OPA1 and cardiolipin team up for mitochondrial fusion

Abstract

Fusion between the inner membranes of two mitochondria requires the GTPase OPA1, but the molecular mechanism is poorly understood. A study now shows that fusion of two liposomes can be performed by OPA1 tethered to just one liposome, through an interaction with the phospholipid cardiolipin on the opposing liposome.

Main text

As dynamic organelles, mitochondria require a balance between fusion and fission events for their proper function¹. Mitochondria have double membranes, and the merger of two mitochondria involves outer membrane (OM) fusion followed by inner membrane (IM) fusion. Mechanoenzymes from the dynamin superfamily of large GTPases act sequentially to mediate these membrane remodeling reactions¹. In mammals, fusion of the OM is carried out by mitofusins (MFN1 and MFN2), whereas subsequent fusion of the IM is carried out by optic atropy 1 (OPA1). These reactions are critical for cell function. Mutations in MFN2 cause peripheral neuropathy, and mutations in OPA1 cause dominant optic atrophy, the most common inherited optic neuropathy. The biochemical basis of OPA1 action is poorly understood. In this issue², Ban et al. achieve a breakthrough in this area by developing a liposome-based assay to measure the membrane fusion activity of recombinant OPA1, leading to key biochemical insights.

Mitochondrial fusion is a homotypic membrane fusion event. This aspect is mirrored in the mechanism of OM fusion, which requires OM-bound mitofusin molecules to be present on both of the opposing membranes. Structural studies have provided models of how mitofusin dimers might form between opposing OMs to tether mitochondria together^3–5. Analogously, homotypic fusion of the endoplasmic reticulum (ER) requires trans complexes of atlastin⁶, a large GTPase embedded in the ER membrane. This molecular symmetry, however, does not apply to mitochondrial IM fusion (Figure 1). In contrast to the requirement for mitofusins, mitochondrial fusion assays involving cells or isolated mitochondria indicate that mitochondria from wild-type cells are perfectly capable of fusing with mitochondria from OPA1-null cells^{7, 8}. Complicating things further, OPA1 exists in the mitochondrial intermembrane space as both an IM-bound ‘long form’ (L-OPA1), and a soluble, ‘short form’ (S-OPA1) produced from proteolytic cleavage of L-OPA1. It is clear that optimal fusion of mitochondria requires a combination of L-OPA1 and S-OPA1, and that S-OPA1 alone is insufficient⁹. L-OPA1 alone does not have significant fusion activity under normal circumstances⁹, but appears sufficient for fusion when cells are placed under specific stress conditions¹⁰. In addition, cells lacking the proteases responsible for generating S-OPA1 are left with only L-OPA1, but nevertheless display some level of fusion activity¹¹. Therefore, OPA1 fuses the IM in an asymmetric manner, and some of the evidence describing the relative contribution of L- versus S-OPA1 is difficult to reconcile.

Figure 1. L-OPA1 and cardiolipin interactions mediate membrane fusion.

Whereas homotypic interactions between mitofusin (MFN) molecules in trans are required for OM fusion, OPA1 is required in only one of the organelles for IM (inner membrane) fusion. Ban et al. show that L-OPA1 partners with cardiolipin on the opposing IM; this heterotypic interaction is sufficient for IM fusion. CL, cardiolipin; IMS, intermembrane space.

In order to gain insight into the mechanism of OPA1-mediated membrane fusion, Ban et al. reconstituted OPA1-mediated fusion in vitro to allow dissection of the molecular requirements. By expressing, purifying, and incorporating recombinant L-OPA1 into liposomes, the authors successfully developed a FRET-based membrane fusion assay to detect fusion between liposomes containing L-OPA1 and cardiolipin. This fusion activity increased with increased L-OPA1, required GTP, and was absent in liposomes containing an OPA1 mutant lacking GTP hydrolysis activity. These experiments generated two key insights. First, they demonstrate that L-OPA1 alone is clearly sufficient to mediate membrane fusion. Second, they satisfyingly explain why OPA1 is required on only one mitochondrion to mediate IM fusion. Liposomes containing L-OPA1 can fuse with liposomes devoid of L-OPA1, as long as the latter liposomes contain sufficient amounts of the phospholipid cardiolipin. Cardiolipin is enriched in the mitochondrial IM, where it constitutes about 20% of the phospholipid content, compared to the OM, where it is only present at about 4%. This striking result suggests that mitochondrial IM fusion involves a heterotypic interaction between OPA1 on one side and cardiolipin on the other (Figure 1). This protein-phospholipid interaction may be sufficient to ensure specificity, because the mitochondrial IM is not exposed to other membranes until OM fusion by mitofusins.

