|Home | About | Journals | Submit | Contact Us | Français|
The classic myelin basic protein (MBP) splice isoforms range in nominal molecular mass from 14 to 21.5 kDa, and arise from the gene in the oligodendrocyte lineage (Golli) in maturing oligodendrocytes. The 18.5-kDa isoform that predominates in adult myelin adheres the cytosolic surfaces of oligodendrocyte membranes together, and forms a two-dimensional molecular sieve restricting protein diffusion into compact myelin. However, this protein has additional roles including cytoskeletal assembly and membrane extension, binding to SH3-domains, participation in Fyn-mediated signaling pathways, sequestration of phosphoinositides, and maintenance of calcium homeostasis. Of the diverse post-translational modifications of this isoform, phosphorylation is the most dynamic, and modulates 18.5-kDa MBP’s protein-membrane and protein-protein interactions, indicative of a rich repertoire of functions. In developing and mature myelin, phosphorylation can result in microdomain or even nuclear targeting of the protein, supporting the conclusion that 18.5-kDa MBP has significant roles beyond membrane adhesion. The full-length, early-developmental 21.5-kDa splice isoform is predominantly karyophilic due to a non-traditional P-Y nuclear localization signal, with effects such as promotion of oligodendrocyte proliferation. We discuss in vitro and recent in vivo evidence for multifunctionality of these classic basic proteins of myelin, and argue for a systematic evaluation of the temporal and spatial distributions of these protein isoforms, and their modified variants, during oligodendrocyte differentiation.
The ‘classic’ 18.5-kDa isoform of myelin basic protein (MBP) was first isolated more than 50 years ago as an encephalitogenic determinant (Roboz-Einstein et al. 1962; Carnegie and Lumsden 1966). This protein is an integral component of central nervous system (CNS) myelin, adhering the cytoplasmic leaflets of the oligodendrocyte (OLG) membrane to each other to form the major dense line observed in electron micrographs (Trapp and Kidd 2004). This fundamental role is demonstrated by the shiverer mutant mouse which has an ablation in the MBP gene, and whose CNS myelin is sparse and relatively unstructured (Dupouey et al. 1979; Privat et al. 1979). The compact myelin phenotype can be “rescued” in these mice transgenically (Popko et al. 1987; Readhead et al. 1987; Campagnoni and Macklin 1988; Kimura et al. 1989; Readhead and Hood 1990). The 18.5-kDa isoform, which predominates in adult CNS myelin is thus generally considered to be essential for its development and stability (Readhead et al. 1990; Fitzner et al. 2006; Simons and Trotter 2007), and forms a ‘molecular sieve’ that restricts diffusion of some membrane proteins from paranodal loops into compact myelin (Pedraza et al. 2001; Aggarwal et al. 2011a, b; Simons et al. 2012). This protein is only a minor component of peripheral nervous system (PNS) myelin. The 18.5-kDa MBP has been termed an ‘executive protein’ because of its necessity for CNS myelin formation (Moscarello 1997), unlike other structural myelin proteins such as proteolipid protein (PLP) and myelin-associated protein (MAG). On that note, this review could end here.
But really the situation is not so simple. To this day, the determination of MBP’s ‘structure’ and ‘function’ continues, in part because it is not a single protein. The basic proteins of myelin constitute an extraordinarily varied family of developmentally regulated splice isoforms with myriad, combinatorial, post-translational modifications. The variety of small ligand and protein-protein interactions is equally diverse. These phenomena are indicative of more than mere membrane adhesion, and the complex structure-function relationships of these proteins have been reviewed and speculated on numerous occasions, especially within the last decade (Kies et al. 1972; Kies 1982; Moscarello 1990; Smith 1992; Kursula 2001, 2008; Campagnoni and Campagnoni 2004; Harauz et al. 2004, 2009; Tzakos et al. 2005; Boggs 2006, 2008a; DeBruin and Harauz 2007; Harauz and Libich 2009). Here, again, we return to a question posed already over 30 years ago (Martenson 1980): ‘What does myelin basic protein do?’.
In the CNS, myelin arises from oligodendrocytes (OLGs), a highly plastic lineage (Baumann and Pham-Dinh 2001; Barres 2008; Bradl and Lassmann 2010; Miron et al. 2011). The OLGs proceed through a regulated pathway culminating in the assembly of the components of the myelin membrane (Pfeiffer et al. 1993; Miller 1996), beginning with differentiation of the bipolar early oligodendrocyte progenitor cell (OPC) into a multipolar late progenitor (pro-oligodendroblast) that expresses pro-oligodendroblast antigen. The onset of terminal differentiation marks the immature OLG stage; the cells now synthesize the glycolipids galactosylceramide (GalC) and sulfatide, and express the proteins 2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNP) and MAG. The OLG is at the mature stage when classic MBP isoforms and PLP are expressed, and extensive processes form and extend around an axon.
The developmental changes of OLGs are context-specific, involving homo- and hetero- cell-cell interactions (Sherman and Brophy 2005; Quarles et al. 2006; Trajkovic et al. 2006; Emery 2010; Stys 2011; Wake et al. 2011; Yang et al. 2011). The OPCs first migrate from the subventricular zone to all areas of the brain and spinal cord, and then proliferate before they differentiate further, with copious synthesis of components for and extension of membrane processes. These membrane extensions make physical contact with axons, exchange signals, and then form sheets that ensheathe the axon. Compact myelin is formed by flattening of the multiple, spirally wrapped lamellae with the extrusion of cytoplasm, a process modeled in various ways, e.g., the ‘liquid croissant’ and ‘corkscrew’ models (Bauer et al. 2009; Sobottka et al. 2011; Ioannidou et al. 2012; Simons et al. 2012). Myelin continues to be formed until the early twenties in humans (Fields 2005), and there are gender differences in OLG physiology (Cerghet et al. 2006). In demyelinating diseases such as multiple sclerosis, inherent attempts at repair prove futile because of the inability of the brain to replicate the proper early developmental signaling cues (Franklin 2002; Kuhlmann et al. 2008; Miron et al. 2011), and strategies to enhance remyelination remain elusive (Dubois-Dalcq et al. 2005; Miller and Fyffe-Maricich 2010). Other white matter pathologies also require an intimate understanding of myelin (Matute and Ransom 2012). We discuss here, toward this end, the roles of the primary family of highly basic proteins of myelin, those that arise from the gene in the oligodendrocyte lineage (Golli).
The Golli gene complex has different transcription start sites expressing the Golli-specific and classic protein families at different stages of OLG differentiation (Fig. 1a) (Campagnoni et al. 1993; Pribyl et al. 1993; Givogri et al. 2001; Campagnoni and Campagnoni 2004; Rasband and Macklin 2012). The so-called Golli proteins (denoted BG21, J37, TP8 in the mouse) are expressed from transcription start site 1 in neuronal and myelin-producing cells (Landry et al. 1996, 1997, 1998), but also in macrophages and T-cells of the immune system (Kalwy et al. 1998; Feng et al. 2006). The highest levels of expression of Golli proteins are in OLGs at intermediate stages of differentiation (Givogri et al. 2001). The Golli isoforms have many functions depending on the tissue in which they are expressed, but in OLGs promote migration and process extension, and enhance potassium-induced calcium influx, a phenomenon that is dependent on plasma membrane targeting (Paez et al. 2007, 2009a, b; Fulton et al. 2010a, b). Transgenic mice with a knockout in the Golli protein isoforms only can still form myelin but with deficiencies in timing in vivo, and an inability of primary OLGs to elaborate membrane sheets in culture (Jacobs et al. 2005). Over-expression of Golli proteins results in severely abnormal myelination (Jacobs et al. 2009).
The classic MBP isoforms arise from transcription start site 3 of Golli, and are differently regulated than the Golli ones arising from transcription start site 1 (Fig. 1b) (Givogri et al. 2001; Campagnoni and Campagnoni 2004; Rasband and Macklin 2012). Their synthesis overlaps temporally with that of Golli proteins, and they begin to be synthesized in large quantities in more highly differentiated and mature OLGs. The classic basic proteins comprise splice isoforms ranging in nominal molecular mass from 14.0 to 21.5 kDa. [The situation is actually much more complex than stated here, as a unique 14-kDa MBP population arises from transcription start site 2 (Kitamura et al. 1990), and 19.7-kDa and 21.0-kDa isoforms have also been reported (Aruga et al. 1991; Nakajima et al. 1993), to give just two examples]. The intron-exon structures of classic isoforms of human MBP were identified by cDNA cloning before the discovery of Golli (de Ferra et al. 1985; Kamholz et al. 1986; Campagnoni and Macklin 1988). The designation ‘MBP’ usually refers to the classic 18.5-kDa isoform which predominates in the adult CNS, being peripheral membrane-associated and found in cell processes and compact myelin. This isoform is extremely positively charged (+19 at neutral pH), with a myriad of post-translational modifications, primarily deimination, phosphorylation, and methylation (Table S1). Here, our focus will be on the classic 18.5-kDa and 21.5-kDa isoforms of MBP, which differ by a 26-residue segment encoded by classic exon-II (Fig. 1b). The 18.5-kDa isoform is representative of peripheral membrane-associated ones like the 14-kDa protein; indeed, the latter occurs in high proportion in rodent compact myelin, but not in human or bovine brain. The exon-II-containing murine 21.5-kDa and 17.22-kDa isoforms share similar nuclear-trafficking patterns.
