Structural basis for the phosphorylation of FUNDC1 LIR as a molecular switch of mitophagy

ABSTRACT

Mitophagy is a fundamental process that determines mitochondrial quality and homeostasis. Several mitophagy receptors, including the newly identified FUNDC1, mediate selective removal of damaged or superfluous mitochondria through their specific interaction with LC3. However, the precise mechanism by which this interaction is regulated to initiate mitophagy is not understood. Here, we report the solution structure of LC3 in complex with a peptide containing the FUNDC1 LC3-interacting region (LIR) motif. The structure reveals a noncanonical LC3-LIR binding conformation, in which the third LIR residue (Val20) is also inserted into the hydrophobic pocket of LC3, together with the conserved residues Tyr18 and Leu21. This enables Tyr18 to be positioned near Asp19 of LC3, and thus phosphorylation of Tyr18 significantly weakens the binding affinity due to electrostatic repulsion. Functional analysis revealed that mitochondrial targeting of the LIR-containing cytosolic portion of FUNDC1 is necessary and sufficient to initiate mitophagy when Tyr18 is unphosphorylated, even in the absence of mitochondrial fragmentation. Thus, we demonstrated that phosphorylation of Tyr18 of FUNDC1 serves as a molecular switch for mitophagy. This may represent a novel target for therapeutic intervention.

Introduction

Mitophagy is defined as a critical process for selective removal of damaged mitochondria via autophagic degradation. Mitophagy acts as a primary mechanism of mitochondrial quality control, which is fundamental for normal cellular homeostasis and function. Defective mitophagy has been linked with the pathogenesis of cancers, metabolic disorders and aging-related degenerative diseases,^1-3 and the accumulation of dysfunctional mitochondria is a common feature of the etiology of these diseases. Several mitophagy receptors mediate mitophagy through their interaction with LC3. Early genetic screening in yeast identified Atg32 as a degron for mitophagy. Atg32 interacts with Atg11 and Atg8 through its conserved I/V112-L-S114 and W86-X-X-I89 sequences, respectively,^4,5 to recruit autophagy machinery and phagophore (autophagosome precursor) membranes toward mitochondria, while the Atg32-Atg8 interaction serves as an auxiliary to facilitate formation of autophagosomes. In mammalian systems BNIP3L/NIX and its homolog BNIP3 interact with LC3 and mediate mitophagy during red blood cell maturation.^4,6,7 The activity of BCL2L13/BCL-RAMBO, another mediator of mitophagy in mammalian cells, is thought to be regulated by induction and/or phosphorylation.⁸ In addition to these receptor-mediated mitophagy pathways, the PINK1-PARK2/Parkin pathway induces mitophagy when mitochondrial membrane potential is lost.^9,10 Recently, we have identified FUNDC1 (FUN14 domain containing 1) as a specific mitophagy receptor during hypoxia-induced mitophagy.^11,12 It was found that hypoxia-induced mitophagy is mostly abolished by FUNDC1 knockdown with siRNA, which can be restored by transient expression of wild-type FUNDC1. FUNDC1, an outer mitochondrial membrane protein containing 3 transmembrane domains, has its N-terminal and C-terminal regions exposed to the cytosol and the intermembrane space, respectively.

Autophagy receptors such as SQSTM1/p62 and OPTN bind to LC3 through a typical motif containing a consensus 4-residue LC3-interacting region (LIR) AxxB (Fig. S1). In the LIR motif, A is an aromatic amino acid and B is a large aliphatic hydrophobic amino acid. Both A and B insert into deep hydrophobic pockets on the surface of LC3.^13,14 FUNDC1 contains a typical LIR (Y18-E19-V20-L21) motif in its cytosol-exposed N-terminal region. FUNDC1-induced mitophagy is highly dependent on the direct interaction between its LIR motif and LC3. Under normal conditions, the LIR motif of FUNDC1 is phosphorylated at Ser13 by CSNK2/CK2 kinase and at Tyr18 by SRC kinase, which inhibits its interaction with LC3 and prevents mitophagy. In contrast, hypoxia induces dephosphorylation of FUNDC1, which can then bind to LC3 and induce mitophagy.^11,12 This dephosphorylation-induced binding is in contrast to other LIR motif-containing proteins, in which phosphorylation normally increases their LC3 binding affinities and promotes autophagy. Previous studies in yeast showed that phosphorylation of Ser71 of Atg32 promotes Atg32-Atg8 interaction.^15,16 Similarly, BNIP3 mediates mitophagy by directly interacting with LC3 in mammalian cells, and phosphorylation of the BNIP3 LIR motif at Ser17 and Ser24 promotes its binding to LC3B and GABARAPL2/GATE-16,¹⁷ although the relevant kinases and phosphatases have yet to be identified. In addition, phosphorylation of serine residues (especially Ser177) immediately preceding the LIR in the N-terminal region of OPTN also increases the affinity of the interaction with LC3.¹³

To investigate the molecular details of how the phosphorylation regulates FUNDC1-induced mitophagy, we have solved the solution structure of LC3 in complex with a FUNDC1 peptide containing the LIR Y(18)EVL(21). We provide both structural and functional evidence indicating that the reversible phosphorylation of FUNDC1 LIR residue Tyr18 functions as a molecular switch for mitophagy.

