A molecular switch that governs mitochondrial fusion and fission mediated by the BCL2-like protein CED-9 of Caenorhabditis elegans

Abstract

Depending on the cellular context, BCL2-like proteins promote mitochondrial fusion or fission. What determines which of these two opposing processes they promote has so far been unknown. Furthermore, the mechanisms through which BCL2-like proteins affect mitochondrial dynamics remain to be fully understood. The BCL2-like protein CED-9 of Caenorhabditis elegans has previously been shown to promote mitochondrial fusion by physically interacting with the mitochondrial fusion protein FZO-1. Here, we report that CED-9 also physically interacts with the mitochondrial fission protein DRP-1 and that this interaction can be enhanced when CED-9 is associated with the BH3-only protein EGL-1. In addition, we show that the EGL-1–CED-9 complex promotes mitochondrial fission by recruiting DRP-1 to mitochondria and that the egl-1 gene is required for CED-9-dependent mitochondrial fission in vivo. Based on these results, we propose that EGL-1 converts CED-9 into a mitochondrial receptor for DRP-1, thereby shifting its activity from profusion to profission. We hypothesize that BCL2-like proteins act as mitochondrial receptors for DRP-1–like proteins in higher organisms as well and that BH3-only proteins play a general role as modifiers of the function in mitochondrial dynamics of BCL2-like proteins. We speculate that this function of BCL2-like proteins may be as couplers of mitochondrial fusion and fission.

Mitochondria are highly dynamic organelles that constantly fuse and divide. Members of the family of dynamin-related GTPases (Drps) are critical for mitochondrial fusion and fission (1–8). Specifically, Fzo1 and Mgm1 of Saccharomyces cerevisiae; Mfn1, Mfn2, and Opa1 of mammals; and FZO-1 and EAT-3 of Caenorhabditis elegans are required for the fusion of the outer and inner mitochondrial membrane, respectively. Conversely, Dnm1 of S. cerevisiae, Drp1 of mammals, and DRP-1 of C. elegans are required for mitochondrial fission. Drps required for mitochondrial fusion are transmembrane proteins whereas Drps required for mitochondrial fission are cytosolic proteins that are recruited to mitochondria to promote fission. The recruitment of S. cerevisiae Dnm1 involves three different sets of proteins—Fis1-Mdv1, Fis1-Caf4, and Num1, among which only Fis1 is conserved in mammals and C. elegans (9–12). However, the roles of mammalian Fis1 and the two C. elegans paralogues FIS-1 and FIS-2 in the recruitment of Drp1/DRP-1 to mitochondria remain unclear. Specifically, mammalian Fis1-KO cells do not show any specific mitochondrial morphology defects (13). Similarly, mutations in the C. elegans fis-1 or fis-2 gene fail to alter mitochondrial morphology (14). Hence, how Drp1-like proteins are targeted to mitochondria in higher eukaryotes remains unclear.

BCL2-like proteins, which are crucial regulators of apoptosis, have recently been implicated in the control of mitochondrial dynamics (15). For example, the BCL2-like proteins BAX and BAK of mammals and CED-9 of C. elegans play critical roles in Drp1/DRP-1–dependent mitochondrial fission in apoptotic cells (16–20). More recently, the same BCL2-like proteins have been shown to also promote mitochondrial fusion through their interactions with Mfn1, 2/FZO-1 (21–24). Furthermore, the mammalian BCL2-like protein BCL-xL interacts with Drp1 in neurons and can affect the rate of both mitochondrial fusion and fission (25, 26). Together, these observations suggest that BCL2-like proteins are regulators of mitochondrial dynamics that can have both profission and profusion activities. However, the molecular mechanism that determines whether they promote mitochondrial fission or fusion remains to be elucidated.

C. elegans CED-9 localizes predominantly to the outer mitochondrial membrane and plays a critical role in apoptosis (27–30). By binding to a dimer of the APAF1-like adaptor protein CED-4, CED-9 blocks the activation of the procaspase proCED-3 in cells that are programmed to live. Conversely, in cells that are programmed to undergo apoptosis, the binding of the proapoptotic BH3 (BCL2 homology domain 3)-only protein EGL-1 to CED-9 causes a major conformational change of CED-9, which results in the release of the CED-4 dimer and, consequently, CED-3–dependent cell death. We have previously demonstrated that CED-9 also plays a role in mitochondrial dynamics and promotes both mitochondrial fission and fusion (18, 24). As CED-9 is known to form complexes with CED-4 and EGL-1 (31–36), we investigated whether CED-4 or EGL-1 affects the ability of CED-9 to promote either one of these two processes. Based on the results presented here, we propose that EGL-1 acts as a molecular switch that converts CED-9 from a protein with profusion activity to a protein with profission activity and that the EGL-1–CED-9 complex acts as a mitochondrial receptor for DRP-1.

Results

CED-9 Physically Interacts with DRP-1.