Structurally, all isoforms of MBP are intrinsically disordered proteins (IDPs) (Hill et al. 2002; Harauz et al. 2004). This property means that they have a higher proportion of charged, and lower proportion of hydrophobic amino acids, than classically folded proteins with a compact core, and with clear ‘lock and key’ mechanisms of function, as we envisage with enzymes (Tompa 2009). In general, IDPs are highly extended and flexible, interact with a variety of binding partners, and often act as ‘hubs’ in structural or signaling networks (Uversky 2010). The interactions of IDPs with other biomolecules involve local disorder-to-order transitions, either by induced fit, or fixation of transient fluctuations that form recognizable target motifs (Tompa et al. 2009; Uversky 2011; Disfani et al. 2012). The varied and multiple interactions of IDPs are still specific, even when they are sometimes weak (Hsu et al. 2012; Das et al. 2013).
The 18.5-kDa MBP isoform is a quintessential IDP – it has long been known to be highly flexible, behaving like a randomly coiled polymer in aqueous buffer (Moscarello et al. 1973; Martenson 1978). It gains ordered secondary structure in association with detergents and lipids, consistent with its classification as a peripheral membrane protein [reviewed in (Harauz et al. 2004; Polverini et al. 1999; Smith 1992)]. We know now that as an IDP, MBP’s extended nature allows it to associate rapidly with various membrane surfaces, other proteins, or small ligands, undergoing local conformational transitions as it does so. These interactions are summarized in Table S2. Using biophysical techniques such as fluorescence, electron paramagnetic resonance (EPR), and nuclear magnetic resonance (NMR) spectroscopy, we continue to map the specific targets and partial, induced disorder-to-order transitions of this protein as it forms complexes with various surfactant and protein partners (Fig. 2). Although highly flexible in vitro, 18.5-kDa MBP comprises several α-helical or poly-proline type II (PPII) molecular recognition fragments (MoRFs) (Harauz et al. 2004; Harauz and Libich 2009; Libich et al. 2010); cf., (Mittag et al. 2009). Nevertheless, in any reconstituted system that we have so far studied in vitro, the protein structure remains fuzzy. This concept was first introduced and is described more fully in (Tompa and Fuxreiter 2008; Fuxreiter and Tompa 2012). For 18.5-kDa MBP, it means that the protein in situ in any reconstituted complex still has many segments that are mobile and available for further interactions with other surfaces or proteins (Fig. 2f) (Harauz and Libich 2009; Harauz et al. 2009; Libich et al. 2010).
The main function of 18.5-kDa MBP – membrane adhesion – has been a subject of investigation for decades (Boggs et al. 1982), and investigators continue to dissect molecular details of how this protein is positioned within two membrane bilayers to form the major dense line of myelin. The inner leaflet of the myelin membrane, where MBP is located, has a different composition than the outer leaflet, and comprises roughly 44% cholesterol, 27% phosphatidylethanolamine, 13% phosphatidylserine, 11% phosphatidylcholine, 3% sphingomyelin, and 2% phosphatidylinositol, as estimated from biophysical measurements of surface charge (Inouye and Kirschner 1988; Taylor et al. 2004; Rasband and Macklin 2012). The MBP-lipid interactions rely on a balance of interactions between the basic residues of MBP and the acidic headgroups of the lipid bilayer, to assemble the proper multilamellar structure seen in myelin sheaths (Boggs et al. 1982; Jo and Boggs 1995; Hu et al. 2004; Rispoli et al. 2007; Min et al. 2009). Of note also is the relatively high proportion of phosphoinositides, particularly phosphatidylinositol-4,5-bis-phosphate (PI(4,5)P2). In vitro and in cellula, MBP has been shown to sequester PI(4,5)P2 inmembranes (Musse et al. 2008a, 2009; Harauz et al. 2009; Nawaz et al. 2009). Cholesterol is also important in myelin formation and stabilization (Saher et al. 2005, 2011), and specifically facilitates 18.5-kDa MBP’s ability to stack membranes (Suresh et al. 2010).
The molecular details of how 18.5-kDa MBP molecules interact with the lipid bilayer and with each other within the sheath are still being deciphered. When associated with a phospholipid membrane, MBP interacts via at least three amphipathic α-helical regions, as ascertained by EPR and NMR spectroscopy (Fig. 2a and e) (Bates et al. 2004; Libich and Harauz 2008; Homchaudhuri et al. 2010; Bamm et al. 2011). The three-dimensional conformation of MBP even when sandwiched between lipid bilayers is dynamic, with large segments mobile and free for interaction with other partners – a ‘fuzzy’ ensemble of conformations (Fig. 2e and f) (Zhong et al. 2007; Libich et al. 2010). With our present knowledge, we have depicted schematic models of adhesion in which the N-terminal and C-terminal segments of 18.5-kDa MBP interact independently, like a hairpin or paperclip, with separate apposed leaflets of the myelin membrane [see Figure 8 in reference (Homchaudhuri et al. 2010), and Figure 1 in reference (Ahmed et al. 2010)], in a major dense line of roughly 3 nm thickness. The electrostatic interaction of MBP with the lipid surface neutralizes the negative surface charge and decreases the hydration of the lipid polar head groups (Homchaudhuri et al. 2010), allowing close adhesion of these surfaces (Min et al. 2009, 2011).
In this context, two phenomena have been probed to form a better picture of myelin architecture at the molecular level: (i) self-assembly of MBP (Smith 1980, 1982, 1985; Beniac et al. 2000; Ishiyama et al. 2001; Kattnig et al. 2012), and (ii) the effects of specific, additional interactions with divalent metal ions like Zn2+ and Cu2+ (Earl et al. 1988; Berlet et al. 1994; Riccio et al. 1995; Baran et al. 2010; Bund et al. 2010; Smith et al. 2010). Cell biological studies involving immunofluorescence microscopy have suggested the idea of MBP forming a ‘molecular sieve’ on the myelin membrane, implying the formation of a two-dimensional network of the protein (Aggarwal et al. 2011a, b). Independent and contemporary biophysical studies involving double electron-electron resonance (DEER) spectroscopy have been used to determine the distance distribution characterizing the mutual orientation of adjacent proteins in locally enriched regions of 18.5-kDa MBP on a membrane surface (Kattnig et al. 2012). Using these data, we have modeled the non-specific, lateral self-assembly process as a distribution of ellipsoids in two dimensions [ibid.]. This planar array could explain how MBP forms a size barrier that restricts lateral diffusion of membrane proteins with large cytosolic domains from the paranodal loops into compact myelin (Pedraza et al. 2001; Aggarwal et al. 2011b).
The strong plasma membrane association of 18.5-kDa MBP requires careful spatial organization of its synthesis in myelinating OLGs, and is achieved as follows. The mRNA for the 18.5-kDa classic isoform comprises a minimal 21-nucleotide 3′-untranslated region (3′UTR), which increases the translational efficiency of the mRNA (Ueno et al. 1994a, b), and which causes it to be trafficked to the peripheral processes of developing OLGs (Ainger et al. 1997). The translocation is believed to aid in the immediate insertion of the highly positively charged MBP directly into the OLG processes forming compact myelin, and preventing it from interacting deleteriously with the plasma membrane of the cell body or organellar or nuclear membranes (Hardy et al. 1996; Carson et al. 1998; Barbarese et al. 1999; Maier et al. 2008). [Free, cytoplasmic MBP has been demonstrated to cause aberrant internal membrane fusion in pancreatic islet cells, for example (Kolehmainen and Sormunen 1998)]. It has also been suggested that the strong association of the basic protein with PI(4,5)P2 plays a further role in its peripheral membrane-targeting (Nawaz et al. 2009). This differential targeting of MBP mRNAs to the plasma membrane is essential for establishing and maintaining OLG polarity (Baron and Hoekstra 2010; Simons et al. 2012).
