Glycobiology Advance Access originally published online on May 7, 2003
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Glycobiology, 2003, Vol. 13, No. 8 55R-66R
© 2003 Oxford University Press
REVIEW |
Dystroglycan glycosylation and its role in matrix binding in skeletal muscle
Department of Neuroscience, Glycobiology Research and Training Center, University of California, San Diego, School of Medicine, La Jolla, CA 92093-0691
accepted on April 24, 2003
 |
Abstract |
|---|
Dystroglycan is an essential component of the dystrophin–glycoprotein complex. Three glycan sequencing studies have identified O-linked mannose chains, including NeuAc
2, 3Galß1,4GlcNAcß1,2Man-O, on dystroglycan. Chemical deglycosylation of dystroglycan, antibody blocking studies, and glycan blocking studies all suggest that the O-linked glycans on dystroglycan mediate the binding of extracellular matrix proteins in skeletal muscle. Structural data on laminin G domains and agrin-binding studies also suggest this is the case. Dystroglycan, however, is able to bind proteins via mechanisms that do not involve O-linked glycans. Moreover, laminin and other matrix proteins can bind cell adhesion molecules via their glycan chains. Thus although complex and sometimes not overly convincing, these data suggest that glycosylation plays an important role in dystroglycan binding and function in skeletal muscle. Key words: agrin / dystroglycan / integrin / laminin / muscular dystrophy
|
Introduction |
|---|
Skeletal muscle is a complex tissue made up of many cell types (for a review, see Engel and Franzini-Armstrong, 1994)
. Foremost among these is the skeletal myofiber, whose primary function is to generate force and induce movement. Skeletal myofibers are large, multinucleated cells that form from the fusion of single cell myoblast precursors during embryonic development. Myofibers within individual muscles are connected to a basal lamina of extracellular matrix (Figure 1). Basal lamina proteins bind to a complex of proteins within the sarcolemmal membrane. Sarcolemmal membrane proteins in turn bind to the actin/myosin complex and intermediate filaments that define the I-band, A-band, and Z-line of individual sarcomeres. The actin-myosin system is responsible for force generation in myofibers, but this force is transmitted in complex ways at differing points throughout the cell (see Patel and Lieber, 1998). The actin/myosin system and titin transmit force via the Z-line, and the sarcomere transduces force via its connection to the extracellular matrix and costameres. Force between adjacent sarcomeres (or myofibrils) is transduced via intermediate filaments. Myofibers may also transduce force via the endomysial connective tissue and ultimately the tendon.
|
Given the incredible complexity of these interactions and the baroque molecular architecture used to achieve this complexity, it is not surprising that defects in a large number of proteins cause muscle pathology. Here, the genetic and biochemical evidence that muscle disorders are related to defects in glycosylation will be reviewed. This review will focus on forms of muscular dystrophy related to defects in the glycosylation of dystroglycan, an important matrix-binding protein in skeletal muscle and in other tissues. In doing so, it is hoped that glycobiologists will realize the incredible potential and need for research in this area. A detailed elucidation of the structure–function relationships and cell biology of the glycosyltransferases and carbohydrate structures involved in these disorders could dramatically improve both the diagnosis and treatment of these diseases.
|
Muscular dystrophy and the dystrophin–glycoprotein complex |
|---|
Muscular dystrophy is a term that encompasses a variety of genetic disorders that result in muscle wasting. Many forms of muscular dystrophy are caused by defects in genes that result in loss of expression of members of the dystrophin–glycoprotein complex. The dystrophin–glycoprotein complex is a macromolecular structure involved in linking the extracellular matrix, through the sarcolemmal membrane, to the actin cytoskeleton (for a review, see Blake et al., 2002
; Henry and Campbell, 1999; Durbeej et al., 1998) (Figure 2A). The basal lamina of the extracellular matrix that ensheathes each myofiber is composed of laminin, collagen IV, and other matrix proteins (for a review, see Hall and Sanes, 1993; Sanes and Lichtman, 1999). These proteins are made and deposited by the muscle during embryonic and early postnatal development.
|
Laminin-2 is associated with the dystrophin–glycoprotein complex in skeletal muscle, where it is expressed ubiquitously along the myofiber. Laminin-2, like all laminins, is a cruciform trimer made from three laminin chains (in this case ß1,
2, and 
1) (see Engel, 1992). The 2 chain contains five G domains at its C-terminus that can interact with muscle membrane receptors, including dystroglycan and integrins (Henry and Campbell, 1999; Timpl et al., 2000). The interaction of laminin with dystroglycan occurs via G domains within the laminin protein (Talts et al., 1999). Laminin G domains are defined by a pattern of repeating cysteines, and all known laminin chains contain five such domains at the C-terminal end of the protein.
Dystroglycan is a highly glycosylated matrix-binding protein that is posttranslationally cleaved into two peptides, and ß dystroglycan (Ervasti and Campbell, 1991). Dystroglycan resides on the outer surface of the myofiber membrane and binds tightly but noncovalently to the ß chain, which is a transmembrane protein (Ervasti and Campbell, 1991). ß Dystroglycan, in turn, binds via its cytoplasmic domain to dystrophin (Chung and Campanelli, 1999), a large protein that associates with cytoplasmic and transmembrane members of the dystrophin–glycoprotein complex and with actin (Ervasti and Campbell, 1993). A number of other transmembrane proteins and cytoplasmic proteins are associated with the dystrophin–glycoprotein complex, including sarcoglycans, sarcospan, dystrobrevins, syncoilin, and dysbindin (Henry and Campbell, 1999; Benson et al., 2001; Poon et al., 2002).
