Molecular recognition and substrate mimicry drive the electron-transfer process between MIA40 and ALR

Abstract

Oxidative protein folding in the mitochondrial intermembrane space requires the transfer of a disulfide bond from MIA40 to the substrate. During this process MIA40 is reduced and regenerated to a functional state through the interaction with the flavin-dependent sulfhydryl oxidase ALR. Here we present the mechanistic basis of ALR–MIA40 interaction at atomic resolution by biochemical and structural analyses of the mitochondrial ALR isoform and its covalent mixed disulfide intermediate with MIA40. This ALR isoform contains a folded FAD-binding domain at the C-terminus and an unstructured, flexible N-terminal domain, weakly and transiently interacting one with the other. A specific region of the N-terminal domain guides the interaction with the MIA40 substrate binding cleft (mimicking the interaction of the substrate itself), without being involved in the import of ALR. The hydrophobicity-driven binding of this region ensures precise protein–protein recognition needed for an efficient electron transfer process.

Disulfide bonds are important for the structure and function of proteins in eukaryotes, prokaryotes, and even viruses. Several enzymes are known to catalyze dithiol-disulfide transfer reactions between proteins to efficiently form or disrupt disulfide bonds (1). A class of them includes sulfhydryl oxidases that are capable of forming disulfide bonds de novo (2, 3). In general, these enzymes exist as homodimers, depend on FAD as a cofactor, and use oxygen as final electron acceptor. They typically contain a CXXC motif, involved in the redox-reaction and close to the FAD molecule, and amino- and/or carboxy-terminal segments having Cys-conserved residues or motifs. The Saccharomyces cerevisiae protein Erv1 (essential for respiration and vegetative growth) and the human homologue ALR (augmenter of liver regeneration) are sulfhydryl oxidases working in the intermembrane space of mitochondria (4). ALR is found in a large number of different cell-types and tissues (5). Its activity is essential for the survival of the cell, for the biogenesis of mitochondria and for the maturation of cytoplasmic proteins with mitochondrially assembled iron–sulfur clusters (4–7). ALR is found in two main alternatively spliced forms (8). The long form of the oxidase (lf-ALR, 23 kDa) exists predominantly in the mitochondrial intermembrane space (IMS) and contains an 80-amino acid N-terminal extension with respect to the short form (sf-ALR, 15 kDa) (9) (Fig. S1) that is an extracellular cytokine and also participates in intracellular redox-dependent signaling pathways (10). In yeast, there is only the spliced form equivalent to the long form of ALR, Erv1 (Fig. S1), which is localized exclusively in mitochondria (11). The N-terminal parts of full-length ALR and Erv1 (N-terminal shuttle domain, hereafter) are very distinct, in contrast to their more similar (40% identity) C-terminal parts that include a FAD-binding domain with a proximal CXXC motif (Fig. S1). Both N-termini of ALR and Erv1 contain an additional CXXC motif (also called distal disulfide; Fig. S1). Previous genetic studies in S. cerevisiae demonstrated that all Cys residues of Erv1 were required for its function in vivo (12). These motifs at the N- or C-terminus have been proposed to work as electron shuttling from the substrate to the FAD molecule. Recent Cys-specific mutagenesis data on Erv1 support this proposal (13–16).

The up-to-now reported physiological partners of ALR are human Mia40 (MIA40 hereafter) (13) and cytochrome c (17), both of which are mitochondrial proteins. MIA40 is an oxidoreductase that promotes the import and oxidative folding of proteins in the IMS (18–22). This process produces the formation of disulfide bonds in MIA40-substrates, specifically containing twin CX₉C or CX₃C motifs, with a consequent reduction of the disulfide bond in the CPC (C53 and C55) active site of MIA40 (20). ALR has been proposed to be responsible for the reoxidation of the disulfide present in the CPC motif of MIA40, thus regenerating MIA40 to a functional state, ready to accept electrons from another mitochondrial imported substrate molecule (13). Cytochrome c is also part of this electron-transfer pathway; its interaction with ALR is proposed to restore the oxidized redox state of the cysteines of ALR that are involved in the oxidation of the CPC motif of MIA40 (17). In this process oxidized cytochrome c accepts electrons through the FAD molecule of ALR (17). This cascade of interactions is not yet fully understood at the structural level and the mechanism of these electron-transfer reactions still remains largely speculative. We present here the mechanistic basis of ALR–MIA40 molecular recognition at atomic resolution by biochemical and structural analyses of the mitochondrial ALR isoform and its covalent mixed disulfide intermediate with MIA40.

