Effects of targeting signal mutations in a mitochondrial presequence on the spatial distribution of the conformational ensemble in the binding site of Tom20

Abstract

The 20‐kDa TOM (translocase of outer mitochondrial membrane) subunit, Tom20, is the first receptor of the protein import pathway into mitochondria. Tom20 recognizes the mitochondrial targeting signal embedded in the presequences attached to mature mitochondrial proteins, as an N‐terminal extension. Consequently, ~1,000 different mitochondrial proteins are sorted into the mitochondrial matrix, and distinguished from non‐mitochondrial proteins. We previously reported the MPRIDE (multiple partial recognitions in dynamic equilibrium) mechanism to explain the structural basis of the promiscuous recognition of presequences by Tom20. A subset of the targeting signal features is recognized in each pose of the presequence in the binding state, and all of the features are collectively recognized in the dynamic equilibrium between the poses. Here, we changed the volumes of the hydrophobic side chains in the targeting signal, while maintaining the binding affinity. We tethered the mutated presequences to the binding site of Tom20 and placed them in the crystal contact‐free space (CCFS) created in the crystal lattice. The spatial distributions of the mutated presequences were visualized as smeared electron densities in the low‐pass filtered difference maps obtained by X‐ray crystallography. The mutated presequence ensembles shifted their positions in the binding state to accommodate the larger side chains, thus providing positive evidence supporting the use of the MPRIDE mechanism in the promiscuous recognition by Tom20.

Abbreviations

ALDH

aldehyde dehydrogenase

CCFS

crystal contact‐free space

MBP

maltose binding protein

MPRIDE

multiple partial recognition in dynamic equilibrium

pALDH

the presequence of ALDH

TOM

translocase of the outer mitochondrial membrane

Tom20

20‐kDa subunit of the TOM complex

1. INTRODUCTION

Most mitochondrial matrix proteins are synthesized as precursor proteins in the cytosol and imported into mitochondria. ¹ , ² , ³ The precursor proteins bear a cleavable N‐terminal amino acid sequence that contains the targeting signal directing them to mitochondria. ³ The N‐terminal extension is referred to as a “presequence.” The presequences do not share distinct sequence homology, but instead display common physicochemical properties. The presequences are typically 15–40 residues long and have an amphiphilic pattern of hydrophobic and positively charged residues. ⁴ Protein import across the two mitochondrial membranes is mediated by the coordinated actions of the protein assemblies, TOM and TIM (translocases of outer and inner mitochondrial membranes, respectively). ⁵ Among the many subunits, Tom20 (a 20‐kDa subunit of TOM) is the general receptor in an early step of protein translocation into mitochondria (Figure 1a). ⁶ Animal and fungal Tom20 proteins are anchored to the mitochondrial outer membrane by an N‐terminal hydrophobic transmembrane (TM) segment, and the C‐terminal soluble domain of Tom20 is exposed to the cytosol and recognizes the mitochondrial targeting signal embedded in the presequences. ⁷ , ⁸

FIGURE 1.

Structure and function of Tom20 and its hypothetical molecular mechanism to enable the promiscuous recognition. (a) The Tom20 protein resides in the mitochondrial outer membrane and recognizes the targeting signal embedded in the presequences, which are the extra N‐terminal segments of mitochondrial preproteins. The presequences are removed by limited proteolysis after transportation into the matrix. (b) Schematic diagram of the three poses of the presequence helix (depicted as a cylinder) in a dynamic exchange. The three poses were obtained as snapshots in previous crystallographic studies. ¹⁶ , ¹⁷ Three poses were assumed, but more poses could be involved in the dynamic equilibrium. Tom20 is equipped with two hydrophobic pockets: the ϕ₁/ϕ₅ site and the ϕ₄ site