In addition to its interaction with lipids, MBP also interacts with a variety of proteins (Table S2). The in vitro and some in vivo evidence for these interactions has been extensively reviewed in references (Harauz et al. 2004, 2009; Boggs 2006, 2008a, b; Galiano et al. 2008; Bauer et al. 2009; Harauz and Libich 2009). Some of these complexes involve electrostatic interactions, and have been considered to be non-specific because of the high positive charge of this protein. However, electrostatic interactions with other proteins are an important property of IDPs that allows them to form signaling hubs, and are used by other IDPs to ‘accelerate’ coupled binding and folding to other proteins (Dagliyan et al. 2011; Dogan et al. 2012; Ganguly et al. 2012). We have argued that as MBP is so abundant in OLGs, it is able to participate in these interactions even though other more ubiquitous but numerically less plentiful proteins are present which can also do so. More recently, we have utilized transfected N19-oligodendroglial cells to show that these interactions also occur in cellula and have a physiological role there. The conditionally immortalized N19-OLG cell line is valuable for exploring MBP’s multifunctionality in glia, because it closely resembles an immature OLG, before MBP begins to be synthesized in large quantities (Foster et al. 1993, 1995; Verity et al. 1993; Byravan et al. 1994).
Oligodendrocytes in the CNS undergo dramatic morphological changes during myelination. Time lapse imaging of myelin formation has revealed that OLG processes contact and retreat from axons repeatedly (Asou et al. 1994). When a contact is established, they wrap spirally around the axon, and then widen until the internodal axon surface is covered with a membrane layer (Ioannidou et al. 2012). Compaction and elimination of cytosol occurs after that stage. During these dynamic events, OLGs must remodel their internal microfilament and microtubule networks continuously (Kramer et al. 2001; Richter-Landsberg 2001, 2008; Song et al. 2001; Sherman and Brophy 2005; Fitzner et al. 2006; Simons and Trotter 2007; Bauer et al. 2009). Several lines of in vivo and in vitro evidence implicate the MBPs as additional regulators of the glial cytoskeleton, in addition to their main function of maintaining the spacing of myelin lamellae.
First, it has been demonstrated in vitro that MBP polymerises and bundles actin, both associations being modulated by post-translational modifications and binding to calmodulin (Fig. 2c and d) [(Bamm et al. 2011; Boggs et al. 2005, 2006; Boggs and Rangaraj 2000; Hill and Harauz 2005), and further references in Table S2]. The protein also links actin microfilaments to the lipid membranes (Boggs and Rangaraj 2000; Boggs et al. 2005, 2006, 2012; Boggs 2006). Second, MBP behaves like a microtubule-associated protein (MAP), and has many similarities to other MAPs, such as tau. Classic MBP isoforms polymerise and bundle tubulin in vitro (Hill et al. 2005), bind microtubules to lipid membranes in similar ways as they do microfilaments (Boggs et al. 2011), stabilize microtubules in OLGs (Galiano et al. 2006, 2008; Galiano and Hallak 2008), and cross-link microfilaments and microtubules to each other in vitro (Boggs 2006; Boggs et al. 2011). These interactions involve electrostatic associations and can be regulated by post-translational modifications (phosphorylation and deimination) that decrease the net positive charge of 18.5-kDa MBP, and by changes in lipid composition that increase the net negative surface charge of the lipid bilayer.
Studies with various classic splice isoforms and deletion mutants of MBP have failed to reveal distinct motifs that interact with the cytoskeletal proteins (Boggs et al. 2005; Hill and Harauz 2005; Hill et al. 2005). Solid-state NMR spectroscopy suggested that the entire polypeptide chain appears to be involved in the association with actin microfilaments (Fig. 2d) (Ahmed et al. 2009), consistent with the paradigm of intrinsically disordered proteins that function as springs or as linkers [and which are referred to as ‘entropic chains’ (Tompa 2002, 2005)]. In the case of 18.5-kDa MBP, this analogy means that the protein has a degree of flexibility to form a structural scaffold in OLGs. Such biophysical studies show that different segments of the protein can interact alternately with diverse binding partners such as actin and Ca2+-CaM, not solely one (Bamm et al. 2010, 2011). This ‘cross-talk’ and ‘moonlighting’ characteristic of IDPs does not preclude specificity of the interaction (Tompa et al. 2005). For example, the C-terminal domain of 18.5-kDa MBP interacts with membrane surfaces, with actin, and with Ca2+-calmodulin (Fig. 2), but the latter can displace MBP from its complexes with microfilaments (Boggs and Rangaraj 2000).
To investigate these associations in an in vivo context, we have recently examined the cytoskeletal interactions of classic 18.5-kDa MBP in early developmental N19-OLGs transfected with fluorescently tagged MBP, actin, and tubulin (Smith et al. 2012b). We showed that 18.5-kDa MBP redistributed to distinct ‘membrane-ruffled’ regions of the plasma membrane containing actin and tubulin and rapidly co-localized with them, when stimulated with PMA (phorbol-12-myristate-13-acetate), a potent activator of the protein kinase C pathway (Fig. 3). Moreover, using phospho-specific antibody staining, we showed an increase in 18.5-kDa MBP that was phosphorylated at Thr98 MBP (human 18.5-kDa sequence numbering – murine Thr95; see Table S1) in membrane-ruffled N19-OLGs. The MBP and actin also appeared simultaneously, and were co-localized in newly formed membrane domains resembling focal adhesion contacts induced by IGF-1 (insulin-like growth factor 1) stimulation in cells grown on laminin-2. This study thus supported a role for the classic 18.5-kDa MBP isoform in cytoskeletal as well as membrane remodeling in OLGs, as suggested by the many prior in vitro investigations (Table S2). Membrane domains enriched in cholesterol and MBP, and associated with cytoskeletal proteins, have been isolated from myelin by Triton X-100 extraction and sucrose density gradient centrifugation, suggesting also that these associations are pronounced in specific regions of the sheath [(Arvanitis et al. 2005); cf. (Gillespie et al. 1989; Wilson and Brophy 1989)].
Transient calcium influx is an integral part of events such as OPC proliferation and migration, and one of the more intriguing newly suggested MBP-protein functions may be its effect on calcium channels on the OLG plasma membrane. Pioneering study by Paez and colleagues in the group of Dr. Anthony Campagnoni had shown that the early developmental Golli isoforms stimulated Ca2+-influx into immortalized N19 and primary oligodendroglial progenitor cells in culture (Paez et al. 2009a, b, 2011; Fulton et al. 2010a). In analogous experiments, we further showed that the classic MBP isoforms, in contrast to Golli isoforms, decreased Ca2+-influx through voltage-operated Ca2+ channels (VOCC) in both cell populations (Smith et al. 2011). As for Golli isoforms, membrane-targeting of MBP was required to have the effect, suggesting that there is either a direct or indirect interaction of MBP with the channel, or with another protein which creates a signal that affects the channel. One possibility is simple stabilization of the cortical actin cytoskeleton by MBP (see Fig. 3), as observed for annexins (Monastyrskaya et al. 2009). As the Golli and classic MBP proteins arise from the same gene complex and are differentially expressed throughout OLG maturation, it is reasonable to propose that one of the roles of classic MBP may be to down-regulate intracellular Ca2+ concentrations following Golli-protein expression.
These investigations suggest that binding of classic MBP isoforms to the plasma membrane, and concomitant regulation of voltage-gated Ca2+-channels, is important also for maintaining low intracellular calcium concentrations in mature OLGs. Excess influx of calcium into OLGs results simply in cell death (Tzeng et al. 1995), but smaller amounts can have longer-term effects. Increased calcium levels would also activate calmodulin, which can bind one of several targets on 18.5-kDa MBP (Fig. 2c) (Libich et al. 2003a, b; Libich and Harauz 2008; Bamm et al. 2011), and thus regulate the protein’s associations with both cytoskeletal proteins (Boggs and Rangaraj 2000) and with the myelin membrane (Homchaudhuri et al. 2010). Dysregulation of this process could have potentially serious structural consequences. Moreover, activation of peptidylarginine deiminases by increased intracellular calcium levels would result in increased deimination of MBP, which may have a role in developmental myelination (Moscarello et al. 1994), but which could also contribute to structural destabilization of adult myelin (Harauz and Musse 2007; Musse et al. 2008c). This latter phenomenon may play a role in the pathogenesis of multiple sclerosis (Musse et al. 2006, 2008b; Moscarello et al. 2007; Mastronardi and Moscarello 2008; Koch et al. 2012). The sequestration of PI(4,5)P2 by MBP, has also been shown to be disturbed by increased calcium concentrations (Nawaz et al. 2009). Thus, the regulation of VOCCs by 18.5-kDa MBP could have a regulatory and protective role in many ways.
