Myristoyl group-aided protein import into the mitochondrial intermembrane space

The MICOS complex mediates formation of the crista junctions in mitochondria. Here we analyzed the mitochondrial import pathways for the six yeast MICOS subunits as a step toward understanding of the assembly mechanisms of the MICOS complex. Mic10, Mic12, Mic26, Mic27, and Mic60 used the presequence pathway to reach the intermembrane space (IMS). In contrast, Mic19 took the TIM40/MIA pathway, through its CHCH domain, to reach the IMS. Unlike canonical TIM40/MIA substrates, presence of the N-terminal unfolded DUF domain impaired the import efficiency of Mic19, yet N-terminal myristoylation of Mic19 circumvented this effect. The myristoyl group of Mic19 binds to Tom20 of the TOM complex as well as the outer membrane, which may lead to “entropy pushing” of the DUF domain followed by the CHCH domain of Mic19 into the import channel, thereby achieving efficient import.

its own, thereby bending the IM, and a subpopulation of Mic10 molecules also associate with the dimeric form of ATP synthase, thereby contributing to crista rim formation 21 . The IMS domain of Mic60 functions as a platform for interactions with OM proteins including the TOM and TOB/SAM complex proteins, thereby transiently forming contacts between the OM and IM. Mic19 was found to associate with cytochrome oxidase subunit IV (CoxIV), as well 22 . However, precise mechanisms of how each MICOS sub-complex is made from their constituent proteins and how the two sub-complexes assemble together to form CJ structures are largely unclear. ΔΨ (open circles) or without ΔΨ (closed circles) for the indicated times at 25 °C. After proteinase K (PK) treatment, mitochondria were subjected to SDS-PAGE and radioimaging. Imported, protease-protected proteins were quantified, and the amounts of the radiolabeled proteins added to each reaction were set to 100%. Values are mean ± SEM (n = 3). Full-length gel images are presented in Supplementary Fig. S4.
In the present study, we analyzed the import pathways for the six yeast MICOS subunits as a first step toward understanding of the assembly mechanisms of the MICOS complex. In contrast to the other five subunits, which used the presequence import pathway involving the TOM complex in the OM and TIM23 complex in the IM, import of Mic19 required the TOM complex and Tim40/Mia40 23,24 , which mediates oxidative folding in the IMS [25][26][27] . Besides, the Mic19 import depended partly on the previously uncharacterized pathway that requires the myristoyl group attached to the N-terminal domain 23,28 . We will discuss how N-myristoylation contributes to the efficient import of Mic19 into mitochondria.

Results and Discussion
Import of the MICOS subunits other than Mic19 is mediated by the TIM23 complex. The five yeast MICOS subunits, Mic10, Mic12, Mic26, Mic27, and Mic60, are integral IM proteins with one or two TM segments while Mic19 is a peripheral IM protein on the IMS side ( Fig. 1A) [11][12][13] . Except for presequence-containing Mic60, import pathways for the other 5 MICOS subunits are not obvious from their amino-acid sequences as they lack presequences (Fig. 1A). We thus performed in vitro import of those proteins into mitochondria in the presence or absence of the membrane potential across the IM (ΔΨ). Mic60 was efficiently imported into the protease-protected location of mitochondria with ΔΨ, but not at all in the absence of ΔΨ, as expected for the presequence-pathway import (Fig. 1B). Mic60 import required Tim50 of the TIM23 complex (Fig. S1), but not the mitochondrial Hsp70 import motor, Ssc1 (Fig. S2A). Since Mic60 contains a TM segment starting from 82-residue downstream to the presequence cleavage site, the TM segment is probably released laterally into the IM by the stop-transfer mechanism 29 . It is to be noted that presequence-containing proteins for lateral release from the TIM23 complex may not require Ssc1 if a TM segment is close to the presequence-cleavage site and the following C-terminal domain does not fold tightly 30 . Import of Mic10, Mic12, Mic26, and Mic27 was significantly hampered by dissipation of ΔΨ, yet there remained slight residual import into the PK-protected location even in the absence of ΔΨ. Similar ΔΨ-independent import of the TIM23 pathway substrates was reported for a small single-spanning IM protein like a subunit of the F 1 F o -ATPase, Su-e 31 . Import of Mic10, Mic12, Mic26, and Mic27 required Tim50 of the TIM23 complex (Fig. S1), but not the import motor Ssc1 (Fig. S2A). Since Mic12 spans the IM once with the N in -C out topology (Fig. 1A), it is probably inserted into the IM via the TIM23 complex by the stop-transfer mechanisms. On the other hand, Mic10, Mic26, and Mic27 span the IM twice and take N out -C out topology with a loop between the first and second TM segments exposed to the matrix (Fig. 1A). A possible mechanism to achieve the membrane topology of these proteins would be that their first TM segment reaches the matrix completely through the TIM23 complex and then inserts into the IM from the matrix side by Oxa1 32 , while the second TM segment being released laterally into the IM from the TIM23 complex by the stop-transfer mechanism 33 , although this possibility needs to be experimentally tested. On the other hand, import of Mic19 was not affected by the absence of ΔΨ at all or by functional defects of Tim50 (Fig. S1) or Ssc1 (Fig. S2A).
