Microtubule dynamics regulates mitochondrial fission

Mitochondria are organized as tubular networks in the cell and undergo fission and fusion. While several of the molecular players involved in mediating mitochondrial dynamics have been identified, the precise cellular cues that initiate fission or fusion remain largely unknown. In fission yeast, mitochondria are organized along microtubule bundles. Here, we employed deletions of kinesin-like proteins to perturb microtubule dynamics, and determined that cells with long microtubules exhibited long, but fewer mitochondria, whereas cells with short microtubules exhibited short, but several mitochondria due to reduced mitochondrial fission in the former and elevated fission in the latter. Correspondingly, upon onset of closed mitosis in fission yeast, wherein interphase microtubules assemble to form the spindle within the nucleus, we measured increased mitochondrial fission. We determined that the consequent rise in the mitochondrial copy number was necessary to reduce partitioning errors while stochastically partitioning mitochondria between daughter cells. We discovered that the association of mitochondria with microtubules physically impeded the assembly of the fission protein Dnm1 around mitochondria, resulting in inhibition of mitochondrial fission. Taken together, we demonstrate a novel mechanism for regulation of mitochondrial fission that is dictated by the interaction between mitochondria and the microtubule cytoskeleton.

. Mitochondrial number is inversely proportional to microtubule length. a, Maximum intensity projections of deconvolved Z-stack images of microtubules (left), mitochondria (centre) and their composite (right) in wild-type ('WT', top, strain KI001, see Table S1), Klp5Δ/Klp6Δ ('MT long ', strain G3B, see Table S1) and Klp4Δ ('MT short ', strain G5B, see Table S1) cells. b, Box plot of length of anti-parallel microtubule bundles in WT, MT long and MT short cells (n=40, 37 and 63 bundles respectively). c, Box plot of number of mitochondria per cell in WT, MT long and MT short cells (n=10, 12 and 13 cells respectively). d, Box plot of the total volume of mitochondria per cell in WT, MT long and MT short cells normalised to mean total wild-type mitochondrial volume (n=10, 12 and 13 cells respectively). Light grey crosses represent outliers, asterisk represents significance (p<0.05) and 'n.s.' indicates no significant difference.
The fission yeast protein Mmb1 has been identified to associate mitochondria with dynamic microtubules 9 . Upon microtubule depolymerisation using methyl benzimidazol-2-ylcarbamate (MBC, see Supplementary Information) or deletion of Mmb1, mitochondria have been observed to undergo rapid fragmentation [9][10][11] . We observed that mitochondria did indeed undergo increased fission in the absence of microtubules, but did not subsequently aggregate as reported previously (Extended Data Fig. 1a-c). Instead the fragmented mitochondria were mobile and frequently in close contact with each other (Extended Data Fig. 1b, c, Supplementary Video S1). Since the depolymerisation of microtubules had a direct effect on mitochondrial fission, we set out to study the consequence of modification of microtubule dynamics on mitochondrial dynamics. To this end, we visualised the mitochondria and microtubules of fission yeast strains carrying deletions of antagonistic kinesin-like proteins, Klp5/Klp6 and Klp4 in high-resolution deconvolved images (Fig. 1a, Supplementary Videos S2, S3 and S4).
The heteromeric Klp5/Klp6 motor is required for maintenance of interphase microtubule length by promoting catastrophe at microtubule plus ends 12,13 . Cells lacking Klp5 and Klp6 exhibited long microtubules ('MT long ', Fig. 1b) as reported previously due to a decreased catastrophe rate 12 . Klp4 is required for polarised growth in fission yeast and has been suggested to promote microtubule growth 14,15 . As a result, in the absence of Klp4, microtubule anti-parallel bundles were only half the length of wild-type bundles ('MT short ', In contrast, Klp4Δ cells had 10 mitochondria per cell on an average (Fig. 1c). This indicated that the number of mitochondria per cell was inversely related to the length of microtubule bundle. So too, cells carrying a deletion of the EB1 homologue Mal3 also produced small microtubule bundles 16 and additionally contained several mitochondria as reported previously 9 (Extended Data Fig. 1d). However, the decrease in the number of mitochondria in Klp5Δ/Klp6Δ cells and increase in Klp4Δ cells were not at the expense of mitochondrial volume, since the net mitochondrial volume in both cases was comparable to wild-type mitochondrial volume (Fig. 1d, Extended Data Fig. 1e).
