Disease-associated mutations in Drp1 have fundamentally different effects on the mitochondrial fission machinery

Abstract

Patient mutations have been identified throughout dynamin-related protein 1 (Drp1), the key protein mediator of mitochondrial fission. These changes generally impact young children and often result in severe neurological defects and, in some instances, death. Until now, the underlying functional defect leading to patient phenotypes has been largely speculative. We therefore analyzed six disease-associated mutations throughout the GTPase and middle domains (MD) of Drp1. The MD plays a role in Drp1 oligomerization, and three mutations in this region were predictably impaired in self-assembly. However, another mutant in this region (F370C) retained oligomerization capability on pre-curved membranes despite being assembly-limited in solution. Instead, this mutation impaired membrane remodeling of liposomes, which highlights the importance of Drp1 in generating local membrane curvature before fission. Two GTPase domain mutations were also observed in different patients. The G32A mutation was impaired in GTP hydrolysis both in solution and in the presence of lipid but remains capable of self-assembly on these lipid templates. The G223V mutation also exhibited decreased GTPase activity and was able to assemble on pre-curved lipid templates; however, this change impaired membrane remodeling of unilamellar liposomes similar to F370C. This demonstrates that the Drp1 GTPase domain also contributes to self-assembly interactions that drive membrane curvature. Overall, the functional defects caused by mutations in Drp1 are highly variable even for mutations that reside within the same functional domain. This study provides a framework for characterizing additional Drp1 mutations to provide a comprehensive understanding of functional sites within this essential protein.

Introduction

Mitochondria serve a crucial role in ATP production, and cells with high energy demands, such as cardiomyocytes and neurons, are especially sensitive to mitochondrial dysregulation (1–5). To preserve mitochondrial function and sustain cell health, a necessary balance between organelle fission and fusion regulates ultrastructural changes that govern bioenergetic activity and energy homeostasis. Correspondingly, mutations in proteins responsible for regulating mitochondrial dynamics are associated with several human diseases, including neurodegenerative disorders (6,7), ischemia-reperfusion injury (8–10) and optic atrophy (11). Dynamin-related protein 1 (Drp1) is the principal regulator of mitochondrial fission (12–14) and point mutations in the coding region of its gene (DNM1L) are especially harmful as patients present with encephalopathy due to defective mitochondrial and peroxisomal fission-1 (EMPF1) (15–21). This disease is characterized by neurodevelopmental delay, microcephaly, refractory epilepsy and hypotonia that may lead to death in childhood (Fig. 1). Patients with Drp1 mutations exhibit a hyperfused mitochondrial network because of unopposed fusion, but the underlying basis for how defects in the mitochondrial fission machinery progress to disease is ill-defined.

Figure 1.

Patient mutations characterized in this study. (A) A schematic depicting Drp1 domain architecture is presented. The patient mutations are located throughout the GTPase (green) and middle (dark blue) domains, which are highlighted in green and navy blue, respectively. This model was generated using AlphaFold. The structure shows the relative position of each domain, including the BSE (purple) and GED (light blue). The intrinsically disordered VD is represented as an orange line. (B) A list of Drp1 patient mutations included in this study is provided alongside the patient presentation and patient outcome if it is known.

Within cells, mitochondrial division is a multistep process initiated by the enhanced recruitment and oligomerization of Drp1 on the mitochondrial surface. The importance of this process in mammalian cells has been implicated in several essential cellular events, including removal of organelle damage, regulation of apoptosis and maintaining an appropriate distribution of mitochondria to promote signaling pathways and sustain energy demands (22–24). EMPF1 mutations identified thus far are located throughout the catalytic and self-assembly domains of Drp1, which are referred to as the GTPase and middle domains (MD), respectively (Fig. 1A). Because the complete loss of Drp1 causes embryonic lethality in mammals (2), we propose these patient mutations represent the most severe disruptions to fission possible while still allowing for survival beyond embryonic development. As such, these point mutations provide unique examples of specific defects in the mitochondrial fission machinery and investigation can reveal key roles associated with these disrupted sequences.

Like other dynamin superfamily proteins (DSPs), Drp1 is composed of four distinct domains that play critical roles in mitochondrial membrane remodeling, namely, the GTPase domain, the MD, the GTPase effector domain (GED) and the variable domain (VD). The GTPase domain is highly conserved among DSPs and mediates nucleotide binding and hydrolysis (25). It contains four key consensus nucleotide binding elements, which are referred to as G1–G4 motifs. The G1 motif, also known as the P-loop, is responsible for binding the phosphates present in GTP. One well-defined synthetic mutation in this region, K38A, is known to completely attenuate GTP hydrolysis by Drp1 (14). Without catalytic activity, Drp1 cannot constrict the underlying membrane (26), thereby inhibiting mitochondrial fission. The G2–G4 motifs are involved in coordinating the Mg²⁺ ion as well as in binding the nucleotide base.

The MD forms an extended helical structure that interacts with the GED, to form a four-helix bundle, called the stalk. This structure regulates Drp1 self-assembly interactions and the formation of larger oligomers (27). Drp1 exists predominantly as a mixture of dimers and tetramers in the cell (28), and the dimer represents the functional assembly unit of Drp1 (29). Higher-order multimers, namely tetramers and above, exist in a dynamic equilibrium with the dimers in solution, and these species represent storage forms of the protein that can serve as ‘reservoirs’ (29). The synthetic mutation G363D is located in the MD and results in an obligate dimer that prevents Drp1 self-assembly interactions (30,31). By preventing functional assembly of the fission machinery, this mutation impairs mitochondrial division.

