Charcot–Marie–Tooth disease and intracellular traffic

Highlights

► Intracellular traffic defects cause important human diseases. ► Mutations causing CMT often alter intracellular trafficking. ► Alterations of intracellular traffic are one of the major causes of CMT neuropathy.

Abstract

Mutations of genes whose primary function is the regulation of membrane traffic are increasingly being identified as the underlying causes of various important human disorders. Intriguingly, mutations in ubiquitously expressed membrane traffic genes often lead to cell type- or organ-specific disorders. This is particularly true for neuronal diseases, identifying the nervous system as the most sensitive tissue to alterations of membrane traffic. Charcot–Marie–Tooth (CMT) disease is one of the most common inherited peripheral neuropathies. It is also known as hereditary motor and sensory neuropathy (HMSN), which comprises a group of disorders specifically affecting peripheral nerves. This peripheral neuropathy, highly heterogeneous both clinically and genetically, is characterized by a slowly progressive degeneration of the muscle of the foot, lower leg, hand and forearm, accompanied by sensory loss in the toes, fingers and limbs. More than 30 genes have been identified as targets of mutations that cause CMT neuropathy. A number of these genes encode proteins directly or indirectly involved in the regulation of intracellular traffic. Indeed, the list of genes linked to CMT disease includes genes important for vesicle formation, phosphoinositide metabolism, lysosomal degradation, mitochondrial fission and fusion, and also genes encoding endosomal and cytoskeletal proteins. This review focuses on the link between intracellular transport and CMT disease, highlighting the molecular mechanisms that underlie the different forms of this peripheral neuropathy and discussing the pathophysiological impact of membrane transport genetic defects as well as possible future ways to counteract these defects.

1. Introduction

Compartmentalization is an essential feature of eukaryotic cells and intracellular trafficking is the process responsible for the transport of material between organelles. Indeed, membrane traffic comprehends a complex network of pathways connecting different kinds of organelles and mediating the exchange of components between them. Membrane traffic presents two main routes, the biosynthetic (or exocytic) and the endocytic route, and it is fundamental for the development and the homeostasis of all mammalian tissues. Thus, alterations of intracellular traffic often result in the development of diseases and in the last decade a number of disorders have been linked to genetic defects in intracellular transport (Olkkonen and Ikonen, 2000, 2006).

Regarding the numerous membrane traffic disorders identified to date, the genetic defects affect components of the machinery for cargo sorting and biogenesis of vesicles, components of the cytoskeletal machinery for motility of transport vesicles, or components of the machinery for tethering, docking and fusion of vesicles with their targets (Olkkonen and Ikonen, 2000, 2006). As alterations of membrane traffic events have important consequences on different key cellular processes such as signal transduction, proliferation, migration, apoptosis and mitosis, it is not surprising that mutations in membrane traffic genes often give rise to severe human disorders. In this respect, it is interesting to note that these diseases frequently affect the nervous system, possibly because this is a tissue highly sensitive to any kind of interference (Olkkonen and Ikonen, 2000, 2006; Tarabeux et al., 2010; Wang and Brown, 2010). Also, the morphology of neuronal cells, bearing long axons and thus requiring an extremely efficient and organized intracellular vesicular trafficking to transport material meters away from the cell body, could explain this sensitivity (Olkkonen and Ikonen, 2000, 2006).

It is now clear that a number of neuronal disorders are due to intracellular traffic defects or, at least, that alterations of membrane traffic are an important causative component (Morfini et al., 2009; Salinas et al., 2008; Schweitzer et al., 2009). This is not surprising if we consider that a number of neuronal processes, for instance axon growth, repair and regeneration, heavily depend on membrane traffic and, in particular, on iterative events of endocytosis and exocytosis (Bloom and Morgan, 2011). In this respect, it is worth noting that in a number of membrane traffic diseases, mutations actually affect ubiquitously expressed genes but the defect is restricted to specific cell types, for instance certain kinds of neurons (Chen et al., 2004b; Olkkonen and Ikonen, 2000, 2006). With regard to neurons, this may be explained by a greater sensitivity of neuronal cells to the altered function of the mutated protein due to their specific morphological characteristics, by the existence of specific membrane traffic pathways that would be affected, or by the existence of neuronal-specific interactors of the mutated protein whose function would be impaired.

In this review, we focus on the impact of intracellular traffic alterations on the insurgence of the most common hereditary peripheral neuropathy, Charcot–Marie–Tooth (CMT) disease. We discuss recent research regarding the cellular and molecular mechanisms underlying the different forms of the neuropathy due to alterations of intracellular traffic.

2. Membrane traffic

2.1. Steps of membrane trafficking

Intracellular membrane trafficking is the cellular process responsible for the transport of material between different cellular organelles. This process, which has to ensure efficiency, directionality, specificity and fidelity, is extremely complex and highly regulated. During the last two decades, much effort was directed towards uncovering the molecular mechanisms underlying the different steps of membrane trafficking and several advances were made allowing the identification of many components of the molecular machinery. Thus, many of the molecular mechanisms for the maintenance of organelle identity and for the transport of material between organelles are now known. For each transport event, there are at least five steps: formation of the vesicle and cargo sorting, vesicle motility, tethering of the vesicle to the target compartment, vesicle uncoating, docking and fusion of the vesicle with the target compartment (Fig. 1). The process starts with the formation of a vesicle from the donor compartment. The budding of the vesicle is mediated by a protein coat, which is responsible for shaping of the membrane and for cargo sorting. Having selected for the cargo, which also comprises proteins of the membrane traffic machinery important for the subsequent steps of transport, the vesicle pinches off the membrane. After pinching off, the vesicle moves towards the target membrane and, eventually, physically links itself (tethers), although loosely, to the target compartment and loses its coat. The vesicle then docks and fuses with the target compartment, unloading into the target compartment its soluble and membrane-bound cargo (Fig. 1).

Fig. 1.

Membrane traffic steps in a transport event. In order to achieve transport of membrane-bound and soluble molecules from one compartment to another, a vesicle buds from the donor compartment selecting, with its coat, the cargo. Pinching off the membrane is accomplished by dynamins or dynamin-related proteins. By moving on cytoskeletal tracks, the vesicle is then transported to the proximity of the target compartment and tethers to it. Following tethering, the vesicle, through the interaction of v-SNAREs and t-SNAREs, docks and fuses to the target compartment, unloading its membrane and soluble content.

2.2. The membrane traffic machinery

The membrane traffic machinery is extremely complex and there are various molecules responsible for the regulation of the different transport steps. We describe here the main components of the membrane traffic machinery, underscoring their role in membrane traffic and their involvement in human diseases.

2.2.1. Vesicle biogenesis: role of coats

Fundamental for vesicle biogenesis and cargo selection is the vesicle's coat (Bonifacino and Lippincott-Schwartz, 2003; Schekman and Orci, 1996). The coat is composed by proteins that cover the cytoplasmic side of a membrane segment, from which the vesicle will originate. Distinct coat proteins mediate different budding events and the coat is important initially to shape the transport vesicle as the addition of coats to membranes causes changes in membrane curvature (Bonifacino and Lippincott-Schwartz, 2003; Schekman and Orci, 1996). In addition, the coat selects by direct or indirect interaction the cargo molecules. The most studied coat protein is clathrin, which complexes with adaptins to form the clathrin coat (Bonifacino and Lippincott-Schwartz, 2003; Schekman and Orci, 1996). Adaptins bind to transmembrane proteins to select them to be part of the nascent vesicle and also bind to transmembrane receptors that select for soluble ligands in the vesicle (Bonifacino and Lippincott-Schwartz, 2003; Schekman and Orci, 1996). Indeed, the inner shell of the coat is formed by different adaptins that interact to form the adaptor complex, which, in turn, interacts with membrane proteins that can then also select for the soluble cargo of the vesicle. Thus, the membrane of the vesicle is highly enriched with membrane proteins selected by the adaptor complex while the lumen is enriched with molecules sorted by membrane receptors bound to the adaptor complex (Fig. 1). Apart from the clathrin coat, there are a number of other vesicle coats in cells. It is worth mentioning caveolins that bind to cholesterol and coat caveolae, flask-shaped invaginations present in the plasma membrane of several cell types (Hansen and Nichols, 2010).

After pinching off the membrane, it was believed until recently that the coat is lost from the vesicle. However, data reported in the last few years indicate that coat proteins and components of the tethering and docking machinery interact, strongly suggesting that the vesicle coat is maintained for a longer time period, and that it plays a role not only in vesicle biogenesis but also in the subsequent steps of membrane trafficking (Cai et al., 2007b; Guo et al., 2008; Wassmer et al., 2009; Zink et al., 2009).

2.2.2. Vesicle biogenesis: role of dynamins

The vesicle pinch-off is regulated by dynamin, a large GTPase of 100 kDa (Damke et al., 1994). The dynamin superfamily includes a number of different families of proteins: the classic dynamins, such as dynamin 1 (DNM1) and dynamin 2 (DNM2), the dynamin-like proteins, the Mx proteins (GTPases with antiviral activity), the optic atrophy proteins (OPA), the guanylate-binding proteins (GBP) and mitofusins (Haller and Kochs, 2002; Praefcke and McMahon, 2004). These proteins function in various cellular processes, such as cytokinesis, division of organelles and budding of vesicles, and share a common mechanism of action: dynamin rings are formed at the vesicle neck, then following GTP hydrolysis, membrane fission of vesicles from the parent membrane occurs. Membrane fission is often mediated by members of the dynamin superfamily, although each member participates in specific transport steps. In particular, classic dynamins function in the budding of clathrin-coated vesicles at the plasma membrane, cleavage furrow, Golgi, endosomes, caveolae and phagosomes. Dynamin-like proteins are involved in the division of organelles such as mitochondria and peroxisomes. OPA1 and mitofusins are responsible for the fusion of mitochondria (Chen et al., 2003; Legros et al., 2002; Praefcke and McMahon, 2004). There are also other fission events that are dynamin-independent and that require the C-terminal-binding protein/brefeldinA-ADP ribosylated substrate (CtBP/BARS) (Bonazzi et al., 2005; Weigert et al., 1999). Alterations in members of the dynamin superfamily are responsible for a number of human diseases and, in particular, cause a number of neuropathies (Reddy et al., 2011).

2.2.3. Vesicle biogenesis: role of phosphoinositides

Phosphoinositides (PIs) are important components of biological membranes. They are phospholipids that derive from the phosphorylation of phosphatidylinositol (PtdIns). Differential phosphorylation at different positions on the inositol ring leads to the formation of different PIs, and each PI has a unique localization in the cell. PI metabolism is finely regulated by kinases and phosphatases, and they serve not only for the generation of second messengers but also for membrane traffic, as they are spatio-temporal regulators specifying membrane identity (De Camilli et al., 1996; Di Paolo and De Camilli, 2006; Liu and Bankaitis, 2010). Indeed, local production of a single PI on a membrane represents the signal for the recruitment or the activation of key membrane traffic proteins that will initiate, for instance, vesicle formation (De Camilli et al., 1996; Di Paolo and De Camilli, 2006).

PIs recruit and activate specific effector proteins, which contain conserved PI-binding domains such as PH (pleckstrin homology), PX (phox homology), FYVE (Fab1p-YOTB-Vac1p-EEA1) and ENTH (epsin N-terminal homology) domains, to specific membrane locations (Balla, 2005).

In the classic PI turnover pathway, class III PI 3-kinase (vacuolar protein sorting 34 (Vps34) in yeast) phosphorylates PtdIns into PtdIns3P and type III PI 5-kinase (PIKfyve in mammals and Fab1p in yeast) phosphorylates PtdIns3P into PtdIns(3,5)P₂ (Fig. 2). Reverse reactions are catalyzed by the phosphatase FIG4 (Fig4p in yeast) which dephosphorylates PtdIns(3,5)P₂ into PtdIns3P and by the 3-phosphatase myotubularins that dephosphorylate PtdIns(3,5)P₂ into PtdIns5P and PtdIns3P into PtdIns (Nicot and Laporte, 2008) (Fig. 2).

Fig. 2.

Mechanisms of action of phosphatases involved in CMT neuropathy. FIG4 and MTMRs are phosphatases responsible for the conversion of PIs, as shown in the upper part of the figure. Phosphatidylinositol 3-phosphate (PtdIns3P) is present on the early endosomal (EE) membrane and on the intraluminal vesicles of multivesicular bodies (MVBs), whereas phosphatidylinositol 3,5-phosphate (PtdIns(3,5)P₂) is present on the limiting membrane of MVBs. The different distribution of PIs is fundamental to determine and maintain membrane identity and to guarantee the correct flux of material between the different endocytic compartments.

PIs are also important regulators of cytoskeletal dynamics, cell adhesion, cell motility and cytokinesis and thus they are involved in several human disorders (Takenawa and Itoh, 2001). In particular, small changes in PI metabolism induce neurodegeneration, have detrimental effects on the nervous system and cause developmental disorders (Skwarek and Boulianne, 2009; Wen et al., 2011).

2.2.4. Vesicle motility: role of cytoskeletal proteins

Once the vesicle has been formed, it moves within the cytosol to reach the target or acceptor compartment. The movement is mediated by the actin and tubulin cytoskeleton and, in particular, by cytoskeletal motor proteins that are able to move along cytoskeletal tracks (myosins for actin filaments, kinesins and dyneins for microtubules) and are powered by the hydrolysis of ATP (DePina and Langford, 1999; Gill et al., 1991; Hehnly and Stamnes, 2007; Holzbaur and Vallee, 1994; Sablin, 2000; Schroer et al., 1989; Urrutia et al., 1991). Indeed, the vesicle, through other components of the molecular machinery, interacts with the correct motor in order to be transported to its final destination.

The myosin superfamily includes 18 classes of motor proteins that move along actin filaments. They consist of a motor domain, a neck region and a tail region, and can be dimeric. They are important not only for intracellular membrane trafficking but also for muscle contraction, cytokinesis, cell migration and signal transduction (Foth et al., 2006; Hirokawa et al., 2010).

The dynein superfamily includes cytoplasmic dyneins and axonemal dyneins. Axonemal dyneins have a role in the bending of cilia and flagella of eukaryotic cells (Mallik and Gross, 2004; Scholey, 2003). Cytoplasmic dyneins are homodimers consisting of two heavy chains (∼520 kDa) with ATPase activity, two intermediate chains (∼74 kDa), four intermediate light chains (∼33–59 kDa) and several light chains (∼10–14 kDa) (Hirokawa et al., 2010; Karki and Holzbaur, 1999; Pfister et al., 2005). Cytoplasmic dyneins mediate the transport of different intracellular cargoes, such as mRNA, endosomes and viruses, as well as the transport within the flagellum and neurons (Aniento et al., 1993; Bremner et al., 2009; Levy and Holzbaur, 2006; Merino-Gracia et al., 2011; Pazour et al., 1999; Schnorrer et al., 2000).

The kinesin superfamily of motor proteins (KIF proteins) uses microtubules as ‘rails’ to transport cargoes. Kinesins consist of two 120-kDa heavy chains and two 64-kDa light chains organized in two globular heads, a stalk and a tail region. The globular domain (motor domain) has an ATP-binding domain that produces energy by hydrolyzing ATP for the movement along microtubules, and a microtubule-binding site. The different KIFs share high sequence homology in their motor domains, whereas the remaining parts of the molecules contain binding sites to different cargoes and are consequently relatively divergent. Depending on its specificity, the variable region may bind cargoes like mitochondria, lysosomes, endosomes, tubulin oligomers, intermediate filament proteins, mRNA complexes and other macromolecular complexes (Hirokawa et al., 2010; Hirokawa and Takemura, 2003). KIFs are therefore important for a wide variety of intracellular transport steps, including axonal and intraflagellar transport.

Microtubules are typically oriented with their ‘minus-ends’ towards the nucleus and their ‘plus-ends’ towards the cell periphery. Dynein motors mediate movements directed to the microtubule minus-end, whereas most of the kinesins move towards the plus-end (Hirokawa, 1998; Mallik and Gross, 2004). The microtubules in axons are lined up with their plus-ends towards the direction of the synapse. Axonal transport supplies essential organelles and materials to nerve endings mainly by using molecular motors like kinesins (Hirokawa et al., 2010; Hirokawa and Takemura, 2003).

Recently, the role of a third cytoskeletal component, the intermediate filaments, in membrane traffic has been investigated. Neurofilaments are the major intermediate filaments of neurons and are composed of three subunits, heavy, medium and light, consisting of an N-terminal head, an α-helical central rod and a C-terminal tail domain (Liem, 1993). It has been established that intermediate filaments have a key role not only in organelle positioning but also in organelle transport and function, indicating that they actually regulate intracellular vesicular traffic (Chang et al., 2009; Styers et al., 2005, 2004).

Alterations of the cytoskeleton and, in particular, of cytoskeletal motors cause a number of disorders and, in particular, are responsible for the defective removal of intracellular aggregates that are a common cause of neurodegeneration (Ravikumar et al., 2005; Rubinsztein et al., 2005). Intermediate filament mutations also cause a number of disorders, and as they are expressed in a tissue-specific manner, they are important in development and differentiation (Fuchs, 1994; Fuchs and Weber, 1994). In several cases, intermediate filament disorders appear to be caused by disruption of organelle positioning or signaling, as intermediate filaments may also organize signal transduction (Eriksson et al., 2009). Given the recently discovered role of intermediate filaments in membrane traffic (Chang et al., 2009; Minin and Moldaver, 2008; Styers et al., 2005), the molecular mechanism underlying a number of intermediate filament disorders could also be due to an impairment of their role in membrane traffic.

2.2.5. Vesicle tethering, docking and fusion: role of tethers and SNAREs

How does the vesicle recognize the target compartment? This is accomplished with the help of tethering proteins and of SNAREs (soluble N-ethylmaleimide-sensitive factor attachment protein receptors) (Cai et al., 2007a; Pfeffer, 1999; Söllner, 2002; Sztul and Lupashin, 2009).

