MFN2 agonists reverse mitochondrial defects in preclinical models of Charcot-Marie-Tooth disease type 2A

See allHide authors and affiliations

Science  20 Apr 2018:
Vol. 360, Issue 6386, pp. 336-341
DOI: 10.1126/science.aao1785

An innovative approach for a rare disease

Charcot-Marie-Tooth disease type 2A (CMT2A) is a rare, inherited neurodegenerative condition. Affected individuals develop severe progressive muscle weakness, motor deficits, and peripheral neuropathy. Although defects in the gene encoding mitofusin 2 (MFN2) are known to cause CMT2A, the disease remains incurable. Rocha et al. identified specific MFN2 residues contributing to the disease and developed a class of MFN2-agonist drugs. The small molecules restored mitochondrial fusion and activity in the sciatic nerves of mice; they may also help in other diseases linked to mitochondrial trafficking.

Science, this issue p. 336


Mitofusins (MFNs) promote fusion-mediated mitochondrial content exchange and subcellular trafficking. Mutations in Mfn2 cause neurodegenerative Charcot-Marie-Tooth disease type 2A (CMT2A). We showed that MFN2 activity can be determined by Met376 and His380 interactions with Asp725 and Leu727 and controlled by PINK1 kinase–mediated phosphorylation of adjacent MFN2 Ser378. Small-molecule mimics of the peptide-peptide interface of MFN2 disrupted this interaction, allosterically activating MFN2 and promoting mitochondrial fusion. These first-in-class mitofusin agonists overcame dominant mitochondrial defects provoked in cultured neurons by CMT2A mutants MFN2 Arg94→Gln94 and MFN2 Thr105→Met105, as demonstrated by amelioration of mitochondrial dysmotility, fragmentation, depolarization, and clumping. A mitofusin agonist normalized axonal mitochondrial trafficking within sciatic nerves of MFN2 Thr105→Met105 mice, promising a therapeutic approach for CMT2A and other untreatable diseases of impaired neuronal mitochondrial dynamism and/or trafficking.

Mitochondria are organelles that generate a rich energy source for cells, which requires their continuous subcellular redistribution via mitochondrial trafficking and mutual repair via mitochondrial fusion (1). Mitochondrial fusion and subcellular trafficking are mediated in part by mitofusin 1 (MFN1) and MFN2. Genetic mutations in Mfn2 that suppress mitochondrial fusion and motility cause Charcot-Marie-Tooth disease type 2A (CMT2A), the most common inheritable axonal neuropathy (2). Because no therapeutics exist that directly enhance mitochondrial fusion or trafficking, this disease is unrelenting and irreversible.

Computational modeling based on the closed structure of bacterial dynamin–related protein (BDRP) and the more open structure of optic atrophy-1 suggested that MFN2 can change conformation according to how closely the first and second heptad repeat (HR) domains interact (fig. S1). A closed conformation is fusion incompetent, whereas an open conformation favoring mitochondrial fusion can be induced by a competing peptide analogous to amino acids 367 to 384 within the MFN2 HR1 domain (3). We identified amino acids controlling these events, first by truncation analysis to define the smallest fusion-promoting minipeptide (residues 374 to 384) (Fig. 1, A and B) and then through functional investigation of this minimal peptide by alanine (Ala) scanning. Substitution of Ala for Met376, Ser378, His380, and Met381, which are highly conserved across vertebrate species (figs. S2 and S3), impaired minipeptide-stimulated mitochondrial fusion, as measured by an increase in the mitochondrial length/width ratio (aspect ratio) (Fig. 1C). The structural model of human MFN2 in a closed conformation on the basis of homology with BDRP predicted a helical interaction between HR1 and HR2 domains, with alignment of Met376 and His380 side chains in the HR1 domain to Leu727 and Asp725 in the HR2 domain (fig. S1). This arrangement suggested that Met376 and His380 stabilize the MFN2 HR1-HR2 interaction, potentially explaining their critical function as defined by minipeptide Ala scanning. By contrast, Ser378 was modeled as extending from the noninteracting surface of the HR1 α helix (fig. S1), implying a different mechanism for its involvement in mitochondrial fusion.