One of the highly conserved segments of mammalian 18.5-kDa MBP is a proline-rich region (murine sequence –T92 P93R94T95P96P97P98S99–) that contains a minimal XP-X-XP SH3-ligand domain and two mitogen-activated protein kinase (MAPK) phosphorylation sites (Table S1). We have previously shown that 18.5-kDa MBP binds to several SH3-domains in vitro, including those of Fyn, cortactin, and zonula occludens 1 (ZO-1) (Polverini et al. 2008; Smith et al. 2012a, b). Fyn is a member of the Src family of tyrosine kinases with important roles in OLG differentiation and myelination [reviewed in (Kramer-Albers and White 2011)]. Cortactin is an actin-binding protein that plays a role in regulation of actin dynamics in cell lamellipodia and ruffles (Ammer and Weed 2008). The scaffold protein ZO-1 is associated with numerous signaling proteins, tight and gap junctions, and the cytoskeleton (Fanning et al. 1998). It may be associated with gap junctions in OLGs (Li et al. 2004; Penes et al. 2005), and in the radial component of myelin, which contains other tight junction proteins (Kosaras and Kirschner 1990; Karthigasan et al. 1994; Arroyo and Scherer 2000; Devaux and Gow 2008; Gow and Devaux 2008). The target site on 18.5-kDa MBP was shown by CD spectroscopy to be able to form a canonical poly-proline type II (PPII) structure usually recognized by SH3-domains (Polverini et al. 2008; Harauz and Libich 2009). Moreover, 18.5-kDa MBP has been shown to tether the Fyn-SH3 domain to membranes in vitro in a manner regulated by surface charge, i.e., the association diminished with higher proportions of phosphoinositides in the bilayer (Homchaudhuri et al. 2009).
In our study of 18.5-kDa MBP’s cytoskeletal associations in N19-OLGs described above (Smith et al. 2012b), we observed additionally that this isoform co-localized with the SH3-domain-containing proteins cortactin and ZO-1, when stimulated with PMA (Fig. 3). We further used these N19-OLGs to demonstrate the interaction and the physiological role of the binding of 18.5-kDa MBP with the SH3-domain of Fyn, in cellula (Smith et al. 2012a). We focussed on Fyn because it is involved in a number of signaling pathways during OLG development and myelination (Umemori et al. 1994, 1999; Osterhout et al. 1999; Seiwa et al. 2000, 2007; Sperber and McMorris 2001; Sperber et al. 2001; Klein et al. 2002; Lu et al. 2005; White et al. 2008; Perez et al. 2009; Wake et al. 2011). In particular, Fyn has been postulated to be a key regulatory element of the myelination process that triggers phosphorylation of hnRNPA2 (heterogeneous nuclear ribonucleoprotein A2), responsible for efficient transport of MBP mRNA to the site of developing OLG processes in contact with neurons (Seiwa et al. 2000, 2007; White et al. 2008; Laursen et al. 2011).
Classic 18.5-kDa MBP variants with Pro-to-Gly substitutions to disrupt the structure of the PPII SH3-ligand motif, or Thr-to-Glu substitutions to mimic MAPK phosphorylation of Thr92 and Thr95, had lower affinity for the SH3-domain of Fyn in vitro, as measured by isothermal titration calorimetry (Smith et al. 2012a). Co-expression of classic 18.5-kDa MBP with a constitutively active form of Fyn-kinase, p59Fyn-Tyr527Phe, caused extensive membrane elaboration, and branching complexity at the forefront of extending N19-OLG membrane processes (Fig. 4), a phenomenon that was abolished by substituting either proline residue within its XP-X-XP SH3-ligand consensus motif. Over-expression of corresponding green fluorescent protein (GFP)-tagged MBP variants in cultured N19-OLGs expressing only endogenous Fyn produced aberrant elongation of membrane processes and increased branching complexity, in contrast to GFP-tagged wild-type MBP, which caused longer but unbranched processes. Immunostaining demonstrated co-localization of Fyn with 18.5-kDa MBP in the cell body, as well as in the process tips.
Furthermore, these substitutions to the XP-X-XP SH3-ligand-domain inhibited the ability of MBP to decrease L-type VOCC-mediated calcium-influx (Smith et al. 2012a). This effect could have been because of the reduced interaction of these variants with Fyn or other SH3-domain proteins. This altered function of MBP variants suggests that interaction of 18.5-kDa MBP with SH3-domain proteins such as Fyn is an important step in the modulation of OLG Ca2+-uptake by classic MBP isoforms. Phosphorylation of MBP might then result in an increase in calcium concentration in OLG processes. Local increases in calcium concentrations at the process extremities of developing OLGs would provide an explanation for the phenotypic differences in process extension and branching complexity observed in cells expressing pseudo-phosphorylated SH3-ligand variants of 18.5-kDa MBP.
These results in total suggest that MBP’s SH3-ligand domain plays a key role in intracellular protein interactions in vivo, and that these interactions may be required for proper membrane elaboration of developing OLGs (Fitzner et al. 2006; Simons et al. 2012). Specifically, a physiological role of MBP’s association with SH3-domain proteins such as Fyn is to increase process length preferentially to more complex morphological changes. Although the 18.5-kDa MBP splice isoform is produced in large quantities in OLG processes in situ in the brain only as they begin to ensheathe the axon (Butt et al. 1997; Barbarese et al. 1999), significant membrane production and process development still must occur before the axon is fully myelinated.
Consistent with previous studies of Fyn-tau interactions (Klein et al. 2002; Perez et al. 2009), the observed morphological changes in N19-OLG differentiation produced by MBP-Fyn interactions required the XP-X-XP SH3-ligand-domain of classic 18.5-kDa MBP isoforms. Phosphorylation of Thr92 and/or Thr95 in 18.5-kDa murine MBP could also be concluded to regulate this function of the protein. In addition to causing increased branching, the Thr92Glu pseudo-phosphorylation caused the 18.5-kDa MBP isoform to traffic in an oscillatory manner to the nucleus, but this effect was reversed in high-density cultures. Molecular dynamics simulations in silico indicate that phosphorylation at these sites can alter protein-protein interactions, modulate the local conformation of the protein, and the degree of penetration of the central membrane-anchoring segment into a lipid bilayer (Polverini et al. 2011). We have thus proposed that this region of the protein constitutes an important molecular switch (Fig. 5) (Harauz and Libich 2009; Harauz et al. 2009; Polverini et al. 2011), and disruption of it during development has been suggested to result potentially in myelin instability (Rubenstein 2008; Tait and Straus 2008; Bessonov et al. 2010, 2013).
The early developmental Golli isoforms were found shortly after their discovery to have nuclear-localized subpopulations in both neuronal cells (Landry et al. 1996; Pribyl et al. 1996) and in developing OLGs (Givogri et al. 2001). Using various deletion-constructs, it was determined that a non-traditional nuclear localization sequence within a 56-residue stretch of the classic MBP portion of the Golli-MBP was essential for nuclear targeting (Reyes and Campagnoni 2002). It was later shown that N-terminal myristoylation of Golli-MBPs resulted in plasma membrane targeting, which was essential for stimulating Ca2+-influx via VOCCs (Paez et al. 2007).
The classic MBP isoforms were known well before the unearthing of Golli to have intricate mRNA translocation and regulation properties (Colman et al. 1982). The microinjection of exogenous MBP mRNA into primary OLGs showed it to form granules in the perikaryon, which were then transported to the peripheral membrane processes where they became dispersed (Gillespie et al. 1990; Ainger et al. 1993; Brophy et al. 1993). Transport was determined to be dependent on a 21-nucleotide sequence termed the mRNA transport signal (RTS) in the 3′UTR (untranslated region) of MBP mRNA – this minimally sufficient RTS packaged the mRNA into ‘granules’ that were trafficked in a microtubule-dependent manner to the peripheral processes of developing OLGs (Ainger et al. 1997). The OLG cytoskeleton (microtubules and kinesin motors) was necessary for this transport (Barbarese et al. 1988; Carson et al. 1997). Disruption of this transport pathway has been shown to result in failure to form compact myelin, e.g., as seen in quaking viable mice in which the MBP is retained in the nucleus and perikaryon (Larocque et al. 2002). Other CNS proteins are similarly trafficked as part of a highly coordinated process in both OLGs and neurons (Smith 2004; Carson and Barbarese 2005; Carson et al. 2008).