Tethering is the initial attachment of the transport vesicle to the target compartment and precedes the interaction between SNAREs, present on vesicle and target membranes, that will lead to fusion. Tethering proteins or tethers are thus responsible for the initial molecular recognition between the vesicle and the target compartment. There are different kinds of tethering molecules and different tethers control the same intracellular vesicular traffic event, suggesting that they could have different roles in promoting recognition. Tethers can be long proteins with extensive coiled-coil domains and either dimers such as early endosomal antigen 1 (EEA1) or multisubunit complexes such as HOPS and the exocyst (Brown and Pfeffer, 2010; Christoforidis et al., 1999; Hickey and Wickner, 2010; Lipschutz et al., 2000; TerBush et al., 1996). A number of tethering proteins and tethering complexes have been identified and, importantly, it has been demonstrated that they are able to interact with components of the vesicle biogenesis machinery and of the fusion machinery. Indeed, it has been demonstrated that tethers interact with components of the coat, with SNAREs and with Rab proteins, being Rab effectors and Rab exchange factors (Cai et al., 2007a; Pfeffer, 1999; Söllner, 2002; Sztul and Lupashin, 2009).

After the initial loose interaction through tethers, the vesicle docks to the target compartment. Docking is a much closer interaction of the vesicle with the target membrane that is mediated by SNAREs present on the vesicle membrane (v-SNAREs) and SNAREs present on the target membrane (t-SNAREs) (Jahn and Scheller, 2006). v-SNAREs and t-SNAREs interact and bring the vesicle in close contact with the target membrane, catalyzing vesicle fusion with the help of N-ethylmaleimide-sensitive factor and of soluble NSF attachment proteins (SNAPs) (Wickner, 2010).

The dysfunction of tethers and SNAREs is involved in a number of disorders. For example, a SNARE protein redistribution has been reported in a mouse model of Parkinson's disease (Garcia-Reitböck et al., 2010) and a mutation in a SNARE protein, SNAP29, causes a neurocutaneous syndrome (Gissen et al., 2004; Sprecher et al., 2005).

2.2.6. Rab proteins and membrane traffic

Rab proteins are small GTPases important for the regulation of membrane traffic. The Rab family in human cells includes more than 60 different members involved in the regulation of all the key steps of intracellular vesicular trafficking. Indeed, Rab proteins control formation, budding, uncoating, motility, tethering, docking and fusion of vesicles, thus being coordinators of membrane traffic (Hutagalung and Novick, 2011; Stenmark, 2009). Rab proteins’ functions profoundly affect cell proliferation, cell nutrition, innate immune response, mitosis and apoptosis. The action of Rab proteins in all these cellular processes is possible through the interaction with multiple effector proteins such as molecular motors, sorting adaptors, kinases, phosphatases or tethering and fusion factors (Hutagalung and Novick, 2011; Stenmark, 2009). Rab GTPases cycle between an active GTP-bound, membrane-associated form and an inactive GDP-bound form which is mostly cytosolic. After translation, the GDP-bound Rab protein interacts with Rab escort protein (REP), which presents the protein to geranylgeranyl transferase (GGT), thus adding a geranylgeranyl group to the two C-terminal cysteines. This post-translational modification allows membrane anchoring of Rab protein. Then, on membranes, a guanine nucleotide exchange factor (GEF) stimulates Rab nucleotide exchange, causing Rab activation (Hutagalung and Novick, 2011; Stenmark, 2009). The activated GTP-bound Rab then recruits several effectors, and since any given Rab interacts with and regulates the function of different membrane traffic machinery components, it contributes to many, if not all, aspects of intracellular trafficking (Bucci and Chiariello, 2006; Grosshans et al., 2006; Hutagalung and Novick, 2011; Markgraf et al., 2007; Stenmark, 2009). Then, a GTPase-activating protein (GAP) induces Rab to hydrolyze GTP and to return to the inactive GDP-bound state. Rab in the GDP-bound state is recognized by the GDP dissociation inhibitor (GDI), which removes Rab from the membrane (Hutagalung and Novick, 2011; Stenmark, 2009).

The Rab cycle is finely regulated and it is fundamental for the proper functioning of Rab protein and, thus, for correct regulation of membrane traffic events. Any kind of interference or perturbation of the cycle alters the regulation of intracellular trafficking and may lead to diseases. Indeed, a number of disorders have been linked to Rab proteins and to their regulators. For example, choroideremia is caused by mutations in REP-1, X-linked nonspecific mental retardation is caused by mutations in GDI, Warburg Micro and Martsolf syndromes are caused by mutations in Rab3GAP and nonsyndromic autosomal recessive mental retardation is caused by mutations in TRAPPC9, a GEF for Rab1 (Aligianis et al., 2005, 2006; D’Adamo et al., 1998; Mir et al., 2009; Mochida et al., 2009; Seabra et al., 1993). Rab proteins have been implicated in many genetic and acquired disorders (such as infectious diseases and cancer) (Hutagalung and Novick, 2011; Mitra et al., 2011; Seabra et al., 2002). Interestingly, recent data indicate that dysfunction of Rab proteins is a cause of neurological diseases. For instance, Rab7 is mutated in CMT2B and Rab23 is mutated in Carpenter syndrome (Jenkins et al., 2007). Rab proteins are also implicated in Parkinson's and Huntington's disease (Dalfó et al., 2004; Gitler et al., 2008), and mutations in huntingtin protein inhibit trafficking from the trans-Golgi network (TGN) to late endosomes by interfering with Rab8 and its effector protein optineurin. In addition, huntingtin is important for nucleotide exchange and activation of Rab11, thereby impairing Rab11-regulated membrane trafficking and leading to oxidative stress and cell death (del Toro et al., 2009; Hattula and Peränen, 2000; Li et al., 2008, 2010).

2.3. The ubiquitin/proteasome system in membrane trafficking

Misfolded or abnormal proteins are normally recognized by chaperone molecules that refold them correctly. Heat shock proteins (HSPs) are molecular chaperones that prevent the formation of protein aggregates and assist proteins in the acquisition of their native structures. The name ‘heat shock protein’ refers to their increased expression in response to elevated temperatures, although other stressful conditions are also capable of inducing their expression. HSPs can be classified into two groups: the high molecular weight HSPs and the small HSP superfamily. The human genome encodes 10 different small HSPs (HSPB1–HSPB10). Members of this superfamily are characterized by low molecular mass (between 12 and 43 kDa), an α-crystallin domain consisting of 80–100 amino acids in the C-terminal region, and variable N- and C-terminal ends (Kappe et al., 2003; Koga et al., 2011). Small HSPs associate into oligomers and have a chaperone-like activity, interacting with partially denatured proteins and preventing protein misfolding and aggregation (Dierick et al., 2005; Haslbeck et al., 2005; Stromer et al., 2003). They are also involved in other cellular activities such as modulation of actin and intermediate filament dynamics, apoptosis, cellular growth and differentiation (Arrigo, 2005; Gober et al., 2003; Gusev et al., 2002; Mehlen et al., 1997; Mounier and Arrigo, 2002). Given the role of small HSPs in many cellular processes, it is not surprising that a number of diseases, including neurodegenerative disorders, are connected to mutations in these proteins (Dierick et al., 2005; Sun and MacRae, 2005).

However, when it is not possible to repair a damaged protein, chaperone molecules favor its degradation. Autophagy and the ubiquitin/proteasome system are two different processes that mediate the degradation of abnormal proteins (Koga et al., 2011). In chaperone-mediated autophagy, cytosolic proteins that need to be degraded are recognized by a chaperone and targeted to lysosomes where the chaperone binds to a membrane receptor (Cuervo and Dice, 1996; Koga et al., 2011). In conditions such as acute oxidative stress and heat shock, or when the cell's ability to refold or degrade abnormal polypeptides is exceeded, proteins aggregate (Dubois et al., 1991; Johnston et al., 1998). Aggregates can be degraded by autophagy, also referred to as aggrephagy (Yamamoto and Simonsen, 2011). Alterations of autophagy have been identified in many human neuropathies, for example Parkinson's, Alzheimer's or Huntington's disease (Cuervo et al., 2004; Wong and Cuervo, 2010).

In proteasome-mediated degradation, chaperone molecules interact with the ubiquitin/proteasome machinery, promoting the degradation of aberrant proteins. The proteasome system, a multicatalytic ATP-dependent complex, recognizes and degrades proteins that have been tagged by a small molecule, ubiquitin. The process of covalent attachment of ubiquitin to a substrate is known as ubiquitination and it is mediated by three different enzymes that work sequentially: an ubiquitin-activating enzyme E1, an ubiquitin-conjugating enzyme E2 and an ubiquitin ligase E3 (D’Azzo et al., 2005). As well as binding covalently to misfolded cytoplasmic proteins and thereby priming them for proteasome-mediated proteolysis, ubiquitin also directs membrane proteins to the endocytic pathway (Aguilar and Wendland, 2003; D’Azzo et al., 2005). The target protein can be ubiquitinated in various ways and the type of ubiquitin linkages determine the protein's fate. Ubiquitin can be attached to a single site (monoubiquitination) or to multiple sites on a substrate (multiubiquitination). In addition, ubiquitin contains seven lysine residues that can be further ubiquitinated (polyubiquitination). Monoubiquitination and K63-linkage are normally implicated in sorting and degradation in the lysosome, whereas K48-, K11-linked chains and chains of at least four ubiquitin molecules are usually a signal for proteasomal degradation (Aguilar and Wendland, 2003; Jin et al., 2008a; Raiborg and Stenmark, 2009; Thrower et al., 2000). Ubiquitination is a reversible modification and deubiquitinating enzymes (DUBs) are responsible for the disassembling of polyubiquitin chains from the substrate before its degradation, recycling ubiquitin molecules and maintaining a pool of free ubiquitin in the cell (Lee et al., 2011a; Reyes-Turcu et al., 2009). Alterations in the ubiquitin/proteasome system lead to the accumulation of protein inclusions in the cytosol and can cause neurodegenerative disorders (Bedford et al., 2008; Guthrie and Kraemer, 2011).

2.4. Mitochondrial dynamics

Mitochondria are unique organelles bounded by a double membrane. They are responsible for many essential cellular functions, for example energy production, calcium homeostasis, cell growth, development and apoptosis. They form a dynamic network whose proper distribution is important for cell survival. Their correct positioning is regulated by cytoskeletal elements, and like transport vesicles, mitochondria move along cytoskeletal tracks. This movement occurs through the interaction with molecular motors and adaptors that connect mitochondria to the cytoskeleton: kinesins and cytoplasmic dyneins mediate transport along microtubules, whereas myosins mediate transport along actin filaments (Frederick and Shaw, 2007; Hollenbeck and Saxton, 2005). Milton is a kinesin-associated protein that associates with the adaptor Miro, a Rho GTPase localized on the cytosolic side of the mitochondrial outer membrane, to mediate anterograde mitochondrial transport (Fransson et al., 2006; Glater et al., 2006; Stowers et al., 2002). Syntabulin, an adaptor for microtubule tracks that binds to the kinesin KIF5B, also promotes anterograde mitochondrial transport, while APLIP1, a kinesin-associated protein, promotes dynein-mediated retrograde movement (Cai et al., 2005; Horiuchi et al., 2005).

The transport and positioning of mitochondria in neurons are very important due to the cell-specific morphology and to the large amount of energy required at the synapses. Mitochondria are transported from the cell body to synapses using kinesins to move along microtubules (Frederick and Shaw, 2007; Tanaka et al., 2011). They can also move back to the cell body; however, the mechanisms involved in this process remain to be fully characterized (Zinsmaier et al., 2009). Disruption or alteration of these processes causes various human neuropathies (Hollenbeck and Saxton, 2005).

Mitochondria also change their morphology through fusion and division. In physiological conditions, there is a balance between fusion and fission. Defects in this balance affect organelle morphology, distribution and mobility, often contributing to the development of neurodegenerative diseases (Bossy-Wetzel et al., 2003; Suen et al., 2008; Wang and Hong, 2002). The processes of fusion and fission are not yet fully understood; however, they are known to be regulated by mitofusins and dynamin-related proteins, members of the dynamin superfamily responsible for vesicle biogenesis. Mitofusins are transmembrane GTPases located in the mitochondrial outer membrane. There are two different mitofusins in mammals: MFN1 and MFN2. They have a conserved GTPase domain in the N-terminal region, a bipartite transmembrane domain, an internal HR1 (heptad repeat domain) region and a conserved α-helical region forming a coiled-coil structure (or HR2) at the C-terminus. The transmembrane domain of mitofusins anchors the proteins to the mitochondrial outer membrane, exposing both the N- and C-ending to the cytoplasm (Koshiba et al., 2004; Rojo et al., 2002; Santel, 2006). The C-terminal coiled-coil domain of mitofusins is responsible for self-interaction of the molecules, allowing the formation of homo- and heterotypic oligomeric complexes. This interaction tethers adjacent mitochondria, initiating the fusion process that requires GTP hydrolysis (Koshiba et al., 2004). Another mitochondrial protein with a dynamin-related GTPase domain involved in the fusion process is OPA1. OPA1 resides in the intermembrane space and it is anchored to the inner mitochondrial membrane (Olichon et al., 2002). Since mitochondria are double membrane organelles, four membranes have to fuse during the fusion process: mitofusins mediate the fusion of the outer membranes, whereas OPA1 mediates the fusion of the inner membranes (Westermann, 2010). In the fission process, the outer membrane protein FIS1 recruits DRP1 (dynamin-related protein 1) from the cytosol. DRP1 contains a GTPase domain that hydrolyzes GTP, inducing membrane constriction and scission, probably by a similar mechanism to the one utilized by dynamin to pinch off vesicles (James et al., 2003; Smirnova et al., 2001).

It is well established that close relationships exist between mitochondria and other cellular organelles. For instance, the endoplasmic reticulum (ER) and mitochondria closely communicate in order to regulate a number of physiological processes such as mitochondrial energy production, lipid metabolism, apoptosis and calcium signaling. The interaction between the ER and mitochondria is mediated by mitofusins and is important for autophagosome biogenesis, since it provides membranes for the formation of this organelle during starvation (Hailey et al., 2010). In addition, there is an abundance of evidence demonstrating that the machinery responsible for the regulation of mitochondrial dynamics has several common components with the machineries responsible for intracellular vesicular trafficking. A Rab protein, Rab32, is localized to mitochondria and regulates mitochondria-associated membranes modulating apoptosis (Alto et al., 2002; Bui et al., 2010; Guan et al., 1993; Pitts et al., 1999). Rab32 is also important for the correct positioning of mitochondria in the cell (Alto et al., 2002; Bui et al., 2010).

Given the importance of mitochondrial dynamics, aberrant mitochondrial fusion, fission, movement and autophagy have been detected in a number of neurodegenerative disorders such as Parkinson's, Alzheimer's and Huntington's disease (Chen and Chan, 2009; Shirendeb et al., 2011, 2012). In addition, defects in mitochondrial axonal transport have been detected in a Drosophila model of Friedreich's ataxia and in amyotrophic lateral sclerosis (ALS) (De Vos et al., 2007; Shi et al., 2010; Shidara and Hollenbeck, 2010).

2.5. Membrane traffic in neurons

In neurons, membrane trafficking and cargo delivery are essential for the growth, remodeling and maintenance of neurites, as well as for the proper functioning of synapses. Establishment and maintenance of neuronal polarity are ensured by the cytoskeleton and membrane trafficking machinery (Foletti et al., 1999; Horton and Ehlers, 2003). As mentioned in Section 2.2.4, the cytoskeleton and molecular motors are indispensable in driving the movement of intracellular components along neurites in both directions: from and towards the cell body. Post-Golgi vesicles, recycling endosomes, late endosomes and lysosomes contribute to membrane addition and protein/receptor trafficking.

The surface area and cytoplasmic volume of neurons are 10,000 times greater than most eukaryotic cells and the length of axons can be more than one meter (Horton and Ehlers, 2003). Most of the proteins that are necessary for the maintenance and function of the axon and synaptic terminal after their synthesis in the cell body need to be transported along the axon. Energy, organelles and cargo molecules also need to travel a long distance to reach the axonal tip. At the axon terminal, synaptic vesicles containing neurotransmitters are exocytosed and endocytosed after releasing their content at cell–cell contact sites, the synapses. It is important to point out that motor and sensory neurons have generally very long axons that extend far out from the cell body and thus they have a greater need for energy, transport of organelles and molecules as well as myelin.

Intracellular transport in neurons is also important in the process of myelination. Myelin is a multilayered membrane structure generated in the CNS by oligodendrocytes and in the peripheral nervous system by Schwann cells which extend spirals of membrane around the axon of neurons. Schwann cells myelinate only one axonal segment, whereas oligodendrocytes extend several processes, myelinating various axons simultaneously (Nave, 2010). Myelin is not continuous along axons and in the discontinuous sites, termed nodes of Ranvier, the propagation of action potentials occurs. The biogenesis and maintenance of myelin require a tight control of the intracellular transport of myelin proteins and lipids (Anitei and Pfeiffer, 2006; Baron and Hoekstra, 2010; Krämer et al., 2001). Following synthesis at the ER and transport to the Golgi apparatus, various trafficking routes appear to direct myelin components to their final destination: direct transport from the TGN, indirect via endosomes, or via transcytosis (Krämer et al., 2001; Maier et al., 2008). Myelin ensures fast saltatory conduction along vertebrate axons and perturbations in myelin protein trafficking and/or turnover are associated with a number of neurological disorders (Krämer et al., 2001; Nave, 2010; Scherer and Wrabetz, 2008).

