Inside cells, structures called mitochondria supply the energy needed to carry out the processes that sustain life. Mitochondria constantly divide (a process known as fission) or fuse together, which helps to keep them in good working condition and well distributed around the cell. Several neurological disorders, including Parkinson’s disease and Alzheimer’s, are associated with problems that affect mitochondrial fission.

Many different molecules work together to help mitochondria divide, including a protein called Drp1. A number of Drp1 molecules can associate with each other to form an “oligomer” in the shape of a ring around a mitochondrion. The ring then constricts to split the mitochondrion in two.

It is often assumed that Drp1 molecules are recruited to the mitochondria immediately before fission and then form the oligomer ring. However, by using microscopy to track the movement of fluorescently labeled Drp1 molecules in human cells, Ji, Hatch et al. now suggest that Drp1 is continuously binding to and releasing from mitochondria, regardless of the need for fission. The experiments showed that when bound to surface of the mitochondrion, Drp1 switches between assembling and disassembling the oligomer ring. This process of Drp1 assembly and oligomerization on mitochondria is called maturation.

Specific signals for fission can push Drp1 toward maturation, which then leads to fission. Ji, Hatch et al. found that one such signal is the assembly of filaments of a protein called actin. Preventing actin filaments from forming reduced the amount of Drp1 that accumulated at mitochondria, and resulted in the mitochondria dividing less frequently. Further biochemical experiments also revealed that actin interacts directly with Drp1 and stimulates Drp1 activity, helping the ring to organize and assist mitochondrial fission.

The formation of actin filaments is not the only mechanism that can recruit Drp1 to mitochondria. Future work should investigate whether other mechanisms work with actin to recruit Drp1. As with actin filaments, other signals might be predicted to influence the balance of maturation and disassembly of Drp1 oligomers.

Mitochondrial fission and fusion create a dynamic network that is necessary for mitochondrial distribution and homeostasis. During mitosis, fission is necessary to distribute mitochondria between daughter cells. In neurons, fission and transport maintains mitochondrial distribution throughout the length of these highly polarized cells. Mitochondrial fission is also a fundamental step during mitophagy, a quality control mechanism that removes damaged mitochondrial segments. Dysregulated mitochondrial dynamics are linked to multiple neurological disorders, such as Alzheimer’s, Huntington’s, Parkinson’s, hereditary ataxia, and Charcot-Marie Tooth disease (Chen and Chan, 2009; Nunnari and Suomalainen, 2012; Girard et al., 2012; Niemann et al., 2005).

The dynamin-like GTPase Drp1 is a central component of the fission machinery, with Drp1 oligomerization and constriction providing a driving force for the process. Drp1 is a cytosolic protein that translocates to the outer mitochondrial membrane (OMM) (Smirnova et al., 2001). In solution, Drp1 exists in a number of oligomeric states, including dimers, tetramers, and higher-order oligomers (Fröhlich et al., 2013; Macdonald et al., 2014). Membrane-bound oligomeric Drp1 induces tubulation of associated membrane, and constricts the membrane in the presence of GTP (Francy et al., 2015; Mears et al., 2011). Drp1 can bind to the OMM through several factors, including the mitochondrial-specific lipid cardiolipin (Macdonald et al., 2014; Stepanyants et al., 2015; Bustillo-Zabalbeitia et al., 2014; Ugarte-Uribe et al., 2014); and several ‘receptors’ on the OMM, such as Mff, MiD49, and MiD51 (Richter et al., 2015).

What mechanisms target specific sites on mitochondria for fission? Recent findings have shed light on this question. The endoplasmic reticulum (ER) forms close contacts with mitochondria at the eventual fission site (Friedman et al., 2011). Actin polymerization at the ER-mitochondrial interface may be a key component, controlled by the ER-bound formin INF2 (Korobova et al., 2013) and the mitochondrially-bound Spire1C protein (Manor et al., 2015). Myosin II activity also plays a role in this pathway (DuBoff et al., 2012; Korobova et al., 2014). Inhibition of actin polymerization, myosin II activity, or INF2 causes decreased levels of mitochondrially-bound Drp1 oligomers, suggesting that actin and myosin are acting at the level of Drp1 recruitment. Two additional actin binding proteins, cortactin and cofilin, have also been linked to mitochondrial fission (Li et al., 2015).

Given the number of factors that can mediate Drp1 recruitment to the OMM, as well as the necessity for Drp1 oligomerization at the fission site, a fundamental question concerns the relationship between Drp1 recruitment, oligomerization, and fission. We envision two possible models for the process, termed de novo assembly and targeted equilibrium. In de novo assembly, cytosolic Drp1 is recruited specifically to fission sites, where it oligomerizes and mediates fission in a concerted series of steps in response to fission signals (which include OMM receptors, cardiolipin and actin). In targeted equilibrium, Drp1 maintains a constant equilibrium between cytosolic and mitochondrial pools, as well as between dimer and oligomer. These equilibria are independent of fission. Fission signals shift these equilibria toward stable oligomerization at targeted fission sites on the OMM.

While few studies address the order of events during Drp1 recruitment, de novo assembly is often tacitly assumed (Otera et al., 2013; Elgass et al., 2015; Labbé et al., 2014). Here, we show evidence supporting targeted equilibrium. We observe that mitochondrially-bound Drp1 oligomers mature into stable oligomers through a series of merging events. Some of these Drp1 oligomers are motile—capable of translocating on the mitochondrial surface before engaging in fission. Notably, not all mitochondrially-bound Drp1 oligomers engage in fission. Using ionomycin treatment to induce mitochondrial fission, we show that actin polymerization precedes Drp1 oligomerization, and that actin enriches with Drp1 at fission sites. Inhibition of actin polymerization, INF2 or myosin II decreases ionomycin-induced Drp1 accumulation and mitochondrial fission. Biochemically, actin filaments bind Drp1 directly and enhance its GTPase activity, suggesting that actin filaments can organize Drp1 into a productive oligomer. These results suggest that actin promotes fission by facilitating productive oligomerization of Drp1 on the OMM.

