Charcot-Marie-Tooth (CMT) disease is one of the commonest inherited peripheral neuropathies with a prevalence of about 10-30 per 100,000, depending on the country of origin 1 . CMT neuropathies are characterized clinically by progressive sensory loss, weakness, muscle atrophy, loss of deep tendon reflexes, and foot deformities. To date, about forty genes have been incriminated in CMT. Some of these genes encode mitochondrial proteins, i.e. mitofusin 2 (MFN2) [MIM:608507] and the ganglioside-induced differentiation-associated-protein 1 (GDAP1) [MIM:606598]. GDAP1 mutations are responsible: for CMT4A [MIM:214400] 2 , the most frequent recessive subtype of demyelinating CMT; for AR-CMT2 [MIM:607706] 3 , the axonal recessive subtype; for CMTRIA [MIM:608340] 4 , the intermediate recessive CMT; and for CMT2K [MIM:607831] 5 6 7 , the rare dominant subtype.

The gene locus, first identified in cases of CMT4A in Tunisian families on chromosome 8q13-q21 (CMT4A), is associated with slow motor nerve conduction velocity, and hypomyelination as detected by nerve biopsy 8 . The corresponding gene, GDAP1, was identified by Baxter et al. (2002). GDAP1 was previously identified as one of ten cDNAs expressed in a differentiated Neuro2a mouse neuroblastoma cell line, in which cholinergic differentiation with neurite sprouting had been induced by transfection of GD3 synthase (a2,8 sialyltransferase) cDNA 9 .

The GDAP1 gene spans ~17 kilobases of genomic DNA, containing six exons and five introns. Two transcript variants of GDAP1 have been identified. Transcript variant 1, representing the longer variant, encodes the longer isoform, designated as isoform a (358 amino acids). Transcript variant 2, encoding isoform b, contains an alternate in-frame exon, uses an alternate splice site in the 5'-coding region and a downstream start codon, compared to transcript variant 1. Isoform b (290 amino acids) has a shorter N-terminal sequence compared to isoform a. mRNA studies using RT-PCR have demonstrated the ubiquitous expression of GDAP1, predominantly in nervous tissues 3 .

Phylogenetic and structural analyses suggest that GDAP1 belongs to a subfamily of glutathione-S-transferases (GSTs) 3 10 11 . GDAP1 has two typical GST domains, domain I (GST-N) at the N-terminal region and domain II (GST-C) at the C-terminal region. The existence of these domains was predicted by using bioinformatics tools 3 10 but no functional GST activity has been demonstrated so far 11 12 . It has also been shown that GDAP1 has a single transmembrane domain (TMD) at the extremity of the C-terminal, and a hydrophobic domain (HD) in the flanking C-terminal region 10 13 . Moreover, GDAP1 has two additional regions between amino acids 152-164 and 169-195 that are predicted to represent two helices, α4 and α5, constituting the α4-α5 loop.

Several studies have suggested that GDAP1 plays a role in mitochondrial fission 14 15 . It was shown that the TMD plays a key role in mitochondrial targeting and mitochondrial membrane insertion 13 . Positively charged amino acids flanking the TMD at the C-terminal are involved in mitochondrial targeting and the fission function of GDAP1. The HD seems essential for the mitochondrial fission mediated by GDAP1. We have recently shown that GDAP1 mutations in fibroblasts from CMT2K patients are associated with defective mitochondrial complex I activity 5 .

The examination of the relationship between the genotype and the phenotype of the various forms of CMT disease involving GDAP1 mutations shows that the recessive forms are generally far more severe than the dominant forms. In particular, the truncating mutations have a more severe phenotype in the recessive forms with early onset and rapid progression of the disease leading to the inability to walk by the age of 10, generally followed by wheelchair-dependence before the end of the third decade 16 .

However, the phenotypic presentations of patients carrying GDAP1 mutations are heterogeneous. There is not only a large clinical variability concerning the ages of onset and functional disability among patients carrying the same mutations 17 but also a significant intra-familial variability 18 . Since the majority of mutations have been described in only a few unrelated patients, it is difficult to determine phenotype-genotype correlations for GDAP1 gene mutations. The analysis of a larger number of cases of CMT involving GDAP1 mutations, along with detailed clinical, electrophysiological and pathological data, will be required to establish a reliable relationship between the genotype and phenotype of the different forms of the disease.