3. CMT disease

CMT disease is the most common hereditary peripheral neuropathy with a prevalence of 1:2500 (Skre, 1974). CMT disease, also classified as hereditary motor and sensory neuropathy (HMSN), is highly heterogeneous comprising a number of genetically distinct disorders that exhibit similar clinical symptoms. CMT neuropathy was first described in 1886 by Jean Martin Charcot and his student Pierre Marie in France, and by Howard Henry Tooth in Cambridge. CMT disease affects both motor and sensory nerves. Following the first identification of a duplication of the peripheral myelin protein 22 (PMP22) gene as the cause of one form of CMT disease, CMT1A (Lupski et al., 1991; Raeymaekers et al., 1991), several other genetic causes of CMT neuropathy have been discovered. More than 30 CMT disease-causative genes are now known, allowing accurate genetic diagnosis in about 70% of patients (Table 1) (Banchs et al., 2009; Berciano, 2011; Pareyson and Marchesi, 2009; Reilly et al., 2011). Although some clinical trials are in progress, no specific treatment for CMT disease currently exists and rehabilitative strategies are presently the most helpful therapies to patients (Schenone et al., 2011). A recent trial on the use of ascorbic acid in patients affected by CMT1A, based on evidence that in transgenic mice the severity of the neuropathy is reduced by this treatment, unfortunately showed no significant effect in humans (Pareyson et al., 2011; Passage et al., 2004).

Table 1.

Genetic defects associated with the different clinical forms of CMT disease and the proposed pathogenetic mechanisms.

Type	Gene/locus	Gene function	Disease mechanism
Demyelinating autosomal dominant – AD CMT1
CMT1A	PMP22	Myelination	Duplication/gene dosage/altered myelination
CMT1B	MPZ	Myelination	PM/altered myelination
CMT1C	LITAF/SIMPLE	Protein degradation	PM/altered protein degradation?
CMT1D	EGR2	Transcription of genes involved in myelination	PM/altered myelination
CMT1E	PMP22	Myelination	PM/altered myelination
CMT1F	NEFL	Cytoskeleton	PM/defective transport and assembly of neurofilaments/delayed neuroregeneration
HNPP	PMP22	Myelination	Deletion/gene dosage/altered myelination
Demyelinating autosomal recessive – AR CMT1 or CMT4
CMT4A	GDAP1	Mitochondrial dynamics	PM/altered mitochondrial distribution in axons
CMT4B1	MTMR2	Dephosphorylation of PIs	PM/reduced phosphatase activity/altered membrane recycling in neurons
CMT4B2	MTMR13	Dephosphorylation of PIs	PM/reduced phosphatase activity/altered membrane recycling in neurons
CMT4C	SH3TC2	Membrane traffic	PM/altered recycling in neurons
CMT4D	NDRG1	Membrane traffic	PM/altered membrane traffic
CMT4E	EGR2	Transcription of genes involved in myelination	PM/altered myelination
CMT4F	PRX	Maintenance of peripheral nerve myelin	PM/altered myelination/altered ensheathing of regenerating axons
CMT4G	HK1	Energy production	PM/alteration of cell survival?
CMT4H	FGD4 (Frabin)	Regulation of actin cytoskeleton	PM/abnormal cytoskeletal transport?/altered PI metabolism?
CMT4J	FIG4/SAC3	Dephosphorylation of PIs	PM/altered PI metabolism
CCFDN	CTDP1/FCP1	Dephosphorylation of RNA polymerase II	PM/aberrant splicing/altered transcription of myelin genes?
AR CMT1	PMP22	Myelination	PM/altered myelination
AR CMT1 or DSN/CH	MPZ	Myelination	PM/altered myelination
Axonal autosomal dominant – AD CMT2
CMT2A1	KIF1B	Movement on microtubules	PM/altered axonal transport
CMT2A2	MFN2	Mitochondrial dynamics	PM/altered axonal mitochondrial transport and mitochondrial dynamics
CMT2B	RAB7A	Regulation of membrane traffic	PM/altered axonal transport
CMT2C	TRPV4	Cation channel	PM/changes in calcium concentration?
CMT2D	GARS	Protein translation	PM/altered translation in the axons?
CMT2D or SS	BSCL2/Seipin	ER transmembrane protein?	PM/ER stress?
CMT2E	NEFL	Cytoskeletal transport	PM/defective transport and assembly of neurofilaments/delayed regeneration
CMT2F	HSPB1	Protein degradation	PM/altered protein degradation
CMT2G	12q12-q13 (FGD4?)	?	?
CMT2I/J	MPZ	Myelination	PM/altered myelination
CMT2H/K	GDAP1	Mitochondrial dynamics	PM/altered mitochondrial distribution in axons
CMT2L	HSPB8	Protein degradation	PM/altered protein degradation?
CMT2M	AARS	Protein translation	PM/altered translation in the axons?
Axonal autosomal recessive – AR CMT2
CMT2B1	LMNA	Nuclear architecture	PM/decreased resistance to mechanical stress in neurons?
CMT2B2	MED25	Transcription	PM/altered transcription of myelin genes?
AR CMT2	GDAP1	Mitochondrial dynamics	PM/altered mitochondrial distribution in axons
AR CMT2	LRSAM1	Protein degradation	PM/altered protein degradation?
Dominant intermediate – DI-CMT
DI-CMTA	10q24. 1-25.1	?	PM/?
DI-CMTB	DNM2	Vesicle budding	PM/alterations of membrane traffic?
DI-CMTC	YARS	Protein translation	PM/altered translation in the axons?
Slow NCV	ARHGEF10	Regulation of actin cytoskeleton	PM/myelin defects during development
X-linked CMT or CMT5
CMTX1	GJB1	Gap junctions	PM/impaired interactions between glia and neurons/myelination defects
CMTX2	Xp22.2	?	?
CMTX3	Xq26	?	?
CMTX4	Xq24-q26.1	?	?
CMTX5	PRPS1	Nucleotide biosynthesis	PM/reduced nucleotide availability/myelination or G-protein defects?

AARS, alanyl-tRNA synthetase; ARHGEF10, Rho guanine nucleotide exchange factor 10; BSCL2, Berardinelli-Seip congenital lipodystrophy 2; CCFDN, congenital cataracts facial dysmorphism neuropathy; CH, congenital hypomyelination; CTDP1, C-terminal domain phosphatase 1; DNM2, dynamin 2; DSN, Déjèrine–Sottas neuropathy; EGR2, early growth response protein 2; FCP1, TFIIF-associating RNA polymerase C-terminal domain phosphatase 1; FGD4, FYVE, RhoGEF and PH domain-containing protein 4; GARS, glycyl-tRNA synthetase; GDAP1, ganglioside-induced differentiation-associated protein 1; GJB1, Gap junction protein β1; HK1, hexokinase 1; HNPP, Hereditary Neuropathy with liability to Pressure Palsies; HSPB1, heat shock protein beta-1; HSPB8, heat shock protein beta-8; KIF1B, kinesin family member 1β; LITAF, lipopolysaccharide-induced TNF factor; LMNA, lamin A/C; LRSAM1, leucine rich repeat and sterile alpha motif 1; MED25, Mediator complex subunit 25; MFN2, mitofusin 2; MTMR, myotubularin-related protein; MPZ, myelin protein zero; NCV, nerve conduction velocity; NDRG1, N-myc downstream regulated gene 1; NEFL, neurofilament light chain; PM, point mutation; PMP22, peripheral myelin protein 22; PRPS1, phosphoribosylpyrophosphate synthetase 1; PRX, periaxin; SAC3, SAC domain containing protein 3; SH3TC2, SH3 domain and tetratricopeptide repeats-containing protein 2; SIMPLE, small integral membrane protein of lysosome/late endosome; SS, Silver syndrome; TRPV4, transient receptor potential cation channel subfamily V member 4; YARS, tyrosyl-tRNA synthetase; ?, unknown.

3.1. Clinical features of CMT disease

The age of onset of CMT disease is within the first or second decade of life (although it has been reported to be as late as the seventh decade), with no race predilection. Lifespan is not affected, although CMT neuropathy is characterized by slowly progressive weakness of the distal muscles that can lead to muscle atrophy (Barisic et al., 2008; Pareyson et al., 2006; Skre, 1974). The weakness usually starts in the legs and feet, then subsequently affects hands and arms. Patients first experience difficulties in walking as the disorder affects lower leg muscles first. If foot muscles are heavily affected, this also leads to foot deformities such as ‘pes cavus’ with high arches and hammertoes (where the middle joint of a toe bends upwards) (Barisic et al., 2008; Pareyson et al., 2006; Skre, 1974). Other features are decreased or absent tendon reflexes, the inability to hold the foot horizontal (foot drop) and a high-stepped gait with frequent tripping and falls. If the weakness also affects the hands and arms, this results in hand deformities with poor finger control and increasing difficulties in writing and manipulating small objects (Barisic et al., 2008; Pareyson et al., 2006; Skre, 1974). The different extents of sensory loss that accompany the various forms of the disorder are generally more serious distally, with numbness at the feet or legs frequently recorded. Some forms of the disease are termed ulcero-mutilating as they are characterized, in addition to prominent sensory loss, by frequent toe and foot ulcers, with recurrent infections often leading to amputations (Barisic et al., 2008; Pareyson et al., 2006; Skre, 1974). Thus, complications of CMT neuropathy are represented by progressive weakness, progressive inability to walk and manipulate objects, and frequent injuries in areas of the body displaying decreased sensation. Other associated clinical symptoms are deafness, hand tremors, diaphragm palsy, vocal cord palsy, pyramidal signs, papillary abnormalities, optical nerve atrophy, mental retardation and renal failure (Barisic et al., 2008; Patzkó and Shy, 2011; Schenone et al., 2011).

3.2. Classification of the different forms of CMT disease: axonal versus demyelinating

CMT disease includes several clinically and genetically distinct disorders that have been classified mostly according to nerve conduction velocities. In 1968, Dyck and Lambert started to classify neuronal peripheral disorders as HMSN and divided them into two groups: type 1 with low nerve conduction velocities and type 2 with normal or near-normal nerve conduction velocities (Dyck and Lambert, 1968). In 1980, Harding and Thomas, studying 228 HMSN patients, noted that nerve conduction velocities exhibited a bimodal distribution and thus decided to set a threshold of 38 m/s to separate patients into the two categories (Harding and Thomas, 1980). Thus, on the basis of electrophysiological properties and neuropathology, CMT neuropathy has been divided into two main types: CMT1 (or HMSN type I) with nerve conduction velocities below 38 m/s, caused by abnormalities in the myelin sheath and called demyelinating, and CMT2 (or HMSN type II) with nerve conduction velocities greater than 38 m/s, caused by abnormalities in the axon and thus called axonal. In demyelinating CMT forms, the defect generally affects Schwann cells first and, as a consequence, causes axonal loss. Indeed, in the peripheral nervous system, Schwann cells tightly communicate with axons in order to regulate their development, function and maintenance (Corfas et al., 2004). Failing of this interaction due to damaged Schwann cells results in axonal damage and degeneration (as in demyelinating CMT) and axonal neurofilaments become more dense; it has been proposed that denser packing of neurofilament is the cause of axonal injury although this has not been proved yet (de Waegh and Brady, 1990; de Waegh et al., 1992). In axonal CMT forms, it is axonal transport that is affected, subsequently causing degeneration of the axon. Biopsies of sural nerves from patients affected by demyelinating CMT forms show segmental demyelination, whereas the same type of biopsies from patients with the axonal form present axonal loss but not demyelination (Szigeti and Lupski, 2009). However sometimes the electrophysiological distinction between demyelinating and axonal forms may be difficult. Indeed in the neuropathies with primary involvement of myelin or Schwann cells, secondary axonal degeneration can occur, while in primarily axonal neuropathies, secondary demyelination may also be present (Hanemann and Gabreels-Festen, 2002; Krajewski et al., 2000; Tankisi et al., 2007). Although this classification is still used, the different CMT forms are classified considering not only the electrophysiological and anatomical findings, but also considering inheritance genetic patterns and the causative mutant genes (Reilly, 2007; Reilly et al., 2011). Consequently, an increased number of CMT forms has been identified, including the demyelinating autosomal dominant CMT1 (AD CMT1) form, the demyelinating autosomal recessive CMT1 (AR CMT1 or CMT4) form, the axonal autosomal dominant CMT2 (AD CMT2) form, the axonal autosomal recessive (AR CMT2) form and the X-linked (CMTX or CMT5) form. In addition, dominant intermediate (DI-CMT) forms of the disease with intermediate median motor nerve conduction velocities have also been described. A further division of each type into subtypes depends on the genetic defect (Table 1) (Reilly, 2007; Reilly et al., 2011).

Mutated proteins in neurons are generally the cause of both dominantly and recessively inherited forms of axonal CMT. In these cases, mutations have a cell-autonomous effect, i.e. mutant cells exhibit an altered phenotype independently of mutations in other cell types which interact with the affected neurons. Dominantly and recessively inherited forms of demyelinating CMT are due to a cell-autonomous effect in Schwann cells, where only mutated proteins expressed in Schwann cells are responsible for the demyelination. However, the expression of the mutant proteins by other cell types in these cases can contribute to the altered phenotype in peripheral neuropathies and CMT disease in a non-cell-autonomous manner (Suter and Scherer, 2003). In a non-cell-autonomous disorder, mutations affect cells in the proximity of the target neurons (such as glial cells, astrocytes, oligodendrocytes and microglia), eventually leading the target cells to exhibit a mutant phenotype (Boillee et al., 2006; Custer et al., 2006; Di Giorgio et al., 2007; Nagai et al., 2007; Yazawa et al., 2005). For example, mutations in proteins expressed in glial cells can either cause the release of toxic components or alter neuronal support functions, leading to damages to the neighboring neurons (Lobsiger and Cleveland, 2007).

4. Genetic causes of CMT disease

Since the initial discovery of PMP22 as the causative gene for CMT1A (Matsunami et al., 1992; Patel et al., 1992; Timmerman et al., 1992; Valentijn et al., 1992), more than 30 genes have been linked to CMT neuropathy, and for several loci associated with different forms of the disease, the identification of the responsible gene is still in progress. In this review, we decided to group the identified CMT disease-causative genes on the basis of their function in order to obtain a clearer idea of the multiple processes that, if altered, cause the disorder. Of course, the optimum way of doing this would be to classify the genes on the basis of the altered function that gives rise to the disorder. However, as genes usually have more than one function or influence more than one cellular process, and for many of the CMT disease-causative genes the molecular mechanism leading to the disorder is still not known, this was not always possible. Therefore, upon consideration of all the proven functions of each gene, we have grouped the genes based on the altered function most likely to be responsible to lead to the neuropathy. We believe that this classification is a useful starting point to fully understand the molecular causes of the different forms of the disorder. We list all the genes, pathways and processes involved in the pathogenesis of CMT disease in order to provide a comprehensive representation of all the genetic defects involved. In the following sections, however, we focus on the alterations of membrane traffic genes and discuss in detail their putative mechanisms of action.

4.1. Defects in vesicle budding: DNM2

DNM2 is a large GTPase which is ubiquitously expressed (Diatloff-Zito et al., 1995). It consists of a GTPase domain at the N-terminus, a middle domain (MD), a PH domain, a GTPase effector domain (GED) and a proline-rich domain (PRD) at the C-terminus. The catalytic GTPase domain binds and hydrolyzes GTP in order to deform membranes; the MD binds to γ-tubulin and is responsible for DNM2 self-assembly (Durieux et al., 2010; Thompson et al., 2004). The PH domain is involved in the interaction with membrane PIs, especially phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P₂), targeting DNM2 to membranes (Klein et al., 1998; Zheng et al., 1996). The GED participates in the self-assembly of DNM2 and acts as a GAP (Sever et al., 1999). The PRD is the binding site for proteins that interact with dynamin via Src-homology 3 (SH3) domains (Dong et al., 2000; Soulet et al., 2005). DNM2 is subjected to several post-translational modifications. It can be phosphorylated by Src kinase, leading to albumin endocytosis and membrane vesiculation at the TGN, or dephosphorylated, leading to dopamine-induced Na⁺K⁺-ATPase endocytosis (Efendiev et al., 2002; Shajahan et al., 2004; Weller et al., 2010). DNM2 can undergo S-nitrosylation by nitric oxide, with subsequent increase of its GTPase activity and endocytosis (Kang-Decker et al., 2007). Finally, cathepsin L cleaves cytoplasmic DNM2 at positions 355–360 in the MD (Sever et al., 2007).