To monitor Drp1 dynamics in live cells, we developed a U2OS cell line stably expressing GFP-Drp1 with partial shRNA suppression of endogenous Drp1 (gDrp-U2OS). By quantitative western blotting, gDrp-U2OS cells maintain 1.64-fold total Drp1 levels compared to control U2OS cells, with 56% endogenous Drp1 and 44% GFP-Drp1 (Figure 1—figure supplement 1A,B). Mitochondrial morphology appears similar between control U2OS and gDrp-U2OS cells (not shown). We also developed a live-cell assay to measure mitochondrial fission rate, in which peripheral cellular regions are imaged over 10 min for dynamics of a fluorescent mitochondrial marker. Fission events are scored based on stable separation of the marker, and similar fission rates are obtained with either a matrix marker or an OMM marker (Figure 1—figure supplement 1C). Using this assay, the fission rates of control U2OS and gDrp1-U2OS cells are statistically similar (1.26 ± 1.01 versus 1.51 ± 0.69 fission events per mm mitochondrion per min, p = 0.49) (Figure 1—figure supplement 1D). Additionally, a second clone displaying undetectable endogenous Drp1 levels and GFP-Drp1 levels ~3.5-fold higher than control Drp1 levels displays similar mitochondrial fission frequency to control cells (Figure 1, Figure 1—figure supplement 1E–G), suggesting that GFP-Drp1 functionally compensates for endogenous Drp1 in these cells. We used the gDrp1-U2OS line for subsequent studies, due to its overall Drp1 levels being closer to those of control cells.

By live-cell confocal microscopy, much of the GFP-Drp1 in gDrp-U2OS cells appears diffuse in the cytoplasm (Figure 1—figure supplement 2A). However, adjusting the fluorescence threshold to eliminate this diffuse signal reveals abundant Drp1 ‘puncta’ of varying intensity (Figure 1A, Figure 1—figure supplement 2A,B), likely representing Drp1 oligomers. We arbitrarily defined two categories of puncta: ‘total’ puncta, including all Drp1 signal over the diffuse background; and ‘high-threshold’ puncta, representing the brightest 30% of Drp1 puncta. These designations do not represent two distinct pools of intensity, as the distribution of punctum intensity is continuous (Figure 1—figure supplement 2B). Here, we use these categories to distinguish between smaller puncta and those approaching maximal size.

Figure 1 with 2 supplements see all

Download asset Open asset

We employed a particle-tracking algorithm to analyze punctum position and lifetime relative to a mitochondrial matrix marker (Figure 1C). Approximately 49% of total puncta and 70% of high-threshold puncta co-localize with mitochondria at individual frames in the viewing period (Figure 1B). For most puncta, mitochondrial association is transient (Video 1), with mean lifetimes of 45.0 ± 104.8 s and 64.4 ± 113.1 s for total and high-threshold puncta, respectively (3582 and 1601 puncta analyzed). The largest population of puncta remains bound < 12 s in both cases (Figure 1C), although the high-threshold puncta are significantly more stably associated with mitochondria (p<0.001 for lifetime comparison). Puncta density on mitochondria does not vary systematically with mitochondrial length, with approximately 0.41 total puncta and 0.17 high-threshold puncta per μm mitochondrial length (Figure 1—figure supplement 1H).

Video 1

Download asset

How do Drp1 puncta assemble on mitochondria? We observed three distinct dynamic behaviors associated with mitochondrially-bound Drp1 puncta: punctum merging, punctum motility, and mitochondrial fission. Mitochondrially-bound puncta frequently merge together, with successive merging events resulting in punctum growth (Figure 2A,B, Figure 2—figure supplement 1A, Video 2,3). Merging is a reversible process, and can be followed by splitting of the merged punctum (Figure 2—figure supplement 1B, Video 4). The combination of these results suggests that a population of Drp1 is in rapid equilibrium between the cytosol and the OMM, and that one mechanism for assembly of larger Drp1 oligomers is through merging of mitochondrially-bound oligomers. We refer to this process as ‘maturation’. Imaging at higher resolution by Airyscan microscopy suggests that maturing puncta are loosely associated with the mitochondrion, and progressively encircle the mitochondrion as they mature (Figure 2C, Video 5).

Figure 2 with 1 supplement see all

Download asset Open asset

Video 2

Download asset

Video 3

Download asset

Video 4

Download asset

Video 5

Download asset

Low-intensity puncta are motile along the OMM, allowing them to interact during merging events (Figure 2A,B, Figure 2—figure supplement 1A,B). In addition, a sub-set of high-threshold Drp1 puncta translocates directionally along mitochondria for short distances at a mean velocity of 46.9 ± 12.8 nm/sec (Figure 3A,B). Assessing the fraction of motile puncta is challenging due to the underlying motility of the mitochondrion itself, thus our analysis is limited to situations where mitochondria are visibly stationary during the viewing period. Motile high-threshold puncta are capable of stopping abruptly and engaging in fission (Figure 3A, Video 6).

Figure 3 with 1 supplement see all

Download asset Open asset

Video 6

Download asset

We next assessed the correlation between Drp1 puncta and mitochondrial fission, and found that all observed fission events are associated with high-threshold Drp1 puncta (49 events, examples in Figure 1A, Figure 2A, Figure 2—figure supplement 1A, Figure 3A). However, only a low percentage of Drp1 puncta (2.5% of total puncta and 6.3% of high threshold puncta) engages in observable fission events during the 10 min imaging period. Even for productive puncta, the lag between punctum establishment and mitochondrial fission varies significantly, with a mean of 76.0 ± 40.3 s (Figure 3D) that appears to be independent of mitochondrial length (Figure 3E). These results suggest that establishment of an apparently stable mitochondrially-bound Drp1 oligomer is not sufficient, and that additional steps are required to drive fission.

We made two additional observations concerning mitochondrial fission. First, mitochondria undergoing fission tend to be longer than the mean (Figure 3—figure supplement 1A). Second, Drp1 remains at newly created mitochondrial ends for a significant time after fission. This residual Drp1 can be asymmetrically distributed, with one end inheriting most or all residual Drp1 (Figure 3—figure supplement 1B–D).

