PINK1-Phosphorylated Mitofusin 2 Is a Parkin Receptor for Culling Damaged Mitochondria

Mitochondria are endosymbiotic organelles derived from primitive aerobic bacteria. Healthy mitochondria are essential energy generators for most metazoan cell functions, whereas senescent or damaged mitochondria are sources of toxic reactive oxygen species. Thus, mitochondrial biogenesis and mitophagic elimination are carefully orchestrated, and their mutational disruption causes chronic degenerative diseases (1). Genetic studies have linked Parkinson’s disease to mutations of two mitophagy genes, the E3 ubiquitin ligase Parkin and the serine-threonine protein kinase PINK1 (PTEN-induced putative kinase protein 1; PTEN is phosphatase and tensin homolog) (2). Loss of the mitochondrial inner membrane electrochemical gradient stabilizes PINK1 on damaged organelles, tagging them for Parkin binding, ubiquitination, and mitophagic elimination (3). The specific nature of the molecular interaction between PINK1 and Parkin is unclear, and Parkin receptor proteins on damaged mitochondria have not been identified.

Mitofusins (Mfn) 1 and 2 are mitochondrial outer-membrane fusion proteins and Parkin ubiquitination substrates (4, 5). Combined genetic ablation of Mfn1 and Mfn2 in mouse hearts induces mitochondrial dysfunction and fragmentation that should stimulate mitophagic removal but instead results in proliferation of abnormal organelles (6). Because mitophagy is stimulated by Parkin-mediated ubiquitination of mitochondrial proteins (7, 8), we tested whether Mfn1 or Mfn2 might mediate signaling activity of the PINK1-Parkin pathway. Parkin co-immunoprecipitated with Mfn2, but not Mfn1, from extracts of human embryonic kidney (HEK) cells transfected with tagged mitofusins and Parkin. The association of Mfn2 and Parkin was greatly enhanced in cells also transfected to overexpress PINK1 (Fig. 1, A and B, and fig. S1).

• Download high-res image
• Open in new tab
• Download Powerpoint

The oxidative phosphorylation inhibitor carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP) depolarizes mitochondria and stimulates PINK1-mediated translocation of Parkin to mitochondria, thus targeting damaged organelles for mitophagy (3) (fig. S2). Consistent with a role for Mfn2 in this process, FCCP treatment stimulated Parkin binding to endogenous fibroblast Mfn2 (fig. S3). Overexpression of PINK1 enhanced, whereas PINK1 suppression with RNA interference inhibited, this effect of FCCP (fig. S3). We developed an ex vivo model of this pathway in cardiomyocytes because they express endogenous Mfn2 and Parkin in large amounts and have abundant mitochondria. In genetically normal cardiomyocytes, mitochondrial uncoupling with FCCP (fig. S4) directed redistribution of Parkin to intracellular punctae especially profuse in the mitochondrial-rich perinuclear region (Fig. 1C, left). Parkin translocation was stimulated by FCCP in Mfn1-null, but not Mfn2-null, cardiomyocytes (Fig. 1C, middle and right). In Mfn2-null cells, Parkin remained diffusely cytosolic, demonstrating a requirement for endogenous Mfn2 to promote Parkin localization to depolarized mitochondria. Because Mfn1 and Mfn2 are each expressed in normal amounts in hearts lacking their opposite number (fig. S5) and recombinant Mfn1 does not bind Parkin (Fig. 1A), Parkin recruitment appears to be a unique property of Mfn2.

Mfn2 can be a Parkin ubiquitination substrate (8). This might require PINK1-stimulated association of Parkin and Mfn2 with mitochondria, so we tested the functional importance of PINK1-Mfn2-Parkin interactions on ubiquitination. We observed complete concordance of ubiquitination with PINK1-stimulated Mfn2 binding to Parkin in HEK cells (Fig. 2A) and with FCCP-stimulated binding of endogenous cardiac Mfn2 to Parkin in isolated perfused mouse hearts (Fig. 2B). Conversely, Mfn2 (but not Mfn1) gene ablation in cardiac myocytes decreased mitochondrial ubiquitination stimulated by mitochondrial depolarization, following the same pattern seen for Parkin translocation (Fig. 2C and fig. S6). Moreover, the mitophagic response, measured as punctal accumulation of the mitophagy adaptor protein p62 [also called sequestosome 1 (SQSTM1)], was impaired in FCCP-treated Mfn2-deficient cardiac myocytes (Fig. 2D and fig. S6). Together, these data support a model in which Mfn2 functions as a receptor to which cytosolic Parkin binds on depolarized mitochondria, provoking ubiquitination of mitochondrial proteins that target the organelle for autophagic elimination.

• Download high-res image
• Open in new tab
• Download Powerpoint

Enhancement of Mfn2-Parkin association by PINK1 indicated that Mfn2 might be phosphorylated. We noted a slight electrophoretic mobility shift in a fraction of Mfn2 when PINK1 was cotransfected, suggesting a possible posttranslational modification (Fig. 3A, top). This mobility shift was augmented on Phos-tag (Wako Chemicals USA, Incorporated, Richmond, Virginia) gels, indicating that Mfn2 might be phosphorylated at one or more sites (Fig. 3A, bottom, and fig. S7). Indeed, phosphoserine immunoreactivity of immunoprecipitated Mfn2 was also increased in cells overexpressing PINK1 (Fig. 3B). If PINK1-mediated phosphorylation of Mfn2 is required for Mfn2-Parkin interactions, then catalytically inactive PINK1 should be ineffective. Consistent with this notion, the catalytically inactive PINK1 Lys²¹⁹, Asp³⁶², Asp³⁸⁴ →Ala mutant (9) failed to promote Mfn2-Parkin binding or to induce the characteristic change in Mfn2 mobility on Phos-tag gels (Fig. 3C).