DNM2 is mainly involved in membrane trafficking and in the formation and release of vesicles from membranes (Fig. 1). It participates in clathrin-dependent and clathrin-independent endocytosis, in intracellular trafficking from endosomes and the Golgi apparatus and regulates both the actin and tubulin cytoskeleton not just in connection with membrane trafficking processes (Durieux et al., 2010). DNM2 co-localizes with clathrin-coated vesicles and plays a role during their maturation (Loerke et al., 2009; Rappoport and Simon, 2003; Warnock et al., 1997). It forms a complex with sorting nexin 9 (SNX9) and fructose-1,6-bisphosphate aldolase. Phosphorylation of SNX9 induces the release of aldolase from the SNX9–DNM2 complex, recruiting DNM2 to the plasma membrane (Lundmark and Carlsson, 2004). DNM2 binds to PtdIns(4,5)P₂ and proteins containing BAR domains (such as amphiphysins and SNX9) on the membrane and forms an oligomer helical structure around the neck of the nascent vesicles (Shin et al., 2008b; Warnock et al., 1997). Finally, SNX9 promotes dynamin GTPase activity, and GTP hydrolysis results in membrane constriction and vesicle scission (Lundmark and Carlsson, 2005; Soulet et al., 2005). DNM2 is also involved in clathrin-independent endocytosis, like micropinocytosis and macropinocytosis, as well as in the formation of phagosomes and caveolae (Cao et al., 2007; Gold et al., 1999; Henley et al., 1998; Liu et al., 2008; Predescu et al., 2003). In addition, DNM2 localizes to the TGN where it associates with cortactin and syndapin 2 and regulates vesicle formation (Cao et al., 2005; Jones et al., 1998; Kessels et al., 2006; Kreitzer et al., 2000; Maier et al., 1996). Recent studies have also demonstrated an involvement of DNM2 in exocytosis, even though its role in the exocytic machinery is not clear (Arneson et al., 2008; Jaiswal et al., 2009; Min et al., 2007). DNM2 interacts with and is involved in the regulation of actin and microtubule networks. It interacts with Abp1 (actin-binding protein 1), a Src kinase that physically links the endocytosis machinery to the cortical actin network, and cortactin, a component of the cortical actin cytoskeleton that regulates actin polymerization (Kessels et al., 2001; Mooren et al., 2009; Schafer et al., 2002). Interaction between DNM2 and the actin cytoskeleton is important not only for endocytosis and vesicle formation, but also for the formation of membrane tubules and protrusions and for cell migration. Indeed DNM2, together with cortactin and Arp2/3, reorganizes actin at the edge of migrating cells, allowing lamellipodia formation (Krueger et al., 2003). Additionally, it is present in other structures important for cell migration, such as cortical ruffles, podosomes and invadopodia, as well as in focal adhesions and actin-stress fibers (Baldassarre et al., 2003; Ezratty et al., 2005; McNiven et al., 2000; Ochoa et al., 2000; Schlunck et al., 2004; Yoo et al., 2005). The PRD region of DNM2 interacts with microtubules, regulating their polymerization–depolymerization (Hamao et al., 2009; Lin et al., 1997; Tanabe and Takei, 2009; Warnock et al., 1997). Moreover, DNM2 binds to γ-tubulin and it has been described as a component of the centrosome (Thompson et al., 2004). It has also been reported that DNM2 participates in all the phases of mitosis, suggesting a role in the regulation of many different microtubule-dependent processes (Liu et al., 2008; Thompson et al., 2002). Furthermore, DNM2 is capable of triggering apoptosis and the GTPase domain of DNM2 is important in this function (Fish et al., 2000; Soulet et al., 2006).

Four isoforms are expressed by the DNM2 gene using a combination of two alternative splice sites. Each DNM2 isoform appears to perform specific functions: data suggest that isoforms 1 and 3 are preferentially involved in clathrin and caveolae-dependent endocytosis, whereas isoforms 2 and 4 are preferentially involved in uncoated endocytosis and trafficking from the Golgi (Durieux et al., 2010; Liu et al., 2008). The DNM2 gene has recently been described as a susceptibility gene for late-onset Alzheimer's disease (Aidaralieva et al., 2008). Furthermore, mutations in the DNM2 gene cause rare forms of DI-CMT type B (DI-CMTB) peripheral neuropathy and autosomal dominant centronuclear myopathy (CNM) (Bitoun et al., 2005, 2008; Fabrizi et al., 2007b; Gallardo et al., 2008; Zuchner et al., 2005). In DI-CMTB, five different DNM2 mutations have been identified in the N-terminal region of the PH domain and one in the MD (Bitoun et al., 2008; Fabrizi et al., 2007b; Gallardo et al., 2008; Zuchner et al., 2005). Due to the ubiquitous expression of DNM2, DNM2 mutations appear to affect both Schwann cells and neurons (Niemann et al., 2006).

To date, the mechanisms involved in the pathophysiology of disorders caused by DNM2 mutations are unknown, even though many reports suggest that an impairment in membrane trafficking contributes to the pathogenesis of DI-CMTB. Impairment of clathrin-mediated endocytosis has been demonstrated in cultured cells expressing CNM or CMT-DNM2 mutants, and one of the DI-CMTB mutants was shown to alter the intracellular trafficking of the transferrin-containing compartment (Bitoun et al., 2009; Tanabe and Takei, 2009). In addition, DI-CMTB mutants can disorganize the microtubule cytoskeleton, and one of them has been shown to impair microtubule-dependent membrane transport (Tanabe and Takei, 2009; Zuchner et al., 2005). This is consistent with observations that suggest neuropathies, including CMT neuropathy, are caused by defects in membrane transport steps such as endocytosis, axonal transport, or protein degradation (Suter and Scherer, 2003). In DI-CMTB, DNM2 mutations that alter the microtubule network may lead to abnormal axonal transport and protein trafficking, a pathophysiological mechanism previously described in various forms of CMT disease. Interestingly, DNM2 has recently been found on late endosomes in a complex with CIN85 and Rab7, and Rab7 mutations are also responsible for a form of CMT neuropathy (Schroeder et al., 2010; Verhoeven et al., 2003a).

The fact that DNM2 is involved in a wide variety of functions and interacts with various proteins makes the identification of the pathogenetic mechanisms in DI-CMTB difficult. Additionally, the phenotype of DI-CMTB patients could be due to impairment of the various functions of the protein. How DNM2 mutations alter cell function in tissue-specific disorders is currently an important issue that awaits to be resolved. However, it is interesting to note that inhibition of DNM1 expression, as well as inhibition of expression of its interacting partner amphiphysin 1, prevents neurite formation in cultured hippocampal neurons, indicating that DNM1 function is required for normal neuronal morphogenesis (Mundigl et al., 1998; Torre et al., 1994). Thus, the role of DNM2 in neurite outgrowth and neuritogenic signaling and the effects of expression of DNM2 mutant proteins causing CMT disease on these processes should be investigated. The findings of such studies could contribute to establishing the mechanism of CMT neuropathy.

4.2. Defects in PI metabolism: MTMRs and FIG4

4.2.1. MTMRs

Myotubularin-related proteins (MTMRs) are a family of ubiquitously expressed PI 3-phosphatases consisting of catalytically active or inactive members in humans. Active MTMRs possess 3-phosphatase activity towards both PtdIns3P and PtdIns(3,5)P₂ polyphosphoinositides, suggesting an involvement in intracellular trafficking and membrane homeostasis (Fig. 2) (Robinson and Dixon, 2006). PtdIns3P is produced by a class III PI 3-kinase in mammals, which corresponds to the Vps34 protein in yeast, and is important for endosome function. PtdIns3P is highly enriched on early endosomes and on the internal vesicles of multivesicular bodies (MVBs), and is involved in membrane transport (Lindmo and Stenmark, 2006; Schu et al., 1993). It recruits effector proteins containing FYVE, PX or PH motifs such as EEA1 that cooperates with the activated Rab5 GTPase to regulate early endosomal fusion and hepatocyte growth factor-regulated tyrosine kinase substrate (Hrs) that controls the first steps of receptor sorting and internalization within the MVBs (Gillooly et al., 2000; Raiborg et al., 2001; Roth, 2004; Simonsen et al., 1998). The other MTMR substrate, PtdIns(3,5)P₂, is generated by the phosphorylation of PtdIns3P by PIKfyve in mammals and by Fab1p in Saccharomyces cerevisiae (Gary et al., 1998; Sbrissa et al., 1999). PIKfyve is localized on early endosomes and regulates retrograde transport; however, PtdIns(3,5)P₂ subcellular localization and function have not been fully elucidated (Cabezas et al., 2006; De Lartigue et al., 2009; Ikonomov et al., 2003; Rutherford et al., 2006). PtdIns(3,5)P₂ is present at a very low abundance in cells and it was originally discovered in yeast, where its levels increase in response to hyperosmotic shock, leading to a reduction in vacuole size (Dove et al., 1997). PtdIns(3,5)P₂ is implicated in a number of cellular processes, including control of the size and acidification of endosomes and lysosomes, regulation of retrograde membrane trafficking from lysosomes and late endosomes to the Golgi complex and ubiquitin-dependent sorting of some cargo proteins into MVBs (Dove et al., 1997, 2004; Efe et al., 2005; Michell et al., 2006; Mollapour et al., 2006; Odorizzi et al., 1998; Rudge et al., 2004; Rusten et al., 2006; Shisheva, 2008). Since MTMR substrates function within the endosomal–lysosomal pathway, it is not surprising that MTMRs themselves also play a role in endocytosis and trafficking of membranes and proteins (Naughtin et al., 2010; Tsujita et al., 2004).

Among the MTMRs, MTMR2 is a catalytically active phosphatase of 643 amino acids with a PH-GRAM (pleckstrin homology, glucosyltransferase, Rab-like GTPase activator and myotubularin) domain, which binds PIs; a PTP (protein tyrosine phosphatase) domain; a coiled-coil domain for homo- and heterodimerization with other members of the family, such as MTMR13; and a PSD-95-Dlg-ZO-1-binding domain (PDZ-BD) at the C-terminus (Begley et al., 2003; Robinson and Dixon, 2005). Many studies have shown that MTMR2 localization is mainly cytosolic, with enrichment in the perinuclear region (Berger et al., 2003; Franklin et al., 2011; Laporte et al., 2002; Previtali et al., 2003; Robinson and Dixon, 2005). However, it has also been shown that both overexpressed and endogenous MTMR2 significantly co-localizes with the late endocytic protein Rab7, and binds to the hVps34/hVps15 complex, suggesting that MTMR2 functions within the endocytic pathway regulating PtdIns3P levels (Cao et al., 2008). Under certain conditions, such as changes in the levels of intracellular PIs, dephosphorylation and/or after interaction with an inactive partner like MTMR5, MTMR2 localizes to precise subcellular compartments (Kim et al., 2003). Indeed, unphosphorylated MTMR2 localizes to endocytic compartments where it dephosphorylates PtdIns3P, and when hypoosmotic stress is induced in COS7 cells (a condition that increases PtdIns(3,5)P₂ levels), MTMR2 relocalizes to the membranes of intracellular vacuoles formed under this condition (Berger et al., 2003; Franklin et al., 2011).

MTMR13 is a catalytically inactive phosphatase of 1849 amino acids with a PH-GRAM domain, a PTP domain with substitutions in the Cys and Arg residues of the Cys-X₅-Arg site for catalytic activity, and a coiled-coil domain and PDZ-BD as in MTMR2. MTMR13 also possesses a DENN (differentially expressed in neoplastic versus normal cells) domain at the N-terminus and a canonical PH domain at the C-terminus (Azzedine et al., 2003). Recent studies have revealed that DENN domains interact directly with Rab proteins and that they are Rab-specific GEFs, thus regulating Rab function (Allaire et al., 2010; Marat et al., 2011; Marat and McPherson, 2010; Niwa et al., 2008; Yoshimura et al., 2010). MTMR13 has GEF activity towards Rab28 (Yoshimura et al., 2010). Rab28 is a poorly characterized small GTPase that co-localizes with endosomal sorting complex required for transport I (ESCRT-I) in the unicellular organism Trypanosoma brucei, where it may play a role in the turnover of ubiquitinated endocytosed proteins and in the lysosomal delivery of cargo, suggesting that MTMR13 could also be involved in endocytic traffic (Lumb and Field, 2010). Interestingly, MTMR2 and MTMR13 interact and form heterotetramers by the association of homodimers, consisting of two MTMR2 and two MTMR13 molecules (Berger et al., 2006a; Bolis et al., 2007; Robinson and Dixon, 2005). MTMR13 is thought to regulate both MTMR2 subcellular localization and phosphatase activity, increasing the enzymatic activity of MTMR2 towards PtdIns3P and PtdIns(3,5)P₂ (Berger et al., 2006a).

Although ubiquitously expressed, mutations in either MTMR2 or MTMR13 cause CMT type 4B1 and 4B2 neuropathy, respectively (Azzedine et al., 2003; Bolino et al., 2000; Senderek et al., 2003b). CMT4B1-associated mutations are MTMR2 loss-of-function mutations: base-pair insertions or deletions that cause frameshifts, missense mutations that create stop codons or amino acid changes (Bolino et al., 2000; Previtali et al., 2007). Most of the identified MTMR2 mutations affect the PTP domain, resulting in loss of enzymatic activity (Berger et al., 2002; Parman et al., 2004; Previtali et al., 2007). Despite the ubiquitous expression of MTMR2, Schwann-cell-specific depletion of MTMR2 is sufficient for the development of myelin outfolding in transgenic mice (Bolino et al., 2004; Bolis et al., 2005; Bonneick et al., 2005). Since dysmyelination and axonopathy are not observed in the motor-neuron-null mouse but loss of MTMR2 phosphatase activity in Schwann cells is sufficient and necessary to cause myelin outfolding as in CMT4B1, MTMR2 has a Schwann-cell-autonomous role (Bolis et al., 2005). However, because MTMR2 is expressed at high levels by peripheral neurons and axonopathies are not easily reproduced in mice, a potential cell-autonomous role of MTMR2 in neurons too cannot be excluded (Berger et al., 2002; Suter and Scherer, 2003). CMT4B2 results instead from the loss of the DENN domain of MTMR13 (Senderek et al., 2003b). The precise mechanisms by which MTMR mutations lead to CMT neuropathy are not known; however, it is worth noting that the DENN domain of MTMR13 interacts with Rab proteins, implying a link between CMT disease mutations and alterations in endosomal trafficking.

Although the role of MTMR2 and MTMR13 in intracellular trafficking and the exact nature of the intracellular compartments to which they are associated remain to be assessed, several of their interacting partners have been identified. MTMR2 associates with MTMR13 (Robinson and Dixon, 2005), and with the neurofilament light chain (NEFL) in neurons and Schwann cells (Previtali et al., 2003). Interestingly, mutations in the gene encoding the NEFL protein also cause CMT disease (Abe et al., 2009; Jordanova et al., 2003; Shin et al., 2008a; Yum et al., 2009). MTMR2 is ubiquitously expressed, whereas NEFL is specifically expressed in the nervous system. Thus, the role of the MTMR2/NEFL interaction in neurons and/or Schwann cells may explain why mutations in the ubiquitously expressed MTMR2 specifically affect the nerves (Previtali et al., 2003). Interactions between MTMR2 and Dlg1 (discs large 1, also known as SAP97, synapse-associated protein 97) or PSD-95, two members of the membrane-associated guanylate kinase (MAGUK) family, have also been reported (Bolino et al., 2004; Lee et al., 2010). Dlg1 is a scaffolding protein which links transmembrane proteins with the intracellular cytoskeleton. It is located at adherens junctions of epithelial cells and at pre- and postsynaptic sites in neurons (Fujita and Kurachi, 2000). Dlg1 interacts with kinesin 13B (kif13B) and the Sec8 exocyst complex component in Schwann cells (Bolis et al., 2009). Kif13B is a plus-end motor protein that transports PtdIns3P-containing vesicles along microtubules in neurons, whereas the exocyst is an octameric protein complex that tethers secretory vesicles at the plasma membrane for exocytosis (He and Guo, 2009; Horiguchi et al., 2006). A model has been proposed whereby kif13B transports Dlg1 to sites of membrane remodeling to control and balance myelination: interaction of Dlg1 with Sec8 would promote membrane addition, whilst interaction of Dlg1 with the phosphatase MTMR2 would negatively regulate membrane formation (Bolino et al., 2004; Bolis et al., 2009). Therefore, loss of the MTMR2/Dlg1 interaction in Schwann cells may impair membrane homeostasis, leading to myelin defects (Bolino et al., 2004).

4.2.2. FIG4

FIG4 is a PtdIns(3,5)P₂ 5-phosphatase that dephosphorylates PtdIns(3,5)P₂ at the 5th position of the inositol ring converting it to PtdIns3P (Fig. 2) (Gary et al., 2002; Marks, 2008; Rudge et al., 2004).

It contains a polyphosphoinositide phosphatase domain called the Sac1 domain (Gary et al., 2002). Fig4p (Fig4 in yeast) is required to activate Fab1p (the PtdIns3P 5-kinase that generates PtdIns(3,5)P₂) and it physically associates with Fab1p and Vac14p in a protein complex that regulates the overall concentration of PtdIns(3,5)P₂. This complex, that also includes Vac7p and Atg18p, works at the vacuole membrane (the yeast vacuole corresponds to late endosomes/lysosomes in mammalian cells) to sustain both basal and hyperosmotic shock-induced PtdIns(3,5)P₂ synthesis (Botelho et al., 2008; Jin et al., 2008b; Michell and Dove, 2009; Sbrissa et al., 2007). Vac14p (also known as ArPIKfyve in mammals) and Vac7p are Fab1p activators which regulate PtdIns(3,5)P₂ synthesis and turnover (Bonangelino et al., 2002; Duex et al., 2006; Efe et al., 2007). Interestingly, mutation in the VAC14 mouse gene induces neurodegeneration and neurological defects similar to those produced by the lack of FIG4 (Chow et al., 2007; Ferguson et al., 2009). In addition, deletion of FAB1, VAC14 or VAC7, as well as yeast cells lacking Atg18p, show enlarged vacuoles and acidification defects similar to FIG4 deletion (Bonangelino et al., 2002; Dove et al., 2002; Efe et al., 2007; Gary et al., 2002; Rudge et al., 2004). In yeast, the transport of membrane proteins from the vacuole is dependent on PtdIns(3,5)P₂. Therefore, the enlarged vacuole could originate from a failure in membrane recycling from the vacuole to the pre-vacuolar endosomes (Dove et al., 2004). Concordantly, PIKfyve (the equivalent of Fab1p in mammals) regulates endosome-to-TGN retrograde transport in mammalian cells (Ikonomov et al., 2003; Rutherford et al., 2006).