The bundle signaling element (BSE) is a region composed of three helices derived from the N- and C-termini of the GTPase domain and the C-terminus of GED. This structurally distinct segment connects the GTPase domain to the stalk and bridges these catalytic and assembly regions, respectively. In dynamin, the BSE changes conformation in response to nucleotide binding, and in this way the nucleotide-state of the GTPase domain can drive stalk reorganization of the assembly and impart contractile force on the membrane (32). Interestingly, GTP binding does not mediate constriction by Drp1; rather, hydrolysis is required (26,33).

Lastly, the VD is comprised of a ~130 amino acid sequence connecting the membrane-proximal ends of the MD and GED, which positions this region at the opposite end of the stalk from the GTPase domain in the 3D structure (Fig. 1). This domain is intrinsically disordered and, consequently, was deleted in the crystallographic studies (34). The VD mediates lipid interactions and directly interacts with lipid membranes containing the mitochondria-specific lipid cardiolipin (CL) (29,35,36). It also acts as a negative regulator of mitochondrial fission by impeding the interaction between Drp1 and mitochondrial fission factor at the outer mitochondrial membrane (OMM) (37). Collectively, nucleotide interactions with the GTPase domain and lipid interactions mediated by the stalk and VD promote functional assembly of Drp1 to form the fission machinery. In cells, this process follows membrane remodeling through mitochondrial-endoplasmic reticulum contact sites that impinge on the OMM (38), pre-shaping local mitochondrial constriction sites. The extent to which Drp1 contributes to membrane remodeling through functional self-assembly remains unclear.

While synthetic mutations have helped characterize the roles of functional domains, patient mutations have also been identified throughout the Drp1 sequence. Still, the mechanism by which Drp1 activity and mitochondrial fission are impaired is unclear and detailed structural and biochemical characterization has not yet been performed. Visualization of mitochondria in patient-derived fibroblasts revealed that these missense mutations all invoke a hyperfused mitochondrial network. Here, we isolate recombinant proteins that incorporate several Drp1 point mutations to identify specific defects using a series of biochemical and structural methods. Given the severe detrimental effects of Drp1 mutations in patients, we expect that all the mutations will be impaired in some essential aspect of the fission machinery. Based on the position of the mutation, these reconstitution experiments will identify key sites that contribute to assembly of Drp1 polymers and/or GTPase activity. As such, this study highlights essential roles for Drp1 in promoting mitochondrial fission.

Collectively, we present an unappreciated variation in functional defects caused by mutations in Drp1, including at distinct sites located within the same domain. We examined four patient mutations located in the MD and two in the GTPase domain. Most MD mutations were assembly-incompetent and dimer-limited, though one was still able to oligomerize on a pre-curved lipid template but unable to tubulate large, unilamellar vesicles (LUVs). Adding to the complexity of functional interactions, we found that the two mutations in the GTPase domain possessed different defects. One blunted GTP hydrolysis, whereas a second was defective in membrane remodeling. These results underscore the importance of Drp1’s ability to recognize a lipid template and to drive membrane curvature, and this remodeling event is coordinated through distinct domains. Collectively, patient mutations in Drp1 allow us to probe the unique functions of specific regions within the protein.

Results

To assess the functional impact of different disease-associated mutations in Drp1, a set of established assays was utilized to ascertain protein self-assembly properties along with membrane binding and remodeling capabilities. Initially, the oligomerization state(s) of each protein in solution was determined using size-exclusion chromatography coupled to multi-angle light scattering (SEC-MALS). In parallel, negative-stain EM was used to assess the oligomerization properties of WT and mutant proteins in the presence of GMP-PCP, a non-hydrolysable nucleotide analog that mimics the GTP-bound state, or lipid nanotubes containing the mitochondria-specific phospholipid CL, which is known to promote Drp1 self-assembly and stimulate GTP hydrolysis (39,40). Nanotubes provide a rigid template with a pre-defined curvature for Drp1 self-assembly, so LUVs, also referred to as liposomes, were used to examine membrane remodeling capabilities without any pre-defined curvature. After imaging the proteins on these lipid templates, the abundance and morphology of the Drp1 oligomers could be assessed. As a complement to structural studies, GTPase assays were performed to evaluate the enzymatic activity of WT and mutant Drp1 in solution and in the presence of lipid templates. Drp1 self-assembly on a lipid scaffold results in a stimulated rate of GTP hydrolysis because of intermolecular GTPase domain interactions in the helical lattice (39).

Drp1 self-assembly properties are impaired by the A395D, R403C and C431Y mutations

The first case of a pathogenic variant in Drp1 was described in 2007 (16). This patient, a newborn girl, exhibited a defect in mitochondrial fission and died at age 37 days. A heterozygous mutation in Drp1 was identified, which set a precedent for the description of missense mutations identified later (16). This newborn presented with microcephaly, abnormal brain development, optic atrophy, hypoplasia, persistent lactic acidemia and a mildly elevated plasma concentration of long chain fatty acids, which were attributed to altered mitochondrial and peroxisomal dynamics. This dominant-negative mutation (c.1184C → A) resulted in the introduction of a negatively charged amino acid (aspartate) in place of an alanine at position 395 (A395D). This site is within the MD of Drp1 and is conserved among DSPs, including dynamin (Fig. 2A). It has been previously noted that this domain is critical for intermolecular interactions that drive Drp1 oligomerization (27,30,41). Thus, introducing a negatively charged aspartic acid residue at this position was predicted to disrupt self-assembly.

Figure 2.