But there is another factor for classic MBP isoforms that overrides the 3′UTR. The full-length classic 21.5-kDa transcript includes 26 residues encoded by exon-II (exon-6 in Golli numbering). Previously, the exon-II-containing MBP isoforms (21.5-kDa and 17.22-kDa in the mouse, Fig. 1b) had been shown to be expressed at high levels only in developing OLGs (Barbarese et al. 1978). Despite the caveat that trafficking of MBP isoforms differs between glial and non-glial cells (Barbarese et al. 1988; Boccaccio and Colman 1995), their distribution pattern has been studied primarily in HeLa cells (Colman et al. 1990; Staugaitis et al. 1990). In these studies, the exon-II-lacking 14-kDa and 18.5-kDa MBPs (the predominant forms in compact myelin) were distributed primarily in the perinuclear regions of the cell, in configurations highly suggestive of close association with membranes, whereas the exon-II-containing 17.22- and 21.5-kDa isoforms were distributed diffusely in both the cytoplasm and the nucleoplasm, and often accumulated within the nucleus (ibid.). This distribution was suggested to be correlated with the presence of the protein segment encoded by exon-II, which is unique to these isoforms (Fig. 1b).
In shiverer mouse OLGs (Allinquant et al. 1991; Pedraza and Colman 2000), as in HeLa cells (Colman et al. 1990; Staugaitis et al. 1990) that have been transfected with MBP variants, the exon-II-containing MBPs were found distributed in the cytosol and nucleus, in contrast to exon-II-lacking isoforms that were confined to the plasma (and some internal) membranes. The mRNAs for exon-II-containing MBPs have been detected only in the cell body of OLGs in culture (de Vries et al. 1997). In transfected N19-cultures, the full-length 21.5-kDa isoforms were targeted primarily to the nucleus, despite the presence of the 3′UTR, in contrast to the 18.5-kDa isoforms, which were primarily cytoplasmic and plasma membrane-associated, because of this RTS (Smith et al. 2011, 2012a, b). It appears then that there are two processes for localization of exon-II-containing MBPs, one involving trafficking of the mRNAs to the perikaryon, and the other for trafficking of the proteins into the nucleus (Boccaccio et al. 1999; Raju et al. 2008; Percipalle et al. 2009).
The transport of exon-II-containing MBP into the nucleus was initially determined to be an active process (meaning that it was time-, temperature-, and energy-dependent) by a series of experiments performed in HeLa cells (Pedraza et al. 1997). Passive diffusion (as would be expected for a small protein of less than 20 kDa or so) did not occur for 21.5-kDa MBP, presumably because it still has such a strong membrane-associating capacity. Active transport was shown to be higher at low cell density, and negligible at high density when cells came into contact, a situation perhaps mimetic of membrane-membrane interaction that occurs when an OLG process wraps around an axon. Other proteins (IgG –immunoglobulin G) could be carried by 21.5-kDa MBP into the nucleus, and it was suggested that exon-II-lacking isoforms could thus be co-transported (Hardy et al. 1996). Energy depletion of the culture diminished nuclear localization, showing the process to be active. Phorbol ester stimulation inhibited the nuclear transport; if the protein was then phosphorylated by kinase pathways that were thus activated, then this modification might have played a role. As exon-II does not have a classic nuclear localization signal (NLS), it was concluded that the NLS was novel or cryptic, or involved multiple sites. The NLS for the 21.5-kDa MBP isoform has recently been shown to reside within the highly conserved 26-residue segment (VPWLKQSRSPLPSHARSR PGLCHMYK in the mouse) encoded by exon-II, particularly via two putative non-traditional PY-NLS motifs identified by the pattern of ZX2-5PB, where ‘Z’ is a basic residue, ‘X’ is any residue, and ‘B’ is a hydrophobic residue (Smith et al. 2012c). Site-directed mutagenesis of selected residues within this segment in red fluorescent protein (RFP)-tagged constructs, prevented nuclear localization in N19-OLGs, and highlighted the importance of the arginine and lysine residues within these motifs for trafficking this protein to the nucleus.
However, Thr92Glu substitution (pseudo-phosphorylation) of 18.5-kDa MBP caused it to translocate to the nucleus from the cytoplasm, and back in a cyclic manner, in N19-OLGs, suggesting a role for Thr92 phosphorylation [(Smith et al. 2012a), cf., (Boggs et al. 2006; Nardozzi et al. 2010)]. Furthermore, at high N19-OLG density in culture, the 21.5-kDa isoform translocated from the nucleus to the cytoplasm (Smith et al. 2012b), and also after phorbol ester treatment of low-confluence cultures (Smith et al. 2012a). Such nuclear-cytoplasmic shuttling under different conditions is a characteristic of many signaling and regulatory proteins; for many proteins, the phosphorylation status is one means to promote nuclear trafficking [e.g., (Nardozzi et al. 2010)]. It has recently been suggested that the DEAD-box RNA helicase Ddx54 may increase the rates of expression and of nuclear trafficking specifically of the 21.5-kDa isoform, but not of exon-II-lacking isoforms (Ueki et al. 2012; Zhan et al. 2013). Classic MBP isoforms are, thus, further examples of proteins that play roles both at the plasma membrane and in the nucleus (Benmerah et al. 2003), and this subcellular localization is controlled both during and after synthesis.
The expression of exon-II-containing MBP transcripts is increased during active early myelination in the human and mouse, and in immature OLGs in culture (Barbarese et al. 1978; Kamholz et al. 1988; Jordan et al. 1989; Pedraza 1997; Pedraza et al. 1997), and has been suggested to be associated specifically with differentiation (de Vries et al. 1997). In multiple sclerosis, potentially during remyelination attempts, the exon-II-containing isoforms have been shown to be up-regulated (Capello et al. 1997). In geriatric rats, exon-II-lacking isoforms are virtually absent (Sugiyama et al. 2002). Curiously, 21.5-kDa MBP is temporo-spatially expressed in developing rat molars (Kim et al. 2008). The functional implications of this observation remain unresolved, but it was suggested that this full-length classic isoform may play a regulatory role in the terminal histo-differentiation of odontogenic cells, perhaps involving the release of neurotrophins and/or neurotransmitters. All in all, such observations are consistent with the original proposal that the exon-II-containing MBP isoforms may have regulatory effects on the myelination program, because of their active transport into the nucleus (Staugaitis et al. 1996; Pedraza et al. 1997).
Using the N19-OLG cell culture system, we have observed a number of distinct functional differences between the exon-II-lacking 18.5-kDa and exon-II-containing 21.5-kDa iso-forms of classic MBP, first that over-expression of the latter changed the cellular distribution of L-type VOCCs in cultured N19-OLGs (Smith et al. 2011), and second with respect to cell proliferation and survival (Smith et al. 2013). In contrast to 18.5-kDa MBP, 21.5-kDa MBP over-expression increased proliferation of early developmental immortalized N19-OLGs by increasing phosphorylation of ERK1/2 and Akt1 kinases, and of small ribosomal subunit protein S6 (rpS6). Activated ERK1/2 MAP kinases (i.e., phosphorylated at Thr202/Tyr204) have been shown to respond to a number of extracellular stimuli or cues including mitogens and growth factors, and are implicated in oligodendroglial differentiation (Younes-Rapozo et al. 2009; Fyffe-Maricich et al. 2011; Ishii et al. 2012). In our study, significant increases in ERK1/2 activation were observed in cells that were undergoing mitosis in N19-OLG cultures that were transfected with 21.5-kDa MBP. Indeed, there was an increase in ERK1/2 activation during mitosis in all dividing N19-OLGs in the culture, even those not expressing 21.5-kDa. These data suggested that this proliferative response may be because of a secreted factor, or to cell-cell mediated contact.
The p90 ribosomal protein S6 kinases are downstream of the Ras/ERK1/2 signaling cascade, and have been shown to exclusively phosphorylate rpS6 at Ser235/236 in vitro and in vivo using an mTOR (mammalian target of rapamycin)-independent mechanism (Roux et al. 2007). As we observed increases in the phosphorylation of rpS6 at Ser235/236 in N19-OLG cultures expressing 21.5-kDa MBP, it is reasonable to hypothesize that this isoform may be modulating glial cell proliferation and cell cycle progression through the ERK1/2-regulated signaling pathway and an mTOR-independent mechanism (Tyler et al. 2009; Guardiola-Diaz et al. 2012).
Co-culture of N2a neuronal cells with N19-OLGs transfected with the 21.5-kDa isoform, but not the 18.5-kDa isoform, caused the N2a cells to have increased neurite outgrowth and process branching complexity (Smith et al. 2013). The same phenomenon was observed in N2a cells treated with conditioned medium from these N19-OLGs. Insertion of a double-NES (nuclear export signal) at the N-terminus of 21.5-kDa MBP abolished these effects. On the other hand, insertion of a double-NLS at the N-terminus of 18.5-kDa MBP caused its nuclear import, but there was no induction of either cell proliferation or neurite outgrowth. Taken together, these results indicated that 21.5-kDa MBP localization within the nucleus was responsible directly or indirectly for causing release of a secreted factor that stimulated N19-OLG proliferation and N2a neurite outgrowth and branching. These effects were thus isoform-speciflc and required the 26-residue protein segment encoded by exon-II, not just nuclear localization.
