The roles of dystroglycan in the nervous system: insights from animal models of muscular dystrophy

Dystroglycan is a receptor of extracellular matrix (ECM) proteins in many developing and adult mammalian tissues (Ibraghimov-Beskrovnaya et al., 1993; Blaeser et al., 2018), and is composed of two protein subunits translated from a single mRNA transcript of the DAG1 gene (Ibraghimov-Beskrovnaya et al., 1992). The α-subunit, designated as α-dystroglycan (αDG), resides at the outer surface of the plasma membrane, where it shares a tight noncovalent bond with the membrane-spanning β-subunit (βDG) (Holt et al., 2000; Akhavan et al., 2008). The intracellular domain of βDG interacts with cytosolic proteins, most notably those of the dystrophin family (Xin et al., 2000; Palmieri et al., 2017). Together, αDG, βDG and dystrophin represent the core functional unit of the dystrophin-glycoprotein complex, physically linking ECM and cytoskeletal elements (see Box 1 for an overview of dystroglycan structure and interactions).

The expression of this protein complex is widespread; αDG is found in cells of the skeletal muscle, nervous system, digestive tract, kidney, skin and reproductive organs (Durbeej et al., 1998). Many functions have been ascribed to αDG, depending on developmental and cell-specific contexts. αDG participates in basement membrane formation (see Box 2 for a glossary of terms) and in signal transduction from the ECM (Gracida-Jiménez et al., 2017). Further, through physical anchorage with the surrounding matrix, αDG protects muscle cell membranes against contraction-induced damage (Han et al., 2009).

Aside from binding endogenous extracellular ligands, αDG is a receptor for Mycobacterium leprae – the organism responsible for leprosy – and for the Lassa hemorrhagic fever virus and the lymphocytic choriomeningitis virus (Cao et al., 1998; Rambukkana et al., 1998). The interaction between αDG and many of its extracellular binding partners is mediated by glycosylation (Box 2). Loss-of-function mutations in certain glycosyltransferase enzymes, discussed in more detail below, cause αDG hypoglycosylation (Box 2), which diminishes the binding affinity between αDG and its extracellular ligands (Fig. 1A,B).

In humans, the hypoglycosylation of αDG is characterized by progressive muscular dystrophy frequently associated with brain malformation and intellectual disability – a spectrum of recessive genetic disorders referred to as α-dystroglycanopathies. Researchers frequently use animal models of α-dystroglycanopathies to study the dysfunction of αDG in the brain, eye and peripheral nerve, because these are the least accessible components of the human pathology. Here, we summarize findings from these animal models and derive key inferences regarding the functions of αDG in the nervous system, with particular focus on their relevance to the human disorders.

αDG is so heavily glycosylated that carbohydrates constitute over half of the glycoprotein's mass. Of these carbohydrate structures, matriglycan (Box 2) is directly responsible for the binding of αDG to its ECM ligands. It is comprised of a massive repeating disaccharide (-3Xylα1-3GlcAβ1-) that tightly binds LG domains (Box 2) in the laminin, agrin (Box 2), perlecan, slit, neurexin and pikachurin extracellular proteins (Briggs et al., 2016). Matriglycans are synthesized by the enzyme Large and are bound to αDG through a tandem ribitol phosphate and a core O-linked trisaccharide (GalNAcβ1-3GlcNAcβ1-4Man-) (Kanagawa et al., 2016). The maximum quantity and repeat length of matriglycan molecules on αDG is not known.

To date, 16 genes have been identified as definitive contributors to the construction of functional matriglycans on αDG (Table 1, Fig. 2). A 17th gene, POMGNT1, is not directly involved in matriglycan synthesis but is thought to regulate its installation on αDG (Yoshida-Moriguchi et al., 2010). Loss of function in any of these genes compromises matriglycan structure and impairs αDG ligand-binding affinity, leading to a clinical condition called secondary α-dystroglycanopathy. In such cases, αDG maintains normal tissue expression and localization, but is hypoglycosylated (Box 2).