Chow et al. found a spontaneous mutation of the FIG4 gene in the mouse that is responsible for the ‘pale tremor’ phenotype, so named because of the light pigmentation and the severe tremor and abnormal gait (Chow et al., 2007). As in the yeast studies, FIG4-deficient mice show reduced levels of PtdIns(3,5)P₂ and enlarged late endosomes–lysosomes. The observation of the ‘pale tremor mouse’ phenotype allowed the identification of FIG4 mutations in patients with CMT neuropathy. Indeed, many symptoms in the mutant mice, such as neuronal degeneration in the central and peripheral nervous systems, large myelinated axons in the sciatic nerve and large vacuole accumulation within neurons that precedes neuronal loss, are similar to those of humans with CMT4J disease (Chow et al., 2007; de Leeuw, 2008). But how do mutations in FIG4 cause neurodegeneration? The endosome-lysosome system performs a key role in membrane and protein homeostasis and controls their degradation. A correct trafficking is essential for cell survival. Alterations in this process particularly affect neurons with long axons (Baloh, 2008). Abnormal transport of intracellular organelles has been observed by time-lapse imaging in fibroblasts from CMT4J patients, suggesting that a defective trafficking of intracellular organelles due to obstruction by endosomes could be a potential mechanism causing this disorder (Zhang et al., 2008a). FIG4 has a role in the regulation of PtdIns(3,5)P₂ and proper PtdIns(3,5)P₂ levels are essential for retrograde membrane trafficking from lysosomes and late endosomes (Michell and Dove, 2009; Rutherford et al., 2006). Alteration in PtdIns(3,5)P₂ levels due to FIG4 mutations could therefore inhibit this recycling, leading to the accumulation of large endosomes (Zhang et al., 2008a). In addition, it has been observed that FIG4 exists in complex with ArPIKfyve in mammalian cells and that ArPIKfyve knockdown significantly reduces FIG4 protein levels, suggesting a role for ArPIKfyve in the attenuation of FIG4 proteasome-dependent degradation (Ikonomov et al., 2009, 2010; Sbrissa et al., 2007). It has been hypothesized that, when associated with ArPIKfyve, FIG4 is protected from degradation, whereas when it is in isolation, FIG4 is unfolded and therefore easily degraded (Ikonomov et al., 2010). A mutation at position 41 in FIG4 present in CMT4J patients appears to be responsible for the loss of ArPIKfyve protective activity against degradation. The resulting rapid degradation of FIG4 in these patients might alter PtdIns(3,5)P₂ homeostasis, thus causing defects in membrane trafficking (Chow et al., 2007; Ikonomov et al., 2010).

Importantly, alterations in PI signaling and vesicle trafficking have been implicated in other forms of CMT disease. Altered levels of PtdIns(3,5)P₂ and PtdIns3P, due to mutations in MTMR2 and MTMR13 which dephosphorylate PtdIns3P and PtdIns(3,5)P₂ at the 3rd position of inositol, cause CMT4B1 and CMT4B2, respectively, characterized by excessive myelin outfolding probably due to a defective transport towards myelin sheaths (Nicot and Laporte, 2008). It has recently been demonstrated that loss of FIG4 rescues myelin outfolding caused by MTMR2 deficiency in Schwann cells as well as neurons, probably balancing the increase of PtdIns(3,5)P₂ in Mtmr2-null cells and thus reducing myelin outfolding (Vaccari et al., 2011). The presence of cytoplasmic inclusions in Schwann cells and the typical demyelinating features of CMT4J patients support a Schwann-cell-autonomous role for FIG4 (Scherer and Wrabetz, 2008; Vaccari et al., 2011; Zhang et al., 2008a).

Neurodegeneration may be related not only to alterations in the delivery of membrane components from endosomes but also to abnormal accumulation of proteins due to impaired degradation. Indeed, neurons are long-lived, terminally differentiated cells, and accumulation of components that should be degraded easily causes damage in these cells that cannot renew themselves. It is known that alterations in the endosomal/degradative pathway lead to neurodegeneration in other disorders such as Niemann–Pick type C disorder (Karten et al., 2009). In addition, the late endosomal–lysosomal system has a role in autophagy, the process by which cells degrade their own components. Autophagy is altered in many human diseases, including several neurodegenerative disorders where the accumulation of misfolded proteins is a result from defects in autophagy (Huang and Klionsky, 2007; Martinez-Vicente and Cuervo, 2007). Vps34 and its product PtdIns3P are involved in the control of autophagic vesicles, and PtdIns(3,5)P₂ has an essential role in autophagy in the mammalian nervous system (Ferguson et al., 2009; Petiot et al., 2000; Simonsen and Tooze, 2009). Interestingly, in mice with mutations in FIG4 and VAC14 that cause severe neurodegeneration, a reduced number of myelinated axons in the sciatic nerve and loss of neurons are observed, and autophagy intermediates accumulate in the brain and spinal cord. Enlarged late endosomes/lysosomes are present in the cytoplasm of cultured fibroblasts and neurons from these mice and autophagy appears to be blocked. Concordantly, Ferguson et al. have recently proposed that CMT4J represents a type of autophagy-related disease caused by mutations in the autophagy machinery itself (Ferguson et al., 2009, 2010).

4.3. Defects in cytoskeletal transport: KIF1B, NEFL and FGD4/Frabin

4.3.1. KIF1B

KIF1Bβ is a plus-end-directed motor that transports synaptic vesicle precursors in the axon from the cell body to the synapse. It is generated by a splicing variation in the cargo-binding domain of the KIF1B gene. KIF1Bβ contains a C-terminal PH domain with a preference for binding to PtdIns(4,5)P₂, a phospholipid widely distributed throughout the plasma membrane, Golgi, endosomes and ER, as well as within the nucleus (Hirokawa et al., 2010; Watt et al., 2002). KIF1Bβ is expressed in both neurons and glia and it is essential for proper localization of myelin protein mRNA in zebrafish oligodendrocytes in order to elaborate the correct amount of myelin around axons (Lyons et al., 2009). It also acts as a tumor suppressor and induces apoptosis in neurons (Munirajan et al., 2008; Schlisio et al., 2008). KIF1Bβ is essential for the transport of DENN/MADD and Rab3 vesicles (Niwa et al., 2008). Rab3 is a small GTPase located to synaptic vesicles that is implicated in synaptic vesicle dynamics and exocytosis (Geppert et al., 1997; Takai et al., 1996). DENN/MADD is a Rab3 GEF, designated as GEP, consisting of an N-terminal MADD domain and a conserved C-terminal domain (Niwa et al., 2008; Sakisaka and Takai, 2005). DENN/MADD binds to the stalk domain of KIF1Bβ and interacts with Rab3 on cargo membranes, therefore acting as a linker between KIF1Bβ and Rab3-carrying vesicles (Niwa et al., 2008).

Like Kif1b knockout mice, DENN/MADD knockout mice die after birth because of a respiratory problem and exhibit a reduced number and size of synaptic vesicles (Tanaka et al., 2001; Zhao et al., 2001). In kif1b heterozygous mice, both the number of synapses and the density of synaptic vesicles are reduced, consistent with a defect of synaptic vesicle precursor axonal transport (Zhao et al., 2001). A low survival rate of kif1b⁻/⁻ neurons from mice co-cultured with wild type glia has been reported, suggesting that KIF1Bβ acts cell-autonomously in neurons (Zhao et al., 2001). A mutation of KIF1Bβ has been shown to cause type 2A of CMT disease in a Japanese family. CMT2A patients have a loss-of-function point mutation in the ATP-binding site of the motor region of KIF1Bβ that causes significant reduction in ATPase and in vitro motor activities (Zhao et al., 2001). However, the fact that KIF1Bβ mutation causing CMT2A has been identified in only a single family questions whether this gene should be considered as a significant candidate for the etiology of CMT2 (Zuchner and Vance, 2006).

4.3.2. NEFL

Mutations in the neurofilament light polypeptide gene (NEFL) cause autosomal dominant axonal CMT2 (CMT2E) or demyelinating CMT1 (CMT1F) (De Jonghe et al., 2001; Jordanova et al., 2003; Mersiyanova et al., 2000; Shin et al., 2008a). Recent reports show that NEFL mutations can also cause an autosomal recessive form of CMT neuropathy (Abe et al., 2009; Yum et al., 2009). In the demyelinating forms of CMT disease due to NEFL mutations, the demyelination may only be a consequence of a primary axonopathy (Fabrizi et al., 2004, 2007a). Neurofilaments play important roles in the maintenance of the cytoskeleton and axonal structure. Therefore, it is not surprising that NEFL mutations can cause an axonal neuropathy. Indeed, CMT disease-associated NEFL mutations affect the formation of neurofilament networks, the assembly of neurofilaments and both anterograde and retrograde transport (Brownlees et al., 2002; Pérez-Ollé et al., 2002, 2005). Some mutations can also cause fragmentation of the Golgi apparatus, altered mitochondrial distribution and degeneration of neuritic processes in cultured neuronal cells (Brownlees et al., 2002; Pérez-Ollé et al., 2005). Furthermore, mutations result in the formation of neurofilament aggregates (Fabrizi et al., 2004; Pérez-Ollé et al., 2002; Sasaki et al., 2006). Accumulation of neurofilaments may prevent the transport of proteins and cellular components either by creating a barrier or by trapping them within the inclusions. Also, it has been suggested that NEFL mutants may disrupt the interactions with mitochondria and the formation of neurofilament aggregates traps mitochondria within these inclusions (Pérez-Ollé et al., 2005). Mitochondria accumulation could therefore interfere with the correct supply of energy to the rest of the cell.

Neurofilament transport is dependent on the motor proteins KIF5A and dynein, and both the kinesin and dynein families of motors require ATP (Uchida et al., 2009; Wagner et al., 2004). NEFL mutations could perturb the correct functioning of the motors or the interaction with molecular motors. It is interesting to note that mutations in KIF5A disrupt neurofilament transport and cause hereditary spastic paraplegia (Musumeci et al., 2011; Wang and Brown, 2010). In addition, mice null for KIF5A display abnormal accumulations of neurofilaments (Xia et al., 2003). As mentioned in the previous section, mutation of the molecular motor KIF1Bβ causes another form of type 2 CMT disease, providing further evidence that alterations of molecular motors, and in general of axonal transport, result in neurodegeneration (Xia et al., 2003; Zhao et al., 2001). Axonal transport is essential for the survival of neurons and it is responsible for the transport of organelles and ligands, such as neurotrophins and other growth factors, and transport defects appear to be a cause for the development of neuropathies (De Vos et al., 2008; Goldstein and Yang, 2000).

Disruption of the assembly and aggregation of neurofilaments that interferes with axonal transport is also induced by expression of mutant HSP27 (or HSPB1), a chaperone protein that causes an autosomal recessive form of distal motor neuropathy in CMT2F (Ackerley et al., 2006; Lin and Schlaepfer, 2006; Zhai et al., 2007). It is not surprising that mutant HSP27 causes protein aggregation and loss of viability of transfected neuronal cells, since its normal function is to bind and prevent misfolding and aggregation of nascent proteins and it is known to interact with intermediate filaments (Perng et al., 1999; Vos et al., 2008).

NEFL interacts with MTMR2 in Schwann cells as well as in neurons (Previtali et al., 2003). Interestingly, mutations in MTMR2 cause an autosomal recessive demyelinating form of CMT neuropathy referred to as CMT4B (Bolino et al., 2000). As a consequence of the interaction with NEFL, MTMR2 mutants can lead to neurofilament aggregation (Goryunov et al., 2008). The exact relationship between MTMR2 catalytic activity and NEFL aggregation is not known. However, catalytically inactive CMT disease-related MTMR2 mutants lead to NEFL assembly defects and to pathologies similar to the one caused by NEFL mutations, suggesting that MTMR2 and NEFL may function in a common pathway in the development and maintenance of peripheral axons.

4.3.3. FGD4/Frabin

FGD4/Frabin (FYVE, RhoGEF and PH domain-containing protein 4) is a member of the Cdc42 GEF family (Obaishi et al., 1998). Cdc42, together with the Rac and Rho subfamilies, belongs to the Rho family of small G-proteins, important regulators of actin cytoskeleton organization. Cdc42 regulates actin cytoskeleton dynamics influencing cell migration, adhesion and cytokinesis. Its activity is tightly regulated by the interconversion between a GDP-bound inactive and a GTP-bound active form (Heasman and Ridley, 2008; Jaffe and Hall, 2005). GEFs, like Frabin, activate Cdc42 through the binding of GTP (Umikawa et al., 1999). Frabin can either activate Cdc42 directly, Rac indirectly, or its activity can be Cdc42/Rac independent (Ikeda et al., 2001; Nakanishi and Takai, 2008; Ono et al., 2000; Umikawa et al., 1999).

In fibroblasts, Cdc42 activation by Frabin causes filopodia formation, while Rac activation induces the formation of lamellipodia (Ono et al., 2000). In addition, Frabin can activate both Cdc42 and Rac, inducing microspike formation (Umikawa et al., 1999; Yasuda et al., 2000). Mutations in FGD4 are associated with CMT4H, an autosomal recessive demyelinating form of CMT disease, suggesting that Frabin is involved in the myelination process. Although the molecular mechanisms by which FGD4 mutations cause CMT4H are completely unknown (Delague et al., 2007; Stendel et al., 2007), the molecular structure of Frabin allows for speculation. FGD4 consists of an F-actin-binding (FAB) domain at the N-terminal region, important for the association with F-actin, followed by a Dbl homology (DH) domain, and two PH domains separated by a FYVE domain (Nakanishi and Takai, 2008; Obaishi et al., 1998). DH domains are conserved in the GEFs for Rho proteins, and when in proximity to a PH domain, they are important for the GEF nucleotide-exchange activity (Rossman et al., 2005). Interestingly, PH domains bind PIs on membranes and FYVE domains associate with PtdIns3P, a phosphatidylinositol mainly localized on early endosomes and MVBs (Kutateladze, 2006; Lemmon, 2008). Therefore, the presence of both PH and FYVE domains in Frabin suggests that it may act as a bridge between membranes containing PtdIns3P and actin. The fact that Cdc42 functions in vesicle transport by regulating actin supports this hypothesis (Harris and Tepass, 2010; Luna et al., 2002; Musch et al., 2001). Actin is known to be important for vesicle trafficking in several ways, and it provides tracks for motor protein-based vesicle transport. In mammalian cells, a significant fraction of Cdc42 localizes to the Golgi apparatus where it binds to the coat protein I (COPI) vesicle coat protein, thus promoting vesicle formation and regulating actin dynamics (Chen et al., 2005, 2004a; Luna et al., 2002; Matas et al., 2004). In addition, at the cell cortex, Cdc42 promotes actin assembly, thereby regulating exocytosis (Momboisse et al., 2009; Zhang et al., 2008b). Finally, the PH and FYVE domains may also connect Frabin to MTMRs by mediating the binding to myotubularin substrates and products. The finding that nerve biopsy samples from patients with MTMR-associated CMT neuropathy show abnormalities in myelin folding similar to those observed in CMT4H patients is in line with this hypothesis (Stendel et al., 2007). In conclusion, mutations of Frabin could affect PI metabolism or cytoskeleton dynamics, thereby interfering with membrane traffic and/or myelin deposition.

4.4. Defects in the regulation of membrane traffic events: Rab7, NDRG1 and SH3TC2

4.4.1. Rab7

Rab7 is a small GTPase of the Rab family, first identified in a rat liver cell line (Bucci et al., 1988). Rab7 is localized to late endosomes and lysosomes and regulates late endocytic traffic. It is thus a key protein for the biogenesis of lysosomes and phagolysosomes, and for the maturation of late autophagic vacuoles (Bucci et al., 2000; Harrison et al., 2003; Jager et al., 2004) (Figs. 3 and 4). Rab7 is also important for cell nutrition and apoptosis (Edinger et al., 2003; Snider, 2003) (Fig. 4).

Fig. 3.

Role of endocytic membrane traffic proteins involved in CMT disease. DNM2 is important for budding of vesicles from the plasma membrane. Rab7 controls transport from late endosomes (LE) to lysosomes (Lys). LRSAM1 binds and ubiquitinates Tsg101, a protein involved in formation of MVBs and in the sorting of signaling receptors in intraluminal vesicles in order to be degraded; ubiquitination of Tsg101 inactivates its sorting function and thus inhibits signaling receptor degradation. A role for LITAF/SIMPLE in the sorting and degradation of signaling receptors is strongly suggested by the finding that this protein is ubiquitinated by Nedd4 and then interacts with Tsg101.

Fig. 4.

Intracellular trafficking proteins involved in CMT neuropathy. DNM2 regulates vesicle budding. KIF1B controls vesicle motility on microtubules. LITAF/SIMPLE and LRSAM are present in the endocytic pathway and probably regulate protein degradation. Myotubularin-related proteins (MTMR2 and MTMR13) and FIG4 regulate PI metabolism at the level of early endosomes and late endosomes, respectively. Rab7 is present on late endosomes and regulates transport to lysosomes. SH3TC2 regulates endosomal recycling together with Rab11, while NDRG1 regulates membrane traffic at the level of early endosomes together with Rab4 and PRA1. HSPs regulate proteasomal degradation and associate with neurofilaments and actin filaments. FGD4 associates with and regulates actin filaments. MFN2 and GDAP regulate mitochondrial dynamics and mitochondrial axonal transport.

A number of Rab7 effector proteins have been identified. For instance, Rab7 recruits RILP (Rab-interacting lysosomal protein) on endosomal membranes, which in turn recruits the dynein/dynactin complex, thereby allowing microtubule minus-end-mediated transport of endosomes and lysosomes (Cantalupo et al., 2001; Harrison et al., 2003; Johansson et al., 2007). Also, FYCO1 (FYVE and coiled-coil domain containing 1) forms a complex with Rab7 and directs plus-end transport of autophagosomes along microtubules (Pankiv et al., 2010). Rab7 also recruits the retromer complex to endosomes, and interacts with hVps34/p150 regulating PI 3-kinase activity in the endo-lysosomal pathway (Rojas et al., 2008; Stein et al., 2003).

