CMT disease: historical perspectives

In their first clinical observations in 1886, Jean-Martin Charcot, Pierre Marie10 and Howard Tooth11 described a slowly progressive peroneal muscular atrophy (PMA) that does not affect the upper extremities until several years.12 A few years afterwards, Jules Dejerine and Jules Sottas described the ‘hypertrophic interstitial neuritis’ (now called Dejerine–Sottas syndrome),13 which corresponds to a hereditary motor and sensory neuropathy (HMSN) characterised by onset in infancy or early childhood. Later on, Gustave Roussy and Gabriel Levy14 described another variant of HMSN with tremor and ataxia, called ‘Roussy–Levy syndrome’ (RLS),15 now considered as a variant of CMT1 (the demyelinating form of CMT),16 further increasing the nosologic complexity of CMT.17 In 1927, Sergeij Dawidenkov presented a clinical classification in which these hereditary neuropathies were classified into 12 categories,18 but it has never been routinely used. In 1968, Peter Dyck and Edward Lambert proposed the use of clinicopathological data, natural history, inheritance pattern and motor nerve conduction velocities (MNCV) measured on median nerves to categorise the various forms of PMA.19 ,20 Since then, the electrophysiological dichotomy of demyelinating (MNCV <38 m/s in median nerve) and axonal (MNCV >38 m/s in median nerve) hereditary neuropathies is the mainstay of the classification, although it is well known that both groups are heterogeneous.21 With time, it appeared reasonable to individualise an ‘intermediate form’, which corresponds to patients who have MNCV between 30 and 40 m/s (in median nerve).22 ,23

In 1975, the ‘hypertrophic’ and ‘neuronal’ (or axonal) forms of PMA were respectively named HMSN type I and type II;24 actually, in this classification, HMSN were divided into seven groups: HMSN I (dominantly inherited hypertrophic neuropathy), HMSN II (neuronal type of peroneal muscular atrophy), HMSN III (hypertrophic neuropathy of infancy corresponding to DSS), HMSN IV (Refsum's disease), HMSN V (neuropathy associated with spastic paraplegia (SPG)), HMSN VI (neuropathy associated with optic atrophy) and HMSN VII (neuropathy with retinitis pigmentosa).2 Over the years, the term ‘HMSN’ was replaced with ‘CMT’ (but HMSN is still widely used by neurologists), the purpose being to stress that most of the hereditary motor sensory neuropathies are in fact CMT. After the discovery of the first genetic abnormality in CMT (duplication of the PMP22 gene),25 mutations in numerous other genes were progressively identified in patients with CMT. Because of the great number of genes involved (and the constant discovery of new ones), the current classification needs to be permanently updated.

CMT disease: a heterogeneous genetic syndrome

To date, numerous genes have been implicated in hereditary neuropathies, particularly in CMT and dHMN, which share a clinical and molecular overlap with CMT: altogether, >60 genes have been identified.26 In spite of this genetic heterogeneity, all forms of CMT have quite a homogeneous clinical phenotype: PMA, skeletal deformities (including pes cavus), and usually decrease or absence of tendon reflexes.17 ,27 Other signs, such as pyramidal tract involvement (CMT5) or optic neuropathy (CMT6), may also be observed. Some specific features (clinical, biological or pathological) can be suggestive of some particular genes. For example, the association of a CMT phenotype with proteinuria and/or abnormal elongated expansions of Schwann cell cytoplasms (as evidenced by electron microscopic examination of a nerve sample) is strongly suggestive of INF2-related CMT.28 When additional signs are found, some authors have proposed the term of ‘CMT plus’ (eg, as CMT5 and CMT6), but the individualisation of such a supplemental subtype does not add any meaningful information in our opinion.