A395D, R403C and C431Y are dimer-limited mutations located in the MD. (A) Each mutation is marked by a red sphere and labeled on the crystal structure monomer (colored as in Fig. 1). The adjacent dimer is shown in gray. (B) GTPase activity of WT Drp1 and the A395D, R403C and C431Y (0.5 μM final for each) mutants in solution (blue) as well as in with lipid nanotubes (red). Significance was evaluated using a two-way ANOVA. ^****P < 0.0001. ns, not significant. Error bars represent the SEM. (C–E) SEC-MALS was performed to analyze the multimeric state of each mutant protein (red) compared with WT (blue). The dotted lines shown on the right y-axis indicates the size of the Drp1 multimers based on the predicted molar mass. (F–M) Negative-stain EM images showing WT protein (F and G), A395D (H and I), R403C (J and K) and C431Y (L and M) in the presence of the non-hydrolysable nucleotide analog GMP-PCP (F, H, J, L) and in the presence of lipid nanotubes (G, I, K, M). Solid white arrowheads highlight protein striations on the nanotube resulting from decoration by WT protein. White lines indicate undecorated areas on the nanotube. Scale bar (black), 200 nm in all images.

Our studies with recombinant A395D confirmed a defect in self-assembly, consistent with a previous study (30) using affinity-tagged protein that augments Drp1 assembly (42). First, we measured the GTPase activity of A395D in the presence and absence of lipid nanotubes containing CL (Fig. 2B). These nanotubes are rigid and possess pre-existing curvature, so Drp1 binding and oligomerization can be measured without the need to induce membrane remodeling. A395D displayed an average rate of 3.8 ± 0.2 min⁻¹ in solution and a similar rate of 6 ± 1 min⁻¹ when nanotubes were present. Comparatively, WT exhibited rates of 1.2 ± 0.1 min⁻¹ and 15 ± 1 min⁻¹ in the absence and presence of CL nanotubes, respectively. While the measured difference between the enzymatic activity of A395D and that of WT in solution was not significant (Fig. 2B), the average rate of GTP hydrolysis for A395D was about 3-fold higher than WT. When nanotubes were added to the reaction, the GTPase activity of WT was characteristically stimulated ~13-fold, whereas no significant change in the rate of GTP hydrolysis was observed when A395D was added to nanotubes. Notably, we observed some association of this mutant with nanotubes (Supplementary Material, Fig. S1), though we did not find any organized decoration by EM. Therefore, A395D was unable to assemble into a functional oligomer on the lipid template, which would also prohibit membrane remodeling.

As a complement to enzymatic characterization, SEC-MALS analysis revealed that A395D was limited to a dimer in solution. Comparatively, WT Drp1 existed primarily as a mixture of dimers and tetramers (Fig. 2C). Our negative-stain EM images further support these results. Unlike WT, A395D was unable to form rings or spirals in solution with GMP-PCP (Fig. 2F and H). Additionally, A395D could not form functional helical polymers on lipid nanotubes (Fig. 2G and I), which was consistent with the lack of lipid-induced stimulation of GTPase activity. Overall, A395D was defined as an assembly-defective mutation, which is consistent with the important role of the MD in intermolecular Drp1 interactions.

In 2016, an additional MD mutation was identified corresponding to a c.1207C → T mutation, resulting in the substitution of arginine with cysteine at position 403 (R403C) (17). Interestingly, this mutation was discovered in two unrelated patients, both of whom developed normally until ~4–5 years of age before presenting with refractory epilepsy, encephalopathy, developmental regression and myoclonus. As with A395D, the R403C mutation is also located in the MD relatively close to the base of the assembly stalk (Fig. 2A). While this residue is in a loop region that was unresolved within the crystal structure (34), a cryo-EM structure with MiD49 (43) resolved this loop structure and likely informs the AlphaFold model (Fig. 2A). Overall, R403C showed similar attributes when compared with A395D. Specifically, the R403C mutant was found to be dimer-limited in solution and was unable to decorate CL-nanotubes (Fig. 2D, J and K). The basal GTPase activity exhibited by R403C in solution (3.3 ± 0.2 min⁻¹) was also increased compared with WT (1.2 ± 0.1 min⁻¹), although this change was not statistically significant (Fig. 2B). With nanotubes, the activity rate increased to 6.5 ± 0.4 min⁻¹, and similar to A395D, this mild stimulation was not significantly different, consistent with the lack of functional assembly on CL-nanotubes observed by EM. Likewise, nucleotide binding did not promote assembly into spiral polymers.

In 2018, another pathogenic variant in Drp1 was identified in the MD (15). This patient, a newborn girl, presented with developmental delay and regression, seizures, muscle tone abnormalities and gastroesophageal reflux. The mutation was lethal at age 10 months. Clinical whole-exome sequencing revealed the mutation c.1292G → A, which resulted in an amino acid substitution of cysteine to a tyrosine at position 431 (C431Y) in the MD (Fig. 2A). Relative to the two previously described mutations at the base of the stalk (A395D and R403C), the C431Y mutation is located more toward the center of the assembly stalk (Fig. 2A). In solution, the rate of GTP hydrolysis for C431Y measured 2.1 ± 0.1 min⁻¹ (Fig. 2B). As conventionally predicted for mutations in the MD, C431Y did not display an enhanced rate of GTP hydrolysis in the presence of nanotubes (2.8 ± 0.5 min⁻¹) and was dimer-limited in solution (Fig. 2E). Our EM images further support these results, with C431Y being unable to form rings or spirals in the presence of GMP-PCP (Fig. 2L) and also being unable to decorate lipid nanotubes (Fig. 2M).