Rare mutations in DAG1 (Box 2), the gene encoding αDG and βDG, have also been reported to interfere with the process of matriglycan addition, either through inhibiting the extension of matriglycan (Hara et al., 2011) or by perturbing the maturation and trafficking of αDG (Geis et al., 2013; Signorino et al., 2018). Clinical conditions arising from mutations in DAG1 itself are referred to as primary α-dystroglycanopathies. Frameshift mutations in DAG1 can completely abrogate αDG and βDG, resulting in particularly severe clinical manifestations (Riemersma et al., 2015). Often, but not always, the degree of αDG hypoglycosylation correlates with disease severity (Jimenez-Mallebrera et al., 2009; Alhamidi et al., 2017).

New genes continue to be implicated in α-dystroglycanopathies and further expand our view of αDG biosynthesis. Mutations in the Golgi membrane trafficking proteins TRAPPC11 and GOSR2 (Table 1) were recently reported to cause hypoglycosylation of αDG linked to muscular dystrophy, epilepsy and brain abnormalities (Larson et al., 2018). Given the more general nature of an ER-to-Golgi transport defect, it is likely that hypoglycosylated αDG is not the only disease mechanism for these two genes. Because many patients still lack a molecular diagnosis (Godfrey et al., 2007; Mercuri et al., 2009; Graziano et al., 2015), there are probably additional genes involved in α-dystroglycanopathies waiting to be identified.

α-Dystroglycanopathies encompass a wide spectrum of disease severities that include disrupted nervous system development and progressive muscular dystrophy (see Box 3 for details on the muscular phenotypes of α-dystroglycanopathies). The most severely affected patients present with profound brain malformation and ultimately do not survive infancy – a phenotype referred to as Walker–Warburg syndrome (Cormand et al., 2001). Affected individuals can display profound cognitive deficits, hydrocephaly (Box 2), and brain and retinal dysplasia, some of which may be observed by ultrasound during gestation (Vohra et al., 1993; Trkova et al., 2015). Other syndromes recognized in the spectrum of α-dystroglycanopathy include (in order of decreasing severity) muscle-eye-brain (MEB) disease, Fukuyama-type congenital muscular dystrophy, and several forms of congenital and limb-girdle muscular dystrophies (reviewed by Godfrey et al., 2011).

A common feature of severe α-dystroglycanopathies is cobblestone lissencephaly (Box 2) and an array of neurological defects (Fig. 3A). Light and electron microscopy of postmortem tissue establish a general theme of aberrant cell migration underlying such nervous system dysplasia (Nakano et al., 1996). Examination of the cerebral cortex, cerebellum, retina, brainstem and spinal cord shows gross morphological abnormalities due to displacement of neurons both within the tissue and beyond its normal borders (Yamamoto et al., 1997, 2008). In some cases, there is evidence of atrophy in the spinal cord and retina (Meilleur et al., 2014), sometimes accompanied by progressive dysfunction of ocular physiology (Santavuori et al., 1989; Pihko et al., 1995; Cormand et al., 2001; von Renesse et al., 2014).

Over two decades of accumulated research triangulates αDG dysfunction as a primary causative mechanism for this class of muscular dystrophies with central nervous system involvement. In mice, disruption of the genes involved in the synthesis of matriglycans on αDG, or in Dag1 itself, results in animals that recapitulate the human condition in many respects (Michele et al., 2002; Moore et al., 2002). Despite our knowledge regarding the genetic etiology of α-dystroglycanopathies, the pathogenic events leading to such a heterogeneous spectrum of disease remain unclear.

An increasingly rich assortment of model systems is currently available for investigating α-dystroglycanopathies. Interpreting the literature necessitates considering the strengths and weaknesses of each model, and selecting the best model requires factoring these caveats into the goals of future studies.

Traditionally, mice have been the preferred animal model for α-dystroglycanopathies. There are at least 41 distinct genetic mouse models: 15 that directly mutate or delete Dag1 itself (Table 2), and 26 that model the hypoglycosylation of αDG through mutation or deletion of glycosyltransferase genes (Tables 3 and 4). Global deletion of Dag1 or the glycosyltransferases Pomt1, Pomt2, Fktn or Fkrp is embryonic lethal in mice due to disruption of Reichert's membrane, a murine basement membrane barrier between embryonic and placental tissue (Williamson et al., 1997; Willer et al., 2004; Kurahashi et al., 2005; Chan et al., 2010; Hu et al., 2011a,b). Defects in Reichert's membrane also occur upon knockout of glycosyltransferase enzymes involved in matriglycan synthesis (Willer et al., 2004; Kurahashi et al., 2005; Chan et al., 2010; Hu et al., 2011a). This membrane is not present in human embryos, and thus these models are only practical for assessing the role of αDG in early murine development.

