Mutations in mitofusin 2 (MFN2) that are associated with the pathology of the debilitating neuropathy Charcot–Marie–Tooth type 2A (CMT2A) are known to alter mitochondrial morphology. One such abundant MFN2 mutation, R364W, results in the generation of elongated, interconnected mitochondria. However, the mechanism leading to this mitochondrial aberration remains poorly understood. Here, we show that mitochondrial hyperfusion in the presence of R364W-MFN2 is due to increased degradation of DRP1 (also known as DNM1L). The E3 ubiquitin ligase MITOL (also known as MARCHF5) is known to ubiquitylate both MFN2 and DRP1. Interaction with and subsequent ubiquitylation by MITOL is stronger in the presence of wild-type MFN2 than with R364W-MFN2. This differential interaction of MITOL with MFN2 in the presence of R364W-MFN2 renders the ligase more available for DRP1 ubiquitylation. Multi-monoubiquitylation and proteasomal degradation of DRP1 in R364W-MFN2 cells in the presence of MITOL eventually leads to mitochondrial hyperfusion. Here, we provide a mechanistic insight into mitochondrial hyperfusion, while also reporting that MFN2 can indirectly modulate DRP1 – an effect not shown previously.
A balanced fission–fusion dynamic is the key to maintaining integrity, electrical and biochemical connectivity, and quality control of mitochondria, as well as protecting and segregating mitochondrial DNA (mtDNA). Fission creates a number of discrete non-networked mitochondria, whereas fusion increases their connectivity, allowing sharing of matrix proteins and mtDNA (Liu et al., 2020). Fission is mediated by several proteins. A cytosolic protein of the dynamin guanosine triphosphatase (GTPase) family, dynamin-related protein 1 (DRP1, also known as DNM1L) plays a pivotal role in this process; other molecular players of scission, like FIS1, MiD49 (MIEF2), MiD51 (MIEF1) and MFF act as adaptors for DRP1 on mitochondria. DRP1 binds to these adaptors and forms higher-ordered oligomeric complexes that wrap around mitochondrial tubules to mediate fission. DRP1 is known to be regulated by various post-translational modifications, including phosphorylation, ubiquitylation and sumoylation (Chang and Blackstone, 2010). Studies have shown that mice lacking Drp1 have developmental abnormalities and die after embryonic day 12.5 (Ishihara et al., 2009), and mutations in this gene lead to microcephaly, abnormal brain development or even cardiomyopathy (Waterham et al., 2007; Ashrafian et al., 2010). Mutations in human DRP1 also affect the overall mitochondrial morphology, forming a hyperfused mitochondrial network and disrupted cellular homeostasis (Longo et al., 2020; Whitley et al., 2018; Smirnova et al., 2001).
Mitochondrial fusion, on the other hand, is primarily regulated by proteins of the inner and outer mitochondrial membranes. Mitofusins 1 and 2 (MFN1 and MFN2) are essential for fusion of the outer mitochondrial membrane (OMM) (Chen et al., 2003). Though highly similar in sequences, these proteins have distinct roles in regulating mitochondrial dynamics and function (Zorzano and Pich, 2006). MFN1 plays a critical role in mitochondrial docking and fusion. MFN2 affects fusion, controls mitochondrial oxidation (Chen et al., 2003) and stabilizes interactions between mitochondria and endoplasmic reticulum (ER)–mitochondria junctions. MFN2 present at the mitochondria-associated membranes (MAMs) has been reported to have roles in insulin signalling, cardiomyocyte bioenergetics, glucose homeostasis and lipid metabolism (Koshiba et al., 2004; Ishihara et al., 2004; De Brito and Scorrano, 2008; Sugiura et al., 2013; Sebastián et al., 2012; Area-Gomez et al., 2012; Chen et al., 2011; Dorn and Maack, 2013; Dorn et al., 2015; Hailey et al., 2010; Merkwirth and Langer, 2008; Bui et al., 2010). Mutations in MFN2 are associated with neurodegenerative diseases and debilitating neuropathies, where mitochondrial morphology and dynamics are affected, inter-organellar associations are altered, and axonal positioning of mitochondria is disrupted – all features that ultimately promote axonal degeneration (Misko et al., 2012; Cartoni et al., 2010; Detmer et al., 2008). Studies using Mfn2 knockout models and mutants show abnormalities, like uncoordinated limb movements with reduced cerebellum size (Chen et al., 2007). In yeasts, however, there is a single homologue of the mitofusins, Fuzzy onion 1 (Fzo1), which regulates the fusion process. Fusion of the OMM is closely followed and succeeded by fusion of the inner mitochondrial membrane (IMM). This is primarily executed by optic atrophy 1 (OPA1), a GTPase (Okamoto and Shaw, 2005).
An event closely associated with the fission–fusion balance is mitochondrial hyperfusion, which results in the formation of long filamentous mitochondria. While hyperfusion has been known for quite some time, the molecular mechanisms governing this remain only partially explored. Hyperfusion is suggested to be one of the ways to mitigate stress (Tondera et al., 2009; Zahedi et al., 2019; Redpath et al., 2013; Sgarbi et al., 2014). Reduced mitochondrial fission can also trigger hyperfusion (Das and Chakrabarti, 2020). Phosphorylation of DRP1 at Ser637 residue leads to its sequestration in the cytosol, functional inactivation and decreased mitochondrial fission – this in turn promotes hyperfusion. Generation of elongated mitochondria is suggested to be associated with regular cellular physiology as well as pathological states. The hyperfused mitochondrial network present at G1-S transition supports higher ATP production and affects entry into S phase by controlling the levels of a cyclin E (Mitra et al., 2009). Modest levels of cellular stress, well below those needed for apoptosis, lead to the formation of a reticular network of filamentous mitochondria (Tondera et al., 2009; Zahedi et al., 2019; Redpath et al., 2013; Sgarbi et al., 2014). This is referred to as stress-induced mitochondrial hyperfusion (SIMH). This helps reduce the cellular stress burden by optimizing mitochondrial ATP production. However, continued exposure to stress ultimately results in mitochondrial fragmentation (Das and Chakrabarti, 2020). Recent studies have reported the prevalence of mitochondrial hyperfusion in the debilitating neuropathy Charcot–Marie–Tooth type 2A (CMT2A) (Larrea et al., 2019; El Fissi et al., 2018).
