Dual role of Miro protein clusters in mitochondrial cristae organisation and ER-Mitochondria Contact Sites

Mitochondrial Rho (Miro) GTPases localize to the outer mitochondrial membrane and are essential machinery for the regulated trafficking of mitochondria to defined subcellular locations. However, their sub-mitochondrial localization and relationship with other critical mitochondrial complexes remains poorly understood. Here, using super-resolution fluorescence microscopy, we report that Miro proteins form nanometer-sized clusters along the mitochondrial outer membrane in association with the Mitochondrial Contact Site and Cristae Organizing System (MICOS). Using knockout mouse embryonic fibroblasts (MEF) we show that Miro1 and Miro2 are required for normal mitochondrial cristae architecture and endoplasmic reticulum-mitochondria contacts sites (ERMCS). Further, we show that Miro couples MICOS to TRAK motor protein adaptors to ensure the concerted transport of the two mitochondrial membranes and the correct distribution of cristae on the mitochondrial membrane. The Miro nanoscale organization, association with MICOS complex and regulation of ERMCS reveal new levels of control of the Miro GTPases on mitochondrial functionality.


Introduction
Mitochondria are the powerhouse of cells, generating ATP to drive key cellular functions including ion pumping, intracellular trafficking and cellular signalling cascades 1-7 . The mitochondrial population are trafficked to where they are needed to meet local energy and Ca 2+ buffering demands, by motor and adaptor proteins found on their outer surface that tether these organelles to the cytoskeleton [8][9][10][11] . Miro proteins form complexes with the TRAK adaptors and dynein/kinesin motors to regulate the microtubule-dependent transport of mitochondria 7,12 . Recently, an actindependent transport of mitochondria has also been linked to Miro regulation through the recruitment and stabilization of the mitochondrial myosin 19 (Myo19) to the outer mitochondrial membrane 13,14 . In yeast Miro exists as a single orthologue, Gem1, important for correct mitochondrial inheritance and cellular viability [15][16][17] . In mammals there are two Miro family members, Miro1 and Miro2, that share ~60% sequence identity, comprising two GTPase domains flanking two EF-hand Ca 2+ -binding domains and a C-terminal transmembrane domain that targets them to the Outer Mitochondrial Membrane (OMM) 12, 18 . Although their role in mitochondrial transport is well established, far less is known about their interactions with other key protein complexes located at the OMM or the Inner Mitochondrial Membrane (IMM).