Mitochondria undergo fragmentation in apoptotic cells in C. elegans, and this fragmentation is dependent on DRP-1 and CED-9 (18). Therefore, we investigated whether CED-9 can physically interact with DRP-1. By using GST pull-down experiments, we found that in vitro translated, [³⁵S]methionine-labeled DRP-1 protein copurifies with GST-tagged CED-9 (GST::CED-9) protein but not with GST alone (Fig. 1A and Fig. S1). Conversely, in vitro translated, [³⁵S]methionine-labeled CED-9 copurifies with GST-tagged DRP-1 (GST::DRP-1) but not GST alone (Fig. 1B) (24). To validate these results in vivo, with the use of an anti–CED-9 antibody (24), we immunoprecipitated CED-9 from embryonic lysates generated from transgenic animals expressing the WT ced-9 gene under the control of a heat-inducible promoter [P_HSced-9(wt)]. The level of CED-9 protein expressed is comparable in all the transgenic lines used throughout our studies (Fig. S2). By using an anti–DRP-1 antibody (Fig. S3), we found that endogenous DRP-1 coimmunoprecipitates with CED-9 (Fig. 1C). We estimate that in these experiments, the anti–CED-9 antibody immunoprecipitated most of the CED-9 protein (>70%). However, only approximately 1.2% of the total DRP-1 protein coimmunoprecipitated with CED-9. As shown later, most CED-9 but only approximately 10% of DRP-1 localizes to a mitochondria-enriched cell fraction (Figs. S4 and S5). Hence, CED-9 interacts with approximately 12% of mitochondria-associated DRP-1. In contrast, an unrelated mitochondrial protein, the protein bc1 (a component of complex III of the electron transport chain), failed to coimmunoprecipitate with CED-9 (Fig. 1C). Furthermore, an anti-GFP antibody failed to precipitate DRP-1 (Fig. 1C). Based on these results we conclude that CED-9 can physically interact with DRP-1 in vitro and in vivo and that this interaction is most probably direct.

Fig. 1.

CED-9 physically interacts with DRP-1. (A and B) CED-9 and DRP-1 interact in vitro. Purified GST::CED-9 (10 μg/mL), GST (10 μg/mL), or GST::DRP-1 (5 μg/mL) was used in pull-down experiments in the presence of in vitro translated [³⁵S]methionine-labeled DRP-1, luciferase, or CED-9. (C) CED-9 and DRP-1 interact in vivo. Embryonic lysates from animals expressing P_HSced-9(wt) were immunoprecipitated by using anti–CED-9 (IP CED-9) or anti-GFP (IP GFP) antibodies and analyzed by Western blot using anti–DRP-1, anti-bc1, and anti–CED-9 antibodies. (D) CED-9 interacts with the middle domain of DRP-1. (Left) Schematics of truncated DRP-1 proteins. The GTPase domain, the predicted coiled-coil regions (CC1 and CC2), the middle domain, and the GED domain are indicated. (Right) GST::CED-9 pull-down experiments using truncated DRP-1 proteins. The average relative percentages of truncated DRP-1 proteins pulled down by GST::CED-9 compared with full-length DRP-1 are indicated (three data sets from three independent experiments). (E) Mutations in the BH1 and the BH2 domain of CED-9 affect the ability of CED-9 to interact with DRP-1. (Left) Schematic of different CED-9 mutant proteins. (Right) Pull-down experiments using DRP-1 and GST::CED-9(wt), GST::CED-9(N158A,A159G,Q160A), or GST::CED-9(R211E,N212G). The average relative percentages of DRP-1 pulled down by mutant GST::CED-9 proteins compared with GST–CED-9(wt) are indicated (six data sets from two independent experiments). The statistical significance was determined by using the Student t test.

To identify the domains of DRP-1 required for binding to CED-9, we tested fragments of the DRP-1 protein for their ability to bind to CED-9 in vitro. As shown in Fig. 1D, we found that DRP-1 interacts with CED-9 through its middle domain (amino acids 280–502), which is not predicted to adopt any particular structure; however, it has been shown that this domain is required for the ability of S. cerevisiae Dnm1p to form oligomers (37, 38). Interestingly, we found that the interaction between the middle domain alone (amino acids 280–502) or in combination with the GED domain (amino acids 280–705) and CED-9 was increased compared with the interaction between full-length DRP-1 and CED-9, which suggests that the GTPase domain of DRP-1 antagonizes the interaction between DRP-1 and CED-9. Furthermore, we found that two CED-9 mutants, CED-9(R211E, N212G) and CED-9(N158A, A159G, Q160A), which we previously analyzed for their abilities to interact with FZO-1 (24), have significantly reduced capabilities to interact with DRP-1 in vitro (Fig. 1E). Amino acids 158, 159, and 160 and amino acids 211 and 212 are located in the BH1 and BH2 domain of CED-9, respectively. Therefore, the BH1 and BH2 domain of CED-9 and the middle domain of DRP-1 may play an important role in the interaction between CED-9 and DRP-1.

EGL-1 But Not CED-4 Affects Interactions of CED-9 with DRP-1 and FZO-1.

Next we tested whether EGL-1 or CED-4, known binding partners of CED-9, affect CED-9's ability to interact with DRP-1. To test this, we preincubated in vitro translated, [³⁵S]methionine-labeled CED-9 in the presence or absence of in vitro translated, [³⁵S]methionine-labeled EGL-1 (EGL-1-CED-9; 1:1 molar ratio) or CED-4 (CED-9–CED-4; 1:2 molar ratio). We then performed in vitro GST pull-down experiments by using GST-tagged DRP-1. As shown in Fig. 2A, we found that, compared with CED-9 alone, the amount of CED-9 pulled down per unit of GST::DRP-1 is 1.8-fold higher on average when CED-9 is bound to EGL-1. The amount of GST::DRP-1 pulled down with glutathione resin was unaffected by the presence of EGL-1 (Fig. S6A); we also determined that EGL-1 by itself is unable to bind to GST::DRP-1 (Fig. S7). We have previously shown that CED-9 also interacts with FZO-1, which is required for mitochondrial fusion (24). Interestingly, we found that compared with CED-9 alone, the amount of CED-9 pulled down per unit of GST-tagged FZO-1 (GST::FZO-1) is 1.4-fold lower on average when CED-9 is bound to EGL-1 (Fig. 2C). The amount of GST::FZO-1 pulled down with glutathione resin was unaffected by the presence of EGL-1 (Fig. S6B). Based on these in vitro data, we propose that the binding of EGL-1 to CED-9 can increase the interaction between CED-9 and DRP-1 and decrease the interaction between CED-9 and FZO-1. In contrast, the binding of CED-4 to CED-9 changed in a statistically significant manner neither the amount of CED-9 pulled down per unit of GST::DRP-1 (Fig. 2B), nor the amount of CED-9 pulled down per unit of GST::FZO-1 (Fig. 2D). Hence, the binding of CED-4 to CED-9 does not appear to affect the interactions of CED-9 with DRP-1 or FZO-1.