Results

The Electron-Transfer Mechanism between MIA40 and ALR.

Wild-type sf-ALR and lf-ALR were obtained from Escherichia coli (SI Text) with one FAD per molecule of protein (assessed by the typical flavin absorbance peaks at 455 and 375 nm, with a 280/450 nm absorbance ratio of approximately 5; Fig. S2). To avoid unwanted oligomerisation due to the presence of the nonconserved cysteines (C74 and C85 in sf- ALR and C154 and C165 in lf-ALR), we mutated these to Ala (Fig. S2). These mutants still behaved as a dimer but did not show any oligomeric species (SI Text and refs. 13 and 17). The C74A/C85A sf-ALR and C154A/C165A lf-ALR mutants exhibited no difference in the UV/visible spectra as well as activity toward DTT comparable to that of the wild-type sf- and lf-ALR (13, 17). We therefore used these mutants in the rest of this study.

The electron-transfer process between either long or short forms of oxidized ALR (lf-ALR_4S-S and sf-ALR_3S-S, respectively) and partially reduced MIA40 (MIA40_2S-S) have been investigated in vitro monitoring the reaction by ¹H-¹⁵N heteronuclear single quantum coherence (HSQC) NMR spectra. lf-ALR_4S-S can quantitatively oxidize MIA40_2S-S to fully oxidized MIA40_3S-S (Fig 1 A and B). Upon addition of lf-ALR_4S-S, the amide (NH) resonances pattern of MIA40_2S-S changes to that of MIA40_3S-S [i.e., to the form of MIA40 where C53 and C55 are oxidized (Fig 1 A and B)]. Similarly, sf-ALR_3S-S can also oxidize MIA40_2S-S but only partially. Both lf-ALR_4S-S and sf-ALR_3S-S are stable dimers in solution being able to receive four electrons from two MIA40_2S-S molecules. The reaction of lf-ALR_4S-S with MIA40_2S-S proceeds to completion at 0.5∶1 molar ratio, whereas with sf-ALR_3S-S the electron-transfer reaction with MIA40_2S-S proceeds only to a 50% with the same molar ratio (Fig. S3). Previously, Thorpe and coworkers (13) reported that, during in vitro assays of oxidation catalyzed by ALR, both sf- and lf-ALR have an enzymatically active core, but MIA40 can be oxidized only by the long form of ALR. An important role for the N-terminal shuttle domain in the yeast homologue, Erv1, has also been demonstrated recently in vitro (14, 16), in organello (14) and in vivo (15).

Fig. 1.

Oxidation process of MIA40_2S-S by lf-ALR_4S-S as followed by NMR. The ¹H-¹⁵N HSQC spectrum of a 1∶0.5 ¹⁵N-labeled MIA40_2S-S/unlabeled C154/165A lf-ALR_4S-S mixture (in red) is superimposed with the ¹H-¹⁵N HSQC spectra of (A) MIA40_2S-S (in black) or (B) MIA40_3S-S (in black). NH resonances of cysteine residues of the CPC motif and of some surrounding residues are indicated.

Our NMR data suggested that there must be a specific noncovalent protein-protein recognition between the long form of ALR and MIA40 driving the electron-transfer process. We measured this directly by isothermal titration calorimetry using lf-ALR and sf-ALR in binding experiments in solution (SI Text). Yeast Mia40 (Mia40 hereafter) mutated in its CPC motif into SPS was used to avoid intermolecular disulfide bond formation that would mask the noncovalent interaction. Such an approach was previously used to assay the noncovalent protein–protein interaction (i.e., with reduced cysteines) between Mia40 and its substrates (23), or between Mia40 and Erv1 (14). We found that lf-ALR interacts with high affinity with Mia40 (Kd = 25 μM; Fig. S4), whilst sf-ALR has only a marginal binding affinity, 14 times weaker than lf-ALR (Kd = 350 μM; Fig. S4). This data, together with our NMR experiments, support the idea that the N-terminal shuttle domain mediates a specific protein–protein interaction of ALR with MIA40.

