and further processed for immunoblotting.

For pulse-chase experiments, stable Halo-Myo19 or Myo19-Halo HeLa cells were transfected with siRNA for Miro1/2. 24 h later, the cells were pulse labelled with 2 µM HaloTag^® TMR ligand (Promega) for 30 min at 37°C, 97% humidity and 5% CO₂. Free ligand was removed by several washing steps and a 30 min incubation with complete DMEM supplemented with 600 µg/ml G418. Next, the cells were incubated with 5 µM HaloTag^® Biotin ligand, to prevent further labelling of the Halo-tag with any remaining fluorescent ligand. Cells were harvested at 0, 4, 8 and 12 h time points. Proteins were separated by SDS-PAGE, electrophoretically transferred to a PVDF membrane and analysed for TMR fluorescence followed by immunoblotting. Fluorescent bands were quantified using ImageJ.

Prior to mitochondria purification, HeLa wt and HeLa myc-BirA* Myo19 tail cells were treated overnight with 10 mM sodium butyrate. Mitochondria were purified with Qproteome Mitochondria Isolation Kit (QIAGEN) according to the manufacturer's protocol for high-purity mitochondria.

To label mitochondria, cells were incubated with 75 nM MitoTracker Red CMXRos (M7512, Thermo Fisher Scientific) in complete DMEM at 37°C, 95% humidity and 5% CO₂ for 10 min. Immediately afterwards, cells were washed three times with warm PBS and fixed with 4% (w/v) paraformaldehyde (PFA) in PBS at 37°C for 25 min. Free aldehydes were quenched with 0.1 M glycine in PBS at RT for 10 min. Subsequently, cells were washed three times with PBS and mounted with Mowiol [3.4 mM Mowiol 4-88, 105 mM Tris-HCl, pH 8.5, 18.4% (v/v) Glycerin and 223 mM DABCO]. To stain cells for endogenous Myo19, they were permeabilized at RT for 15 min with 0.1% (v/v) Triton X-100 in PBS, followed by washing with PBS for 2×5 min. Unspecific binding sites were blocked by incubation with blocking buffer [5% (v/v) normal goat serum in PBS] at RT for 10 min. The primary antibody (HPA021415, Sigma-Aldrich) was diluted 1:100 (1.8 μg/ml) in blocking buffer. Coverslips were placed on a drop of antibody solution in a humid chamber and incubated overnight at 4°C. Thereafter, cells were washed 3× for 5 min with PBS at RT and incubated in the dark with secondary antibody (111-545-003, Jackson ImmunoResearch) diluted 1:500 (3 μg/ml) in blocking solution for no longer than 1 h at RT. Subsequently, cells were washed three times with PBS and mounted with Mowiol [3.4 mM Mowiol 4-88, 105 mM Tris-HCl, pH 8.5, 18.4% (v/v) Glycerin, 223 mM DABCO].

HeLa cells grown on coverslips were serum-starved for 30 min in MEM medium at 37°C, 5% CO₂ and 95% humidity. AF 488-conjugated human transferrin (30 µg/ml) (ThermoFisher) in MEM medium was added for 1 h at 37°C. Cells were washed with ice cold 1× PBS and subsequently fixed with 4% PFA for 20 min at RT.

HeLa cells were fixed 24 h after transfection and analysed by spinning disk microscopy (UltraVIEW VoX, PerkinElmer; Nikon Eclipse Ti). Images were acquired with a 60×/1.49 NA oil objective and an EMCCD camera using Volocity software. Z-stacks covering the entire depth of cells with intervals of 0.5 μm were acquired. To quantify retargeting of tail constructs to early endosomes, maximum intensity projections were obtained using ImageJ. To separate mitochondria from background a ‘top hat’ filter was applied. Pearson's correlation coefficient was used to express the intensity correlation of co-localizing objects. Analysis was restricted to custom ROI of each cell, with a thresholding factor of 2. The result is +1 for perfect correlation, 0 for no correlation and −1 for perfect anti-correlation.

To assess mitochondria distribution, 20,000 HeLa cells per well were plated on glass coverslips in 24-well plates. The next day, cells were transfected with appropriate plasmids using the transfection reagent Lipofectamine™ LTX and PLUS™ reagent (Invitrogen) according to the manufacturer's instructions. 24 h after transfection, the cells were processed further: Cells that had been transfected with Halo constructs were labelled with 5 µM HaloTag^® TMR ligand (Promega) in full medium for 30 min at 37°C, followed by several washing steps with PBS and full medium before fixation. Cells were washed with PBS and fixed with 4% PFA, stained for DNA with DAPI and, if appropriate, for F-actin with fluorophore-labelled phalloidin, mounted on microscope slides and left to harden at 4°C overnight. For experiments involving protein depletion, 100,000 HeLa cells per well were seeded in 6-well plates. 24 h after seeding, the cells were transfected with appropriate siRNAs using the transfection reagent Lipofectamine™ RNAiMAX according to the manufacturer's instructions. The next day, the cells were replated into 24 well plates and transfected, fixed and stained as described above. Imaging was performed with a 63× N.A. 1.4 Plan-Apochromat oil immersion objective on a Zeiss LSM 510 confocal laser scanning microscope equipped with Argon and Helium-Neon lasers and three 12-bit R/FI detectors. Z-stacks of mitochondria and, if appropriate, other labelled structures were recorded at a plane interval of 0.5 µm; upper and lower boundaries of the stacks were set manually for each field of view. Bright-field images were recorded with 12-bit R/FI detectors along with the z-stacks. DAPI fluorescence was excited with a mercury vapour short-arc HBO 50 lamp and detected with an 8-bit AxioCam MRm camera.

Image processing and analysis were performed in ImageJ (NIH, USA) with customized macros. Briefly, z-stacks of fluorescence channels were converted to sum projections, and sum projections of the mitochondria were top-hat filtered and background corrected. Size and orientation of the DAPI images were adjusted to the confocal images. For each cell, the outline was determined manually based on the bright-field image. Shape descriptors routinely available in ImageJ were measured for the cell outline. The nucleus outline was determined via automated thresholding of the DAPI image and was dilated to determine the boundary of the perinuclear region, which was defined as a 4 µm wide ring around the nucleus. When the dilated nucleus outline exceeded the cell outline, it was cut accordingly. For each of the cell's regions (in/above/below nucleus, perinuclear, peripheric), the proportion of mitochondria signal was measured as well as the fraction of the total cell area occupied by this region. The perinuclear mitochondria accumulation factor was defined as the proportion of mitochondrial signal within the perinuclear region, divided by the fraction of the total cell area occupied by the perinuclear region.

Data are displayed as box plots with the box ranging from 25th to 75th percentiles, whiskers ranging from 10th to 90th percentiles, the median marked as a line, and the mean marked as a square. Data were analysed by Kruskal-Wallis ANOVA to locate statistically significant differences in sets of more than two samples followed by a Mann-Whitney U-test for pairwise comparisons; the significance levels were set to *P≤0.05, **P≤0.01 and ***P≤0.001.

We are grateful to P. Aspenström, M. Falasca, A. Stephenson and S. E. Rice for the gift of plasmids, M. Horsthemke for help with artwork and to P. Hanley for valuable comments on the manuscript. A preliminary account of this work was presented at the 2016 ASCB Annual Meeting.