Structural bases for the Charcot–Marie–Tooth disease induced by single amino acid substitutions of myelin protein zero

Myelin protein zero (MPZ or P0) is a major transmembrane protein expressed in peripheral compact myelin and functions to glue membranes to form multiple layered membranes characteristic of myelin. Intermembrane adhesion is mediated by homophilic interactions between the extracellular domains (ECDs) of MPZ molecules. Single amino acid substitutions in an ECD cause demyelinating neuropathy, known as Charcot–Marie–Tooth disease (CMT); however, the mechanisms by which such substitutions induce the disease are not well understood. To address this issue, we constructed a novel assay to evaluate the membrane-stacking activity of ECD using ECD-immobilized nanodiscs. Using this novel “nanomyelin” system, we found that octameric (8-meric) ECDs with a stacked-rings-like configuration are responsible for membrane adhesion. Two inter-ECD interactions, cis and head-to-head, are essential to constituting the 8-mer and, consequently, to gluing the membranes. This result was further reinforced by the observation that the CMT-related N87H substitution at the cis interface abolished membrane-adhesion activity. In contrast, the CMT-related D32G and E68V variants of ECD retained membrane-stacking activity, whereas their thermal stability was reduced compared to that of the WT. Reduced thermal stability may lead to impairment of the long-term stability of ECD and the layered membranes of myelin.


Introduction
Myelin covering a neuronal axon functions as an electrical insulator that is essential for rapid signal conduction along the axon, known as saltatory conduction. In the peripheral nervous system (PNS), myelin is formed by Schwann cells that spiral around axons to form layered membranes. The membranes are regularly and densely packed in the region called compact myelin, where the spaces between extracellular membranes and between intracellular membranes are maintained to be approximately 50 and 30-40 Å, respectively (1). Myelin protein zero (MPZ or P0) is the most abundant protein expressed in the compact myelin of the PNS (2) and functions as an adhesion molecule to stack membranes (3). MPZ is essential for proper myelin formation, and genetic mutations leading to amino acid substitutions in MPZ cause inherited demyelinating neuropathies known as Charcot-Marie-Tooth disease (CMT) (4). MPZ is a type 1 membrane protein composed of an immunoglobulin-like extracellular domain (i.e., ECD), a single transmembrane helix, and an intracellular region. Reportedly, MPZ exists as a mixture of monomers, dimers, and tetramers; however, the biological significance of these multimers in myelin is poorly understood (5)(6)(7). Two crystal structures of ECD have been determined: an ECD structure from human MPZ (hMPZ-ECD) fused with maltose-binding protein (MBP) (8) and an ECD structure from rat MPZ (rMPZ-ECD) (6). HMPZ-ECD exists as a monomer in the crystal because of the presence of fused MBP, and the structure provides little information about the molecular interactions of ECD (8). In contrast, the rMPZ-ECD crystallized as a multimer, and three types of inter-ECD interactions were observed. The cis interaction formed an ECD tetramer by binding four ECDs with the same orientation. The trans and head-to-head interactions laterally and vertically connect the two ECDs in opposite directions, respectively (6). The trans and headto-head interactions potentially adhere two membranes via ECD, and the former interaction has been proposed to be responsible for the compact myelin formation since the resultant ECD multimer fits the approximately 50 Å gap between the membranes histologically observed in myelin (6). Curiously, CMT-related amino acid substitution sites have been found at these three interaction sites (4), suggesting that all these sites are related to proper myelin formation.
However, the significance of these substitutions in membrane adhesion has not yet been well examined, partially because cell biology or animal experiments are time-consuming and unsuitable for comparing the activities of many ECD variants (9,10). Therefore, an alternative experimental tool for analyzing the activity of ECD variants must be developed.
In this study, we established a new method for analyzing the membrane-stacking activity of human MPZ-ECD (hMPZ-ECD; hereafter referred to as ECD) by monitoring the oligomerization of ECD-immobilized nanodiscs (NDs) (11); Using this "nanomyelin" system in combination with ECD variants with single amino acid substitutions, we identified the residues responsible for the membrane adhesion activity of ECD. We also found that some CMT-related amino acid substitutions did not affect the membrane-stacking activity of ECD but reduced the thermal stability of the protein. Based on these results, we provide molecular insights into CMT initiated by amino acid substitutions in the ECD.