Loss of Miro1 and Miro2 alters mitochondrial ultrastructure
We recently showed that loss of both Miro proteins in MEFs leads to altered mitochondrial distribution and morphology 13 . To further explore the effects of Miro loss on mitochondrial morphology and structure we performed Structured Illumination Microscopy (SIM), a superresolution imaging procedure that provides an optical resolution almost twice the diffraction limit 43 . Using SIM, wild type (WT) cells expressing a mitochondrial matrix targeted GFP (mtRo GFP ) showed a predominance of thin and long mitochondria, with individual mitochondria having continuous GFP staining with occasional cells presenting short and round mitochondria ( Fig. 1A and B). In contrast, about 70% of Miro1/2 double knockout (DKO) cells presented a discontinuous and often hollow matrix-targeted GFP signal ( Fig. 1A and B) which correlated with a predominance of shorter mitochondria with enlarged and more rounded mitochondrial segments.
The discontinuous mitochondrial matrix suggests a role for Miro proteins in maintaining the architecture of the inner mitochondrial membranes. Importantly, re-expression of either Myc Miro1 or Myc Miro2 in DKO cells rescued mitochondrial matrix continuity (Fig. 1B, Supplementary Fig.   1).
To further explore the effects of depleting Miro proteins on mitochondrial structure and morphology, we carried out ultrastructural analysis of MEF cells using Transmission Electron Microscopy (TEM). In WT MEFs, the majority of mitochondria showed normal cristae structure and a homogeneous cristae distribution throughout the mitochondrial segments (Fig. 1C, D and E). In contrast, the majority of DKO cells presented an altered cristae architecture with frequent vesiculated mitochondrial matrix ( Fig. 1C and D). Importantly, DKO cells showed a non-uniform arrangement of the mitochondrial cristae, with some mitochondrial regions having the normal density of cristae alternated with regions that appeared enlarged and devoid of cristae (Fig. 1E).
Together, these data indicate that Miro proteins play a role in regulating mitochondrial morphology and in the maintenance of the internal structure of the mitochondrial membranes and mitochondrial cristae architecture.
To test whether these changes in mitochondrial structure could be a consequence of reduced levels of protein components known to regulate cristae structure we carried out western blot analysis of lysates from three different WT and DKO MEF lines. We did not observe any significant change in the levels of several MICOS components tested (Mic19/CHCHD3, Mic60/Mitofilin and Mic25/ApooL). Outer mitochondrial membrane protein Sam50, known to closely associate to the MICOS complex to form the MIB complex and bridge the IMM and the OMM, also did not show any significant changes in DKO cells. In addition, we observed no changes in the levels of proteins related to Miro transport function (TRAK1 or TRAK2) or mitochondrial components like ATP5 or Tom20 (Fig. 1F). In contrast, we report a striking 4-fold increase in the levels of Inositol 1,4,5trisphosphate (IP3) receptor (IP3R), a known regulator of the contact sites between ER and mitochondria that regulates Ca 2+ communication between the two organelles ( Fig. 1F). Other components of ERMCS that act in coordination with IP3R like GRP75 and the OMM channel VDAC1 were also found to be moderately upregulated, although not to a statistically significant level (Fig. 1F). IP3Rs located at the ER are one of the main Ca 2+ -release channels 44 and upon activation by Inositol   1,4,5-trisphosphate (IP3) can transfer Ca 2+ to the mitochondria through a IP3R-VDAC complex 45 . In mammalian cells, IP3R acts as an adapter present on the ER that forms a complex with GRP75 and VDAC to maintain ERMCS 46 . At steady state, IP3R levels are tightly regulated and alteration of IP3R expression has been implicated with changes in ER morphology and ER Ca 2+ release 47,48 .

Loss of Miro1 and Miro2 alters ER/Mitochondrial communication
To study whether Miro accomplishes a role in regulating the connectivity of the ER membranes we performed a fluorescent recovery after photobleaching (FRAP) assay in our MEF cell lines.
FRAP analysis of ER-luminal targeted DsRed ( DsRed ER) showed no significant difference in fluorescence recovery between WT and DKO cells (WTM ~ 87 ± 7.5% against DKOM ~86 ± 8.2 %), with just a delay in the initial recovery time ( Fig. 2A . 2C and D), which was specific to the loss of Miro as re-expression of either Myc Miro1 or Myc Miro2 in DKO cells rescued the amount of overlap between ER and mitochondria ( Fig. 2C and D). This was further confirmed using TEM by transfecting an ER-targeted HRP construct (KDEL-HRP) 49 to enhance the contrast of ER structures and allow the identification and quantification of ERMCS (defined by proximity of ER and mitochondria within 35 nm, Fig. 2E). DKO cells showed a decreased number of contacts between the two organelles ( Fig. 2E and F; WT: 0.48 ± 0.19 and DKO: 0.36 ± 0.13; ER/mitochondria contacts; mean ± SD), confirming that Miro proteins accomplish a role in regulating ER and mitochondria association. To test if the altered ERMCS affects the communication between both the organelles, we measured mitochondrial Ca 2+ uptake upon ATP induced Ca 2+ release from ER stores (Fig. 2G). WT cells required 9.7s (9.73±0.75s, mean ± SEM) to reach maximum amplitude (F/FMin = 1.44±0.03 fold mean ± SEM) while DKO cell showed a significantly delayed uptake (15.96±1.27s, mean ± SEM) and reduced amplitude (F/FMin = 1.26±0.04, mean ± SEM) ( Fig. 2G and H). This suggests that the ER/mitochondrial handling of Ca 2+ is severely affected as a consequence of a decrease in ER/mitochondrial contact sites in DKO cells. We also observed that upon treatment with ATP, there was a significantly larger loss of Ca 2+ from ER stores in DKO cells than in WT cells (Fig. 2G). This is probably due to increased level of IP3R receptors in DKO cells as more IP3R may result in a significantly larger Ca 2+ release upon stimulation 48 .