Mutations that lead to loss of one of these proteins result in the loss of other members of the dystrophin–glycoprotein complex along the sarcolemmal membrane, and deficits in most of these proteins cause muscular dystrophy. Loss of dystrophin results in Duchenne muscular dystrophy (DMD), loss of sarcoglycans results in forms of Limb-Girdle muscular dystrophy, and loss of laminin results in merosin (laminin-2)-dependent congenital muscular dystrophy (for a review, see Durbeej et al., 1998; Blake et al., 2002). In addition, loss of dystrobrevins results in muscular dystrophy in mice (Grady et al., 1999); defects in other proteins, including collagen VI, titin, desmin, and caveolin-3, cause severe myopathies in humans (for a review, see Carlsson and Thornell, 2001; Nishino and Ozawa, 2002). In contrast to these genes, null mutations in dystroglycan (Dag1) are lethal at an early embryonic stage in mice (Williamson et al., 1997). Thus complete loss of dystroglycan protein in humans may not allow viability. Chimeric skeletal muscles made with dystroglycan-deficient embryonic stem cells do have muscular dystrophy. Thus muscular dystrophy can arise from the absence of dystroglycan in skeletal muscle (Cote et al., 1999). As will be described in detail later in this article, a host of diseases have now been linked to the aberrant posttranslational modification of dystroglycan, and all of these disorders involve changes in dystroglycan glycosylation. Thus a large number of genetic and biochemical studies have shown that members of the dystrophin–glycoprotein complex form an essential scaffold that links the extracellular matrix to the actin cytoskeleton. Moreover, studies suggest that glycans on dystroglycan are essential for the formation of this complex.
|
The neuromuscular junction and the utrophin–glycoprotein complex |
|---|
 TopAbstractIntroductionMuscular dystrophy and the... The neuromuscular junction and... Glycosylation of {alpha}...Role of {alpha} dystroglycan...Other transmembrane scaffolds to...ConclusionsReferences |
|---|
Most of the proteins described are expressed along the entirety of the sarcolemmal membrane in muscle. As such, they are also expressed in regions of the myofiber apposed to the motor nerve terminal (the neuromuscular junction) and tendons (the myotendinous junction). Other molecules related to dystrophin and the dystrophin–glycoprotein complex, however, are uniquely expressed at the neuromuscular junction and are not present along extrasynaptic regions of the myofiber (Figure 2B). These include utrophin, a paralog of dystrophin (Ohlendieck et al., 1991
), dystrobrevin-1 (Peters et al., 1998), and ß2 syntrophin (Peters et al., 1997). In addition, uniquely synaptic forms of laminin are present in the synaptic cleft at the neuromuscular junction (Chiu and Sanes, 1984; Patton et al., 1997). These include at least three laminin chains—ß2, 4, and 5. Thus at least three laminin forms, laminin-4 (ß2, 2, 1), laminin-9 (ß2, 4, 1), and laminin-11 (ß2, 5, 1), could be uniquely expressed at the neuromuscular junction. Although synaptic homologs of dystroglycans and sarcoglycans are not known, the neuromuscular junction does contain unique glycan structures, such as terminal ßGalNAc linkages (Sanes and Cheney, 1982; Martin et al., 1999) that can be present on dystroglycan (Xia et al., 2002).
Thus an additional protein scaffold exists at the neuromuscular junction that is very similar in structure to the dystrophin–glycoprotein complex but contains novel proteins. This likely is associated with some proteins that have been glycosylated with uniquely synaptic glycans. These include synaptic forms of laminin, which may interact with a uniquely glycosylated form of dystroglycan. ß Dystroglycan, in turn, can interact with utrophin, which binds to actin filaments via protein motifs that are distinct from those used by dystrophin (Rybakova et al., 2002). The expression of the utrophin–glycoprotein complex at the neuromuscular junction is independent of the extrasynaptic dystrophin–glycoprotein complex. In the mdx mouse (a model for DMD), dystrophin and many dystrophin-associated glycoproteins are absent or very reduced along the extrasynaptic muscle membrane; however, utrophin and its associated glycoproteins are still expressed at the neuromuscular junction (Matusumura et al., 1992). Similarly, mice lacking utrophin maintain expression of dystrophin and the dystrophin-associated glycoproteins in the extrasynaptic membrane (Grady et al., 1997; Deconinck et al., 1997).
|
Glycosylation of dystroglycan
|
|---|
Glycan structures have been reported for
dystroglycan from three different tissues in three different species (Chiba et al., 1997; Smalheiser et al., 1998; Sasaki et al., 1998), but none are from mouse or human tissue. All of the reported structures used dystroglycan that had been purified using carbohydrate-binding lectins. Because these sequences are derived from material that has been enriched for particular glycans, they may only represent a fraction of the total number of glycans that can be present. Therefore it is likely that as-yet-unreported glycan structures will still be identified on this protein.
Preliminary characterization of dystroglycan glycosylation was primarily based on its migratory pattern on sodium dodecyl sulfate gels. Early studies by Campbell and colleagues demonstrated that dystroglycan is a heavily glycosylated protein that migrates in a heterogeneous pattern on such gels (Ibraghimov-Beskrovnaya et al., 1992). The dystroglycan gene encodes about a 70-kD chain polypeptide, however, this protein migrates at 120 kDa in brain and peripheral nerve, 140 kDa in cardiac muscle, 150 kDa in lung, 156 kDa in skeletal muscle, and 190 kDa in Torpedo electric organ (see Durbeej et al., 1998; Henry and Campbell, 1999). ß Dystroglycan, in contrast, migrates as a 43-kDa protein in all of these tissues, though it, too, clearly is a glycoprotein. The molecular weight of dystroglycan from chick skeletal muscle differs as development progresses, suggesting that glycosylation is also regulated during development within the same tissue (Leschziner et al., 2000). Likewise, glycosylation of dystroglycan is affected by muscle denervation (Leschziner et al., 2000). Therefore glycosylation of dystroglycan in skeletal muscle is also regulated by neural activity. Because of its migration pattern on gels and because the primary sequence contains two potential sites for glycosaminoglycan addition, dystroglycan was originally thought to be a proteoglycan. However, a number of studies using both enzymatic (glycosaminoglycan lyase) and chemical methods (nitrous acid) failed to prove the presence of any glycosaminoglycan (GAG) side chains (Smalheiser and Schwartz, 1987; Smalheiser, 1993; Ervasti and Campbell, 1993; Yamada et al., 1996; Gee et al., 1993; Bowe et al., 1994). The definition absence of GAGs, however, was based on a lack of altered migration on one-dimensional sodium dodecyl sulfate–polyacrylamide gel electrophoresis gels. Therefore, it is still possible that short GAG chains that terminate prior to the GAG elongation reactions (see Esko and Selleck, 2002) decorate these sites, as these would not have greatly altered the molecular weight of the protein.