Oligomerization state of ECD in solution
We prepared a recombinant ECD with a C-terminal histidine-tag (ECD-His) by refolding and analyzed its multimeric states using size-exclusion chromatography (SEC) with a multi-angle light scattering (MALS) detector (Fig. 1). The protein eluted in two main peaks, and MALS revealed that it contained particles with apparent sizes of 126 (±2) kDa and 17 (±0.03) kDa, which correspond to octameric (8-meric) and monomeric ECD-His, respectively. We independently reapplied the second SEC analysis to the 8-meric and monomeric ECD-His eluates and confirmed that they produced two peaks, similar to the elution pattern of the first SEC (Fig. S1). These results indicated the existence of an 8-mer-monomer equilibrium for ECD-His. Notably, the retention volume of monomeric ECD-His (19-20 mL in Fig. 1) was larger than the expected volume of 16-17 mL based on the molecular weight (MW) of ECD-His, indicating that monomeric ECD-His interacted with the gel matrix (agarose and dextran) of the SEC column.

Crystal structure of the multimeric human MPZ-ECD
We determined the crystal structure of ECD (without a histidine-tag) with a 2.1 Å resolution to obtain the molecular basis for the multimerization of human MPZ-ECD. The structure matched well with the previously reported structure of rMPZ-ECD with an RMSD value of 0.387 Å (115 Cα atoms in the amino acid residues 1-119 excluding the missing 4 Cαs in rMPZ-ECD). The F-G loop (N102-G108), which was disordered in the rat structure, was visible in the human structure ( Fig.   S2) (6). The molecular contacts of human ECDs were similar to those of rat ECDs in the crystal, where each ECD contacted the other four ECDs through three types of inter-ECD interactions ( Fig. 2A and S3). A ring-shaped ECD tetramer was formed through cis interactions, and the ring was laterally and vertically attached to other tetramers in the opposite direction via trans and head-to-head interactions, respectively ( Fig. 2A). The residues involved in the inter-ECD interactions are shown in Fig. 2B-D. Aromatic residues play central roles in all the interfaces, as seen in the rMPZ-ECD structure: W24 makes a π-stacking interaction with H86 of the facing ECD at the cis interface (Fig. 2B), H52 forms a cation-π interaction with R45 at the trans interface (Fig.   2C), and W28 makes hydrophobic interactions with a cavity formed by D32 and K55 at the headto-head interface (Fig. 2D). Detailed structural differences between human and rat MPZ-ECDs caused by the three amino acid substitutions are described in the Supplemental Results.
Two possible interactions are responsible for the membrane-stacking activity of the ECD.
The trans interaction laterally connects two ECDs, whose C-terminus is linked to the transmembrane helix via a short linker in the native protein, and the trans-interacted ECDs separate the two membranes with a distance of more than 44 Å (inter-W28-Cζ2 distance along the membrane-plane normal vector) ( Fig. 2A). The head-to-head interaction is the other option to glue the membranes, and the inter-membrane distance becomes more than 67 Å in this case (inter-E119-Cα distance along the membrane-plane normal) ( Fig. 2A). The simultaneous formation of trans and head-to-head interactions is unlikely because it disturbs the planar structure of the membranes.
The oligomerized ECDs showed differently shaped B′-C (S25-I33) and F-G (N102-G108) loops from those of the monomeric ECD fused with MBP (8), although the overall ECD structures were similar with an RMSD of 1.550 Å ( Fig. S4A and B). In the multimeric ECD, the B′-C loop is involved in both the cis and head-to-head interactions, and the F-G loop is involved in the trans interaction. In monomeric ECD, some residues on these loops form crystal contacts with the surrounding MBPs, so the resultant structure does not necessarily reflect the native monomeric conformation. Nevertheless, structural differences in these loops between the two ECD structures suggest that the B'-C and F-G loops are flexible, and structural changes occur upon the formation of inter-ECD interactions.