Miro proteins associate with MICOS components
Mitochondrial cristae are maintained by the interplay between the MICOS complex located at cristae junctions and Sam50 located at the OMM 50,51 . Alteration of MICOS complex proteins by knockout or knockdown has revealed their importance for the maintenance of mitochondrial cristae ultrastructure [51][52][53] . The disruption of cristae architecture in DKO MEFs ( Fig. 1C and E) indicates a possible link between MICOS and Miro. Indeed, immunoprecipitation of GFP Miro1 and GFP Miro2 in HeLa cells revealed robust interactions with the core components of the MICOS complex Mic60/Mitofilin and Mic19/CHCHD3 and with the MIB complex forming component Sam50, whereas an unrelated OMM protein, Tom20, did not co-immunoprecipitate with either Miro1 or Miro2 (Fig. 3A). Interestingly, we did not detect interaction between Miro and Mitofusins or Mtx1, suggesting that these interactions might be low affinity, transient or cell type dependent (Fig. 3A). Control experiments with EGFP or the mitochondrially targeted GFP Su9 did not coimmunoprecipitate any of the MICOS components tested ( Fig. 3A and Supplementary Fig. 2A and B) confirming the specificity of the interactions. Importantly, we confirmed endogenous association between Miro2 and Sam50 using specific antibodies against Miro2 in WT mouse brains, while as expected, anti-Miro2 antibodies did not co-immunoprecipitate Sam50 in brain lysates from Miro2 KO animals (Fig. 3B). Furthermore, Mic60/Mitofilin was also specifically coimmunoprecipitated with Miro2 in lysates from WT brains (Fig. 3B). We further confirmed the interaction of endogenous Miro2 with Sam50 and Mic60/Mitofilin in situ using a Proximity Ligation Assay (PLA), which allows to test association of proteins that reside in close proximity (30 -40 nm) within the same complex 54 . In these experiments the presence of the antibody pair showed a 4-fold enrichment of fluorescent signals compared to the single antibodies confirming that native Miro2 and Sam50 as well as Miro2 and Mic19/CHCHD3 can be found in the same complex ( Fig. 3C, with quantification shown in Fig. 3D and E). Moreover, Blue Native (BN)-PAGE followed by western blotting with Miro1 and Miro2 antibodies revealed the appearance of complexes at 1000 /1200 and 500 -700 kDa that are similar to complexes detected with Mic19/CHCHD3, Sam50 or Mic60/Mitofilin antibodies (Fig. 3F) 50,53,55,56 . Thus, Miro proteins can be detected in a complex with key components of the MICOS complex on mitochondria.