Mic19 import depends on Tim40/Mia40 and the presequence receptor Tom20, but not Tom70/71. Tim40/Mia40 and Erv1 mediate import and disulfide bond formation of soluble proteins with a twin CX 3 C or CX 9 C motif in the IMS without requiring ΔΨ (TIM40/MIA pathway) 8,[34][35][36] . Previous studies reported that human Mic19 (ChChd3) 37 and yeast Mic19 are imported into mitochondrial IMS via the TIM40/ MIA pathway. Mic19 proteins from various organisms consist of the N-terminal segment containing a myristoylation motif followed by a domain with an unknown function called a DUF (domain of unknown function) domain and the C-terminal CHCH (coiled-coil helix coiled-coil helix) domain, which often contains a twin CX 9 C motif ( Fig. 2A) 28 . Although fungal Mic19 proteins including yeast Mic19 do not have a typical twin CX 9 C motif in the CHCH domain, they instead have a CX 10 C motif, which was shown to be important for its import 24 . Indeed, the C146S/C157S mutation in the CX 10 C motif of yeast Mic19 impaired its import into the IMS (Fig. 2B). Under non-reducing conditions, Mic19C146S, not Mic19C157S or Mic19C146S/C157S, generated a 75kD band (Fig. 2C, uppermost panel), which was confirmed to be a mixed-disulfide intermediate with Tim40/ Mia40 (Fig. 2C, lowermost panel) through C298 of Tim40/Mia40 38 . A mix-disulfide intermediate was slightly observed for even wild-type Mic19, which mainly formed an oxidized form upon its import (Fig. 2C). Our results thus strongly support the model that the TIM40/MIA pathway is the primary import pathway for Mic19 and that Mic19 forms a mix-disulfide intermediate with Tim40 through the second cysteine residue, Cys157.
Substrates for the TIM40/MIA pathway usually do not depend on either the presequence receptors Tom20 and Tom22 39 or receptors Tom70/71 for presequence-less proteins. We thus analyzed if in vitro import of Mic19 depends on the mitochondrial surface receptors. In vitro import of Mic19, like fraction (Elute) were analyzed by SDS-PAGE and radioimaging. Band shifts by the attached FLAG tag and pull-down by the anti-FLAG antibody confirmed that the 75 kD proteins (indicated with arrowheads in the uppermost panel), which were derived from WT Mic19 or Mic19C146S, contained Tim40 and thus represented Mic19-Tim40 conjugates. (D) The indicated proteins were analyzed by SDS-PAGE followed by immunoblotting for the wild-type strain (WT) and those with Tom20-AID*-9 × Myc and Tom20-3 × miniAID instead of Tom20 after cultivation in lactate medium (+0.05% glucose) at 30 °C. (E) The indicated radiolabeled proteins were incubated with the indicated mitochondria for the indicated times at 25 °C. After treatment with or without PK (50 μg/ml) for 20 min on ice, mitochondria were subjected to SDS-PAGE and radioimaging. Bound proteins (-PK) and imported, protease-protected proteins (+PK) were quantified, and the amounts of the radiolabeled proteins added to each reaction were set to 100%. Values are mean ± SEM (n = 3). Full-length blot/gel images are presented in Supplementary Fig. S5.
Mic19G2A Mic19Δ20  presequence-containing pSu9-DHFR but unlike presequence-less ADP-ATP carrier (AAC), did not depend on the receptors for presequence-less proteins, Tom70 and Tom71 (Fig. S2B,C). We then tested involvement of the presequence-receptor Tom20 in the Mic19 import. For this purpose, we tried to delete the Tom20 receptor domain by an auxin-based degron (AID) system in vivo 40 . However, we found that introduction of the AID tag or 3 × miniAID tag to Tom20 destabilized Tom20 even in the absence of auxin, leaving the partner receptor Tom22 level less significantly affected (Fig. 2D). We thus isolated mitochondria with a reduced level of Tom20 from the strains expressing Tom20 with the AID tag or 3 × miniAID tag and performed in vitro import. Surprisingly, import of Mic19 was retarded in mitochondria lacking Tom20 while that of the canonical TIM40/MIA substrate Tim9 was not affected (Fig. 2E).