Many Cre-driver mouse lines (Box 2) are used for tissue-specific deletion of Dag1, allowing embryonic survival and therefore the study of phenotypes in older animals. Particularly useful have been the GFAP-Cre, NEX (also known as Neurod6)-Cre and Nes (also known as nestin)-Cre mice, which drive recombination in glia, neurons and neural stem cells, respectively. Crossing these lines with floxed Dag1 mice is a common strategy for investigating dystroglycan's functional contributions in distinct cell types (Satz et al., 2010). Neuron subtype-selective approaches are also possible – for example, Pcp2-Cre mice exhibit recombination primarily in cerebellar Purkinje neurons (Nguyen et al., 2013).

A recent comparison of α-dystroglycanopathy mouse models demonstrated significant phenotypic differences depending on the Cre-driving promoter. Gene knockout driven by Nes-Cre showed a considerably milder effect compared with that driven by Emx1-Cre (Sudo et al., 2018). Neural stem cells of the cerebral cortex express both Nes and Emx1, but Nes-Cre-driven knockout was inefficient in early stages of development (Liang et al., 2012). Because the timing of αDG loss of function has important phenotypic consequences, Emx1-Cre mice might be a preferred system for investigating the functional contribution of αDG during fetal brain development.

Knockdown of glycosyltransferase genes, rather than complete ablation, is another common approach to generate viable postnatal mouse models of α-dystroglycanopathies. For example, various levels of knockdown in Fkrp – a glycosyltransferase critical for matriglycan synthesis – mimic the broad clinical variability seen in α-dystroglycanopathies (Chan et al., 2010; Blaeser et al., 2013). A comparison between Fkrp, Pomgnt1 and Large (also known as Large1; Box 2) mutant mouse strains has elucidated gene-specific phenotypes and exemplifies the utility of these diverse models (Booler et al., 2015).

In addition to mouse models, zebrafish are commonly used owing to their short life cycle and feasibility of targeted gene knockdown with morpholinos (Table 5). Although comparisons between zebrafish and human phenotypes are limited, it is a useful strategy for rapidly confirming the functional significance of newly identified genes in α-dystroglycanopathies. Other non-mammalian models include chick, Xenopus, Drosophila and Caenorhabditis elegans, as these systems are amenable for high-throughput genetic screens or allow greater access for experimental manipulation. This abundance of animal models allows for a flexible approach to studying the genetic and phenotypic spectrum of α-dystroglycanopathies.

Patients with α-dystroglycanopathy show disruption of basement membranes in both muscle and brain (Saito et al., 1999; Vajsar et al., 2000; Goddeeris et al., 2013). Basement membrane synthesis is believed to begin with soluble extracellular laminin binding to cell membrane galactosyl-sulfatide glycolipids (Li et al., 2005). At high local concentrations, laminins polymerize to form a network at the cell surface, recruiting additional basement membrane proteins and mobilizing integrin and αDG to strengthen the nascent matrix. Thus, αDG is not required for initial basement membrane formation, but likely participates in its maintenance.

As in humans, loss of functional αDG in experimental animal models results in gross malformation to the nervous system, including lissencephaly and hydrocephaly (Satz et al., 2010). Such phenotypes are often accompanied by microscopic ruptures in basement membranes of the neocortex, hippocampus, cerebellum, retina and spinal cord, with abnormal migration of neurons through these breaches (Ackroyd et al., 2009; Li et al., 2011; Wright et al., 2012; Nguyen et al., 2013).

Because Dag1-null mouse embryos do not survive gastrulation, other methods have been sought to probe the role of αDG in basement membrane integrity. Mouse embryonic stem cells can be cultured in suspension to form embryoid bodies (Box 2). An initial report observed complete absence of basement membrane in embryoid bodies derived from Dag1-null embryonic stem cells (Henry and Campbell, 1998), but later studies show that such Dag1-null embryoid bodies can, in fact, form a basement membrane (Li et al., 2002).