Monitoring chemical shift changes on ¹⁵N-labeled sf-ALR_3S-S upon titration with unlabeled MIA40_2S-S, we also found that the CEEC motif close to the FAD molecule and the neighboring residues are affected by the reaction with MIA40 (Fig. S3). On the contrary, the chemical shifts of the NHs of the other four cysteines C91, C108, C15, C124 (involved in disulfide bonds, see later) remain unchanged, indicating that these sites are not involved in the electron-transfer reaction. It also appears that the MIA40_2S-S signal intensity decreased with increasing amounts of lf-ALR_4S-S or sf-ALR_3S-S and, concomitantly, signals corresponding to MIA40_3S-S appeared and increased in intensity, without the detection of additional signals belonging to a possible protein–protein complex. This behavior therefore indicates that the latter does not accumulate in solution in a sufficient amount to be detected by NMR. Accordingly, ¹⁵N heteronuclear relaxation rates of sf-ALR in the final protein mixture provide a value of 15.1 ± 0.9 ns for the correlation time for protein tumbling, which is comparable with the value for the isolated protein (15.6 ± 0.7 ns).

In conclusion, all the data indicate that the CPC motif of MIA40 is involved in electron-transfer with ALR and that the N-terminal segment of ALR containing the distal CRAC motif specifically recognizes MIA40 to efficiently reactivate it.

Structural Basis for the Electron Shuttle Mechanism in ALR.

The crystal and solution structural characterization of sf-ALR shows that the protein consists of a 30-kDa homodimer connected by two intermolecular disulfide bonds, namely C15-C124′ and C15′-C124, where ′ denotes the other molecule of the dimer (SI Text and Fig. S5). During preparation of this manuscript Thorpe and colleagues reported an X-ray structure for a His-tagged construct of sf-ALR (24) that is essentially identical to ours. The lf-ALR isoform has not provided any crystal. Therefore, an NMR approach is necessary for its characterization. The ¹H-¹⁵N HSQC map of lf-ALR shows NH signals belonging to the FAD-binding domain with chemical shifts similar to those of sf-ALR (Fig. 2A) plus a large number of signals clustered in the central region of the spectrum (amide proton resonances between 8 and 8.5 ppm, Fig. 2A), several of the latter showing negative or low (< 0.5) ¹⁵N{¹H} NOE values, both properties being characteristic of unfolded polypeptides. These data indicates that lf-ALR is constituted by the same cone-shaped five-helical bundle fold (α1–α5) present in sf-ALR plus an unstructured and highly flexible N-terminal segment. A number of NH cross-peaks belonging to residues close to the CEEC-FAD redox active center in the FAD-binding domain show, however, significant chemical shift variations with respect to those in sf-ALR (Fig. 2B), indicating that the N-terminal shuttle domain interacts with the FAD-binding domain in the proximity of the CEEC-FAD redox active center. To identify which residues of the N-terminal shuttle domain are involved in the interaction with the FAD-binding domain, we cloned and expressed a set of constructs of ALR with different lengths at the N-terminus (see Materials and Methods). Then, by comparing ¹H-¹⁵N HSQC experiments of them with that of lf-ALR we found that the construct with the additional residues ASRRRPCRACVDFKTW with respect to sf-ALR (named CRAC sf-ALR hereafter) is sufficient to provide a ¹H-¹⁵N HSQC spectrum with all cross-peaks perfectly superimposed with the corresponding ones of ¹H-¹⁵N HSQC of lf-ALR (Fig. S6). We can therefore conclude that only these sixteen residues of the N-terminal shuttle domain are specifically involved in the interaction with the FAD-binding domain. ¹H-¹⁵N HSQC spectra of CRAC sf-ALR at different temperatures compared with that of sf-ALR shows that the interaction between the two regions is largely abolished at 312K, indicating its transient nature (Fig. S7). This is in agreement with the electron shuttling role of this specific region, which has to recognize, on one side, MIA40 to acquire electrons and, on the other side, the CEEC-FAD redox active center to donate electrons. Secondary structure analysis based on backbone chemical shifts (HN, N, CO, CA, CB) also shows that the residues FKTWM of CRAC sf-ALR have a high propensity to form an α-helical turn, as also predicted through bioinformatic tools (25, 26) for ALR as well as for the Erv1 corresponding region. This information, together with the NMR chemical shift perturbation data, allowed us to build a model of the interaction between the CRAC-containing N-terminal segment and the FAD-binding domain of ALR (SI Text for details). We found that the interactions holding together the two latter regions are mainly of electrostatic nature involving several charged residues (Fig. 2B). The hydrophobic residues that are downstream of CRAC motif (F, W, and M), are all solvent-exposed on the same side of the α-helical turn formed downstream the CRAC motif (Fig. 2B).