In Drosophila models, mutant alleles of Marf that mimic human R364W-MFN2 and L76P-MFN2 mutations (which are prevalent in CMT2A) lead to the formation of enlarged mitochondria, some with elongated morphology, as well as depletion of this organelle at the neuromuscular junctions (El Fissi et al., 2018). However, expression of R364W-MFN2 in mouse embryonic fibroblasts does not elicit mitochondrial hyperfusion (El Fissi et al., 2020). Recently, a rat model with the R364W mutation of the Mfn2 gene has been reported to exhibit multiple axonal abnormalities and progressively worsening motor defects (De Gioia et al., 2020). Although these studies provide evidence that the R364W-MFN2 mutation alters mitochondrial morphology and affects general organismal health, the underlying molecular players remain obscure. Biophysical studies of MFN2 with mutations in the helical domain 1 (HD1; R364W and R364P) have detected elevated GTP turnover (Li et al., 2019). Another MFN2 mutation in the HD1 domain (S378P) can rescue a reticular network of mitochondria in Mfn2-null fibroblasts (Samanas et al., 2020). Mitochondrial hyperfusion could be a result of enhanced fusion or reduced fission. Here, we examine this hypothesis using ex vivo and in silico approaches and show that increased degradation of DRP1 in the presence of the R364W-MFN2 mutant leads mitochondrial hyperfusion. R364W-MFN2 promotes DRP1 multi-monoubiquitylation and its proteasomal degradation, thus enabling hyperfusion. The E3 ubiquitin ligase MITOL (also known as MARCHF5 or MARCH5) is instrumental in this regulation. MITOL ubiquitylates both MFN2 and DRP1 to regulate mitochondrial dynamics. Decreased interaction of MITOL with R364W-MFN2 as compared to the wild-type MFN2 (WT-MFN2) renders it more available to ubiquitylate other substrates. This differential interaction of MITOL with WT-MFN2 versus the mutant governs the increased availability of the ligase for DRP1. Hence, we also identify a potential mechanism by which MFN2 indirectly regulates the levels of DRP1.
R364W-MFN2 promotes mitochondrial hyperfusion
We first verified whether expressing the CMT2A disease-associated MFN2 mutant R364W in various human cell lines led to mitochondrial hyperfusion. HeLa cells exogenously expressing R364W-MFN2 showed a significant increase in the number of interconnected filamentous mitochondria (Fig. 1A,B; Fig. S1A). The control cells, as well as those overexpressing WT-MFN2, had more tubular mitochondria than filamentous ones. Similarly, HeLa cells stably expressing R364W-MFN2 showed a significant increase in average mitochondrial length (Fig. 1C,D; Fig. S1B). Whereas overexpression of WT-MFN2 marginally but significantly increased the average mitochondrial length when compared with that of the controls, exogenous R364W-MFN2 expression further increased this length (Fig. S1C). Protein and transcript levels of MFN2 were comparable across the stable cell lines (Fig. 1C; Fig. S1D). To further verify the effect of R364W-MFN2 on mitochondria, HeLa cells were depleted of MFN2. This led to mitochondrial fragmentation, as detected by decreased mitochondrial length (Fig. 1E,F; Fig. S1E). Exogenous expression of WT-MFN2 in the MFN2-depleted cells rescued the mitochondrial morphology such that it appeared similar to that of control cells. R364W-MFN2 expression in the mock as well as MFN2-depleted cells led to a significant increase in filamentous mitochondria. Enhanced mixing of mitoRFP- and mitoGFP-labelled mitochondria in heterokaryons was observed in the presence of R364W-MFN2 as compared with that in the WT-MFN2 cells (Fig. 1G,H). Furthermore, increased mitochondrial length and a higher percentage of filamentous mitochondria were observed in U87MG glioblastoma cells transiently overexpressing R364W-MFN2 (Fig. 1I,J; Fig. S1F). This indicates a cell line-independent phenomenon. These results together suggest that the presence of R634W-MFN2 in human cells leads to mitochondrial hyperfusion.
Decreased levels of DRP1 in R364W-MFN2 cells leads to mitochondrial hyperfusion
The phenomenon of mitochondrial hyperfusion is mediated by a number of molecular players, including cellular stress, and deletions or mutations in genes associated with the fission–fusion balance. Accordingly, hyperfusion can be initiated either by increased fusion (as in the case of SIMH) or decreased fission events (Mitra et al., 2009; Qian et al., 2012; Larrea et al., 2019; El Fissi et al., 2018). Recently, biophysical studies using truncated forms of MFN2 have shown an increased turnover of GTP in the presence of the R364W-MFN2 mutation. However, this mutation did not significantly perturb protein dimerization at the GTPase interface (Li et al., 2019). While this could be one of the determinants for hyperfusion, we hypothesized that there could be other contributing factors. We therefore examined the protein levels of some of the players involved in mitochondrial dynamics. STOML2 and OPA1 are among the key regulators of SIMH (Tondera et al., 2009; Lebeau et al., 2018). Western blot analyses of whole-cell lysates from the stable cells did not show any significant change in the levels of STOML2 across samples (Fig. 2A,B). No significant difference was observed in the proteolytic processing of OPA1 across the samples, suggesting that SIMH may not be the key player in R364W-MFN2-mediated hyperfusion. Similar protein levels of MFN1 and FIS1 were detected across samples; however, a significant decrease in the levels of DRP1 was detected in R364W-MFN2 lysates when compared with the levels in WT-MFN2 lysates or the control samples. DRP1 activity is facilitated by phosphorylation on two major sites – Ser637 and Ser616 (Cereghetti et al., 2008). Like DRP1, phospho-DRP1 (both Ser616 and Ser637) levels were significantly lower in the presence of R364W-MFN2 when compared with levels in the presence of WT-MFN2. However, the ratio of phospho-DRP1 to DRP1 across the samples remained unchanged, suggesting that the decrease in phospho-DRP1 was merely a reflection of the reduced total DRP1 levels (Fig. 2A,B). The transcript levels of DRP1, however, remained unchanged across samples (Fig. S2A). Imaging results showed that the numbers of DRP1 puncta on mitochondria were significantly lower in cells expressing R364W-MFN2 than in cells expressing WT-MFN2 or the control cells (Fig. 2C). A similar result was also obtained in lattice structured illumination microscopy (SIM) super-resolution images (Fig. S2B,C). Furthermore, co-immunostaining cells for phospho-DRP1 (S616) and DRP1 showed a comparable decrease in puncta [positive for DRP1 and phospho-DRP1 (S616)] (Fig. 2D,E; Fig. S2D). The number of mitochondrial fission events was also significantly lower in R364W-MFN2 cells when compared with control cells (Fig. S2E,F). Taken together, these results suggest that lower levels of DRP1 could be an important player potentiating mitochondrial hyperfusion in cells with R364W-MFN2. To further validate the involvement of DRP1 in R364W-MFN2-mediated hyperfusion, stable cells were transiently transfected with DRP1–mCherry and scored for the mitochondrial length (Fig. 2F,G; Fig. S3A). A rescue of the hyperfusion phenotype was observed in the R364W-MFN2 cells expressing DRP1–mCherry, with mitochondrial length similar to that of WT-MFN2 cells. Thus, increased levels of DRP1 could prevent mitochondrial hyperfusion. Furthermore, DRP1 depletion in cells stably expressing WT-MFN2 led to a significant increase in filamentous mitochondria. Since cells stably expressing R364W-MFN2 already had lower DRP1 levels, further depletion did not significantly alter the filamentous mitochondrial phenotype (Fig. S3B,C).