Dystroglycan contains several N-linked glycans, and enzymatic removal of N-linked glycans alters its molecular weight by about 4 kDa (Ervasti and Campbell, 1991). The remainder of dystroglycan glycosylation presumably occurs on O-linked sites in the serine-theronine-rich "mucin" domain. This mucin region is located roughly in the middle third of the protein (Figure 3). Electron microscopy studies show that dystroglycan has a dumbbell-like appearance, with a globular N-terminal domain, a long stalk-like region in the mucin domain, and a globular C-terminal domain (Brancaccio et al., 1995). Analysis of the primary sequence suggests that the O-linked glycosylation may be particularly complex. There are about 50 serines and threonines that could be glycosylated in the mucin region of roughly 170 amino acids. Wilson et al. (1991) have compared O-glycosylated sequences in many proteins and have identified that a proline is very often found at the -1 or +3 position in proteins where a single O-linked glycan is present on serine or threonine. Dystroglycan has 18 P(-1) sites and 8 P(+3) sites. All but four of these potential sites reside in the mucin region (amino acids 317 to 488 based on sequence reported in Ibraghimov-Beskrovnaya et al., 1991). Surprisingly, the extracellular domain of ß dystroglycan also has a hot spot for O-linked glycosylation with two P(-1) sites and two P(+3) sites in the span of 11 amino acids. In proteins where multiple O-linked sites are present, however, there appears to be no propensity for proline spacing as is found at singly glycosylated sites (Wilson et al., 1991). Indeed, there are also several small repeating sequences of serine and/or threonines that are devoid of prolines in the mucin domain of dystroglycan. Because all proteins used for comparative sequence analysis were likely glycosylated with O-linked GalNAc, it may be that the unusual proliferation of proline-spaced sites may define sites for O-linked mannose on this protein (Sasaki et al., 1998; Smalheiser et al., 1998; Chiba et al., 1997). Other repeating sequences are also present and may or may not be important, including a V(x/T)(x/T)P motif, which is present five times in the dystroglycan sequence. There are no apparent consensus sites for O-linked fucosylation or glucosylation.
|
Although the glycans present on any particular serine or threonine is unknown, the sequence of O-linked glycans as a group has been studied in several tissues. O-linked glycans from the mucin region have been sequenced from three different sources, sheep brain (Smalheiser et al., 1998
), bovine peripheral nerve (Chiba et al., 1997), and rabbit skeletal muscle (Sasaki et al., 1998). In all three cases, an O-linked mannose structure was found, which was elaborated with a sialyl-N-acetyllactosamine to yield NeuAc2,3Galß1,4GlcNAcß1,2Man-O-Ser/Th. In rabbit skeletal and bovine peripheral nerve, a significant amount of Galß1,3GalNAc-O was also identified, suggesting that multiple O-linked structures exist. These findings were consistent with the study using sheep brain as well. Both the rabbit skeletal muscle and bovine peripheral nerve studies used 2-aminobenzamide labeling of oligosaccharides liberated by mild hydrazinolysis, followed by digestion with glycosidases and high-performance liquid chromatography analysis of glycan chains. Chiba et al. (1997) also used ß-elimination with sodium borohydride to liberate glycans. In both studies, glucose was also present, possibly as a contaminant. Chiba et al. (1997) also performed 1,2-diamino-4, 5-methylenedioxybenzene labeling of sialic acids. They found a 4:1 ratio of N-acetyl versus N-glycolyl neuraminic acid in peripheral nerve, with no O-acetyl derivatives present. Smalheiser et al. (1998) analyzed O-linked glycans on dystroglycan from sheep brain using very different methods. O-linked glycans were liberated by ß-elimination with sodium borohydride but then were methylated and characterized by fast atom bombardment mass spectrometry (FAB-MS). In addition to the structures found in peripheral nerve and skeletal muscle, they found O-mannosyl-linked Lewis x antigen Galß1,4(Fuc1,3)GlcNAcß1,2Man-O in brain. It is not clear whether this structure is not present in the other two tissues studied or if the FAB-MS method is simply more sensitive and therefore would have identified this structure in other tissues had it been used. Lectins that bind fucose, however, do not stain skeletal muscle (Scott et al., 1988), suggesting that fucose simply is not present.
In all three glycan sequencing studies, carbohydrate binding lectins were used to purify the material in question. Sasaki et al. (1998) and Chiba et al. (1997) used wheat germ agglutinin (WGA) chromatography, and Smalheiser et al. (1998) used Concanavalin A and Jacilin agglutinin chromatography. In all cases, laminin-1 affinity chromatography was also used. Although the demonstration of these sequences is extremely important for understanding dystroglycan function, it is entirely possible that some glycan chains were not identified because glycan binding was used in the purification process (Figure 3). For example, dystroglycan from skeletal muscles of mice overexpressing the cytotoxic T cell GalNAc transferase (Galgt2) binds well to ßGalNAc binding lectins, such as Wisteria floribunda agglutinin, but that this material binds poorly, if at all, to WGA (Nguyen et al., 2002).
