Home » Blog » Structural basis for regulated assembly of the mitochondrial fission GTPase Drp1

Structural basis for regulated assembly of the mitochondrial fission GTPase Drp1

by bikku19792
0 comment

[ad_1]

Identification of WT Drp1 dimer structure

Drp1 has been shown to exist as a mixture of dimers and tetramers in solution in a concentration-dependent manner, and the dimer state represents the core unit of the larger helical machinery16. For this study, WT Drp1 (isoform 1) was expressed in and isolated from E. coli using established methods17. This particular isoform is the second longest and includes the B insert (exons 16 and 17). It was selected for the cryoEM experiments since it has previously been shown to more dimeric when compared with other splice variants2. Dimers were isolated for study by diluting the protein to concentrations that would be enriched for dimers (600 nM). Using cryo-EM, the structure of a full-length dimer of WT human Drp1 was resolved to a reported resolution of 5.97 Å and a 3DFSC resolution of 6.07 Å, which prevents the identification of side chains but secondary structure can be observed in regions of the map (Supplementary Table 1, Supplementary Fig. 1). Four conformations were identified with significant conformational heterogeneity conferred primarily through GTPase motions that highlight variability in the position of this domain relative to the stalk. Conformation 1 generated the highest quality map with the most particles, the best fit, and is used as the model dimer after applying a low pass filter of 6 Å to remove over-fitting artifacts (Fig. 1b, c). In this map both BSEs are resolved, generating the most complete model with the highest confidence. The second G domain density is not fully resolved in any of the conformations. We believe this is due to the heterogeneity of each G domain relative to one another. Conformations 2-4 provide insight in the heterogeneity in the dimer interface; however, resolution and map quality suffered due to G domain flexibility.

Overall, the architecture of the model dimer presents a compact organization of the GTPase and stalk domains (Fig. 1d, Movies 1 and 2), resulting in a ~ 100 Å decrease in length of the solution dimer when compared to the length of the crystal structure dimer with the VD deletion and GPRP401-404AAAA mutation (Fig. 1e, Supplementary Fig. 1f). The compaction of Drp1 dimers, as compared to the extended state described in the crystal structure, is achieved through hinge motions. Specifically, hinge 1, comprised of two disordered loops that connect the BSE to the distal end of the stalk, adopts a conformation that places the GTPase domain near its own stalk (Fig. 1). Separately, a pivot between adjacent stalks in the homodimer increases the number of intermolecular contacts as compared to those in the crystal structure.

The largest rearrangement when comparing the solution Drp1 to previous structures was observed in hinge 1. A local resolution of 5.5 Å was observed in this region (Supplementary Fig. 1c), allowing flexible fitting of the crystal structure to the EM density. In comparison with the crystal structure, the fit revealed that the G domain is capable of significant rearrangement to bring the GTPase domain helix α2G adjacent to loop L1NS in the stalk (Fig. 1d). This loop is important for Drp1 assembly, as mutations in this conserved region limit Drp1 and other DSPs ability to assemble to anything larger than a dimer4,16,18,19. The compaction of the GTPase domain against this loop would prevent assembly beyond a dimer, representing a conformational change compared to helical and filament assemblies that exhibit an extended G domain conformation to expose conserved self-assembly interfaces.

The residues that form the discrete dimer interface identified by the Daumke group remain at the core of this dimer conformation;9 however, the density reveals additional burial of surface area between adjacent stalks in this region. Additional residue contacts are formed through stalk motions that close the angle between the monomers. The orientation of head and stalk is generally heterogenous and asymmetric, suggesting that the hinge motion is dynamic in the solution state of the Drp1 homodimer (Supplementary Fig. 1). The major difference in the distinct dimer conformations can be attributed to a “pivoting” motion around the fulcrum of the interface that results in altered angles between adjacent stalks. The solution dimer conformation is more closed, further obscuring potential stalk interfaces. The previous structures in the crystal lattice or assembled helical polymers likely represent an “open” conformation where the core dimer interface is maintained but the peripheral stalk regions are exposed, removing auto-inhibitory interactions that limit assembly in these regions.