Increased proteasomal degradation of DRP1 in R364W-MFN2 cells promotes hyperfusion
As a cytosolic protein, DRP1 can be regulated via both the proteasomal and the autophagic pathways (Purnell and Fox, 2013); under conditions of stress, it is degraded via the proteasomal pathway (Zhu et al., 2019; Sabouny et al., 2017). This promotes mitochondrial hyperfusion as a response to mitigate cellular stress (Zhu et al., 2019; Sabouny et al., 2017). An enrichment of DRP1 levels was detected when R364W-MFN2 cells were treated with MG132, a drug that blocks proteasomal degradation (Fig. 3A,B). A similar accumulation of DRP1 was not detected in the WT-MFN2-expressing cells or the control cells. This suggests increased DRP1 degradation via the proteasomal pathway in the presence of R364W-MFN2. MG132 treatment led to an increase in MFN2 levels across the three stable cell lines, and imaging experiments corroborated these results. Live-cell imaging of MG132-treated samples showed rescue of the hyperfusion phenotype in cells stably expressing R364W-MFN2 (Fig. 3C,D; Fig. S4A). There was a decrease in mitochondrial length, and the resulting size was similar to those of WT-MFN2-expressing cells and control cells. Furthermore, increased numbers DRP1 puncta on mitochondria and decreased hyperfusion were detected in MG132-treated R364W-MFN2 cells (Fig. 3E,F). Thus, restoration of DRP1 levels by either transfection or by blocking DRP1 degradation rescued the normal phenotype.
Increased DRP1 ubiquitylation in the presence of R364W-MFN2 promotes hyperfusion
We next checked whether DRP1 ubiquitylation was altered in the presence of R364W-MFN2. Enhanced DRP1 ubiquitylation was detected in R364W-MFN2 samples when compared with levels in WT-MFN2-expressing cells or control cells (Fig. 4A). U87MG cells overexpressing WT-MFN2 or R36W-MFN2 also showed similar results (Fig. 4B). DRP1 ubiquitylation was thoroughly compromised when cells stably expressing either WT-MFN2 or R364W-MFN2 were treated with siRNAs against DRP1; the effect was more prominent in the mutant cell line (Fig. S4B). It is known that the E3 ubiquitin ligase MITOL regulates mitochondrial fission by ubiquitylating DRP1 (Nakamura et al., 2006; Karbowski et al., 2007; Escobar-Henriques and Joaquim, 2019). Hence, we hypothesized that MITOL-mediated regulation of DRP1 could affect mitochondrial hyperfusion in R364W-MFN2 cells. To address this, stable cells were transfected with MITOL and checked for endogenous ubiquitylation of DRP1 (Fig. 4C). Increased DRP1 ubiquitylation was detected in R364W-MFN2 cells with exogenous MITOL expression. WT-MFN2-expressing cells and control cells showed similar ubiquitylation pattern irrespective of the presence of MITOL. An enhanced ubiquitylation smear along with lower levels of DRP1 in R364W-MFN2 cells, even in the absence of exogenous MITOL, suggested a predisposition of this fission protein for proteasomal degradation. To investigate this further, the stable cell lines were co-transfected with HA-tagged ubiquitin (Ub) along with either MITOL or MITOLC14F (a mutant form that is deficient in ligase activity; Fig. S4C) and checked for DRP1 ubiquitylation (Fig. 4D). In R364W-MFN2 cells in the presence of MITOLC14F, the DRP1 ubiquitylation was restored to levels comparable to those present in the WT-MFN2 cells. Significantly enhanced ubiquitylation was observed in these cells upon MITOL overexpression. Exogenous expression of MITOL in R364W-MFN2 cells led to a significant decrease in DRP1 levels (Fig. 4E). In WT-MFN2 cells, MITOL or MITOLC14F overexpression did not result in any detectable alteration in DRP1 levels. Thus, enhanced degradation of DRP1 by overexpression of MITOL could lead to an increase in filamentous mitochondria. The MITOL levels in the stable cell lines were comparable (Fig. 4F). Furthermore, a marginal reduction in DRP1 ubiquitylation was evident in R364W-MFN2 cells when MITOL was knocked down; a similar discernible change was not evident in WT-MFN2 cells or control cells (Fig. 4G). Higher amounts of DRP1 were detected in MITOL siRNA-treated R364W-MFN2 cells when compared with DRP1 levels in the mock-treated controls. Although this does show an effect of MITOL on DRP1 ubiquitylation, we cannot rule out the role of some other ligase in this context.
R364W-MFN2 cells transfected with MITOLC14F showed a partial rescue of the hyperfusion phenotype (Fig. 5A,B; Fig. S4D). We observed that ∼70% cells transfected with MITOL had interconnected mitochondria, while those transfected with MITOLC14F had ∼58% filamentous mitochondria. Transfecting MITOL into WT-MFN2 cells did not lead to a significant increase in the percentage of cells with hyperfused mitochondria – this could be because merely increasing the levels of MITOL is not sufficient to alter mitochondrial morphology. Other factors, such as ubiquitin, could be rate limiting, and hence the number of cells with hyperfused mitochondria would not significantly increase. Exogenous expression of MITOLC14F in WT-MFN2 cells increased the number of cells with interconnected filamentous mitochondria (Fig. S4D), similar to previous reports (Karbowski et al., 2007; Park et al., 2014). This could be because MITOL is an active modulator of mitofusins; this regulation is potentially perturbed in the presence of MITOLC14F. Secondly, it is also possible that a dominant-negative effect of MITOLC14F on the endogenous protein is more evident in WT-MFN2 cells. Furthermore, knocking down MITOL in R364W-MFN2 cells led to significant reduction in mitochondrial hyperfusion and a lesser number of cells with interconnected mitochondria (Fig. S4E–G). Similar to the effects of MITOLC14F expression, WT-MFN2 cells treated with siRNAs targeting MITOL showed significant increase in the occurrence of interconnected filamentous mitochondria.
We observed an increase in the number of DRP1 puncta on mitochondria in R364W-MFN2 cells in the presence of MITOLC14F; this was not the case in WT-MFN2 cells (Fig. 5C,D). Similarly, knocking down MITOL in R364W-MFN2 cells led to a significant increase in the number of DRP1 puncta on mitochondria (Fig. S4H,I). Co-expression of Ub along with MITOL or its mutant in R364W-MFN2 cells further enhanced the mitochondrial phenotypes (Fig. 5E) – we found that ∼81% of R364W-MFN2 cells had interconnected mitochondria in the presence of MITOL and Ub, while those with MITOLC14F had ∼58% with filamentous mitochondria. Surprisingly, WT-MFN2 cells expressing MITOLC14F and Ub also had significantly increased occurrence of filamentous mitochondria. These results again suggest an involvement of ubiquitylation and degradation of DRP1 as a plausible mechanism of mitochondrial hyperfusion in the presence of the CMT2A-associated MFN2 mutation. Interestingly, although the localization of overexpressed MITOL was more dispersed, MITOLC14F was punctate in its distribution (Fig. 5A,E).