To understand the difference in mitochondrial numbers in wild-type, Klp5Δ/Klp6Δ and Klp4Δ cells, we acquired and analysed time-lapse videos at the single mitochondrion level in all three cases (Fig. 2a, Supplementary Videos S5, S6 and S7). Similar to our observations from high-resolution images, we measured ~5, ~2 and ~8 mitochondria on an average in wild-type, Klp5Δ/Klp6Δ and Klp4Δ cells respectively ( Fig. 1c and 2b). Analysis of evolution of these mitochondrial numbers revealed no significant changes with time ( Fig.   2b). Interestingly, we discovered that the size of the individual mitochondrion correlated with the length of the microtubule bundle in the cell, with Klp5Δ/Klp6Δ cells containing the longest mitochondria and Klp4Δ cells having short mitochondria. Wild-type cells predictably exhibited mitochondrial size ranges that lay between that of Klp5Δ/Klp6Δ and Klp4Δ cells   Table S1), Klp5Δ/Klp6Δ ('MT long ', strain G3B, see Table S1) and Klp4Δ ('MT short ', strain G5B, see Table S1) cells represented in the intensity map indicated to the right of the images. White solid arrowheads point to representative fusion events and open arrowheads to fission events. Time is indicated above the images in mm:ss. b, Evolution of mitochondrial number with time indicated as mean (solid grey line) and standard error of the mean (shaded region) for WT, MT long and MT short cells (n=21, 15 and 8 cells respectively). c, Box plot of the size of mitochondria in WT, MT long and MT short cells, calculated as the length of the major axis of an ellipse fitted to each mitochondrion (n=1613, 739 and 1326 mitochondria respectively). d, Box plot of the fission rate of mitochondria per second in WT, MT long and MT short cells (n=21, 15 and 8 cells respectively). e, Box plot of the fusion rate of mitochondria per second in WT, MT long and MT short cells (n=21, 15 and 8 cells respectively). Light grey crosses represent outliers, asterisk represents significance (p<0.05) and 'n.s.' indicates no significant difference.
Additionally, mitochondria in wild-type cells underwent ~1 fission and ~1 fusion event every minute on an average, whereas Klp5Δ/Klp6Δ cells exhibited a fission rate that was half that of wild-type, and Klp4Δ mitochondria had a fission rate that was almost double that of ( Fig. 2d). The fusion rate of mitochondria in Klp4Δ cells was slightly higher than in wildtype and Klp5Δ/Klp6Δ cells (Fig. 2e), likely due to the increased number of mitochondria in Klp4Δ cells that could participate in fusion. However, the resulting ratio of the mean fission rate to the mean fusion rate was ~1, ~0.5 and ~1.3 in wild-type, Klp5Δ/Klp6Δ and Klp4Δ cells respectively. We therefore concluded that the difference in mitochondrial numbers between wild-type cells and Klp5Δ/Klp6Δ and Klp4Δ arose primarily due to the changes in fission rates of mitochondria. Increase in oxidative stress via reactive oxygen species (ROS) levels has also been described to induce mitochondrial fission 17 . However, we measured no difference in ROS levels between wild-type, Klp5Δ/Klp6Δ, and Klp4Δ cells (Extended Data Fig. 2d).
To test the role of the microtubule in determining the length and dynamics of mitochondria, we depolymerised microtubules in wild-type and Klp5Δ/Klp6Δ cells and visualised the mitochondria in time-lapse movies (Extended Data Fig. 3a). In both cases, upon microtubule depolymerisation, we observed a switch in mitochondrial morphology to increased numbers, increased fission and unaltered fusion rate that was reminiscent of Klp4Δ cells (Extended Data Fig. 3b-h).