Results

The unphosphorylated cytosolic portion of FUNDC1 is sufficient to initiate mitophagy

The LIR motif of FUNDC1 is located within the N-terminal 50 amino acid residues, which are exposed to the cytosol.¹² This prompted us to speculate that the cytosolic portion of FUNDC1 may function as a degron for the recruitment of LC3-positive phagophore membranes. To test this, we constructed a mitochondrially-targeted protein (hereafter referred to as FLAG-FN^Mito) in which the N-terminal (1–50 amino acids) of wild-type FUNDC1 was fused to the transmembrane domain of BCL2L1/BCLXL, which targets the outer membrane of mitochondria (Fig. 1A).¹⁸ When wild-type FLAG-FN^Mito and GFP-LC3 were cotransfected into HeLa cells, we found that FLAG-FN^Mito was indeed localized at the membrane of mitochondria, as indicated by the colocalization with mitochondrial outer membrane protein TOMM20. Then, we subjected the cells to hypoxia for 6 h, which was previously shown to induce the dephosphorylation of FUNDC1.^11,12 Without FLAG-FN^Mito, GFP-LC3 puncta could be observed, but not localized on the mitochondria membrane. However, in the presence of FLAG-FN^Mito, many more GFP-LC3 puncta were observed and were mostly located on the mitochondrial membrane (Fig. S2A). Next, we cotransfected FLAG-FN^Mito and GFP-Cherry-LC3 plasmids into FUNDC1 knockout HeLa cells. Under hypoxia treatment, many more red fluorescent puncta than green ones were observed, all colocalized with FLAG-FN^Mito (in blue), indicating that the green fluorescence was quenched in an acidic environment due to autophagic flux. As expected, the number of green puncta was significantly increased with chloroquine (CQ) treatment to inhibit the fusion of autophagosomes with lysosomes (Fig. 1B; Fig. S2B and C).

Figure 1.

The LIR-containing cytosolic portion of FUNDC1 recruits LC3 toward mitochondria when dephosphorylated. (A) Structure of the recombinant FLAG-FN^Mito expression plasmid, which contains the FUNDC1 N-terminal region (1–50 amino acids) and the transmembrane domain of BCL2L1 (204–233 amino acids, Mito™). Residues 10–26 of FUNDC1 are shown expanded at the top. The 4-amino acid LIR motif is highlighted in yellow. Ser13 and Tyr18 are in red. (B) FUNDC1 knockout HeLa cells were cotransfected with plasmids expressing GFP-Cherry-LC3 and FLAG vector or FLAG-FN^Mito plasmids, and cells were subjected to normoxia or hypoxia (with or without 20 μm CQ) for 12 h before being fixed with 4% paraformaldehyde. Cells were stained with TOMM20 or FLAG antibodies. (C) HeLa and FUNDC1 knockout cells transfected without or with FLAG-FN^Mito and then treated with or without hypoxia for 12 h in the presence or absence of bafilomycin A1 (BAF). Samples were collected for western blotting with the indicated antibodies. The wild type FUNDC1-MYC was used as a positive control. (D) The scramble and FUNDC1 knockdown cells were transfected without or with FLAG-FN^Mito, and then treated with CSNK2/CK2 inhibitor TBBt (50 μM) and SRC inhibitor PP2 (10 μM) for 12 h. The mitochondrial marker proteins were then analyzed with the indicated antibodies.

We also used mt-Keima, an inner mitochondrial membrane-targeted fluorescent protein for detecting mitophagy,¹⁹ which is resistant to lysosomal proteases and shows a low-ratio fluorescence (543 nm:488 nm) at neutral pH and high-ratio fluorescence at acid pH, and the results also indicated that FLAG-FN^Mito restored the hypoxia-induced mitophagy with full autophagic flux in FUNDC1 knockout HeLa cells (Fig. S3). Meanwhile western blot analysis also showed that FUNDC1 knockout inhibited hypoxia-induced TIMM23 and TOMM20 degradation, whereas the degradation was enhanced with the transfection of FLAG-FN^Mito (Fig. 1C). Bafilomycin A₁ (BAF) treatment enhanced the levels of LC3-II (Fig. 1C) also indicating that FLAG-FN^Mito restored the hypoxia-induced mitophagic flux.²⁰

We have previously reported that the CSNK2/CK2 inhibitor TBBt and the SRC inhibitor PP2 can effectively induce dephosphorylation at residues Ser13 and Tyr18 of FUNDC1, respectively.¹¹ TBBt and PP2 treatment could induce mitophagy in the scrambled siRNA-transfected cells as evidenced by the degradation of TOMM20 and TIMM23, whereas the levels of TOMM20 and TIMM23 had almost no change in FUNDC1 knockdown cells with or without treatment of the 2 inhibitors (Fig. 1D). When FLAG-FN^Mito was transfected into FUNDC1 knockdown cells, the ability of TBBt and PP2 treatment in inducing mitophagy could be clearly restored. Western blot analysis indicated that the level of FLAG-FN^Mito with phosphorylated Ser13 was decreased and the SRC kinase was inactivated, due to TBBt and PP2 treatment. All these data clearly indicate that the N-terminal 50 residues of FUNDC1 is the determinant region for FUNDC1-induced mitophagy, and that the reversible phosphorylation plays an important role in switching FUNDC1-induced mitophagy.