Fig. 2.

EGL-1 enhances the interaction between CED-9 and DRP-1. (A) EGL-1 enhances GST::DRP-1-CED-9 interaction in vitro. Purified GST::DRP-1 (5 μg/mL) was used in pull-down experiments in the presence of in vitro translated [³⁵S]methionine-labeled luciferase, CED-9 alone, or EGL-1–CED-9 complex. Average relative pull-down data are indicated (six data sets from three independent experiments). (B) CED-4 does not affect GST::DRP-1–CED-9 interaction in vitro. Purified GST::DRP-1 (2.5 and 5 μg/mL) were used in pull-down experiments in the presence of in vitro translated [³⁵S]methionine-labeled luciferase, CED-9 alone, or CED-9–CED-4 complex. Average relative pull-down data are indicated (four data sets from three independent experiments). (C) EGL-1 reduces GST::FZO-1–CED-9 interaction in vitro. Purified GST::FZO-1 (2.5, 5, and 10 μg/mL) were used in pull-down experiments in the presence of in vitro translated [³⁵S]methionine-labeled CED-9 alone or EGL-1–CED-9 complex. Average relative pull-down data are indicated (six data sets from two independent experiments). (D) CED-4 does not affect GST::FZO-1–CED-9 interaction in vitro. Purified GST::FZO-1 (2.5, 5, and 10 μg/mL) were used in pull-down experiments in the presence of in vitro translated [³⁵S]methionine-labeled CED-9 alone or CED-9–CED-4 complex. Average relative pull-down data are indicated (six data sets from two independent experiments). For all experiments, the statistical significance was determined by using the Student t test.

GTP Affects Interaction Between CED-9 and DRP-1.

GTP binding and hydrolysis are important properties of Drps (39). Therefore, we tested whether CED-9's ability to interact with DRP-1 is affected by the nucleotide bound to DRP-1. We preloaded GST::DRP-1 with nonhydrolysable forms of GTP or GDP (GTPγS or GDPβS, respectively) and performed GST pull-down experiments in the presence of in vitro translated, [³⁵S]methionine-labeled CED-9 protein. We found that the amount of CED-9 pulled down per unit of the GTP-bound form of GST::DRP-1 was 1.7-fold higher on average than the amount of CED-9 pulled down per unit of the GDP-bound form of GST::DRP-1 (Fig. 3A). In contrast, we found that GTP did not affect the amount of CED-9 pulled down per unit of a mutant GST::DRP-1 protein, GST::DRP-1(S41N), which has reduced affinity for GTP (40) (Fig. 3B). Finally, we found that preloading both CED-9 with EGL-1 and GST::DRP-1 with GTP increased the amount of CED-9 pulled down per unit of GST::DRP-1 3.0-fold (Fig. 3C). The amount of GST::DRP-1 pulled down with glutathione resin remained almost constant throughout these experiments (Fig. S6C). Altogether, these data suggest that compared with their interactions with the GDP-bound form of DRP-1, the interactions in vitro of both CED-9 and the EGL-1–CED-9 complex with the GTP-bound form of DRP-1 are enhanced.

Fig. 3.

GTP enhances the interaction between CED-9 and DRP-1. (A) GTPγS enhances GST::DRP-1–CED-9 interaction in vitro. Purified GST::DRP-1 (5 μg/mL) was used in pull-down experiments in the presence of in vitro translated [³⁵S]methionine-labeled CED-9 alone or in the presence of GTPγS or GDPβS. Average relative pull-down data are indicated (four data sets from four independent experiments). (B) GTPγS does not enhance GST::DRP-1(S41N)–CED-9 interaction in vitro. Purified GST::DRP-1(S41N) (5 μg/mL) was used in pull-down experiments in the presence of in vitro translated [³⁵S]methionine-labeled CED-9 alone or in the presence of GTPγS or GDPβS. Average relative pull-down data are indicated (three data sets from three independent experiments). (C) EGL-1 and GTP enhance CED-9–DRP-1 interaction in vitro. Purified GST::DRP-1 (5 μg/mL) was used in pull-down experiments in the presence of in vitro translated [³⁵S]methionine-labeled luciferase, CED-9 alone, and EGL-1–CED-9 in the presence or absence of GTPγS or GDPβS. Average relative pull-down data are indicated (two data sets from one experiment). In A and B, statistical significance was determined by using the Student t test.

EGL-1–CED-9 Complex Recruits DRP-1 to Mitochondria.

By using differential centrifugation, we isolated a mitochondria-enriched cell fraction from total worm lysates and analyzed the subcellular localization for endogenous DRP-1 protein by using an anti–DRP-1 antibody. By using this approach, we found that approximately 10% of total endogenous DRP-1 is found in the mitochondria-enriched fraction in WT animals (Fig. S5). Similar results were obtained with transgenic animals expressing mitogfp (P_HS mitogfp; ^+/+) or mitogfp and ced-9 [P_HS mitogfp, P_HS ced-9(wt); ^+/+] under the control of a heat-inducible promoter before or after heat shock (heat shock⁻ or heat shock⁺; Fig. 4). Hence, in WT animals as well as animals overexpressing ced-9, most DRP-1 localizes to the cytosol and only approximately 10% localizes to the mitochondrial surface. In contrast, we found that the amount of DRP-1 that localizes to the mitochondrial surface increases from approximately 10% to approximately 40% in animals coexpressing ced-9 and egl-1 [P_HS mitogfp, P_HS ced-9(wt), P_HS egl-1(wt); ^+/+; Fig. 4]. (As the level of endogenous EGL-1 protein is unknown, we are unable to estimate the fold over-expression in transgenic lines expressing egl-1.) This increase in DRP-1 localization to mitochondria was not observed in response to egl-1 expression in a ced-9–null mutant background [P_HS mitogfp, P_HS egl-1(wt); ced-9(n2812); Fig. 4]. Therefore, we conclude that the EGL-1–CED-9 complex but not EGL-1 or CED-9 alone can recruit DRP-1 to mitochondria.