Super-resolution imaging reveals that Miro is localized to discrete clusters on the mitochondrial surface
We noticed in our BN-PAGE experiments that both Miro1 and Miro2 antibodies detected strong and specific complexes around 140 kDa (Fig. 3G) Supplementary Fig. 4). To investigate this further we took advantage of dSTORM [57][58][59] imaging (a super-resolution technique which provides almost six-fold higher resolution than SIM) and performed correlative SIM / dSTORM 60 . Using dSTORM, we observed individually resolved clusters in Myc Miro1 expressing HeLa cells along the mitochondrial membrane that were not resolved under SIM imaging ( Supplementary Fig. 5). Similar nanoclusters were also observed upon dSTORM imaging of GFP Miro1 or GFP Miro2 in HeLa cells (Fig. 4C). To confirm that cluster formation is not due to the use of dSTORM, we compared imaging of Miro proteins with mitochondrial matrix targeted GFP Su9 (as a negative control). In contrast to GFP Miro2, GFP Su9 showed a more uniform distribution in the mitochondria (Fig. 4C). To test whether the levels of Miro proteins on the mitochondrial membrane can influence cluster formation, we imaged HeLa cells expressing low to very high amounts of GFP Miro2 and observed that in all conditions GFP Miro2 formed similar clusters along mitochondria, confirming that Miro protein levels on the OMM do not play a significant role in this nanoscale organization ( Supplementary Fig. 6). Further, we performed Density-Based Spatial Clustering of Applications with Noise (DBSCAN) which has been widely used to asses clustering of various membrane proteins 61 . The DBSCAN cluster map showed nanoscale domains formed by both GFP Miro1 and GFP Miro2 in Hela cells (Fig. 4D). We next quantified the sizes of Miro clusters by two complementary methods. First, we analyzed the cluster sizes using Ripley's K function 62 followed by quantification of cluster sizes post reconstruction of dSTORM images using the Feret's diameter (longest distance between any two points along the perimeter of each cluster). Ripley's K function indicated that both GFP Miro1 and GFP Miro2 formed a cluster size around 100-150 nm (Fig. 4E). The distribution of diameters revealed clusters ranging from 50 to 250 nm (Fig. 4F). GFP Miro1 clusters were found to have a median diameter around 100 nm (Median Feret's diameter = 108 nm ± 84-161 nm Interquartile Range (IQR)) very similar to that of GFP Miro2 clusters (Median Feret's diameter = 95 nm ± 67-150 nm (IQR)), both of which were much larger than the localization precision of the instrument (marked as a red bar in Fig. 4F). To test whether the observed nanoscale distribution of the Miro clusters is specific to HeLa cells or conserved across different cell types, we also imaged GFP Miro2 clusters in our MEF cells and in primary cultures of hippocampal neurons. We observed nanocluster-like organization of GFP Miro2, similar to HeLa cells, in both MEFs and in primary hippocampal neurons (Supplementary Fig. 7A and B) demonstrating that the nanoscale distribution of Miro protein complexes appears to be conserved across cell types.

Miro nanodomains are closely associated with MICOS nanoclusters
Previous studies using super-resolution STED imaging revealed discontinuous clusters of MICOS proteins along mitochondria 31  and Tom20 (previously shown to be non-interacting with Miro proteins). Figure 5E showed a significantly lower mean CCF between GFP Miro2 and Tom20 when compared to Miro2 and Mic60/Mitofilin (Fig. 5C, D and E) consistent with the specificity of the association between MICOS and Miro clusters.