Similarly, aside from the presence of Lewis x in the brain, no tissue-specific heterogeneity in dystroglycan glycosylation was identified in these articles. It would be surprising if such differences did not exist. Lectin and antibody blotting suggest that brain and peripheral nerve dystroglycan have terminal GalNAcs and express the HNK-1 epitope (Yamada et al., 1996; Smalheiser and Kim, 1995). Dystroglycan isolated from skeletal muscle, in contrast, contains few if any terminal GalNAcs (Xia et al., 2002), though some are masked by sialic acid (Ervasti et al., 1997), and does not express HNK-1 (Ervasti and Campbell, 1993). Dystroglycan also has very different affinities for different forms of laminin (Talts et al., 1999, 2000). Therefore, use of laminin-1 may have also eliminated glycoforms that do not bind well to this kind of laminin. This is an especially sobering thought, given that laminin-1, though copiously expressed from the Engelbreth-Holm-Swarm (EHS) tumor (and therefore easily purified), has a fairly restricted distribution in the nervous system and skeletal muscle (Lentz et al., 1997).
The finding of O-linked mannose is the first description of this linkage on a purified mammalian glycoprotein, but it has been known for several decades that O-linked mannose exists in mammals. O-linked mannose was reported to be present in mammalian brain by Finne et al. (1979) long before its potential value could be fully appreciated. Several studies have confirmed and expanded on this result (Yuen et al., 1997; Chai et al., 1999; Kogelberg et al., 2001). O-linked mannose has been reported to be present on HNK-1 structures in rabbit brain (Yuen et al., 1997), and structures identical to those reported to be on dystroglycan in sheep brain (Smalheiser et al., 1998) were shown to be abundant in rabbit brain (Chai et al., 1999). Here, glycans were released from pronase-digested lipid/GAG-depleted brain proteins and purified by ion-exchange chromatography. Fractions were then analyzed by liquid secondary ion mass spectrometry. After further fractionation of O-linked pools and derivitization, sequences were identified by gas chromatography mass spectrometry. The authors claim that by this method, 30% of the O-linked glycan in brain is O-linked via mannose (Chai et al., 1999). Interestingly, in addition to identifying NeuAc2,3Galß1,4GlcNAcß1,2Man-O and Gal1,4[Fuc1,3]GlcNAcß1,2Man-O structures that had been reported to be present on dystroglycan in sheep brain (Smalheiser et al., 1998), they also found 1,6-disubstituted Man--O structures, including NeuAc2,3Galß1, 4GlcNAcß1,2[NeuAc2,3Galß1,4GlcNAcß1,6]Man-O (and desialylated variants thereof). Although not yet reported, these branched structures might exist on dystroglycan. If they do not, then such structures must be present on other O-linked glycoproteins. If Chai et al. (1999) are correct in their estimate that 30% of O-linked glycans in brain are linked via mannose, then other proteins with O-linked mannose must exist because dystroglycan is not likely to represent a third of all O-linked glycoprotein in the brain.
|
Role of dystroglycan glycosylation in ligand binding
|
|---|
Dystroglycan has been shown to bind to extracellular matrix proteins, transmembrane proteins, viruses, and bacteria (Henry and Campbell, 1999). Most, but not all, of these binding interactions appear to occur via glycans present in the mucin domain, and most require calcium. Because of the large number of ligands that interact with dystroglycan, I address each group separately.
Laminin
Neither the N- nor C-terminal globular domains of dystroglycan that flank the mucin region bind laminin with high affinity (Brancaccio et al., 1997; Sciandra et al., 2001; DiStasio et al., 1999). Therefore, laminin binding requires the mucin region. It is important to note, however, that there is no direct evidence that O-linked glycopeptides derived from dystroglycan bind directly to any of these ligands. In addition, binding studies often use dystroglycan that has been isolated from different tissue sources or species. Therefore, comparing such studies is problematic.
Studies on native proteins
Many different forms of laminin exist, and many of these have been used to study binding to dystroglycan. Native laminin trimers (laminin-1, ß1,1,1; laminin-2, ß1,2,1) as well as recombinantly produced laminin G domains (1, 2, 4, and 5) have all been used. In skeletal muscle, 1 is only transiently expressed during embryonic development, while 2 is the principal extrasynaptic chain (Sanes et al., 1990). Laminin 4 and 5 both are both expressed around myofibers in the perinatal period but become confined to the neuromuscular and myotendinous junctions at early postnatal ages (Patton et al., 1997).
Using laminin-1 overlays to dystroglycan purified from rabbit skeletal muscle, Ervasti and Campbell (1993) demonstrated that glycosylation was required for laminin binding. Several lines of evidence were provided. First, chemical deglycosylation of dystroglycan with trifluoromethanesulfonic acid (TFMS) eliminated binding. Second, a monoclonal antibody that requires the presence of glycans on dystroglycan for binding, IIH6, also blocked laminin binding. IIH6 also blocks laminin-1 binding to muscle dystroglycan from Torpedo (Sugiyama et al., 1994) and bovine peripheral nerve (Matsumura et al., 1997). Interestingly, neuraminidase digestion of dystroglycan, which removed sialic acids, did not inhibit binding, nor did VIA4-1, a second carbohydrate-dependent monoclonal antibody. Thus there appears to be some specificity to the type of glycan on dystroglycan that binds laminin-1, and it does not appear to require sialic acid.