Fig. 4.

DRP-1 predominantly localizes to the cytoplasm and can be recruited to mitochondria by EGL-1–CED-9. Total lysate (T), cytoplasm (C), and a mitochondrial-enriched fraction (M) were obtained and analyzed as described in Materials and Methods from transgenic lines not exposed to heat shock (heat shock⁻) or exposed to heat shock (heat shock⁺), which induces the expression of the transgenes. The presence of DRP-1, tubulin, and ATP synthase in the different fractions was analyzed by Western blot by using anti–DRP-1, anti-tubulin, and anti-ATP synthase antibodies, respectively. Histograms (Right) show relative amounts of DRP-1 in the different fractions for each condition. Data shown represent one representative data set of two independent experiments. The genotype of the different transgenic lines analyzed and the transgenes they carry are indicated. All strains analyzed were homozygous for the ced-3 mutation n717.

EGL-1–CED-9 Complex Promotes DRP-1-Dependent Mitochondrial Fission.

To determine the effect of ced-9 and egl-1 expression on mitochondrial morphology in vivo, we analyzed mitochondrial morphology approximately 200 min after the first cell division in transgenic lines expressing mitogfp, ced-9, and egl-1 under the control of a heat-inducible promoter. As transgenic animals used throughout our studies carry extrachromosomal arrays and therefore are mosaic for the presence of the transgenes (i.e., the expression of the transgenes), we analyzed mitochondrial morphology in individual cells of transgenic animals and assigned the morphology in an individual cell to one of three categories (“fragmented,” “tubular,” or “highly globular”; Fig. 5A). (All data were scored in blinded fashion by two investigators independently.) Furthermore, because the expression of egl-1 induces apoptosis (34), the experiments were performed in the background of the ced-3 loss-of-function mutation n717, in which apoptosis is blocked in general (41). [The ced-3 gene acts downstream of egl-1 in the apoptotic process and has no known role in mitochondrial dynamics (as detailed later; Fig. S8).] (18, 24). As previously described, the expression of ced-9 promotes mitochondrial fusion (24). Specifically, cells expressing mitogfp alone have tubular mitochondria (96%; n = 30; three independent transgenic lines; P_HSmitogfp; Fig. 5B), whereas the majority of cells expressing mitogfp and ced-9 have one or two large “highly globular” mitochondria per cell, which is indicative of increased mitochondrial fusion (98%; n = 30; three independent transgenic lines; P_HSmitogfp, P_HSced-9(wt); Fig. 5B). In contrast, we found that the expression of both ced-9 and egl-1 promotes mitochondrial fragmentation. Specifically, more than 50% of cells expressing mitogfp, ced-9, and egl-1 have fragmented mitochondria (53%; n = 50; three independent transgenic lines; P_HSmitogfp, P_HSced-9(wt), P_HSegl-1(wt); Fig. 5B) (3D reconstruction of z stacks of wild-type embryos or embryos over-expressing ced-9 or ced-9 and egl-1 are available in Movies S1–S3). In contrast, expressing egl-1 in a ced-9–null mutant background did not cause mitochondrial fragmentation (92% cells with tubular mitochondria; n = 120; three independent lines; P_HSmitogfp, P_HSegl-1(wt) in ced-9(n2812); Fig. 5B), demonstrating that both CED-9 and EGL-1 are required to promote mitochondrial fission. To confirm that this change in mitochondrial morphology is the result of increased mitochondrial fission, we repeated the experiment in animals in which DRP-1 protein had been depleted by using RNA-mediated interference (RNAi; Fig. S3). We found that depleting DRP-1 blocks the appearance of fragmented mitochondria (3% cells with fragmented mitochondria; n = 50; two independent transgenic lines), which indicates that EGL-1–CED-9–induced mitochondrial fragmentation is caused by DRP-1–dependent mitochondrial fission [P_HSmitogfp, P_HSced-9(wt), P_HSegl-1(wt), drp-1(RNAi); Fig. 5B). Based on these findings we propose that, whereas CED-9 promotes FZO-1- and EAT-3–dependent mitochondrial fusion, the EGL-1–CED-9 complex promotes DRP-1–dependent mitochondrial fission.

Fig. 5.

EGL-1 blocks CED-9–dependent mitochondrial fusion and promotes CED-9–dependent mitochondrial fission in vivo. (A) Schematic of the different classes of mitochondrial morphology scored. (B) Transgenic lines were analyzed by Nomarski optics and fluorescent microscopy. The genotype of the different transgenic lines analyzed and the transgenes they carry are indicated. Quantification of the different mitochondrial morphologies observed, and the number of embryonic cells and transgenic lines analyzed are indicated. (Scale bar: 10 μm.) Three-dimensional reconstructions of regions marked with stars are available in Movies S1–S3.

Ability of CED-9 to Recruit DRP-1 to Mitochondria and Promote Mitochondrial Fission Is Dependent on Ability to Bind to EGL-1 and DRP-1.