Given the importance of this small GTPase in many cellular functions (Fig. 4) and its multiple interactions, it is not surprising that Rab7 mutations underlie neuronal diseases. Indeed, four missense mutations in the rab7 gene on chromosome 3q21 are associated with CMT2B (Chiariello et al., 1998; Houlden et al., 2004; Meggouh et al., 2006; Verhoeven et al., 2003a). The four CMT2B-causing Rab7 mutant proteins have been characterized biochemically and show very similar biochemical properties (De Luca et al., 2008; Spinosa et al., 2008). Indeed, they have increased K_off for nucleotides, and these altered nucleotide dissociation rates in turn negatively affect GTPase activity per binding event (De Luca et al., 2008; Spinosa et al., 2008). In particular, the GDP dissociation rate is strongly increased and, accordingly, the mutant proteins are predominantly in the GTP-bound form (De Luca et al., 2008; Spinosa et al., 2008). In addition, their activation is not dependent on GEFs, and they show enhanced interaction with a number of effector proteins (McCray et al., 2010; Spinosa et al., 2008). Furthermore, they are able to rescue Rab7 function following Rab7 silencing, suggesting that they behave as active mutant proteins (De Luca et al., 2008; Spinosa et al., 2008).

In neurons, Rab7 regulates long-range retrograde axonal transport of neurotrophins and neurotrophin receptors (Deinhardt et al., 2006). It also controls endocytic traffic and neuritogenic signaling of the nerve growth factor receptor TrkA. Indeed, after neuronal stimulation by nerve growth factor (NGF), Rab7 interacts with TrkA with effects on receptor signaling and neurite outgrowth (Saxena et al., 2005a). NGF promotes neuronal survival and neurite outgrowth by binding and activating its receptor TrkA on axon tips (Kaplan et al., 1991; Klein et al., 1991). Upon NGF binding, TrkA is internalized into endosomes and retrogradely transported, continuing to signal (Ehlers et al., 1995; Grimes et al., 1997, 1996). Signaling endosomes containing activated TrkA are then transported retrogradely over long distances from the axonal synapse to the cell body (Delcroix et al., 2003; Ehlers et al., 1995; Grimes et al., 1997; Howe and Mobley, 2005; Saxena et al., 2005b). Interestingly, inhibition of Rab7 activity causes the accumulation of TrkA within endosomes and enhanced TrkA signaling in NGF-stimulated PC12 cells, leading to an increase in neurite outgrowth (Saxena et al., 2005a). CMT2B-associated Rab7 mutants are still able to interact with TrkA after NGF stimulation, but TrkA phosphorylation is strongly enhanced and the downstream signaling pathways are altered, leading to the inhibition of neurite outgrowth in PC12 cells (BasuRay et al., 2010; Cogli et al., 2010). Therefore, Rab7 plays an important role in controlling TrkA signaling by regulating its endosomal traffic, possibly through control of the endosomal signaling time, and subsequently promoting neurite outgrowth. Defects in Rab7 activity may interfere with the endosomal residence time of TrkA and thereby with the signal duration. However, this may not be the only pathogenic effect of Rab7 mutations, since defective neurite outgrowth has also been observed in Neuro2A cells where TrkA signaling is not involved (Cogli et al., 2010). This suggests that another alternative pathway may contribute to CMT2B. For instance, Rab7 could interact with an effector selectively expressed in peripheral neurons only, thus regulating a cell-type-specific pathway (Cogli et al., 2009). Notably, impaired neurite outgrowth has also been observed in other cell lines where the effect was reversed by valproic acid, indicating a way to overcome the inhibition of neurite outgrowth (Yamauchi et al., 2010).

Rab7 is a ubiquitously expressed protein (Bucci et al., 1988; Verhoeven et al., 2003a). Why should mutations in a ubiquitous protein selectively affect a specific cell type? It is important to note that axonal transport is very important for neuronal functions, and axons, especially in peripheral neurons, can be particularly long, more than one meter. Therefore, mutations in membrane trafficking may have stronger effects on such cells where the transport of cellular components needs to cover large distances compared to other cell types. Also, CMT2B patients do not show any developmental defects, indicating that the inhibition of neurite outgrowth in this case might affect neuroregeneration, which becomes less successful with age, explaining the late onset of CMT2B (Perlson et al., 2004; Wu et al., 2007). Neuroregeneration of axons consists of the formation of a new growth cone at the cut tip of the axon after damage. An increase of intracellular calcium levels, intracellular signaling pathways, and local protein synthesis and degradation are involved in the formation of the new growth cone (Chierzi et al., 2005; Gitler and Spira, 1998; Liu and Snider, 2001; Verma et al., 2005; Wu et al., 2007). If Rab7 mutant proteins affect axonal regeneration, the specific effect on peripheral neurons could be due to the fact that axonal regeneration in mammals occurs mainly in the peripheral nervous system (Hilliard, 2009; Shim and Ming, 2010). It is worth noting that aberrations of macroautophagy have been observed in several neurodegenerative disorders. Thus, as Rab7 regulates the maturation of autophagic vacuoles, CMT disease-causing mutations in the Rab7 protein could also affect this pathway (García-Arencibia et al., 2010; Jager et al., 2004; Martinez-Vicente et al., 2010; Vogiatzi et al., 2008; Yu et al., 2005).

4.4.2. NDRG1

Mutations in N-myc downstream regulated gene 1 (NDRG1) are responsible for CMT4D, an autosomal recessive demyelinating neuropathy (Hunter et al., 2003; Kalaydjieva et al., 2000, 1996, 1998). The 43 kDa NDRG1 protein is a member of the NDRG family characterized by an α/β hydrolase region without presenting a hydrolytic catalytic site (Melotte et al., 2010; Shaw et al., 2002). NDRG1 is ubiquitously expressed, with high levels in the peripheral nervous system where it is confined to Schwann cells (Berger et al., 2004; King et al., 2011). NDRG1 is repressed by N-myc during mouse development (Shimono et al., 1999), upregulated during cellular differentiation (van Belzen et al., 1997) and positively regulated by p53 which leads to reduced expression in p53-dependent tumors (Kurdistani et al., 1998). It has been proposed as a metastasis suppressor gene (Bandyopadhyay et al., 2006; Guan et al., 2000), and to function in the ER stress response (Segawa et al., 2002). However, how NDRG1 mediates its multiple functions remains largely unknown.

Myelinating Schwann cells and oligodendrocytes express NDRG1; however, sensory and motor neurons as well as their axons lack NDRG1 (Berger et al., 2004). NDRG1 mutations in CMT4D patients lead to a Schwann-cell-autonomous phenotype resulting in defective myelination with secondary axonal degeneration, indicating a role for the wild type protein in the development and/or maintenance of the myelin sheaths in peripheral nerves (Berger et al., 2004; Okuda et al., 2004). Since myelin biogenesis involves coordinated activities of both the endocytic and the exocytic pathways (Anitei and Pfeiffer, 2006; Trajkovic et al., 2006; Winterstein et al., 2008), it is not surprising that several NDRG1 interaction partners with various roles in intracellular trafficking have been described, highlighting that the mechanism of CMT4D pathogenesis is connected to the alterations in the trafficking inside Schwann cells (Hunter et al., 2005; Kachhap et al., 2007).

NDRG1 localizes to the nucleus and the cytoplasm. Its association with adherens junctions suggests a functional involvement in the E-cadherin/catenin complex (Berger et al., 2004; Lachat et al., 2002; Tu et al., 2007). E-cadherin is a 120 kDa transmembrane glycoprotein present in adherens junctions on the surface of epithelial cells (Whittard et al., 2002). Its extracellular domain forms a homodimer with E-cadherin of neighboring cells in the presence of extracellular calcium (Koch et al., 1997; Nagar et al., 1996). The cytoplasmic domain of E-cadherin interacts with α-, β- and γ-catenins (Kobielak and Fuchs, 2004; Piepenhagen and Nelson, 1993). The assembly and turnover of the E-cadherin molecule involve its phosphorylation, ubiquitination, internalization by endosomes, and subsequent lysosomal or proteasomal degradation or recycle back to the cell surface (Fujita et al., 2002). E-cadherin trafficking from the cell surface to the endosomes and back is central to the dynamics and stability of the adhesion complex (Le et al., 1999). It has been demonstrated that NDRG1 is a Rab4a effector involved in the recycling of E-cadherin. NDRG1 is recruited from the cytosol to perinuclear recycling/sorting endosomes by binding to PtdIns4P (Kachhap et al., 2007). Rab4, together with Rab11, is involved in the regulation of endosomal recycling back to the plasma membrane. They also participate in adherens junction dynamics by interacting with α- and β-catenin (Mruk et al., 2007; Sönnichsen et al., 2000). Interestingly, an involvement of the cadherin/catenin complex in the initiation of myelination at the Schwann cell-axon interface has recently been demonstrated (Lewallen et al., 2011). NDRG1s role in CMT4D pathogenesis therefore appears to be connected to its function in cellular trafficking. Additional evidence is provided by the identification of the trafficking protein prenylated Rab acceptor 1 (PRA1) as an interacting partner of NDRG1 (Hunter et al., 2005). PRA1 is required for vesicle formation from the Golgi apparatus, and interacts with members of the Rab family which regulate transport between organelles, including Rab7 and Rab4 (Bucci et al., 1999; Gougeon et al., 2002). Interestingly, as mentioned earlier, Rab7 plays a crucial role in late endosomal traffic and is mutated in another form of CMT neuropathy, CMT2B, and Rab4 interacts with NDRG1 (Hunter et al., 2005; Kachhap et al., 2007; Verhoeven et al., 2003a; Vitelli et al., 1997). NDRG1 may thus be another member of the group of CMT disease-associated proteins with a role in endosomal transport.

Other NDRG1-interacting proteins are the apolipoproteins A-I (APOA1) and A-II (APOA2), suggesting a role for NDRG1 in lipid transport (Hunter et al., 2005). These proteins are components of high-density lipoproteins that regulate lipid distribution within the body (Schmitz and Grandl, 2009). The integrity of this process is very important. Indeed, genetic disorders of cholesterol transport, such as APOA1 deficiency, cause peripheral neuropathy (Ng et al., 1996). Because myelinating Schwann cells have a high demand for lipids, CMT4D may be a trafficking disorder in these cells, resulting in abnormal targeting of lipids/proteins to the myelin membrane (Hunter et al., 2005).

4.4.3. SH3TC2

CMT disease type 4C (CMT4C) is an autosomal recessive form of demyelinating neuropathy characterized by mutations in SH3TC2 (Azzedine et al., 2006; Senderek et al., 2003a). CMT4C mutations lead to truncations at the N- and C-termini and also to amino acid substitutions throughout the SH3TC2 protein. The 1288 amino acid SH3TC2 protein is strongly expressed in neural tissues, including peripheral nerve tissue. It contains two N-terminal SH3 domains and 10 C-terminal tetratricopeptide repeat (TPR) domains (Senderek et al., 2003a). SH3 domains bind to proline-rich regions of other proteins and are involved in clathrin-mediated vesicle endocytosis and synaptic vesicle recycling (Kim and Chang, 2006; McPherson, 1999). TPR domains, usually present in tandem repeats, mediate protein-protein binding and multiprotein complex formation (Blatch and Lassle, 1999). SH3TC2 localizes to recycling endosomes and interacts with the small GTPase Rab11 (Arnaud et al., 2009; Roberts et al., 2010; Stendel et al., 2010). Rab11 regulates the recycling of internalized receptors and membrane back to the cell surface (Ullrich et al., 1996). Interestingly, SH3TC2 mutants causing CMT4C are unable to associate with Rab11, with consequent mistargeting from recycling endosomes towards the cytosol (Roberts et al., 2010; Stendel et al., 2010).

Schwann cell dysfunction and defects in myelination have been observed in SH3TC2 knockout mice. Furthermore, impaired myelination in primary rat Schwann cells expressing the dominant negative Rab11 has been reported. These findings suggest that SH3TC2, together with Rab11, regulates Schwann cell myelination (Arnaud et al., 2009; Stendel et al., 2010).

Recycling endosomes have been shown to sort and redirect some myelin components to the plasma membrane during morphogenesis of the myelin sheath in oligodendrocytes. However, the molecular pathways regulating vesicular transport during myelination are largely unknown (Winterstein et al., 2008). The SH3TC2/Rab11 interaction is therefore relevant for peripheral nerve pathophysiology, highlighting the important role of the endosomal recycling pathway in Schwann cell myelination (Stendel et al., 2010).

4.5. Defects in the regulation of protein degradation: HSPs, LRSAM1 and LITAF/SIMPLE

4.5.1. HSPs

Mutations in the genes HSPB1 (HSP27) and HSPB8 (HSP22), two members of the small HSP superfamily, have been associated with CMT2F and CMT2L, respectively (Evgrafov et al., 2004; Irobi et al., 2004; Tang et al., 2005). HSP22/HSPB8 is ubiquitously expressed, with high expression detected in the spinal cord and in motor and sensory neurons (Irobi et al., 2004). Immunolocalization studies in neuroblastoma cell lines have shown that HSP22 is predominantly localized to the plasma membrane. Furthermore, it possesses chaperone-like activity and prevents protein aggregation (Carra et al., 2005; Chowdary et al., 2004, 2007; Kim et al., 2004). It has recently been demonstrated that HSP22 forms a complex with the co-chaperone Bag3 to target misfolded proteins to degradation by macroautophagy (Carra et al., 2008a,b). In addition, HSP22 possesses pro-apoptotic activity, in contrast to the anti-apoptotic activity of most of the small HSPs (Gober et al., 2003; Li et al., 2007). In line with the fact that small HSPs form homo- and hetero-oligomeric complexes, HSP22 interacts with HSP27 (HSPB1), MKBP (HSPB2), HSPB3, αB-crystallin (HSPB5), HSP20 (HSPB6) and cvHSP (HSPB7) (Fontaine et al., 2005; Sun et al., 2004).

Although mutations of HSP22 are known to cause CMT2L and other neuromuscular disorders, the molecular mechanism underlying these diseases is poorly understood (Irobi et al., 2004; Tang et al., 2005). These missense mutations, located in the central α-crystallin domain of HSP22, decrease the chaperone-like activity and alter the interaction with other small HSPs (Irobi et al., 2004; Kim et al., 2006). Interestingly, it has been shown that some HSP22 mutants are less effective than the wild type protein in preventing Htt43Q aggregation, a pathogenic form of huntingtin responsible for Huntington's disease (Carra et al., 2005). Huntingtin is a ubiquitously expressed protein in mammals that has a role in the intracellular transport of vesicles and organelles along microtubules (Caviston and Holzbaur, 2009; DiFiglia et al., 1995; Hoffner et al., 2002). Huntingtin is also linked to actin-based and endosomal motility; however, its function is not yet fully understood (Caviston and Holzbaur, 2009; Pal et al., 2006). Mutant huntingtin causes defective axonal trafficking, and thus defects in huntingtin protein clearance due to mutations in HSP22 might affect the intracellular trafficking along the axon. This may explain why HSP22 mutations specifically affect neurite length and motor neuron integrity without affecting other cell types (Gunawardena et al., 2003; Irobi et al., 2010; Szebenyi et al., 2003; Trushina et al., 2004).

HSP27/HSPB1 is ubiquitously expressed in human tissues and, as mentioned above, interacts with HSP22 (Sun et al., 2004). HSP27 is important for axonal outgrowth in the peripheral nervous system, and is induced in regenerated axons (Hirata et al., 2003; Read and Gorman, 2009; Williams et al., 2005). HSP27 has a key role in neuronal survival, binding to molecular components of the apoptotic machinery to inhibit neuronal cell death (Benn et al., 2002; Voss et al., 2007). It also functions in proteasome-mediated degradation of proteins. Indeed, HSP27 interacts with components of the proteasome and binds ubiquitin, facilitating protein degradation (Parcellier et al., 2006, 2003).

Although HSP27 regulates cytoskeletal dynamics, the molecular mechanism is not fully understood. It is known that HSP27 associates with and stabilizes the actin cytoskeleton (Jia et al., 2010; Pivovarova et al., 2007). HSP27 affects actin at the cell surface, with effects on membrane ruffling, pinocytosis, cell migration and accumulation of stress fibers (Doshi et al., 2009; Lavoie et al., 1993; Lee et al., 2008; Mounier and Arrigo, 2002; Schneider et al., 1998). Phosphorylation of HSP27 by MAPK pathways regulates actin polymerization, whereas unphosphorylated HSP27 monomers inhibit actin polymerization (Benndorf et al., 1994; Guay et al., 1997; Kostenko et al., 2009; Pichon et al., 2004; Schneider et al., 1998). Blocking HSP27 phosphorylation in dorsal root ganglion neurons by inhibition of MAPK pathways causes aberrant neurite growth, highlighting the importance of phosphorylated-HSP27 for the interaction with actin and neurite outgrowth (Williams et al., 2005).

HSP27 also associates with intermediate filaments, preventing their aggregation (Jia et al., 2010; Perng et al., 1999). Interestingly, HSP27 mutants lead to progressive degeneration of motor neurons which disrupts the neurofilament network with consequent aggregation of NEFL protein, thus providing evidence for the essential role of HSP27 in neurofilament assembly (Evgrafov et al., 2004; Zhai et al., 2007). Importantly, mutations in the NEFL gene also cause a form of CMT2 (CMT2E) (Fabrizi et al., 2007a). Four missense mutations associated with distal hereditary motor neuropathy and CMT2F occur in the HSP27 conserved α-crystallin domain and one is positioned in the variable C-terminal part (Evgrafov et al., 2004). Like HSP22 mutations, HSP27 mutations in the core α-crystallin domain also decrease its chaperone function. Interestingly, high levels of HSP27 have been detected in individuals with neurodegenerative disorders characterized by accumulation of improperly folded proteins, inclusion bodies or plaques in the nervous system such as ALS, Alzheimer's, Parkinson's and Alexander's disease (Head et al., 1993; Renkawek et al., 1999; Shimura et al., 2004; Vleminckx et al., 2002). Small HSPs facilitate the refolding or degradation of misfolded proteins, preventing their aggregation. Alterations in these functions could therefore explain the role of HSP27 in the progression of these disorders.