We previously determined the solution structure of the cytosolic domain of rat Tom20, in complex with a presequence peptide derived from rat aldehyde dehydrogenase (ALDH). ⁹ The bound presequence peptide (pALDH) adopted an amphiphilic helical conformation, with the three hydrophobic leucine residues in the Tom20‐binding consensus sequence interacting with the hydrophobic surface of Tom20. We defined the consensus pattern of the mitochondrial targeting signal as ϕXXϕϕ, where ϕ denotes a hydrophobic residue and X is any amino acid residue. ¹⁰ This five‐residue pattern can represent diverse amino acid sequences. The targeting signals are embedded at various positions within the presequences, and thus their positions must be identified experimentally. ¹¹ The interaction between Tom20 and a presequence is a classic example of the coupled folding and binding of an intrinsically disordered polypeptide (IDP) to its target protein. The presequence has no stable structures in the free state but adopts a helical conformation when bound to Tom20. This binding mode is called templated folding, because the interaction with Tom20 triggers the conformational changes. ¹²

In contrast to the sequence diversity of the presequences, a single Tom20 species is expressed in animal and fungal cells. ¹³ Thus, Tom20 must distinguish about ~1,000 different mitochondrial proteins from other non‐mitochondrial proteins and sort them into mitochondria. The recognition of diverse presequences by Tom20 is a good example of promiscuous recognition. We previously advocated a hypothetical mechanism, “multiple partial recognitions in dynamic equilibrium,” or MPRIDE for short. ¹⁴ In MPRIDE, we first assume that the α‐helical presequence segment behaves as a rigid unit in the bound states (Figure 1b). The presequence interacts with Tom20 in many bound states with different poses relative to Tom20. In each pose, a subset of the features in the targeting signal is recognized by Tom20. Mismatches exist in the numbers of hydrophobic side chains (cyan, green, and orange ovals) in the targeting signal consensus and binding sites (blue and red ovals) on Tom20. To recognize the three hydrophobic residues with just two binding sites, the bound states/poses must rapidly exchange, presumably without the dissociation of the presequence. To enable the MPRIDE mode recognition, the binding site of Tom20 is exclusively composed of hydrophobic aliphatic side chains. The exclusion of aromatic side chains makes the binding surface relatively flat and smooth. The large residual mobility of the presequence in the binding groove confers an entropic advantage to increase the binding affinity, and in parallel, nearly equal affinities for the diverse mitochondrial presequences.

The structures determined by protein crystallography should be regarded as snapshots. The large movement of a mobile part of a protein or a moving ligand in a bound state is arrested by the molecular contacts in the protein crystal lattice. If the mobile part or the ligand in the binding site is located at a site lacking direct crystal contacts, then the movements could be preserved in the crystal lattice. We developed a design method to create crystal contact‐free space (CCFS) in protein crystals, using an MBP (maltose‐binding protein) fusion protein with a rigid connection (Figure 2a). ¹⁵ To overcome the weak affinity (K _d = 250 ± 50 μM ¹⁶ ), a presequence peptide was tethered to Tom20 via an engineered disulfide bond to ensure the full occupancy in the Tom20 binding site. The spacer length was optimized in the previous peptide library study. ¹⁰ Instead of conventional model building of the presequence, we used the smeared electron densities in CCFS to show the spatial distribution of the presequence peptide in the binding groove of Tom20. ¹⁵ In practice, the truncation (to be precise, zero padding) of high‐resolution diffraction data before map generation is effective as a low‐pass filter to increase the signal‐to‐noise ratio of the Fo–Fc difference map (Figure 2b). The improvement depends on the choice of the resolution limit (r _min) of the truncation. We used r _min = 7 Å for the α‐helical presequence peptide. The visualized electron density corresponds to the overlapping volume of the moving presequence peptide in the binding groove of Tom20. Note that the diffraction resolution of the crystals must be sufficient (better than 2 Å) to locate the MBP‐Tom20 protein accurately in the crystal lattice for a good Fo–Fc difference map.

FIGURE 2.