Next, we analyzed mutant EGL-1 and CED-9 proteins for their abilities to recruit DRP-1 to mitochondria and to promote mitochondrial fission. The egl-1(Δ17aa) transgene causes the synthesis of a mutant EGL-1 protein that lacks the BH3 domain and shows a 10-fold reduced interaction with CED-9 in vitro (Fig. S9). We found that, compared with the expression of ced-9 and egl-1, the expression of ced-9 and egl-1(Δ17aa) was less effective in recruiting DRP-1 to mitochondria (∼19% of DRP-1 localized to the mitochondria-enriched fraction; P_HSmitogfp, P_HSced-9(wt), P_HSegl-1(Δ17aa); Fig. 4). Similarly, the expression of ced-9 and egl-1(Δ17aa) was less effective in promoting mitochondrial fission (10% cells with fragmented mitochondria; n = 50; two independent transgenic lines; P_HSmitogfp, P_HSced-9(wt), P_HSegl-1(Δ17aa); Fig. 5B). The residual DRP-1 recruitment and mitochondrial fission observed in these transgenic lines can be explained by the fact that EGL-1(Δ17aa) still can interact with CED-9, but, as mentioned earlier, at a level that is 10-fold reduced (Figs. 4 and 5B). Similarly, the observation that rather than observing predominantly highly globular mitochondria, similar percentages of tubular and highly globular mitochondria are detected in these transgenic lines, is most likely a result of the fact that CED-9–induced mitochondrial fusion is not completely recovered because of the residual binding of EGL-1(Δ17aa) to CED-9 (Fig. 5B).

Next, we analyzed CED-9(N158A, A159G, Q160A), which, in vitro, shows reduced interactions with EGL-1 (Fig. S9) and DRP-1 (Fig. 1E). We found that, compared with the expression of ced-9 and egl-1, the expression of ced-9(N158A, A159G, Q160A) and egl-1 was less effective in recruiting DRP-1 to mitochondria [∼12% of DRP-1 localized to mitochondria-enriched fraction; P_HSmitogfp, P_HSced-9(N158A, A159G, Q160A), P_HSegl-1(wt); Fig. 4]. Furthermore, the expression of ced-9(N158A, A159G, Q160A) and egl-1 was less effective in promoting mitochondrial fission [12% cells with fragmented mitochondria; n = 50; three independent transgenic lines; P_HSmitogfp, P_HSced-9(N158A, A159G, Q160A), P_HSegl-1(wt); Fig. 5B]. Based on these observations, we conclude that the ability of CED-9 to recruit DRP-1 to mitochondria and to promote mitochondrial fission is dependent on its ability to interact with both EGL-1 and DRP-1.

Loss of egl-1 Function Causes Increase in Mitochondrial Length.

Our data support the notion that CED-9 promotes FZO-1–dependent mitochondrial fusion whereas the CED-9–EGL-1 complex recruits DRP-1 to mitochondria, thereby causing DRP-1–dependent mitochondrial fission. However, because our data are based on the over-expression of ced-9 or ced-9 and egl-1, we cannot rule out that the effects observed are a consequence of a CED-9–induced shift of endogenous DRP-1 or FZO-1 from their native binding partners. To address this question, we took a complementary approach and determined the effects of loss-of-function mutations of ced-9 and egl-1 on mitochondrial morphology. The loss of ced-9 function has previously been shown to cause no detectable defect in steady-state mitochondrial morphology in embryonic cells or in body wall muscle cells of larvae of the fourth larval stage (L4 larvae) (23, 24, 42). It has therefore been proposed that, in most cell types and under most cellular conditions, CED-9–dependent mitochondrial fusion and fission are in balance (24). Our data indicate that EGL-1 is required for CED-9–dependent mitochondrial fission. Hence, the loss of egl-1 function should result in the specific block of CED-9–dependent mitochondrial fission. If the notion that CED-9–dependent mitochondrial fusion and fission are in balance were correct, in contrast to the loss of ced-9 function, the loss of egl-1 function should therefore cause a defect in steady-state mitochondrial morphology. To test this, we analyzed mitochondrial length in body wall muscle cells of L4 larvae. (The experiments were performed in the background of the ced-3 loss-of-function mutation n2427.) We found that the egl-1 loss-of-function mutation n1084 n3082 causes an approximately 20% increase in mitochondrial length (Fig. 6 A and B). This observation is consistent with the notion that the loss of egl-1 function reduces mitochondrial fission. To rule out that the reduction in mitochondrial length observed in egl-1(n1084 n3082) animals is the result of the loss of CED-4 and CED-3 activation, we analyzed mitochondrial morphology in animals carrying a ced-3 null mutation, n717. We found that the loss of ced-3 function does not affect mitochondrial length (Fig. S8), which confirms that egl-1 promotes mitochondrial fission by stabilizing the interaction between CED-9 and DRP-1 rather than causing CED-4 and CED-3 activation.

Fig. 6.