BSE lock through loop L3S

Comparing the autoinhibited dimer structure to the published crystal structure, the G domain of the cryo-EM structure is positioned 79 Å closer to the distal end of the stalk. This change in position is accompanied by a 67° rotation and a 61° twist of the G domain (Fig. 2a, b). This results in a “locked” conformation of the G domain against the distal end of the stalk, mediated through interactions between the BSE and loop L3S (Fig. 2c, Movie 3). Absent side chain resolution to identify critical contacts stabilizing this conformation, mutagenesis was pursued to further examine the role of this loop. L3S was not fully resolved in the crystal structure; however, a Drp1-MiD49 filament structure4 identified a small helical segment in the middle of the loop at residues 452-456 QELLR, and AlphaFold predicted a helix in this region as well4,20. A comparison of available DSP structures reveals that mitochondrial fission DSP loops have additional residues (448-451 NYST) when compared with other DSPs (Supplementary Fig. 2c–e). This additional sequence may contribute to the BSE lock, and a similar conformational rearrangement was observed by the Low group in a crystal structure tetramer of Cyanidioschyzon merolae Dmn1 (cmDmn1)21, the red algae mitochondrial fission dynamin (Supplementary Fig. 2e). Dynamin has a shorter loop with a more extended α2S helix, so it remains unclear whether the BSE lock is conserved in all DSPs. If this feature is unique to mitochondrial fission DSPs, this region would affect the wide range of BSE motions depending on its activation state (Supplementary Fig. 2f).

Fig. 2: The BSE lock.
figure 2

a, b Comparing the previous crystal structure9 (PBD ID: 4BEJ, blue) to the dimeric cryo-EM structure (orange) based on overlay of the stalk region. c Loop 3 (L3s) interacts with BSE helix α2b and contains R456. d Negative stain electron microscopy characterization of WT and R456E Drp1 in the presence and absence of non-hydrolysable GMP-PCP and CL-containing nanotubes (CLnts). Multiple grids were made and several images were collected for each condition (WT apo=8, WT + GMP-PCP = 38, WT+CLNTs = 44, R456E apo = 4, R456E + GMP-PCP = 21, R456E+CLNTs=48). Scale bar = 100 nm. e, f Sedimentation assays were used to quantify changes in polymerization based on the relative percent of protein detected in the supernatant (S) and pellet (P) fractions. Data are presented as mean values +/− SEM, and a two-tail t-test was used to determine statistical significance. Each dot represents an experimental replicate (WT(blue) = 9, R456E(pink) = 8. ** P = 0.006). g SEC-MALS analysis of Drp1 WT (blue) and Drp1 R456E (pink) is presented with multimeric states indicated. h GTPase activity was determined for WT (blue) and Drp1 R456E (pink) alone and stimulated with CLnts. Data are presented as mean values +/− SEM, and a two-tail t-test was used to determine statistical significance. Each dot represents an experimental replicate (WT (apo)=12, WT (CLnts)=10, R456E (apo and CLnts)=12. ****P < 0.0001). i, j MitoTracker Orange was used assess mitochondrial morphology in WT MEFs compared to Drp1 knock-out MEFs transfected with an empty pCMV vector (EV) and pCMV vectors containing Drp1 WT and Drp1 R456E. Scale bar 5 µm, inset 10 × 10 µm. Mitochondrial morphology was quantified using a blinded assessment described in the methods section. Two independent experiments were performed to assess the percentage of cells with defined mitochondrial morphologies in each sample (Total cells counted: WT MEF (dark gray) = 60, EV (light gray) = 64, KO+Drp1 (WT, blue) = 30, KO + R456E (pink) = 33).

To determine the effect of this small helical segment in L3S on the activity of Drp1, a charge reversal was introduced through a mutation, R456E, to disrupt this lock and promote the opening of the BSE away from the stalk. In a previously published structure, GMP-PCP-bound complex of Drp1and MiD 49 reveals an extension of the G domain away from L3S (Supplementary Fig. 2f), showing the range of BSE extension possible in this region of Drp14. This use of a non-hydrolysable nucleotide locks the protein in a nucleotide-bound state that promotes self-assembly state. Indeed, when GMP-PCP was added to Drp1 R456E, the protein had an increased propensity to assemble into spiral polymers compared to WT, consistent with this mutation opening the auto-inhibited lock (Fig. 2d). Negative stain images show that after 2 h of incubation, WT forms predominantly rings with short spirals averaging 0.1 µm in length. In comparison, R456E was observed to have a 10-fold decrease in the abundance of rings and a 2.6-fold increase in spiral length (Supplementary Fig. 2g–i). In agreement, 77% of the R456E protein sedimented in the presence of GMP-PCP as compared to only 52% for WT protein, further suggesting that R456E favors assembly in a GTP-bound state (Fig. 2e). The decoration on cardiolipin-containing nanotubes (CLnts) observed by negative stain was similar to WT (Fig. 2d), so the organization of the polymer does not appear to be affected by this mutation; rather, the mutation most likely results in a protein conformation that is more poised to assemble in response to nucleotide binding or lipid interactions with the VD. Importantly, there was no appreciable aggregation with the WT or R456E (Fig. 2e, f) when not in the presence of a nucleotide, and there was no shift in its solution multimeric state when assessed using size-exclusion chromatography coupled to multi-angle light scattering (SEC-MALS, Fig. 2g). Together, these data show the BSE lock did not impact the solution multimer formation. Instead, the charge reversal in L3S weakens this self-regulatory BSE lock, priming the protein for assembly. The lipid sensing regions of the variable domain and the stalk interfaces required to build the helical assembly were not perturbed.

