Subjects and Methods
A total of 726 Japanese patients with clinically diagnosed CMT were enrolled in this study. However, demyelinating patients whose PMP22 duplication or deletion were confirmed using fluorescence in situ hybridization or multiplex ligation probe amplification were not enrolled. The protocol was reviewed and approved by the institutional review board of Kagoshima University (Kagoshima, Japan). All patients and family members provided written informed consent to participate in this study.

Patient Selection of Original Study for WES
To identify novel AR‐CMT disease‐causing genes using WES analyses, we collected DNA samples in a first case series of 372 Japanese patients with clinically diagnosed CMT between April 2007 and April 2012. Subsequently, we excluded 69 patients with pathogenic mutations in the 28 genes known to cause CMT or IPN (Supplementary Table 1), which were detected using the custom MyGeneChip CustomSeq Resequencing Array (Affymetrix, Inc., Santa Clara, CA) following the protocol described previously.14, 15, 16, 17 We performed WES in the remaining 303 patients and excluded 85 patients with pathogenic mutations in the known CMT genes and in other IPN genes (Supplementary Table 1). We also excluded 55 patients suspected of AD or X‐linked inheritance pattern on the basis of family history information as well as undetermined inheritance. After these exclusions, we selected 163 unrelated patients with presumed AR inheritance or with sporadic inheritance and no known genetic etiology (Fig 1).
Figure 1  Flow chart for genetic tests, selection of patients, and exome‐based shared variants detection (ESVD) system among the first case series of 372 Charcot–Marie–Tooth (CMT) patients, we selected 163 patients with presumed autosomal‐recessive or sporadic CMT and no known genetic etiology. Using the ESVD system, we automatically performed variant filtering under the conditions (1) to (6). Subsequently, we identified 5 patients with a mutation in the MME gene by the shared variants/genes pickup system. AD = autosomal‐dominant; AR = autosomal‐recessive; SNV = single‐nucleotide variation; MAF = minor allele frequency.

Patient Selection of Additional Study
To investigate whether MME mutations cause CMT and to investigate the frequency of patients with MME mutations, we further performed targeted resequencing using next‐generation sequencing (NGS) in a 137 second case series between May 2012 and June 2014 and a 217 third case series of unrelated Japanese patients with CMT between July 2014 and March 2015 (Fig 2).
Figure 2  Flow chart for genetic tests in the additional patients. During the second and third case series, we identified 1 and 4 patients with mutations in the MME gene.

Extraction of Genomic DNA
Genomic DNA of the patients and family members was extracted from peripheral blood using QIAGEN's Puregene Core Kit C (QIAGEN, Valencia, CA) or from saliva using the Oragene DNA self‐collection kit (DNA Genotech, Ottawa, Ontario, Canada), according to the manufacturer's protocol.

WES
Three micrograms of genomic DNA were processed using a SureSelect v4+UTRs or v5+UTRs kit according to the manufacturer's instructions. Captured DNA was sequenced using a HiSeq 2000 (Illumina, San Diego, CA). Sequences were aligned to the human reference genome (NCBI37/hg19) using the Burrows–Wheeler Aligner.18 Variant calling was performed using SAMtools.19 Variants were annotated using in‐house scripts, which provided the variants list. Previously known variants were annotated from the 1000 Genomes Project and dbSNP137.

Development of the Exome‐Based Shared Variants Detection System
To extract candidate genes or variants from the WES data, Maze Inc. (Tokyo, Japan) developed user‐friendly analysis software, called the exome‐based shared variants detection (ESVD) system, under our supervision. This system consists of two parts: a “variant‐filtering system” and a “shared variants/genes pickup system.” The variant‐filtering system provides users with many options for filtering the raw variant data. The conditions for filtering variants include mutation type, genotype, quality, read depth, minor allele frequency (MAF), and presence in public databases; dbSNP137, 1000 Genomes project database (http://browser.1000genomes.org), and The Human Genetic Variation Database (HGVD) comprised of exome sequencing of 1,208 Japanese individuals (http://www.genome.med.kyoto‐u.ac.jp/SnpDB/). Moreover, the shared variants/genes pickup system allowed us to detect the variants and genes shared among cases with similar phenotypes, and the number can be custom defined. In this study, using the ESVD system, we automatically performed variant filtering under the following conditions: (1) mutation type (single‐nucleotide variation [SNV]; not including insertion and deletion); (2) mutation classification (nonsynonymous, nonsense, and splicing site variants); (3) genotype (homozygous); (4) Phred‐scaled quality score/mapping quality (both >20); (5) read depth (> 10); and (6) existence in the public database (none or MAF <1%). Subsequently, we manually selected the variants on an autosome. Eventually, after filtering, the variants that were shared by three or more cases were extracted (Fig 1). We also manually extracted presumed compound heterozygous variants of the MME gene from the WES data set of the 163 patients.