Altered MFN2–MITOL interaction and ubiquitylation in the presence of R364W-MFN2
Our results described so far show that MITOL plays a differential role between R364W-MFN2 and WT-MFN2 in the ubiquitylation and degradation of DRP1. However, the question remains as to the mechanism by which R364W-MFN2 could potentially affect DRP1. To address this, we generated and analysed the structures of MFN2 and its mutant in silico. The probable impact of R364W mutation was explored using the 3D structures of closed and extended forms of MFN2 protein via implementation of homology modelling, molecular dynamics simulation analyses and protein–protein docking approaches.
The three-dimensional (3D) structures of the closed and open forms of WT-MFN2 and R364W-MFN2 proteins were generated using homology modelling approaches (see Materials and Methods), and these structures were further utilized in exhaustive all-atom molecular dynamics (MD) simulation to estimate the probable energetic and structural alterations resulting due to the R364W mutation. Energy profiles of WT-MFN2 and R364W-MFN2 (closed form) during the 200 ns MD simulation run suggested that wild-type protein structures tend to be more stable as compared to the mutant counterparts (Fig. 6A) whereas the root-mean-square displacement (RMSD) trajectories showed similar profiles, especially during the latter half of the MD simulation run (Fig. 6B). However, a more in-depth residue-wise structural fluctuation analysis showed segmental and differential fluctuation patterns within the mutant protein compared to WT-MFN2 (Fig. 6C,D). 3D structures of WT-MFN2 and R364W-MFN2 obtained before and after 200 ns simulations were compared. Marked structural alterations at the MITOL-interacting region (residues 400–450) were observed post-simulation between the two MFN2 proteins (Fig. 6E). A similar trend was also observed for the extended form of the MFN2 structure, where WT-MFN2 showed more stable energy profiles compared to those of R364W-MFN2 (Fig. S5A). However, RMSD trajectories showed similar profiles during the initial half of the MD simulation run (Fig. S5B). Interestingly, the extended form of the WT-MFN2 structure showed more residue-wise fluctuation compared to the mutant form (Fig. S5C,D). Subtle structural alterations at the MITOL-interacting region (residues 400–450) were observed post-simulation between the two MFN2 proteins (Fig. S5E). These MD simulation-derived findings encouraged us to investigate whether the probable structural alteration in the MITOL-binding region observed between the WT-MFN2 and R364W-MFN2 proteins could actually impact MITOL binding and docking, as captured through protein–protein docking between MITOL and the WT-MFN2 and R364W-MFN2 proteins (see Materials and Methods for details). We observed that the average Gibbs free energy (ΔG) of the binding for the top five docking solutions was relatively lower (more negative ΔG indicates higher stability) in the complexes generated by WT-MFN2 models extracted from the later time points (150 ns and 200 ns for the closed form and 50 ns, 75 ns and 100 ns for the extended form) of the MD simulation compared to that derived with R364W-MFN2 (Fig. 6F; Fig. S5F).
This probable compromised interaction had two outcomes. Firstly, immunoprecipitation of MFN2 with MITOL indicated lower interaction between the proteins in R364W-MFN2 samples when compared with the WT-MFN2 controls (Fig. 6G). Secondly, reduced ubiquitylation was detected in the R364W-MFN2 stable cell lysates, as compared with that in the WT-MFN2 control lysates (Fig. 6H). It is known that MITOL polyubiquitylates MFN2 via Lys63 (K63) ubiquitin linkages; this alters MFN2 oligomerization status (Sugiura et al., 2013). Hence, we checked for the status of K63-mediated polyubiquitylation of MFN2 in the stable cell lines and detected similar ubiquitylation patterns in the presence of Ub or its mutant (K63Ub, which has arginine substitution at all lysine residues except K63) (Fig. 6H). K63Ub is known to selectively promote K63 ubiquitin linkage (Tan et al., 2008). Next, stable cells were co-transfected with MITOL or its mutant, along with HA-tagged K63Ub. Again, reduced polyubiquitylation of MFN2 was detected in the presence of the R364W-MFN2 mutant when compared to levels in the wild type. Exogenous MITOL expression led to increased MFN2 ubiquitylation in the stable cell lines (Fig. 6I; Fig. S5G). However, the control and WT-MFN2 cell lysates showed a stronger polyubiquitylation smear than lysates from the R364W-MFN2 cells. A compromise in MITOL-mediated K63-mediated polyubiquitylation of MFN2 was detected in the presence of MITOLC14F (Fig. 6J). Similarly, knocking down MITOL reduced K63Ub-mediated polyubiquitylation in the stable cell lines. Here again the reduction in polyubiquitylation was more pronounced in WT-MFN2 samples when MITOL was compromised (Fig. 6K). Immunoprecipitation efficiency between MFN2 and MITOL was not affected by MITOL ligase activity – WT-MFN2 showed a stronger interaction than R364W-MFN2 irrespective of the presence of MITOL or the ligase-deficient MITOLC14F (Fig. 6L). Since direct interaction between MFN2 and DRP1 has also been reported (Huang et al., 2011), we checked for this interaction in the stable lines and found that WT-MFN2 and R364W-MFN2 samples had similar DRP1 immunoprecipitation profiles (Fig. S6A). Our results suggest an indirect effect of R364W-MFN2 on DRP1. It is plausible to hypothesize that reduced association of R364W-MFN2 with MITOL leaves it more available for DRP1. This enables increased association with DRP1, leading to ubiquitylation and degradation.