Fig. 1 MFN2 Ser378 phosphorylation by PINK1 regulates mitochondrial fusion.

(A) Amino acid sequence surrounding the fusion-promoting MFN2 peptide. Side chain characteristics (H, hydrophobic; +, basic; −, acidic) are indicated above. HR1 hinge region amino acids are green. The open box encloses N-terminal residues 367 to 374, and the shaded box encloses the C-terminal minipeptide comprising residues 374 to 384 (minipeptide 374–384). Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr. hMfn2, human MFN2. (B) Mitochondrial fusion stimulated by N- and C-terminal minipeptides. The aspect ratio is the ratio of the mitochondrial long axis to the short axis. (Inset) Fusion in MFN1- or MFN2-null MEFs. White elements correspond to treatment with HR1 minipeptide 367–374, and black elements correspond to treatment with HR1 minipeptide 374–384. Data are means ± SEM of results from three independent experiments; P values (shown above the bars in the inset) were determined by Student’s t test. (C) Alanine (A) scanning of minipeptide 374–384 fusion activity. Veh, vehicle. (D) Ser378 substitution analysis of minipeptide 374–384 fusion activity in MFN2-null MEFs. Data in (C) and (D) are means ± SEM of results from three, five, or eight independent experiments, as indicated. P values in (C) and (D) (shown above the bars) are versus the parent minipeptide 374–384 [analysis of variance (ANOVA)]. (E) Binding of minipeptides with Ser378 substitutions to the HR2 target sequence. Data are means ± SEM of result from six experimental replicates. (F) Binding of the Asp378 minipeptide to the HR2 target sequence before (left) and after (right) Ala substitution for putative interacting amino acids. Data are means ± SEM of results from six experimental replicates. P values were determined by Student’s t test. NS (above bars), not significant; NS (below bars), nonspecific binding; FITC, fluorescein isothiocyanate. (G) Ion chromatograms from assigned MFN2 Ser378 phosphopeptide fragment ions (identified by “y” designations) after incubation with PINK1 kinase (top) and a stable isotope-labeled synthetic counterpart (bottom). Proportional intensities are in adjacent stack plots. The graph shows coelution of the synthetic phosphopeptide with the PINK1 product, as well as the positive identification of the product ions from tandem mass spectrometry. The lowercase “s” in the sequence indicates the phosphorylated residue. (H) Mitochondrial fusion promoted by MFN2 Ser378 mutants with and without PINK1 kinase. An immunoblot of protein expression is shown at the bottom. Data are means ± SEM of results from four independent experiments. P values were determined by ANOVA. β-Gal, β-galactosidase.

To address whether Ser378 might be phosphorylated, we replaced Ser378 (with Ala, Cys, Asn, or Gly) in the minipeptide and found that phosphorylation and fusion activity were abrogated. Functionality was restored by substituting Asp to mimic phosphorylated Ser or by inserting phospho-Ser itself (Fig. 1D and fig. S4). Moreover, in an in vitro binding assay devoid of cellular kinases, the Asp378-containing minipeptide bound to its putative HR2 interacting domain whereas Ser378 and Ala378 minipeptides did not (Fig. 1E). Elimination of minipeptide binding by replacement of HR2 Leu724, Asp725, and Leu727 with Ala confirmed the HR1-HR2 interaction model (Fig. 1F).

Nuclear magnetic resonance spectrometry of the minipeptides showed low conformational stability with a propensity to form helical structures. Ser378 phosphorylation reduced the peptide dynamics most visibly for residues Leu379 to Met381, potentially changing amino acid side chains presented to HR2 (fig. S5, data S1, and movie S1). Recombinant MFN2 mutations that replaced Ser378 with Asp (mimicking MFN2 Ser378 phosphorylation) or substituted Ala for Met376 or His380 (disrupting the putative HR1-HR2 interaction controlled by Ser378 phosphorylation) impaired MFN2-stimulated mitochondrial fusion (fig. S6). By contrast, replacing MFN2 Ser378 with Ala (to prevent phosphorylation) or substituting Ala for neighboring Val372, which is not critical for HR1-HR2 interactions, did not depress MFN2-mediated fusion (fig. S6).