Fig. 2.

Structural characterization of lf-ALR. (A) Superimposition of ¹H-¹⁵N HSQC spectra of ¹⁵N-labeled C154/165A lf-ALR (black) and ¹⁵N labeled C74A/C85A sf-ALR (red); (B) Experimental data-driven docking model of homodimeric CRAC sf-ALR. The N-terminal shuttle domain of one subunit (in red) interacts with the FAD-binding domain of the other subunit (in magenta). The residues of the FAD-binding domain experiencing largest chemical shift differences between lf-ALR and sf-ALR are mapped as black spheres. Proximal and distal disulfides are in yellow; the hydrophobic residues, Phe, Trp and Met dowstream the CRAC motif are in blue. Lys/Arg (in cyano) and Asp/Glu (in green) involved in intersubunit electrostatic interactions are also shown.

In conclusion, the structural model of lf-ALR allows us to identify the residues of the N-terminal shuttle domain specifically interacting with the CEEC-FAD redox active center.

Electron Transfer Is Driven by Shuttling of the N-Terminal of ALR between the Hydrophobic Cleft of MIA40 and the FAD-Binding Domain of ALR.

To shed light on the molecular features of electron transfer between ALR and MIA40, the protein–protein recognition site has been investigated. Due to the transient nature of the MIA40–ALR complex, we designed specific mutations on both proteins that blocked progression of the reaction allowing us to isolate the complex in a stable form. Specifically, we mutated one of the two cysteine residues of the active sites of MIA40 (C53S) and of CRAC sf-ALR (C71S or C74S) to prevent the completion of the reaction. We thus generated two complexes that are stabilized by an intermolecular disulfide bond between one or the other of the two cysteines of the CRAC motif of CRAC sf-ALR construct and C55 of the CPC motif of MIA40 [which is known to be the essential cysteine in vivo and is crucial for mixed disulfide formation with its substrates (20)] (see SI Text for details). These two isolated complexes have a 2∶1 MIA40/ALR stoichiometry (Fig. S8). We have alternatively ¹⁵N-labeled one of the two partners in both covalent complexes so that we can selectively identify the regions of interaction on each partner. The chemical shift perturbation analysis, performed through ¹H-¹⁵N transverse relaxation optimized specroscopy (TROSY)-HSQC experiments, on both CRAC sf-ALR constructs complexed with MIA40, shows that the residues constituting the cone-shaped five-helical bundle fold of ALR are essentially unaffected (Fig. 3 A and C), indicating that they are not involved in the protein–protein recognition process. On the contrary, the residues in the N-terminal shuttle domain are largely perturbed, demonstrating their specific interaction with MIA40 (Fig. 3 A and C). Chemical shift perturbation analysis also shows that the two complexes behave similarly in terms of residues affected by protein–protein interaction, even if in the C71S CRAC sf-ALR/C53S MIA40 adduct the chemical shift changes mapped on MIA40 are more clustered, suggesting a more specific protein–protein recognition (Fig. 3 B and D). In MIA40, NH chemical shifts of the residues constituting the hydrophobic cleft of MIA40 are largely perturbed upon complex formation (Fig. 3 B and D). These hydrophobic residues are essentially the same ones that are involved in protein recognition of the CX₉C or CX₃C substrates (23), suggesting that the recognition process between