Basement membranes produced in Dag1-null embryoid bodies can develop an abnormally thick morphology, suggesting reduced compaction or increased matrix secretion (Li et al., 2002). The epiblast-like cells of embryoid bodies express both αDG and integrin, two major classes of cell surface laminin receptors. In the absence of either receptor, the other can compensate to assemble a basement membrane if high concentrations of laminin proteins are present. However, knockout of both receptors prevents basement membrane formation entirely (Li et al., 2017). Although αDG is not required for the initial polymerization of laminin (Li et al., 2005), it may reinforce and compact the basement membrane through collateral linkages to laminin, agrin and perlecan (Goddeeris et al., 2013).

When Dag1 knockout is restricted to the epiblast of the developing mouse, the embryo successfully progresses through gastrulation but develops severe malformation of the nervous system (Satz et al., 2008). Interestingly, selective deletion of Dag1 from various neuron subtypes in NEX-, PCP2- and malpha6 (also known as Tuba1c)-Cre mice, or from the meninges in Wnt1-Cre mice, does not cause nervous system malformation (Satz et al., 2009, 2010; Hu et al., 2011a; Nguyen et al., 2013). Likewise, removing the cytoplasmic C-terminus of βDG has no apparent effect (Satz et al., 2010).

However, αDG is highly expressed in glial cells contacting the basement membranes that ensheath the brain, retina and neural vasculature (Zaccaria et al., 2001). Knockout of glial Dag1 in GFAP-Cre mice leads to fragmentation of basement membranes overlying the cerebral cortex and cerebellum (Moore et al., 2002; Satz et al., 2010; Nguyen et al., 2013). Various mouse models expressing a hypoglycosylated αDG also show similarly disrupted basement membranes (Tables 3 and 4) (Michele et al., 2002; Lee et al., 2005; Satz et al., 2010; Takahashi et al., 2011). Together, these animal models suggest that αDG has an important receptor function in glial cells involving the integrity of brain ECM structure.

During mammalian brain development, neuroepithelial cells extend from the ventricular zone to contact the pia mater (Box 2) at the brain's surface (McLone, 1980). These neuroepithelial cells, which later differentiate into radial glia, form a layer of endfoot processes, called the glia limitans, in close apposition to the pia mater (Fig. 3B). Meningeal fibroblasts of the pia mater deposit ECM components that are sandwiched to form a basement membrane between the pia mater and underlying glial endfeet (Sievers et al., 1994). Radial glia rely on the basement membrane to maintain their morphology and localization. In turn, they remodel and strengthen its framework through receptor-matrix interactions (Sievers et al., 1986). In vitro cultures of meningeal cells and glia spontaneously form basement membrane structures at the interface between the two cell types, demonstrating that both the meninges and glia are necessary for the formation of the pial basement membrane (Abnet et al., 1991; Struckhoff, 1995). Importantly, the pial basement membrane and its associated glial endfeet act as a scaffold for neuronal migration and establishment of the brain's cellular architecture (Halfter et al., 2002).

Ruptures in the pial basement membrane are characteristic of severe forms of α-dystroglycanopathy (Fig. 3B). Such ruptures correlate with disorganized glial endfeet and abnormal neuronal migration (Hu et al., 2007; Yang et al., 2007). A common feature is the accumulation of neurons beyond the cortical boundary, presumably due to over-migration through basement membrane breaches (Booler et al., 2015). The underlying cortical plate is also disrupted by irregular neuronal orientation and premature termination of migration, interrupting the two major modes of radial neuronal migration: radial glia-guided and radial glia-independent migration (Nakagawa et al., 2015). This is speculatively due to the apparent disorganization of glial processes and fragmented basement membrane, respectively. Similarly, both the hippocampus and the cerebellum experience neuronal under-migration, resulting in abnormal morphology (Liu et al., 2006; Nguyen et al., 2013). Additional neuronal migration defects can be found in the retina, hindbrain and spinal cord across mouse, zebrafish and Drosophila models (Table 5) (Lunardi et al., 2006; Qu et al., 2006; Shcherbata et al., 2007; Thornhill et al., 2008; Kawahara et al., 2009; Gupta et al., 2011; Marrone et al., 2011a,b; Wright et al., 2012).