In contrast to skeletal muscle, binding of laminin-1 or laminin-2 to dystroglycan isolated from bovine peripheral nerve is eliminated by neuraminidase treatment. Moreover, addition of soluble sialic acid or NeuAc2,3Galß1,4GlcNAc inhibited binding (Yamada et al., 1996). Similar results were found when studying the adhesion of RT4 Scwhannoma cells to laminin (Matsumura et al., 1997). Both neuraminidase and soluble sialic acid blocked these effects, as did IIH6 (but not VI4A-1) (Matsumura et al., 1997). It is odd, however, that neuraminidase digestion of muscle dystroglycan does not inhibit binding of the blocking antibody IIH6, nor does it inhibit laminin-1 binding (Ervasti and Campbell, 1993). This would suggest that sialic acid is not important for binding the muscle form, but it is important for binding to the form in peripheral nerve. Of course, these forms could be quite different. Dystroglycan in skeletal muscle has almost twice as much carbohydrate per unit molecular weight as it does in peripheral nerve.
The degree to which the glycan blocking experiments using dystroglycan from peripheral nerve are relevant is uncertain. For example, both N-glycolyl and N-acetylneuraminic acid (as well as colominic acid) can block laminin-1 binding in these studies, but addition of any of these sialic acid forms requires concentrations above 1 mM (Yamada et al., 1996). Likewise, NeuAc2,3Galß1,4GlcNAc can block laminin binding to dystroglycan (again at concentrations above 1 mM), but NeuAc2,3Galß1,4Glc cannot (Chiba et al., 1997). Even at concentrations of 10 mM, NeuAc2, 3Galß1,4GlcNAc only inhibits laminin-1 binding by 35%. One could interpret these data to mean that the subterminal glycans are more important for binding than sialic acid is. Alternately, the valency of the glycans in the mucin region may be very important for high affinity binding. If so, the glycan blocking experiments done in peripheral nerve (Chiba et al., 1997; Yamada et al., 1996) may be reflective of higher-affinity multivalent forms that would work at more reasonable concentrations.
Laminin-1 binding was first described to the brain form of dystroglycan, which was originally called cranin (Smalheiser and Schwartz, 1987). Brain dystroglycan also binds laminin-2 (Tian et al., 1997). Brain dystroglycan migrates at a lower molecular weight than does dystroglycan isolated from skeletal muscle (Ibrighimov-Beskrovnaya et al., 1992) but shows many similar properties. Binding of laminin-1 is blocked by heparin and sulfatides but not chondrotin sulfate (but see Gee et al., 1993) or lactose (Smalheiser and Kim, 1995). Treatment of dystroglycan with 50 mM periodic acid eliminates binding (Smalheiser, 1993), whereas nitrous acid treatment does not block binding (Smalheiser and Schwartz, 1987). These two experiments suggest that laminin binding occurs via glycans but does not occur via GAGs (for which there is no evidence on any dystroglycan form). Digestion with O-sialoprotease altered the migration of brain dystroglycan and inhibited laminin-1 binding, suggesting that O-linked glycans are important (Smalheiser and Kim, 1995). The biochemical data on brain dystroglycan is likely to reflect an amalgam of glycoforms, as dystroglycan is expressed in the vasculature as well a variety of neural cells and structures (Gorecki et al., 1994; Montanaro et al., 1995; Drenckhahn et al., 1996; Tian et al., 1997; Koulen et al., 1998; Moukles et al., 2000; Zaccaria et al., 2001), including synapses (Cavaldesi et al., 1999; Levi et al., 2002; Zaccaria et al., 2001; Smalheiser and Collins, 2000). Therefore, the degree to which the binding studies on whole-brain protein reflect the function or glycosylation of dystroglycan in various types of neurons or glia is unclear.
Recent work by Michele et al. (2002) has lent further credence to the notion that the binding of laminins and (other ligands) to dystroglycan is dependent on glycans. Here gel overlays and solid state binding assays were performed using dystroglycan isolated from the muscles of patients where glycosylation had been altered. These included patients with muscle-eye-brain disease, Fukuyama congenital muscular dystrophy, and the myodystrophy mouse. In muscles from all of these samples, there is no binding of either the IIH6 or VIA4-1 antibody to dystroglycan. There is little or no binding of laminin-1 as well. This study is consistent with the findings of Ervasti and Campbell (1993) and suggests that laminin binding is dependent on glycans. Thus although the glycan structure required by IIH6 is unknown, and high-affinity binding of laminin to particular glycans has yet to be shown, there is a wealth of evidence to suggest that laminins bind dystroglycan via glycan chains.
Studies on recombinant proteins
G domains of laminin 1, 2, and 4 bind to dystroglycan isolated from chick muscle (Talts et al., 1999). Binding of recombinant proteins containing all five G domains as well as smaller groupings of two to three domains and individual domains were studied. Both the G1–3 and the G4–5 fragments of laminin 2 bound dystroglycan with 15–50 nM affinity. The G4–5 fragment of laminin 1 bound at 
200 nM, whereas the G1–3 fragment did not bind at all. Binding of all of these laminin fragments did not differ between dystroglycan isolated from kidney and skeletal muscle. None of the laminin G domains bound when studied as single domains (at concentrations below 500 nM). Thus certain laminin G domains participate in dystroglycan binding, but all require more than one domain for high-affinity binding. This conclusion is made stronger by comparative studies with the normally synaptic G domains of laminin 4 (Talts et al., 2000). The G domains of laminin 4 bind 30–100 times more poorly than those of laminin 2 and were not saturating even at 1 µM. This occurred despite the fact that laminin 4 bound equally well as laminin 2 to fibulin-1, fibulin-2, heparin, and sulfatides. Thus it is unlikely that the protein produced was simply not folded correctly. Binding of laminin 1 and 2, both as G domains and as native proteins, is also differentially sensitive to heparin inhibition, with laminin 1 being more sensitive than laminin 2 (Pall et al., 1996; Talts et al., 1999). Finally, a recombinant form of laminin 5 produced in Escherichia coli has been reported to bind to dystroglycan using a gel overlay (Shimizu et al., 1999). Such studies show that glycosylation of the laminin G domain is not important for binding.