MIA40 and ALR is based on the same molecular grounds found in the MIA40/CX₉C or CX₃C substrate interaction. Considering that (i) the latter interaction is based on hydrophobic contacts established through an amphipathic α-helix of the substrate (23, 27), and (ii) the residues close to CRAC motif show a tendency to form an α-helical amphipatic turn exposing hydrophobic residues to the solvent, we investigated the effect of these hydrophobic residues in complex formation with MIA40 through mutagenesis. MIA40/Mia40 interacts with ALR/Erv1 in two ways: (i) with unfolded and reduced Erv1 during its import (28, 29), and (ii) with folded and oxidized ALR/Erv1 after its import to reoxidize the CPC motif of MIA40/Mia40 (13, 30). Our analysis focused so far exclusively on the second interaction; i.e., between folded/oxidized ALR and MIA40. Because it is not yet known what part(s) of the unfolded ALR/Erv1 target the protein for import, we wanted to ascertain that mutations of the hydrophobic residues downstream of the distal disulfide, do not affect the import process of ALR/Erv1. We tested this in import assays. We generated 3 mutants in the CRSCNTLLDFQ segment of Erv1 (homologous to the ALR CRACVDFKTWM segment), namely mutant LLQ/A, LLFQ/A and LLF/E (SI Text). Import of these as a radioactive precursor into WT mitochondria and their capacity to bind during import to endogenous Mia40 was assessed (Fig. 4). Control WT Erv1 was imported and interacted with Mia40 in a time-dependent manner forming a β-mercaptoethanol sensitive intermediate (“Mia40–Erv1;” Fig. 4A, lane 5). All three mutants (Fig. 4 B and C) were imported at levels comparable to the WT Erv1 and maintained their capacity to make the transient intermediate with Mia40 during import. This shows that these mutations does not affect the Mia40-dependent import into the IMS. Therefore, the specific interactions between this segment and MIA40, monitored by NMR (using the folded ALR and folded MIA40 proteins), must specifically underpin the step of reoxidation of MIA40 by ALR. To directly prove this point, we further analyzed in organello the functional defects of these mutants in the following two-step import reaction: purified mutants or WT Erv1 (as indicated) were first imported in Gal-Erv1 yeast mitochondria lacking endogenous Erv1 (14) and allowed to mature for 30 min. Subsequently, the Erv1-replenished mitochondria were isolated and used for import of radioactive Mia40. In this way we could detect in the IMS environment the interaction of the different folded Erv1 versions with radioactive Mia40 specifically in the oxidation step of the CPC motif of Mia40 (14). Without preimport of Erv1, Mia40 is imported but no complex with Erv1 is formed (lanes 2–5, Fig. 4D). After preimport of WT Erv1 a clear complex of preimported Erv1 with radioactive Mia40 is formed (Mia40–Erv1; lanes 6–9 in Fig. 4D). However, preimporting either the LLFQ/A or the LLF/E mutant (lanes 1–5 and 6–9 respectively, Fig. 4E), the Mia40–Erv1 complex was completely abolished despite an efficient import of Mia40. Import levels of all Erv1 versions are comparable as indicated by the experiment in Fig. 4 A–C. This data therefore supports the idea that the hydrophobic residues of the region downstream of the distal disulfide have a specific and vital role for the interaction of imported/folded Erv1 with the CPC active site of Mia40.

Fig. 3.