A subset of α-dystroglycanopathy patients develop both pachygyria and polymicrogyria – abnormally thick and thin convolutions of the cerebral cortex, respectively – specifically in the frontal and parietal lobes (Aida et al., 1996; Meilleur et al., 2014; Yoshioka et al., 2017). Occipital subcortical heterotopia, cerebellar cysts and other defects across many brain regions also occur (Clement et al., 2008). In contrast, brain malformations in Pomgnt1-null and Fkrp(Y307N) mice follow a lateromedial and rostrocaudal gradient of severity. Large^myd mice show a more consistent abundance of lesions across the midbrain and cortex (Booler et al., 2015). There are no reports of a lesion gradient in α-dystroglycanopathy patients that match that of Pomgnt1-null and Fkrp(Y307N) mice. The underlying mechanisms behind such diversity of neural dysplasia (Box 2) between patients and mice are unknown. Remarkably, despite widespread disorganization to the brain's usually laminar architecture, many neurons still differentially express layer-specific markers and project to appropriate brain regions in mouse models (Myshrall et al., 2012). This indicates that some axon pathfinding mechanisms are independent of cortical lamination and may help explain the rather mild cognitive dysfunction in some patients with otherwise extensive brain malformation (Godfrey et al., 2007; Meilleur et al., 2014).

Detailed examination of α-dystroglycanopathy model embryos has highlighted pathologic features that mark the onset of abnormal brain development. In Pomgnt2-null and Fkrp(Y307N) mice, ectopic movement of Cajal–Retzius cells and subplate neurons through the pial basement membrane occurs as early as embryonic day 10.5 (Nakagawa et al., 2015; Booler et al., 2016). This is followed by cortical plate neurons streaming through the pial ruptures, accumulating beyond the brain's surface in the subarachnoid space. Cajal–Retzius cells orchestrate cortical lamination by secreting reelin (Box 2) (Ogawa et al., 1995). Thus, this early mislocalization of Cajal–Retzius cells might influence later positioning of cortical plate neurons. However, biochemical analyses indicate that there is no perturbation to global reelin signaling in Pomgnt2-null and Fkrp(Y307N) mouse brains (Nakagawa et al., 2015; Booler et al., 2016).

Meningeal cells of the pia mater and their associated pial basement membrane provide signaling cues and a physical substrate for the migration of Cajal–Retzius cells and other neurons (Borrell and Marín, 2006; Sekine et al., 2012). In α-dystroglycanopathies, the pial basement membrane is believed to form abnormally, potentially rendering it prone to discontinuity and creating a permissive environment for migrating cells to cross its boundary (Hu et al., 2007). Indeed, atomic force microscopy reveals that the retina's inner limiting basement membrane in Pomgnt1-null mice is roughly half as stiff and half as thick as that in wild-type mice (Hu et al., 2010; Zhang et al., 2013). Further, Pomgnt1-null neural stem cells demonstrate up to 60% slower accumulation of surface laminin compared with wild type, consistent with reduced ECM assembly by αDG-deficient cells (Zhang et al., 2013).

Interestingly, mutations in ECM receptor genes other than Dag1 can similarly cause basement membrane breaches. Both GPR56 and integrin-β1 – cell surface receptors of collagen III and laminins, respectively – are expressed on radial glial endfeet abutting the pial basement membrane. Knockout of these receptors causes mislocalization of glial endfeet and aberrant neuronal migration at fragmented regions of the pia mater, ultimately leading to a cobblestone lissencephaly-like cortical dysplasia (Graus-Porta et al., 2001; Li et al., 2008).

Directly perturbing the ECM can also result in pathologic conditions reminiscent of α-dystroglycanopathy. Patients with mutations in laminin subunits α2 and β1 often develop localized cobblestone lissencephaly in the occipital lobe (Geranmayeh et al., 2010; Radmanesh et al., 2013). In animal models, neuronal migration defects are elicited through the perturbation of laminin-α1, laminin-β2, laminin-γ1, laminin-γ3, collagen III α1, collagen IV α1 or collagen IV α2 (Halfter et al., 2002; Pöschl et al., 2004; Luo et al., 2011; Ichikawa-Tomikawa et al., 2012; Radner et al., 2013).