Mutation Screening of the MME Gene in the Additional Patients
For the second series, we performed mutation screening of 60 known or candidate CMT‐related genes using the Illumina Miseq platform (Illumina Inc., San Diego, CA, USA) following the protocol described previously by Maeda et al.20 Subsequently, we performed mutation screening of other known CMT and IPN genes, including the MME gene, using WES (Fig 2). For the third series, we performed targeted resequencing of 72 known or candidate CMT‐related genes, including the MME gene, using a custom Ion AmpliSeqM panel (Life Technologies, Carlsbad, CA) (Fig 2 and Supplementary Table 1). Briefly, a custom panel targeting 72 genes was designed using the Ion AmpliSeq Designer tool (http://www.ampliseq.com). Library and template preparation was performed according to manufacturer's instructions. Sequencing was performed on the Ion Proton (Life Technologies) using the Ion PI Chip kit v2 BC. Sequence data were aligned and mapped to the reference sequence, and variants were called using the Torrent variant caller. Variants were transferred to the CLC Genomics Workbench 6 software program (CLC bio, Aarhus, Denmark) for annotation and filtering.

Mutation Confirmation and Segregation Analysis
Using Sanger sequencing, we reconfirmed the pathogenic mutations revealed by microarray or NGS. Furthermore, segregation studies were performed in the available family members whenever possible. In order to rule out the possibility that the prioritized variants in the MME gene may be Japanese‐specific polymorphisms, we confirmed whether the variants that existed in the in‐house database comprised of the WES data from 800 Japanese healthy control subjects and 3,742 disease control subjects excluding patients with CMT.

In Silico Analysis
To determine whether variants in the MME gene are damaging or deleterious, we obtained the predicted functional scores of nonsynonymous variants or deletions with three prediction algorithms, including PolyPhen2,21 SIFT (Sorts Intolerant From Tolerant amino acid substitutions),22 and PROVEAN (Protein Variation Effect Analyzer).23

RNA Extraction and Reverse‐Transcription Polymerase Chain Reaction
To study the messenger RNA (mRNA) expression of the MME gene, we performed reverse‐transcription polymerase chain reaction (RT‐PCR) in individuals from some families with MME mutations. Whole blood was collected into PAXgene Blood RNA Tubes (Qiagen), and total RNA was prepared from blood using the PAXgene Blood RNA Kit (Qiagen) in order to generate a complementary DNA (cDNA) pool by RT‐PCR using the High‐Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Carlsbad, CA), according to the manufacturer's instructions. PCR primers for the β‐actin housekeeping gene were used as the internal control. MME cDNAs were amplified using the following primer pairs: (1) forward primer located in exon 6: 5′‐TGATAGCAGAGGTGGAGAACC‐3′ and reverse primer in exon 9: 5′‐CATCGATGGGCAATCTTTCT‐3′; (II) forward primer in exon 4: 5′‐AATGTCATTCCCGAGACCAG‐3′, reverse primer in exon 7: 5′‐TCATCAGTGCCAACAAACAA‐3′. The expected sizes of the RT‐PCR products were 350bp (base pairs) and 344bp for primer pairs 1 and 2, respectively. PCR products were subjected to agarose gel electrophoresis and sequenced using the Sanger method.

Haplotype Analyses
To determine whether recurring MME mutations (c.654+1G>A) occurred because of independent mutational events or common ancestry, we performed a haplotype analysis using five microsatellite markers (D3S1595, D3S1280, D3S3509, D3S1275, and D3S3692) and seven single‐nucleotide polymorphisms (SNPs; rs1803155, rs12493885, rs12497267, rs9438, rs358733, rs11918974, and rs3816527) flanking the MME gene by an automated fluorescent method on an ABI 3130xL or 3500xL genetic analyzer (Applied Biosystems).

Cognitive Assessment and Neuroimaging
In some patients, we performed brain magnetic resonance imaging (MRI) or computed tomography (CT), and a neuropsychological examination including the Mini–Mental State Examination (MMSE; 0–30 scale, normal >24)24 and the Alzheimer's Disease Assessment Scale‐Cognitive Behavior Section (ADAS‐cog; 0–70 scale, normal <10).25 We also evaluated patient 1 (P1) with amyloid positron emission tomography (PET) imaging with 11C‐PIB (Pittsburgh compound‐B). PET data were acquired on a Discovery PET/CT 710 (GE Healthcare, Tokyo, Japan) after the injection of 11C‐PiB (470 MBq). 11C‐PiB images were derived from dynamic summations of standard uptake values (SUVs) over 60 minutes and were normalized with the cerebellum as a reference region (SUVR).

Histopathological Examinations
Sural nerve biopsies obtained from P1 and P4 were analyzed for morphometric changes using light microscopes. Semithin sections from Epon embedded tissues were stained with Toluidine blue. An immunohistochemical assay was performed using an anti‐NEP/CD10 antibody (ab951; Abcam, Tokyo, Japan) in tissue from human tonsils and peripheral nerves, and the findings were compared to a disease control patient (female, 39 years old, polymyositis).

Western blotting
Western blotting was performed using sodium dodecyl sulfate/polyacrylamide gel electrophoresis. Briefly, 20‐μg homogenates of sural nerve were resuspended in a reduced sample buffer, electrophoresed on a 10% Tris gel with Tris running buffer, blotted to a polyvinylidene difluoride membrane, and sequentially probed with polyclonal rabbit antihuman CD10 antibody (Abcam). The membrane was subsequently incubated with a horseradish peroxidase–labeled polymer‐conjugated antimouse antibody reagent (EnVision+ reagent; Dako, Tokyo, Japan). 3‐3′‐diaminobenzidine was used for chromogenic visualization.