F370C is a dimer-enriched, assembly-competent mutant

In 2019, a patient with the mutation c.1109 T → G resulting in a phenylalanine to cysteine substitution at position 370 (F370C) in Drp1 was identified (19). This patient, a young girl, developed normally from the time she was a newborn until age 5 years when she began experiencing myoclonic seizures. A brain MRI at age 3 years appeared normal, but a second MRI at age 7 years revealed global cortical atrophy. Thus, similar to patients with the R403C mutation, this young girl did not present with symptoms characteristic of other pathogenic Drp1 mutations until she reached age 5 years. Like the three previously discussed mutations, the F370C mutation is also located in the MD (Fig. 3A), though its position may be critical to maintain proper intramolecular interactions between adjacent alpha-helices in the Drp1 stalk (MD–GED interactions). In the absence of nanotubes, the rate of GTP hydrolysis for F370C measured 2.5 ± 0.3 min⁻¹, which was not significantly different from the activity of WT protein in solution (Fig. 3B). In the presence of lipid nanotubes, F370C displayed a stimulated rate of 14.0 ± 0.6 min⁻¹, a ~5.5-fold increase. While the fold-change in lipid stimulation was reduced compared with WT, there is not a significant difference in the nanotube-stimulated activity of F370C versus WT. However, the rate of GTP hydrolysis for F370C was found to be 4.1 ± 0.5 min⁻¹ in the presence of liposomes. This activity was significantly decreased compared with F370C with nanotubes and was not significantly different from that of F370C in its solution state, indicating that the GTPase activity of the F370C mutant was enhanced by nanotubes but unaffected by liposomes. Our SEC-MALS analysis found that this mutation led exclusively to dimers in solution (Fig. 3C). In support of the GTPase and SEC-MALS results, EM imaging showed that F370C does not oligomerize into spirals in the presence of GMP-PCP (Fig. 3D and G). Still, F370C was able to decorate nanotubes but could not tubulate liposomes, consistent with the GTPase activities measured in the presence of these different lipid templates (Fig. 3E, F, H and I). To our knowledge, this is the first Drp1 MD mutant described that is unable to form rings or spirals in the presence of GMP-PCP but that can still bind and self-assemble on a lipid template. This suggests that Drp1 spirals and lipid-induced helices are reliant on distinct intermolecular interactions. Moreover, the dimer unit is sufficient to form a helical polymer on a pre-curved membrane template, leading to stimulated GTP hydrolysis.

Figure 3.

F370C is a dimer-enriched and assembly-competent Drp1 mutant. (A) Residue F370 (red) is located within the MD of Drp1. The adjacent Drp1 monomer is shown in gray. (B) GTPase activity of WT versus F370C (0.5 μM final for each) in solution (blue) as well as in the presence of lipid nanotubes (red) and liposomes (orange). Significance was evaluated using a two-way ANOVA. ^****P < 0.0001. ns, not significant. Error bars represent the SEM. (C) SEC-MALS was performed to analyze the multimeric state of the F370C mutant (red) compared with WT protein (blue). The dotted lines shown on the right y-axis indicate the size of the Drp1 multimers based on the predicted molar mass. (D–I) Negative-stain EM images showing WT protein (D–F) and F370C (G–I) in the presence of the non-hydrolysable nucleotide analog GMP-PCP (D, G), in the presence of lipid nanotubes (E, H) and in the presence of liposomes (F, I). Solid white arrowheads highlight protein striations on the nanotube resulting from protein decoration. The white line indicates an undecorated area on the nanotube. Scale bar (black), 200 nm in all images.

GTP hydrolysis by Drp1 is impaired by the G32A mutation

In a previous study, whole-exome sequencing revealed that a girl aged 7 possessed a mutation in Drp1 at c.95G → C, which resulted in a glycine to alanine amino acid substitution at position 32 (G32A) (15). The patient presented with developmental delay, microcephaly, optic atrophy, nystagmus, dysmorphic facies, muscle tone abnormalities, ataxia and failure to thrive. Patient-derived fibroblasts showed an aberrantly hyperfused mitochondrial network (15). G32A was among the first mutations to be identified in the GTPase domain (Fig. 4A). As would be anticipated for a mutation in this region, the catalytic activity of G32A in solution (0.9 ± 0.1 min⁻¹) was diminished compared with WT (1.2 ± 0.1 min⁻¹) (Fig. 4C). Additionally, the GTPase activity of the G32A mutant was not increased by the presence of nanotubes nor liposomes (Fig. 4D). In solution, G32A existed in essentially the same mixture of multimeric states as WT (Fig. 4B), which was consistent with the solution assemblies being maintained by stalk interactions similar to WT. EM images revealed that G32A formed rings, but not spirals, in the presence of GMP-PCP (Fig. 4E and I). In addition, G32A was able to decorate CL-containing lipid nanotubes (Fig. 4F and J).

Figure 4.

G32A is catalytically impaired and assembly competent. (A) The residue G32 (red) is positioned in the GTPase domain adjacent to the nucleotide (shown in gold). (B) SEC-MALS was performed to analyze the multimeric state of the G32A mutant (red) compared with WT protein (blue). The dotted lines shown on the right y-axis indicate the size of the Drp1 multimers based on the predicted molar mass. (C) GTPase activities of WT versus G32A in solution are compared (0.5 μM final for each). Significance was evaluated using a two-way ANOVA. ^***P = 0.0007. Error bars represent the SEM. (D) GTPase activities of WT and G32A are compared in the presence of lipid nanotubes (red) and liposomes (orange). ^****P < 0.0001. (E–J) Negative-stain EM images depicting WT (E–H) and G32A (I–L) in the presence of GMP-PCP (E, I), decorating lipid nanotubes (F, J), tubulating liposomes (G, K) and tubulating liposomes in the presence of GTP (H, L). Solid white arrowheads highlight protein striations on the nanotube resulting from protein decoration. Open arrowheads highlight constriction sites. Scale bar (black), 200 nm in all images.