Because disruption of either ECM receptors or the pia mater ECM itself results in comparable pathology, receptors such as αDG are likely to be involved in physically strengthening the pial basement membrane. However, given that αDG is expressed in a wide array of brain cells beyond radial glia, as discussed in detail below, additional factors are surely necessary to account for all disease-related phenomena seen in patients. This is particularly true for patients with mutations affecting the earliest steps of O-mannosylation, which may affect other glycosylated proteins in addition to αDG (Lommel et al., 2013; Vester-Christensen et al., 2013). However, recent work indicates that cadherins and plexins, two major cell surface receptor families, are not modified by the same O-mannosylation enzymes as αDG (Larsen et al., 2017). Therefore, we consider that the impairment of pial basement membrane integrity from specific loss of αDG-matrix interaction is a likely pathogenic mechanism for the neuronal migration defects in α-dystroglycanopathies.

Although gross malformation to the nervous system is common in α-dystroglycanopathies, increasing evidence suggests perturbation at the subcellular level. Abnormalities in cell polarity, channel distribution and cellular signaling could influence the disease phenotype and be intimately tied to the development of brain dysplasia. As a bridge between extracellular and intracellular elements, αDG and βDG coordinate cytosolic proteins with external cues. Thus, in addition to ECM organization, binding between αDG and its ligands may localize subcellular specializations through interaction with βDG (Moore and Winder, 2010). There is also accumulating evidence that βDG can translocate to the nucleus, regulating nuclear envelope structure and gene expression (Martínez-Vieyra et al., 2013; Mathew et al., 2013; Gracida-Jiménez et al., 2017).

In the embryonic chick retina, DAG1 knockdown induces neuroepithelial stem cell detachment from the inner limiting basement membrane, with a concomitant loss of elongated cell morphology and stalling of apicobasal interkinetic nuclear migration (Schröder et al., 2007). Displacement of molecular markers of epithelial polarity is commonly observed in Dg-null Drosophila (Deng et al., 2003; Schneider et al., 2006; Shcherbata et al., 2007). Further, binding between laminin and αDG in mouse embryoid bodies is sufficient to induce polarized morphology of epiblast-like cells and regulate apicobasal orientation (Li et al., 2017). Although the molecular mechanisms have not been precisely defined, there are accumulating data that asymmetrically distributed αDG-laminin interactions can determine cell polarization at basement membrane contact points, perhaps involving cytoplasmic microtubules, CLASP or PAR-1 proteins (Masuda-Hirata et al., 2009; Nakaya et al., 2013).

In the brain and retina, αDG colocalizes with aquaporin-4 water channels and Kir4.1 potassium channels at perivascular glial endfoot processes (Guadagno and Moukhles, 2004). Co-immunoprecipitation experiments showed that both aquaporin-4 and Kir4.1 associate with members of the dystrophin-glycoprotein complex, including βDG (Fort et al., 2008). Binding between αDG and laminin and agrin apparently anchors aquaporin-4 and Kir4.1 within specialized domains at glial endfeet, possibly regulating ion and water homeostasis in the brain (Noel et al., 2005). Animal models of α-dystroglycanopathies accordingly show a deficit of aquaporin-4 and Kir4.1 channels in perivascular glial endfeet (Michele et al., 2002; Rurak et al., 2007; Noell et al., 2011). Notably, deletion of the cytoplasmic C-terminus of βDG is sufficient to reduce Kir4.1 channel clustering in retinal radial glia and produce a reduced electroretinogram b-wave (Box 2), indicating abnormal visual signal transmission in the retina (Satz et al., 2009). Downregulation of βDG and aquaporin-4 in glial endfeet also occurs after epileptic activity (Gondo et al., 2014). This channel clustering defect may broadly perturb interstitial brain water homeostasis and be linked to white matter abnormalities in α-dystroglycanopathy patients (Aida et al., 1996; Cormand et al., 2001; Larson et al., 2018).

In addition to the endfeet of radial and perivascular glia, αDG is expressed in myelinating glia at various stages of maturity (Yamada et al., 1994; Colognato et al., 2007). White matter alterations are frequently seen on brain magnetic resonance imaging (MRI) scans of α-dystroglycanopathy patients, particularly in those with intellectual deficits (Bönnemann et al., 2014). As discussed in the preceding section, some of these signal alterations may be related to an abnormal water content of the white matter owing to aberrant channel function in radial and perivascular glial endfeet. However, further observations also suggest a pathology of myelinated fiber tracts in α-dystroglycanopathies.