To complement the assembly assay, enzymatic activity was measured using an endpoint malachite green assay (Fig. 2h). R456E (0.32 min1) exhibited a five-fold decrease in basal activity compared WT (1.7 min1). This change likely reflects conformational differences in the protein since the multimer state in solution is not altered. The stimulated activity of R456E in the presence of CLnts (16.8 min1) was still lower compared to WT (35.1 min1), but there was significant stimulation for the R456E when compared to its basal rate demonstrating its ability to form higher order oligomers. This is consistent with the assembly observed in the presence of GMP-PCP. Again, this mutant is more assembly-potent, and the lipid-bound structures look indistinguishable from WT (Fig. 2d), though we cannot discount small differences that affect this stimulation.

To test the effect of the mutation on mitochondrial fission in cells, we transfected Drp1 knock-out (KO) mouse embryotic fibroblasts (MEFs) with WT Drp1 and R456E (Fig. 2i). When transfected with WT Drp1, cellular mitochondrial networks were mostly fragmented due to the over-expression of the fission protein. Cells transfected with R456E were observed to have fused mitochondria, similar to the KO cells treated with an empty vector (EV) control (Fig. 2j, Supplementary Fig. 2j). Additionally, the R456E mutant protein in these transfected cells formed aggregated puncta and did not exhibit a diffuse signal observed with the WT Drp1 vector (Supplementary Fig. 2k). This likely represents premature assembly of the R456E mutant in cells, suggesting that this BSE lock is critical to sustain an auto-inhibited state that prevents premature assembly of Drp1 polymers. Only when this autoinhibited state is relieved, or unlocked, does the WT protein become primed to form a helical assembly around the outer mitochondrial membrane.

A flexible dimer interface

The dimer interface that orients the monomers relative to one another was found to have a large amount of heterogeneity within the conformations and among available structural data. The crystal structure exhibits a discrete dimer interface with a more acute lateral angle measuring 85° (Fig. 3a). The solution structure presented here suggests that removing the VD and introducing the poly-A mutation within L2S near the membrane proximal interface (previously labeled interface 3) yielded an open conformation more amenable to crystallization. These changes also disrupted key regulatory regions, leading to an “open” conformation and the stalk orientations are consistent with previously reported DSP helical conformations. Within the cell, interactions with lipids, ER contact sites, post-translational modification(s), and partner proteins interactions could all promote an open state alone or in concert with one another. In the solution (i.e., “closed”) state, L2S is juxtaposed to α1Ns, and this interaction requires a conformation with an obtuse angle between adjacent stalks (ranging from 103 to 145°) to form a more continuous interface (Supplementary Fig. 3a). In order to identify intra-monomer conformational rearrangements, the four helices comprising the stalk of the solution structure were aligned with the crystal structure helices (Fig. 3b). Loops within α1 confer a large degree of flexibility, and this is evident when comparing the different chains of the crystal structure (Supplementary Fig. 3b). Helices 2 and 4 were relatively unchanged, and Helix 3 had a shift in alignment at the C-terminal end due to flexion in the middle of the helix. A conserved tyrosine (Y493) was found to bookend the dimer interface based on optimal placement of this helix in the density (Fig. 3c, d, Movie 4). To alter the interactions at this position, mutagenesis substituted a smaller alanine in place of the bulkier tyrosine. As evidenced by negative-stain EM analysis, this mutation showed a decreased ability to form spirals in the presence of GMP-PCP and was unable to uniformly decorate CLnts when compared to WT (Fig. 3e). This finding was further quantified using a sedimentation assay. In apo conditions, both WT and Y493A resulted in similar levels of protein found in the pellet (17% and 20% respectively). No assembly of Y493A was detected when GMP-PCP was added to induce spiral formation, as only 19% of the Y493A sediments with GMP-PCP. This represents no change compared to average sedimentation under apo conditions, while WT Drp1 had a three-fold increase in pelleted protein (52%). A decrease was also observed in the ability of Y493A to decorate CLnts (42% of Y493A was found in the pellet compared to 60% for WT), confirming that this mutant prevents assembly even though lipid binding is likely preserved since the VD is unchanged. This is not surprising, considering the importance of the dimer state, the functional unit of Drp1. If a stable dimer interface is disrupted, the protein will be incapable of forming larger assemblies. Since no significant change in sedimentation was observed when comparing the WT and Y493A Drp1 in the absence of assembly inducers (Fig. 3f, g), SEC-MALS was used to assess the multimer state in solution. Y493A was found to exist in equilibrium between dimers and monomers and was not able to form larger multimers observed with WT Drp1 at the same concentration (Fig. 3h). Knowing that the multimer state for Drp1 is concentration dependent, mass photometry was used at a lower concentration (100 nM) to confirm this difference, and Y493A was found to exist largely as a monomer while WT protein was mostly dimeric (Supplementary Fig. 3c).