The concept of the crystal contact‐free space (CCFS) method. (a) The fusion with MBP (maltose‐binding protein) with a connector α‐helix creates CCFS in crystals. The rigid connection ensures the creation of sufficient room, independent of the packing mode of the protein molecules in the crystal. The covalent tethering guarantees the full occupancy of the presequence (magenta oval) in CCFS. (b) The truncation of high‐angle diffraction spots is effective as a low‐pass filter to improve the signal‐to‐noise ratio of the Fo–Fc difference electron density map, for visualization of a highly mobile presequence in CCFS

The essence of the MPRIDE mechanism is the preservation of the large mobility of a ligand in the bound state. ¹⁴ In accordance, our previous NMR ¹⁵N relaxation analyses suggested motion on a sub‐millisecond timescale at the Tom20–presequence interface. ¹⁷ Importantly, a subset of the ligand features is recognized in each pose of the ligand, but all of the features are collectively recognized in the dynamic equilibrium between several discrete poses. Consequently, the effects of functional mutations in the mitochondrial targeting signal are expected to be adsorbed by the changes in the spatial distribution of the presequence. Here, we mutated the three Leu residues to Phe, Trp, and Ala residues in the Tom20‐binding consensus of pALDH, and applied the CCFS method. We successfully visualized the spatial distributions of the four mutated presequences in the binding groove of Tom20, and found that the positional shifts of the electron density were consistent with the MPRIDE mechanism.

2. RESULTS

2.1. Spatial distribution of the wild‐type pALDH presequence tethered to Tom20

We prepared the MBP‐Tom20 fusion protein according to the same construction design adopted in the previous CCFS study. ¹⁵ A four‐residue spacer sequence, KEAL, was inserted between MBP and Tom20, and the five‐residue segment, AATGD, in the MBP sequence was deleted to enlarge the CCFS. This construct was previously referred to as Δ5MBP<+4>Tom20, but here, simply as MBP‐Tom20. The amino acid sequence of the presequence peptide (pALDH) is G¹²PRLSRLLSYAGC, ²⁴ where the Tom20‐binding consensus is underlined, and the spacer sequence is italicized. Note that the numbering starts from 12, because this peptide sequence corresponds to the C‐terminal half of the rat ALDH (aldehyde dehydrogenase) presequence. ¹⁸ The C‐terminal Cys residue of pALDH was cross‐linked to a single Cys residue in the cytosolic domain of Tom20 via a disulfide bond. The position of the disulfide bond and the number of spacer residues were optimized beforehand by a peptide library experiment, ¹⁰ with consideration given to the preservation of the presequence motions. We refer to the final construct as MBP‐Tom20‐SS‐pALDH(*), where * denotes the wild‐type (wt) or mutations (e.g., L15F).

We crystallized MBP‐Tom20‐SS‐pALDH(wt). The crystals grew with a reservoir solution (26% PEG 3350, 0.1 M MES pH 6.5, and 0.2 M ammonium iodide) at 293 K and were cryoprotected by the addition of glycerol at 18% (v/v). In contrast, the crystals in the previous study were grown with a reservoir solution (20% PEG 3350 and 0.2 M potassium nitrate) at 293 K and cryoprotected by the addition of glycerol at 15%. ¹⁵ We determined the position of MBP‐Tom20 in the crystal lattice by molecular replacement, using the MBP‐Tom20 structure obtained from the previous study (PDB 5AZ9). ¹⁵ Unexpectedly, the long connector helix is straight in the new structure but was bent in the previous structure (PDB 5AZ9), although the overall structures of the MBP‐Tom20 are quite similar (Figure S1). For reasons not fully understood, we suspected the repeated microseedings as the cause of the structural difference. The electron density was visualized in the low‐pass filtered Fo–Fc map (r _min = 7 Å) in the new crystal, as in the original crystal (Figure 3a). The continuous, elongated electron density consists of two parts. One is the overlapping volume of the moving presequence helix (orange mesh) and the other is the spacer segment (blue mesh). As a reference, the snapshot structure of the presequence helix and spacer in the crystal of Tom20‐SS‐pALDH(wt) is shown (Figure 3b). Without fusion to MBP, the crystal contacts with the neighboring molecules effectively fixed the pALDH in a particular pose in the crystal lattice. ¹⁷

FIGURE 3.