Multi-monoubiquitylation of DRP1 by MITOL regulates mitochondrial hyperfusion
Immunoprecipitation of DRP1 with MITOL indicated higher interaction between the proteins in R364W-MFN2 samples, when compared with the interaction in WT-MFN2 control samples (Fig. 7A,B). This was observed despite the protein levels of DRP1 being lower in the presence of R364W-MFN2 (Figs 2–4). Furthermore, increased interaction between DRP1 and MITOL was detected when MFN2 was knocked down, and DRP1 protein levels were lower in MFN2 siRNA-treated cells (Fig. 7C,D). Transcript levels of DRP1 remained unaffected in MFN2 compromised samples (Fig. S6B). Ubiquitylation of DRP1 is known (Nakamura et al., 2006; Karbowski et al., 2007; Wang et al., 2011; Gu et al., 2018), but whether the protein is mono- or poly-ubiquitylated is still far from clear. Among the various lysine mutants of ubiquitin, only transfection of K0 ubiquitin (K0Ub) was able to recapitulate a similar pattern of ubiquitylation of DRP1 as seen with wild-type ubiquitin (Fig. 7E; Fig. S6C,D). K0Ub is a lysineless mutant of ubiquitin that promotes only monoubiquitylation (Majumder and Chakrabarti, 2015). Co-expression of K0Ub along with MITOL or its mutant in R364W-MFN2 cells showed a reduction in DRP1 ubiquitylation in the presence of MITOLC14F (Fig. 7F). Higher levels of DRP1 were seen in these cells with MITOLC14F. The WT-MFN2 cells did not show a discernible change in DRP1 ubiquitylation pattern when either MITOL or its mutant was used. We next used a competitive inhibitor of ubiquitin (ΔG75/76Ub) in similar experiments (Fig. 7G). DRP1 ubiquitylation could not be detected in the samples. Moreover, similar levels of DRP1 across all samples indicated that its MITOL-mediated degradation in the R364W-MFN2 cells was compromised in the presence of ΔG75/76Ub. We further verified the effects of ubiquitin variants and MITOL on mitochondrial morphology. A partial rescue of phenotype in the presence of MITOLC14F was observed in the presence of K0Ub (Fig. 8A), similar to the results with Ub (Fig. 5E). We observed that ∼94% of cells stably expressing R364W-MFN2 had interconnected mitochondria in the presence of MITOL and K0Ub, whereas those with MITOLC14F had ∼78% with filamentous mitochondria (Fig. 8A,B). Surprisingly, WT-MFN2 cells expressing MITOLC14F and K0Ub also had significantly increased occurrence of filamentous mitochondria. The near 100% incidence of the filamentous mitochondrial phenotype and decreased DRP1 levels in R364W-MFN2 cells with MITOL and K0Ub clearly suggested that multi-monoubiquitylation and degradation of DRP1 severely perturbed mitochondrial dynamics. In the WT-MFN2 cells, presence of ΔG75/76Ub with MITOL or its mutant promoted mitochondrial hyperfusion – again suggesting that the absence of ubiquitin-mediated modulation of mitofusins perturbed mitochondrial dynamics in these cells (Fig. 8C,D). However, ΔG75/76Ub in R364W-MFN2 cells was able to completely rescue the mitochondrial hyperfusion phenotype. Taking all these results into consideration, we can conclude that in the presence of the CMT2A disease-associated MFN2 mutant, R364W, enhanced degradation of DRP1 leads to mitochondrial hyperfusion.
This study identifies a mechanism by which the abundantly present CMT2A disease-associated MFN2 mutant, R364W, affects mitochondrial morphology and favours the generation of interconnected filamentous mitochondria. We show that in the presence of R364W-MFN2, excess degradation of DRP1 helps elicit a hyperfusion phenotype. Decreased interaction and subsequent ubiquitylation of R364W-MFN2 with MITOL makes the ligase more available for its other substrates. This differential interaction between MITOL and MFN2 in the presence of the mutation leads to increased degradation of DRP1. Proteasomal turnover of DRP1 in the presence of MITOL via its multi-monoubiquitylation leads to a decrease in DRP1 protein levels when R364W-MFN2 is present (Fig. 8E). This study not only helps advance our understanding of the mitochondrial hyperfusion phenotype, it also shows that MFN2 can indirectly regulate DRP1 levels – an effect not shown previously. While the Drosophila and rat models suggest that presence of R364W-MFN2 or its mimetic allele promotes mitochondrial elongation, we provide a direct mechanistic insight into the molecular players responsible for this.
MFN2 regulates mitochondrial dynamics. It promotes mitochondrial fusion in association with MFN1 and other accessory proteins to maintain mitochondrial homeostasis. Dominant allelic mutations in MFN2 are associated with CMT2A, a type of hereditary motor and sensory neuropathy (Verhoeven et al., 2006; Stuppia et al., 2015; El Fissi et al., 2018). While substitution mutations near or within the GTPase domain (such as R94Q and T105M) aggravate mitochondrial fission and clustering, those within HD1 of MFN2 (such as R364W and L76P) promote mitochondrial hyperfusion and interconnected, filamentous morphology. In a Drosophila model, CMT2A mimetic alleles of R94Q-MFN2 and R364W-MFN2 have been shown to cause mitochondrial depletion at neuromuscular junctions (El Fissi et al., 2018). Overexpression of Drp1 in R364W-MFN2 mimetic Drosophila could reverse the hyperfused or giant mitochondria phenotype. The R364W mutant of the Mfn2 gene in a rat model has been shown to cause multiple motor deficits that worsen over time, with reduced density of myelinated axons and active axonal degeneration in distal nerves (De Gioia et al., 2020). While biophysical studies with the R364W-MFN2 mutant have detected enhanced GTP turnover, protein dimerization at the GTPase interface is not significantly altered (Li et al., 2019). This suggests that mitochondrial hyperfusion in the presence of R364W-MFN2 could be the result of multiple factors, of which downregulation of DRP1 (as detected in this study) could be one of the causative agents.
Mitochondrial hyperfusion is known to be potentiated by global inactivation of the mitochondrial fission machinery, with DRP1 playing a decisive role (Santel and Frank, 2008; Wilson et al., 2013). Hence, it is obvious that the recruitment of DRP1 to the mitochondria is finely regulated by multiple post-translational modifications including phosphorylation, sumoylation and ubiquitylation. Phosphorylation of DRP1 is the best characterized amongst these modifications – phosphorylation on Ser616 promotes mitochondrial fission, while that of Ser637 leads to fusion (Cereghetti et al., 2008). Decreases in phospho-DRP1 (Ser616 and Ser637) levels clearly ruled out the involvement of phosphorylation of DRP1 as the primary regulator in the presence of R364W-MFN2.
Since the DRP1 levels were low in cells expressing R364W-MFN2, it was plausible that ubiquitylation and eventual degradation of DRP1 could be responsible for the mitochondrial phenotype observed in the presence of R364W-MFN2. It has been reported that DRP1 is regulated by two mitochondrial E3 ubiquitin ligases, parkin (PRKN) and MITOL. Studies across species (using Drosophila models as well as mammalian cells) have shown that parkin modulates DRP1-mediated mitochondrial fission (Clark et al., 2006; Park et al., 2006; Lutz et al., 2009; Wang et al., 2011). However, whether this effect is direct or indirect remains contentious. There is evidence to suggest a direct effect, whereby parkin-mediated DRP1 ubiquitylation results in degradation and mitochondrial elongation (Wang et al., 2011). Other studies, conversely, suggest that the PINK1–parkin-mediated regulation of mitochondrial fission does not involve direct interaction with DRP1 (Gegg et al., 2010; Glauser et al., 2011; Rakovic et al., 2011; Tanaka et al., 2010). The other E3 ubiquitin ligase, MITOL is suggested to degrade both DRP1 and FIS1 and cause extensive mitochondrial fusion (Nakamura et al., 2006; Yonashiro et al., 2006). However, later reports have provided evidence that MITOL is essential for subcellular distribution and mitochondrial association of DRP1; it also helps to stabilise MFN1 (Karbowski et al., 2007; Park et al., 2014). We found that in the presence of R364W-MFN2, MITOL-dependent DRP1 degradation drove mitochondrial elongation. However, in the presence of WT-MFN2, MITOL activity was more involved in regulating subcellular localization of DRP1. Unlike MITOL, the PINK1–parkin pathway may not be the regulator, since the cell lines used in this study are low or deficient in parkin (Mukherjee and Chakrabarti, 2016a; Yeo et al., 2012). Most of our results show that in the presence of MITOL, DRP1 levels decrease, while that is not so when its ligase activity is compromised. This indicates that MITOL has a role in the regulation of DRP1. It could, however, be an indirect effect – where MITOL regulates another ubiquitin ligase in trans and the second ligase in turn modulates DRP1 levels. Trans regulation between E3 ubiquitin ligases that affects the functions of a third protein is a known phenomenon (Courbard et al., 2002; Magnifico et al., 2003; Yang et al., 2008; Mukherjee and Chakrabarti, 2016a; Mukherjee et al., 2019).