The increasing importance of GDAP1 in CMT prompted us to develop a web-based locus-specific database (LSDB) for GDAP1 sequence variations with short publication delays. To our knowledge, the only other database collecting GDAP1 mutations is the Inherited Peripheral Neuropathies Mutation Database (IPNMDB) 19 , which includes 29 GDAP1 mutations among the 953 mutations in 44 genes related to inherited peripheral neuropathies. Our LSDB aims at collecting all the GDAP1 mutations incriminated in disorders involving mitochondrial dynamics and bioenergetics, with a full record of clinical, electrophysiological and biochemical data. We here describe the construction of the database, the procedure for data submission, and the presentation of the data.

We have established a locus-specific database for GDAP1 using the Leiden Open-source Variation Database (LOVD) version 2 build 31 system 20 . This database includes a total of 20 variant items and 18 clinical patient items. Figure 1 shows a typical entry. A standardized description of the clinical and molecular items is set up using drop-down lists with predefined variables. The clinical features are based on the usual symptoms of mitochondrial diseases.

The data is openly accessible and should prove a valuable tool for clinicians and researchers alike since it will contain published as well as unpublished sequence variants. Contributors can submit their variants online to the database after registering for a login and password. This contact information is collected for reference purposes and clarification of the data submitted. The description of mutational data is confirmed using the web-based software Mutalyzer, to which the LOVD software is currently also linked 21 . Variations submitted directly to the LSDB are re-checked with these tools.

The GDAP1 database reviews all the clinical and molecular data of patients carrying GDAP1 mutations published in peer-reviewed literature, as well as unpublished contributions that may be directly submitted. While most variants can be described in terms of the latest update of the standard nomenclature 22 23 , some inaccuracies may persist because gene anomalies discovered earlier might have been named according to a convention now out of use. Eventually, the DNA published field of the page dedicated to each variant (Figure 1b) indicates whether the published name of the mutation has been modified by the curator. The GDAP1 LSDB website requires absolute compliance with the rules set out below to describe sequence variants in order to provide uniform and comparable data.

The GDAP1 mutations are described according to the GDAP1 transcript variant 1 [RefSeq:NM_018972.2], containing 6 exons and encoding isoform a containing 358 amino acids. The nucleotide numbering of nuclear genes reflects cDNA numbering with +1 corresponding to the A of the ATG translation initiation codon in the reference sequence, according to the guidelines of the Human Genome Variation Society (HGVS). The initiation codon is codon 1.

The criteria of pathogenicity, dependent on the clinical context and molecular findings, are stated in the Variant data fields: Reported pathogenicity and Concluded pathogenicity (Figure 1b). Putative novel mutations detected in affected patients must segregate according to disease status and not be present in control individuals. Putative mutations are graded according to type of mutation: frameshift and nonsense mutations are considered to be pathogenic; missense mutations are described as being of unknown pathogenicity when detected in single families, or as probably pathogenic when detected in a group of families; the mutations are considered to be pathogenic when so proven by experimental evidence or detected in multiple families.

One of the patients with a GDAP1 mutation referred to our database also carried a mutation in the MFN2 gene 31 . Such cases of multiple gene mutations are now being brought to light more often with the rapid advance of genetic investigations. The GDAP1 database is part of the MITOchondrial DYNamics variation portal (mitodyn.org), aimed at specializing in genes involved in disorders of mitochondrial dynamics 32 . The mitodyn.org will therefore incorporate other genes involved in diseases affecting mitochondrial dynamics and bioenergetics. Thus, the OPA1 LSDB (eOPA1) 33 , curated by our laboratory, will soon be transferred to mitodyn.org. We will then integrate MFN2, responsible for CMT2A2 [MIM:609260] 34 , and DNM1L associated with encephalopathy with lactic acidosis [MIM:603850] 35 . Since the same patients may be affected by these different diseases involving mitochondrial dysfunction, the data collected should interest physicians and researchers alike. Eventually, we hope to be able to refine genotype-phenotype correlations in diseases involving mitochondrial dynamics by comparing and cross-checking the clinical, electrophysiological and biochemical data recorded in the mitodyn.org databases.

The authors declare that they have no competing interests.

JC and MF initiated the project, created the database, monitored data collection and drafted the manuscript. AC participated in writing the article and the analysis of results. DB, CV, VP, PR and MF supervised the project and revised the manuscript. All authors have given final approval of the version to be published.