On the basis of these recent studies, it is becoming increasingly evident that the nuclear localization of early minor isoforms of MBP may play a crucial role in regulating and/or initiating myelin development in the mammalian CNS. The onset of expression of 21.5-kDa MBP occurs at the developmental stage when OPCs transition into immature OLGs and positive regulation of cell proliferation of potential myelinating OLGs would be important. The OPCs respond to a number of growth factors such as platelet-derived growth factor (PDGF-α), neurotrophin (NT-3), and fibroblast growth factor (FGF), that promote self-renewal and proliferation. Further exposure of OPCs to cytokines, such as transforming growth factor TGF-beta, or brain-derived neurotrophic factor (BDNF), can regulate the balance between self-renewal and OLG differentiation (Baron et al. 2000).
It has been reported that non-transfected N19-OLG cells co-cultured with PC12 neuronal cells caused neurite outgrowth because of release of growth factors, both nerve growth factor (NGF) and brain-derived neurotrophic factor (BDNF) (Byravan et al. 1994). However, we did not observe this effect with N2a neuronal cells. It occurred only when the N2a cells were co-cultured with N19-OLGs transfected with 21.5-kDa MBP, or with conditioned medium from 21.5-kDa MBP-transfected N19-OLGs. Nevertheless, NGF and/or BDNF are likely candidates for the secreted factor induced by nuclear-trafficked 21.5-kDa MBP (Du et al. 2006; Van’t Veer et al. 2009; Xiao et al. 2010, 2012).
Finally, it has been reported that transfection of 21.5-kDa MBP into human hepatoma carcinoma (Hep-G2) cells also resulted in increased proliferation of these cells, and blocking of apoptotic pathways (Pan et al. 2009). We also have preliminary indications that over-expression of 21.5-kDa MBP delays cell apoptosis following exposure of N19-OLG cultures to high extracellular potassium concentrations and oxidative stress [unpublished data, cf., (Smith et al. 2012d)]. In summary, 21.5-kDa MBP may thereby contribute further to OLG survival and therefore increased myelin stability, as aberrant apoptosis would induce rapid demyelination (Artemiadis and Anagnostouli 2010; Caprariello et al. 2012).
The 21.5-kDa MBP isoform must interact with some target in the nucleus. Although both 18.5-kDa and 21.5-kDa isoforms are intrinsically disordered, one notable difference may be that the exon-II-encoded segment has a higher positive charge density (proportion of basic residues) than those encoded by the other exons (Hill and Harauz 2005; Hill et al. 2005), which may affect its interactions with nucleic acids or with nuclear proteins. These latter interactions remain to be investigated by techniques such as chromatin precipitation, for example (Sher et al. 2012; Spiro 2012).
As indicated above, we have also shown earlier that, although 21.5-kDa MBP was localized primarily to the nucleus of N19-OLGs, it partially redistributed to the cytosol and to membrane ruffles on stimulation with phorbol ester (Smith et al. 2012b), and also partially redistributed to the cytosol after culture at high cell density (Smith et al. 2012a). The 21.5-kDa isoform also interacts in vitro with actin (Boggs et al. 2005; Hill and Harauz 2005) and tubulin (Hill et al. 2005), and is enriched in detergent-insoluble myelin fractions, which contain the radial component of mature myelin (Karthigasan et al. 1994, 1996), along with CNP (myelin 2′,3′-cyclic nucleotide 3′-phosphodiesterase), actin, and tubulin. The in vitro interactions of 21.5-kDa MBP with cytoskeletal proteins, and with phospholipid membranes, are not appreciably different from those of 18.5-kDa MBP (Boggs et al. 2000, 2005; Hill and Harauz 2005; Hill et al. 2005). Nevertheless, the exon-II-containing 21.5- and 17.22-kDa (murine) MBP isoforms, may have roles in the radial component of mature myelin, which may contain tight junctions [see discussion in (Boggs 2006; DeBruin and Harauz 2007)], and other roles in the cytosol or outer membrane under different conditions.
The expression of 21.5-kDa MBP is altered in demyelin-ating pathologies. First, it has been reported that both human exon-II-containing isoforms of MBP (21.5-kDa and 20.2-kDa) were up-regulated in multiple sclerosis (Capello et al. 1997). In a spontaneously demyelinating mouse model, levels of phosphorylated MBP isoforms, both 18.5-kDa and 21.5-kDa, were increased in detergent-resistant microdo-mains derived from myelin (DeBruin et al. 2006). Decreased levels of phosphorylated 21.5-kDa MBP, relative to the non-phosphorylated isoform, have been observed in mice experiencing demyelination because of advanced age, cuprizone treatment, or double knockout of FcRγ/Fyn genes which are involved in signaling in OLG differentiation (Seiwa et al. 2007). In all of these studies, the cellular mechanisms taking place remain to be determined. Perhaps the expression of classic 21.5-kDa MBP is up-regulated during inherent remyelination attempts in which early events of OLG differentiation and/or proliferation are induced. [During remyelination attempts in multiple sclerosis, Golli genes are also up-regulated (Filipovic et al. 2002; Filipovic and Zecevic 2005)]. Thus, it would be worthwhile to map the spatial distribution of this isoform and its phosphorylated forms, during myelination, and during remyelination of lesions induced by cuprizone or other methods.
We have seen that the classic MBP family comprises a variety of developmentally regulated splice isoforms, that translocate to different cellular compartments (periphery, cell body, nucleus), with multiple biomolecular associations. A ‘dynamic molecular barcode’ of combinatorial post-translational modifications generates further diversity of the family (Kim et al. 2003; Harauz et al. 2004, 2009; Harauz and Libich 2009; Zhang et al. 2012). In particular, phosphorylation has arisen several times in the discussion above, and it is worthwhile to discuss this modification further at this point.
Phosphorylation and dephosphorylation activate or deactivate proteins in cellular signaling events, by modifying their structure, function, subcellular localisation, or interactions (Pawson and Scott 2005). Phosphorylation of MBP (at seryl and threonyl residues, Table S1) is effected by several protein kinase (PK) families (Carnegie et al. 1974; Eichberg and Iyer 1996; Harauz et al. 2004), and yields several of the MBP charge components isolated from myelin (Zand et al. 1998; Kim et al. 2003). Phosphorylation of MBP is not a spurious event – it occurs immediately before and during myelinogenesis (Ulmer and Braun 1983; Vartanian et al. 1986; Stariha et al. 1997), changes in level during development and with age, and is decreased in multiple sclerosis (Kim et al. 2003). Phosphorylation of 18.5-kDa MBP predominates at residues in regions of greater disorder, as previously noted for this protein, and consistent with other IDPs (Iakoucheva et al. 2004; Collins et al. 2008; Harauz and Libich 2009).
The mitogen-activated protein kinases (MAP-kinases, or MAPKs) are especially relevant here. The MAPK family comprises extracellular signal-regulated protein kinases (ERKs), p38 MAPKs, and c-Jun NH2-terminal kinases (Plattner and Bibb 2012), and is important in OLG proliferation, cell survival, differentiation, and apoptosis [reviewed in (Fragoso et al. 2007; Haines et al. 2008; Stariha and Kim 2001a, b)]. The ERKs in particular are increasingly being demonstrated to participate in OLG differentiation (Stariha et al. 1997; Younes-Rapozo et al. 2009; Fyffe-Maricich et al. 2011; Ishii et al. 2012; Sun et al. 2012). The phosphorylation of murine Thr95 (human Thr98, bovine Thr97) within the segment Thr92-Pro93-Arg94-Thr95-Pro96 (the putative SH3-ligand; Table S1) by MAP-kinases is regulated by action potential generation in axons (Murray and Steck 1984; Atkins et al. 1997, 1999), and in OLGs in culture in response to extracellular ligands or depolarization (Vartanian et al. 1986, 1992; Dyer et al. 1994; Soliven et al. 1994). The residue Thr92 is a second MAP-kinase target within the putative SH3-ligand segment (Hirschberg et al. 2003). We consider these two residues to be key post-translationally modified sites that can affect local structure and stability of the protein, its protein-protein interactions, and its intracellular targeting and stability, thereby operating as a ‘molecular switch’ (Fig. 5). There is much evidence to support this conjecture.