The majority of CMT subtypes have AD inheritance, although X-linked and AR transmission also exist.16 Most of the CMT diseases are demyelinating (CMT1 for the AD forms and CMT4 for the AR ones), and up to one-third are primarily axonal (type 2: AD-CMT2 and AR-CMT2). In CMT, information of families in term of diagnostic certainty, mode of inheritance, prognosis and sometimes therapeutic trials (see further) requires an accurate genetic diagnosis.27 ,29 Recently, Saporta et al,30 in a personal series of >500 patients with CMT, reported that the most prominent CMT subtypes identified were CMT1A (duplication of PMP22), CMTX1 (mutations of GJB1), hereditary neuropathy with liability to pressure palsies (deletion of PMP22), CMT1B (mutations of MPZ) and CMT2A (mutations of MFN2); other CMT subtypes accounted for <1% of all patients with genetically defined CMT; only 1.8% of patients with CMT1 were without a genetic diagnosis, although 65.6% of patients with CMT2 were in the same situation.30 Murphy et al31 (on 916 patients) have found that four common genes (PMP22, GJB1, MPZ and MFN2) account for >90% of all CMT molecular diagnoses. Finally, Fridman et al32 (on 997 patients) have confirmed that the most frequent CMT subtypes are CMT1A, CMTX1, CMT2A, CMT1B and HNPP, these genes being affected in 89.2% of all genetically confirmed mutations.

Among CMT2 patients, the most frequent mutated gene is MFN2 and, at the present time, it is said to account for 10–30% of these cases; otherwise, mutations in NEFL gene have been recently found as unexpectedly frequent in CMT2 patients of Chinese origin.33 Concerning AR forms, many of the reported families and patients originated from Mediterranean countries as a consequence of the high rate of consanguinity (especially in Maghreb and the Middle East).

At last, a close relationship between CMT2 and dHMN was noticed before the identification of mutated genes: first, CMT2 is generally a peripheral nerve disease with predominant motor involvement, patients having sometimes no or little sensory symptoms;34 ,35 in addition, families were identified in which some patients had CMT2 and other dHMN.36 Identification of genes responsible for dHMN confirmed the existence of an overlap between the two conditions.

A continuum between earlier onset forms and adult-onset forms of CMT disease

CMT3, or HMSN III, corresponds to DSS, which has been described as an early-onset form of demyelinating HMSN with a delay in motor milestones and very low MNCV (<10 m/s); it is now more regarded as a clinical syndrome rather than as a separate entity.37 ,38 DSS was initially considered as having an AR inheritance, but both AD and AR modes of inheritance were further reported.39 This syndrome can be considered as a continuum between the syndrome of congenital hypomyelinating neuropathy (CHN) and CMT1. In DSS, initially described as ‘hypertrophic demyelinating neuropathy’,13 nerve biopsy shows diffuse and severe demyelinating lesions causing sometimes palpable nerve hypertrophy because of the great number of ‘onion bulb’ Schwann cell proliferations.40 Some DSS cases have been linked to heterozygous mutations of PMP22, MPZ and EGR2 genes, known to be involved in peripheral nerve system (PNS) myelination.41 Severe cases of early-onset hereditary neuropathy have been associated with other genes such as GDAP1,42 NEFL43 and PRX.39 In fact, DSS progressively became synonymous of severe early-onset demyelinating CMT, irrespective of the mode of inheritance.

The most severe form of demyelinating neuropathy corresponds to the entity called ‘congenital hypomyelinating neuropathy’ and its most extreme phenotype (complete absence of myelin) that is present at birth or in the first months of life. CHN is mostly due to defective synthesis and maintenance of myelin and many cases have been reported before the era of molecular genetics, with detailed pathological data.44 During the last 15 years, mutations have been reported in various genes (most of them concerned CMT1 genes), but unfortunately, histopathological data are scarce in CHN patients with determined gene mutations. They generally show a severe loss of myelin sheaths and lesions, suggesting a process of dysmyelination.39 ,44 ,45 A few cases of CHN, characterised by a complete absence of myelin in the patients’ peripheral nervous system, have been described.46 In one patient, born from consanguineous parents, a homozygous deletion encompassing a myelin-specific enhancer of EGR2 was identified. This genetic anomaly was associated with a complete loss of EGR2 expression in the peripheral nervous system.47 By contrast, axonal forms of congenital peripheral neuropathies seem to be very rare.39 We observed only a few cases associated with MFN2 mutations (personal unpublished data).