Stereo views of the electron density of the wild‐type pALDH presequence in CCFS. (a) Electron density contoured at 3σ corresponding to the presequence (orange mesh) and spacer (blue mesh) in the CCFS of the MBP‐Tom20‐SS‐pALDH(wt) crystal. (b) Crystal structure of Tom20‐SS‐pALDH(wt) (PDB entry 2V1T) ¹⁷ as a reference. The presequence trapped in a pose in the binding groove of Tom20 is depicted in the cartoon model. In (a) and (b), the cysteine residue of Tom20 used as the tethering point is red‐colored

2.2. Crystallization of the mutated pALDH presequences tethered to Tom20

Under the assumption of the MPRIDE mechanism, the replacement of the hydrophobic consensus residues with larger or smaller hydrophobic side chains should result in positional shifts of the electron density of the presequence in CCFS. For each mutation, one of the three hydrophobic Leu residues of pALDH was substituted with a Phe or Trp residue. These substitutions have comparable affinities to Tom20. ¹⁰ As a control, the substitution with an Ala residue was performed. The Ala substitutions with a hydrophobic, but smaller side chain reduce the affinity for Tom20.

In total, nine pALDH peptides were synthesized to test three mutations (L → F/W/A) at the three ϕ positions (L15, L18, and L19) (Table 1). We then tethered these peptides to the MBP‐Tom20 protein and performed crystallization screening. We obtained crystals, except for pALDH(L15A), but cross‐seeding using microcrystals of the MBP‐Tom20‐SS‐pALDH(wt) was necessary. The crystals diffracted to resolutions of 1.5–2.0 Å, except for pALDH(L18F) (Tables 1 and S1). The processed diffraction data were separated into two space groups, C2 (L15F, L18W), which is the same as the MBP‐Tom20‐SS‐pALDH(wt) crystal, and P2₁ (L15W, L18A, L19F, L19W, L19A). The asymmetric unit of the C2 crystals contained one MBP‐Tom molecule, and that of the P2₁ crystals contained two MBP‐Tom molecules. The data quality was assessed with phenix.xtriage, and twinning was suggested for the P2₁ crystals. The L15W and L18A crystals were pseudomerohedral‐like but not twinned, whereas the L19F, L19W, and L19A crystals were twinned with twin fractions of 0.20–0.44 and a twin law of h, −k, −h‐l. Initial phases were obtained by the molecular replacement method, and the structural models of MBP‐Tom20 were refined. As in the new MBP‐Tom20‐SS‐pALDH(wt) crystal, the long connector helix between MBP and Tom20 was straight in all of the crystals containing the pALDH(mut) (Figure S1). This rationalizes the use of the MBP‐Tom20 structure in the new MBP‐Tom20‐SS‐pALDH(wt) crystal and its Fo–Fc difference map as a reference.

TABLE 1.

Summary of the crystallographic characteristics of the MBP‐Tom20‐SS‐pALDH crystals

pALDH a	Crystal resolution (Å)	Space group	Molecule in ASU	Twin fraction	Twin law	Low‐pass filtered Fo–Fc omit map
WT	1.9	C2	1	No		Chain A, Figure 3a
L15F	2.0	C2	1	No		Chain A, Figure 4a
L15W	1.8	P21	2	No		Chain B, Figure 4b
L15A	No crystal
L18F	Bad crystals
L18W	1.8	C2	1	No		Chain A, Figure 4c
L18A	1.5	P21	2	No		NCS averaged, Figure 4d
L19F	1.6	P21	2	0.35	h, −k, −h‐l	Low quality
L19W	1.9	P21	2	0.44	h, −k, −h‐l	Low quality
L19A	1.8	P21	2	0.20	h, −k, −h‐l	Low quality

Note: See Table S1 for data collection and refinement statistics.

^a The relative size of the side chain can be calculated from apparent mole volumes, Φ_R (cm³/mol) ²⁸ ; Ala: Leu: Phe: Trp = 0.3:1: 1.2:1.6.