Ubiquitylation of mitochondrial MFN2 in mammals by MITOL regulates ER–mitochondria interactions and affects mitochondrial fusion (Karbowski et al., 2007; Sugiura et al., 2013; Cherok et al., 2017; Escobar-Henriques and Joaquim, 2019). MITOL has been shown to ubiquitylate mitochondrial MFN2 but not its ER-associated counterpart via K63Ub linkages (Sugiura et al., 2013). Ubiquitylation of MFN2 by MITOL is needed for the induction of its GTPase activity. Furthermore, MITOL partially localizes to MAM junctions, where it ubiquitylates MFN2 and is crucial for MAM formation in neurons. While this is reported to trigger fusion between ER and mitochondria, so far there is no evidence to support the relevance of MITOL-dependent MFN2 ubiquitylation in the fusion between two mitochondria. Furthermore, these junctions serve not only as inter-organellar physical tethers but are also hubs for multiple biochemical processes (Rieusset, 2018). During mitochondrial division, ER wraps around a mitochondrion and DRP1 is recruited to these sites. Together they initiate a mechanical strain that drives mitochondrial division (Das and Chakrabarti, 2020; Friedman et al., 2011). Hence, a balance between the regulation of MFN2 by MITOL and that of DRP1 by MITOL is extremely important in regulating healthy mitochondrial dynamics. A decrease in DRP1 protein levels in the presence of the CMT2A-associated R364W-MFN2 mutant suggests an additional layer of complexity to this already critically balanced mechanism. MITOL has different affinities for its substrates (Yonashiro et al., 2006). The lower affinity of R364W-MFN2 for MITOL probably leaves it free to interact with and ubiquitylate DRP1 in excess. This eventually leads to mitochondrial hyperfusion – this is consistent with the previously published data on Drosophila mimetic R364W-MFN2, where overexpression of DRP1 can reverse the enlarged mitochondrial phenotype. However, some questions, though beyond the scope of the present study, still remain. Firstly, presence of MITOLC14F only partially rescues mitochondrial hypertubulation in the R364W-MFN2 mutant cells, thus suggesting a catalytic-independent function of MITOL on DRP1. Secondly, in WT-MFN2 cells, the presence of MITOLC14F causes a decrease in DRP1 puncta on mitochondria; however, it does not significantly affect DRP1 ubiquitylation. This raises the possibility of a dichotomy between ubiquitylation-mediated DRP1 degradation and DRP1-mediated mitochondrial fission.
This study, while trying to understand the reason for mitochondrial hyperfusion in a prevalent form of hereditary motor and sensory neuropathy, actually unravels a unique and indirect mode of regulation of DRP1 by MFN2. It also highlights that the role of MITOL in mitochondrial dynamics is far more complicated than has been so far appreciated.
MATERIALS AND METHODS
Constructs, antibodies and reagents
Myc-tagged MFN2 construct was a gift from Hye-Seong Cho (Department of Biochemistry, Ajou University School of Medicine, Suwon, South Korea). The DRP1–mCherry and GFP-tagged MITOL constructs were purchased from Addgene (Addgene 49152 and 62039, respectively). HA-tagged wild-type ubiquitin and HA-tagged ΔG75/76 ubiquitin were gifts from Rafael Mattera (Neurosciences and Cellular and Structural Biology Division, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD, USA); HA-tagged K0, K48 and K63 ubiquitin were gifts from Kah-Leong Lim (Nanyang Technological University, National University of Singapore, National Neuroscience Institute, Singapore); HA-tagged K6 and K11 ubiquitin were gifts from Tomohiko Ohta (Division of Breast and Endocrine Surgery, St Marianna University School of Medicine, Kawasaki, Japan). OMP25–GFP was a gift from Ramanujan S. Hegde (MRC Laboratory of Molecular Biology, Cambridge, UK). pAcGFP1-Mito (mitoGFP) was a gift from Subrata Banerjee (Saha Institute of Nuclear Physics, Kolkata, India). ER–GFP (GFP–KDEL) was a gift from Erik L. Snapp (Janelia Research campus, VA, USA). mitoRFP was used as described previously (Mookherjee et al., 2020). R364W-MFN2 and C14FMITOL were generated by standard site-directed mutagenesis methods.
Antibodies used and their dilutions for use were as follows: MFN2 (D1E9) (11925; Cell Signaling Technology, USA) at 1:1000 dilution, MFN1 (ab57602; Abcam, Cambridge, UK) at 1:1000 dilution, OPA1 (MA5-16149; Thermo Scientific, Rockford, IL, USA) at 1:2000 dilution, DRP1 (D6C7) (8570; Cell Signaling Technology) at 1:1000 dilution, DRP1 (ab56788; Abcam) at 1:1000 dilution, P-DRP1 (S616) (3455; Cell Signaling Technology) at 1:1000 dilution, P-DRP1 (S637) (4867; Cell Signaling Technology) at 1:500 dilution, MARCH5 (19168; Cell Signaling Technology) at 1:2000 dilution, STOML2 (ab191884; Abcam) at 1:2000 dilution, FIS1 (PA22142; Thermo Fisher Scientific) at 1:1000 dilution, β-actin (ab8226; Abcam) at 1:4000 dilution, TOMM20 (ab56783; Abcam) at 1:100 dilution and β-catenin (ab6302; Abcam) at 1:3000 dilution. GFP (used at 1:5000 dilution) and HA antibodies were gifts of Ramanujan S. Hegde (MRC Laboratory of Molecular Biology, Cambridge, UK). MG132 was from Sigma Aldrich and MitoTracker Red FM was from Invitrogen. MG132 treatment was with 20 μM of the drug for 4 h.