We then constructed S13A, Y18W, and S13A Y18W mutants of FLAG-FN^Mito and tested their abilities in mitophagy induction. Similar to wild-type FLAG-FN^Mito, which failed to induce GFP-LC3 puncta, the S13A mutant of FLAG-FN^Mito, which mimics unphosphorylated Ser13, also failed to induce mitophagy. However, the Y18W mutant of FLAG-FN^Mito, which mimics unphosphorylated Tyr18 but still retains LIR activity, induced the formation of pronounced LC3 aggregates around the mitochondria. The LC3 aggregates were even more abundant in the presence of the S13A Y18W double mutant of FLAG-FN^Mito (Fig. S4A and B). Western blotting analysis showed that the Y18W and S13A Y18W mutants of FLAG-FN^Mito accelerated the degradation of mitochondrial proteins and increased the level of LC3-II (Fig. S4C), whereas wild-type FLAG-FN^Mito and the S13A mutant had no effect. The mitophagy-inducing ability of FLAG-FN^Mito was abolished by the double mutation S13A Y18A. Taken together, these results reveal that the unphosphorylated cytosolic portion of FUNDC1 serves as the determinant for LC3 recruitment, and the dephosphorylation of Tyr18 is the most critical factor for mitophagy induction, whereas the effect of Ser13 phosphorylation is limited.

The LIR motif of FUNDC1 binds to LC3 in a noncanonical conformation

We synthesized a 17-amino acid peptide encompassing the FUNDC1 LIR motif (residues 10–26 of FUNDC1; F-pep), and determined the solution structure of LC3 in complex with F-pep using NMR spectroscopy (Fig. 2A). The complex structure was finally calculated using 5050 nuclear Overhauser effect (NOE) distance restraints, 182 dihedral angle restraints and 34 H-bond restraints. The resonance assignments and structural statistics are shown in Figure S5 and Table 1, respectively (PDB ID code 2n9x and BMRB accession no. 25919). The overall structure of LC3 did not change significantly upon F-pep binding, with a root mean square deviation (RMSD) of 1.0 Å for the backbone heavy atoms of the secondary structure regions between the mean structures of free LC3 and LC3-F-pep. Similar to previously reported LC3-LIR complex structures, F-pep bound LC3 mainly through interaction of the LIR motif with the LIR docking site, which consists of 2 hydrophobic pockets, HP1 and HP2.²¹ The side chains of the 2 critical hydrophobic LIR residues, Tyr18 and Leu21, were inserted into HP1 and HP2, respectively. HP1 is comprised of residues Asp19, Ile23, Lys51, and Leu53, while HP2 is comprised of residues Phe52, Val54, Leu63, Ile66 and Ile67 (Fig. 2B). Unexpectedly, the hydrophobic side chain of Val20, the third residue of the FUNDC1 LIR, was also inserted into HP1 of LC3 together with Tyr18 and made hydrophobic contacts with the side chain of residue Leu53 in HP1, which were consistent with the NOEs between FUNDC1 residues Val20 and Tyr18 side chains (Fig. S6) and the intermolecular NOEs between FUNDC1 Val20 and LC3 residue Leu53 in HP1 (Fig. S7 and Table S1). As a result, the second residue of the LIR, Glu19, protruded away from the binding interface. In addition, several other negatively charged residues of FUNDC1, such as Glu12, Asp15 and Asp22, may have made electrostatic interactions with LC3 (Fig. 2C).

Figure 2.

Solution structure of LC3 in complex with a synthetic FUNDC1 peptide containing the LIR (F-pep). (A) Superimposition of the backbones of the ensemble of 20 structures of the LC3-FUNDC1 complex. The α helixes are colored in red, β sheets are colored in blue, and F-pep is colored in yellow. (B) Ribbon representation of the LC3-FUNDC1 mean structure. LC3 is colored in gray, and the F-pep is colored in yellow. Side chains of LC3 residues that form the 2 hydrophobic pockets are labeled in black, and 3 residues of the FUNDC1 LIR are labeled in red. (C) Electrostatic potential surface of LC3 (blue, positively charged residues; red, negatively charged residues; gray, uncharged residues). The side chains of important residues of F-pep are shown as ball-and-stick models (labeled black) and the important residues of LC3 are labeled in red. (D) Superimposed ribbon structures of the LC3-FUNDC1 and LC3-SQSTM1 complexes. In the LC3-FUNDC1 complex, LC3 is blue and F-pep is dark blue. In the LC3-SQSTM1 complex, LC3 is pink and the SQSTM1 peptide is in magenta. The LIR residues of FUNDC1 and SQSTM1 are labeled.

Table 1.

Structural statistics for the LC3-FUNDC1 complex.

NMR distance and dihedral restraints
Distance restraints
Total NOE restraints	5045
Intra-residue	1364
Sequential	726
Medium-range	443
Long-range	579
Intermolecular	43
Ambiguous	1739
Hydrogen bond restraints	34
Dihedral angle restraints
Φ	115
ψ	83
Structure statistics (20 structures)
Restraints violations
Distance (> 0.2 Å)	1
Dihedral angle (> 5 °)	2
Max. distance constraint violation (Å)	0.26
Max. dihedral angle violation (°)	6.5
RMSD from mean structure (Å)
Backbone heavy atoms of LC3	0.53 ± 0.10
All heavy atoms of LC3	1.16 ± 0.09
Backbone heavy atoms of complex	0.65 ± 0.12
All heavy atoms of complex	1.22 ± 0.10
Backbone heavy atoms of F-pep	1.22 ± 0.44
Ramachandran statistics (%)
Most favored regions	82.9
Additional allowed regions	13.7
Generously allowed regions	1.3
Disallowed regions	2.0

Note. RMSD were calculated only for residues in the secondary structure regions of LC3 and/or residues 17–23 of F-pep.