Structural characterization of the ALR/MIA40 covalent adducts. (A) ¹H-¹⁵N HSQC spectra of the ¹⁵N-labeled C71S CRAC sf-ALR (black) and in the complex with unlabeled C53S MIA40 (red). NH chemical shift changes of some interacting residues are indicated; (B) Residues of C53S MIA40 experiencing backbone chemical shift changes upon complex formation with C71S CRAC sf-ALR are mapped on the C53S MIA40 structure in red or in orange (orange 0.05 < Δ_HN < 0.1 ppm, redΔ_HN > 0.1 ppm); (C) ¹H-¹⁵N HSQC spectra of the ¹⁵N-labeled C74S CRAC sf-ALR (black) and in the complex with C53S MIA40 (red). NH chemical shift changes of some interacting residues are indicated; (D) residues of C53S MIA40 experiencing backbone chemical shift changes upon complex formation with C71S CRAC sf-ALR are mapped on the C53S MIA40 structure in red or in orange (see B).

Fig. 4.

Effect of mutation of hydrophobic residues of Erv1 on the interaction with Mia40 in organello and cell viability in vivo. (A) WT Erv1, (B) LLFQ/A Erv1 mutant, and (C) LLF/E Erv1 mutant. Import of radioactive versions of Erv1 are indicated and their capacity to form the intermediate with endogenous Mia40 (Mia40-Erv1) was analyzed. (D) and (E) Purified Erv1 variants were preimported in Erv-depleted mitochondria and allowed to mature. Subsequently, radioactive Mia40SPC was imported and its interaction with the different mature Erv1 versions was assayed. 10% of the translation mix used for the import was loaded as standard. (F) In vivo yeast complementation assays. Comparison of the growth of a GAL-ERV1 inactivated S. cerevisiae strain supplemented with plasmids expressing WT Erv1 and Erv1 mutants (LLFQ/A and LLF/E) on galactose (SG), glucose (SC), and lactate (SL) containing plates after 24 and 48 h of preincubation in liquid cultures (different dilutions 10⁵, 10⁴, 10³, 10², 10 as indicated). A growth defect is found for the Erv1 mutants grown on glucose and lactate.

We further tested this concept in a reconstituted system in vitro (SI Text and refs. 14 and 30). Briefly, when pure Mia40 partially reduced in its CPC motif was incubated with pure WT Erv1 it gave a covalent complex. This was abolished when the mutants LLF/E, LLF/A and LLFQ/A were used (Fig. S9). Specificity of the interaction was confirmed using a mutant Mia40 in its CPC active site (the SPS mutant), or in six of the hydrophobic residues of the substrate cleft, as these mutants failed to form the covalent adduct with Erv1 (Fig. S9).

To investigate if the mutations of the hydrophobic residues downstream of the CRAC motif affects directly the electron-transfer reaction, we performed an NMR titration between ¹⁵N-MIA40_2S-S and unlabeled FWM/E lf-ALR_4S-S mutant (functionally equivalent to the LLF/E mutant of Erv1). The electron-transfer reaction does not proceed to completion at the 0.5∶1 molar ratio but only to 50%, as observed in the case of sf-ALR (Fig. S3). These NMR data suggest that MIA40 can still exchange electrons with ALR, but through the CEEC motif of ALR, bypassing the CRAC-mediated electron shuttling transfer mechanism. This route is, however, less efficient compared to exchanging electrons via the CRAC shuttling motif, arguing for a specific hydrophobicity-driven recognition between the N-terminal shuttle domain of ALR and MIA40. This “conduit” more efficiently drives the electrons from CPC motif of MIA40 to the distal disulfide and then down to the FAD moiety.

As the above three independent sets of experiments (import assays/interaction in organello, reconstitution assays with pure proteins in vitro and NMR titration experiments) argued very strongly for a vital role of the hydrophobic residues downstream of the distal disulfide, we tested whether these mutants are critical in vivo. To this end, yeast complementation assays were performed (Fig. 4F) where a galactose-inactivated Erv1 strain was tested for viability upon expression of the different Erv1 mutants (LLF/E, LLFQ/A) or the WT Erv1 as a control (SI Text). Each one of the two sets of mutations was lethal arguing for a vital role of these amino acids in vivo.