Crystal structures
The crystal structure of several laminin G domains has been solved and has shed a great deal of light on how these domains might interact with glycans on dystroglycan (Hohenester et al., 1999; Tisi et al., 2000). The structures of the G5 and G4–5 regions of laminin 2 have been solved. These show that individual G domains are folded into a beta sandwich composed of 14 beta strands in two sheets. The resulting G domains are roughly spherical in shape with a diameter of 3.5 nm, which is consistent with earlier electron microscopy studies. The edge of one beta sandwich binds calcium. Calcium is required for binding of laminins to dystroglycan, and mutagenesis studies on laminin G domains suggest that most of the amino acids required for dystroglycan binding reside near the calcium ion in the crystal structure (Timpl et al., 2000). This same region also contains amino acids involved in binding of laminins to heparin and sulfatides (Timpl et al., 2000), which may explain why heparin inhibits laminin binding to dystroglycan. The amino acids required for binding to heparin are not identical to those required for binding to dystroglycan. Thus heparin and dystroglycan binding to laminins are distinct to some extent.
In the G4–5 crystal structure, the G4 and G5 domains associate via an interface of 450 Å, with both calcium ions being at the periphery of the structure. Based on amino acid mutagenesis, laminin residues important for dystroglycan binding span a 3-nm distance within these individual G domains (Timpl et al., 2000). Because at least two domains are required for binding, it is possible that the binding site of laminin for dystroglycan may span as much as 6 nm. Such a large binding site would be well suited to binding multiple glycan chains such as one would find in a mucin domain. Thus G domains of laminins and other dystroglycan binding proteins may be specific to glycoforms of dystroglycan that contain multiple O-linked sites.
Agrin
Several groups simultaneously reported that agrin can bind dystroglycan (Campanelli et al., 1994; Gee et al., 1994; Bowe et al., 1994). Agrin is an extracellular matrix proteoglycan involved in neuromuscular formation (for a review, see Sanes and Lichtman, 1999). Mice lacking agrin fail to properly form neuromuscular junctions and die at birth (Gautam et al., 1996). Agrin is made by both muscle and neurons (Hoch et al., 1993), however, splice forms specific for the nervous system have a high activity in inducing the aggregation of nicotinic acetylcholine receptors (AChRs) in skeletal muscle (Ferns et al., 1992, 1993). Because the concentration of AChRs is an essential step in the development of the neuromuscular synapse, identifying receptors and binding proteins for agrin has been the focus of a number of groups.
The idea that dystroglycan was a functional receptor for agrin-induced AChR clustering was enticing (e.g., see Gee et al., 1994), however, four results argue that dystroglycan is not such a receptor. First, mice lacking dystroglycan in their skeletal muscles have AChR aggregates at the neuromuscular junction (Cote et al., 1999). The structure of the neuromuscular junctions in these animals is perturbed, suggesting a structural role for dystroglycan at the synapse; however, neuromuscular junctions still form. Second, muscles from mice lacking dystroglycan can make AChR aggregates in culture using purified agrin (Grady et al., 2000). Again, these aggregates are less stable, suggesting a role for dystroglycan in AChR cluster stability. Third, a tyrosine kinase called MuSK is stimulated by agrin (Glass et al., 1996) and is essential for agrin signaling and AChR clustering (DeChiara et al., 1996). Fourth, splice forms of agrin that are active in AChR clustering actually bind more poorly to dystroglycan than do splice forms derived from skeletal muscle (Campanelli et al., 1996; Gesemann et al., 1996), which have little or no activity in such assays. The fact that dystroglycan is essential for this important function of agrin does not make its interaction with this ligand unimportant, however, as neuromuscular structure is clearly aberrant in muscles lacking dystroglycan protein (Cote et al., 1999). In addition, muscles lacking one of several putative laminins, which bind dystroglycan, also have aberrant neuromuscular structure (Patton et al., 1999, 2001), as do muscles lacking utrophin (Grady et al., 1997; Deconinck et al., 1997). Thus dystroglycan is an integral part of the synaptic utrophin–glycoprotein complex.
The region of agrin that binds to dystroglycan has been mapped by deletion analysis to the second (G2) of agrin's three G domains, all of which reside in the C-terminal half of the molecule (Gesemann et al., 1996). Like laminins, a tandem of two G domains (G2 and G3) is far superior for binding when compared to a single G domain, and binding requires mM levels of calcium (Gesemann et al., 1996; Campanelli et al., 1996). The third G domain (G3) contains the splice site that defines the bioactive neural-specific agrin and the inactive muscle form (Ferns et al., 1992; Gesemann et al., 1996). The G2 domain also can contain a four amino acid splice insert that conveys heparin binding on dystroglycan (Ferns et al., 1992, 1993). As with the laminin G domains, heparin and dystroglycan binding appear to have significant overlap. The G2–G3 region of neural agrin binds more poorly to dystroglycan and to heparin than does the muscle form (Gesemann et al., 1996; Campanelli et al., 1996), which binds in the 2–5 nM range (Gesemann et al., 1998).