Altogether, these results suggest that the G32A mutation is defective solely in its ability to hydrolyze GTP. The G32 residue is highly conserved among DSPs and is located in the G1 motif, also called the P-loop, which is responsible for binding to the phosphate backbone of the nucleotide (25). The synthetic K38A mutation in this loop prohibits GTP hydrolysis, similar to other GTPases. The G32A mutation is positioned at the start of this loop, and we reason that disrupting the glycine flexibility at this position impaired the GTPase activity of Drp1. GTP binding affinity assessed using FRET between Drp1 Trp (donor) and mant-GTP (acceptor) revealed a ~5-fold higher apparent K_D for G32A compared with WT (~55 and ~9.4 μM, respectively) indicating that the P-loop mutation weakens GTP binding (Supplementary Material, Fig. S2). Yet, this is not a pronounced difference and, consequently, is unlikely to cause the changes we observed under the saturating conditions (1 mm GTP) used in our enzyme assays. Importantly, the self-assembly interactions of G32A were relatively unperturbed on the lipid templates. This was further demonstrated when imaging G32A in the presence of CL-containing liposomes. G32A was still able to recognize and self-assemble on these LUVs, which showed that it retained membrane remodeling properties similar to WT Drp1 even though its enzymatic activity was not enhanced in the presence of liposomes (Fig. 4G and K). In addition, we examined Drp1 constriction sites induced by GTP hydrolysis after pre-assembly of WT and G32A on liposomes (Fig. 4H and L). For WT, distinct constriction sites could be identified (indicated by open arrowheads), and the average diameter of these tubes after GTP addition measured 44 ± 13 nm (n = 139), consistent with previous EM and light microscopy results (44). In contrast, G32A constriction sites after GTP addition were not apparent and the average size of the tube diameters was 66 ± 13 nm (n = 265, Supplementary Material, Fig. S3). Thus, G32A was unable to constrict tubules as efficiently as WT Drp1 because of limitations in GTP hydrolysis.

Membrane remodeling is impaired by the G223V mutation

Another mutation in the GTPase domain was identified that encoded a G223V mutation (Fig. 5A). This resulted from the c.668G → T base change that was identified in a girl patient, who developed normally until about age 2.5 years, when she presented with seizures and ataxia (19). An MRI performed at age 6 years showed global cerebral and cerebellar atrophy and the patient developed refractory epilepsy. Even though the G223V mutation site is not in a conserved GTP binding motif and is positioned ~18 Å from the nucleotide binding site (Fig. 5A), the rate of GTP hydrolysis in solution was also impaired with this change (0.7 ± 0.2 min⁻¹, Fig. 5C), a ~2.5-fold decrease when compared with WT (1.7 ± 0.3 min⁻¹). For comparison, the G32 residue is 6.5 Å from the nucleotide (Fig. 4A), so the G223 site is located at the periphery of the GTPase domain. The lipid nanotube-stimulated rate for G223V was 2.9 ± 0.2 min⁻¹, compared with 13 ± 1 min⁻¹ for WT in the presence of nanotubes (Fig. 5D). This represents only a ~4-fold increase compared with G223V in the apo state. The GTPase activity of G223V with liposomes was also depressed (2.7 ± 0.1 min⁻¹) compared with WT (10 ± 1 min⁻¹), which represents a similar ~4-fold increase. Thus, the already diminished basal rate of GTP hydrolysis for G223V is augmented by the presence of both preformed and unstructured lipid templates, but the resultant stimulated activity was substantially impaired when compared with WT.

Figure 5.

G223V is impaired in membrane remodeling with mild defects in GTP hydrolysis. (A) The residue G223 (red) is positioned in the GTPase domain relatively far from the nucleotide (shown in gold). (B) SEC-MALS was performed to analyze the oligomeric state of the G223V mutant (red) compared with WT protein (blue). The dotted lines shown on the right y-axis indicate the size of the Drp1 multimers based on the predicted molar mass. (C) GTPase activity in solution for WT and the G223V mutant (0.5 μM final for each). Significance was evaluated using a two-way ANOVA. ^*P = 0.0214. Error bars represent the SEM. (D) Lipid-stimulated GTPase activities are compared with WT and G223V in the presence of lipid nanotubes (red) and liposomes (orange). ^*P = 0.0375. ^****P < 0.0001. ns, not significant. (E–J) Negative-stain EM images depicting WT (E–G) and G223V (H–J) in the presence of GMP-PCP (E, H), decorating lipid nanotubes (F, I) and tubulating liposomes (G, J). Solid white arrowheads highlight protein striations on the nanotube resulting from protein decoration. Scale bar, 200 nm in all images.