Given the diversity of the interactions and functions of the small HSPs, the identification of the pathological mechanism responsible for the forms of CMT neuropathy caused by mutations in HSP22 and HSP27 is not straightforward. These mutations could interfere with the chaperone-like activity, causing misfolding and aggregation of other proteins. The fact that patients present CMT neuropathies late in life could be explained by the delayed effects of aggregates accumulating in neurons. Another mechanism for the disease may be that small HSP mutations alter the apoptotic pathway, thereby influencing the pro-survival activity of HSP27 or the pro-apoptotic activity of HSP22. However, the role of HSP27 in neurofilament assembly and the disruption of the neurofilament network caused by HSP27 mutants strongly suggest that the pathologic mechanism for CMT2F is based on alterations of cytoskeletal dynamics and axonal transport. HSP22 has not been shown to interact with cytoskeletal elements; however, it is possible that mutated HSP22 indirectly affects axonal transport through the interaction with its partner HSP27. In line with this hypothesis, it has been demonstrated that mutations in HSP22 increase the interaction with HSP27, leading to the formation of aggregates (Irobi et al., 2004). Thus, HSP22 mutations might interfere with HSP27 function. In conclusion, the mutations in HSP27 and HSP22, which cause two forms of CMT disease, could alter, directly or indirectly, cytoskeletal functions and affect axonal transport in motor and sensory neurons.

4.5.2. LRSAM1

A mutation of the LRSAM1 gene has recently been identified in patients with an autosomal recessive axonal form of CMT disease (Guernsey et al., 2010). LRSAM1 (leucine rich repeat and sterile alpha motif 1), also known as TAL (Tsg101-associated ligase) or RIFLE, is a RING finger E3 ubiquitin ligase that plays a role in endocytosis and in adhesion of neuronal cells in culture (Amit et al., 2004; Li et al., 2003). Tsg101 (tumor susceptibility gene 101) sorts monoubiquitinated cargoes like EGFR into MVBs and retroviral Gag proteins for budding out of the cell (Bishop et al., 2002; Garrus et al., 2001; Katzmann et al., 2001; Slagsvold et al., 2006). LRSAM1 has two domains that independently bind to Tsg101. Bivalent binding is essential for attachment of multiple monomeric ubiquitins to Tsg101 (McDonald and Martin-Serrano, 2008). Following ubiquitination, Tsg101's sorting function is inactivated (Amit et al., 2004) (Fig. 3). A recycling model of ubiquitination/deubiquitination has been proposed whereby multiple monoubiquitination of Tsg101 by LRSAM1 inactivates Tsg101 and deubiquitinating enzymes reactivate its sorting function, thus regulating its shuttling between a membrane-bound active form and an inactive soluble form (Amit et al., 2004). LRSAM1 is also a regulator of Tsg101 expression. Polyubiquitination of Tsg101 C-terminal lysines by LRSAM1 targets excess of the protein to proteasomal degradation (McDonald and Martin-Serrano, 2008). LRSAM1 mutations that make the protein catalytically inactive and its depletion by siRNA both accelerate receptor degradation (Amit et al., 2004). It is interesting to note that the LRSAM1 gene mutation detected in patients with CMT neuropathy appears to be a loss-of-function of the gene product, suggesting that the disease-causing mutation affects the degradative pathway (Guernsey et al., 2010).

4.5.3. LITAF/SIMPLE

The LITAF/SIMPLE gene was identified in 1997 as a p53-induced gene, PIG7 (Polyak et al., 1997). LITAF/SIMPLE is a widely expressed gene encoding a protein involved in protein degradation that has been proposed to localize to early endosomes (Lee et al., 2011b) or to the late endosomal/lysosomal compartments (Eaton et al., 2011; Moriwaki et al., 2001). LITAF/SIMPLE is a non-glycosylated membrane protein that exhibits patches of sequence similarity with major integral membrane proteins of lysosomes, LAMPs, LIMPs, and also with endolyn, mainly in the N-terminal domain. Its C-terminus contains a modified RING finger domain and the carboxyl terminus signal for endocytosis YXXΦ (where Φ is any bulky hydrophobic amino acid) (Eaton et al., 2011; Lee et al., 2011b; Moriwaki et al., 2001). LITAF/SIMPLE contains two domains at the N-terminus that mediate the interaction with WW domain-containing proteins: a PPXY responsible for binding to neuronal precursor cell-expressed developmentally downregulated 4 (Nedd4); and a P(S/T)AP motif that binds with Tsg101 (Shirk et al., 2005). The same domains are also responsible for the binding of another WW domain-containing protein, Itch (Eaton et al., 2011). Nedd4 is an E3 ubiquitin ligase which monoubiquitinates membrane proteins that need to reach the lysosomes in order to be degraded (Ingham et al., 2004). ESCRT is the machinery involved in the sorting of ubiquitinated membrane proteins to lysosomes. Monoubiquitinated substrates are recognized by Tsg101, a component of ESCRT-I that acts downstream of Nedd4 (Blot et al., 2004; Haglund et al., 2003). ESCRT-II and -III complexes subsequently sort target proteins into MVBs. Following fusion with lysosomes, the MVB content is degraded (Raiborg and Stenmark, 2009). Although the function of LITAF/SIMPLE is not yet fully characterized, the following hypothesis based on its interaction with Nedd4 and Tsg101 and on its localization along the endo-lysosomal pathway currently exists: Nedd4 ubiquitinates LITAF/SIMPLE and the ubiquitinated LITAF/SIMPLE interacts with Tsg101, suggesting a role for LITAF/SIMPLE in the ubiquitin-mediated lysosomal degradation pathway (Shirk et al., 2005) (Fig. 3).

Itch is a member of the Nedd4 family that ubiquitinates and induces proteasomal degradation of different substrates (Azakir and Angers, 2009; Azakir et al., 2010; Chang et al., 2006; Rossi et al., 2005). It localizes to the TGN, but after the interaction with LITAF/SIMPLE, it changes its localization to lysosomes (Angers et al., 2004; Eaton et al., 2011). Even though Itch and Nedd4 are very similar proteins and members of a conserved family of ubiquitin ligases, Nedd4 localization is not altered by LITAF/SIMPLE (Eaton et al., 2011).

Mutations of the LITAF/SIMPLE gene are associated with CMT1C, an autosomal dominant demyelinating form of CMT type 1, suggesting that LITAF/SIMPLE may have a critical role in peripheral nerve function (Bennett et al., 2004; Gerding et al., 2009; Latour et al., 2006; Saifi et al., 2005; Street et al., 2003). Despite the ubiquitous pattern of expression, the high expression level of LITAF/SIMPLE in the peripheral nerves and Schwann cells explains why mutations in this protein can cause a demyelinating neuropathy that specifically affects the peripheral nervous system (Lee et al., 2011b; Moriwaki et al., 2001; Street et al., 2003). However, how these mutations cause peripheral nerve demyelination is unknown. It has recently been shown that CMT1C-linked mutants mislocalize from the membrane of early endosomes to the cytosol and that they are unstable, prone to aggregation, and degraded by both the proteasome and aggresome–autophagy pathways (Lee et al., 2011b). These findings, together with the fact that CMT1A (another form of CMT neuropathy caused by gene duplication or point mutations in PMP22) is also characterized by the formation of intracellular ubiquitinated PMP22 aggregates (aggresomes), suggest protein misfolding as a common cause of demyelinating peripheral neuropathies and highlight the importance of the proteasome and autophagy pathways in the clearance of CMT disease-associated mutant proteins (Fortun et al., 2006; Ryan et al., 2002).

4.6. Defects in mitochondrial dynamics and mitochondrial axonal transport: MFN2 and GDAP1

4.6.1. MFN2

Mitofusins mediate the process of mitochondrial fusion and regulate mitochondrial metabolism, apoptosis and cellular signaling (de Brito and Scorrano, 2008b; Santel, 2006). MFN2 is important not only for mitochondrial fusion but also for tethering of mitochondrial and ER membranes (de Brito and Scorrano, 2008a, 2010). MFN2 mutations are associated with type 2A of CMT disease and with hereditary motor and sensory neuropathy type VI (Kijima et al., 2005; Lawson et al., 2005; Züchner et al., 2006, 2004). Both disorders lead to axonal degeneration, and the latter is coupled with visual impairment due to optic atrophy (Züchner et al., 2006). MFN2 mutations in CMT2A are located in the GTPase domain and in the C-terminal coiled-coil domain, suggesting that the mutated proteins are defective either in GTPase activity or in the capacity to tether to fusion partners during mitochondrial fusion (Santel, 2006; Verhoeven et al., 2006). Interestingly, mutations in OPA1, a mitochondrial inner membrane protein important for mitochondrial fusion, also result in neuropathologies, suggesting that alterations in the fusion process may be the cause of neuronal disorders (Alexander et al., 2000). However, it is not clear why defects in mitochondrial fusion would affect neuronal cells only.

The mechanism by which MFN2 mutations cause CMT2A is unknown. Current models propose that it could be the consequence of mitochondrial transport defects in the axons (Cartoni and Martinou, 2009). Mitochondrial transport and distribution are particularly important for neurons, where energy is required far from the cell body, along axons and dendrites. In agreement with this hypothesis that mutations in MFN2 may perturb the dynamics or the axonal transport of mitochondria, expression of CMT2A-related MFN2 mutants in neurons leads to mitochondrial transport defects and aggregation around the nucleus, with few and mostly static mitochondria along axons. Interestingly, the decrease of axonal mitochondrial transport is not caused by alterations in mitochondrial oxidative respiration, indicating that alterations caused by CMT2A mutants are independent of defects in energy production (Baloh et al., 2007). In addition, mitochondria are improperly distributed along the axon in motoneurons of transgenic mice expressing an MFN2 pathogenic allele, and in Purkinje cells of MFN2-deficient mice (Chen et al., 2007; Detmer et al., 2008). It has been demonstrated that a correct mitochondrial distribution in peripheral axons is also important for the proper function of neurons in Drosophila (Guo et al., 2005; Stowers et al., 2002; Verstreken et al., 2005). Finally, in CMT2A patients, mitochondria accumulate in the distal part of sural nerve axons (Verhoeven et al., 2006).

Adaptor proteins connect mitochondria to kinesin and dynein motor proteins that are responsible for anterograde and retrograde transport along axonal microtubules (Fransson et al., 2006; Glater et al., 2006; Hollenbeck and Saxton, 2005; Li et al., 2009). Altered mitochondrial distribution is also seen in cells lacking the kinesin Kif5b, and in cells transfected with mutated Miro adaptor protein (Fransson et al., 2006; Guo et al., 2005; Tanaka et al., 1998). Miro is a transmembrane GTPase, associated with the outer membrane of mitochondria, important for the correct axonal transport of mitochondria in neurons (Guo et al., 2005). It forms a complex with Milton, that in turn binds to the kinesin heavy chain. Therefore, the Milton/Miro complex connects kinesins to mitochondria (Fransson et al., 2006; Glater et al., 2006; Stowers et al., 2002). It was recently demonstrated that MFN2 interacts with mammalian Miro (Miro1/Miro2) and Milton (OIP106/GRIF1), regulating mitochondrial transport in axons (Misko et al., 2010). Taken together, these findings strongly indicate that defect(s) in mitochondrial axonal transport could be the underlying mechanism responsible for the pathophysiology of CMT2A.

4.6.2. GDAP1

More than 40 different mutations in ganglioside-induced differentiation-associated protein 1 (GDAP1) cause different forms of CMT neuropathy (Baxter et al., 2002; Cassereau et al., 2011; Cuesta et al., 2002). Mutations in the GDAP1 gene are usually linked to recessive forms of CMT disease (CMT4A or AR CMT2) and more rarely to a dominant form (CMT2K), the latter being far less severe (Cassereau et al., 2011). Indeed, recessive forms of CMT neuropathy due to GDAP1 mutations show an early onset, usually in the first decade of life, and rapid progression of the disease with assisted walking after the age of 10 and wheelchair requirement in the third decade of life (Cassereau et al., 2011). Missense mutations have been reported in sporadic cases of CMT disease (Kabzińska et al., 2011).

GDAP1 is highly expressed in neurons, in particular in motor and sensory neurons of the spinal cord, and it is localized to the mitochondrial outer membrane (Niemann et al., 2005; Pedrola et al., 2005). GDAP1 has a single transmembrane domain and its targeting to the outer mitochondrial membrane and function are dependent on its tail anchor (Wagner et al., 2009). GDAP1 is important for mitochondrial network dynamics (Niemann et al., 2005; Pedrola et al., 2005). Indeed, silencing of GDAP1 results in tubular mitochondrial morphology, whereas overexpression of GDAP1 induces fragmentation of mitochondria (Pedrola et al., 2008). Interestingly, in the recessively inherited forms of CMT disease, GDAP mutated proteins have reduced fission activities. In the dominantly inherited forms of CMT neuropathy, GDAP mutated proteins negatively influence the fusion of mitochondria (Niemann et al., 2009). Truncated GDAP, resulting from CMT disease-causing mutations, is not targeted to mitochondria and is thus unable to cause mitochondrial fragmentation, inducing perturbation of normal mitochondrial dynamics (Niemann et al., 2005). Alterations of mitochondrial dynamics disturb the integrity of peripheral nerves, leading to both axonal and myelination defects (Niemann et al., 2005). However, CMT disease-associated mutations in GDAP1 appear to lead mainly to an axonal phenotype, although a variable degree of demyelination has been observed associated with the different mutations (Cassereau et al., 2011). Therefore, it is unclear whether GDAP1 mutations affect both neurons and Schwann cells, and whether their effect is cell-autonomous or caused by altered axon-Schwann cell interactions (Suter and Scherer, 2003). As mitochondrial dynamics are affected by GDAP1 mutations, it is believed that mitochondrial motility, in particular in the more distal portion of the axons, could be affected as well as energy production by mitochondria (Cassereau et al., 2011).

4.7. Defects in myelination

As demyelinating defects are one of the major causes of CMT neuropathy, mutations in a number of different genes involved in myelination have been identified. The first two, PMP22 and MPZ, encode structural components of myelin. PMP22, a small protein expressed primarily in Schwann cells, is a major component of the myelin sheath. It is important for correct myelination and maintenance of the myelin sheath and axons (Naef and Suter, 1998; Snipes et al., 1992). Duplication, deletion or point mutations of PMP22 are associated with different forms of CMT disease: CMT1A, Hereditary Neuropathy with liability to Pressure Palsies (HNPP), CMT1E and AR CMT1 (Dubourg et al., 2006; Houlden and Reilly, 2006) (Table 1). Furthermore, the major integral membrane protein of peripheral nerve myelin, MPZ (myelin protein zero), is mutated in CMT1B, AR CMT1 (Déjèrine–Sottas neuropathy) and CMT2I/J (Berger et al., 2006b; Houlden and Reilly, 2006; Shy, 2006). MED25, also known as ARC92 or ACID1, is a component of the Mediator complex that recruits RNA polymerase II to specific gene promoters (Rana et al., 2011). A mutation in MED25 leads to CMT2B2 (Leal et al., 2009). Patients present a classic axonal peripheral neuropathy with mild myelin defects (Leal et al., 2009). Data showing that MED25 expression levels correlate with PMP22 expression levels suggest that one of the genes regulated by MED25 is PMP22, indicating that MED25 is important in myelination (Leal et al., 2009).

Other genes involved in myelination and mutated in CMT disease are early growth response 2 (EGR2), Gap junction β-1 (GjB1) and periaxin (PRX). EGR2 is a zinc finger transcription factor that induces the expression of several proteins involved in myelin sheath formation and maintenance, for example, MPZ (Jang and Svaren, 2009). Mutations of EGR2 have been shown to be associated with CMT1D, CMT4E, Déjèrine–Sottas neuropathy and congenital hypomyelinating neuropathy (Bellone et al., 1999; Mikesová et al., 2005; Timmerman et al., 1999; Warner et al., 1998; Yoshihara et al., 2001). Mutations of EGR2 causing CMT disease inhibit myelin gene expression. It has been demonstrated that one of these mutations decreases the binding of EGR2 to the promoter of GjB1, another gene associated with CMT neuropathy that is important for myelination (Musso et al., 2001; Nagarajan et al., 2001). The GjB1 protein (also called connexin-32) is a transmembrane protein that oligomerizes to form gap junction channels that allow diffusion of small molecules (Rahman et al., 1993). Altered function of this protein results in demyelination as communication between glial cells and neurons is disrupted (Abrams et al., 2002, 2003, 2001; Neuberg and Suter, 1999). GjB1 mutation leads to a form of X-linked CMT disease (CMTX1) (Bergoffen et al., 1993; Fairweather et al., 1994; Ionasescu et al., 1994; Schiavone et al., 1996). PRX is a Schwann cell-specific protein that has a role in axon–glial interactions and is expressed in a developmentally regulated manner (Gillespie et al., 1994; Scherer et al., 1995). PRX is important for the maintenance of peripheral nerve myelin and, in particular, for ensheathing regenerating axons (Gillespie et al., 1994; Scherer et al., 1995). A mouse model lacking functional PRX exhibits morphological changes in the neuromuscular junction. In particular, the terminal portion of peripheral motor axons shows extensive pre-terminal branches in demyelinated regions and axonal swelling, associated with asynchronous failure of action potential transmission at high stimulation frequencies (Court et al., 2008). Mutations in PRX cause CMT4F and Déjèrine–Sottas neuropathy (Boerkoel et al., 2001; Guilbot et al., 2001; Kabzinska et al., 2006; Kijima et al., 2004; Marchesi et al., 2010; Takashima et al., 2002).