Miro1 and Miro2 regulate MICOS complex formation and distribution
Our biochemical and super-resolution data revealed that Miro proteins form clusters in the mitochondrial surface that associate with MICOS clusters and interact with MICOS components and Sam50. In order to test whether Miro proteins are implicated in the correct association of the MICOS components we performed BN-PAGE on WT and DKO MEF cell lines. Miro1 antibodies detected specific bands from 500 to 700 kDa and around 1000 -1200 kDa that were not present in DKO cells (Fig. 5F). These bands correlated with Mic19/CHCHD3 positive bands detected in WT cells supporting that Miro can form part of high order molecular complexes containing MICOS components. Interestingly, whilst the signal of Mic19/CHCHD3 positive complexes that migrated at around 500-700 kDa and 1200 kDa were reduced in DKO cells in comparison to WT (Fig. 5F), a new molecular species positive for CHCHD3, appeared at a molecular weight of about 400 kDa in DKO lysates that was not present in WT cells (Fig. 5F), indicating that Miro proteins may be required for the correct assembly of MICOS complexes.
We aimed to understand whether the loss of Miro proteins affects the interaction between the core components of the MIB/MICOS complexes spanning OMM and IMM. Both Mic19/CHCHD3 and Sam50 pulled down the core components of MIB/MICOS, e.g. Mic60/Mitofilin, Mic19/CHCHD3 and Sam50 (Fig. 5G). These interactions appeared conserved in the absence of Miro (Fig. 5G) indicating that there is no gross alteration of the core MICOS complex. This was further investigated using PLA, which allows the in situ analysis of protein interactions. PLA analysis revealed that in absence of Miro there was a mild but significant decrease in the extent to which the core components of the MIB/MICOS interact ( Fig. 5H and I). This weakening of interaction was consistent between Sam50 and Mic60/Mitofilin and between Mic60/Mitofilin and Mic19/CHCHD3 ( Fig. 5H and I). Thus, while not essential for the assembly of MICOS complexes, Miro may regulate the overall stability of at least some species of these complexes, and its absence, leads to the destabilization of particular forms of the MICOS/MIB complexes.
Since Miro proteins can interact with several MICOS components and the depletion of Miro associates with defects in cristae architecture, we wanted to directly test how MICOS organization at the IMM is affected by the loss of Miro. We carried out dSTORM imaging after staining WT and DKO cells against the core MICOS component Mic19/CHCHD3. In WT cells Mic19/CHCHD3 showed an array of dense localizations evenly distributed throughout the entire mitochondria (Fig. 6A). In contrast, DKO cells showed mitochondrial regions with sporadic localizations of Mic19/CHCHD3 (Fig. 6A). In addition, we carried out DBSCAN analysis which can reveal the extent of cluster formation from the raw localization data 61 . In WT MEFs, DBSCAN analysis revealed the previously reported formation of Mic60/Mitofilin clusters arranged in a "discontinuous rail-like distribution" 31 ( Fig. 6A and B). Importantly, in DKO cells this array of clusters was severely affected, with large areas of mitochondria devoid of Mic19/CHCHD3 clusters (Red circles in Fig. 6B). This heterogeneity of Mic19/CHCHD3 clusters distribution correlated with altered Nearest Neighbor Distances (NND) distribution in DKO cells, compared to WT ( Fig. 6C; p < 0.001, Two Sample KS test). In comparison to WT MEFs, DKO cells showed decreased shorter NND distances (~60-110 nm) while longer NND distances (>110 nm) were increased (Fig. 6C).

MICOS cluster analysis in Miro DKO cells demonstrated that the distribution of MICOS clusters
throughout the mitochondria is affected by the loss of Miro proteins. Due to its role in mitochondrial transport we hypothesized that Miro might well serve as a link between the MICOS clusters and the cytoskeleton. This link would ensure homogeneous distribution of MICOS throughout mitochondria and any alteration might result in deformed cristae architecture. To test this hypothesis, we performed immunoprecipitation assays from lysates from WT and Miro DKO cells with the core MICOS components and the two TRAK adaptor proteins. We observed a strong co-immunoprecipitation of TRAK1 when the IP was carried out with antibodies against Mic19/CHCHD3 or Sam50 (Fig. 6D). Strikingly, TRAK1 was no longer able to coimmunoprecipitate with Mic19/CHCHD3 or Sam50 in Miro DKO cells indicating that the interaction between TRAK1 and Mic19/CHCHD3 (and Sam50) is regulated by Miro (Fig. 6D).
TRAK2 was also observed to co-immunoprecipitate with MICOS components only in WT cells although to a lower extent than that of TRAK1 (Fig. 6D), further supporting the Miro-dependent interaction between TRAK1/2 and MICOS. Reciprocally, both Sam50 and Mic19/CHCHD3 were readily detected in immunoprecipitates using a TRAK1 antibody from WT lysates but not from DKO cell lysates (Fig. 6D). Thus, Miro maintains an association between the cristae structures and the motor machineries through a complex containing MIB/MICOS components and the TRAK motor adaptor proteins.
We have recently shown that TRAK proteins can localize to, and induce the anterograde trafficking  (Fig. 6E). To force the uneven transport of membrane compartments in our MEF models, we expressed TRAK1 and the motor KIF5C (together with Tom70(1-70) GFP to label mitochondria) and observed that mitochondria accumulated in the periphery of the cells both in WT and in DKO cells as expected 13 . Strikingly, in WT cells the abundance of Mic19/CHCHD3 clusters matched the distal accumulation of mitochondria, showing higher signal in distally transported mitochondria when compared with mitochondria that did not reach the periphery, thus suggesting that TRAK1/KIF5C directed trafficking co-transported Mic19/CHCHD3 positive clusters (Fig. 6F). In contrast, in the absence of the TRAK1 / MICOS bridge mediated by Miro, distally transported mitochondria were almost devoid of Mic19/CHCHD3 clusters in DKO cells ( Fig. 6E and F) indicating that a critical role of Miro in regulating mitochondrial transport is to couple the TRAK/kinesin motor machineries to the MIB/MICOS complexes.