N-myristoylation enhances import of Mic19. Unusual dependence of Mic19 import on the presequence-receptor Tom20 suggests that Mic19 may utilize not only the pathway aided by Tim40/Mia40, but also another minor pathway aided by Tom20 for its import. In connection to this, a previous study on ChChd3, a mammalian homolog of Mic19, suggested that N-terminal myristoylation at Gly2 is important for its localization to mitochondria under the microscope in vivo 28 . Since yeast Mic19 also contains a myristoylation motif, MGX 3 S (X stands for any amino acid residue) at the N-terminus ( Fig. 2A) 41 (Fig. 3A). Yeast Mic19 is therefore N-myristoylated through its N-terminal myristoylation motif in reticulocyte lysate in vitro.
We then analyzed the effects of defective N-myristoylation and/or Cys mutations in the CX 10 C motif of Mic19 on its cellular localization (Fig. 3B). Although a defective N-myristoylation (G2A or Δ20) or mutation in the CX 10 C motif (C146S/C157S) alone did not deteriorate the membrane localization of Mic19, their combination (G2A/C146S/C157S) significantly impaired the localization of Mic19 to membranes in vivo (Fig. 3B). These membrane-localized Mic19 derivatives were resistant to externally added proteinase K, indicating their localization within mitochondria (Fig. 3C). Therefore N-myristoylation and the CX 10 C motif function redundantly in mitochondrial targeting of yeast Mic19 in vivo.
Next, we tested the role of N-myristoylation of Mic19 in mitochondrial import in vitro (Fig. 3D). Binding of Mic19 to mitochondria was markedly impaired by inhibition of N-myristoylation by the G2A or Δ20 mutation of Mic19 while import into mitochondria was mildly affected by these mutations. We thus tested the effects of DTT, which impairs the TIM40/MIA pathway import (Fig. 3E), on import of Mic19G2A and Mic19Δ20 in vitro. Although import of Mic19 was significantly impaired by the G2A or Δ20 mutation, the residual import of Mic19G2A and Mic19Δ20 was still sensitive to 22 mM DTT, like wild-type Mic19 (Fig. 3F), suggesting that N-myristoylation and the CX 10 C motif function independently in the import of Mic19.

Myristoylation circumvents the import-impairing effect by the DUF domain.
We next tested the effects of N-myristoylation on the import of canonical TIM40/MIA substrates. While a TIM40/MIA substrate Mdm35 [42][43][44] has the N-terminal myristoylation motif as well as a twin CX 9 C motif, Tim9 has only the twin CX 9 C motif, not the myristoylation motif. Although binding of Mdm35 to mitochondria was significantly reduced by the G2A mutation, its import was not affected by the G2A mutation (Fig. 4A). We then compared Tim9 and the fusion protein consisting of the N-terminal 20-residue segment of Mic19 (Mic19(1-20)), with or without the G2A mutation, followed by Tim9 for their binding to and import into isolated mitochondria (Fig. 4B). The Mic19(1-20) segment, irrespective of the presence of the myristoylation motif, did not enhance mitochondrial binding or import of Tim9, or rather partially impaired binding and import of Tim9.