We compared the LC3-FUNDC1 structure with other reported LC3-LIR complex structures, including SQSTM1 (pdb id: 2K6Q;¹⁴), OPTN (pdb id: 2LUE;¹³), PLEKHM1 (pdb id: 3X0W;²²), ATG4B (pdb id: 2Z0D;²³) and ATG13 (pdb id: 3WAO;²⁴), and found that the binding conformation of the FUNDC1 LIR motif was significantly different from the others. When the LC3 structures were superimposed and FUNDC1 was compared with each of the other proteins, the RMSDs for the backbone heavy atoms of the 4-residue LIR motifs were rather large (FUNDC1:SQSTM1 was 4.6 Å [Fig. 2D], FUNDC1:OPTN was 5.6 Å, FUNDC1:PLEKHM1 was 5.2 Å, FUNDC1:ATG4B was 5.0 Å, FUNDC1:ATG13 was 5.1 Å). In contrast, the average pairwise RMSD for the backbone heavy atoms of the 4-residue LIR motifs of SQSTM1, OPTN, PLEKHM1, ATG4B and ATG13 was only 1.4 ± 0.3 Å (Fig. S8). This conformational difference between the FUNDC1 LIR motif and the LIRs of the other LC3-binding proteins can be explained by the fact that the hydrophobic side chain of the third residue of the FUNDC1 LIR (Val20) was also inserted into HP1, whereas the side chains of the corresponding third LIR residues of all the other proteins are positioned outside of HP1. As a result, the first aromatic LIR residue (Tyr18) of FUNDC1 was pushed toward the α1 helix of LC3 and was not inserted into HP1 as deeply as the first residue of the other LIRs. Therefore, the FUNDC1 LIR motif adopts a rather unusual conformation in binding LC3.

Previous studies revealed that the formation of an intermolecular β-sheet between the LIR and LC3 is also a conserved feature among LC3-LIR complex structures. In the LC3-SQSTM1 structure, the SQSTM1 LIR peptide forms an intermolecular β-sheet with LC3, with the amide proton of LC3 residue L53 forming a H-bond with the carbonyl oxygen of SQSTM1 residue T339, and the amide proton of SQSTM1 residue L341 forming a H-bond with the carbonyl oxygen of LC3 residue L53, which is a feature of parallel β-strands.¹⁴ However, our LC3-FUNDC1 complex structure indicated that FUNDC1 LIR peptides are not in a position to form intermolecular β-sheets with LC3. Also, we found no sign of β-sheet formation in our NMR data for FUNDC1 residues E19 and L21, which correspond to SQSTM1 residues T339 and L341, respectively.

Tyr18 phosphorylation dictates the interaction of FUNDC1 with LC3

In the structure of LC3-F-pep, the side-chain hydroxyl group of FUNDC1 Tyr18 was positioned close to the side-chain carboxyl group of Asp19 of LC3 HP1, while Ser13 of FUNDC1 was located in a hydrophobic area between the side chains of Lys49 and Lys51. It is reasonable to assume that phosphorylation of Tyr18 would result in electrostatic repulsion between the phosphate group and the side chain of Asp19 in HP1, while the phosphorylation of Ser13 should not perturb the binding significantly.

To better understand the effect of phosphorylation on the activation of mitophagy in response to mitochondrial stress, we synthesized 3 phosphorylated peptides, F-Y18p-pep (Tyr18 phosphorylated), F-S13p-pep (Ser13 phosphorylated) and F-S13p/Y18p-pep (both Ser13 and Tyr18 phosphorylated), and then compared their interactions with LC3 to that of unphosphorylated F-pep using NMR titration experiments (Fig. S9A to C). The NH proton chemical shift perturbation pattern of LC3 with F-S13p-pep was similar to that with F-pep (Fig. 3A), indicating that the phosphorylation of S13 only had a minor effect on binding. The chemical shift perturbation patterns of LC3 with F-Y18p-pep and F-S13p/Y18p-pep were similar to each other, but were quite different from that of F-pep. Compared to F-pep, the chemical shift changes of LC3 in the presence of F-Y18p-pep and F-S13p/Y18p-pep were smaller for residues in the α1 helix but larger for residues on the β2 strand and its preceding loop (Fig. 3A). This suggests that phosphorylated Tyr18 moves toward the β2 strand and away from the α1 helix where Asp19 is located.

Figure 3.

Phosphorylation of Tyr18 significantly weakens the binding affinity of FUNDC1 LIR peptides for LC3. (A) Overlay of representative residues (R21 and K49) of the 2D ¹H-¹⁵N HSQC spectra of LC3 titrated by 4 different FUNDC1 peptides at different concentration ratios: black (0), red (0.25), magenta (0.5), cyan (1.0), and blue (2.0). (B) ITC titration of LC3 with 4 different FUNDC1 peptides. The upper panel shows the raw ITC data, and the lower panel shows the integrated heat per titration step (points) and best-fit curves. The dissociation constant (K_d) of each peptide is shown in red.

We further measured the dissociation constant (K_d) for the interaction of these peptides with LC3 by isothermal titration calorimetry (ITC) experiments. The results showed that the K_d for F-pep was 0.40 μM, and the phosphorylation of Tyr18 drastically weakened the binding of F-Y18p-pep with LC3 (K_d 1.72 μM). In contrast, the phosphorylation of Ser13 only had a limited effect on LC3 binding, as the K_d for F-S13p-pep was slightly higher (0.60 μM), while the K_d for F-S13p/Y18p-pep (2.00 μM) was similar to that of F-Y18p-pep (Fig. 3B). These results indicated that the phosphorylation of Ser13 only had a minor effect on the binding affinity of FUNDC1 to LC3, whereas the phosphorylation of Tyr18 significantly weakened the binding affinity.