Binding to chemically deglycosylated dystroglycan has not been reported; however, agrin does bind to multivalent neoglycoconjugates, including NeuAc2,3Galß1, 4GlcNAcß-BSA, in the 10–20 nM range (Xia and Martin, 2002). Because such structures contain three of the four saccharides present on the mannose-O-linked chains of dystroglycan (Sasaki et al., 1998), agrin could bind dystroglycan in large part via such glycans. Surprisingly, agrin bound equally well to multivalent Galß1,4GlcNAcß- (or Galß1,3GalNAc-) as it did to NeuAc2,3Galß1,4GlcNAcß-, suggesting that sialic acid contributed little to binding. This would be consistent with the findings of Ma et al. (1993) in Torpedo. There they showed that peanut agglutinin, a Galß1,3GalNAc-O-binding lectin, blocks most agrin binding to muscle membranes. This lectin binds poorly to glycans containing sialic acid, and therefore would not presumably inhibit binding to any NeuAc2, 3Galß1,3GalNAc that may have been present.
Agrin binding to N-acetyllactosamines (LacNAcs) would also be consistent with the findings that sialidase treatment of muscle cells increases agrin-independent clustering of AChRs (Martin and Sanes, 1995). Neuraminidase treatment mimics agrin signaling, and this activity is also blocked by peanut agglutinin (Parkhomovskiy et al., 2000). AChR clustering induced by either agrin or laminin-1 is also inhibited by treatment of muscle cells with certain glycosidases, again suggesting that glycans are involved in signaling (Martin and Sanes, 1995; McDearmon et al., 2001). Finally, binding of agrin to dystroglycan is severely diminished in muscle homogenates where dystroglycan glycosylation is altered (Michele et al., 2002). All of these studies suggest that agrin, like laminin, can bind dystroglycan via glycan chains.
Agrin could also effect dystroglycan binding by proxy through its many other binding partners. For example, agrin binds both laminin-1 and laminin-2 (Denzer et al., 1997, 1998; Cotman et al., 1999). The structural interplay between these two matrix proteins could alter access of the G domains to receptors on the muscle membrane. This interaction may be very complex, as laminin–agrin interactions are partly via protein–protein interactions and partly via heparan sulfate GAGs, which are present on agrin (Cotman et al., 1999). Both laminin and agrin also bind at least half a dozen other proteins each, and dystroglycan, laminin, and agrin all bind heparin (Talts et al., 1999; Pall et al., 1996; Gesemann et al., 1996; Campanelli et al., 1996). Thus it is important to keep in mind that most of the binding studies to date on laminin and agrin have been done using soluble (and often recombinant) ligands, and these proteins may not reflect the behavior of these polymeric matrix molecules as they exist in the basal lamina.
Neurexins
Neurexins are transmembrane proteins that are expressed in the nervous system and are one of the major binding proteins for laitrotoxin, the venom of the black widow spider (Ushkaryov et al., 1992). There are two families of these proteins ( and ß neurexins), and they contain six or one G domains, respectively. Sugita et al. (2001) have shown that both and ß neurexins can bind to dystroglycan with high affinity. Chemical deglycosylation (with TFMS) of dystroglycan inhibits binding of recombinant G domains from neurexins to dystroglycan isolated from skeletal muscle, heart, lung, or brain. Interestingly, enzymatic deglycosylation with a cocktail of N-glycanase, O-glycanase, and sialidase has no effect on binding to dystroglycan from heart or muscle but increases binding to the lung form. Although there was little logic to the use of this combination of reagents, it nevertheless showed that glycosylation of dystroglycan is heterogeneous in different tissues. Michele et al. (2002) have also shown greatly reduced binding of recombinant neurexin to dystroglycan from patients in whom dystroglycan glycosylation has been altered.
Perlecan
Peng et al. (1998) have shown that perlecan, another proteoglycan present in the muscle basal lamina, can bind to dystroglycan from sheep brain in gel overlays. Perlecan can also be coprecipitated with dystroglycan from Xenopus muscle extracts. Talts et al. (1999) have shown binding of a recombinant fragment of perlecan to chick muscle dystroglycan in the 3 nM range. The first two of the three G domains in perlecan are required for this high-affinity binding. Recombinant perlecan binds with stronger affinity than either recombinant laminin 1 or 2. There is no evidence regarding the extent to which perlecan binds via glycans. Based on comparisons with laminin G domains, it is likely that these interactions are similar, but there is some anecdotal evidence that this might not be the case. First, perlecan binding to dystroglycan is not blocked by heparin, whereas heparin does block binding of some (but not all) laminins (Pall et al., 1996; Talts et al., 1999). Second, a 100-fold excess of laminin only blocks perlecan binding by 50% or less, depending on the fragment used (Talts et al., 1999). Likewise, mice lacking perlecan appear to have relatively normal muscle and neuromuscular structure (Arikawa-Hirasawa et al., 2002). The main deficit at the neuromuscular junction appears to be a complete lack of acetylcholinesterase expression (Arikawa-Hirosawa et al., 2002). This is the enzyme that degrades acetylcholine in the synaptic cleft after it has been secreted by the motor nerve terminal. Thus muscles in these animals could be hyperexcitable due to an inability to degrade neurotransmitter.
Biglycan
Recently, Fallon and colleagues have shown that biglycan, a small chondroitin sulfate proteoglycan, binds dystroglycan (Bowe et al., 2001). Unlike perlecan, laminins, and agrins, biglycan does not have any G domains. There are several aspects of this interaction that delineate it from the binding of these other extracellular matrix ligands. First, biglycan binding to dystroglycan does not require the glycosylation of dystroglycan to occur (Bowe et al., 2001). Biglycan binding does not occur in the mucin region but via the C-terminal domain of the dystroglycan protein. Second, chondroitin sulfate chains on biglycan are required for binding. Recombinant biglycan produced in bacteria does not bind to dystroglycan, and chondrotinase treatment inhibits binding of native biglycan. Thus the conventional wisdom for matrix ligands—that glycans on dystroglycan are required but glycans on the ligand are not—has been turned on its head in this instance. This has profound implications for dystroglycan function because it suggests that dystroglycan could not only bind the extracellular matrix via its glycans but also can bind glycans in the matrix as well. Biglycan is reported to be upregulated in mdx muscles, suggesting that it may be involved in compensatory interactions in muscular dystrophy (Bowe et al., 2001). Expression of dystroglycan, however, is reduced, not increased, in mdx muscle. Thus this result would appear to argue that biglycan does not bind strongly enough to dystroglycan to control its localization on the muscle membrane.