Miro regulated TRAK/MICOS bridge ensures appropriate distribution of inner components of mitochondria
The interaction between MICOS/Miro/TRAK carried an important implication. By linking the mitochondrial transport machinery to MICOS clusters Miro may ensure the concerted transport of the OMM with the IMM containing the complexes responsible for ATP generation. To test this hypothesis, we investigated the relative distribution of an IMM component of the OXPHOS system responsible for energy production (the ATPase subunit ATP5) and an OMM protein, Tom40. We took advantage of our recently developed tools to accurately measure signal distribution in cells with restricted size and shape growing in adhesive micropatterned substrates (see Experimental procedures for details) 13, 66 . We again forced the redistribution of mitochondria to the periphery of the cells by expressing TRAK1 and KIF5C and measured the relative distribution of Tom40 and ATP5 on mitochondria (Fig. 7). In WT cells both ATP5 and Tom40 signal presented a similar distribution in the mitochondrial network, consistent with a coordinated transport of both membranes ( Fig. 7A and B). In stark contrast, DKO cells showed a relative accumulation of the OMM marker, Tom40, in the periphery of the cells while the ATP5 signal appeared more accumulated in more proximal structures (Fig. 7A and B). When we did the projection of all tips of all 32 cells imaged (Fig. 7C and D) we noticed a consistent relative accumulation of the IMM marker (ATP5) in the periphery of WT cells with respect to Tom40 signal, while in DKO cells the relative accumulation of the OMM marker in the periphery was accentuated (Fig. 7C and D).
Further, we also measured the density of signal in mitochondria of ATP5 and Tom40 in concentric rings radiating out from the center of the cell (MitoSholl analysis) 13,67 and calculated the ratios of the normalized signals (ATP5 / Tom40) to plot them as a function of distance (Fig.   7E). The resulting plot shows that, in WT cells, the ATP5 / Tom40 ratio increases towards the periphery indicating that Miro-regulated TRAK1/KIF5C mitochondrial transport preferentially enriches the transported mitochondria with IMM components, perhaps by accumulating them by the pulling forces applied onto the MICOS complexes. In contrast, in DKO cells this ratio sharply drops in the most distal regions of the cell (Fig. 7E) indicating that, without Miro, TRAK directed mitochondrial transport fails to efficiently couple the IMM to the mitochondrial transport pathway.
Altogether these results suggest that Miro acts as a critical adaptor to link the mitochondrial transport machinery to the mitochondrial cristae organization to ensure the concerted transport of the OMM with the IMM components to guarantee the appropriate provision of energy to the regions where mitochondria are delivered.