Together, the data of (i) NMR analysis of the complex, (ii) in organello imports and interaction, (iii) in vitro reconstitution with pure proteins, and (iv) in vivo complementation assays demonstrate that the CRACVDFKTWM segment is the crucial molecular mediator that guides the electron-transfer process from MIA40 to the FAD-binding domain of ALR.

Discussion

The long mitochondrial ALR isoform (48 kDa) is a homodimer connected by two intermolecular disulfide bonds, namely C95–C204′ and C95′–C204, with each subunit constituted by a five-helix folded FAD-binding domain at the C-terminus and an unstructured N-terminal domain of approximately 80 residues. The FAD-binding domain has four cysteine residues (namely C142, C145, C171, and C188) forming intramolecular disulfide bridges flanking the FAD molecule. C142–C145 disulfide is located close to the isoalloxazine ring of the FAD molecule and constitutes the redox active center of the sulfhydryl oxidase. Each N-terminal domain of the ALR homodimer interacts through its CRAC region with the CEEC–FAD redox active center of the other subunit, suggesting an intersubunit electron-transfer process as observed for yeast Erv1 (15). By directly comparing both affinity and electron transfer of the long and short isoforms of ALR with MIA40 we showed a more efficient binding of lf-ALR and that lf-ALR oxidizes the CPC motif of MIA40 more efficiently than sf-ALR. These data indicate that the N-terminal domain specifically recognizes MIA40 thus allowing a more efficient electron-transfer process between them. The CRAC motif is located indeed in a highly flexible protein region that can more easily be recognized by CPC motif of MIA40 with respect to the CEEC motif which, on the contrary, is quite buried. The structural flexibility of the N-terminal domain as well as its relatively weak binding affinity vs. both Mia40 and the FAD-binding domain also allow it to sample appropriate structural conformations able to drive a rapid electron shuttling transfer. Mutagenesis studies provided additional support for this mechanism and showed the specific residues of the N-terminal domain recognizing MIA40/Mia40 to efficiently mediate the electron transfer from MIA40/Mia40 to the FAD-binding domain of an ALR/Erv1 folded molecule. Solvent-exposed hydrophobic residues downstream of the shuttle CRAC motif of ALR specifically recognize indeed the hydrophobic cleft of MIA40. The fact that the mutation of the residues crucial to determine this interaction does not affect import of Erv1 suggests that there are other, as yet unknown, parts of Erv1 that function as targeting signals for the import into mitochondria.

The present data reveal a salient mechanistic feature of the MIA40-ALR disulfide relay system: the cleft of MIA40 seems to accommodate, using the same molecular principles and based on the same type of interactions, either the substrate targeting signal (23, 27) or the shuttle CRAC motif of ALR. Reoxidation of the CPC of MIA40 by the CRAC motif of ALR is therefore based on a “substrate mimicry” of the CRAC region (Fig. 5). An important consequence of this is that the binding of the imported substrate and of folded ALR is therefore mutually exclusive on the same active site of a MIA40 molecule. In addition, the CRAC region is the only segment of ALR with strong affinity with MIA40. Together these results provide independent evidence for a serial interaction between substrate and MIA40, followed by MIA40 and ALR, rather than a concomitant binding of the three partners (Fig. 5). The mitochondrial substoichiometric amount of Erv1 vs. Mia40 (SI Text and Fig. S10) as well as the 2∶1 in vitro stoichiometry of MIA40/ALR complex, both important parameters for the function of the two proteins in vivo, also support the above model. Although the presence of a “ternary” complex between Mia40, the substrate and Erv1 has been reported in yeast mitochondria (31), a later study by Bien et al. (15) disputed this and suggested it rather represents abortive intermediate species. Clearly, further in vivo studies, including possibly in vivo kinetics of both the Mia40-substrate and Mia40-Erv1 subreactions are needed to clarify this issue. Finally, the observation that certain growth defects for Mia40 hydrophobic mutants can be suppressed by Erv1 overexpression (32) could be explained in the light of the importance of hydrophobic interactions in the substrate-mimicry mechanism reported here.

Fig. 5.