2.3. Spatial distributions of the mutated pALDH presequences tethered to Tom20

We attempted to detwin the problematic twinned data, using phenix.morph_model. The quality criterion is whether the electron density corresponding to the spacer segment is visible in the CCFS. This is because the spacer segment is relatively fixed by tethering, independently of the motions of the presequences. Following this criterion, we abandoned the twinned P2₁ crystals (L19F, L19W, and L19A) and selected the two C2 (L15F, L18W) and two non‐twinned P2₁ (L15W, L18A) crystals for further analyses. In the cases of the substitutions with Phe and Trp at position 15, the electron densities corresponding to the presequence part (magenta mesh) moved away from Tom20, and the shifts became larger as the size of the side chain increased from Phe to Trp (Figure 4a,b). As expected, the electron densities corresponding to the spacer regions (yellow–green mesh) remained at the same position. In contrast, the electron density in the CCFS was unchanged for the substitution with Trp at position 18 (Figure 4c). The hydrophobic side chain at position 18 is recognized by the hydrophobic pocket at the ϕ₄ site (Figure 1b). Because this hydrophobic pocket is shallow, the large tryptophan side chain is accepted without shifting the position of the presequence. Finally, the electron density corresponding to the presequence part disappeared for the substitution with Ala at the same position (Figure 4d), indicating that the small Ala side chain cannot maintain the complex state. Even though the mutated peptide was still located in the CCFS, the large motion was beyond the detection limit of the low‐pass filtered difference map.

FIGURE 4.

Stereo views of the electron densities of the mutated pALDH presequences in CCFS. (a)–(d) Electron densities contoured at 3σ corresponding to the presequences (magenta mesh) and spacer segment (yellow‐green mesh) in the CCFS of the MBP‐Tom20‐SS‐pALDH(mut) crystals. The electron density corresponding to the presequence and spacer in the CCFS of the MBP‐Tom20‐SS‐pALDH(wt) crystal is superimposed as a reference (gray mesh). (e) All electron densities and Tom20 structures were superimposed. The root‐mean‐square deviations (RMSDs) after superimposing the Cα atoms of the Tom20 part (residues 369–400) are as small as 0.13–0.22 Å

3. DISCUSSION

Targeting signals that specify the destinations for protein transport in living cells are typically short, linear amino acid motifs contained in cleavable N‐terminal extensions or embedded in mature protein sequences. ¹⁴ In general, a target signal is defined by a consensus sequence containing wild characters, such as ϕ and X, and hence represents diverse amino acid sequences. By contrast, the number of receptors/binding proteins is usually one or a few at most. ³ , ¹³ , ¹⁹ Thus, protein transport is mediated by many‐to‐one recognitions by the receptor/binding proteins. It is conceivable that special molecular mechanisms (beyond the classical lock‐and‐key and induced‐fit/conformational selections) could operate to realize the promiscuous recognitions of the targeting signals. Based on our crystallographic and NMR studies on the mitochondrial Tom20 protein–presequence interactions ⁹ , ¹¹ , ¹⁵ , ¹⁶ , ¹⁷ , ²⁰ we advocated the MPRIDE mechanism ¹⁴ : a receptor protein recognizes a partial feature of the ligand in a pose, and the dynamic equilibrium between many poses enables the receptor protein to recognize the full features of the ligand (Figure 1b). This dynamic, multiple recognition mode enables the Tom20 receptor to recognize diverse mitochondrial presequences (Figure 1a). The MPRIDE mechanism predicts that a mutation in the Tom20‐binding consensus will affect the spatial distribution of a presequence in the bound state. The replacement by a side chain with a different size in the Tom20‐binding consensus, while maintaining the requirement for the mitochondrial targeting signal consensus, should shift the position of the presequence ensemble responding to the size and position of the side chain. We prepared nine MBP‐Tom20‐SS‐pALDH (mut) (Table 1), and obtained interpretable Fo–Fc difference maps for four mutations to visualize the positional shifts of the presequence ensemble in the binding groove of Tom20 (Figures 3 and 4).

Upon mutation, the space group of some crystals changed from C2 to P2₁ (Tables 1 and S1). The same change in the space group occurred in the previous replacement of Leu by Ser in the Tom20‐binding consensus. ¹⁵ In the two crystal lattices, two MBP‐Tom20 molecules have tight contacts (green and blue protomers in Figure 5) and form similar dimeric structures, whereas the packings of the dimeric structures are different between the two space‐group crystals. The difference in the packing mode does not affect the formation of the CCFS, but the P2₁ crystals were prone to crystal twinning. The quality of the resultant electron density maps of the twinned crystals was poor and unsuitable for further analysis.

FIGURE 5.