As the side-chain hydroxyl group of FUNDC1 Tyr18 is close to the side-chain carboxyl group of Asp19 of LC3, we tested whether Asp19 determines the selectivity of LC3 for unphosphorylated and phosphorylated FUNDC1 by constructing a D19N mutant of LC3 (LC3^D19N) (Fig. S10). We first examined the interaction of LC3^D19N with F-pep and F-Y18p-Pep by NMR titration experiments, and found that the NH chemical shift perturbation patterns of LC3^D19N were quite similar for both peptides, suggesting that the binding pattern of LC3^D19N was similar for both unphosphorylated and phosphorylated peptides (Fig. 4A). Consistent with these results, the K_d values for the interaction of LC3^D19N with both peptides were also similar: 0.86 μM for F-pep and 0.85 μM for F-Y18p-pep (Fig. 4B). We also tested the ability of FUNDC1 to induce mitophagy in HeLa cells transfected with wild-type GFP-LC3 or the GFP-LC3^D19N mutant. As expected, compared with wild-type FUNDC1-MYC, overexpression of the Y18W mutant of FUNDC1-MYC enhanced the aggregation of wild-type GFP-LC3. However, in GFP-LC3^D19N-transfected HeLa cells, wild-type FUNDC1-MYC and its Y18W mutant showed similar levels of GFP-LC3^D19N puncta (Fig. 4C and D). Immunoprecipitation analysis also revealed that the unphosphorylated FUNDC1-MYC Y18W mutant bound more GFP-LC3 than wild-type FUNDC1-MYC, whereas FUNDC1-MYC or its Y18W mutant had similar affinities for the GFP-LC3^D19N mutant (Fig. 4E). These results indicate that the D19N mutation almost abolishes the selectivity of LC3 toward unphosphorylated and phosphorylated FUNDC1. Taken together, these results suggest that the phosphorylation status of Tyr18 in the LIR motif Y(18)EVL(21) plays a central role in regulating the binding affinity of FUNDC1 with LC3 and determines the mitophagy-inducing activity of FUNDC1.

Figure 4.

The D19N mutation of LC3 abolishes its selectivity toward the phosphorylated LIR of FUNDC1. (A) Overlay of 2D ¹H-¹⁵N HSQC spectra of wild-type LC3 (upper panel) and the D19N mutant LC3 (lower panel) titrated by a 2-fold concentration of F-pep (black) or F-Y18p-pep (red). (B) ITC titration of the LC3 D19N mutant with F-pep and F-Y18p-pep. The upper panel shows the raw ITC data, and the lower panel shows the integrated heat per titration step (points) and best-fit curves. The dissociation constant (K_d) of each peptide is shown in red. (C) HeLa cells were cotransfected with GFP-LC3 or mutant GFP-LC3^D19N and FUNDC1-MYC or FUNDC1-MYC-Y18W as indicated. Cells were fixed with 4% paraformaldehyde and stained with the indicated antibodies. Boxed areas in the merged images are enlarged in the “zoom” images. (D) The area covered by GFP-LC3 aggregates in cells treated as in Figure 4C was quantified via ImageJ (mean ± SEM; n = 100 cells from 3 independent experiments; *p < 0.05). (E) After overexpression of the indicated plasmids for 24 h, HeLa cells were collected for immunoprecipitation (IP) with anti-MYC and analyzed by western blotting using anti-MYC or anti-GFP antibodies.

The third residue of LIR plays a critical role in LC3 binding

As the hydrophobic side chain of FUNDC1 Val20 is also inserted into LC3 HP1, we mutated Val20 to Ala and examined the binding of F-V20A-pep (V20A mutant F-pep) to LC3. NMR titration experiments indicate that the NH chemical shift perturbation pattern of LC3 in the presence of F-V20A-pep was significantly different from that with F-pep, especially for the residues in HP1 and HP2, such as Gln26, His27, Lys30, Val33, Phe52, Leu53, Val54 and Val58 (Fig. S11A). The affinity of F-V20A-pep for LC3 (K_d 3.72 μM) was ∼10 times lower than that of F-pep (Fig. 5D). To verify the functional importance of Val20, we constructed a vector expressing a V20A mutant of FUNDC1-MYC and co-expressed it with GFP-LC3 in HeLa cells. Compared with wild-type FUNDC1-MYC, the level of mitophagy induced by overexpression of the V20A mutant was significantly reduced, as indicated by fewer GFP-LC3 puncta, reduced conversion of LC3-I to LC3-II, and accumulation of the mitochondrial proteins TOMM20 and TIMM23 (Fig. 5A to C). Consistent with the NMR titration and ITC results, the V20A mutation significantly impaired the mitophagy-inducing ability of FUNDC1, and its effect was similar to that of the Y18A and L21A mutations.¹² These results clearly confirm that in addition to the canonical LIR residues Tyr18 and Leu21, the hydrophobic residue Val20 is also critical for FUNDC1 to bind LC3 and induce mitophagy.

Figure 5.