The hydrophobicity-driven substrate mimicry mechanism. Shown on the right, a specific region (in green) of the N-terminal shuttle domain of lf-ALR guides the interaction with the Mia40 substrate binding cleft (in green), mimicking the interaction of the IMS-targeting signal of Cox17 [in green, (23, 27)] with Mia40 (shown on the left). Cys residues are in yellow. In the middle, the crucial reaction intermediates are depicted.

Our data has allowed the dissection of the region that mediates the electron transfer between MIA40 and ALR and uncovered the fundamental principles of hydrophobicity-driven and substrate mimicry for the molecular interactions that underpin this process. This rationale will function as a springboard to facilitate future studies on this important aspect of mitochondrial machineries and cell physiology.

Materials and Methods

NMR Spectroscopy.

NMR experiments for resonance assignment were carried out on 0.5–1 mM ¹³C,¹⁵N-labeled sf-ALR_3S-S and CRAC sf-ALR_4S-S samples in 50 mM phosphate buffer, pH 7.0, containing 10% (vol/vol) D₂O. All NMR spectra were collected at 298 and 308 K. The ¹H, ¹³C, and ¹⁵N resonance assignment of sf-ALR_3S-S is deposited at the BioMagResBank database. To determine the secondary structure elements, Chemical Shift Index (33) and PECAN (34) programs were used. The ϕ and ψ dihedral angle constraints were derived from the chemical shift analysis by using TALOS+ program (35). Relaxation experiments on ¹⁵N-labeled samples were performed at a 600 MHz Bruker spectrometer measuring ¹⁵N backbone longitudinal (R₁) and transverse (R₂) relaxation rates and the heteronuclear ¹⁵N{¹H} NOEs. ¹⁵N R₂ were also measured as a function of the refocusing time (τ_CPMG) in a Carr–Purcell–Meiboom–Gill (CPMG) sequence, which ranged between 450, 700, 900, and 1,150 μs (36).

To follow the disulfide oxidation of MIA40_2S-S upon addition of lf-ALR_4S-S or sf-ALR_3S-S, we titrated ¹⁵N-labeled MIA40_2S-S with unlabeled lf-ALR_4S-S or sf-ALR_3S-S at 298 K and followed the reaction by ¹H-¹⁵N HSQC spectra. The same procedure was performed between ¹⁵N-labeled sf-ALR and unlabeled MIA40_2S-S and between ¹⁵N-labeled MIA40 and unlabeled FWM/E lf-ALR_4S-S mutant. ¹H-¹⁵N TROSY-HSQC experiments at 312 and 308 K were acquired on C71S CRAC sf-ALR/C53S MIA40 and C74S CRAC sf-ALR/C53S MIA40 complexes, where one of the two partners was ¹⁵N-labeled, and on the corresponding free forms, in order to map the interacting residues.

Import of Wild-Type Erv1 and Erv1 Mutants LLFQ/A and Erv1 LLF/E.

Wild-type Erv1, Erv1 LLFQ/A, and Erv1 LLF/E mutants were imported in wild-type mitochondria (14) for the indicated time points at 30 °C. The reactions were either left untreated or blocked with 25 mM N-ethylmaleimmide (NEM). Mitochondria were resuspended in 1.2 M sorbitol/20 mM Hepes pH 7.4, treated by Proteinase K (PK) for 30 min on ice, followed by addition of PMSF (4 mM). Samples were analyzed by SDS-PAGE and digital autoradiography (Molecular Dynamics).

In Organello Complementation Assay Using a Two-Step Import in Gal–Erv1 Mitochondria after Preimport of Erv1.

Mitochondria depleted of endogenous Erv1 were prepared from a Gal–Erv1 S. cerevisiae strain (14) and replenished by importing purified versions of WT Erv1, Erv1 LLFQ/A, and Erv1 LLF/E. Subsequently, mitochondria were reisolated by centrifugation, resuspended in fresh import buffer, and used for import of ³⁵S-labeled MIA40 at 30 °C for the indicated time points. NEM was added (20 mM) to stop the reaction, followed by incubation with PK and PMSF. Samples were finally analyzed by SDS-PAGE and digital autoradiography (Molecular Dynamics).