Our SEC-MALS analysis for G223V was essentially identical to WT (Fig. 5B) and G32A, demonstrating that G223V was in essentially the same dynamic equilibrium of multimeric states as WT in solution. EM imaging showed that G223V formed rings, but not extended spirals in solution. Therefore, GTPase domain interactions that stabilize ring-to-ring interactions in nucleotide-induced spirals are likely impaired, which may explain the mild decrease in GTPase activity in solution (Fig. 5C). G223V was able to decorate nanotubes similar to WT (Fig. 5F and I), consistent with the enhanced rate of hydrolysis in the presence of lipid nanotubes. When we analyzed this mutant in the presence of liposomes, there was an increase in activity compared with the rate in solution without lipid, but the stimulated activity was ~4-fold less than that observed with WT Drp1. Consistent with this observation, we found that G223V is less efficient in its ability to tubulate liposomes, and the few protein-lipid tubes that were identified were less uniform when compared with WT (Fig. 5G and J). As such, G223V is defective in generating membrane curvature needed to form lipid-induced helical structures.

Discussion

To begin probing the impact of patient mutations on Drp1 structural and biochemical properties, individual point mutations were introduced into our expression construct and each variant was isolated using an established protocol (42). We predicted that each mutant would display enzymatic activity, but many would be diminished relative to WT since the mutations are not embryonic lethal in patients. Indeed, all mutants possess GTPase activity, but mutations in the GTPase domain impair activity. Surprisingly, mutations in the MD generally exhibited a higher basal rate when compared with WT Drp1. Traditionally, mutations in the MD of Drp1 have been automatically assumed to confer a self-assembly defect. However, our studies demonstrate that there are different modes of assembly that require MD engagement to form multimers in solution and larger oligomers in response to nucleotide binding and/or lipid interactions.

Based on the experiments with A395D, R403C and C431Y, assembly-incompetent mutations are associated with changes in the MD (Fig. 6). Each of these three mutants were limited to dimers in solution, were unable to form larger multimers (tetramers and hexamers) based on SEC-MALS and could not self-assemble into spirals or helical polymers when nucleotide or lipid were introduced, respectively. Notably, A395D and R403C hydrolyzed GTP at a ~3-fold higher rate than WT in solution, and C431Y displayed a GTPase activity ~2-fold higher than WT in solution. While the measured changes were not statistically significant, these results suggest that the ability of Drp1 to form larger multimers in solution, such as tetramers and hexamers, limits its basal enzymatic activity (45). A recent study also described novel patient mutations in the MD of Drp1, G363D and G401S (46). Notably, the G363D and G401S mutations were found to be dimer-limited and had elevated rates of GTP hydrolysis in solution (46). Currently, it is unknown whether G401S can self-assemble into larger oligomers, but we predict that this is unlikely based on the position in the same loop at the base of the Drp1 stalk as R403C, which was also mutated in the crystal structure to limit Drp1 assembly properties (residues 401–404, GPRP to AAAA). Based on similarity to other DSPs, it was proposed that the formation of spirals in solution contributed to a higher GTP hydrolysis rate (41). Our findings demonstrate that assembly-limited dimers are incapable of forming spirals when GMP-PCP was added, but they still exhibited an elevated rate in solution. Therefore, the interchange in multimeric state (i.e. dimer to tetramer) impacts Drp1-mediated hydrolysis as well.

Figure 6.

Patient mutations result in defects in unique aspects of Drp1 function. Drp1 (blue) exists in a dynamic mixture of multimeric species in solution. Some patient mutations interfere with this equilibrium by limiting the protein to assembly-incompetent dimers. However, as demonstrated by the F370C mutant, shifting the equilibrium toward more dimers in solution does not necessarily preclude protein lipid recognition and subsequent self-assembly on a lipid template. In fact, both F370C and G223V are able to decorate pre-curved templates but are unable to remodel membranes at constriction sites. The G32A mutation in the GTPase domain limits hydrolysis, which prevents membrane constriction.

Negative-stain EM was used to evaluate the effects of nucleotide and lipid interactions on Drp1 self-assembly into large, helical oligomers. The rings and spirals formed by WT in solution with GMP-PCP have been considered functional mimics of oligomeric structures that form around mitochondria during fission (41). Indeed, WT Drp1 formed spirals in a nucleotide-bound state, but these nucleotide-bound Drp1 oligomers appear to be structurally and functionally distinct from membrane-bound oligomers, which exhibit stimulated enzymatic activity. In agreement, the F370C mutant was unable to form spirals in the presence of GMP-PCP, but it had no apparent defect in assembling on a CL-enriched lipid template. Moreover, this mutant protein was limited to a dimer in solution, which was still capable of assembling into an extended helical lattice on a preformed lipid template. In agreement, previous studies have suggested that cytosolic oligomers must first cycle down to dimers by hydrolyzing GTP or via an alternate mechanism (29,45). Then, dimeric Drp1 can interact with its partners at the OMM. Some MD mutants, A395D, R403C and C431Y, may possess a slightly higher basal rate of GTP hydrolysis compared with WT, in line with previous findings with other dimer-restricted mutants such as G350D and G363D (45,46). However, they were unable to form helices on lipid templates. There are no mutations in the VD, which confers membrane binding, and A395D did exhibit some affinity for lipid templates (Supplementary Material, Fig. S1). Therefore, these mutations likely prevent structural rearrangements required to self-assemble into ordered polymers, rendering them trapped in an assembly-limited state, while retaining high basal GTPase activity from productive dimer–dimer collisions in transient, amorphous complexes (45). Collectively, these results demonstrate that only lipid-induced oligomers exhibit robust stimulated GTPase activity.