The loss of egl-1 function causes elongated mitochondria. (A) The egl-1(n1084 n3082) loss-of-function mutation enhances mitochondrial length in body wall muscle cells. Transgenic animals expressing P_myo-3mitogfp in control, egl-1(n1084 n3082), or ced-9(n2812); egl-1(n1084 n3082) animals were analyzed by fluorescent microscopy. All strains analyzed were homozygous for the ced-3 mutation n2427. (Upper) Fluorescent images of mitochondria in body wall muscle cells of indicated transgenic animals. (Bottom) Schematic of the mitochondrial morphologies observed. The locations of nuclei (nuc) are indicated. (Scale bar: 5 μm.) Three-dimensional reconstructions of animals of the three genotypes are available in Movies S4–S6. (B) Average length of mitochondria in body wall muscle cells. The length was measured by using Metamorph software. Each blue dot represents the average mitochondrial length of one muscle cell. The red dots represent the average mitochondrial length for the respective genotype. The number of muscle cells analyzed and the average length are indicated. Mitochondria in egl-1(n1084 n3082) animals are statistically longer than mitochondria in control animals (P = 0.015) or in ced-9(n2812); egl-1(n1084 n3082) animals (P = 0.025). Statistical significance was determined by using the Student t test. (C) The loss of egl-1 function partially blocks CED-4- and CED-3–induced apoptosis of ALM neurons. The presence of ALM neurons after expression of the P_mec-7ced-4 or P_mec-7ced-3 transgene was determined in an egl-1(+) background or the egl-1(n1084 n3082) background as described in Materials and Methods. *P_mec-7ced-4 and P_mec-7ced-3 transgenes in the egl-1(n1084 n3082) background were crossed back into the egl-1(+) background and the presence of ALM neurons was reanalyzed. The percentage of ALMs present is indicated (n, number of ALMs analyzed).

Furthermore, we found that the mitochondrial morphology phenotype caused by egl-1(n1084 n3082) is completely suppressed by the ced-9 null mutation n2812, which suggests that blocking CED-9–dependent mitochondrial fusion compensates for the reduction in mitochondrial fission caused by the loss of egl-1 function (Fig. 6 A and B) (3D reconstruction of z stacks of ced-3(n2427), ced-3(n2427); egl-1(n1084 n3082) and ced-9(n2812); ced-3(n2427); egl-1(n1084 n3082) body wall muscle cells mitochondria are available in Movies S4–S6). Based on these observations, we conclude that CED-9–dependent mitochondrial fusion and fission are in balance and that egl-1 is specifically required for CED-9–dependent mitochondrial fission.

EGL-1 Activity Can Be Detected in Cells Programmed to Live.

By using transcriptional reporters, it has been shown that the egl-1 gene is expressed at detectable levels specifically in cells programmed to die during C. elegans development (43, 44). However, the finding that egl-1 mutants have longer mitochondria in body wall muscle cells suggests that egl-1 might be expressed in cells programmed to live, at least at a low level. (With the use of available reporters, this broad expression of egl-1 might be below the level of detection. Consequently, experiments that use such reporters to express a gene of interest specifically in cells destined to die should be taken with caution, as the gene of interest might also be expressed in cells destined to live, albeit at a low level.) To test this, we determined the killing activity of the proapoptotic genes ced-4 or ced-3 in a WT and egl-1 mutant background. (All data were scored in blinded fashion.) It has previously been shown that the expression of ced-4 and ced-3 under the control of the mec-7 promoter (P_mec-7ced-4 and P_mec-7ced-3), which is active in the ALM neurons, two neurons that normally survive, can cause the ALMs to inappropriately undergo apoptosis (45). We found that egl-1(n1084 n3082) blocks approximately 50% of CED-4- and CED-3–induced ALM death (Fig. 6C). This observation suggests that low levels of EGL-1 protein are present in the ALMs and, hence, cells that are not programmed to die during development. For this reason, EGL-1–CED-9 complexes might be present in most cells throughout C. elegans development and adult life and, hence, capable of promoting mitochondrial fission through their interaction with DRP-1.

Discussion

Complex of EGL-1 and CED-9 Can Act as Receptor for DRP-1 on Mitochondria.

It has previously been shown that the BH3-only protein EGL-1 and the BCL2-like protein CED-9 of C. elegans promote DRP-1–dependent mitochondrial fission in apoptotic cells (18). However, the molecular mechanism through which EGL-1 and CED-9 promote DRP-1–dependent mitochondrial fission has so far been unknown. We demonstrate that a complex of EGL-1 and CED-9 directly interacts with DRP-1. Furthermore, we provide evidence that the EGL-1–CED-9 complex can act as a mitochondrial receptor for DRP-1 (Fig. 7). The loss of drp-1 function in C. elegans causes a severe “hyperfusion” phenotype (3). In contrast, the loss of egl-1 function only causes an increase in mitochondrial length of approximately 20%. Therefore, mitochondrial receptors other than EGL-1–CED-9 must exist for DRP-1 in C. elegans. It has recently been suggested that the outer mitochondrial membrane protein Mff can act as a receptor for Drp1 in mammalian cells (13, 46). Hence, the C. elegans homologues of mammalian Mff, MFF-1, and MFF-2 might act as such additional mitochondrial receptors for DRP-1 in C. elegans.

Fig. 7.

Control of mitochondrial dynamics by CED-9 and EGL-1. A simplified molecular model is shown. Mitochondrial morphology is maintained by a balance between CED-9–dependent mitochondrial fusion and fission. CED-4–CED-9 complexes or CED-9 alone promote FZO-1- and EAT-3–dependent mitochondrial fusion by directly interacting with FZO-1. EGL-1–CED-9 complexes promote DRP-1–dependent mitochondrial fission by recruiting GTP-bound DRP-1 to mitochondria. Further details are provided in the text.

Mammalian BCL-xL has been shown to interact with Drp1 (26). Therefore, we hypothesize that BCL2-like proteins act as receptors for Drp1 in mammals as well. Interestingly, it has been demonstrated in HeLa cells that the stable association of Drp1 with mitochondria in response to apoptotic signals is dependent on BAX and BAK (19). Furthermore, the knock-down in HeLa cells of the mammalian BH3-only protein Puma can attenuate the relocalization of Drp-1 to mitochondria in response to doxorubicin-induced DNA damage (47). These findings suggest that, at least in HeLa cells, BAX and BAK as well as a complex of Puma and a BCL2-like protein can act as receptors for Drp1.