Discussion
Here we demonstrate the nanoscale spatial organization and protein complex formation of the Miro mitochondrial GTPases and their dual role in regulating the formation and functionality of the ERMCS and in connecting the MIB/MICOS complexes, responsible for maintaining cristae architecture, to the mitochondrial transport pathway. Miro proteins link the TRAK motor adaptors to the MICOS complexes to ensure the correct distribution of MICOS throughout the mitochondria and to facilitate the coordinated delivery of both membranes during mitochondrial transport.
ERMCS are key structures for the regulation of Ca 2+ communication between ER and mitochondria and play important roles in the regulation of mitochondrial division and the segregation of mitochondria and mtDNA in newly generated mitochondrial tips 68 . In yeast, the Miro homolog Gem1 is associated with the regulation of ER-mitochondria connections 15 . In mammals, Miro interacts with Mitofusins and DISC1 which are known to be associated with ERMCS in mammalian cells [69][70][71] . Our data shows that, in addition, the absence of Miro proteins associates with a decrease in contacts between ER and mitochondria, which correlates with alterations in mitochondrial Ca 2+ uptake and in the intraluminal concentration of Ca 2+ in the ER.
This role of Miro proteins in maintaining the ER-mitochondrial Ca 2+ homeostasis is supported by recent reports that link dMiro to the control of the VDAC1-IP3R complexes that regulate Ca 2+ communication in Drosophila 41,42 . Interestingly, the increase in the levels of IP3R that we report here are associated with alterations in Ca 2+ homeostasis. Increased levels of IP3R have been previously associated with increased ER Ca 2+ release which in turn has an impact on muscle contractility, induction of apoptosis and in the regulation of mitosis and that has been associated with multiple human diseases [72][73][74][75] . Altered Ca 2+ communication between the ER and mitochondria due to a decrease in ERMCS might be responsible for the increase in the protein levels of IP3R in Miro DKO cells. This provides striking evidence for the existence of a regulatory feedback mechanism that can control the number and composition of ER-mitochondrial contacts depending on the activity of the ERMCS complexes. It is worth noting that the upregulation of IP3R might be a direct result of altered ER-associated degradation (ERAD) at ERMCS 76  Furthermore, we demonstrate that the loss of Miro proteins disrupts the previously reported "discontinuous rail-like" distribution of MICOS complexes throughout mitochondria 31 as seen by the changes in our nearest-neighbor distances analysis.
The disruption of mitochondrial cristae architecture by loss of Miro, whilst widespread, does not perfectly match that observed upon deletion of MICOS components or Sam50, which are usually described as "onion-like" membranous structures 29  leading to the loss of their "discontinuous rail-like" distribution 31 . In addition, it is possible that MIB/MICOS complexes that are not anchored to the cytoskeleton are rendered less stable and dissociate to a certain extent, explaining why we observe a small but significant decrease in the interaction between core components of the MIB/MICOS complexes in situ in our PLA assays.
Miro proteins are critical regulators of mitochondrial trafficking from yeast to mammals 7 . The accepted model of mitochondrial trafficking presumed that Miro proteins provided a link between the OMM and the microtubule motors kinesin and dynein through the recruitment of the adaptors TRAK1 and TRAK2 [81][82][83][84] . We have recently challenged this idea by showing that Miro is not essential for kinesin/TRAK directed mitochondrial movement but rather regulates its activity 13 . In the present paper we demonstrate that an important function of Miro proteins is to regulate the association of the mitochondrial transport machinery to the MICOS complexes. By regulating this association, Miro proteins facilitate the concerted transport of both mitochondrial membranes to the cellular regions where they are needed. These findings are supported by a recent report linking Mic60/Mitofilin function, a core component of the MICOS complex, to mitochondrial motility in Drosophila 37 . Although the mechanism of such regulation remains unknown, in that work the authors report a decrease in dMiro levels upon genetic deletion of Mic60/Mitofilin. It remains to be tested if Miro is stabilized by the formation of assemblies between TRAK adaptors and MIB/MICOS complexes and whether the dissociation of these high order assemblies renders Miro prone to be degraded.
A major consequence of this regulation is that the mitochondrial cristae that are associated to the OMM through the MIB complexes could potentially be distributed in a regulated manner by the concerted action of Miro and TRAK proteins. Thus, by connecting the transport machinery to the MICOS complexes Miro opens the door to a new way of mitochondrial regulation by controlling the distribution of mitochondrial cristae within mitochondria. In addition, the direct association with the cristae molecular architecture may provide a mechanism for the transport machinery to sense the functionality of the mitochondria to be transported.