We examined Drp1 puncta at higher spatial resolution by live-cell 3D-structured illumination microscopy (3D-SIM), using Tom20-mCherry as the mitochondrial marker in order to label the OMM. We analyzed diameters of both mitochondria and associated Drp1 puncta for ‘productive’ events (leading to fission) and non-productive events (sites of stable Drp1 puncta that did not undergo fission during observation). The mean mitochondrial diameter in the absence of Drp1 is 322 ± 64 nm (Figure 4B). Non-productive high-threshold Drp1 puncta still cause significant constriction (Figure 4—figure supplement 1A), with mean mitochondrial diameter of 215 ± 50 nm (Figure 4B) and Drp1 diameter of 270 ± 38 nm (Figure 4C) at these sites. This constriction is maintained over time (Figure 4D). Productive Drp1 puncta undergo significant contraction immediately prior to fission, with mean Drp1 diameter of 200 ± 31 nm in the frame prior to fission (Figure 4C,D, Figure 4—figure supplement 1B, Video 7,8). The mitochondrial diameter in this frame is 134 ± 25 nm, which is at the limit of resolution for 3D-SIM. Our results are qualitatively similar to those obtained from fixed-cell PALM studies (Rosenbloom et al., 2014), while adding the dynamic component of contraction prior to fission.

Figure 4 with 1 supplement see all

Download asset Open asset

Video 7

Download asset

Video 8

Download asset

We also observed motile puncta by 3D-SIM. These puncta appear to distort as they translocate, periodically separating into strands (Figure 5A, Figure 5—figure supplement 1A, Video 9,10). Dynamic shape changes are not characteristic of stable puncta, which can be most easily appreciated when viewing a punctum transitioning from motile to stationary (Figure 5B). Motile puncta appear to encircle the mitochondrion to a significant degree, although the spatial resolution achieved here is not sufficient to answer this question definitively. Punctum velocity in 3D-SIM is similar to that measured by confocal microscopy (48.5 ± 17.1 nm/sec, Figure 5—figure supplement 1B).

Figure 5 with 1 supplement see all

Download asset Open asset

Video 9

Download asset

Video 10

Download asset

What signal might stimulate Drp1 punctum maturation? We previously showed that mitochondrial length increased upon three cellular treatments: inhibition of actin polymerization by Latrunculin A (LatA), siRNA suppression of the formin protein INF2, or inhibition of myosin II by chemical inhibitors or siRNA (Korobova et al., 2013; Korobova et al., 2014). The increase in mitochondrial length suggests a fission defect, but could also reflect changes in mitochondrial fusion. Here, we establish that the effect is on mitochondrial fission, using our live-cell fission assay. Suppression of either INF2 or myosin IIA causes an ~50% decrease in fission rate, whereas Drp1 suppression or expression of the dominant-negative Drp1 K38A mutant cause near-complete inhibition (Figure 6—figure supplement 1A,B). In addition, suppression of either INF2 or myosin IIA reduces mitochondrially-associated Drp1 puncta, with a moderate reduction of total Drp1 puncta (2.7-fold for both INF2 and myosin IIA) and a larger reduction in high threshold puncta (6.7-fold and 8.3-fold, respectively, Figure 6—figure supplement 1C). LatA treatment also causes a reduction in mitochondrially-associated Drp1 puncta, although the degree of high threshold reduction is not as great as for INF2 or myosin IIA suppression (Figure 6—figure supplement 1D). These results support the previous results suggesting a role for actin polymerization in mitochondrial Drp1 assembly.

We next examined the relationship between actin filaments and Drp1 during mitochondrial fission by live-cell microscopy in gDrp-U2OS cells. One challenge is the abundance of actin-based structures in U2OS cells. To mitigate this issue, we imaged cells at a focal plane significantly above the ventral surface, where there are fewer actin-based structures such as stress fibers. In this region, we observed frequent examples of actin filaments accumulating prior to fission (Figure 6—figure supplement 2A, Video 11), but the overall actin filament abundance in the region remained a challenge for determining whether its presence might contribute to fission. To assess the significance of the actin accumulation, we calculated the mean percentage of mitochondrial area covered by actin (Figure 6—figure supplement 2B), and compared this number to the percentage of fission events at which actin was present immediately prior to fission. We found that a significantly higher percentage of fission events (56%) were associated with actin filaments than would be expected by random correlation (28%, n = 59 fission events, Figure 6— figure supplement 2C).

Video 11

Download asset

While this correlation is suggestive, the low frequency of fission events is challenging for temporal correlation of actin and Drp1 accumulation prior to fission. To examine this relationship in more detail, we used ionomycin, a calcium ionophore, to increase mitochondrial fission rate (Tan et al., 2011; Sanmartin et al., 2014). A recent publication showed that 3T3 cells respond to increased cytosolic calcium with an acute burst of actin polymerization throughout the cytosol, mediated by INF2 (Shao et al., 2015). We find a similar ionomycin-induced actin burst in gDrp-U2OS cells (16.9 ± 8.7% signal increase, t_1/2 28.1 ± 19.8 s) followed by a rapid decline (Figure 6A). Ionomycin treatment also increases mitochondrial fission rate 3.9-fold (2.61 ± 1.01 vs 10.23 ± 5.41 events mm^-1 min^-1) (Figure 6B). Similar fission rate increases occur when using mitochondrial matrix markers tagged with either dsRed or BFP (Figure 6—figure supplement 3).

Figure 6 with 4 supplements see all

Download asset Open asset

We next asked whether the quantity of Drp1 oligomers also increased following ionomycin treatment, initially monitoring total Drp1 puncta. Indeed, ionomycin treatment causes Drp1 oligomer signal increases from 10-20%, depending on the experiment, with a consistent t_1/2 of 99.9 ± 19.0 s (Figure 6A, also see Figure 7A,C and D). In contrast to the transient actin increase, the Drp1 oligomer signal remains elevated for >10 min. These measurements are rapid to conduct, allowing analysis of every time point in the sequence, but do not provide information on the change in mitochondrially-bound Drp1 puncta. For this reason, we also quantified the change in the number of mitochondrially-associated Drp1 puncta upon ionomycin treatment on a more limited set of time points, and found ~40% increase for low- or high-threshold puncta (Figure 6C).

Figure 7

Download asset Open asset

In addition, we visually monitored the dynamics of actin and Drp1 at individual ionomycin-induced fission sites by Airyscan microscopy. Prior to ionomycin treatment, a number of low intensity Drp1 puncta dynamically associate with mitochondria, and a meshwork of actin exists throughout the cytoplasm (Figure 6D, Figure 6—figure supplement 4, Video 12). This actin meshwork enriches at certain positions on mitochondria, and Drp1 puncta tend to be more abundant at these sites but are dynamic. Upon ionomycin treatment, the actin signal intensifies throughout the cytosol, particularly at sites of mitochondrial crossover. Drp1 puncta mature at these actin-enriched sites, and some of these sites undergo fission while others undergo constriction but no fission during the viewing period (Figure 6D, Figure 6—figure supplement 4, Video 12).