Viruses and bacteria
Lymphocytic chorionic meningitis virus (LCMV), Lassa fever virus (LSV), and mycobacterium leprae have all been shown to bind to dystroglycan (Cao et al., 1998; Rambukkana et al., 1998). Mycobacterium leprae requires laminin 2 as a cofactor for binding. Treatment of dystroglycan with 0.1 mM periodate, which will remove N-acyl side chains from sialic acids (Diaz and Varki, 1985), blocks mycobacterium leprae binding but not heparin, suggesting that there is no overlap between the mycobacterium leprae and heparin binding sites, but that glycans are required (Rambukkana et al., 1998). Dystroglycan also inhibits binding of mycobacterium leprae–laminin complexes to rat and human Schwann cells. LCMV and LSV bind directly to dystroglycan. Both viruses require native protein for binding and cannot bind to recombinant forms of the protein, suggesting that glycans are involved. LCMV binding also can be abolished by pretreatment of dystroglycan with 10 mM sodium periodate (Kunz et al., 2001).
Neuraminidase treatment or addition of as much as 100 mM sialic acid, in contrast, does not block binding, nor does heparin (Kunz et al., 2001). These data would appear to be inconsistent, though no analysis of the neuraminidase-treated material was done to confirm that the enzyme actually worked. By expressing recombinant fragments of dystroglycan in embryonic stem cells lacking dystroglycan, Kunz et al. (2001) showed that a portion of the mucin domain and an adjoining region of the N-terminal domain are required for virus binding. Thus, unlike laminins, some additional polypeptide structure in dystroglycan may be required for virus binding. Nevertheless, laminin can block most virus binding in competition experiments. Although not entirely consistent, these data for the most part suggest that glycans mediate pathogen binding to dystroglycan.
|
Other transmembrane scaffolds to the extracellular matrix potentially involving glycans |
|---|
|
TopAbstractIntroductionMuscular dystrophy and the...The neuromuscular junction and...Glycosylation of {alpha}...Role of {alpha} dystroglycan...Other transmembrane scaffolds to...ConclusionsReferences |
|---|
Most studies of laminin binding have suggested that glycosylation of laminin itself is not required for binding to
dystroglycan. Nevertheless, laminin is a highly glycosylated protein, and glycans on laminin could modulate its binding to other cell surface receptors. The only glycan sequences known for laminin involve EHS tumor-derived laminin-1 (Arumgham et al., 1986; Fujiwara et al., 1988; Knibbs et al., 1989; Tanzer et al., 1993). Laminin-1 has approximately 40 N-linked oligosaccharides per molecule, and this makes up roughly 20% of the molecular weight of the protein. Little if any O-linked glycan is present. N-linked structures include bi-, tri-, and tetraantennary complex type oligosaccharides, high-mannose-type oligosaccharides, polylactosaminyl side chains with repeating Galß1,4GlcNAcß1,3- units, and complex type oligosaccharides with nonreducing termini containing either sialic acid, -Gal, ß-Gal, or ßGlcNAc, as well as -Gal linked to N-acetyllactosamine.
Aside from dystroglycan, integrins are the main binding proteins for laminin. Integrins are heterodimeric transmembrane proteins that mediate many important processes involving laminin signaling (Hynes, 2002). There a large number of integrin and ß chains, and different ß integrin heterodimers bind matrix proteins with varying specificities. In skeletal muscle, the principal integrin along the myofiber membrane and at the neuromuscular and myotendinous junction are splice forms of integrin 7ß1 (Martin et al., 1996). Mutations in integrin 7 cause a form of congenital myopathy (Hayashi et al., 1998). In addition, antibodies to the integrin 7 block AChR clustering induced by agrin or laminin-1 in skeletal muscle cells (Burkin et al., 1998, 2000), as do antibodies to integrin ß1 (Martin and Sanes, 1997). Recently, antibodies to integrin 6ß1 were also shown to block laminin-2-induced AChR clustering in skeletal muscle cells (Smirnov et al., 2002). Integrin 6, however, is not significantly expressed in adult human skeletal muscle (Martin et al., 1996).
Although there is little evidence that laminins bind integrins via glycans, there are studies that suggest that glycans on laminin modulate integrin binding and/or activity. Integrin 6 can utilize glycans in part to bind to laminin (Chammas et al., 1991). There is no evidence that integrin 7 binds laminin via glycans; however, both laminin-1 (Zhou and Cummings, 1993) and integrin 7 (Gu et al., 1994) bind the lactose lectin L14. L14 is a member of the S-type lectin family, some members of which are highly expressed in skeletal muscle (Catt et al., 1987; Poirer et al., 1992; Cooper and Barondes, 1990; Cooper et al., 1991). L14 binds both laminin-1 and integrin 7, and it can inhibit the binding of these molecules to one another (Gu et al., 1994). Therefore, integrin 7-laminin interactions could be bridged by lactosamine moeities on laminin via L14 or another lactose-binding lectin. Laminin also binds a number of glycans directly. These include heparin (Talts et al., 1999), sulfatides (Talts et al., 1999), and sulfoglucuronyl glycolipids (Mohan et al., 1990). Thus glycans on receptors other than dystroglycan may serve as scaffolds to laminin, and glycans on laminin may form scaffolds to membrane receptors.