Comparison of the packing modes of the MBP‐Tom20 molecules in the C2 and P2₁ crystals. (a) MBP‐Tom20‐SS‐pALDH(L15F) crystal as a representative of the C2 crystals. (b) MBP‐Tom20‐SS‐pALDH(L15W) crystal as a representative of the P2₁ crystals. The orange and purple symbols mark the positions of the crystallographic two‐fold rotation axis in the C2 crystal, and the non‐crystallographic pseudo two‐fold axis in the P2₁ crystal, respectively. The magenta squares depict the frame of the ASU units

We found the significant positional shifts of the electron densities of the presequence peptides bearing mutations at position 15 in the binding site of Tom20 (Figure 4a,b). In the previous peptide library study, the wild‐type amino acid sequence (LSRLL) of the ALDH presequence represented the most preferred amino acid residues by Tom20 (see figure 6 in Reference 10). This suggests that the residues 15–19 of the ALDH presequence are locally optimized in the sequence space with respect to Tom20 binding. The L15F (FSRLL) and L15W (WSRLL) had comparable affinities to the WT (LSRLL) sequence, but we have not tested whether they are functional as a mitochondrial targeting signal in cells. Instead, in an exhaustive search of a data set of experimentally confirmed presequences, ²¹ 14 statistically significant six‐residue consensus patterns were identified (see figure 3 in reference 21). In their report, the first and second major consensus patterns contained F at the positions corresponding to position 15. This indicates that the L15F (FSRLL) sequence does exist at the N‐termini of mitochondrial preproteins and may function as a mitochondrial targeting signal. In contrast, the L15W (WSRLL) sequence is very rare, if any, at the N‐termini of mitochondrial preproteins.

The replacements of a Leu residue at the ϕ₁ position of the Tom20‐binding consensus by larger hydrophobic side chains push the presequence helix away from the Tom20 molecule in poses 2 and 3, in which the side chain of the ϕ₁ position in the Tom20‐binding consensus is recognized by one of the two hydrophobic sites on Tom20 (Figure 6a). In contrast, the replacement has little effect in pose 1 because the side chain of the ϕ₁ position is unrecognized. The smeared electron density in the CCFS does not fully cover the entire distribution of the moving presequence peptide, but rather corresponds to the overlapping volume of the electron densities of the moving presequence peptide in the binding site of Tom20. Taken together, the electron density is expected to move upward and tilt up at the C‐terminal side (Figure 6b). This argument is oversimplified but could explain the changes in the electron densities in the CCFS. The magenta‐colored densities around the C‐termini of the α‐helical structures increase, and the displacements of the mass centers of the electron densities are roughly comparable to the increases in the size of the side chains by mutations from Leu to Phe/Trp (Figure 6c).

FIGURE 6.

Interpretation of the positional shifts of the electron densities in the CCFSs of the MBP‐Tom20‐SS‐pALDH(L15F) and MBP‐Tom20‐SS‐pALDH(L15W) crystals. (a) Molecular basis of the electron density changes observed in CCFS, induced by a mutation in the targeting signal. The sizes of the ovals represent the volumes of the side chains. The flat blue and concave red circles represent the hydrophobic binding sites on Tom20. The large side chain at the ϕ₁ position in the Tom20‐binding consensus pushes the presequence peptide up in poses 2 and 3. (b) The overlapping volume (transparent magenta) of the electron densities of the moving α‐helical presequence in CCFS moves upward and tilts up at the C‐terminal side. The horizontal dotted lines indicate the positional relations of the three poses. (c) Changes in the size of the mutated side chains (the blue arrows) and those in the positions of the mass centers of the electron densities in the CCFS (the yellow arrows)

Finally, the presequence recognition by Tom20 should be discussed in light of its role in the TOM complex. Recently, the cryo‐EM structure of the human TOM complex containing Tom22 and Tom20 subunits has been reported. ²² Unfortunately, in contrast to the structure and position of Tom22, those of Tom20 are much less accurate and seem unusable for detailed discussion.

In conclusion, the present study has provided positive experimental evidence for the use of the MPRIDE mechanism in the Tom20‐presequence system. In the MRIDE recognition, the dynamic movement exerts buffering effects against the structural differences in ligands by shifting the spatial distribution of the ligand in the bound state. We expect that the MPRIDE mechanism will be universally applicable to other promiscuous ligand recognitions by proteins.