Video 12

Download asset

We tested the association between actin and Drp1 further by inhibiting actin polymerization, followed by ionomycin stimulation. Pre-treatment for 15 min with LatA abolishes both Drp1 accumulation (Figure 7A,B) and the increase in mitochondrial fission (Figure 6B). Suppression of INF2 by siRNA treatment potently inhibits the actin increase, but causes a smaller decrease in whole-cell Drp1 oligomer accumulation than that achieved by LatA (Figure 7C). However, the effect of INF2 suppression on mitochondrially-associated high threshold Drp1 puncta is dramatic, with almost complete elimination of the ionomycin-induced increase (Figure 6C Right). INF2 suppression also causes a decrease in ionomycin-induced mitochondrial fission (Figure 6B). Interestingly, INF2 suppression has a negligible effect on Drp1 puncta of lower intensity (Figure 6C Left). These results suggest that INF2-mediated actin polymerization plays a role in Drp1 maturation on mitochondria.

We next tested the role of myosin II in this process. Suppression of myosin IIA does not alter the actin response to ionomycin (signal intensity increase 20.0 ± 6.4% with t_1/2 of 22.8 ± 12.7 s), but strongly reduces Drp1 oligomerization (Figure 7D) and mitochondrial fission (Figure 6B). These experiments suggest that both INF2-induced actin polymerization and myosin II activity are important pre-requisites for productive Drp1 oligomer assembly.

The mechanism by which actin might stimulate Drp1 accumulation is unclear. One possibility is that Drp1 binds directly to actin filaments at the fission site, similar to the direct binding between dynamin and actin (Gu et al., 2010; Palmer et al., 2015; Mooren et al., 2009). Interestingly, a previous study demonstrates that actin accumulates at clathrin-mediated endocytosis sites prior to dynamin 2 arrival (Grassart et al., 2014). We tested the possibility of a direct interaction using purified proteins. Through co-sedimentation assay, Drp1 binds actin with an apparent K_d of 0.8 ± 0.4 μM (Figure 8A, Figure 8— figure supplement 1A). Interestingly, less than 50% of the Drp1 binds actin filaments even at saturating actin concentrations (Figure 8A). This situation is not caused by a significant pool of ‘inactive’ Drp1 in our purified preparation, since >90% of the Drp1 pellets when incubated with a non-hydrolyzable GTP analogue (Figure 8—figure supplement 1B). We postulate that partial binding is due to differential affinities of Drp1 oligomeric states for actin filaments. In these assays, Drp1 does not change the critical concentration of actin, since actin polymerizes to a similar extent in the absence or presence of Drp1 (Figure 8—figure supplement 1C).

Figure 8 with 1 supplement see all

Download asset Open asset

By total internal reflection (TIRF) microscopy, GFP-tagged Drp1 binds evenly throughout the actin filament (Figure 8B), with no apparent preference for specific filament regions (Figure 8C, Video 13). By negative-stain electron microscopy, Drp1-bound filaments have mean widths of 27.2 ± 1.3 nm, compared to 8.9 ± 0.2 nm for actin alone (Figure 8D). It is unclear whether this increased width is due to Drp1 binding a single filament or to Drp1-mediated filament bundling. The width of a Drp1 dimer is 22.9 nm (Fröhlich et al., 2013), similar to our measured width. However, Drp1 can induce filament bundling, as evidenced by TIRF microscopy (Figure 8E, Video 14). Indeed, time-lapse imaging of the bundling process suggests that filaments have a preferred bundled orientation, and can be end-bound to the sides of Drp1-bound filaments prior to bundling (Figure 8F, Video 15).

Video 13

Download asset

Video 14

Download asset

Video 15

Download asset

Drp1 binding to anionic lipids such as cardiolipin increases its GTPase activity (Macdonald et al., 2014), presumably by placing its GTPase domains in close proximity as is found for other dynamin family proteins (Bui and Shaw, 2013). We tested whether actin filaments could also increase Drp1’s GTPase activity and found a ~3.5-fold increase (Figure 9A). An additional question is whether actin filaments can synergize with Drp1 receptors on the OMM. We tested this possibility using the cytosolic portion of Mff. Mff alone causes only a slight increase in Drp1 GTPase activity at the concentrations tested. However, the combination of Mff and actin filaments causes a substantial increase in Drp1 activity, far beyond the additive effects of either Mff or actin alone (Figure 9B). These results show that Drp1 binds actin filaments in a manner that stimulates its catalytic activity, and that actin filaments can synergize with Mff to increase productive Drp1 oligomerization.

Figure 9

Download asset Open asset

Overall, our results support several important and novel features of Drp1 oligomerization. Rather than being an ‘all or none’ process in which Drp1 oligomerization at the fission site is induced de novo by fission signals, we postulate that both Drp1 oligomerization and mitochondrial association are in constant equilibrium even in the absence of fission signals. Fission signals serve to ‘target’ this equilibrium to fission sites. Actin filaments are one such fission signal.

We illustrate these features as a three-stage model for fission-productive Drp1: recruitment, maturation, and conversion (Figure 10). In recruitment, ‘units’ of Drp1 are in rapid equilibrium between cytosolic and mitochondrially-bound pools. These Drp1 units diffuse on the OMM, and merge to form larger units. Maturation is the progressive association of mitochondrially-bound Drp1 units, leading to assembly of an oligomer that fully encircles the mitochondrion. The size of the mature oligomer is unclear, but might be similar to the 26-40 subunits found for dynamin2 (Grassart et al., 2014; Cocucci et al., 2014). Mature oligomers are capable of movement along the OMM. Conversion represents a distinct process in which a site containing a stable Drp1 oligomer is rendered fission-competent.

Figure 10

Download asset Open asset

This model raises many new testable questions. First, what is the complete list of “fission signals”? There is evidence that OMM proteins such as Mff, MiD49/51, and Fis1, as well as cardiolipin exposure on the OMM, can serve as fission signals. We postulate that actin filaments in proximity to the mitochondrion are another such signal. Other potential factors that might influence fission could include mitochondrial length and mitochondrial branches. We find that fission events occur disproportionately on longer mitochondria, but we do not find an increase in Drp1 puncta density on longer mitochondria, suggesting that maturation is not favored on long mitochondria. It should be realized, however, that most mitochondria exist in a branched network in these cells. The relationship between mitochondrial branching and mitochondrial fission is a fascinating question still to be addressed.