8
In vitro assay to examine the membrane-stacking activity of ECD To identify the residues that are responsible for the membrane-stacking activity of ECD, we developed an assay system to analyze the membrane-stacking activity of ECD by monitoring the oligomerization of ECD-immobilized nanodiscs (NDs) (11). We prepared NDs containing Ni 2+chelating lipids on which ECD-His molecules were bound (ECD-His-NDs) (Fig. S5). Fig. 3A shows the SEC profile of ECD-His-NDs. Fig. S6 shows the proteins present in each elution peak. ECD-His-NDs were eluted at the void volume (8.2 mL), which corresponded to a MW >1120 K. Because the MW of monomeric ECD(W24A)-His-ND was 300 K (described below; Fig. 4D), it was assumed that ECD-His-NDs would form at least a tetramer in the eluate. The multimerization of ECD-His-NDs was partially inhibited by the coexistence of ECDs (without a histidine-tag) and completely impaired by the addition of imidazole ( Fig. 3B and S6), suggesting that multimerization is mediated by interactions between ECD-His attached to the lipids.  Fig. 3D shows an expanded view of an EM image of a typical rod-like structure where structural images of an ND (height/width = 50/97 Å (11)) and the 8-meric ECDs assembled by the cis and head-to-head interactions (height/width = 67/80 Å, Fig. 2A) were manually superimposed to guide the size of the particles. The sizes of the thicker and thinner parts roughly correspond to those of the ND and 8-meric ECDs, respectively, suggesting repetitive assembly of the NDs and 8-meric ECDs (Fig. S5). Because these stacked NDs connected by ECD 8-mers mimic the layered membranes glued by MPZ in myelin, hereafter, we describe them as nanomyelin. The turns and branches of nanomyelin may be caused by a tilted ECD-8-mer attached to the edge of an ND via one or two histidine-tags and by the attachment of multiple ECD-8-mers on the surface of a single ND, respectively.

The cis and head-to-head interactions were responsible for the membrane-stacking
To identify the inter-ECD interaction site(s) responsible for nanomyelin formation, we prepared four representative ECD-His variants with alanine substitutions at the residues involved in inter-ECD interactions in the crystal structure (W24A at the cis interface, R38A and R45A at the trans interface, and W28A at the head-to-head interface) ( Fig. 2B-D).
Using these ECD variants, we analyzed their multimerization by SEC and their membranestacking activity using nanomyelin experiments.  SEC experiments revealed that the W24A (a cis interface substitution) and W28A (a head-to-head interface substitution) variants existed solely as monomers ( Fig. 4A and C) and that these variants did not induce multimerization of NDs (Fig. 4D, 4F, 5A, and 5I). In contrast, the two variants with

Effects of the CMT-related amino acid substitutions on the membrane-stacking activity of ECD
We prepared ECD-His variants with CMT-related substitutions (E68V and N87H at the cis interface; D46V, A47V, and H52R at the trans interface; and D32G and D32N at the head-to-head interface) to determine the effects of these substitutions on nanomyelin formation ( Fig. 2B-D).
Among disease-related substitutions, H52R and N87H are linked to Dejerine-Sottas Syndrome (DSS; a severe type of demyelinating neuropathy), and D32G, D46V, A47V, and E68V are causative of CMT Type 2 (CMT2) (4). The D32N substitution produces an additional glycosylation site on ECD, leading to CMT Type 1B (CMT1B) (12). The recombinant protein used in this study was non-glycosylated. Therefore, the results obtained for the D32N variant solely reflected the effect of side-chain neutralization.
As shown in Fig. 4

Thermal stability of the ECD variants
In the nanomyelin experiments, the amino acid substitutions D32G, D32N, D46V, A47V, H52R, and E68V did not impair the membrane-stacking activity of ECD, although these substitutions induced CMT. To examine the effect of these substitutions on the thermal stability of ECD, we analyzed the temperature-dependent denaturation of the ECD-His variants using circular dichroism (CD) spectroscopy (Fig. 6). For all ECD-His variants, the protein samples precipitated in the heating processes, and the reversed cooling processes did not show apparent changes in [θ]217. These results indicate that the observed processes were irreversible. The data were fitted to a theoretical curve based on a two-state fold-unfold equilibrium (13) to obtain the midpoint temperatures (Tms) for denaturation. The Tm values are listed in Table 1. The E68V (cis interface; (head-to-head interface), and three variants with substitutions at the trans interface (D46V, A47V, and H52R) showed marginal (< 3 °C) differences in Tm from WT (Fig. 6).