In addition to mitochondrial fusion, OPA1 is important for maintaining the structure of cristae membranes, infoldings of the IM that have high membrane curvature. Interactions between OPA1 molecules on the IM have been proposed to maintain proper cristae shape¹². Ban et al. modified the liposome assay to evaluate the ability of OPA1 to mediate liposome tethering. When L-OPA1 was incorporated into two sets of liposomes containing low cardiolipin, the liposomes could not fuse, but were tethered together. The authors propose that the two types of L-OPA1 interactions may lead to different functional outcomes. In one scenario, non-fusogenic trans complexes of L-OPA within a single mitochondrion may help to maintain cristae structural integrity. In the other, L-OPA1 becomes fusogenic when it interacts with cardiolipin-rich domains from an opposing IM.

Although L-OPA1 alone is sufficient for fusion in vitro, fusion events in vivo appear more complex, because they are clearly affected by S-OPA1. To illuminate the function of S-OPA1, the authors examined its effect on liposome fusion under two different conditions – one in which L-OPA1 levels are held constant, and one in which total OPA1 levels are held constant. Addition of S-OPA1 to a constant concentration of L-OPA1 enhanced both liposome binding and fusion activity, consistent with a model where L-OPA1 cleavage into S-OPA1 promotes fusion⁸. However, when total OPA1 levels were held constant, and S-OPA1 levels were increased at the expense of L-OPA1, the authors observed a decrease in fusion activity. The interpretation of these results is complicated, because it is not clear whether addition of soluble S-OPA1 faithfully mimics physiological proteolytic cleavage of L-OPA1 to S-OPA1. To address these issues, it will be important to reconstitute liposomes with engineered versions of L-OPA1 that can be artificially cleaved. Consistent with the critical role of S-OPA1 for mitochondrial fusion in mammalian cells⁹, the yeast ortholog Mgm1 requires proteolytic processing for activity. Together, L-Mgm1 and S-Mgm1 act as a heterodimer to mediate fusion. In a division of labor, L-Mgm1 functions as an IM anchor, while S-Mgm1 drives fusion via GTP hydrolysis^{13, 14}. In fact, a heterodimer containing GTPase defective L-Mgm1 is functional for fusion, as long as the associated S-Mgm1 has GTPase activity. Such results underscore the need to further examine the collaboration between L- and S-OPA1.

This study raises additional interesting questions. The demonstration that L-OPA1 is sufficient for fusion raises the issue of why L-OPA1 has little activity when expressed in OPA1-null cells. In contrast, L-OPA1 appears to be fusogenic when cells are treated with cycloheximide or other stressors, in a phenomenon termed stress-induced mitochondrial hyperfusion¹⁰. These observations suggest that unknown regulatory mechanisms exist to control the activity of L-OPA1 in vivo. In addition, it will be important to address how the heterotypic interaction between L-OPA1 and cardiolipin leads to activation of membrane fusion. Conversely, given that cardiolipin is present in cis, how is the fusogenic activity of OPA1 suppressed until OM fusion brings it face-to-face with the opposing IM? The GTP hydrolysis activity of OPA1 is increased by cardiolipin binding, and a disease allele of OPA1 has been shown to abrogate this effect and impair fusion¹⁵. Moreover, like other dynamins, OPA1 is known to self-assemble, and S-OPA1 is capable of forming large assembles that can tubulate membranes in vitro¹⁵. These biochemical properties are likely relevant to the fusion process, and it will be interesting to determine whether cardiolipin can regulate OPA1 conformation and self-assembly.