The current working model of how Drp1-like GTPases such as C. elegans DRP-1 cause mitochondrial fission is that they are targeted to mitochondria as dimers. Once associated with mitochondria, they undergo GTP-driven multimerization and GTP hydrolysis-driven conformational changes, which lead to membrane constriction and ultimately mitochondrial fission (5). As our data indicate that EGL-1–CED-9 preferentially binds to GTP-bound DRP-1, we propose that EGL-1–CED-9 specifically recruits DRP-1 dimers that are “primed” for GTP-driven multimerization. Interestingly, it has recently been shown that in mammals, mitochondria-associated Drp1 can induce changes in the outer mitochondrial membrane that promote BAX oligomerization, cytochrome c release, and, hence, apoptosis (48). In analogy, we speculate that, when it has been recruited to the outer mitochondrial membrane through the EGL-1–CED-9 complex, C. elegans DRP-1 in turn may stimulate the proapoptotic activity the EGL-1–CED-9 complex has been proposed to acquire in apoptotic cells (18, 49).

EGL-1 Can Act as a Molecular Switch for CED-9's Function in Mitochondrial Dynamics.

Depending on the cellular context, BCL2-like proteins have been shown to promote mitochondrial fusion or fission (15). However, it remains unclear how these two opposing activities are supported by the same proteins. Our data indicate that the BH3-only protein EGL-1 can act as a molecular switch for the functions in mitochondrial dynamics of the BCL2-like protein CED-9 (Fig. 7). Specifically, we demonstrate that the binding of EGL-1 to CED-9 can increase the interaction between CED-9 and DRP-1, which is required for mitochondrial fission, but decrease the interaction between CED-9 and FZO-1, which is required for mitochondrial fusion. It has previously been shown that the binding of EGL-1 induces a dramatic conformational change in CED-9 (50). We speculate that this conformational change alters CED-9's binding sites for FZO-1 and DRP-1 and hence CED-9's affinities for FZO-1 and DRP-1. Alternatively, the EGL-1–induced conformational change may alter the stoichiometries between CED-9 and DRP-1 or FZO-1 or other aspects of their interactions.

BH3-only proteins are binding partners of BCL2-like proteins in mammals as well. Therefore, we hypothesize that the function of BH3-only proteins as molecular switches for the function in mitochondrial dynamics of BCL2-like proteins is conserved. In support of this notion, mammalian BH3-only proteins have recently also been implicated in mitochondrial dynamics. For example, the coexpression in HeLa cells of either BCL-xL or Mcl-1 with the BH3-only genes Bid, Bim, or Puma results in extensive mitochondrial fragmentation independently of apoptosis induction (51). In addition, the expression of the BH3-only gene Noxa in HeLa cells induces mitochondrial fragmentation. Conversely, mitochondrial fragmentation in response to the proteasome inhibitor bortezomib (Velcade) is greatly attenuated in Noxa-deficient (Noxa^−/−) immortalized baby mouse kidney cells (52). Finally, peptides spanning the BH3 domains of BID or BIM are capable of inducing mitochondrial fragmentation in BAX-, BAK-deficient (BAX^−/−, BAK^−/−) mouse embryonic fibroblast cells (53). Interestingly, in the same study, it was also demonstrated that BH3 peptides can increase by approximately twofold the binding of over-expressed BCL-xL to Drp1 (53). Based on these observations, we propose that, like in C. elegans, the binding of BH3-only proteins to BCL2-like proteins in mammals affects the interactions between BCL2-like proteins and Drp1 and hence the capability of BCL2-like proteins to modulate mitochondrial dynamics.

Physiological Roles for CED-9–Dependent Mitochondrial Fusion and Fission.

Our data support the model that CED-9–dependent mitochondrial fusion and fission are in balance (Fig. 7). What might the physiological role of CED-9–dependent mitochondrial fusion and fission be? Time-lapse analyses of individual, photo-labeled mitochondria in mammalian cells have revealed that fusion and fission events are often temporarily linked and that mitochondria undergo approximately five fusion/fission cycles per hour (54, 55). It has furthermore been proposed that these fusion/fission cycles help maintain the functionality of mitochondria. However, how fusion and fission are linked during these cycles has so far been unclear. As CED-9 directly interacts with components of the mitochondrial fusion and fission machineries and affects both processes, we speculated that CED-9 may act as a coupler between mitochondrial fusion and fission during these fusion/fission cycles, thereby helping to maintain mitochondrial integrity.

We also demonstrate that the balance of CED-9–dependent mitochondrial fusion and fission, and hence steady-state mitochondrial morphology can be affected by changes in the amount of EGL-1 protein (Fig. 7). Specifically, we present evidence that decreasing the amount of EGL-1 promotes CED-9–dependent mitochondrial fusion and thereby causes an increase in the average length of mitochondria in body wall muscle cells in L4 larvae. Consistent with the work of Breckenridge et al. (42), we did not observe a mitochondrial morphology defect in embryos carrying an egl-1 loss-of-function mutation (Fig. S10). The absence of a detectable defect in embryos may be a result of our inability to detect subtle differences in mitochondrial length in embryonic cells. Alternatively, it is possible that EGL-1 does not play a role in the control of mitochondrial morphology at this early stage of embryonic development. We also present evidence that increasing the amount of EGL-1 promotes DRP-1–dependent mitochondrial fission and thereby causes a change in steady-state mitochondrial morphology from tubular to fragmented. Such a fragmented mitochondrial phenotype is normally observed in cells undergoing apoptosis during C. elegans development, in which it is known that the egl-1 gene is transcriptionally up-regulated (18, 28). Therefore, physiological signals such as apoptotic signals affect steady-state mitochondrial morphology by modulating the amount of EGL-1 protein.

The goal of future studies is to identify additional signals that affect the balance between CED-9–dependent mitochondrial fission and fusion and to determine the molecular mechanisms through which they act. Ultimately, we hope to understand how multicellular organisms modulate mitochondrial dynamics in response to intra- and extracellular signals.