It did not escape our attention that in the case of F370C, and also R403C, an additional cysteine residue is being introduced into the Drp1 protein sequence. These mutants, along with all the other proteins used in this study, were maintained in a reduced state using abundant reducing agent (10 mm BME) to avoid protein oxidation and the formation of disulfide bonds between solvent-exposed cysteine residues. Despite this, it could still be possible for solvent-inaccessible residues located in the core of the stalk region to form disulfide bonds. Such bonds could potentially influence Drp1 subunit conformational dynamics and, possibly, the overall shape of the molecule. For the disease-associated cysteine residue introduced at the 370 site, another cysteine residue (C431) is within disulfide bonding distance. Yet, our SEC-MALS results reveal that both the F370C mutant and the C431Y mutant elute similarly as dimers, suggesting that there are no drastic changes in the overall structure of the molecule when cysteines are present or absent. Additionally, a previous study has shown that Drp1 does not form disulfide-based oligomers even though the native sequence contains nine cysteine residues (47). For this reason, we do not anticipate that these mutations would alter Drp1 activity in response to cellular redox state, but we cannot exclude this possibility.

Patients with mutations in the GTPase domain (G32A and G223V) also presented at later stages of development despite the essential role for Drp1 enzymatic activity in mitochondrial fission. Additionally, we have previously demonstrated that both stalk and GTPase domain dimerization are necessary for Drp1 oligomerization, and interactions between adjacent GTPase domains enhance the enzymatic activity of the protein (39). Consistent with their position in the GTPase domain, we observe that G32A and G223V exhibit decreased GTPase activity in solution compared with WT. Still, they retain some enzymatic function, which is less severe when compared with synthetic mutations identified in conserved G boxes. For G32A, the enzymatic defect is more detrimental, likely owing to its position at the beginning of the P-loop. However, the assembly properties of the mutant still allow it to bind lipid templates, form oligomers on preformed nanotubes and remodel liposomes into ordered tubules similar to WT. Even so, the catalytic activity was not robustly stimulated by oligomerization on these lipid species, highlighting that the G32A defect is primarily enzymatic. Moreover, this defect prevents robust constriction of liposomes (Supplementary Material, Fig. S2), which highlights the critical role of GTP hydrolysis in membrane remodeling. Notably, there is another mutation located within a conserved motif that has been recently identified. The G149R mutation is positioned in the G3 motif, or Switch 2 region, of the GTPase domain (48). Like the P-loop where the G32 site is located, Switch 2 is also highly conserved among DSPs and the G149 residue in particular is responsible for binding the γ-phosphate during hydrolysis. Thus, we predict that G149R is likely defective in hydrolysis in a similar manner as G32A. However, the replacement of a flexible glycine residue with a bulky, positively charged arginine residue may also lead to additional complications.

The G223V mutation is not in a position that would obviously impact GTP binding or hydrolysis. Still, the decrease in activity in solution may reflect changes in local conformational sampling because of the change from a flexible glycine to a more structured valine residue adjacent to Switch I. Similar to G32A, this mutant was able to form ordered helices on preformed lipid templates, but for G223V, this decoration was accompanied by a modest (4-fold) increase in GTPase activity. Given the mild decrease in solution activity and the partial recovery on a pre-curved template, we questioned whether this mutation might exhibit a defect in membrane tubulation because of weakened intermolecular interactions. Indeed, G223V was unable to form regular tubules with CL-enriched liposomes. This observation suggests that the major defect for this mutant is likely an inability to reshape membranes through intermolecular GTPase domain interactions. Importantly, this mutation, along with F370C, highlights a critical role for Drp1 in inducing membrane curvature. Without this ability, the G223V protein cannot mediate functional fission. While other factors in the cell, including cytoskeletal proteins and ER-mitochondria interactions (38), can help shape mitochondrial fission sites, the ability of Drp1 to impart and stabilize membrane curvature is a critical role preceding the GTP-induced membrane constriction.

It is important to note that all the mutations included in this study were found to be heterozygous in patients. Previous studies suggested the G32A, A395D and R403C mutations are likely dominant-negative (15–17). In the studies wherein the other three mutations were first identified, it was unclear whether the mutations were dominant-negative, although the C431Y mutation could be as well (15). In future studies, the impact of these mutations on WT Drp1 function should be examined. Still, the observed dominant-negative effects are most likely related to the inherent functional defects identified in this study.

Ultimately, this work elucidates important structure–function relationships for Drp1 and demonstrates the variability the exists between distinct mutation sites. In fact, the specific functional defect caused by each mutation cannot simply be presumed based on the domain in which it is located. Going forward, we hope that this study serves as a foundation for determining the specific defects associated with individual mutations in Drp1.

Materials and Methods

Protein constructs and mutagenesis

Drp1 Isoform 3 (UniProt ID O00429-4) was cloned into the pCal-N-EK vector as described previously (37,42). Site-directed mutagenesis was completed to introduce each mutation into this construct individually using the QuikChange Lightning kit (Agilent) with primers from Integrated DNA Technologies (Coralville, IA). The Drp1 monomer structure was generated by AlphaFold (49,50), and this monomer was aligned to the dimer interface in the crystal structure to ensure that features in this model are consistent with available structures (PDB IDs: 4BEJ and 5WP9 and EMDB ID: EMD-8874) (34,43). The VD (residues 503–637) was not shown since no structural features were identified by this software or resolved in any structure.