Inter-ECD interactions responsible for the membrane-stacking
Based on the previously reported crystal structure of multimeric rMPZ-ECD (6), the trans interaction, which laterally combines ECDs in opposite directions, has been considered responsible for the formation of membrane layers in compact myelin (14). In this study, we confirmed that the trans interaction was formed in the crystal (Fig. 2), but this trans interaction was unlikely to be formed in solution because the variants with substitutions at the trans interaction site did not affect 8-mer formation or nanomyelin formation ( Fig. 4 and 5). Rather, ECDs in solution formed cis and head-to-head interactions to construct stacked-rings like 8-mer (the bluish and greenish ECDs in Fig. 2A, hereafter the stacked-rings-8-mer), and this 8-mer was responsible for membrane-stacking because some amino acid substitutions at these interfaces impaired the oligomerization and membrane-adhesion activities of ECD ( Fig. 3-5). An emerging issue here is that the 67 Å vertical length of the stacked-rings-8-mer exceeds the known extracellular inter-membrane distance of 47 Å in compact myelin (15), and the 8-mer does not seem to fit the inter-membrane gap. A possible explanation for this is that the stacked-rings-8-mer plays a role in the early stages of myelination. It has been reported that myelin-associated glycoproteins (MAGs) primarily attach the mesaxon membranes in the myelinating Schwann cells with the inter-membrane distance of approximately 120 Å, and these loosely layered membranes are eventually compacted to make the inter-membrane distance of 47 Å by replacing the adhesion proteins from MAG to MPZ (16). The stacked-rings-8-mer with the height of 67 Å may be formed during the myelin compaction to facilitate the transition of the inter-membrane gap from the MAGmediated 120 Å to the final 47 Å which, in this case, would be achieved by the trans interaction between ECDs.
In contrast, a recent X-ray diffraction study reported that the extracellular inter-membrane distance in compact myelin is not uniform and varies from 40 to 65 Å (17). Taking into account the size of the polar heads of lipids (18), the stacked-rings-8-mer may be accommodated in the compact myelin area with a relatively large intermembrane distance and work to stabilize the membrane layers. This is consistent with the freeze-fractured image of compact myelin, in which particles with an averaged diameter of 86 Å, of which 50-70% would be MPZ (2), are distributed randomly on the membrane surface, with no lattice-like regular distribution of MPZ required for the formation of the trans interaction (19,20).
In summary, we do not exclude the possibility that the trans interaction between ECDs stabilizes the membrane layers in compact myelin; however, we propose that the stacked-rings-8-mer of ECDs would also contribute to glue membranes, either as transient membrane stackers during the compaction process of PNS myelin or as major adhesion molecules in compact myelin regions with relatively longer intermembrane distances.

Mechanism by which the N87H substitution extinguishes the membrane-stacking activity of ECD
The N87H substitution was found in a patient with Dejerine-Sottas syndrome (DSS), a severe form of demyelinating neuropathy, with three simultaneous amino acid substitutions in ECD: I85T, N87H, and D99N (21). In this study, we found that the N87H substitution induces monomerization of ECD and impairs its membrane-stacking activity (Fig. 4A, 4D, and 5C). In the crystal structure of the multimeric ECDs, N87 was located at the cis interface and faced V2 of the adjacent ECD (Fig. 2B). The distance between the Cβ of N87 and the Cγ2 of V2 is 4.2 Å, indicating these atoms form a van der Waals interaction. The introduction of a bulk imidazole ring at N87 would induce a steric crash with V2 on the facing ECD, preventing the formation of cis interactions. In addition to the cis interaction, the N87H substitution prevents head-to-head interactions and causes monomerization of the protein. The W24A substitution at the cis interface also induced monomerization (Fig. 4A), suggesting that cis and head-to-head interactions are linked. In the crystal structure of the multimeric ECD, the B′-C loop constitutes both the cis and head-to-head interaction sites, and the structure of the loop is different from that of the monomeric ECD ( Fig.   S4A) (8). The cis interaction-induced conformational change of the B′-C loop would be a requisite for the formation of the head-to-head interaction. The head-to-head interaction in turn will stabilize the conformation of the B′-C loop to strengthen the cis interaction because the W28A substitution at the head-to-head interface also induces the monomerization of ECD (Fig. 4C). A subsequent kinetic study is required to fully understand the multimerization process of ECD. Notably, the transmembrane helix of MPZ contains a glycine-zipper oligomerization motif, which may stabilize the dimer and tetramer of intact MPZ, although we did not observe dimeric or tetrameric forms of ECD in the SEC experiment (22).
The N87H substitution induced a major loss-of-function effect on ECD, whereas DSS caused by heterozygous MPZ mutations have been associated with dominant-negative cytotoxic effects induced by aberrantly folded variant proteins (21,23). Since the in vitro folding efficiency of the N87H variant observed in this study was higher than that of WT ECD, the N87H substitution may have a minor impact on the folding property of MPZ and work synergistically with the other substitutions (I85T and D99N) to cause the severe phenotype of this patient. To understand the role of the N87H substitution in DSS, an integrated analysis with the other substitutions may be necessary.