MFN2 can be phosphorylated by mitochondrial PTEN-induced putative kinase 1 (PINK1) (4, 5). Targeted mass spectrometry demonstrated phosphorylation of MFN2 Ser378, as well as MFN2 Thr111 and Ser442, which were previously reported (4, 5), by PINK1 kinase (Fig. 1G, figs. S7 and S8, and table S1) but not by software-nominated G-protein receptor kinase 2 (fig. S9). We expressed MFN2 Ser378 mutants with or without PINK1 in MFN1 and MFN2 doubly deficient cells (designated MFN1−/− MFN2−/−). Fusion-defective mitochondria in these cells were abnormally short at baseline, but forced expression of wild-type (WT) Ser378 MFN2 resulted in elongation from restoration of fusion (Fig. 1H and fig. S10). Coexpression of PINK1 with MFN2 or mutational replacement of MFN2 Ser378 with Asp (which mimics PINK1-mediated Ser378 phosphorylation) restrained MFN2-stimulated elongation (Fig. 1H and fig. S10). By contrast, MFN2 Ala378 (which cannot be phosphorylated) promoted mitochondrial fusion resistant to PINK1 suppression (Fig. 1H and fig. S10). The effects of MFN2 Ser378 mutants were recapitulated in in vitro assays of fusion-mediated mitochondrial content exchange (fig. S11).

We assessed fusogenic activity of commercially available small-molecule candidate pharmacophores (data S2 and supplementary materials and methods), focusing on those having structures that mimicked Ser378-phosphorylated (class A) and -unphosphorylated (class B) minipeptide amino acid side chains (fig. S12 and data S3). We reasoned that simultaneous application of class A and B agonists could enhance mitofusin function by acting on both MFN2 Ser378 phosphorylation states. Lead compound A (Cpd A) and Cpd B acted synergistically to promote mitochondrial fusion (fig. S13; compare aspect ratios with those in fig. S12C). Therefore, we assimilated Cpd A and Cpd B functionality into asingle molecule by creating Cpd A–Cpd B chimeras (Fig. 2A, fig. S14, and data S4). Chimera B-A/long (B-A/l) potently stimulated mitochondrial fusion in MFN2-deficient cells (Fig. 2B), competed for minipeptide binding at the MFN2 HR2 interaction site (Fig. 2C), and was as effective as the combination of Cpd A and Cpd B in reversing mitochondrial dysmorphology provoked by the fusion-defective CMT2A mutant MFN2 Thr105→Met105 (MFN2 T105M) (Fig. 2D). Fusogenic effects of Cpd A were specific for the Asp378 mutant of MFN2 that mimicked Ser378 phosphorylation, whereas Cpd B and chimera B-A/l were nonselective for the phosphomimic Asp378 and nonphosphorylatable Ala378 mutants (Fig. 2E). Because they mimic the WT MFN2 HR1 sequence and interact with HR2, mitofusin agonists evoked fusion to proportionally similar degrees in mitochondria expressing mutants of HR1 that are fusion deficient (Fig. 2F; compare with Fig. 1H and fig. S6). Small-molecule mitofusin agonists required endogenous MFN1 or MFN2 to promote mitochondrial fusion, exhibited no detectable promiscuous activity for structurally related dynamin, and did not compromise cell viability (fig. S15). On the basis of fluorescence resonance energy transfer (FRET) analysis of MFN2 labeled at the N and C termini (supplementary materials and methods), mitofusin agonists promoted an open MFN2 conformation favoring mitochondrial fusion (3), with a rank order paralleling that for HR2 binding and mitochondrial fusion (Fig. 2G; compare with Fig. 2, B and C), supporting allosteric activation.