Second, do the various fission signals act in series or in parallel? In other words, do the signals work together within the same fission pathway, or act in alternate pathways? Both options could be true, depending on the fission signal. For example, we postulate that actin filaments and cardiolipin might represent alternate maturation factors, since both are polyvalent anions. In contrast, actin filaments could work in concert with Mff, either in “coincidence detection” by Drp1 (Hatch et al., 2014) or by actin filaments serving as a reservoir for delivery of Drp1 oligomers to Mff. Recent work has shown that Mff binds preferentially to oligomerized Drp1 (Liu and Chan, 2015), which is consistent with this possibility. We show here that actin filaments and Mff synergize in stimulating Drp1’s GTPase activity. A final possibility is that specific fission factors mediate distinct steps in the process. For example, a recent study on the membrane-altering ability of the Drp1-cardiolipin interaction might suggest that cardiolipin triggers a conversion step (Stepanyants et al., 2015).

Third, what is the size of the Drp1 ‘unit’ that is recruited to mitochondria? We cannot detect the smallest oligomers (dimers, tetramers) in this study for several reasons, including: our threshold procedure forcibly removes these from analysis, and only ~ half of the Drp1 in these cells is GFP-labeled. The available data suggest that the recruited unit may be mixed oligomers of GFP-labeled and endogenous Drp1. Recent biochemical studies have shown that Drp1 exists in equilibrium between oligomeric states in solution, with some evidence that dimers are the relevant membrane-recruited unit in the case of cardiolipin binding (Fröhlich et al., 2013; Macdonald et al., 2014). On the other hand, in the present study we observe clear instances of Drp1 puncta apparently translocating from cytosol to mitochondrion. We state ‘apparently’ because it is possible that these oligomers transfer from another membrane or a cytoskeletal structure. Based on these findings, we postulate that a range of oligomeric species can be recruited to mitochondria, but that their on- and off-rates might vary considerably.

Fourth, what steps are stimulated by actin? Our data show that inhibiting actin polymerization inhibits the accumulation of Drp1 oligomers, suggestive of roles in either recruitment or maturation. Direct Drp1 binding to actin filaments near mitochondria might enhance initial mitochondrial binding, or might accelerate oligomerization steps of cytosolic Drp1 prior to mitochondrial interaction, both of which would aid recruitment. Alternately, actin binding to mitochondrially-bound Drp1 might enhance oligomerization on the mitochondrial surface, which would represent maturation. The two possibilities are not mutually exclusive. A related question is: what is myosin II’s role in the process? One possibility is that myosin II activity might lead to Drp1-independent ‘pre-constriction’ of the OMM, as we have proposed previously (Korobova et al., 2014; Hatch et al., 2014). It is also possible that myosin II organizes the INF2-assembled actin filaments in a manner optimal for Drp1 recruitment and/or maturation.

Fifth, what are the roles of other actin binding proteins that have been identified as contributing to mitochondrial fission, such as cortactin, cofilin, and Spire 1C (Manor et al., 2015; Li et al., 2015)? There is evidence that mitochondrially-bound Spire 1C and ER-bound INF2 work through the same pathway, suggesting that Spire 1C might serve as a nucleation factor and INF2 as an elongation factor and/or a severing protein. Spire proteins cooperate with other formin proteins in a similar manner (Vizcarra et al., 2011; Quinlan, 2013). Roles for cortactin and cofilin in the same pathway as Spire 1C/INF2 are possible, but an alternative is that they contribute to a distinct actin-dependent pathway. The fact that steady-state fission rate is only partially reduced by INF2 or myosin IIA suppression (Figure 6—figure supplement 1) suggests some degree of functional redundancy.

Sixth, does the conversion step represent a change in Drp1 structure/activity, or changes in other fission components? One possibility is recruitment of additional necessary molecules. For example, recent publications show that endophilin co-operates with actin and dynamin in clathrin-independent endocytosis (Boucrot et al., 2015; Renard et al., 2015). Alternately, exposure of cardiolipin on the OMM could trigger a conversion step, as mentioned above (Macdonald et al., 2014).

Seventh, how does ionomycin stimulate mitochondrial fission? We postulate that, in addition to the known effect of calcium on Drp1 phosphorylation state (Cribbs and Strack, 2007), increased cytosolic calcium activates INF2, as demonstrated in recent work (Shao et al., 2015) (R. Wedlich-Söldner, personal communication). These effects are likely to be similar to those induced by the Listeria protein LLO (Stavru et al., 2013), which is also a calcium ionophore.

Eighth, what is the role of the ER in mitochondrial fission? The tight association of ER with mitochondria at fission sites (Friedman et al., 2011) could contribute in two ways. First, ER supplies the INF2 isoform responsible for actin dynamics at the fission site (Korobova et al., 2013; Manor et al., 2015). Second, ER might contribute to the calcium dynamics necessary for ionomycin-mediated fission, considering the complex relationship between extracellular calcium entry, calcium release from ER, and ER-mitochondrial calcium communication (Horne and Meyer, 1997; Csordás et al., 1999).

Finally, what is the mechanism of Drp1 motility on mitochondria? Our 3D-SIM time-lapse movies suggest that motile Drp1 oligomers are rings that encircle the mitochondrion, and that these rings change significantly during the motility process, in a manner similar to ‘walking’ along the mitochondrial surface. Another possible mechanism could be ‘treadmilling’ of Drp1 units, adding to one side of the Drp1 oligomer while dissociating from the other side, akin to cytoskeletal polymers. Furthermore, what is the purpose of this motility? One possibility is something we call the ‘night watchman’ hypothesis, in which motility allows mature Drp1 oligomers to ‘patrol’ the mitochondrion in search of fission signals.