In vitro and in silico, phosphorylation of MBP has potential structure-stabilizing effects (Ramwani et al. 1989; Deibler et al. 1990; Polverini et al. 2008, 2011; Harauz and Libich 2009), renders it less susceptible to proteolysis (Ramwani and Moscarello 1990; Medveczky et al. 2006), and modifies its interactions with phospholipid bilayers (Boggs et al. 1997; Polverini et al. 2011). Moreover, charge components of MBP that are phosphorylated at Thr92/Thr95 have a decreased ability to assemble actin, and to bind actin filaments and microtubules to a lipid bilayer (Hill and Harauz 2005; Boggs et al. 2006, 2011). In contrast, those present in natural MBP (which may also be phosphorylated at other sites by other kinases – see Table S1) have an enhanced ability to polymerise and bundle tubulin (Hill et al. 2005).
In vivo, phosphorylation of 18.5-kDa MBP causes it to be associated with less compact myelin (Schulz et al. 1988), and the overall level tends to be decreased in myelin derived from multiple sclerosis patients (Kim et al. 2003). We have shown that MBP phosphorylated at the murine Thr95 site is developmentally partitioned into myelin microdomains called ‘lipid rafts’, some of which may represent signaling domains in mature myelin [(DeBruin et al. 2005; DeBruin and Harauz 2007); cf. (Gielen et al. 2006)]. Using an ND4 transgenic mouse line over-expressing the PLP variant DM20 and which exhibit spontaneous demyelination, we have shown that, compared with healthy mice, there are significant changes in the distribution of MBP splice isoforms, and of their phosphorylated and methylated components (at murine Thr95 and Arg104, respectively; Table S1) in myelin-derived detergent-resistant microdomains (DeBruin et al. 2006). This observation may reflect inherent attempts at repair. Finally, as described above, we have shown that pseudo-phosphorylation of the MAP-kinase targets of 18.5-kDa MBP affects the protein’s trafficking and interactions with Fyn in vitro and in cellula (Smith et al. 2012a). Although both Thr92 and Thr95 have been phosphorylated in vitro (Boggs et al. 2006, 2011) and in silico (Polverini et al. 2011), MBP variants that were pseudo-phosphorylated by Glu substitution at both sites did not express in N19-cells (Smith et al. 2012a). This observation might reflect a drastic change in either protein structure or intracellular trafficking, or both.
Hub proteins, in particular, have large numbers of phosphorylation sites (Batada et al. 2006). So does MBP, with numerous other seryl, threonyl, and tryosinyl phosphorylations because of protein kinases A and C, cAMP-dependent kinases, et cetera (Table S1). Even histidyl and arginyl phosphorylation has been reported, but without identification of specific sites (Smith et al. 1976; Ciesla et al. 2011). A systematic screen by pseudo-phosphorylation at each of these sites will help to reveal their effects on MBP subcellular trafficking and function.
The almost complete absence of CNS myelin, and the lack of compaction at the major dense line of what myelin is formed in the shiverer mutant mouse, can be attributed to the absence of MBP to cause adhesion of the cytosolic surfaces. The formation of compact myelin in the PNS in the shiverer mutant mouse (Kirschner and Ganser 1980) can be attributed to the presence of other adhesive proteins such as P0 and P2, although increased numbers of Schmidt-Lanterman incisures in PNS compact myelin may be because of the absence of the adhesive properties of MBP (Gould et al. 1995). However, other abnormalities of OLGs, Schwann cells, and PNS myelin occur in the shiverer mouse, which may be because of the absence of other functions mediated by classic MBP isoforms. Shiverer OLGs have been described as smaller than normal and irregular in shape, with large numbers of microtubules often assembled in wide bundles or veins, F-actin distributed in punctate foci rather than associated with microtubular veins, and numerous short processes, both in cell culture (Dyer et al. 1995, 1997), and in situ in the CNS (Privat et al. 1979; Inoue et al. 1981, 1983; Cammer et al. 1985; Billings-Gagliardi et al. 2001). Inhibition of MBP synthesis by inhibition of mRNA translation in OPCs had similar effects (Bauer et al. 2012). Half of shiverer OLGs in culture fail to form membrane sheets and about 10% are multinucleated, indicating incomplete cytokinesis (Dyer et al. 1997). The OLGs were found to have a normal or increased rate of proliferation in situ, but to differentiate over a longer time period than wild type OLGs (Seiwa et al. 2002; Bu et al. 2004). Aberrant myelination occurred, with formation of myelin layers around OLG soma, indicating a defect of shiverer OLG processes in distinguishing targets (Privat et al. 1979; Inoue et al. 1981).
In the PNS, increased irregularities in the myelin sheath and at the axon-Schwann cell interface have been noted in the shiverer mouse (Rosenbluth 1980), with shiverer Schwann cells myelinating smaller diameter axons than wild-type cells in a chimera (Peterson and Bray 1984). Both the lipid and protein composition of PNS myelin were abnormal with increased amounts of sulfatide, hydroxy-fatty acid forms of galactosylceramide (GalC), and phosphatidylcholine, and decreased amounts of non-hydroxy-fatty acid forms of GalC and sphingomyelin (Inouye et al. 1985), a decreased amount of the 20-kDa form of myelin oligodendrocyte basic protein (MOBP), but not other size isoforms (Montague et al. 1999), decreased expression of connexin32, myelin-associated glycoprotein (MAG), and cadherin, but not their mRNA levels, in the PNS (Smith-Slatas and Barbarese 2000), and markedly increased amounts of annexin II in Schmidt-Lanterman incisures and paranodal loops of PNS myelin (Hayashi et al. 2007). Shiverer OLGs in the CNS had an increased amount and activity of carbonic anhydrase (Cammer et al. 1985).
The lipid composition of shiverer OLGs has not been examined, but the GalC in them does not cluster into large ordered domains when cultured with neurons, unlike wild-type OLGs, indicating a requirement for MBP for GalC clustering and ordering (Fitzner et al. 2006). Furthermore, GalC and sulfatide-mediated signaling to the cytoskeleton triggered by cross-linking by anti-GalC or anti-sulfatide antibodies, or by carbohydrate-carbohydrate interactions with multivalent galactose and sulfated galactose, does not occur in the absence of MBP in shiverer OLGs, or after MBP knockdown with siRNA for MBP (Dyer 1997; Dyer et al. 1997; Boggs et al. 2008b, c). Thus, 18.5-kDa MBP may be required for transmembrane transmission of extracellular signals to the cytoskeleton. The diverse effects of the lack of MBP in shiverer OLGs and Schwann cells may be because of aberrant signaling, protein and lipid trafficking, or altered cytoskeletal organization, reflecting the ability of 18.5-kDa MBP to modulate Ca2+ entry, interact with cytoskeletal proteins, SH3-domain proteins, and Ca2+-calmodulin, or to a role of 21.5-kDa MBP in the nucleus.
The most complex organ in the body is the brain, and the central nervous system encompasses a plethora of global molecular interaction networks (Agnati et al. 2006), with a high degree of redundancy (Zawadzka and Franklin 2007). Therein lies a great challenge for understanding proteins such as MBP – mature myelin has a tremendous amount of MBP within it, and compact myelin architecture is also difficult to replicate using cell culture, requiring ingenious new experimental approaches [see comments by (Zuchero and Barres 2011)].
That multiple families of intrinsically disordered proteins seem to do similar things confounds attempts to prove that MBP also participates in more than membrane compaction in developing and mature myelin. For instance, the neuronal microtubule-associated protein (MAP) tau is expressed also in OLGs (LoPresti et al. 1995; Gorath et al. 2001); tau and MBP have many physicochemical and phenomenological properties in common, and co-localize with microtubules at the tips of cellular processes (Muller et al. 1997). Other MAPs are also found in OLGs (Bauer et al. 2009). So why are tau and other MAPs required, if MBP can also assemble microtubules? Moreover, the Golli and classic MBP isoforms have many segments in common and their expression profiles overlap temporally (Campagnoni et al. 1993; Givogri et al. 2000, 2001), with potentially similar interaction partners as could be predicted by analysis of MoRFs, as in Fig. 2. The same can be said about the numerous classic splice isoforms of MBP (Fig. 1b, and others); we have only discussed the 18.5-kDa and 21.5-kDa variants here. Apparent redundancy in developmental gene function is common, as evident from the lack of significant phenotype for many genes when knocked out (Cooke et al. 1997). The precise roles of (numerically) minor and major MBP isoforms must be defined by their (partially overlapping) expression during OLG development, their subcellular targeting, and type and degree of post-translational modification, especially phosphorylation.