4. MATERIALS AND METHODS

4.1. Proteins and peptides

The expression and purification of the MBP‐Tom20 fusion protein with the four‐residue spacer and five‐residue deletion inside the MBP part (i.e., ∆5MBP < +4 > Tom20) were performed as described. ¹⁵ The wild‐type pALDH peptide, GPRLSRLLSYAGC, and its amino acid substitution variants, GPRϕ₁SRϕ₂ϕ₃SYAGC, where ϕ₁ϕ₂ϕ₃ = FLL, WLL, ALL, LFL, LWL, LAL, LLF, LLW, and LLA, were synthesized with an N‐terminal acetyl group and a C‐terminal amide group by Hokkaido System Science (Sapporo, Japan) and Toray Research Center (Japan). An intermolecular disulfide bond was formed between the C‐terminal Cys residue of the presequence and a single cysteine residue (Cys100) in the cytosolic domain of Tom20 in high pH buffer (0.1 M Tris–HCl, pH 9.0). The cross‐linked proteins were buffer exchanged and concentrated with an Amicon Ultra‐4 unit (10 kDa NMWL) to 10 mg mL⁻¹ in 100 mM MES buffer, pH 6.5, for crystallization.

4.2. Crystallization

Hanging drops were prepared manually by mixing 1‐μl protein solution and 1‐μl reservoir solution. Each hanging drop was placed over a 0.4 ml reservoir solution. The final crystallization conditions were the same for all crystals: 26% PEG 3350, 100 mM MES, pH 6.5, and 0.2 M ammonium iodide. Cross‐seeding with the MBP‐Tom20‐SS‐pALDH(wt) microcrystals was necessary to obtain the MBP‐Tom20‐SS‐pALDH(mut) crystals. The seed stock was prepared as follows: A crystal was picked up and transferred to a 4‐μl drop of reservoir. A total of 8‐μl of reservoir was added to the 4‐μl drop, by addition in 2 μl increments. After 2 min, the crystal was broken apart with a pipette tip. A 2‐μl portion of the resulting seed solution was added to an 8‐μl reservoir and mixed well with gentle pipetting. Then, 10‐fold dilution was repeated three times. A 0.1‐μl portion of the diluted seed solution was added to a 4‐μl drop and mixed well with gentle pipetting while preparing the crystallization plates.

4.3. Data collection and processing

X‐ray diffraction data were collected with a wavelength of 0.9000 Å at beamline BL44XU of SPring‐8 (Harima, Japan). The crystals were cryoprotected by the addition of glycerol at 18% (v/v) and cryocooled in a nitrogen‐gas stream. The diffraction images were recorded with the 180° total rotation range, 1°per image, with the MAR300HE detector for the MBP‐Tom20‐SS‐pALDH(wt) crystal, and 0.1°per image, with the EIGER16M detector for the MBP‐Tom20‐SS‐pALDH(mut) crystals. The diffraction images from the EIGER in the HDF5 (Hierarchical Data Format) format were converted to the miniCBF format using eiger2cbf, obtained from https://github.com/biochem-fan/eiger2cbf. The data were indexed and integrated with iMosflm ²³ and then scaled and converted from I to F with Aimless in CCP4. ²⁴ The data collection statistics are summarized in Table S1. The data quality was assessed with phenix.xtriage. ²⁵ , ²⁶

4.4. Structure solution and refinement

Initial phases were estimated by the molecular replacement method, using the program phaser in CCP4. ²⁴ For the MBP‐Tom20‐SS‐pALDH(wt) crystal, we used the structures of MBP and Tom20 extracted from the previous structure (PDB 5AZ9) as the two separate search models, and for the MBP‐Tom20‐SS‐pALDH(mut) crystals, we used the MBP‐Tom20 structure obtained from the new MBP‐Tom20‐SS‐pALDH(wt) crystal as the search model. We did not build a model for the presequence peptide including the spacer and the C‐terminal Cys residues, to avoid model bias. The coordinates of MBP‐Tom20 were first adjusted with phenix.morph_model and then refined with phenix.refine. ²⁵ , ²⁶ Automatic water picking (addition, deletion, and refinement) was performed during the phenix.refine run with the option of ordered_solvent = True. R‐free flags were used to avoid overfitting. The water molecules in and around the binding groove of Tom20 were then removed with a PyMOL script, which improves the visual sensitivity of an omit map, in the same way as the “polder map” available in the phenix program suite. ²⁷ To erase the effects of the water removal, we ran phenix.refine again without water picking. During the second run, all reflections were used in the refinement with the option of ignore_r_free_flags = True to maximize the quality of the final electron density maps. Because the atom coordinates were correct after the first run with R‐free flags, the risk of overfitting is very low even without R‐free flags. For confirmation, another run with R‐free flags turned on was performed to check the convergence of the refinement calculations (See the legend of Table S1). The refinement statistics are summarized in Table S1. Refinement using pseudomerohedral twinned data was performed using phenix.refine with the option of twin_law = “h, ‐k, ‐h‐l”.