Giant axonal neuropathy (GAN), a devastating childhood disorder affecting both the peripheral nerves and the central nervous system, is due to mutations in the GAN gene encoding gigaxonin, a protein implicated in the cytoskeletal functions and dynamics. In the majority of the GAN series reported to date, patients had the classical clinical phenotype characterised by a severe axonal neuropathy with kinky hair and early-onset central nervous system involvement including cerebellar and pyramidal signs.48 However, patients with milder CMT-like phenotype have been reported, demonstrating a clinical heterogeneity.49 Nerve biopsy study reveals a severe axonal loss and the hallmark of this condition, the ‘giant axons’ filled with intermediate filaments. So, some authors estimate that GAN could be a continuum with CMT2.49

Early-onset severe forms of CMT2 have been described under the names of ‘severe early-onset axonal neuropathy’ or ‘early-onset hereditary motor sensory neuropathy of axonal type’.50 ,51 They represent an heterogeneous phenotype first delineated by Ouvrier et al,50 characterised by a slow axonal degeneration of peripheral nerves with gait problems often progressing to wheelchair requirement and later respiratory involvement. Inheritance is frequently unclear since most cases are sporadic; some were found to harbour heterozygous de novo dominant mutations of the MFN2 gene, which encodes the mitochondrial fusion protein mitofusin 2.52 ,53 The patients reported by Yum et al51 had compound heterozygous (AR) mutations of the NEFL gene. At present, SEOAN is mostly regarded as a particular clinical syndrome rather than a subtype of CMT, as occurs with DSS.37

Genetic analysis in CMT disease: from phenotype to genotype (and back)

A negative family history does not exclude CMT; sporadic cases induced by de novo mutations would occur in about 10% of the CMT1A cases.54 In addition, nuclear families are usually non-consanguineous and are of small-sized in most Western countries, which makes most cases of AR diseases appear sporadic. Otherwise, mild clinical forms might not be recognised when first-degree relatives of an index case are not systematically clinically and electrophysiologically assessed; a few patients may present only areflexia and nerve conduction anomalies. CMT clinical phenotypes present common characteristics of progressive distal muscle weakness and atrophy, foot deformities, distal sensory loss and depressed tendon reflexes. Despite these similarities, a thorough clinical examination may identify other clinical data that may orientate genetic testing. In table 1, we have mentioned some clinical and biological features, which are characteristic of different subtypes of CMT. This list is not exhaustive, and most of these clinical and biological data must not be considered as specific but as evocative. For example, a recent study has demonstrated that the concomitant occurrence of AD-CMT, age-related macular degeneration and hyperelasticity of the skin is caused by mutations in FBLN5.55 It has also been shown that INF2 mutations are responsible for a dominant-intermediate CMT associated to a focal segmental glomerulosclerosis in a significant number of families,28 ,56–58 so we think that a simple screening test for proteinuria in AD-CMT patients without a known gene mutation may be useful.

Taken together, recent epidemiological data may help clinicians to manage genotyping in CMT.30 Indeed, it could be suggested that clinicians in a first step orientate genotyping by taking into account clinical findings, hereditary transmission, electrophysiological data and country of origin to look for an anomaly in the following genes: PMP22, GJB1, MPZ and MFN2. Then, if these first tests are negative, patients should be referred to reference centres to consider carefully some clinical, biological and optionally pathological features.

It has to be outlined that the rapid development of next-generation sequencing (NGS) techniques and their applications to clinical diagnosis has changed this situation deeply. It is now possible to test all known CMT genes for several patients in a single experiment.59 A new problem arising from these techniques is the interpretation of the identified variants: for example, the causative nature of a new sequence variant or the presence of several genetic variants in the same individual (oligogenic inheritance or rare ‘private’ polymorphisms in more than one CMT gene). On the other hand, the power of NGS might allow the identification of gene mutations that have been missed by classic Sanger sequencing because of technical problems (low efficacy of primer hybridisation on the mutated strand because of a nucleotide variant in the oligonucleotide complementary sequence). Currently, for molecular diagnosis, NGS associated with Sanger strategy is a really efficient method: while NGS allows the screening of a large number of genes, Sanger strategy can be used to sequence the supplement regions missed by NGS, allowing a full screening of the targeted genes.