The V20A mutation significantly impairs the mitophagy-inducing ability of FUNDC1. (A) HeLa cells were cotransfected with plasmids expressing GFP-LC3 and full-length FUNDC1-MYC, FUNDC1-MYC-E12Q, FUNDC1-MYC-D15N, FUNDC1-MYC-V20A or FUNDC1-MYC-D22N for 24 h, and then fixed in 4% paraformaldehyde and stained with anti-MYC antibody before analysis by immunofluorescence microscopy. Boxed areas in the merged images are enlarged in the “zoom” images. (B) The area covered by GFP-LC3 aggregates in cells treated as in Figure 5A was quantified using ImageJ (mean ± SEM; n = 100 cells from 3 independent experiments; *p < 0.05). (C) HeLa cells were transfected with full-length FUNDC1-MYC or the indicated mutants, then analyzed by western blot to detect mitochondrial markers (TOMM20, TIMM23) and the change in LC3-II levels. (D) ITC titration of wild-type LC3 with F-V20A-pep. The upper panel shows the raw ITC data, and the lower panel shows the integrated heat per titration step (points) and best-fit curves. The dissociation constant (K_d) is shown in red.

As residues Glu12, Asp15, and Asp22 of FUNDC1 are located at the FUNDC1-LC3 interface and may interact electrostatically with LC3, we also synthesized 3 mutant peptides, F-E12Q-pep, F-D15N-pep and F-D22N-pep, and examined their interaction with LC3 by NMR titration experiments (Fig. S11B to D). The NH chemical shift perturbation patterns of LC3 in the presence of F-E12Q-pep and F-D22N-pep were quite similar to that with F-pep, whereas the difference in titration patterns between F-D15N-pep and F-pep was a little larger. We also examined the induction of mitophagy in HeLa cells transfected with E12Q, D15N or D22N mutants of FUNDC1-MYC together with wild-type GFP-LC3. These 3 mutations do not have a statistically significant impact on the mitophagy-inducing activity of FUNDC1 (Fig. 5A to C). These results further demonstrate that the binding of FUNDC1 to LC3 relies principally on the 4-residue LIR of FUNDC1 (Y18-E19-V20-L21).

Discussion

We have shown that the ability of the newly identified mitophagy receptor FUNDC1 to bind LC3 and induce mitophagy is negatively regulated by phosphorylation. This is in sharp contrast to other autophagic LC3 receptor proteins, in which phosphorylation normally enhances LC3 binding and autophagy induction.^11,12 Our structural and functional analyses reveal that the negative regulation of the FUNDC1-LC3 interaction is primarily due to the phosphorylation of Tyr18, the first conserved residue of the LIR. Unlike other mitophagy receptors, phosphorylation of the conserved Tyr18 of the LIR motif serves as a unique molecular switch to control the level of mitophagy.

In all the previously published structures of LC3 or its homologs in complex with LIR motifs, it is a common feature that the side chains of the first aromatic residue and the fourth hydrophobic residue of the LIR are inserted into 2 hydrophobic pockets of LC3, while the side chains of the 2 middle LIR residues are floating outside. However, it is evident that the LIR of FUNDC1 adopts a noncanonical conformation when binding LC3, characterized by insertion of the side chain of the third LIR residue (Val20) into the same hydrophobic pocket as the first LIR residue (Tyr18). As a result, the side-chain hydroxyl group of Tyr18 is positioned very close to the side-chain carboxyl group of Asp19. Therefore, phosphorylation of Tyr18 will result in electrostatic repulsion between the phosphate group and the carboxyl group of Asp19, which will weaken the binding between LC3 and FUNDC1. Indeed, LC3 has a higher affinity for the FUNDC1 LIR motif when Tyr18 is unphosphorylated than when it is phosphorylated. This preference is abolished when Asp19 of LC3 is mutated to Asn. The same mutation also eliminates the effect of phosphorylation on the degree of mitophagy induced by FUNDC1. In contrast, our structural and biochemical analysis showed that phosphorylation of Ser13 of FUNDC1 alone only slightly lowers the binding affinity of LC3 for the FUNDC1 LIR. There is a difference in binding affinity from published results, most likely due to a purity problem with the synthetic peptides used in the previous study.¹¹ However, our findings further highlight the critical role of Tyr18 phosphorylation in regulating LC3 binding.

We have previously shown that FUNDC1 directly interacts with LC3B and other LC3 paralogs, and it appears that LC3B has the highest affinity.¹² As the structures of LC3A, LC3B, LC3C, GABARAP, and GABARAPL2 are very similar, and the residues comprising the 2 hydrophobic pockets are mostly conserved among them, it is likely that FUNDC1 can interact with other LC3 paralogs in a similar fashion, although their binding affinities could be significantly different from that of LC3B.

In accordance with our structural analysis, we found that the LIR-containing cytosolic N-terminal 50 residues of FUNDC1, when fused with a mitochondrial targeting sequence, are necessary and sufficient to induce mitophagy when FUNDC1 is dephosphorylated due to hypoxia or the treatment with kinase inhibitors TBBt and PP2. Consistent with these findings, a genetically engineered Y18W mutant, which mimics the unphosphorylated form of LIR while retaining its binding activity, can also directly induce mitophagy in the absence of any stimuli (Fig. S4A to C). Our functional analysis also shows that the S13A mutation of an artificial LC3 receptor (FLAG-FN^Mito) has little effect on the mitophagy induction ability, whereas it can enhance mitophagy induced by the Y18W mutation (Fig. S4A to C). This finding indicates that the mechanism underlying the inhibitory role of S13 phosphorylation on FUNDC1-induced mitophagy is not through directly affecting FUNDC1-LC3 interaction. We have recently shown that full-length FUNDC1 induces mitochondrial fragmentation, most likely through direct interaction of its cytosolic N-terminal domain and intermembrane space region with the mitochondrial fission factor DNM1L/DRP1 and the mitochondrial fission/fusion factor OPA1, respectively.²⁵ It was found that phosphorylation of residue S13 of full-length FUNDC1 reduces its interaction with DNM1L and enhances its interaction with OPA1, which downregulates FUNDC1-induced mitophagy. Conversely, an S13A mutation can enhance the interaction of FUNDC1 with DNM1L and reduce its interaction with OPA1, while FUNDC1-induced mitophagy is upregulated as a result. Taken together, phosphorylation of the FUNDC1 Y18 residue is responsible for regulating its interaction with LC3, whereas phosphorylation of the FUNDC1 S13 residue may be auxiliary for enhancing mitophagy via mitochondrial fission factors.