Cell culture, transfection and siRNA-mediated knockdown
Cell lines used for the experiments were HeLa (human cervical cancer cell line), U87-MG (human glioblastoma cell line), and HeLa cells stably expressing WT-MFN2 or R364W-MFN2. The stable expression lines were generated under neomycin selection using standard procedures (Ning and Tang, 2012). HeLa cells were a gift from Ramanujan S. Hegde (MRC Laboratory of Molecular Biology, Cambridge, UK); U87-MG cells were a gift from Debashis Mukhopadhyay (Saha Institute of Nuclear Physics, Kolkata, India). Maintenance of cells in culture was as described previously (Srivastava and Chakrabarti, 2014). Briefly, cells were grown in 10% fetal bovine serum (FBS; Himedia, Mumbai, India) in Dulbecco's modified Eagle's medium (DMEM; Himedia, Mumbai, India) at 37°C and 5% CO2. For transfections of cells, Lipofectamine 2000 was used (Invitrogen, Carlsbad, CA, USA) as per the manufacturer's instructions. 24 h after transfection, cells were lysed using suitable lysis buffer. For all siRNA-mediated knockdown, Dharmacon ONTARGETplus SMARTpool and Ambion Silencer SelectTM were used. siRNAs used in the study were as follows: MFN2 (L-012961-00-0005), DNM1L (4390824), MITOL (4392420), non-targeting siRNA (D-001810-01-20) and GFP (D-001300-01-20). All tissue culture plasticware and Lab-Tek 8-well chambered slides used for microscopy were from Nunc; coverglass-bottom dishes used for microscopy were from SPL Lifesciences.
For immunocytochemistry, cells were fixed with either 4% formaldehyde or methanol as per the requirement of the antibody, as described previously (Chakrabarti and Hegde, 2009; Srivastava and Chakrabarti, 2014). Cells were permeabilized using phosphate-buffered saline (PBS) containing 10% FBS and 0.1% saponin (Sigma-Aldrich) for 60 min, followed by overnight staining in primary antibody at 4°C and 60 min incubation in secondary antibody at room temperature. The samples were then imaged using confocal or super-resolution microscopes.
Western blotting and immunoprecipitation
The protocol for western blotting was as described previously (Kaul and Chakrabarti, 2017). 10% Tris-tricine gels or 7.5% Tris-glycine gels were used for SDS–PAGE followed by western blotting. Quantification of western blots was performed using GelQuant (Thermo Fisher Scientific) and ImageJ (NIH, Bethesda, MD) software. At least three independent experiments were performed, and band intensities were normalized to loading controls. P-values were determined using Student's t-test. For immunoprecipitation, cells were lysed in immunoprecipitation buffer [50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.1% Triton X-100, 1% IGEPAL, 1 mM PMSF and protease inhibitor cocktail (Sigma Aldrich)], and immunoprecipitation was performed under denaturing conditions as described previously (Kaul and Chakrabarti, 2017). In immunoprecipitation assays using GFP antibody, immunoblots were developed using VeriBlot (Abcam, Cambridge, UK) as per the manufacturer's instruction.
Polyethylene glycol cell fusion assay
Stable cells were transfected with mitoGFP or mitoRFP. The cell fusion assay was performed as described previously (Mukherjee and Chakrabarti, 2016b). In short, 16 h post transfection, cells expressing mitoGFP and mitoRFP were mixed in a 1:1 ratio and replated in 35 mm dishes with coverslips. They were treated with 40 μg/ml cycloheximide for 30 min. This was followed by a 90 s treatment with 50% (v/v) polyethylene glycol (PEG) to induce cell fusion. Fused cells were co-cultured for 5 h in medium containing cycloheximide, fixed with 10% formalin and imaged using the Zeiss LSM710-ConfoCor 3 microscope.
In vivo ubiquitylation assay
In vivo ubiquitylation assay was performed as previously described (Srivastava and Chakrabarti, 2014). Briefly, stable cells co-transfected with HA-tagged wild-type ubiquitin (or ubiquitin mutants), MITOL or MITOLC14F constructs were lysed in immunoprecipitation buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% NP40, 0.1% Triton X-100, 2 mM PMSF and protease inhibitor cocktail) and immunoprecipitated under denaturing conditions with MFN2 or DRP1 antibody. Ubiquitylated DRP1 or MFN2 were detected by immunoblotting with HA antibody. Ubiquitin constructs used were as previously described (Mattera et al., 2004; Nishikawa et al., 2004; Tan et al., 2008).
Confocal imaging and image analyses
Confocal imaging was performed using Zeiss LSM510-meta, LSM710 ConfoCor 3 and Nikon A1R+Ti-E with N-SIM and FCS microscope systems. An Ar-ion laser (GFP excitation or Alexa Fluor 488 excitation with the 488 nm line) and He-Ne lasers (RFP, Alexa Fluor 546 and Alexa Fluor 594 excitation with the 543 nm line; Alexa Fluor 633 and MitoTracker Red excitation with the 561 nm line) were used with either 63×1.4 NA water or 60×1.4 NA oil immersion objectives. mitoGFP, OMP25–GFP and DRP1–mCherry transfected cells were imaged in CO2-independent medium (Thermo Fisher Scientific), maintaining conditions of live-cell imaging as described previously (Mitra and Lippincott-Schwartz, 2010).
Super resolution images were taken using the Zeiss ELYRA 7 microscope system. Image analyses were done using ImageJ software.
Each layer of the z-stack images (1068×1773 pixels) as collected above was converted into a 534×886 pixel matrix. Images with various colour channels were processed using MATLAB image processing toolbox (Gonzalez et al., 2003). 3D models of the cellular substructures were created by combining the z-stack slides and converting the 3D pixel matrix into Cartesian coordinates. PyMOL software (The PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC) was used to convert the Cartesian coordinates of each z-stack image into a 3D surface model.
Mitochondrial length measurement was done manually by drawing lines along the entire length of each mitochondrion and measuring them using Fiji (https://imagej.net/software/fiji/). DRP1 puncta were counted on mitochondria along its length and were normalized to 5 μm length of mitochondria. 2D confocal micrographs of cells of resolution 1024×1024 at 300 dpi were used for quantitative image analysis, where mitochondria were stained with MitoTracker Red. Images were also processed using MATLAB image processing toolbox (Gonzalez et al., 2003) to calculate the mitochondrial length. For this analysis, images were sharpened in order to isolate mitochondria separately by replacing the edge pixel (intensity less than average of its one-pixel neighbour window) by the maximum neighbour intensity. This was followed by an edge-detection step using edge(), which finds edges at points where the gradient of given image is maximum, using the Roberts approximation to the derivative (Parker, 1997). Next, we used dilation by imdilate(), followed by imfil() functions to repair breakages and fill holes where mitochondria were separated from the background. A minimum length threshold of pixel size 2000 (1 pixel=0.00877 µm) was implemented to filter and classify elongated mitochondria. Finally, bwconncomp() and regionprops() were used to calculate the length of the mitochondrial filaments.