Mic19, Mdm35, and Tim9 are all TIM40/MIA pathway substrates, but why does the N-terminal myristoylation enhance import of only Mic19, not Mdm35 or Tim9? We noted that Mic19 possesses a ~100-residue DUF domain between the N-terminal myristoylation domain and the CHCH domain. We attached the DUF domain i.e. residues 21-127 of Mic19 to the N-terminus of Tim9 (Mic19(21-127)-Tim9), and tested its import into mitochondria (Fig. 4C). Mic19(21-127)-Tim9 was hardly imported into mitochondria, suggesting that the N-terminal attachment of the DUF domain inhibited import of Tim9. On the other hand, when residues 1-21 containing the myristoylation motif were further attached N-terminally to Mic19(21-127)-Tim9, the resultant Mic19(1-127)-Tim9 was efficiently bound to and imported into mitochondria. However G2A mutation impaired the import ability of Mic19(1-127)-Tim9. These results suggest that the myristoylation motif is important for the import ability of the DUF-domain containing TIM40/MIA substrates. myristoylation domain mutants (open circles and squares) were incubated with mitochondria for the indicated times at 25 °C. Bound and imported proteins were analyzed as in Fig. 2E. (E) The indicated radiolabeled proteins were incubated with mitochondria in the presence of different concentrations of dithiothreitol (DTT) for the indicated times at 25 °C. As a control, 2 mM DTT was added to the reaction to prevent disulfide formation of the substrate proteins outside mitochondria. Imported proteins were analyzed as in Fig. 2E. (F) Radiolabeled Mic19 (Mic19WT) and its N-myristoylation motif mutants (Mic19G2A, Mic19Δ20) were incubated with mitochondria for the indicated times at 25 °C in the presence of different concentrations of DTT. Imported proteins were analyzed as in Fig. 2E. Full-length blot/gel images are presented in Supplementary Fig. S6 Mic19WT (  C-terminally attached FLAG tag, which was solubilized with Triton X-100 from mitochondria, with different concentrations of proteinase K (PK), the DUF domain appeared more sensitive to PK digestion than the C-terminal CHCH domain (Fig. S3). We thus attached various lengths of the unrelated unfolded segment of PhoA 46 at the N-terminus of the Mic19 CHCH domain and tested their import (Fig. 4D). The presence of an increasing length of the PhoA segment retarded import of the Mic19 CHCH domain and made the imported proteins unstable within mitochondria. However, attachment of the Mic19(1-20) segment with the myristoylation motif, but not the one with the G2A mutation, enhanced mitochondrial binding and import of the CHCH domain with the 100-residue PhoA segment (Fig. 4E). Therefore, the N-terminal myristoylation motif of Mic19 is important for the import of the CHCH domain via the TIM40/MIA pathway in the presence of an N-terminal long unfolded segment.
What is the target of the N-terminal myristoylated segment of Mic19 in its import? Since Mic19 was, unlike conventional TIM40/MIA pathway substrates, recognized by the presequence receptor Tom20, Tom20 could be a candidate for the factor that interacts with the myristoyl group of Mic19. We thus tested binding of myristoylated Mic19 with the isolated receptor domain (residues 51-145) of rat Tom20 (Tom20 sol ) 47 (Fig. 4F). After incubation of Mic19 and Tom20 sol with the C-terminal His 6 tag, affinity pull-down for the His 6 tag revealed specific binding of Mic19 to Tom20 sol . The G2A mutation in the myristoylation motif of Mic19 inhibited its binding to Tom20 sol , suggesting that the myristoyl group of Mic19 is directly recognized by Tom20. The myristoyl group of Mic19 was previously shown to interact with Tob55/Sam50 inside mitochondria 23,48 , but besides, our results here suggest that the myristoyl group of Mic19 is recognized by the cytosolic receptor domain of Tom20. Since Tom20 recognizes the hydrophobic side of the amphiphilic mitochondrial targeting signal like presequences through its hydrophobic groove 47 , the hydrophobic myristoyl group of Mic19 may well bind to the hydrophobic groove of Tom20.
Entropic pushing model for the import of Mic19. Why does the attachment of a DUF domain N-terminally to the CHCH domain decrease the import efficiency of Mic19, and how does the N-terminal myristoylation counteract this effect? Generally, entry of a reduced TIM40/MIA substrate domain, which is not tightly folded in the cytosol, into a narrow Tom40 import channel is a thermodynamically unfavorable process because it will decrease the conformational entropy of the substrate due to the increased excluded-volume constraint between the substrate polypeptide and the import channel (Fig. 5, step 1). This entropy decrease can be circumvented by weak binding of the substrate polypeptide to the inner wall of the import channel and subsequently by Tim40/Mia40 binding in the IMS at the outlet of the channel (Fig. 5, step 2). Further translocation of the substrate polypeptide could be partly driven by the increase in the conformational entropy in the IMS by the mechanism called entropic pulling 49 . Presence of a long disordered segment (the DUF domain for Mic19) in front of the TIM40/MIA substrate domain (the CHCH domain for Mic19) would further increase the conformational entropy of Mic19 in the cytosol, which should make the entry of the CHCH domain into the Tom40 import channel much less favorable (Fig. 5, step 3). However, binding of the N-terminal segment to Tom20 as well as the OM through a myristoyl group will make the DUF domain closely apposed to the membrane, thereby increasing the excluded-volume constraint between the DUF domain and the membrane (Fig. 5, step 4). This will counteract the increase in the conformational entropy arising from the attached DUF domain of Mic19 in the cytosol, thereby driving the entry of the CHCH domain of Mic19 into the Tom40 channel (Fig. 5, step 5). After this "entropic pushing", the CHCH domain will be trapped by Tim40 in the IMS, which should function as the trans side trap, like canonical substrates for the TIM40/MIA pathway substrates (Fig. 5, step 6). Dissociation of the myristoyl group from Tom20 in the cytosol followed by subsequent binding to the TOB/SAM complex in the IMS 23 may also contribute to driving the further translocation of Mic19 across the OM. Efficient sequestration of the unfolded DUF domain from the cytosolic side of the outer membrane may be also important for preventing it from activation of the mitochondrial stress response due to accumulation of unfolded proteins on the mitochondrial surface 50 . Indeed, replacement of the motif for myristoylation in Mic19 with the one for more hydrophobic palmitoylation increased binding to mitochondria, but decreased import efficiency (data not shown), suggesting that optimized hydrophobicity of the N-terminally attached acyl chain is important for efficient binding and dissociation of Mic19 from the mitochondrial surface. The scenario shown in Fig. 5 can be further tested experimentally, and the enigmatic function of the DUF domain should be addressed in future studies. The DUF domain could facilitate optimized distribution of Mic19 to distinct regions of the IM, in addition to crista junction, for interactions with outer membrane components like Tob55/Sam50 or with inner membrane components like CoxIV after reaching the intermembrane space 22 .