Mechanism by which the D32G and E68V substitutions reduce the thermal stability of ECD
The D32G and E68V variants did not show apparent changes in membrane-stacking activity ( Fig.   4 and 5); however, they exhibited reduced thermal stability relative to WT ECD ( Fig. 6 and Table   1). These changes in thermal stability would result from changes in the unfolding process of the ECD-monomer, and the 8-mer-monomer transition will not be reflected in the changes in [θ]217 because changes in the β-sheet structure upon the 8-mer formation are subtle (β strand-fractions are 47 and 50% for the multimer and monomer, respectively).
E68 was located at the cis interface and faced S26 of the interacting ECD ( Fig. 2B and S4A).
These residues are 3.7 Å apart (between E68-C α and S26-Cβ) and make a van der Waals interaction, while no polar interaction is formed within these residues. The E68V substitution retains the van der Waals contact with S26, so it is convincing that the substitution did not alter ECD oligomerization (Fig. 4A). Rather, the E68V substitution exposed the hydrophobic side chain on top of the C″-D loop to the solvent (Fig. S2), which may facilitate partial unfolding of the loop and subsequent denaturation of the entire protein.
D32 is located at the head-to-head interface, and its Oδ1 is at a 3.6 Å distance from the Nζ of K55 (Fig. 2D and S4C). The side-chains of these two residues are connected by an electrostatic interaction and form the wall of the cavity that accommodates the indole ring of W28 of the interacting ECD (Fig. S4C). D32G seems to destabilize this cavity and prevent the 8-merization of ECD by erasing the electrostatic interaction with K55; however, the variant formed an 8-mer and retained membrane adhesion activity, suggesting that the electrostatic interaction between D32 and K55 makes little contribution to the head-to-head interaction (Fig. 4C, 4F, and 5J). Rather, the electrostatic interaction may be important for stabilizing the β-sheet structure by bridging the edges of the strands C and C′ (Fig. S4C), and loss of the interaction by the D32G substitution may facilitate transient partial unfolding (the breathing motion) of β-sheet (24), which would contribute to lowering the Tm. The fact that the D32N substitution did not affect the Tm may indicate that the hydrogen bond formed between N32 and K55 compensated for the electrostatic interaction.

Conclusive remarks
We showed that hMPZ-ECDs form stacked-rings-8-mer via cis and head-to-head interactions in solution and that this 8-mer is responsible for the membrane-stacking activity of ECD. It is unclear whether the stacked-rings-8-mer with a 67 Å height forms transiently during myelin compaction or stably in compact myelin regions with an average inter-membrane extracellular gap of 47 Å.
Further studies using Schwann cells or mutant animals are required to clarify this issue.
The N87H substitution, which is one of three simultaneous amino acid substitutions leading to DSS, induced pronounced effects on ECD activity; it eliminated multimerization and membranestacking activities. N87 is located at the cis interface and is not directly related to membrane adhesion. However, the N87H substitution allosterically affected the head-to-head interaction through the conformational change of the B′-C loop, which led to the loss of the membranestacking activity (Fig. S4A). This suggests that chemical compounds that bind to the cis interface may allosterically affect the head-to-head interface and revive membrane-stacking activity.
The effects of the CMT2-related D32G and E68V substitutions at the head-to-head and cis interfaces, respectively, were relatively moderate; they retained 8-mer-forming and membranestacking activities, while their thermal stabilities were lower than that of the WT. The reduced thermal stability of these variants may influence the long-term stability of these proteins in myelin and trigger the onset of CMT at middle-to-late ages (25,26).
In this study, amino acid substitutions at the trans interface, including three disease-related substitutions (D46V, A47V, and H52R), did not result in any apparent changes in ECD activity.
Our results do not necessarily exclude the possibility that this interaction occurs in compact myelin to stack membranes; however, another possibility is that ECD interacts with other myelin proteins essential for myelin compaction via this surface.

