A role for Mitochondrial Rho GTPase 1 (MIRO1) in motility and membrane dynamics of peroxisomes

Peroxisomes are dynamic organelles which fulfil essential roles in lipid and ROS metabolism. Peroxisome movement and positioning allows interaction with other organelles and is crucial for their cellular function. In mammalian cells, such movement is microtubule‐dependent and mediated by kinesin and dynein motors. The mechanisms of motor recruitment to peroxisomes are largely unknown, as well as the role this plays in peroxisome membrane dynamics and proliferation. Here, using a combination of microscopy, live‐cell imaging analysis and mathematical modelling, we identify a role for Mitochondrial Rho GTPase 1 (MIRO1) as an adaptor for microtubule‐dependent peroxisome motility in mammalian cells. We show that MIRO1 is targeted to peroxisomes and alters their distribution and motility. Using a peroxisome‐targeted MIRO1 fusion protein, we demonstrate that MIRO1‐mediated pulling forces contribute to peroxisome membrane elongation and proliferation in cellular models of peroxisome disease. Our findings reveal a molecular mechanism for establishing peroxisome‐motor protein associations in mammalian cells and provide new insights into peroxisome membrane dynamics in health and disease.


| INTRODUCTION
Peroxisomes are dynamic, multifunctional organelles that vary in size, number and shape depending on cell type, environmental stimuli and metabolic demand, 1 but the underlying molecular mechanisms which govern this versatility are not fully understood. Similar to mitochondria, peroxisomes are oxidative organelles that fulfil important functions in lipid metabolism and ROS homeostasis rendering them essential for human health and development. 2,3 Peroxisomes metabolically cooperate and physically interact with a variety of subcellular organelles including the ER, mitochondria, lipid droplets and other peroxisomes. [4][5][6] These functions require peroxisome positioning and movement within eukaryotic cells.
Whereas in yeast and plant cells peroxisome motility depends on actin filaments and myosin motors, 7,8 in mammalian cells peroxisomes move bidirectionally via microtubules, using both kinesin and dynein motors. [9][10][11][12] The shape and number of peroxisomes is controlled by PEX11β, a peroxisomal membrane protein, which induces elongation and remodelling of the peroxisomal membrane and acts as a GTPase activating protein on the large fission GTPase DNM1L. [13][14][15] Loss of PEX11β was recently linked to spindle misorientation and peroxisome mislocalisation in mitosis causing imbalances in epidermal differentiation. 16 These findings underline the importance of peroxisome multiplication, distribution and inheritance for cell fate decisions.
Abbreviations: TA, tail-anchored; TMD, transmembrane domain; WT, wild type; ROS, reactive oxygen species; ER, endoplasmic reticulum Although key factors required for peroxisome dynamics and multiplication have been identified, it is currently unclear to what extent cytoskeletal tracks, docking factors and pulling forces mediated by associated motor proteins contribute to these processes, in particular in mammals. 17 In baker's yeast, peroxisome distribution and inheritance depends on actin, the myosin motor Myo2 and specific adaptor proteins, Inp1 and Inp2, at the peroxisomal membrane. 7 Furthermore, the peroxins Pex3 and Pex19 have been found to interact with myosin motors. 18,19 In contrast, little is known about the recruitment of microtubule motors to peroxisomes in mammalian cells. 20 Here, we identify the Ras GTPase MIRO1 as a potential adaptor for microtubule-based peroxisome motility in mammalian cells. MIRO proteins were initially identified on the outer mitochondrial membrane 21 where they, together with TRAK1/2, link the microtubule motors kinesin and dynein to mitochondria, [22][23][24][25] and play key roles in mitochondrial motility, homeostasis and inheritance. 26,27 Mammalian MIRO1 and MIRO2 share 60% similarity and an analogous structure containing 2 GTPase and 2 EF-hand calcium binding domains. 21,28 Studies on mammalian MIRO proteins have focused mainly on MIRO1 due to its clear role in mitochondrial motility, particularly in neurons. 22,25 Loss of MIRO1-directed mitochondrial movement and distribution result in neurological defects. 26 MIRO1-mediated mitochondrial positioning is also suggested to shape intracellular energy gradients required for cell migration. 29 We show that MIRO1 localises to peroxisomes and mitochondria, and alters peroxisome distribution and motility. Furthermore, we demonstrate that an exclusively peroxisome-targeted MIRO1 can mediate pulling forces which contribute to peroxisome membrane elongation and proliferation in a cell type-dependent manner. To better understand the versatility of peroxisomes in mammalian cells, we build a first mathematical model of peroxisome dynamics. This model helps to explain the underlying principles of peroxisome morphologies induced by MIRO1-mediated pulling forces and other factors which influence peroxisomal membrane dynamics.