Fission yeast undergoes closed mitosis, wherein the nuclear envelope does not undergo breakdown during cell division 18 . Upon onset of mitosis in fission yeast, the interphase microtubules that were previously in the cytoplasm are reorganised to form the spindle inside the closed nucleus. This natural situation mimics the depolymerisation of microtubules via the chemical inhibitor MBC. Therefore, we set out to study the changes in the mitochondrial network upon cell entry into mitosis. We first obtained high-resolution deconvolved images of the microtubule and mitochondria in fission yeast cells undergoing cell division (Fig. 3a, Extended Data Fig. 4a, Supplementary Video S8). We observed that dividing wild-type cells had ~4x the number of mitochondria in interphase cells (Fig. 3b).
Moreover, similar to what was seen in cells lacking microtubules or Klp4 (Fig. 1a), mitochondria in dividing cells were shorter and more rounded (Fig. 3a, Extended Data Fig.   4a). There was no relationship between length of the mitotic spindle and the number of mitochondria (Fig. 3b, Extended Data Fig. 4a), indicating that the increased fission likely occurred fairly early upon entry into mitosis. Analysis of time-lapse videos of wild-type cells before and 10min after entry into mitosis revealed a doubling of mitochondrial numbers in this time (Fig. 3c, d, Supplementary Video S9). Note that this is likely an underestimate since we were unable to reliably resolve some of the mitochondria in these lower-resolution images. In this same period of time, non-dividing interphase cells did not show any change in mitochondrial numbers (Extended Data Fig. 4b, c).  Table S1) undergoing division. b, Scatter plot of the length of the mitotic spindle vs. the number of mitochondria per cell in dividing cells (n=13 cells). c, Montage of maximum intensity projected confocal Zstack images of the microtubules (top) and mitochondria (bottom) in a wild-type cell (strain KI001, see Table S1) undergoing cell division represented in the intensity map indicated to the right of the images. White open arrowheads point to representative fission events. Time is indicated above the images in mm:ss. d, Bar plot of the mean number of mitochondria per cell before ('00:00') and 10mins after ('10:00') the onset of mitosis. Solid grey lines represent data from individual cells (n=16 cells).
Taken together, our results suggest that mitochondrial fission is inhibited by the association of mitochondria with microtubules. The mitochondrial outer-membrane fission protein in yeast is the dynamin-related GTPase, Dnm1 10 . Dnm1 brings about the fission of mitochondria by self-assembling into a ring around the diameter of the mitochondria and employing its GTPase activity to effect the scission 19 . In the absence of Dnm1, mitochondria are organised as extended, fused 'nets' 10,20 . We hypothesised that the binding of mitochondria to microtubules physically hinders the formation of a complete ring of Dnm1 around the mitochondrion. To test this hypothesis, we introduced GFP-tagged Dnm1 under the control of the thiamine-repressible nmt1 promoter 21 into wild-type cells expressing fluorescent microtubules (see Table S1). C-or N-terminal tagging of Dnm1 renders it inactive 10 and therefore the scission activity of Dnm1 could not be directly observed (Extended Data Fig. 5a). However, we visualised the localisation of Dnm1 'spots', representing Dnm1 assembly, with respect to microtubules. Although almost 100% of the Dnm1 spots colocalised with mitochondria (Extended Data Fig. 5a), 85% of the spots did not localise on the microtubule (Fig. 4a, Supplementary Video S10), indicating that the presence of microtubule had an inhibitory effect on the assembly of the Dnm1 ring.
Moreover, we counted a slight decrease in Dnm1 spots in Klp5Δ/Klp6Δ, and ~2x increase in Klp4Δ cells as compared to wild-type cells (Extended Data Fig. 5b), corroborating our previous observations indicating decreased fission in the former and increased fission in the latter (Fig. 2d).
We discovered that microtubule dynamics is a strong determinant for mitochondrial dynamics and thereby, morphology (Extended Data Fig. 5c). Mitochondrial fission rate was proportional to microtubule catastrophe rate, whereas the fusion rate was almost completely insensitive to microtubule dynamics (Extended Data Fig. 5c). While previous studies discounted the role of motors in the determination of mitochondrial positioning in fission yeast 11,[22][23][24] , we have identified kinesin-like proteins that regulate mitochondrial morphology through their control over microtubule length.