• Download high-res image
• Open in new tab
• Download Powerpoint

To establish that Mfn2 is a substrate of PINK1 and the mitochondrial binding partner for Parkin, we mapped the Mfn2 PINK1 phosphorylation sites and examined their functional consequences. Bioinformatics analysis identified three highly conserved potential Mfn2 phosphorylation sites: Thr¹¹¹ (T111), Ser⁴⁴² (S442), and Tyr⁴⁴⁸ (Y448) (figs. S8 and S9). Because PINK1 is a serine-threonine kinase, we mutated Mfn2 T111 and S442 to alanine, preventing their phosphorylation. Both Mfn2 mutations decreased PINK1-stimulated Mfn2-Parkin binding without eliminating it (Fig. 3D), whereas simultaneous mutation of both residues (T111A/S442A) completely abrogated PINK1-stimulated interactions between Mfn2 and Parkin (Fig. 3E). In agreement with these loss-of-function data, Mfn2 mutations that mimic PINK1 phosphorylation of Mfn2 (T111E/S442E) conferred PINK1-independent binding activity to Parkin (Fig. 3E). Collectively, these studies reveal a potential mechanism by which Mfn2 orchestrates the PINK1-Parkin mitochondrial quality control apparatus.

We tested whether decreased Parkin translocation caused by the absence of its mitochondrial binding partner would adversely affect long-term mitochondrial homeostasis. Because it has no Parkin binding activity, deletion of Mfn1 had no effect on mitochondrial morphometry or respiratory function of cardiomyocytes (fig. S10, A and B). However, deletion of Mfn2 provoked mitochondrial enlargement, which we measured by ultrastructural examination (Fig. 4A and fig. S10A) and flow-cytometric forward light scatter (fig. S10C). Mfn2-null cardiac myocytes also exhibited decreased substrate-dependent O₂ consumption, pointing to mitochondrial respiratory impairment (Fig. 4B).

• Download high-res image
• Open in new tab
• Download Powerpoint

Mitochondrial function deteriorates over time in senescent hearts, contributing to increased prevalence of heart failure with age (10). Because mitochondrial abnormalities caused by Mfn2 ablation resemble those of age-related cardiomyopathy, we assessed in vivo cardiac chamber dilation and contractile function over time in hearts from cardiac-specific Mfn1 and Mfn2 knockout mice. Hearts lacking Mfn1 appeared normal (fig. S11A) and had normal heart function assessed through noninvasive echocardiography (fig. S11B) or invasive hemodynamic measurements of contractility (+dP/dt, the maximal rate of increase in left ventricular pressure over time) (fig. S11C). However, hearts lacking Mfn2 dilated with increasing age (fig. S11D), developed impaired contractile performance assessed by echocardiography (Fig. 4C and fig. S11E), and were insensitive to β-adrenergic stimulation (fig. S11F). If cardiomyopathy in hearts lacking Mfn2 is caused by the defect in Parkin signaling, then primary Parkin deficiency should recapitulate the organelle and organ phenotypes. Indeed, in Drosophila lacking Parkin (11) cardiomyocyte mitochondria were enlarged (Fig. 4D), and heart tubes exhibited impaired respiration (Fig. 4E) with chamber dilation and contractile impairment (Fig. 4F and fig. S12). Parkin-deficient cardiomyocyte mitochondria were not effectively ubiquitinated after FCCP stimulation (Fig. 4G). The common properties of Mfn2-deficient mouse hearts and Parkin-deficient Drosophila heart tubes are characteristic of maladaptive remodeling leading to dilated cardiomyopathy. The inability of Parkin to promote ubiquitin-dependent mitophagy in both models may contribute to accumulation of abnormal mitochondria that ultimately impairs cellular respiration and compromises cardiac function.

Our results implicate the mitochondrial fusion protein Mfn2 as a Parkin receptor on damaged mitochondria, linking mitochondrial regeneration with selective organelle culling. Studies have suggested a role for Mfn2 in mitophagy downstream of Parkin (4, 5), but it also appears to function upstream to help translate the PINK1 signal for Parkin translocation. We envision that PINK1 stabilized in depolarized mitochondria phosphorylates Mfn2, which attracts and binds Parkin to promote mitophagy. In the absence of Mfn2, the PINK1-Parkin pathway of mitophagic quality control is interrupted, abnormal mitochondria accumulate, and cardiac toxicity ensues. A broad role for Mfn2 as Parkin receptor is indicated by mitochondrial abnormalities also observed with liver- and neuron-specific Mfn2 ablation (12, 13). Parkin translocation to mitochondria in cultured embryonic fibroblasts lacking Mfn2 (8) indicates activity for one or more compensatory Parkin binding mechanisms, although respiratory compromise suggests that these secondary mechanisms are incompletely effective (fig. S13). Like tethering to endoplasmic reticulum (14), the unique Parkin receptor function of Mfn2 distinguishes it from Mfn1. Defective Parkin binding may have importance in hereditary neuropathies genetically linked to mutations of Mfn2 and characterized by accretion of abnormal mitochondria (15). The adverse consequences of interrupting Parkin-mediated mitophagy in the heart might also provide insight into the previously unexplained epidemiological link between Parkinson’s disease and cardiomyopathy in the elderly (16).