Materials and methods

Protein expression and purification

A cDNA encoding LC3B (residues 1–120) was subcloned into the PMCSG7 (BioVector, EvNO00084271) vector with an N-terminal 6×His tag followed by an rTEV protease cleavage site. The recombinant plasmid was transformed into E. coli strain BL21 (DE3). For expression and purification, the bacteria were cultured in LB medium at 35°C until OD₆₀₀ = 0.8. Cells were harvested and resuspended in ¹⁵N, ¹³C-labeled M9 medium (3 g/l KH₂PO₄; 17.2 g/l Na₂HPO₄•12H₂O; 0.5 g/l NaCl; 1 g/l NH₄Cl; 1 mM MgSO₄; 1 g/l ¹⁵NH₄Cl and 4 g/l ¹³C₆-glucose) and incubated for another 1 h, then 100 μM IPTG (Inalco, 1758–1400) was added and the cells were incubated for another 8 h at 35°C to induce protein expression. Cells were harvested and resuspended in lysis buffer (50 mM sodium phosphate, 300 mM NaCl, pH 8.0), and lysed by freezing and thawing, followed by sonication. After centrifugation (10,625 × g, 30 min, 4°C), the supernatant fraction was applied to a Ni-NTA affinity column (Qiagen, 30250) and eluted by elution buffer (50 mM sodium phosphate, 300 mM NaCl, 250 mM imidazole, pH 8.0). The eluted protein was cleaved overnight with rTEV protease (Solarbio, P2060) at room temperature to remove the His-tag, and further purified by a superdex-75 column (GE Healthcare, 17-1044-02), using 25 mM sodium phosphate, 100 mM NaCl, pH 7.0.

NMR spectroscopy

The NMR samples contained approximately 0.4 mM ¹⁵N, ¹³C-labeled LC3 protein and 0.8 mM synthesized FUNDC1 peptide (SciLight Peptide) in 25 mM sodium phosphate, 100 mM NaCl, pH 7.0 with 90% H₂O:10% D₂O and 0.01% 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS, an NMR standard). All NMR experimental data were collected on a Bruker Avance 700 or 800 MHz NMR spectrometer with a cryoprobe at 298 K. For backbone assignment of LC3, 2D ¹H-¹⁵N HSQC, 3D HNCACB, 3D CBCA(CO)NH, and 3D HNCO experiments were performed. 3D HBHA(CO)NH, 3D (H)CCH-COSY, 3D (H)CCH-TOCSY, 3D H(C)CH-COSY, and 3D H(C)CH-TOCSY experiments were performed for side-chain assignments of LC3. Distance restraints for structure calculation of LC3 were derived from the 3D ¹H-¹⁵N NOESY-HSQC and 3D ¹H-¹³C NOESY-HSQC, with a mixing time of 100 ms. A 2D F1, F2-¹³C/¹⁵N-filtered NOESY experiment with a mixing time of 200 ms was performed to get the intramolecular NOEs of the FUNDC1 peptide, while a 3D F1-¹³C/¹⁵N-filtered, F2-¹³C-edited NOESY experiment also with a mixing time of 200 ms was performed to obtain the intermolecular NOEs of LC3 and the FUNDC1 peptide. All NMR spectra were processed using NMRPipe and analyzed with NMRView.

Structure calculation

For the structure calculation, initial structures were generated by CYANA with restraints from the CANDID module.²⁶ The distance restraints, dihedral angle restraints and H-bond restraints were determined by SANE, TALOS and secondary structure elements in the initial structure, respectively. All these restraints were further used in DYANA to gain refined structures, which were used as filter models in another round of SANE-DYANA calculation.^27,28 When there was no distance violation bigger than 0.5 Å and no angle violation bigger than 5°, 100 structures with the lowest target function values among the 200 CYANA-calculated structures were selected for further refinement using AMBER 12.²⁹ SANE-AMBER calculation was carried out until no distance violation was bigger than 0.2 Å and no angle violation was bigger than 5°. The top 20 structures with the lowest AMBER energies were selected and a mean structure was generated by SUPPOSE. The quality of the structures was analyzed using PROCHECK-NMR.³⁰

NMR titration

0.1 mM ¹⁵N-labeled LC3 protein (wild-type or mutants) was dissolved in 25 mM sodium phosphate, 100 mM NaCl with 90% H₂O:10% D₂O and 0.01% DSS. The titration experiments were performed at pH 7.0 and monitored by 2D ¹H-¹⁵N HSQC experiments. The concentration ratio of FUNDC1 peptides (SciLight Peptide) to LC3 ranged from 0.25 to 2, except for F-V20A-pep, where the molar ratio ranged from 0.25 to 3.

Isothermal titration calorimetry

ITC experiments were performed using a MicroCal iTC₂₀₀ system (GE Healthcare) at 298 K. 0.1 mM LC3 protein or D19N mutant was placed in the cell and titrated with FUNDC1 peptide (1 mM) or its mutants (1.5 mM) injected in 2.0-μL aliquots. Corresponding “peptide to buffer” controls were performed for background correction. ITC titration data were analyzed using Origin 7.0 (OriginLab) provided with the instrument. Standard deviation was calculated by Origin according to the fit.