Fig. 3: Dimeric cryo-EM structure dimer interface.
figure 3

a Comparison of the stalk orientations in the crystal structure dimer interface, (PDB code: 4BEJ, dark blue – chain A, light blue – chain B) and the dimeric cryo-EM structure (dark orange – chain A, light orange – chain B). b Alignment of each helix comprising the stalk domain (PDB code: 4BEJ, blue; cryo-EM structure, orange). Y493 is highlighted in light green. c Helices forming the dimer interface are fit within the cryo-EM density (chain A, orange; chain B, yellow). Y493 (green) bookends the interface. d Alignment of crystal structure dimer and dimeric cryo-EM structure stalks, reference chain is chain A. Zoomed in view, Y493 in light green. e Negative stain electron microscopy characterization of WT and Y493A Drp1 in the presence and absence of non-hydrolysable GMP-PCP and CL-containing nanotubes (CLnts). Multiple grids were made and several images were collected for each condition (WT apo=8, WT + GMP-PCP = 38, WT+CLNTs = 44, Y493A apo = 9, Y493A + GMP-PCP = 25, Y493A+CLNTs=46). Scale bar = 100 nm. f, g Sedimentation assays were used to quantify changes in polymerization based on the relative percent of protein detected in the supernatant (S) and pellet (P) fractions. Data are presented as mean values +/− SEM, and a two-tail t-test was used to determine statistical significance. Each dot represents an experimental replicate (WT (blue) = 9, Y493A (green) = 8.***P = 0.0005; *P = 0.04). h SEC-MALS analysis of Drp1 WT (blue) and Drp1 Y493A (green) is presented with multimeric states indicated. i GTPase activity was assessed for WT (blue) and Drp1 Y493A (green) alone and stimulated with CLnts. Data are presented as mean values +/− SEM, and a two-tail t-test was used to determine statistical significance. Each dot represents an experimental replicate (WT (apo) = 12, WT (nts) = 10, Y493A (apo) = 12, Y493A (nts)=11. ****P < 0.0001). j, k MitoTracker Orange was used assess mitochondrial morphology in WT MEFs compared to Drp1 knock-out MEFs transfected with an empty pCMV vector (EV) and pCMV vectors containing Drp1 WT and Drp1 Y493A. Scale bar 5 µm. Inset 10 µm. Mitochondrial morphology was quantified using a blinded assessment described in the methods section. Two independent experiments were performed to assess the percentage of cells with defined mitochondrial morphologies in each sample (Total cells counted: WT MEF (dark gray) = 60, EV (light grey) = 64, KO+Drp1 WT (blue) = 30, KO + Y493A (green) = 29).

The GTPase activity was assessed for Y493A and WT Drp1 (Fig. 3i), and the basal rates in solution were comparable (2.3 min1 for Y493A versus 2.4 min1); however, CLnts stimulation was only observed with the WT protein (35 min1), while Y493A was not stimulated (1.2 min1). Therefore, Y493A maintained basal GTPase activity but was incapable of functional assembly.

Transfecting Drp1 KO MEF cells with Y493A saw no change in the interconnected, fused mitochondrial morphology when compared to the empty vector control. Conversely, WT Drp1 transfection resulted in a fragmented mitochondrial network (Fig. 3j, k, Supplementary Fig. 3d). This observation is consistent with the destabilization of the continuous dimer interface by the Y493A mutation that was introduced, which prevents functional assembly and limits mitochondrial fission in cells. Therefore, the Y493 residue is critical for stabilizing interface 2, bookending either side of the interface to accommodate flexibility from a continuous interface to a more discreet intermolecular interaction required for spiral and helical oligomerization.

[ad_2]

Source link

You may also like

Leave a Comment