Fig. 2 Small-molecule mimetics of MFN2 HR1 amino acid side chains that interact with HR2 are mitofusin agonists.

(A) (Top) Three-dimensional representations of hypothetical minipeptide conformations driven by Ser378 phosphorylation, and (bottom) their respective small-molecule mimetics. (p)Ser378, phospho-Ser378. (B) Dose-dependent mitofusin agonism by small-molecule agonists. Data are means ± SEM of results from six independent experiments. EC50, mean effective concentration. (C) Displacement of minipeptide 374–384 from its HR2 binding site by mitofusin agonists. Data are means ± SEM of results from three independent experiments. IC50, mean inhibitory concentration. (D) Restoration of MFN2 T105M-impaired mitochondrial fusion in MFN2−/− MEFs by mitofusin agonists. Data are means ± SEM of results from three or four independent experiments, as indicated. P values were measured by ANOVA. (E) Selectivity of a class A but not a class B mitofusin agonist for Ser378-phosphorylated MFN2. Data are means ± SEM of results from four independent experiments. P values were measured by ANOVA. (F) Impaired basal function but normal proportional agonist responsiveness of MFN2 with mutations altering HR1-HR2–interacting amino acids. The absolute fusogenicity of these MFN2 mutants is depicted in fig. S6. Data are means ± SEM of results from four independent experiments. P values were determined by ANOVA (left) or Student’s t test (right). (G) Change in FRET evoked by mitofusin agonists in isolated mitochondria (left) and intact cells (right). Decreased FRET reflects conformational opening. Data are means ± SEM of results from 3 independent (left) and 14 to 16 replicate (right) experiments. P values versus the vehicle were measured by ANOVA. Cer, Cerulean.

In CMT2A, MFN2 mutants produce mitochondrial “fragmentation” (defined as decreased aspect ratio) and loss of normal membrane polarization through dominant inhibition of normal mitofusins. Experiments using MFN1−/− MFN2−/− murine embryonic fibroblasts (MEFs) showed that in the absence of normal mitofusins, small-molecule mitofusin agonists did not improve mitochondria of cells expressing the guanosine triphosphatase (GTPase)–crippled MFN2 Arg94→Gln94 (R94Q) or Lys109→Ala109 (K109A) mutant (Fig. 3A). However, mitofusin agonists corrected the mitochondrial dysmorphology and reversed the mitochondrial hypopolarization induced by these MFN2 mutants when endogenous MFN1 was present (Fig. 3B). Mitofusin agonists also reversed mitochondrial fragmentation and hypopolarization in cultured mouse neurons expressing (in addition to endogenous mitofusins) CMT2A mutant MFN2 R94Q (Fig. 3, C and D) or CMT2A mutant MFN2 T105M (Fig. 3E). Thus, mitofusin agonists do not restore function of CMT2A MFN2 GTPase domain mutants. Rather, by stabilizing the fusion-permissive open conformation of endogenous normal MFN1 or MFN2, mitofusin agonists can overcome dominant suppression of mitochondrial fusion by these disease-causing dysfunctional proteins.

Fig. 3 Mitofusin agonists correct mitochondrial damage induced by nonfunctioning MFN2 mutants by activating endogenous mitofusins.

(A and B) Effects of mitofusin agonists in MFN1−/− MFN2−/− cells (A) and in MFN1+/+ MFN2−/− cells (B) expressing recombinant WT or mutant MFN2. Data are means ± SEM of results from three independent experiments. P values were measured by ANOVA. Mito, mitochondria. (C) (Top) Genotyping of MFN2 R94Q individual mouse pups from which primary cultured neurons were derived. TG, transgenic MFN2 R94Q; NTG, nontransgenic. (Bottom) Representative confocal images of mitochondrial pathology in cultured neonatal mouse neurons expressing MFN2 R94Q, and correction by mitofusin agonists. Magnified views are from white squares. Ctrl, control; MitoGFP, mitochondrion-targeted green fluorescent protein; TMRE, tetramethylrhodamine ethyl ester. (D) Group data for mitochondrial aspect ratio and polarization studies in cultured neonatal mouse neurons expressing MFN2 R94Q, as depicted in (C). Data are means ± SEM of results from three independent experiments. P values were measured by ANOVA. (E) Results of mitochondrial aspect ratio and polarization studies in cultured neonatal mouse neurons expressing MFN2 T105M. Data are means ± SEM of results from four independent experiments. P values were measured by ANOVA.