Conclusion
As an essential step toward understanding of the mechanisms of assembly of the MICOS complex and the following crista junction formation, we analyzed here the import pathways of the six yeast MICOS subunits into mitochondria. Among those MICOS subunits, only Mic60 possesses a cleavable presequence followed by a TM segment. Mic60 was found to follow the presequence pathway via the TOM40 complex and TIM23 complex, which exclusively requires ΔΨ, to be inserted laterally into the IM by the stop-transfer mechanism. Mic10, Mic12, Mic26, and Mic27 also follow the TIM23 pathway for import, but depending only mildly on ΔΨ. In contrast to these 5 MICOS subunits, Mic19 consists of an N-terminal myristoylation domain followed by a DUF domain and a CHCH domain, and its import did not depend on the TIM23 complex or ΔΨ, but instead used the TIM40/MIA pathway for the import of the CHCH domain into the IMS. The N-terminal attachment of the DUF domain or an unrelated long unfolded segment to a TIM40/MIA pathway substrate reduced its import competence, but that the presence of the myristoylation domain in front of the unfolded segment circumvented the import-impairing effect (Fig. 4). Although the presence of a long unfolded segment (the DUF domain for Mic19) in front of the TIM40/MIA substrate domain (the CHCH domain for Mic19) would increase the conformational entropy of Mic19 in the cytosol, which is unfavorable for the entry of the CHCH domain into the Tom40 import channel, this entropy decrease will be counteracted by binding of the myristoyl group to Tom20 and the OM, which would increase the excluded-volume constraint between the DUF domain and the TOM complex and the OM (Fig. 5). We thus propose this "entropy pushing" as a previously overlooked, new mechanism of importing proteins with a long disordered segment into mitochondria. Since proteins can be often myristoylated at their N-terminus, this mechanism could operate for other proteins destined for other organelles as well as mitochondria.

Materials and Methods
Yeast strains. Yeast strains used in this study are listed in (Table S1A). The MIC19 gene was deleted with kanMX4, which was amplified by PCR with pBluescript-kanMX4 as a template. A yeast strain expressing OsTIR1 was made by integration of pNHK53 linearized with StuI into the URA3 locus. PCR cassettes for the AID*-9xMYC tag or 3xmini-AID tag were amplified from pKan-AID*-9xMYC or pMK151, respectively, and were integrated into the 3′-end of the TOM20 locus of the OsTIR1 strain. DNA construction for attaching the 3 × FLAG-tag to the C-terminus of Tim40 was performed by homologous recombination of the W303-1A strain using appropriate gene cassettes from pFA6a-FLAG-kanMX6.
Plasmids. Plasmids and PCR primers used in this study are listed in Tables S1B,C, respectively. Plasmids for in vitro translation were constructed by standard molecular cloning techniques with materials listed in Table S1. Site-directed mutagenesis and partial deletion mutagenesis were performed by QuickChange.
In vitro protein import into isolated mitochondria. Radiolabeled proteins were synthesized in a cell-free transcription/translation system with rabbit reticulocyte lysate in the presence of [ 35 S]-methionine. Import of radiolabeled precursor proteins into isolated yeast mitochondria was performed at 25 or 30 °C in import buffer (250 mM sucrose, 10 mM MOPS-KOH, pH 7.4, 80 mM KCl, 2 mM ATP, 20 mM NADH, 12 mM creatine phosphate, 120 μg/ml creatine kinase, 2 mM methionine, 5 mM MgCl 2 , 2 mM DTT, 2.5 mM KPi, and 1% BSA). Import reactions were stopped by addition of 500 μl ice-cold SEM buffer (250 mM sucrose, 10 mM