The Golli protein isoforms come first developmentally, being expressed in OLGs at intermediate stages of differentiation (Givogri et al. 2001; Tosic et al. 2002; Campagnoni and Campagnoni 2004; Fulton et al. 2010a). There are both karyophilic and plasma membrane-targeted forms, and these proteins promote OLG migration and process extension, and enhance potassium-induced calcium influx via multiple kinase signaling pathways (Paez et al. 2007, 2009a, b, 2010, 2012; Fulton et al. 2010b).
The 21.5-kDa MBP isoform is the first of the classic isoforms to be synthesized, starting with OPCs. It is nuclear-targeted and serves to increase OLG proliferation, and induce secretion of soluble factors (potentially NGF and/or BDNF) to enhance neurite outgrowth, as discussed above. It is also up-regulated in remyelination attempts in multiple sclerosis (Capello et al. 1997), like Golli (Filipovic et al. 2002; Filipovic and Zecevic 2005), so presumably involved in early events of OLG differentiation and/or proliferation. This isoform continues to be produced even by highly differentiated OLGs and may have a protective role in inhibiting apoptosis and maintaining myelin turnover. It is enriched in the radial component of mature myelin where it may interact with tight junctions, for example.
The 18.5-kDa MBP isoform comes next, being synthesized in copious amounts as the membrane processes are being extended and ensheathing the axon. Myelin basic protein is observed only in OLGs that have migrated into axonal pathways, and it is produced just before commencement of axonal ensheathment (Asou et al. 1995; Butt et al. 1997). In mature myelinating OLGs, 18.5-kDa MBP has been considered to redistribute from the soma and primary processes into the myelin sheaths, reflecting a change in the site of MBP mRNA expression (Brophy et al. 1993; Barbarese et al. 1999). However, in younger mice, MBP is also detected in the cytoplasm and plasma membrane of cell bodies and proximal processes of mature OLGs in situ, both with and without connections to myelin sheaths, and remains there during myelinogenesis (Schwob et al. 1985; Hardy et al. 1996). Myelin basic protein must exert many of its putative diverse scaffolding and signaling functions when cytosol and cytosolic constituents are still present, i.e., during the initial association of processes with axons, axonal ensheathment and remodeling, and early stages of myelination. Cytosol is present in the first few layers of ensheathment before compaction and appearance of the major dense line occurs. These early ensheathments actively elongate and undergo extensive remodeling (Ioannidou et al. 2012). Varicosities containing MBP are also observed in immature myelin sheaths, indicating the presence of cytosol (Hartman et al. 1982). Examination of living myelin sheaths on CNS axons showed that the compact myelin is fenestrated by a network of diffusionally interconnected cytoplasmic pockets (Velumian et al. 2011).
Classic 18.5-kDa MBP may participate in galactosylceramide/sulfatide-mediated signaling pathways induced by contact between the extracellular surfaces of OLGs as they wrap around the axon, which may be a signal for compaction by initiating depolymerization of the actin cytoskeleton and microtubules (Boggs et al. 2004, 2008b). At the tips of the extending membrane processes to form the compact regions of myelin, 18.5-kDa MBP may also tether the cytoskeleton to the membrane, bundle microfilaments and microtubules together and perhaps to each other, and help regulate process extension and retraction and axonal ensheathment (Boggs 2006; Harauz et al. 2009). [The reader is referred to reference (Bauer et al. 2009) for a fuller discussion of cytoskeletal dynamics during OLG differentiation and myelination]. These varied associations may be modulated by local fluctuations in pH or potassium ion concentration (Jo and Boggs 1995; Ro and Carson 2004; Boggs 2006). The specific targeting of 18.5-kDa MBP to the cell periphery, concomitant with synthesis and trafficking of myelin membrane components, serves to establish OLG polarity (Maier et al. 2008; Simons et al. 2012).
Finally, the extrusion of cytoplasm upon flattening of the myelin membrane sheets leaves little space for anything other than small, thin proteins such as 18.5-kDa MBP. Owing to its extreme net positive charge and high degree of flexibility and conformational adaptability, the classic 18.5-kDa MBP isoform serves as an adhesive molecule in CNS myelin sheaths, bringing together the two apposing faces of the cytoplasmic leaflets of the cell membrane processes of myelinating cells, OLGs, to effect myelin compaction around axons. The close planar packing and consequent sieving effect prevents the migration of some other components into these regions (Pedraza et al. 2001; Aggarwal et al. 2011b; Kattnig et al. 2012; Simons et al. 2012). Within compact myelin, MBP may also be able to act as a sensor for galactosylceramide/sulfatide-mediated signaling between myelin layers (termed a glycosynapse; see Fig. 6), which may allow it to transmit signals from the axon throughout compact myelin (Boggs et al. 2004, 2008b, c; Hakomori 2004a, b). The variety of dynamically modified forms of this protein, and phosphorylation-induced nuclear targeting and partitioning into membrane microdomains, suggest that it continues to be involved in mature myelin maintenance and turnover [see Figure 4 in (Harauz and Musse 2007)].
Future investigations of the temporal and spatial distribution of classic phosphorylated MBP isoforms will be pivotal in enhancing our understanding of myelinogenesis – a ‘hierarchy’ of time and place (Campagnoni and Macklin 1988; Pedraza et al. 2001; Simons et al. 2012). This goal has not yet been accomplished because of the immense difficulty of seeing what is happening in myelin in vivo (Zuchero and Barres 2011), although imaging techniques are being developed to detect changes at the cytoplasmic surfaces in compact myelin (Micu et al. 2007; Velumian et al. 2011). A suitable glial system in which the copious amounts of 18.5-kDa MBP at the plasma membrane does not swamp the signal of numerically lesser components is essential (Aggarwal et al. 2011b; Zhang et al. 2011; Jarjour et al. 2012; Liazoghli et al. 2012). Novel methods of imaging at the single molecule level are also required [e.g., (Toulme and Khakh 2012; Velumian et al. 2011)], as are specific approaches to subcellular and molecular sub-fractionation to probe the phospho-proteome at high resolution (Goswami et al. 2012; Mausbacher et al. 2012; Gopalakrishnan et al. 2013). Here, studies on modifications of the highly basic histones and their roles in epigenetics serve as paradigms for future work on classic MBP isoforms [e.g., (Brunner et al. 2012; Copray et al. 2009; Huynh and Casaccia 2010; Sidoli et al. 2012; Tweedie-Cullen et al. 2009)]. If histone H2A variants can be seen as ‘organizing the genome’ (Millar 2013), then analogously, classic MBP splice and modification variants can be thought of as organizing the myelin sheath.
The old question of what myelin basic protein does remains open, and requires new technological and methodological advances to be answered in full. As written by Martenson (Martenson 1980) more than 30 years ago: ‘In any event, I would think that the most fruitful approach to the question of myelin basic protein function would be to consider the protein as an essential element of some vitally important metabolic system that operates within the myelin sheath and to find out what the system does. In such a role the basic protein would act like a respectable protein, rather than like some sort of polymeric “glue”.’ We suggest that (mainly 18.5-kDa) MBP’s ability to form a scaffold underlying the cytoplasmic leaflet of the myelin membrane, connecting it to the canonical actin-tubulin cytoskeleton underneath and to SH3-domain proteins (see schematic in Fig. 6) in OLGs, and possibly to tight or gap junctions in compact myelin, could constitute such a system. In addition, this protein may serve to regulate local calcium concentration influx. All of these associations could be modulated by post-translational modifications, by local fluctuations in pH and ionic strength, and further by interactions with proteins such as Ca2+-calmodulin or SH3-domains. Moreover, the roles of exon-II-containing MBP, like the 21.5-kDa isoform, in the nucleus remain even more enigmatic. Further investigation is necessary to establish what stages of myelinogenesis and/or myelin dynamics are regulated by MBP’s seeming multi-functionality.
The work in our laboratories has been supported by the Canadian Institutes of Health Research (Operating Grants MOP #43982, #74468, and #86483), the Natural Sciences and Engineering Research Council of Canada (Discovery Grant RG121541, and various equipment grants), and the Multiple Sclerosis Society of Canada (Operating Grants, Postdoctoral Fellowships, and Doctoral Studentships). The authors are grateful for stimulating and productive collaborations with many colleagues over the years, and especially with the many former and present group members who have worked on this protein. We dedicate this review to Dr. Celia Campagnoni, in gratitude for her kindness and generosity to so many of us.
The authors have no conflicts of interest to declare.
Additional supporting information may be found in the online version of this article at the publisher’s web-site.
Table S1. Summary of significant sites of post-translational modification of the 18.5-kDa isoform of myelin basic protein, with reference to the human sequence.
Table S2. Summary of protein and small-ligand interactions of classic 18.5-kDa MBP.