| MIRO1 is dually targeted to peroxisomes and mitochondria
Previous studies revealed a dual mitochondrial and peroxisomal localisation of several C-tail-anchored (TA) membrane proteins including FIS1, MFF and GDAP1, which function in peroxisomal and mitochondrial division. [30][31][32][33] In a recent study on the targeting of TA proteins to different organelles, we provided preliminary evidence for a dual peroxisomal and mitochondrial localisation of the Ras GTPases MIRO1 and MIRO2. 34 MIRO1 was initially identified on the outer mitochondrial membrane, 21 and forms a protein complex with TRAK1/2 that includes both kinesin and dynein motors, promoting mitochondrial movement through the microtubule cytoskeleton. [22][23][24][25] A dual mitochondrial and peroxisomal localisation of MIRO1 was confirmed by immunofluorescence after expression of Myc-MIRO1 in COS-7 cells ( Figure 1A). Furthermore, we previously reported endogenous MIRO1 in highly purified peroxisomal and mitochondrial fractions, 34 in agreement with proteomics data. 35,36 The targeting of all known TA proteins to peroxisomes requires the peroxisomal import receptor/chaperone PEX19. 34 For MIRO1, PEX19 binding was shown by immunoprecipitation after coexpression of Myc-MIRO1 and HA-PEX19 in COS-7 cells ( Figure 1B) suggesting a role for PEX19 in the targeting of MIRO1 to peroxisomes. Additionally, in a high-throughput interaction study, MIRO1 was identified as a PEX19 interaction partner. 37 These findings are also consistent with the known organelle targeting signals: MIRO1 possesses a transmembrane domain (TMD) with relatively low hydrophobicity (GRAVY, 1.3) and a moderate net charge in the tail region (1.9), which based on our previous work would be indicative of a TA protein that localises predominantly to mitochondria but has a potential for peroxisomal targeting. 34 Overall, our findings support a dual localisation of MIRO1 at mitochondria and peroxisomes.