4.8. Other defects: PRPS1 and ARHGEF10

Although more than 30 genes have been shown to be associated with different forms of CMT neuropathy, the disease-gene of several forms has yet to be identified. Furthermore, the exact nature of the involvement of some identified disease-genes is still unclear. For these genes, when possible, the putative molecular mechanisms underlying the disease are discussed.

4.8.1. PRPS1

Mutations in PRPS1 (phosphoribosylpyrophosphate synthetase 1) cause a number of different syndromes, one of which is an X-linked form of CMT disease termed CMT5 or Rosenberg-Chutorian syndrome (de Brouwer et al., 2010). Patients show peripheral demyelination and axonal loss. PRS1 is an enzyme required for nucleotide biosynthesis. There are a number of hypotheses regarding how mutations in PRS1 can affect peripheral neurons (de Brouwer et al., 2010). One hypothesis is linked to the fact that for myelin biosynthesis, lipid esters of nucleotides are required as well as S-adenosylmethionine as a co-factor (de Brouwer et al., 2010). Thus, mutations harming PRS1 would reduce the amount of nucleotides, thereby affecting myelination. Another hypothesis is that mutations would decrease the amount of GTP that is required by a number of proteins that regulate membrane traffic or cytoskeletal dynamics, as, for instance, Rab proteins, dynamins/dynamin-like proteins and the Rho GTPase family of actin regulators (de Brouwer et al., 2010). In this case, mutations in PRS1 would affect membrane traffic. However, there is currently no evidence for this hypothesis nor for other alternative hypotheses.

4.8.2. ARHGEF10

ARHGEF10 (Rho guanine nucleotide exchange factor 10) is a GEF for members of the Rho superfamily of small GTPases involved in the regulation of the actin cytoskeleton (Mohl et al., 2006). A mutation in ARHGEF10 causes a mild dominant intermediate form of CMT disease, characterized by slowed nerve conduction velocities and thin myelination (Verhoeven et al., 2003b). The phenotype is not progressive, suggesting that ARHGEF10 is involved in myelination during development (Verhoeven et al., 2003b). ARHGEF10 is thus a regulator of the actin cytoskeleton, and although there is no evidence supporting this hypothesis, mutations in ARHGEF10 could alter actin-dependent membrane traffic events.

4.9. Defects not directly related to trafficking: aminoacyl-tRNA synthetases, LMNA, BSCL2, TRPV4, CTDP1 and HK1

4.9.1. Aminoacyl-tRNA synthetases

Mutations in glycyl-, tyrosyl- and alanyl-tRNA synthetases (GARS, YARS and AARS) cause the autosomal dominant CMT2D, DI-CMTC and CMT2M forms, respectively (Antonellis et al., 2003; Jordanova et al., 2006; Latour et al., 2010). Aminoacyl-tRNA synthetases (ARS) catalyze the transfer of amino acids onto the appropriate tRNA during translation. Mutations of tRNA synthetases cause neurodegeneration and are responsible for other neurological diseases such as spinal cord disorders, leukoencephalopathy and distal spinal muscular atrophy (Antonellis and Green, 2008; Edvardson et al., 2007). The fact that peripheral neurons are affected in a cell-autonomous manner indicates that this type of neuron is more sensitive to protein translation defects. Although the mechanism by which mutations in these genes cause peripheral neuropathies is unknown, several hypotheses have been put forward (Antonellis and Green, 2008). The mutations could affect the ability of the enzymes to charge the amino acids on tRNAs or could alter their intracellular localization (Antonellis and Green, 2008). However, a recent study utilizing mouse models established that mutations in ARS do not cause peripheral neuropathies through amino acid mischarging or through a defect in their known functions in translation (Antonellis and Green, 2008; Stum et al., 2011). This suggests that the mutations could affect RNA charging occurring specifically in axons, leading to neurodegeneration (Antonellis and Green, 2008; Stum et al., 2011). Alternatively, the mutations could affect non-canonical functions of ARS. In this respect, it is worth noting that ARS possess additional functions not directly related to their canonical function. For example, some ARS are involved in transcription silencing, inflammatory responses, signaling or apoptosis (Antonellis and Green, 2008; Park et al., 2005, 2009).

4.9.2. LMNA

The LMNA gene encodes the lamin A/C nuclear envelope protein and is mutated in CMT2B1 and a number of other diseases, including Emery–Dreifuss muscular dystrophy and cardiomyopathy (Bonne et al., 1999; De Sandre-Giovannoli et al., 2002; Fatkin et al., 1999; Worman et al., 2010). Lamins are intermediate filament proteins that form the nuclear lamina, which provides the nuclear envelope and nuclear components with structural support. They are important for DNA replication, gene expression, nuclear transport, apoptosis and signaling (Hutchison and Worman, 2004). The A-type lamins are important in the protection of the cell from mechanical damage. Thus, mutations in LMNA could negatively influence this protection, leading to neuronal (axonal) degeneration (Hutchison and Worman, 2004; Niemann et al., 2006).

4.9.3. BSCL2

BSCL2 (Berardinelli-Seip congenital lipodystrophy 2) protein, also called seipin, has been found mutated in autosomal dominant axonal CMT2D and Silver syndrome, as well as in a number of other disorders including spastic paraplegia (Ito and Suzuki, 2009; Windpassinger et al., 2004). Seipin is a transmembrane protein localized to the ER and degraded by the ubiquitin–proteasome system (Ito and Suzuki, 2009; Windpassinger et al., 2004). Mutants of seipin induce ER stress-mediated cell death, suggesting that ER stress could be the cause of neurodegeneration (Ito and Suzuki, 2009). Seipin is known to regulate adipocyte differentiation, lipid droplet formation and motor neuron development (Fei et al., 2011). Furthermore, seipin mutants induce the formation of aggregates that could lead to degeneration (Ito and Suzuki, 2009; Windpassinger et al., 2004). The exact roles of seipin and of its disease-causing mutants remain to be elucidated.

4.9.4. TRPV4

TRPV4 (transient receptor potential cation channel subfamily V member 4) is a member of the TRP superfamily of cation channels. Mutations in the TRPV4 gene cause CMT2C, skeletal dysplasia and scapuloperoneal muscular atrophy (Deng et al., 2010; Landouré et al., 2010). TRPV4 plays a key role in osmosensation, temperature sensation and mechanosensation (Köttgen et al., 2008). Although TRPV4 is poorly expressed in neurons, alterations of this protein are highly toxic in neuronal cells, causing important changes in calcium concentrations. Thus, several neuronal processes, such as neurite outgrowth and synaptic transmission, are affected, leading to neurodegeneration (Deng et al., 2010; Landouré et al., 2010).

4.9.5. CTDP1

A mutation in the CTDP1 (C-terminal domain phosphatase 1) gene causes congenital cataracts facial dysmorphism neuropathy syndrome (CCFDN) (Varon et al., 2003). One of the symptoms patients show is hypomyelination of peripheral nerves. The CTDP1 gene encodes the protein phosphatase FCP1, which is a component of the transcription machinery. The CCFDN-associated mutation affects a nucleotide in intron 6 causing aberrant splicing (Varon et al., 2003). FCP1 dephosphorylates a serine in the C-terminal domain of the largest RNA polymerase II subunit, regulating gene expression (Varon et al., 2003). It is not known why a mutation in FCP1 causes CCFDN. It is possible that defects in the expression of specific genes could cause this neuropathy, or alternatively the mutation could impair unknown functions of FCP1.

4.9.6. HK1

HK1 (hexokinase 1) is mutated in CMT4G (also referred to as hereditary motor and sensory neuropathy-Russe, HMSNR) (Hantke et al., 2009). Hexokinases are sugar kinases that catalyze the phosphorylation of glucose, the first step in glucose metabolism. HK1 binds to mitochondria and it is the major regulator of the production of ATP by the cell's energy metabolism (Wilson, 2003). On mitochondria, HK1 is also involved in the regulation of cell survival. Mutations in HK1 cause hexokinase deficiency and severe nonspherocytic hemolytic anemia (Hantke et al., 2009). Interestingly, HK1 is highly expressed in the nervous system, in particular at the level of dorsal root ganglia, and it is involved in NGF-mediated neurite outgrowth (Hantke et al., 2009). The molecular mechanism underlying CMT4G is currently unknown; however, it could involve alterations of apoptotic activity or alterations of unknown alternative functions of HK1 specific for the peripheral nervous system (Hantke et al., 2009).

5. Conclusions and future directions

Several neuropathies are caused by functional alterations of intracellular traffic proteins. We have reviewed here the known genetic causes of CMT neuropathy, highlighting genes with functions related, directly or indirectly, to intracellular trafficking. Interestingly, many CMT disease-associated mutations alter genes involved in the regulation of the endomembrane system, strongly suggesting ‘problems in intracellular trafficking’ to be a major cause of CMT neuropathy (Fig. 4). Although each gene has more than one role in the cell, these genes can be grouped on the basis of the function most likely to be altered and, thus, most likely to constitute the basis of the molecular mechanism of action underlying the neuropathy (Table 2). CMT disease-associated mutations in the proteins indicated in Table 2 alter the regulation of PI metabolism, cytoskeletal organization and transport, endosomal trafficking, protein degradation, mitochondrial dynamics and mitochondrial axonal transport (Table 2; Fig. 4). However, other CMT disease-associated proteins could also be involved in the regulation of membrane traffic, although indirectly. For example, defects in PRS1 could alter the availability of nucleotides and thus affect the Rab and/or Rho GTPase cycle, impairing membrane traffic or cytoskeleton organization. Altered ARHGEF10 function could affect the regulation of the actin cytoskeleton and thus, indirectly, trafficking. To date, more than 40 different genes have been linked to CMT neuropathy. In the near future, due to the introduction of next generation sequencing systems, it is very likely that many other genes – which may be mutated in only a few individuals – will be identified.

Table 2.

Functions affected by CMT disease-associated mutations.

Function	Protein	Domains	Interactors
Regulation of membrane trafficking: vesicle fission	DNM2 (also regulates actin cytoskeleton)	GTPase, MD, PH, GED, PRD	Abp1, amphiphysins, CIN85, cortactin, γ-tubulin, microtubules, Rab7, SNX9, syndapin 2
Regulation of membrane trafficking: polyphosphoinositide phosphatases	MTMR2 MTMR13 FIG4	PH-GRAM, PTP, CC, PDZ-BD PH-GRAM, PTP, CC, PDZ-BD, DENN, PH SAC1	Dlg1, MTMR13, MTMR5, NEFL, PSD-95, Vps34/Vps15 MTMR2 ArPIKfyve, PIKfyve
Cytoskeletal transport Regulation of cytoskeletal organization and maintenance	KIF1Bβ NEFL FGD4/Frabin	PH Head, Rod, Tail FYVE, PH, FAB, DH	DENN/MADD MTMR2 Actin
Regulation of membrane trafficking: endosomal maturation	Rab7 (also regulates protein degradation)	GTPase	FYCO1, PRA1, retromer, RILP, TrkA, Vps34/Vps15
Regulation of membrane trafficking: endosomal recycling	NDRG1 SH3TC2	α/β hydrolase SH3, TPR	APOA1, APOA2, E-cadherin/catenin, PRA1, Rab4a Rab11
Regulation of protein degradation: chaperone-like activity	HSP22/HSPB8 HSP27/HSPB1 (also regulates actin cytoskeleton)	α-Crystallin α-Crystallin	αB-crystallin, Bag3, cvHSP, HSPB3, HSP20, HSP27, MKBP Actin, HSP22, intermediate filaments
Regulation of protein degradation: ubiquitin ligase	LRSAM1	LRRs, CC, ERM, SAM, RF, PTAP	Tsg101
Regulation of protein degradation: ubiquitin-mediated lysosomal degradation pathway	LITAF/SIMPLE	RF, PPXY, P(S/T)AP	Itch, Nedd4, Tsg101
Regulation of mitochondrial dynamics and mitochondrial axonal transport	MFN2 GDAP1	GTPase, CC GST	Miro/Milton

It is puzzling that, despite the large variety of genes involved, some of which are even ubiquitous, all patients with CMT neuropathy show similar and specific defects, mainly limited to peripheral motor and sensory neurons. As alterations in a number of different genes lead to similar phenotypes in CMT disease patients, we hypothesize that defects in the potential to regenerate axons following injury could be responsible for the different forms of CMT neuropathy. In this respect it is worth noting that, in contrast to peripheral neurons, neurons in the adult CNS do not regenerate their axons following injury (Huebner and Strittmatter, 2009). Thus, all the genes causing CMT disease could be important regulators of axonal regeneration, explaining why mainly peripheral neurons are affected when the mutated gene encodes a ubiquitous protein. Indeed, ubiquitous proteins may have a specific additional role in the regeneration of axons or in other neuronal-specific processes. In addition, the large variety of genes involved in the disease could be a reflection of the fact that axonal regeneration is a very complex process under the control of a number of other cellular processes. First of all, axonal maintenance and regeneration depend heavily on both myelinating and non-myelinating Schwann cells that respond to a number of different neurotrophic factors which signal to transcription factors (Bhatheja and Field, 2006). Thus, alterations of Schwann cell function are responsible for a number of neuronal disorders, including CMT neuropathy. A number of different signal transduction pathways, initiated mainly by neurotrophic factors and acting not only on Schwann cells but also directly on neurons, are responsible for the correct maintenance and regeneration of axons (Cui, 2006; Tucker and Mearow, 2008). Peripheral neurons are capable of spontaneous axon regeneration, but this property is strictly reliant on signaling pathways (Cui, 2006; Tucker and Mearow, 2008). Signaling pathways control a number of other important cellular processes, such as autophagy, neurite outgrowth and membrane traffic, that influence greatly the ability of the axon to regenerate.

Defects identified in CMT disease include dysregulation of PIs (see Section 4.2), Rab7 (see Section 4.4.1) and HSP22 (see Section 4.5.1), molecules that are involved in the regulation of the autophagy process. Autophagy dysfunction contributes to various neurodegenerative disorders, as both defective and excessive autophagy lead to neurite degeneration and neuronal atrophy (García-Arencibia et al., 2010; Gumy et al., 2010; Rubinsztein et al., 2005; Tooze and Schiavo, 2008). In particular, autophagy appears to be important for axonal maintenance and regeneration. In addition, neurons present a constitutive autophagy process that shows peculiar features and, possibly, molecular mechanisms not common to other cell types (Komatsu et al., 2007; Yue et al., 2008). Thus, alterations of neuronal-specific autophagy events could explain the clinical features of CMT disease patients, considering also that axonal regeneration occurs mainly in the peripheral nervous system (Huebner and Strittmatter, 2009).

DNM (see Section 4.1), Rab7 (see Section 4.4.1), HSP22 and HSP27 (see Section 4.5.1), TRPV4 (see Section 4.9.4) and HK1 (see Section 4.9.6) are involved in the control of neurite outgrowth and are mutated in CMT neuropathy. Alterations of neurite outgrowth impair efficient axonal regeneration (Raivich and Makwana, 2007; Rishal and Fainzilber, 2010). Neurite outgrowth occurs mainly during development; however, development is not impaired in CMT disease patients. Thus, we hypothesize that defective neurite outgrowth specifically affects axonal regeneration of peripheral nerves in CMT disease patients. How can this be explained? Defects affecting axonal development and regeneration in CMT disease patients are initially efficiently counteracted by other factors that become less effective with age. Indeed, it is known that older animals are less successful in axonal regeneration, although the molecular bases for these changes are not yet understood (Hilliard, 2009; Perlson et al., 2004; Wu et al., 2007). The age-related decline in the capacity of peripheral neurons to regenerate their axons due to defects in neurite outgrowth may not be strong enough to have an effect on development, thus explaining why CMT disease-causing mutations affect mainly peripheral neurons and why the onset of CMT neuropathy is often in the second to third decade of life.

Axon repair, axon growth and axon regeneration are processes that are dynamically dependent on cytoskeletal reorganization and intracellular trafficking events (Bloom and Morgan, 2011). For correct axon regeneration, it is important to maintain axon polarity, to initiate growth cone formation, and to promote outgrowth and correct synapse formation. All these steps require iterative events of endocytosis and exocytosis and extensive cytoskeletal reorganization (Bloom and Morgan, 2011). Thus, the CMT disease-causing proteins involved in intracellular trafficking and cytoskeletal organization (Table 2; Fig. 4) could affect axonal regeneration by affecting one or more intracellular trafficking events necessary for this process.

In the near future, we expect to see an increase in the number of identified genes that cause CMT neuropathy. It will be interesting to establish whether other intracellular traffic genes will be identified as causative of CMT disease. It will be important to investigate, at the molecular level, the mechanism by which each gene contributes to the disease in order to start to identify possible therapeutic agents. In this respect, it will be valuable to continue to individuate the genes whose defects affect primarily Schwann or neuronal cell function in order to identify the correct target.

The limited knowledge on the molecular mechanism underlying the different forms of the disorder is a consequence of a lack of animal models. Clearly, the availability of mouse, rat or monkey models would greatly contribute to our knowledge of the disease. However, we should also consider using simpler models that could generate rapid and straightforward answers. For instance, as has been the case for other disorders, the use of Drosophila models or, even better, the vertebrate zebrafish could help identify the exact role in the disease of known genes with multiple cellular functions.

The complexity of the endomembrane system together with the multitude of mutated genes that are causative factors for CMT neuropathy makes it difficult to envisage a common cure/intervention of the disease process. However, further dissection of the molecular mechanism of action of every single gene involved, together with a better general understanding of intracellular trafficking in neurons and an improved knowledge on how alterations of intracellular trafficking specifically affect peripheral neurons should open up a way for the development of specific therapeutic strategies against specific CMT disease targets.

Conflict of interest

The authors have no conflict of interest.