In the peripheral nervous system, αDG localizes to the outer aspect of Schwann cell membranes in contact with the overlying basement membrane and is upregulated during the myelination of regenerating nerves (Masaki et al., 2000). Proper folding of the myelin sheath relies on the collective laminin binding activity of αDG and integrin-α6β4 (Nodari et al., 2008). Additionally, αDG interactions with laminin-211 may guide Schwann cell microvilli to axoglial junctions and thus indirectly mediate sodium channel clustering at the node of Ranvier (Box 2) (Saito et al., 2003). Loss of αDG-laminin binding in mice impairs nerve regeneration, myelination and signal conduction with associated behavioral deficits. In multiple mouse models of α-dystroglycanopathies, researchers reported an axon-sorting defect in the sciatic nerve characterized by amyelinated axon bundles, some of which were large enough to expect myelination (Kelly et al., 1994; Rath et al., 1995; Levedakou et al., 2005; Saito et al., 2007). This type of pathology might not be a major feature of the human disease, as it has not yet been documented in patients.

In the central nervous system, αDG is expressed on developing oligodendrocytes, where it associates with laminin and IGF-1 signaling proteins, suggesting a role in oligodendrocyte morphological maturation and differentiation (Galvin et al., 2010). Knockout of Dag1 in neural stem cells compromises the ECM structure of the subventricular zone, a major region of origin for oligodendrocytes, and impedes the maturation of ependymal cells and oligodendrocytes (McClenahan et al., 2016). Further, in vitro disruption of αDG-laminin binding leads to decreased filopodia formation and myelination by oligodendrocytes (Colognato et al., 2007; Eyermann et al., 2012). Collectively, these data place αDG as a physical link between the myelinating cell membrane and the ECM, perhaps acting to regulate its proliferation and differentiation or even to stabilize and guide the process of myelination itself.

αDG is expressed at peripheral nervous system synapses, where it serves apparent scaffolding functions distinct from that of glial-expressed αDG (Bewick et al., 1993; Zaccaria et al., 2001). A subset of α-dystroglycanopathy patients with mutations in GMPPB (Table 1) exhibit striking decreases in muscle action potentials during repeated nerve stimulation. Pyridostigmine (Box 2) treatment reportedly improves motor function in these patients, suggesting an abnormality of neuromuscular transmission at the postsynaptic neuromuscular junction (NMJ; Box 2) (Belaya et al., 2015; Rodríguez Cruz et al., 2016). However, other α-dystroglycanopathy patients do not show abnormalities, and the reason for these differences remains unknown.

αDG resides at the NMJ postsynaptic membrane and was the first identified receptor of agrin, a master organizer of the NMJ that stimulates acetylcholine receptor (AChR) clustering through its interaction with postsynaptic receptors (Bowe et al., 1994; Campanelli et al., 1994; Gee et al., 1994; Sugiyama et al., 1994). Knockout of Dag1 in mice does not prevent NMJ formation – instead, NMJs are smaller with fragmented AChR clusters (Côté et al., 1999; Jacobson et al., 2001). This is likely because MuSK (Box 2) is the primary agrin receptor required for NMJ formation, while αDG and βDG appear to be involved in a MuSK-independent mechanism to stabilize AChR clusters (DeChiara et al., 1996; Glass et al., 1996; Cartaud et al., 1998; Jacobson et al., 1998).

In addition to agrin, αDG binds extracellular perlecan at the NMJ (Peng et al., 1998). Perlecan, in turn, binds and stabilizes acetylcholinesterase (AChE) in the synaptic basement membrane (Peng et al., 1999). Because perlecan-null mice show complete absence of AChE at the NMJ, it is possible that neuromuscular transmission in α-dystroglycanopathies may also be impaired by a disruption of an AChE-perlecan-αDG complex (Arikawa-Hirasawa et al., 2002). Chimeric Dag1-null mice exhibit reduced levels of AChE at the NMJ (Côté et al., 1999; Jacobson et al., 2001); however, a functional consequence of AChE disruption has not yet been shown in the α-dystroglycanopathies or its models.