Clinical CMT2A classically affects long nerves innervating the lower and upper limbs (6, 7). It is unclear how a principal defect in mitochondrial fusion would cause length-dependent neuronal disease. Conversely, disruption of axonal mitochondrial trafficking (8) would be predicted to preferentially affect cells requiring mitochondrial transport over the greatest physical distance, such as the sciatic nerves originating in the spine and terminating in the foot. MFN2 interacts with the Miro/Milton complex to promote mitochondrial motility in neurons (9), so we tested the effects of mitofusin agonism on murine neuronal mitochondrial trafficking. Chimera B-A/l reversed mitochondrial “clumping” (formation of static mitochondrial aggregates) and restored mitochondrial motility in cultured mouse neurons expressing the CMT2A mutant MFN2 T105M (Fig. 4A and movie S2). Mitochondrial hypopolarization and increased autophagy (Fig. 4B and fig. S16) and mitochondrial dysmorphology (Fig. 4C and fig. S16) were concomitantly ameliorated. Thus, a small-molecule mitofusin agonist enhanced organelle and cell fitness in CMT2A neurons by promoting mitochondrial fusion and subcellular transport.

Fig. 4 A mitofusin agonist restores axonal mitochondrial trafficking suppressed by CMT2A mutant MFN2 T105M.

(A to C) Effects of chimera B-A/l on mitochondrial mobility (A), function (B), and morphology (C) in cultured CMT2A MFN2 T105M–expressing mouse neurons. Data are means ± SEM of results from four independent experiments. P values were measured by ANOVA. (D) Representative kymograph of mitochondrial trafficking in a Ctrl mouse sciatic nerve. Scale bar, 10 μm. (E) Representative serial kymographs of mitochondria in an MFN2 T105M mouse sciatic nerve before and after treatment with chimera B-A/l. m, minutes. (F) Quantitative data for sciatic nerve mitochondrial motility studies. (G) Sizes of motile and static mitochondria in Ctrl and B-A/l–treated (60 min) sciatic nerves. Data in (F) and (G) are means ± SEM of results from four or five independent experiments. P values were measured by ANOVA.

We evaluated the concept of activating mitofusins to stimulate in vivo axonal mitochondrial trafficking in sciatic nerves of mice expressing the CMT2A mutant MFN2 T105M. In normal sciatic nerves, ~30% of axonal mitochondria exhibited robust bidirectional transport (Fig. 4D and movie S3). Mitochondria of MFN2 T105M sciatic nerves were severely hypomotile (Fig. 4E and movie S4), but application of chimera B-A/l to MFN2 T105M sciatic nerves restored mitochondrial motility to within normal levels (Fig. 4F and movie S4). Mobile mitochondria in WT and B-A/l–treated MFN2 T105M axons were smaller than static mitochondria (Fig. 4G), supporting in vitro observations discriminating between MFN2-mediated mitochondrial dysmotility and defective fusion in CMT2A (10).

In this study, we found that PINK1 phosphorylation of MFN2 at Ser378 can alter the positions of Met376 and His380 (in the HR1 domain), which normally interact with HR2 domain amino acids to orchestrate MFN2 toggling between conformations that modulate mitochondrial fusion. These findings establish a mechanistic basis for clinical observations that MFN2 Met376 mutations to Ile, Thr, and Val can cause CMT2A (7, 11, 12).