The dual role of Miro in regulating the number of ERMCS and the distribution of MIB/MICOS complexes has parallels with the ER-mitochondria organizing network (ERMIONE) in
Saccharomyces cerevisiae 85,86 . ERMIONE in yeast is formed by ERMES (ER-Mitochondria Encounter Structure) and MICOS which then recruit the TOM complex and Sam50 and is involved in lipid homeostasis, mitochondrial biogenesis and maintenance of mitochondrial morphology 86 .
However, there is little evidence of a structural molecular assembly in mammals that is 24 homologous to the yeast ERMIONE. Our work supports a central role for Miro proteins in coordinating and integrating different mitochondrial functions by organizing and controlling a mitochondrial signalling network that includes the mitochondrial transport pathway, the MIB/MICOS complexes and the ERMCS and that might be the functional equivalent of the ERMIONE in mammalian cells.

Materials and Methods
Plasmid DNA, Cell culture and transfection GFP Miro1 and GFP Miro2 were generated as described earlier 87

Confocal, SIM, correlated SIM, dSTORM and 3D dSTORM imaging
Confocal imaging was performed on a Zeiss LSM 700 confocal microscope, Structured Illumination Microscopy was performed on Zeiss Elyra PS.1, correlated SIM and dSTORM imaging was performed on the same microscope with 100 × 1.46 NA oil immersion objective. All dSTORM imaging was conducted using a custom-built microscope and analyzed using software written in C++ and Python 94 . Further details about microscopes used in this study can be found in supplementary experimental procedures.

Image processing and analysis
Post reconstruction, images were first corrected for X-Y drift using 1 to 3 fiducials present in the images. Images were either binned using 20 nm pixel size (for dSTORM and colocalization with MICOS components) or 30 nm pixel size (for dimer formation). The reconstructed image was blurred with a Gaussian function with a sigma radius of 0.75 (which translate to 20-30 nm) using 'Accurate Gaussian blur' plugin. For measuring the sizes of nanoclusters, first images were thresholded, and then each particle was detected using particle analyzer algorithm followed by particle size measurement using Feret's diameter plugin present in ImageJ. For colocalization of dual color STORM images, images in 555 nm and 647 nm channels were blurred equally then both channels were aligned using 'Align images FFT' plugin present within GDSC ImageJ plugin (freely downloadable from University of Sussex) which uses a Gaussian for sub-pixel alignment. Van steensel's cross correlation was calculated from the aligned images using plugin JACoP with Xshift of 1 m. DBSCAN and Ripley's K function was determined according to a previously published protocol 95 .

Statistical Analysis
Excel Software (Microsoft), Origin (OriginLab Corporation) and GraphPad Prism (GraphPad Software, Inc) were used to analyze the data. Statistical significance was calculated using either Student's t-test or Mann-Whitney U test (for non-parametric two independent samples) unless otherwise stated. Statistical differences between multiple conditions were performed by one-way ANOVA followed by post hoc Tukey's tests (for comparison between two conditions) or Bonferroni test (for comparison between multiple conditions). Statistical significance was pre-28 fixed at P < 0.05, described as *P < 0.05; **P < 0.01 and ***P < 0.001. All values in text are given as Mean ± S.E.M unless specified.        Significance: * p < 0.05; ** p < 0.01 and *** p < 0.001