Expression and purification of recombinant ECDs
The synthesized gene encoding human MPZ-ECD (I1 through V121), optimized for Escherichia coli expression, was obtained from ATUM and inserted into the pET-28a vector (Merck) using the restriction sites for Nco1 and Xho1. The resulting vector, pET-28a-hMPZ-ECD, produced ECD with two artificial amino acids (MG) at its N-terminus. Using the protocol attached to the QuikChange Site-Directed Mutagenesis Kit (Agilent), the vector pET-28a-hMPZ-ECD was modified to pET-28a-hMPZ-ECD-His, in which a hexa-histidine-tag (His6-tag) was linked to the Cterminus of the ECD via a linker with the PTR sequence. The amino acid sequences of these proteins are listed in Table S1 The protein concentration of solubilized ECD or purified ECD-His was estimated by measuring the A280. A molar extinction coefficient of 36,370 was used for the ECD variants, except for the W24A and W28A variants, whose molar extinction coefficient was 30,680 (27). The protein solution was diluted by the solubilization buffer 1 or 2 to 0.5 mg/mL for ECD or 0.1 mg/mL for ECD-His, and the solution was dialyzed against the dialysis buffer 1 containing 20 mM Tris-HCl (pH 8.2) and 50 mM NaCl with 20 times as much volume as the inner solution at 4 °C overnight.
The outer buffer was exchanged for dialysis buffer 2, which contained 20 mM Tris-HCl (pH 8.2) and 60 mM NaCl, and the dialysis was continued at 4 °C for a day.

SEC-MALS analysis of ECD-His
The SEC-MALS experiment was carried out at room temperature using an Alliance HPLC system (Waters) combined with a DAWN HELEOS II detector (Wyatt) and a Superdex 200 Increase 10/300 GL column (Cytiva) equilibrated with the standard buffer. An ECD-His solution with a concentration of 52 µM was analyzed in the standard buffer. The flow rate was 0.5 mL/min. The injection volume was 0.2 mL. The data were processed using ASTRA 6.1 software (Wyatt) with a dn/dc value of 0.185 ml/g, which is typical for non-glycosylated proteins (28).

Crystallization and structure determination of ECD
The second SEC eluate of ECD, with a peak at 102 mL, was collected and dialyzed against a mM Tris-HCl (pH 8.5) and 300 mM magnesium formate within one week. The crystal was soaked in an increased concentration of magnesium formate with 30% v/v glycerol and frozen in liquid nitrogen. Diffraction data were collected at 100 K at a wavelength of 1.0 Å on the SPring-8 BL32XU using the Eiger X 9M detector. The datasets (360° in 0.1° steps) were processed and scaled using the XDS program package (30) and Aimless in CCP4 suite (31). The crystal belongs to the space group I422, with one molecule in the asymmetric unit. Phase information was obtained by molecular replacement with the PHASER MR (Phaser crystallographic software) (32) using the crystal structure of rMPZ-ECD (PDB ID 1NEU) as the search model. The replacement model was manually constructed using COOT (33) and refined to a resolution of 2.1 Å using the PHENIX package (34). Extra electron density was found on the surface of ECD adjacent to the trans interaction site. This might correspond to a glycerol molecule used as the cryoprotecting agent. The final model contained the ECD residues I1-V121 and a glycerol. Data collection and refinement statistics are presented in Table S2. The figures were prepared using PyMOL (Schrödinger).

Preparation of ECD-attached NDs
The plasmid for the expression of MSP1D1 was obtained from Addgene, and the MSP1D1 protein was produced and purified as previously described (11). Powders of 1,2-dimyristoyl-sn-glycero- In an inhibition experiment to analyze the contribution of the inter-ECD interactions in membrane-stacking, a sample containing NDs, ECD-His, and the 8-meric ECD (without a histidine-tag) with a molar ratio of 1:8:80 was prepared and analyzed by SEC as described above.
To confirm the contribution of the interaction between Ni 2+ -chelated lipids and ECD-His, a mixture of ND, ECD-His, and 250 mM imidazole was prepared and analyzed by SEC in the standard buffer supplemented with 250 mM imidazole.

Electron microscopic studies of ECD-attached NDs (Nanomyelin)
The ND-ECD-His variant complexes eluted at the void volume (8.

CD analyses of ECD-His
CD experiments were performed using a J-720 spectrometer (JASCO) equipped with a Peltiertype thermal controller. The concentrations of the ECD-His variants were adjusted to 6-22 µM in the standard buffer. The temperature-dependent change in the CD signal at 217 nm was monitored for each ECD-His variant in a quartz cuvette with a 0.2 cm path length. The temperature was increased from 20 to 90 °C at intervals of 1 °C and a gradient of 3 °C per min. Midpoint temperatures of denaturation (Tms) were determined using CDpal software (13).

Data and code availability
Atomic coordinates for human MPZ-ECD have been deposited in the PDB (https://www.rcsb.org/) with an ID of 8IIA.   Residues related to CMT1B, CMT2, and DSS are colored lime green, blue, and magenta, respectively. Amino acid substitution sites examined in this study are boxed.