In the era of NGS, clinical (table 1) and neurophysiological findings are critical to evaluate sequencing data, particularly when more than one variant has been identified. The indication of a nerve biopsy is quite rare in this context, relies on a case-by-case discussion and its interest must not be overemphasised; in a few cases, its role may be limited to the differential diagnosis. Although biopsy and fixing techniques of specimens can be performed anywhere provided that adequate procedures are employed, the preparation for microscopic examination can only be carried out in specialised laboratories where nerve samples can be easily sent by regular mail. Some of the characteristic ultrastructural lesions are mentioned in table 2. For instance, PMP22 duplication and mutations induce a homogeneous demyelination and numerous large ‘onion bulbs’ corresponding to the abnormal proliferation of Schwann cells; in cases of P0 mutations, there is also a homogeneous demyelination with ‘onion bulbs’ and sometimes numerous and extensive outfoldings of myelin sheaths that are different from typical ‘tomacula’ formations (characterised by focal hypermyelinations with smooth external contours that have been described in HNPP, a disease stemming from a deletion of PMP22). In about 50% of nerve biopsies of P0 mutated cases that we have carefully studied ultrastructurally, we detected a significant number of myelin sheaths (with or without outfoldings) that exhibit severe abnormalities of myelin compaction. GJB1 mutations are associated to numerous clusters constituted by small myelinated fibres with scattered ‘onion bulbs’; MTMR2, SBF2 and FGD4 induce many infoldings and outfoldings of the myelin lamellae; SH3TC2 and INF2 usually present characteristic anomalies of the cytoplasm of unmyelinated Schwann cells; PRX mutations induce subtle anomalies of the nodes of Ranvier; NEFL mutations are associated with modifications of neurofilament density and loss of microtubules; in all nerve biopsies of MFN2 and GDAP1 mutations that we have observed by electron microscopy (about 10 cases of each subtype), we were able to detect significant and unusual modifications of intra-axonal mitochondria, which are abnormally focally aggregated, round instead of tubular, with irregularities in their external and internal membranes, and disruption of their cristae; NDGR1 mutations are characterised by the accumulation of amorphous material in the adaxonal space; LMNA mutations induce a severe rarefaction of myelinated fibres, without any sign of regeneration.60–63

In summary, a detailed clinical examination and a complete neurophysiological evaluation are highly useful in patients with CMT in order to orientate genetic testing or to evaluate genetic test results. Moreover, additional pathological findings might add a great value in selected cases to confirm the pathogenicity of a given mutation.

CMT disease: what are the limits of the current classification? Proposition of a new classification

In the classification currently used, CMT1 is synonym of HMSN I, CMT2 of HMSN II, CMT3 of HMSN III, but this correspondence is not actually valid for CMT4, which does not correspond to HMSN IV (Refsum's disease) but to AR demyelinating CMT. The fact of having kept the term ‘CMT4’ was due to the confirmation of a linkage of a locus (in the AD demyelinating form of CMT called ‘CMT4A’) to chromosome 8 (8q13-q21) by Ben Othmane et al,64 at a time when only three forms of CMT (CMT1, CMT2 and CMT3) were well categorised: the identification of the gene GDAP1 led to individualise a first subcategory called CMT4A. Afterwards, the discovery of mutations of new genes associated to this particular pattern ended in the name of new forms of CMT4 such as ‘CMT4A’ and ‘CMT4B’.26 Over the years, the discovery of many genes susceptible to be responsible for genetic PMA made the classification of CMT more and more complicated.26 Some subtypes, such as CMTX3, are finally not preserved.65 Subclasses have been created such as CMT2B1 and CMT2B2; only five letters are still available for other CMT2 subtypes (V–Z). In CMTX, numbers are used instead of letters to categorise the subcategory of CMT (eg, CMTX1 and CMTX2). Progressively, we have learnt that mutations in a same gene may cause different CMT phenotypes; for example, mutations in MPZ usually present as a dominant demyelinating form (CMT1B), more rarely as a dominant axonal form (CMT2I/J) and also as DSS.21 It also became evident that the same gene may harbour mutations characterised by different patterns of transmission; for example, mutations in NEFL and in GDAP1 may lead to AD-CMT or AR-CMT.66