| MIRO1 alters peroxisome distribution in COS-7 cells
MIRO1 has been shown to play a key role in mitochondrial motility and distribution in mammalian cells. 26 To determine if MIRO1 also plays a role in peroxisome positioning we expressed Myc-tagged wild type (WT) and mutated versions in COS-7 cells, and analysed their effect on peroxisome distribution (Figures 1A,C and S1). As previously described, 21,38 the expression of Myc-MIRO1 resulted in abnormal mitochondrial morphologies ( Figures 1A and S1). To avoid potential secondary effects due to dysfunctional mitochondria, we generated an exclusively peroxisomal set of MIRO1 proteins by altering the C-terminal TMD using a previously described PEX26/ ALDP construct ( Figure 1C). 39 Expression of the resulting Myc-MIR-O1 Pex fusion protein in COS-7 cells revealed an exclusively peroxisomal localisation, with no effects on mitochondrial morphology and distribution ( Figure 1D,E). Peroxisomes in COS-7 cells usually distribute uniformly throughout the cytoplasm. 30,40 Interestingly, expression of Myc-MIRO1 Pex or Myc-MIRO1 V13-Pex , a constitutively active GTPase mutant, induced peroxisome redistribution and accumulation at the cell periphery ( Figure 1D,F,G). On the other hand, expression of dominant negative Myc-MIRO1 N18-Pex and EF-hand mutant Myc-MIRO1 KK-Pex resulted in peroxisome accumulations which were scattered throughout the cytoplasm ( Figure 1F,G). Comparable results were obtained with the dually targeted MIRO1 versions ( Figure S1B). Myc-MIRO1 ΔTM , a version lacking the TMD/tail sequence, localised to the cytoplasm and had no effect on peroxisome distribution, indicating that membrane anchorage is required for MIRO1 function ( Figure S1). Furthermore, depolymerisation of microtubules with nocodazole in Myc-MIRO1 Pex expressing cells abolished accumulation of peroxisomes in the cell periphery, suggesting that an intact microtubule cytoskeleton is required for peroxisome distribution via MIRO1 ( Figure S2A). Our findings indicate that, similar to its role on mitochondria, MIRO1 can alter peroxisome distribution and positioning by affecting microtubuledependent peroxisome motility. to be unaffected. 43 In agreement with those findings, we did not detect any alterations in peroxisome distribution ( Figure S2C) or motility ( Figure S2D,E). These findings indicate that when targeted to peroxisomes in COS-7 cells, active MIRO1, a known adaptor for the microtubule plus-end motor kinesin, can redistribute peroxisomes to the cell periphery (where microtubule plus ends are located) in a microtubule-dependent manner. However, MIRO1 may not be the only adaptor for microtubule-dependent motor proteins at peroxisomes, as its loss is apparently not essential to maintain peroxisome distribution and motility. It is possible that MIRO2, which also localises to peroxisomes, 34 can complement loss of MIRO1. Furthermore, peroxisomes may tether to or "hitch-hike" other moving organelles to maintain their distribution. The latter process has been observed in filamentous fungi. 44

| MIRO1 induces peroxisome proliferation in human skin fibroblasts
The peroxisome-targeted MIRO1 represents a new tool to manipulate peroxisome motility and to exert motor-driven pulling forces at peroxisomes under control and disease conditions. Peroxisomes in fibroblasts from patients with peroxisomal disorders are often enlarged and reduced in number, and tend to cluster and detach from microtubules. 45 We first expressed Myc-MIRO1 Pex in human skin fibroblasts from a healthy control and examined its effect on the peroxisomal compartment ( Figure 2D). Surprisingly, in these cells peroxisomes did not accumulate at the cell periphery but instead proliferated, presenting a significant increase in number (mean peroxisome number/cell: control 740 AE 50; Myc-MIRO1 Pex 1040 AE 100, n = 24; Figure 2E). In addition, the percentage of motile peroxisomes that moved in a microtubule-dependent manner was significantly increased ( Figure 2F; Figure S2F; Videos S3 and S4). These findings indicate that MIRO1-bound motor proteins can exert forces at peroxisomes, which result in peroxisome division, thus increasing peroxisome number. Separation by pulling forces is only possible when the peroxisome is tethered to another structure, as it would otherwise simply move in the direction of the pulling force ( Figure 4B). This untethered motion is observed in COS-7 cells, where MIRO1 expression accumulates peroxisomes in the cell periphery where microtubule-plus ends are located ( Figures 1 and 4B). We recently revealed that peroxisome-ER membrane contacts are mediated by peroxisomal ACBD5 that interacts with ER-resident VAPB to form a peroxisome-ER tether. 46 Loss of ACBD5 increased the movement of peroxisomes in human skin fibroblasts, indicating that peroxisome-ER membrane contacts restrict peroxisome motility. In line with this, our analyses reveal that the percentage of fast moving peroxisomes in control fibroblasts is lower than that in control COS-7 cells (4.5% AE 0.4% vs 5.2% AE 0.7%). We suggest that peroxisome-ER tethering is celltype specific and that MIRO1/motor-mediated pulling forces can induce peroxisome proliferation in fibroblasts, whereas in COS-7 cells peroxisomes are dragged towards the cell periphery ( Figure 4B). These findings indicate that a close interplay between tethering and motile forces modulates not only peroxisome distribution but also proliferation.
To analyse the impact of MIRO1 expression on peroxisomes in patient fibroblasts, we expressed Myc-MIRO1 Pex in PEX5 and PEX14 deficient cells. PEX5 and PEX14 are proteins of the peroxisomal matrix protein import machinery, and loss of function leads to "empty" membrane structures (so called "ghosts") that lack peroxisomal enzymes and are metabolically inactive. Peroxisomes in those cells are often enlarged and reduced in number ( Figure 2D). Expression of Myc-MIRO1 Pex in both PEX5 and PEX14 deficient cells induced peroxisome proliferation, but many peroxisomes remained enlarged ( Figure 2D,E). MIRO1 expression also significantly increased peroxisome motility in patient cells ( Figure 2F; Figure S2G-H), most prominently for the smaller peroxisomes (Videos S5-S8). In contrast to a recent report, we observed that peroxisomes in PEX14 deficient cells are motile. 47 These findings show that MIRO1-mediated pulling forces can at least partially induce the proliferation of metabolically inactive peroxisomes, indicating that membrane components are the most relevant factors for this process.
2.5 | Peroxisome-targeted MIRO1 promotes the formation of extended membrane protrusions in PEX5 deficient fibroblasts Peroxisomes are highly dynamic organelles that can be found as spherical or elongated structures and also form membrane      forces, but proliferation is reduced, likely due to altered membrane lipids. 51 Despite their fundamental importance to cell physiology, the mechanisms that mediate and regulate peroxisomal membrane