4.5. Low‐pass filtered Fo–Fc map generation

A sigma A‐weighted Fo–Fc (mFo–DFc) difference electron density map was calculated, using Fourier amplitudes (FOFCWT) and phases (PHFOFCWT) outputted by phenix.refine. The program SFTOOLS in CCP4 was used to perform the truncation and zero padding of the high‐resolution structure factor amplitudes, by using the keywords “SELECT RESOL <= 7” and “CALC COL FOFCWT = 0” where 7 is the truncation threshold r _min. Since the truncation of high‐resolution reflections results in a coarse mesh size of the electron density map, zero‐padding maintains the original mesh spacing size, simply for ease of visualization. The truncation threshold was empirically selected. FFT and MAPMASK in CCP4 were used for map generation.

Figures were generated with the program PyMOL (Schrödinger), version 2.4.2 (https://pymol.org/2/, last accessed in April 2022). The Cα atoms of the Tom20 part (residues 369–400) of two MBP‐Tom20 structures were superimposed using the align command. After superimposition, the map object belonging to the moving structure was moved into the same frame of the reference, using the matrix_copy command. The drawing of the electron density was limited to the volume surrounding the bound presequence by using the isomesh command, so the electron densities of other presequences at the symmetry‐related positions were not drawn for clarity. For this purpose, the coordinates of the presequence atoms in the two snapshot Tom20‐SS‐pALDH(wt) structures ¹⁷ (PDB 2V1S and 2V1T) were used to set the drawing limit of the isosurface of the electron density.

AUTHOR CONTRIBUTIONS

Xiling Han: Conceptualization (equal); data curation (equal); formal analysis (equal); investigation (equal); visualization (equal); writing – original draft (equal). Nobuo Maita: Data curation (equal); formal analysis (equal); methodology (equal); software (equal). Atsushi Shimada: Data curation (equal); formal analysis (equal); methodology (equal). Daisuke Kohda: Conceptualization (equal); data curation (equal); formal analysis (equal); funding acquisition (lead); project administration (lead); software (equal); supervision (lead); validation (lead); visualization (equal); writing – original draft (equal); writing – review and editing (lead).

CONFLICT OF INTEREST

The authors declare no competing interests.

Supporting information

TABLE S1. Protein design, X‐ray data collection, and refinement statistics.

FIGURE S1. Structural comparison of the MBP‐Tom20 fusion proteins in the different crystals.

Han X, Maita N, Shimada A, Kohda D. Effects of targeting signal mutations in a mitochondrial presequence on the spatial distribution of the conformational ensemble in the binding site of Tom20. Protein Science. 2022;31(10):e4433. 10.1002/pro.4433

DATA AVAILABILITY STATEMENT

The data (Table S1 and Figure S1) that support the findings of this study are available in the supplementary material of this article. The source data files (in a zipped folder containing 6 subfolders, 58.4 MB) are available from the corresponding author upon reasonable request. The zipped folder contains 5 subfolders named pALDH_wt, pALDH_L15F, pALDH_L15W, pALDH_L18W, and pALDH, L18A. Each subfolder contains PDB, mtz, and map (CCP4 format) files of the MBP‐Tom20‐SS‐pALDH(wt|L15F|L15W|L18W|L18A) crystals. The folder also contains a subfolder named run_MR_Refine, which contains a shell script, run_MR_Refine.com, to run phaser and phenix, and the PDB, mtz, and sequence files for input.