Protein expression and purification

All Drp1 constructs were expressed in BL21-(DE3) Star Escherichia coli. Cells were grown in LB containing 100 μg/mL ampicillin at 18°C with shaking at 200 rpm for 24 h after induction with 1 mm isopropyl-1-thio-β-d-galactopyranoside. Then, cells were harvested via centrifugation at 4300 × g for 20 min at 4°C. The resulting pellet was resuspended in CalA Buffer (0.5 M L-Arginine pH 7.4, 0.3 M NaCl, 5 mm MgCl₂, 2 mm CaCl₂, 1 mm imidazole, 10 mm β-mercaptoethanol) with 1 mm Pefabloc-SC and 100 μg/mL lysozyme. Cells were lysed by sonication on ice. Next, the cell debris was pelleted via centrifugation at 150 000 × g for 1 h at 4°C. First, the CBP-tagged Drp1 was purified by affinity chromatography using calmodulin agarose resin (Agilent) that had been pre-equilibrated with CalA Buffer. After the supernatant was loaded onto the column, the resin was washed with 25 column volumes of CalA Buffer. Next, eight fractions of eluent were collected using 0.5 column volumes of CalB Buffer (0.5 M L-Arginine pH 7.4, 0.3 M NaCl, 2.5 mm EGTA, 10 mm β-mercaptoethanol). Protein-containing fractions were pooled and incubated with GST-tagged PreScission Protease (HRV-3C) overnight at 4°C to remove the CBP-tag. This solution was concentrated using a 30 000 molecular weight cut-off centrifugal filter (Amicon). This concentrated pool of Drp1 was further purified by size exclusion chromatography (SEC) with an ÄKTA Purifier FPLC (GE Healthcare) and a HiLoad 16/600 Superdex 200 Prep Grade column that had been pre-equilibrated with SEC Buffer (25 mm HEPES (KOH) pH 7.5, 0.15 M KCl, 5 mm MgCl₂, 10 mm β-mercaptoethanol). All elution fractions containing Drp1 were pooled and concentrated once again, and glycerol (5% final) was added. The purified Drp1 was aliquoted, flash frozen in liquid nitrogen, and stored at −80°C until use.

Malachite green colorimetric assay

The basal GTPase activity of Drp1 was measured using a colorimetric assay to detect released phosphate, as described previously (37,42). Briefly, Drp1 (500 nM final) was diluted to 2.4X with Assembly Buffer (25 mm HEPES (KOH) pH 7.5, 150 mm KCl, 10 mm β-mercaptoethanol). To start the reaction, 3X GTP/MgCl₂ (1 and 2 mm final, respectively) was added to the Drp1 with either 4X lipid (150 μM final) to calculate the lipid-stimulated rates or only Assembly Buffer to calculate the rate for the protein alone in solution. The reaction was carried out at 37°C. At the chosen time points, a sample aliquot was taken and quickly added to EDTA (100 mm final) to stop the reaction. After collecting all time points, Malachite Green Reagent (1 mm malachite green carbinol, 10 mm ammonium molybdate tetrahydrate, 1 N HCl) was added to each sample, and the absorbance at 650 nm was measured using a VersaMax microplate reader (Molecular Devices). The raw absorbance values obtained, which correspond to phosphate levels, were used to calculate rates in Excel (Microsoft). These rates were then used to determine the k_obs when taking the Drp1 concentration into account. The resulting k_obs values were plotted with GraphPad Prism 8. Statistical significance was calculated using a two-way ANOVA.

Size exclusion chromatography with multi-angle light scattering

SEC-MALS experiments were performed as before (37). Briefly, 5 μM Drp1 in a volume of 500 μL was injected onto a Superose 6 10/300GL column in an ÄKTApure FPLC system (GE Healthcare) connected in line with DAWN Heleos-II 18-angle MALS and Optilab T-rEX differential refractive index detectors from Wyatt Technology. Data were analyzed with ASTRA 7 software from Wyatt Technology.

Lipid nanotube and liposome (LUV) preparations

All lipid nanotubes utilized here were comprised by 40% D-galactosyl-beta-1’-N-nervonyl-erythro-sphingosine (GC), 35% phosphatidylethanolamine (PE), 25% bovine heart CL molar fractions. All liposomes consisted of 40% PC, 35% PE and 25% CL. All lipids were purchased from Avanti Polar Lipids, Inc. (Alabaster, AL). Lipids were added to a glass test tube and slowly dried to a thin film using nitrogen gas. The lipid film was then stored in a desiccator overnight to ensure any trace solvent remaining was removed. The following day, the lipid film was rehydrated with a buffer (200 μL for nanotubes and 250 μL for liposomes) containing 50 mm HEPES (KOH) pH 7.5 and 0.15 M KCl and heated in a 37°C water bath for ~40 min with gentle vortexing every 10 min. With these volumes, the final concentration for both the nanotubes and liposomes was 2 mm. For nanotubes, the lipid film was placed in a water bath sonicator for 30 s and the resulting nanotubes were stored on ice until use. For liposomes, three freeze-thaws were performed with liquid nitrogen prior to usage.

Negative-stain electron microscopy

All samples were added to carbon-coated grids and stained using 2% uranyl acetate. Each grid was made using 2 μM protein and either 1 mm GMP-PCP or 150 μM nanotubes or liposomes as indicated. Sample images were acquired on either a Tecnai T12 or TF20 electron microscope (FEI Co.) at 100 or 200 keV, respectively. The T12 was equipped with Gatan Eagle (4 × 4 k) camera and images were acquired at a magnification of ×20 000. The TF20 was equipped with TVIPS F-416 CMOS (4 × 4 k) camera and images were acquired at a magnification of ×30 000.

Conflict of Interest statement

None declared.

Funding

The National Institutes of Health (T32 GM008803 to B.L.B., F31 GM139324 to K.R., R01 GM121583 to R.R., R01 CA208516 and R01 GM125844 to J.A.M.).

Data Availability

The authors affirm that all data necessary for confirming the conclusions of this article are represented fully within the article and its tables and figures.