Cell culture and antibodies

HeLa cells were grown in DMEM supplemented with 10% fetal bovine serum (Thermo, SV30087.02) and 1% penicillin-streptomycin at 37°C under 5% CO₂. Hypoxic conditions were achieved in a hypoxia chamber (Billups-Rothenberg, Del Mar, CA, USA) flushed with a preanalyzed gas mixture of 1% O₂, 5% CO₂ and 95% N₂. The CRISPR/Cas9 system was used to generate a FUNDC1 knockout HeLa cell line. The FUNDC1 knockdown cell line was screened by Dr. Lei Liu in Dr. Quan Chen's lab. The following antibodies were used for western blot immunofluorescence: anti-TIMM23 (BD, 611222), anti-TOMM20 (BD, 612278), anti-LC3 monoclonal antibody (MBL, PM036); anti-MYC monoclonal antibody (Santa Cruz Biotechnology, B1115); anti-FLAG polyclonal antibody (Sigma, F7425); anti-GFP monoclonal antibody (Santa Cruz Biotechnology, A2115); anti-ACTB monoclonal antibody (Sigma, A5441); anti-SRC (Cell Signaling Technology, 2108S); anti-SRC (p-Tyr416) polyclonal antibody (Cell Signaling Technology, 6943S); and anti-FUNDC1 polyclonal antibody (AVIVA, ARP53280_P050). The anti-p-FUNDC1 (Ser13) polyclonal antibody (1:1,000) was affinity purified after immunizing rabbits with purified FUNDC1 phosphopeptides. The following fluorescent secondary antibodies were used: goat anti-rabbit 555 (Invitrogen, A31572); goat anti-mouse Cy5 (Invitrogen, M35011).

Immunofluorescence microscopy

Cells were grown to 60% confluence on a coverslip. After treatment, cells were washed twice with phosphate-buffered saline (136.9 mM NaCl, 2.7 mM KCl, 1.5 mM KH₂PO₄, 8.1 mM Na₂HPO₄•12H₂O) and fixed with freshly prepared 4% paraformaldehyde at 37°C for 30 min. Antigen accessibility was increased by treatment with 0.2% Triton X-100 (Shanghai Sangon Biotech, TB0198). Cells were incubated with primary antibodies for 2 h at room temperature and, after washing with phosphate-buffered saline, stained with a secondary antibody for a further 1 h at room temperature. Cell images were captured with a TCS SP5 Leica confocal microscope. The images for statistical analysis of the GFP-LC3-positive area were captured by Zeiss Axio Imager Z1 immunofluorescence microscopy.

Western blotting

Cells were treated with the indicated conditions, then were lysed in 1% SDS. Equivalent protein quantities (20 µg) were subjected to SDS-PAGE, and transferred to nitrocellulose membranes. Membranes were probed with the indicated primary antibodies, followed by the appropriate HRP-conjugated secondary antibodies (KPL, 1912I). Immunoreactive bands were visualized with a chemiluminescence kit (Millipore, WBKLS0500).

Immunoprecipitation

Cells were transiently transfected using polyethylenimine (Polysciences,23966). At 24 h post transfection, the cells were lysed with 0.5 ml of lysis buffer (20 mM Tris, pH 7.4, 137 mM NaCl, 2 mM EDTA, 10% glycerol, 1% NP-40 (Shanghai Sangon Biotech, A510110) plus cocktail protease inhibitors (Roche Applied Science, 11206893001) for 50 min on ice. After centrifugation at 12,000 g for 10 min, the lysates were immunoprecipitated with specific antibody and then protein A/G-agarose beads (Abmart, A10001) overnight at 4°C. Thereafter, the beads were washed 5 times with lysis buffer, and the immune complexes were eluted with sample buffer containing 1% SDS for 5 min at 95°C and analyzed by SDS-PAGE.

Statistical analysis

For quantitative analyses of cultured cells represented as histograms, values were obtained from 3 independent experiments and expressed as the mean ± SEM. Statistical analyses were performed using the Student t test, with p values < 0.05 being considered significant. Significance levels relative to the controls are indicated by *p < 0.05, **p < 0.01 and ***p < 0.001in the figures. All statistical analyses were performed with GraphPad Prism software.

Abbreviations

ACTB

actin β

FPLC

fast-flow protein liquid chromatography

FUNDC1

FUN14 domain containing 1

HP

hydrophobic pocket

HSQC

heteronuclear single quantum coherence

ITC

isothermal titration calorimetry

LIR

LC3-interacting region

NMR

nuclear magnetic resonance

NOE

nuclear overhauser effect

PDB

protein data bank

RMSD

root mean square deviation

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgments

All NMR experiments were performed at Beijing NMR Center and the NMR facility of National Center for Protein Sciences at Peking University. We are grateful to Prof. Luhua Lai and Tongqing Li for their assistance in ITC experiments, and to Dr. Hongwei Li and Dr. Xiaogang Niu for their assistance in NMR experiments.

Funding

This research was supported by Grant 2012CB910703 and 2016YFA0501202 from Ministry of Science and Technology of China to BX, and Grant 2016YFA0500201 from Ministry of Science and Technology of China, Grant 31471300 from Natural Science Foundation of China, Grant 5161002 from Natural Science Foundation of Beijing, Grant QYZDJ-SSW-SMC004 from CAS Key project of Frontier Science to QC, and Grant 201520103904 from Natural Science Foundation of China to YZ.