Overlap of co-immunostained phospho-DRP1 (S616) (green) and DRP1 (red) puncta was calculated through quantitative image analysis using the 2D confocal micrographs of cells of resolution 1024×1024 (65.64 μm×65.64 μm). Green and red puncta were separated using MATLAB image processing toolbox (Gonzalez et al., 2003). Partial or whole overlap was calculated between DRP1 and phospho-DRP1 puncta within a close neighbourhood of ∼0.5 μm (8 pixels) across the periphery of each puncta. MATLAB imadjust() tool was used to adjust the puncta intensity between 0.01 and 0.6 followed by conversion of the greyscale image to a binary image, with a threshold luminance level of 0.2 and a lower cutoff area threshold of ∼0.04 μm2 (10 pixels) using bwareaopen(). Next, bwconncomp() and regionprops() functions were used to isolate each puncta and imdilate() was used to create a close neighbourhood of ∼0.5 μm (8 pixels) radius from the periphery of each identified puncta. Finally, overlap between green and red puncta, and its neighbourhood regions was calculated and represented as a fraction of overlap.
Quantitative reverse transcription and real-time PCR
Cells were harvested and RNA was extracted as previously described (Mookherjee et al., 2020). Quantitative reverse transcription-PCR (qRT-PCR) was used to compare the expression of DRP1 (forward primer, 5′-GTGAGGCAGGAGAATTGCTT-3′; reverse primer, 5′-TTGAGACGGAGTTTCGCTCT-3′) and MFN2 (forward primer, 5′-ATGCATCCCCACTTAAGCAC-3′; reverse primer, 5′-CCAGAGGGCAGAACTTTGTC-3′) genes. The two primers used for the housekeeping gene GAPDH were: forward primer, 5′-GACAGTCAGCCCGCATCTTCT-3; reverse primer, 5′-GCGCCCAATACGACCAAATC-3.
The 3D structures of open and closed forms of WT-MFN2 protein were generated using the I-TASSER (Iterative Threading ASSEmbly Refinement) protein structure prediction server, which used a bacterial dynamin-like protein BDLP (PDB ID: 2J69; Low and Löwe, 2006) as template structure. The root-mean-square deviation (RMSD) between the MFN2 model and the template structure was quite low (1.18 Å) despite low sequence similarity between them (sequence identity of 11%). MFN2 is structurally and perhaps functionally similar to BDLPs and therefore may exist as both closed and extended form depending on the rearrangements in the hinge region governed by the bound nucleotide states. So, the open/extended form of MFN2 protein was also generated based on the crystal structure of a BDLP bound with 5′-guanylyl imidodiphosphate (GMPPNP) (PDB ID: 2W6D; Low et al., 2009) via the I-TASSER server. Similar 3D models of MFN2 were used in a previous study (Samanas et al., 2020). The open and closed structures for R364W-MFN2 were also generated using the same protocol described above.
MITOL 3D model was generated using a combination of template-based and ab-initio modelling. The N-terminal zinc finger domain structure (residues 1–70) of MITOL was generated using comparative modelling via the MODELLER program (Šali et al., 1993), with the template structure of RING domain of E3 ubiquitin ligase Doa10 (PDB ID: 2M6M) while the remaining part was modelled using ab initio modelling approach by the Robetta server (Kim et al., 2004).
Molecular dynamics simulation
The 3D structures of WT-MFN2 and R364W-MFN2 proteins were subjected to MD simulation to study the impact of R364W mutation on the structural dynamics using the Desmond (Bowers et al., 2006) MD simulation package. The OPLS_2005 force field parameters (Jorgensen et al., 1996; Kaminski et al., 2001) were used to generate the coordinates and topology of the molecules. TIP3P (Mark and Nilsson, 2001) water was used to solvate the system and an orthorhombic simulation box was created with boundaries 10 Å apart from the molecule. Steepest descent and the limited-memory Broyden–Fletcher–Goldfarb–Shanno (LBFGS) algorithm (Saputro and Widyaningsih, 2017) was used to minimize the energy of the system. The production run was performed at 300 K temperature and 1 atmospheric pressure with a time step of 2 fs for 200 ns (for the closed structures) and 100 ns (for the extended structures) simulations, respectively. Nosé–Hoover Chain thermostat (Nosé, 1984) and Martyna–Tobias–Klein barostat (Martyna et al., 1994) were used to regulate the temperature and pressure of NPT (N, number of particles; P, pressure; T, temperature) ensemble, respectively. The RESPA (reversible reference system propagator algorithms) (Tuckerman et al., 1992) was used for integrating the equations of motion. The energy profile during simulation was analysed by ‘simulation quality analysis’ tool of the Desmond package. The RMSD and root-mean-square fluctuation (RMSF) of the protein residues were analysed by using the ‘simulation event analysis’ module of the Desmond package.
The structures for WT-MFN2 and R364W-MFN2 for both closed and extended forms were extracted at regular time points from the 200 ns and 100 ns simulations, respectively, and were docked with the full-length MITOL 3D model using protein–protein docking mode of the PatchDock (Schneidman-Duhovny et al., 2005) docking program. The region consisting of residues 400–450 of MFN2 was considered to be important for binding (Sugiura et al., 2013), whereas the C-terminal region (residues 254–278) of MITOL was utilized as the seed region for the docking. ΔG of the binding for the top five docking solutions was calculated using the PDB ePISA server (Krissinel and Henrick, 2007) and was compared across WT and mutant MFN2 structure-based docking.
We thank Subrata Banerjee, Hye-Seong Cho, Tomohiko Ohta, Erik L. Snapp, Rafael Mattera and Kah-Leong Lim for plasmids; Ramanujan S. Hegde for cells, plasmids and antibodies; Debashis Mukhopadhyay for cells. We thank Rishi Kant (Zeiss, India) and A. Banerjee (Zeiss, India) for help with the 3D lattice SIM and Zeiss 980 LSM with Airyscan2 super-resolution microscopy, respectively. We acknowledge O.C. and S.C. laboratory members for their help and support throughout the study, especially R. Mukherjee and D. Chatterjee.
Conceptualization: R.D., O.C.; Methodology: R.D., S.D., S.C.; Software: I.M.K., S.D., S.C.; Validation: R.D., O.C.; Formal analysis: R.D., I.M.K., S.D., S.C., O.C.; Writing - original draft: R.D., O.C.; Writing - review & editing: R.D., S.C., O.C.; Visualization: R.D., O.C.; Supervision: O.C.; Project administration: O.C.; Funding acquisition: O.C.
O.C. is supported by the ‘Integrative Biology on Omics Platform Project’, intramural funding of the Department of Atomic Energy, Government of India. O.C. is partially funded by the Science and Engineering Research Board (SERB), Department of Science and Technology, Government of India (EMR/2016/002706) and a National Women Bioscientist Award grant from the Department of Biotechnology, Ministry of Science and Technology, India. I.M.K. acknowledges financial support from the Department of Biotechnology, Ministry of Science and Technology, India (DBT/2019/IICB/1213). S.C. acknowledges CSIR - Indian Institute of Chemical Biology for infrastructural support, and S.D. acknowledges the Council of Scientific and Industrial Research, India for a PhD fellowship.
The authors declare no competing or financial interests.