| Plasmids and antibodies
For cloning of peroxisome-targeted MIRO1, the C-terminal TMD and tail of Myc-MIRO1 were exchanged by a PEX26/ALDP fragment previously shown to target proteins to the peroxisomal membrane. 39 See Table S1 for details of plasmids used in this study, Table S2 for plasmids generated in this study and Table S3 for details of primers used.
All constructs produced were confirmed by sequencing (Eurofins Genomics). Details on all antibodies used in this study can be found in Table S4.

| Immunofluorescence and microscopy
Cells were processed for immunofluorescence 24  frames, from each data set. Next, the trajectories were re-centred such that each trajectory started at (0,0), and subsequently smoothed applying a simple moving-average algorithm using a Hann window.
The first 20 time-frames for these trajectories were then plotted starting at a centre. For cumulative distribution function (CDF) plots, basic instantaneous trajectory speed profiles were estimated by calculating the distance moved between each time-point in the trajectory. These speeds were then pooled and converted into an ECDF.
By pooling the speeds for all data sets for a given condition a single ECDF for each condition was generated. Trajectories for the tracked peroxisomes were analysed by splitting their instantaneous speeds into 2 groups, using a cut-off for linear motion speed of 0.24 μm/s. 42 The relative populations of the 2 groups of peroxisome speeds were used as an indication of the amount of linear motion for each data set, and compared against all trajectories to obtain a percentage of microtubule-dependent motility per cell. The number of peroxisomes per cell was obtained from the motility analysis output, and determined by the detected peroxisome from the first frame of each analysed cell. Peroxisome protrusion lengths were obtained from live-cell imaging data and manually measured using MetaMorph 7. Each observed protrusion was measured at the longest point of extension.
Kymographs were generated using ImageJ (developed at the National Institutes of Health).

| Mathematical modelling
Each peroxisome was described by its body radius r and elongation length L. Simulations were started with 250 peroxisomes, each with a random initial radius and no elongation. After each time step (Δt = 1 second), we implemented 3 processes. First, lipid flow from the ER into the body: the body surface area was increased by αΔt with probability e −γA , where A is the total area of all peroxisomes. Second, if the body radius was above r min , the elongation was increased by length vΔt, with the extra elongation area taken from the body. Third, when the elongation length was longer than L min , peroxisomes underwent division with probability βLΔt. In addition, during each time step, each peroxisome had probability Δt/τ of being removed by turnover. Simulations were carried out in C++ and MATLAB. See Supporting information for full details. The MATLAB code can be made available upon request.