Materials and Methods

General Methods and Strains.

C. elegans strains were cultured as described previously (56). Bristol N2 was used as the WT strain. Mutations used in this study are listed later and were described previously by Riddle and coworkers (57), except where noted otherwise: (LG III) ced-9(n2812) (58); (LG IV) drp-1(tm1108) (National BioResource Project), ced-3(n717), ced-3(n2427) (58); (LG V) egl-1(n1084 n3082) (34); and (LG X) lin-15(n765ts).

Transgenic Animals.

A summary of all transgenic lines generated for this study is provided in Table S1.

In Vitro Interactions.

In vitro interactions were performed by GST pull-down essentially as described previously (24). Briefly, GST::DRP-1 was purified by using the NETN buffer. CED-9, DRP-1, and EGL-1 were in vitro translated in the presence of [³⁵S]methionine using the TNT Rabbit Reticulocyte Lysate System (Promega). To analyze the effect of EGL-1 or CED-4 on the CED-9-DRP-1 or CED-9-FZO-1 interaction, CED-9 was preincubated in the presence or absence of EGL-1 or CED-4 for 30 min at 30 °C. Purified GST::DRP-1 or GST::FZO-1 was then incubated with CED-9, EGL-1-CED-9, or CED-4-CED-9 for 2 h at 4 °C. To test the effect of GTP on the CED-9-DRP-1 interaction, purified GST::DRP-1 was preincubated in 250 μL of binding buffer in the presence of 2 mM EDTA and 0.8 mM GTPγS or GDPβS (Sigma) for 1 h at 25 °C. After addition of 5 mM MgCl₂, the reaction was incubated for 15 min on ice. Subsequently, CED-9 was added to the reaction and incubated for 2 h at 4 °C. Input and pull-down were analyzed by SDS/PAGE and quantified using a PhosphorImager (Storm Imager; GE Healthcare). The percentage of input pulled down was normalized to the amount of GST-fusion protein pulled down by glutathione resin in each reaction.

In Vivo Coimmunoprecipitation Experiments.

In vivo interactions were analyzed by coimmunoprecipitation essentially as described previously (24). Briefly, embryonic lysates from transgenic animals over-expressing ced-9 were immunoprecipitated by using anti–CED-9 (1:33 dilution) (24) or anti-GFP (Invitrogen) antibodies. The input and immunoprecipitates were analyzed by SDS/PAGE. To detect CED-9, DRP-1, and bc1, we used anti–CED-9 (1:2,000), anti–DRP-1 (1:1,000), and anti-bc1 (1:1,000) (59) antibodies, respectively.

Mitochondrial Extraction and DRP-1 Localization Studies.

For mitochondrial extractions, transgenic animals were cultured on egg plates at 20 °C and then heat-shocked at 32 °C for 1 h 15 min to allow the expression of the transgenes. Animals were collected and washed with M9 buffer. Worm pellets were resuspended in 30% sucrose, washed with 0.1 M NaCl, and resuspended in two volumes of IB buffer (1 M Tris HCl, pH 7.4; 0.5 M mannitol; 1.75 M sucrose; 0.5 mM EDTA, pH 8), and homogenized with 20 strokes by using a 7-mL tight Dounce tissue grinder (Fisher Scientific). Low-speed centrifugation (800 × g) was used to remove cell debris and nuclei. High-speed centrifugation (12,000 × g) was subsequently used to separate cytoplasm from the mitochondria-enriched fraction. Total lysate, cytoplasm, and mitochondria-enriched fraction were analyzed by SDS/PAGE. To detect DRP-1, tubulin, and ATP synthase, we used anti–DRP-1 (1:1,000), anti–α-tubulin (1:3,000; Sigma) and anti-ATP synthase (1:2,000; MitoSciences) antibodies, respectively. The data were quantified by using ImageJ.

Analysis of Mitochondrial Morphology in Embryos.

Analysis of mitochondrial morphology in embryos was performed as described previously (24) with the following modifications. For each genotype or set of transgenes, mitochondrial morphology was scored in 10 cells per embryo in three to five embryos from two or three independent lines. Scoring was performed in blinded fashion twice by different individuals by using three different categories of mitochondrial morphology (fragmented, tubular, and highly globular). Schematics of these representative categories are indicated in Fig. 5A. Data shown are the average of the two independent scores and represent the average percentage of cells showing a certain morphology.

RNAi.

RNAi by feeding was performed as described previously (60). Briefly, L4 larvae were grown at 20 °C on drp-1 RNAi plates containing 6 mM IPTG for 24 h. Adults were heat-shocked at 32 °C for 45 min and incubated at 20 °C for 1 h 15 min. Embryos were analyzed by Nomarski optics and fluorescent microscopy.

Analysis of Mitochondrial Morphology in Body Wall Muscle Cells.

Transgenic L4 larvae were analyzed by Nomarski optics and fluorescent microscopy. The length of mitochondria in muscle cells was measured by using Metamorph software. Scoring was performed in blinded fashion. The different datasets follow a log-normal distribution (P = 0.89 for ced-3(lf), P = 0.98 for ced-3(lf); egl-1(lf), P = 0.96 for ced-9(lf); ced-3(lf); egl-1(lf) by Kolmogorov–Smirnov test). The data were log-transformed and analyzed by Student t test.

Presence of ALM Neurons.

The presence of ALM neurons after expression of the P_mec-7 ced-4 or P_mec-7 ced-3 transgene was determined in an egl-1(+) background or the egl-1(n1084 n3082) background as described previously (45). All analyses were performed in blinded fashion.

Image Acquisition and Processing.

Image acquisition and processing were performed as described previously (24).