1. 1. Boucrot E
   2. Ferreira APA
   3. Almeida-Souza L
   4. Debard S
   5. Vallis Y
   6. Howard G
   7. Bertot L
   8. Sauvonnet N
   9. McMahon HT

   (2015) Endophilin marks and controls a clathrin-independent endocytic pathway

   Nature 517:460–465.

   https://doi.org/10.1038/nature14067

   • Google Scholar
2. 1. Bui HT
   2. Shaw JM

   (2013) Dynamin assembly strategies and adaptor proteins in mitochondrial fission

   Current Biology 23:R891–R899.

   https://doi.org/10.1016/j.cub.2013.08.040

   • Google Scholar
3. 1. Bustillo-Zabalbeitia I
   2. Montessuit S
   3. Raemy E
   4. Basañez G
   5. Terrones O
   6. Martinou J-C
   7. Strack S

   (2014) Specific interaction with cardiolipin triggers functional activation of dynamin-related protein 1

   PLoS ONE 9:e102738.

   https://doi.org/10.1371/journal.pone.0102738

   • Google Scholar
4. 1. Chen H
   2. Chan DC

   (2009) Mitochondrial dynamics-fusion, fission, movement, and mitophagy-in neurodegenerative diseases

   Human Molecular Genetics 18:R169–R176.

   https://doi.org/10.1093/hmg/ddp326

   • Google Scholar
5. 1. Cocucci E
   2. Gaudin R
   3. Kirchhausen T

   (2014) Dynamin recruitment and membrane scission at the neck of a clathrin-coated pit

   Molecular Biology of the Cell 25:3595–3609.

   https://doi.org/10.1091/mbc.E14-07-1240

   • Google Scholar
6. 1. Cribbs JT
   2. Strack S

   (2007) Reversible phosphorylation of Drp1 by cyclic AMP-dependent protein kinase and calcineurin regulates mitochondrial fission and cell death

   EMBO Reports 8:939–944.

   https://doi.org/10.1038/sj.embor.7401062

   • Google Scholar
7. 1. Csordás G
   2. Thomas AP
   3. Hajnóczky G

   (1999) Quasi-synaptic calcium signal transmission between endoplasmic reticulum and mitochondria

   The EMBO Journal 18:96–108.

   https://doi.org/10.1093/emboj/18.1.96

   • Google Scholar
8. 1. Dagda RK
   2. Cherra SJ
   3. Kulich SM
   4. Tandon A
   5. Park D
   6. Chu CT

   (2009) Loss of PINK1 function promotes mitophagy through effects on oxidative stress and mitochondrial fission

   Journal of Biological Chemistry 284:13843–13855.

   https://doi.org/10.1074/jbc.M808515200

   • Google Scholar
9. 1. DuBoff B
   2. Götz J
   3. Feany MB

   (2012) Tau promotes neurodegeneration via DRP1 mislocalization in vivo

   Neuron 75:618–632.

   https://doi.org/10.1016/j.neuron.2012.06.026

   • Google Scholar
10. 1. Elgass KD
    2. Smith EA
    3. LeGros MA
    4. Larabell CA
    5. Ryan MT

    (2015) Analysis of ER-mitochondria contacts using correlative fluorescence microscopy and soft x-ray tomography of mammalian cells

    Journal of Cell Science 128:2795–2804.

    https://doi.org/10.1242/jcs.169136

    • Google Scholar
11. 1. Francy CA
    2. Alvarez FJD
    3. Zhou L
    4. Ramachandran R
    5. Mears JA

    (2015) The mechanoenzymatic core of dynamin-related protein 1 comprises the minimal machinery required for membrane constriction

    Journal of Biological Chemistry 290:11692–11703.

    https://doi.org/10.1074/jbc.M114.610881

    • Google Scholar
12. 1. Friedman JR
    2. Lackner LL
    3. West M
    4. DiBenedetto JR
    5. Nunnari J
    6. Voeltz GK

    (2011) ER tubules mark sites of mitochondrial division

    Science 334:358–362.

    https://doi.org/10.1126/science.1207385

    • Google Scholar
13. 1. Fröhlich C
    2. Grabiger S
    3. Schwefel D
    4. Faelber K
    5. Rosenbaum E
    6. Mears J
    7. Rocks O
    8. Daumke O

    (2013) Structural insights into oligomerization and mitochondrial remodelling of dynamin 1-like protein

    The EMBO Journal 32:1280–1292.

    https://doi.org/10.1038/emboj.2013.74

    • Google Scholar
14. 1. Gandre-Babbe S
    2. van der Bliek AM

    (2008) The novel tail-anchored membrane protein mff controls mitochondrial and peroxisomal fission in mammalian cells

    Molecular Biology of the Cell 19:2402–2412.

    https://doi.org/10.1091/mbc.E07-12-1287

    • Google Scholar
15. 1. Girard M
    2. Lariviere R
    3. Parfitt DA
    4. Deane EC
    5. Gaudet R
    6. Nossova N
    7. Blondeau F
    8. Prenosil G
    9. Vermeulen EGM
    10. Duchen MR
    11. Richter A
    12. Shoubridge EA
    13. Gehring K
    14. McKinney RA
    15. Brais B
    16. Chapple JP
    17. McPherson PS

    (2012) Mitochondrial dysfunction and purkinje cell loss in autosomal recessive spastic ataxia of charlevoix-saguenay (aRSACS)

    Proceedings of the National Academy of Sciences of the United States of America 109:1661–1666.

    https://doi.org/10.1073/pnas.1113166109

    • Google Scholar
16. 1. Grassart A
    2. Cheng AT
    3. Hong SH
    4. Zhang F
    5. Zenzer N
    6. Feng Y
    7. Briner DM
    8. Davis GD
    9. Malkov D
    10. Drubin DG

    (2014) Actin and dynamin2 dynamics and interplay during clathrin-mediated endocytosis

    The Journal of Cell Biology 205:721–735.

    https://doi.org/10.1083/jcb.201403041

    • Google Scholar
17. 1. Gu C
    2. Yaddanapudi S
    3. Weins A
    4. Osborn T
    5. Reiser J
    6. Pollak M
    7. Hartwig J
    8. Sever S

    (2010) Direct dynamin–actin interactions regulate the actin cytoskeleton

    The EMBO Journal 29:3593–3606.

    https://doi.org/10.1038/emboj.2010.249

    • Google Scholar
18. 1. Gurel PS
    2. Ge P
    3. Grintsevich EE
    4. Shu R
    5. Blanchoin L
    6. Zhou ZH
    7. Reisler E
    8. Higgs HN

    (2014) INF2-mediated severing through actin filament encirclement and disruption

    Current Biology 24:156–164.

    https://doi.org/10.1016/j.cub.2013.12.018

    • Google Scholar
19. 1. Hatch AL
    2. Gurel PS
    3. Higgs HN

    (2014) Novel roles for actin in mitochondrial fission

    Journal of Cell Science 127:4549–4560.