We propose that equilibrium between fission and fusion of mitochondria in interphase cells can be maintained solely by the inherent dynamic instability of microtubules. Perturbation of microtubule dynamics shifts the balance between fission and fusion rates to a different set point, leading to modification of mitochondrial numbers and morphology. Fission yeast cells likely employ this strategy to fragment interphase mitochondria by emptying the cytoplasm of its microtubules upon onset of mitosis, thereby ensuring stochastic 25 albeit equal partitioning of the several small mitochondria between future daughter cells (Fig. 4b).  Table S1). A majority of Dnm1 spots did not localise on the microtubule (n=14 cells) b, Model of mitochondrial dynamics mediated by microtubule dynamics. Microtubules polymerise and depolymerise at their plus ends ('+'). Absence of microtubule bundles in the cytoplasm during cell division enables the fragmentation of mitochondria. c, When mitochondria are bound to microtubules, Dnm1 ring assembly might be inhibited. Upon microtubule depolymerisation, this inhibition is alleviated and Dnm1 can effectively mediate scission of mitochondria.
We observed that Dnm1 spots were excluded from microtubules (Fig. 4a). We also propose the failure of Dnm1 ring assembly in the presence of microtubules as the nexus between microtubule dynamics and mitochondrial dynamics (Fig. 4c). Whether the microtubule is a simple physical impediment to the formation and activity of the Dnm1 ring or a more complicated chemical inhibitor would be the focus of future studies.
Destabilisation of microtubules is a hallmark of several neurodegenerative disorders including Alzheimer's, Parkinson's and Huntington's disease 26 . So too, colorectal cancer has been characterised by the absence of the tumour suppressor gene APC, which also modulates microtubule dynamics 27 . Remarkably, all these disease states have also been correlated with dysfunction of mitochondrial dynamics 28,29 . In future, it would be interesting to investigate the effect of restoration of microtubule dynamics on the mitochondrial fission and fusion in these scenarios, and consequently on the disease states.  Table S1) and Klp5Δ/Klp6Δ ('MT long mbc ', strain G3B, see Table S1)  Light grey crosses represent outliers, asterisk represents significance (p<0.05) and 'n.s.' indicates no significant difference. Note that the WT data represented in this figure is the identical to the wild-type data plotted in Fig. 2 and Extended Data Fig. 2 and has been re-used for comparison.  Table S1) with short (top) and long (bottom) mitotic spindles. The length of the spindle is indicated to the left of the images. b, Montage of maximum intensity projected confocal Z-stack images of the microtubules (top) and mitochondria (bottom) in a nonmitotic wild-type cell (strain KI001, see Table S1) represented in the intensity map indicated to the right of the images.. c, Bar plot of the mean number of mitochondria per cell before ('00:00') and after ('10:00') the same time window for which mitotic cells were monitored in Fig. 3c, d. Solid grey lines represent data from individual cells (n=7 cells). Figure 5. Dnm1 spots colocalise with the mitochondria and their number is inversely proportional to the length of microtubules. a, Maximum intensity projections of deconvolved Z-stack images of Dnm1 (left), microtubules (centre) and their composite (right) in wild-type ('WT', strain FY7143 transformed with pREP41-Dnm1-Cterm-GFP, see Table S1), Klp5Δ/Klp6Δ ('MT long ', strain FY20832 transformed with pREP41-Dnm1-Cterm-GFP, see Table S1) and Klp4Δ ('MT short ', strain McI438 transformed with pREP41-Dnm1-Cterm-GFP, see Table S1) cells. b, Box plot of the number of Dnm1 spots in WT, MT long and MT short cells (n=13, 14, and 18 cells respectively). Note that the mitochondrial morphology in WT, MT long and MT short cells appears abnormal due to inactivity of the GFP-tagged Dnm1. Light grey crosses represent outliers, asterisk represents significance (p<0.05) and 'n.s.' indicates no significant difference c, Measured parameters relating to mitochondrial numbers and morphology in WT, MT long and MT short interphase cells represented as mean ± s.e.m.

Table S1
Strains and constructs used in this study