In order to simplify the actual classification and to avoid further increasing complexity, we first propose to include the pattern of inheritance in the denomination of each subtype. CMT1, which stands for autosomal-dominant demyelinating CMT, should be thus renamed autosomal-dominant CMT1 (AD-CMT1). CMT4 should then be renamed autosomal-recessive demyelinating CMT (AR-CMT1). A similar reasoning should apply to axonal CMT (CMT2). The spectrum of axonal CMT includes AD and AR cases that should rather be named AR-CMT2 and AD-CMT2. For CMTX, we think that the term ‘XL-CMT’ is more appropriate (XL=X linked). The term ‘spo’ might be used in case of sporadic patients with undetermined molecular cause (spo-CMT).

A concomitant proposition may consist to replace the classical numerals ‘1’ (for CMT1) and ‘2’ (for CMT2) by ‘de’ for demyelinating (‘CMT1’ becoming ‘CMTde’) and ‘ax’ for axonal (‘CMT2’ becoming ‘CMTax’), respectively. We also suggest the use of ‘CMTin’ instead of ‘intermediate CMT’. We think that the term ‘CMT+’ should be definitively suppressed (so the various forms of CMT5 and CMT6 may be classified into AD-CMT, AR-CMT, XL-CMT and spo-CMT) (table 1).

We also propose to insert the name of the implicated gene instead of the use of the traditional letters (eg, A, B, C) aimed to define the different subtypes until now. For example, CMT1B and CMT2I/J could be then named AD-CMTde-MPZ and AD-CMTax-MPZ, respectively. We think that this denomination is more logical and easy to use (particularly for the layman). By the way, it could also be possible to specify whether the mutation is a duplication by adding ‘dup’ after the name of the gene (eg, AD-CMTde-PMP22dup=CMT1A in the classical classification) or a point mutation by adding nothing after the name of the gene (eg, AD-CMTde-PMP22=CMT1E in the classical classification); in case of a deletion of the PMP22 gene (HNPP), we could add the term ‘del’ after the name of the gene as well, even if we think that we have to keep the term ‘HNPP’ for this particular entity; for example, ‘CMT1A’ will then become ‘AD-CMTde-PMP22dup’, ‘CMT4A’ and ‘AR-CMTde-GDAP1’ (figure 2). In cases of two simultaneous mutations in different genes in the same patient, a common label may group the two forms of CMT: for example, in case of the simultaneous occurrence of a CMT2A2 (mutations in MFN2) and a CMT2K (mutations in GDAP1),67 ,68 we would use the term AD-CMTax-MFN2-GDAP1; in the same way, mutations of PMP22 were reported in association with another neuromuscular genetic mutation in a second gene, such as GJB1 or DMPK.69 Moreover, it could be also decided to conclude that the terms ‘DSS’, ‘CHN’, ‘RLS’ and ‘SEOAN’ should be definitely abandoned for greater simplicity, except HNPP, which is definitely a separate entity.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Explanations of our suggestions for a modified classification of Charcot–Marie–Tooth (CMT) diseases and distal hereditary motor neuropathies (dHMN) (α, any gene; dup., duplication).

Finally, with these suggestions, although the acronym for each entity is obviously longer, the main characteristics of the neuropathy are clearly pointed out, easily understood, and can be remembered by anyone. We speculate these changes would probably be more readable for non-experts in the field and at the same time provide useful information for genetic counselling.

So, only the neuropathies with a clear clinical phenotype of HMSN should be included in this classification. In our classification of HMSN (excluding HNPP and dHMN), there are now 68 variants of CMT diseases, for a total of 54 known causative genes and 3 unknown genes (table 3). Of course, dHMN have also a clinical phenotype of PMA, but are characterised by an exclusive involvement of the motor part of the peripheral nervous system. These dHMN are also clinically and genetically heterogeneous; in the usual CMT classifications, these cases are apart, which seems quite reasonable. So we propose to classify them in a separate classification (table 4): 26 variants of dHMN, for a total of 22 known causative genes and 4 unknown genes. Both CMT and dHMN comprise 94 different forms, for a total of 67 known causative genes and 7 unknown genes (among all these genes, 9 are found in both CMT and dHMN).