    https://doi.org/10.1242/jcs.153791

    • Google Scholar
20. 1. Horne JH
    2. Meyer T

    (1997) Elementary calcium-release units induced by inositol trisphosphate

    Science 276:1690–1693.

    https://doi.org/10.1126/science.276.5319.1690

    • Google Scholar
21. 1. Johnson HW
    2. Schell MJ

    (2009) Neuronal IP3 3-kinase is an f-actin-bundling protein: role in dendritic targeting and regulation of spine morphology

    Molecular Biology of the Cell 20:5166–5180.

    https://doi.org/10.1091/mbc.E09-01-0083

    • Google Scholar
22. 1. Koirala S
    2. Guo Q
    3. Kalia R
    4. Bui HT
    5. Eckert DM
    6. Frost A
    7. Shaw JM

    (2013) Interchangeable adaptors regulate mitochondrial dynamin assembly for membrane scission

    Proceedings of the National Academy of Sciences of the United States of America 110:E1342.

    https://doi.org/10.1073/pnas.1300855110

    • Google Scholar
23. 1. Korobova F
    2. Gauvin TJ
    3. Higgs HN

    (2014) A role for myosin II in mammalian mitochondrial fission

    Current Biology 24:409–414.

    https://doi.org/10.1016/j.cub.2013.12.032

    • Google Scholar
24. 1. Korobova F
    2. Ramabhadran V
    3. Higgs HN

    (2013) An actin-dependent step in mitochondrial fission mediated by the ER-associated formin INF2

    Science 339:464–467.

    https://doi.org/10.1126/science.1228360

    • Google Scholar
25. 1. Labbé K
    2. Murley A
    3. Nunnari J

    (2014) Determinants and functions of mitochondrial behavior

    Annual Review of Cell and Developmental Biology 30:357–391.

    https://doi.org/10.1146/annurev-cellbio-101011-155756

    • Google Scholar
26. 1. Li S
    2. Xu S
    3. Roelofs BA
    4. Boyman L
    5. Lederer WJ
    6. Sesaki H
    7. Karbowski M

    (2015) Transient assembly of f-actin on the outer mitochondrial membrane contributes to mitochondrial fission

    The Journal of Cell Biology 208:109–123.

    https://doi.org/10.1083/jcb.201404050

    • Google Scholar
27. 1. Liu R
    2. Chan DC

    (2015) The mitochondrial fission receptor mff selectively recruits oligomerized Drp1

    Molecular Biology of the Cell 26:4466–4477.

    https://doi.org/10.1091/mbc.E15-08-0591

    • Google Scholar
28. 1. Macdonald PJ
    2. Stepanyants N
    3. Mehrotra N
    4. Mears JA
    5. Qi X
    6. Sesaki H
    7. Ramachandran R

    (2014) A dimeric equilibrium intermediate nucleates Drp1 reassembly on mitochondrial membranes for fission

    Molecular Biology of the Cell 25:1905–1915.

    https://doi.org/10.1091/mbc.E14-02-0728

    • Google Scholar
29. 1. Manor U
    2. Bartholomew S
    3. Golani G
    4. Christenson E
    5. Kozlov M
    6. Higgs H
    7. Spudich J
    8. Lippincott-Schwartz J

    (2015) A mitochondria-anchored isoform of the actin-nucleating spire protein regulates mitochondrial division

    eLife 4:.

    https://doi.org/10.7554/eLife.08828

    • Google Scholar
30. 1. Mears JA
    2. Lackner LL
    3. Fang S
    4. Ingerman E
    5. Nunnari J
    6. Hinshaw JE

    (2011) Conformational changes in Dnm1 support a contractile mechanism for mitochondrial fission

    Nature Structural & Molecular Biology 18:20–26.

    https://doi.org/10.1038/nsmb.1949

    • Google Scholar
31. 1. Mooren OL
    2. Kotova TI
    3. Moore AJ
    4. Schafer DA

    (2009) Dynamin2 GTPase and cortactin remodel actin filaments

    Journal of Biological Chemistry 284:23995–24005.

    https://doi.org/10.1074/jbc.M109.024398

    • Google Scholar
32. 1. Niemann A
    2. Ruegg M
    3. La Padula V
    4. Schenone A
    5. Suter U

    (2005) Ganglioside-induced differentiation associated protein 1 is a regulator of the mitochondrial network: new implications for charcot-marie-tooth disease

    The Journal of Cell Biology 170:1067–1078.

    https://doi.org/10.1083/jcb.200507087

    • Google Scholar
33. 1. Nunnari J
    2. Suomalainen A

    (2012) Mitochondria: in sickness and in health

    Cell 148:1145–1159.

    https://doi.org/10.1016/j.cell.2012.02.035

    • Google Scholar
34. 1. Otera H
    2. Ishihara N
    3. Mihara K

    (2013) New insights into the function and regulation of mitochondrial fission

    Biochimica Et Biophysica Acta 1833:1256–1268.

    https://doi.org/10.1016/j.bbamcr.2013.02.002

    • Google Scholar
35. 1. Palmer SE
    2. Smaczynska-de Rooij II
    3. Marklew CJ
    4. Allwood EG
    5. Mishra R
    6. Johnson S
    7. Goldberg MW
    8. Ayscough KR

    (2015) A dynamin-actin interaction is required for vesicle scission during endocytosis in yeast

    Current Biology 25:868–878.

    https://doi.org/10.1016/j.cub.2015.01.061

    • Google Scholar
36. 1. Quinlan ME

    (2013) Direct interaction between two actin nucleators is required in drosophila oogenesis

    Development 140:4417–4425.

    https://doi.org/10.1242/dev.097337

    • Google Scholar
37. 1. Renard H-F
    2. Simunovic M
    3. Lemière J
    4. Boucrot E
    5. Garcia-Castillo MD
    6. Arumugam S
    7. Chambon V
    8. Lamaze C
    9. Wunder C
    10. Kenworthy AK
    11. Schmidt AA
    12. McMahon HT
    13. Sykes C
    14. Bassereau P
    15. Johannes L

    (2015) Endophilin-A2 functions in membrane scission in clathrin-independent endocytosis

    Nature 517:493–496.

    https://doi.org/10.1038/nature14064

    • Google Scholar
38. 1. Richter V
    2. Singh AP
    3. Kvansakul M
    4. Ryan MT
    5. Osellame LD