By combining in silico pharmacophore modeling with structural and functional interrogation of MFN2 HR1 domain–derived minipeptides, we developed a novel small-molecule mitofusin agonist that reversed the neuronal mitochondrial dysmorphometry and impaired mobility evoked by two CMT2A Mfn2 mutants. CMT2A is the prototypical clinical disorder of defective mitochondrial fusion, but impaired mitochondrial trafficking may play as great a role as mitochondrial fragmentation in CMT2A axonal degeneration (810). Individuals with CMT2A express one mutant MFN2 allele in combination with one normal MFN2 allele and harbor two normal MFN1 alleles (13). It is therefore possible that a therapeutic substrate exists for mitofusin agonists to “supercharge” normal mitofusins and overcome dominant inhibition by MFN2 mutants. Our observation that in vivo mitochondrial dysmotility provoked by a CMT2A mutant can be normalized by mitofusin agonism mechanistically links abnormal mitochondrial trafficking in experimental CMT2A to MFN2 dysfunction. Mitofusin agonists may also have therapeutic potential for neurological conditions other than CMT2A, such as Alzheimer’s, Parkinson’s, and Huntington’s diseases, wherein mitochondrial dysmotility and fragmentation are contributing factors (1416).

Supplementary Materials

Materials and Methods

Figs. S1 to S31

Table S1

References (1730)

Movies S1 to S4

Data S1 to S4

References and Notes

Acknowledgments: We gratefully acknowledge discussions with P. Needleman and the assistance of L. Zhang, P. Erdmann-Gilmore, Y. Mi, and R. Connors. Funding: This work was supported by NIH grants R35HL135736 (G.W.D.), R01HL128071 (R.N.K. and G.W.D.), and R01CA178394 and P30CA013330 (E.G.); a McDonnell Center for Cellular and Molecular Neurobiology postdoctoral fellowship (A.F.); and the Washington University Proteomics Shared Resource, supported by National Center for Advancing Translational Sciences grants UL1TR000448, NIGMSP41, GM103422, and NCIP30 CA091842. G.W.D. is the Philip and Sima K. Needleman–endowed professor. Author contributions: G.W.D., A.M.K., D.M.-R, and R.R.T. conceived of or designed the research, except the initial in silico screen. E.G., R.N.K., N.B., and E.Z. conceived of the small-molecule screen, designed the original pharmacophore model, and performed the initial in silico screen. G.W.D. wrote the manuscript. J.M.R. and R.R.T. performed phosphoprotein mass spectroscopy analyses. A.M.K. performed peptide nuclear magnetic resonance studies. A.G.R. screened compounds for activity and characterized agonists. A.G.R., A.F., J.M.A., and W.C.K. performed mitochondrial studies. A.F. performed cultured neuron and ex vivo sciatic nerve studies. J.W.J. analyzed and purified compounds. R.N.K., E.G., and R.H.B. provided essential reagents. Competing interests: D.M.-R. and G.W.D. are inventors on patent application 15/710,696, submitted by Stanford University, which covers the use of peptide regulators of mitochondrial fusion and small-molecule peptidomimetics derived from them. G.W.D. is an inventor on provisional patent applications 62/488,787 and 62/584,515, submitted by Washington University, which cover the use of novel small-molecule mitofusin agonists to treat chronic neurodegenerative diseases. E.G., R.N.K., N.B., and E.Z. are inventors on patent application 62/573,217, submitted by Albert Einstein College of Medicine, which covers compositions of mitofusin agonists and their uses for the treatment of diseases and disorders. D.M.-R. is the founder of Mitoconix Bio, a company focused on improving mitochondrial health as a therapeutic approach for neurodegenerative diseases. None of the research conducted in D.M.-R.’s laboratory is supported by or performed in collaboration with Mitoconix. The other authors declare no competing interests. Data and materials availability: All data are available in the manuscript or the supplementary materials. There are no material transfer agreements associated with this study.

Stay Connected to Science

Navigate This Article