    (2015) Splitting up the powerhouse: structural insights into the mechanism of mitochondrial fission

    Cellular and Molecular Life Sciences 72:3695–3707.

    https://doi.org/10.1007/s00018-015-1950-y

    • Google Scholar
39. 1. Rosenbloom AB
    2. Lee S-H
    3. To M
    4. Lee A
    5. Shin JY
    6. Bustamante C

    (2014) Optimized two-color super resolution imaging of Drp1 during mitochondrial fission with a slow-switching dronpa variant

    Proceedings of the National Academy of Sciences of the United States of America 111:13093–13098.

    https://doi.org/10.1073/pnas.1320044111

    • Google Scholar
40. 1. Sanmarti­n CD
    2. Paula-Lima AC
    3. Garci­a A
    4. Barattini P
    5. Hartel S
    6. Núñez MT
    7. Hidalgo C

    (2014) Ryanodine receptor-mediated Ca2+ release underlies iron-induced mitochondrial fission and stimulates mitochondrial Ca2+ uptake in primary hippocampal neurons

    Frontiers in Molecular Neuroscience 7:.

    https://doi.org/10.3389/fnmol.2014.00013

    • Google Scholar
41. 1. Shao X
    2. Li Q
    3. Mogilner A
    4. Bershadsky AD
    5. Shivashankar GV

    (2015) Mechanical stimulation induces formin-dependent assembly of a perinuclear actin rim

    Proceedings of the National Academy of Sciences of the United States of America 112:E2595.

    https://doi.org/10.1073/pnas.1504837112

    • Google Scholar
42. 1. Smirnova E
    2. Griparic L
    3. Shurland D-L
    4. van der Bliek AM

    (2001) Dynamin-related protein Drp1 is required for mitochondrial division in mammalian cells

    Molecular Biology of the Cell 12:2245–2256.

    https://doi.org/10.1091/mbc.12.8.2245

    • Google Scholar
43. 1. Spudich JA
    2. Watt S

    (1971)

    The regulation of rabbit skeletal muscle contraction. i. biochemical studies of the interaction of the tropomyosin-troponin complex with actin and the proteolytic fragments of myosin

    The Journal of Biological Chemistry 246:4866.

    • Google Scholar
44. 1. Stavru F
    2. Palmer AE
    3. Wang C
    4. Youle RJ
    5. Cossart P

    (2013) Atypical mitochondrial fission upon bacterial infection

    Proceedings of the National Academy of Sciences of the United States of America 110:16003–16008.

    https://doi.org/10.1073/pnas.1315784110

    • Google Scholar
45. 1. Stepanyants N
    2. Macdonald PJ
    3. Francy CA
    4. Mears JA
    5. Qi X
    6. Ramachandran R

    (2015) Cardiolipin's propensity for phase transition and its reorganization by dynamin-related protein 1 form a basis for mitochondrial membrane fission

    Molecular Biology of the Cell 26:3104–3116.

    https://doi.org/10.1091/mbc.E15-06-0330

    • Google Scholar
46. 1. Strack S
    2. Wilson TJ
    3. Cribbs JT

    (2013) Cyclin-dependent kinases regulate splice-specific targeting of dynamin-related protein 1 to microtubules

    The Journal of Cell Biology 201:1037–1051.

    https://doi.org/10.1083/jcb.201210045

    • Google Scholar
47. 1. Tan AR
    2. Cai AY
    3. Deheshi S
    4. Rintoul GL

    (2011) Elevated intracellular calcium causes distinct mitochondrial remodelling and calcineurin-dependent fission in astrocytes

    Cell Calcium 49:108–114.

    https://doi.org/10.1016/j.ceca.2010.12.002

    • Google Scholar
48. 1. Ugarte-Uribe B
    2. Müller H-M
    3. Otsuki M
    4. Nickel W
    5. García-Sáez AJ

    (2014) Dynamin-related protein 1 (drp1) promotes structural intermediates of membrane division

    Journal of Biological Chemistry 289:30645–30656.

    https://doi.org/10.1074/jbc.M114.575779

    • Google Scholar
49. 1. Vizcarra CL
    2. Kreutz B
    3. Rodal AA
    4. Toms AV
    5. Lu J
    6. Zheng W
    7. Quinlan ME
    8. Eck MJ

    (2011) Structure and function of the interacting domains of spire and fmn-family formins

    Proceedings of the National Academy of Sciences of the United States of America 108:11884–11889.

    https://doi.org/10.1073/pnas.1105703108

    • Google Scholar
50. 1. York AG
    2. Chandris P
    3. Nogare DD
    4. Head J
    5. Wawrzusin P
    6. Fischer RS
    7. Chitnis A
    8. Shroff H

    (2013) Instant super-resolution imaging in live cells and embryos via analog image processing

    Nature Methods 10:1122–1126.

    https://doi.org/10.1038/nmeth.2687

    • Google Scholar

Author details

1. Wei-ke Ji

   Department of Biochemistry, Geisel School of Medicine at Dartmouth, Hanover, United States

   Contribution

   W-KJ, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   Contributed equally with

   Anna L Hatch

   Competing interests

   The authors declare that no competing interests exist.

2. Anna L Hatch

   Department of Biochemistry, Geisel School of Medicine at Dartmouth, Hanover, United States

   Contribution

   ALH, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   Contributed equally with

   Wei-ke Ji

   Competing interests

   The authors declare that no competing interests exist.

3. Ronald A Merrill

   Department of Pharmacology, The University of Iowa, Iowa City, United States

   Contribution

   RAM, Ronald Merrill made a key reagent (Mff) essential to experiments conducted for the revision. His insight into the design and productive of this reagent was beyond that of mere technical assistance., Acquisition of data, Contributed unpublished essential data or reagents

   Competing interests

   The authors declare that no competing interests exist.

4. Stefan Strack

   Department of Pharmacology, The University of Iowa, Iowa City, United States

   Contribution

   SS, Drafting or revising the article, Contributed unpublished essential data or reagents

   Competing interests

   The authors declare that no competing interests exist.

5. Henry N Higgs

   Department of Biochemistry, Geisel School of Medicine at Dartmouth, Hanover, United States

   Contribution

   HNH, Conception and design, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   henry.higgs@dartmouth.edu

   Competing interests

   The authors declare that no competing interests exist.

• 7,393

  views

• 1,829

  downloads

• 255

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.