In apparent similarity to the NMJ, αDG is also expressed at postsynaptic specializations of the central nervous system (Zaccaria et al., 2001). Cognitive impairment and epilepsy are features in many α-dystroglycanopathies (Messina et al., 2009; Di Rosa et al., 2011; Astrea et al., 2018; Larson et al., 2018). For some patients, MRI shows grossly normal brain morphology despite significant cognitive deficits (Godfrey et al., 2007; Clement et al., 2008; Jimenez-Mallebrera et al., 2009). This observation suggests a neuronal dysfunction below the detection limit of current brain imaging tools, such as central synaptic dysfunction, for example.

The expression of αDG in central neurons is restricted to a subset of postsynaptic sites in pyramidal cells of the cerebral cortex and hippocampus as well as cerebellar Purkinje cells (Zaccaria et al., 2001). Immunolabeling of murine hippocampal cultures reveals a selective association between αDG and postsynaptic proteins of γ-aminobutyric acid (GABA) inhibitory synapses: dystrophin, GAD, gephyrin and GABA_A receptor subunits α1, β2/3 and γ2 (Brünig et al., 2002; Lévi et al., 2002; Pribiag et al., 2014). Because αDG is an extracellular membrane-associated protein, it is reasonable to speculate that it may act as part of a trans-synaptic protein complex, perhaps facilitating synapse formation or maintenance. Indeed, characterized ligands of αDG, agrin and the neurexin proteins, are expressed at presynaptic terminals in the brain where they may interact across the synapse with matriglycans on postsynaptic αDG (Ferreira, 1999; Sugita et al., 2001).

The NEX-Cre/DG-null mouse, in which Dag1 is conditionally deleted from pyramidal neurons, shows virtually complete loss of cholecystokinin (CCK) GABAergic presynaptic terminals (Fruh et al., 2016). Despite this selective loss of presynaptic input, there is no change in the total number of GABAergic presynaptic terminals, suggesting a compensatory effect. This is consistent with the in vitro observation that Dag1 knockout does not affect overall GABAergic synapse numbers (Lévi et al., 2002). Knockout of Dag1 after synapse formation in adult animals likewise causes selective loss of CCK terminals, but there is no such phenotype in the Dag1 T190M knock-in mouse – a model of α-dystroglycanopathy expressing a neurexin-binding-deficient αDG (Fruh et al., 2016). Because this process appears independent from αDG-neurexin binding, αDG may interact with additional unidentified presynaptic receptors or indirectly control other trans-synaptic proteins to facilitate CCK GABAergic innervation.

In primary hippocampal cultures, a pharmacological increase in neuronal firing rate induces a compensatory boost in inhibitory synaptic strength, accompanied by increased clustering of αDG and GABA_ARs at the postsynaptic surface (Pribiag et al., 2014). Interestingly, this inhibitory upscaling is partly reproduced by the addition of agrin to the culture medium, and it is blocked by knockdown of Dag1 or the glycosyltransferase Large, demonstrating that the receptor function of αDG is required for this form of homeostatic synaptic plasticity. The inhibitory synaptic defects identified in mouse and cellular models may explain the epilepsy seen in patients (Pribiag et al., 2014; Fruh et al., 2016). However, future work must delineate the various biochemical interactions of αDG and βDG and their potential roles in synapse formation versus synapse plasticity.

While αDG resides at the postsynaptic apparatus of some central synapses and the NMJ, it assumes a unique presynaptic position in the retina. Early work identified αDG and βDG, in association with dystrophin, within the outer plexiform layer, where presynaptic photoreceptor terminals contact postsynaptic bipolar and horizontal cells (Montanaro et al., 1995). Interestingly, both α-dystroglycanopathy patients and animal models present abnormal electroretinograms (ERGs; Box 2), indicating retinal dysfunction (Holzfeind et al., 2002; Takeda et al., 2003; Lee et al., 2005). Some patients, in particular those with the POMGNT1 mutations prevalent in Finnish MEB patients, show overt evidence for retinal degeneration (Santavuori et al., 1989; Pihko et al., 1995).

The synaptic ligand of αDG in the retina, encoded by pikachurin (also known as Egflam), was discovered in a transcriptomic screen comparing retinas from wild-type and Otx2-null mice, a mouse strain lacking photoreceptor cells (Sato et al., 2008). Pikachurin localizes to the synaptic cleft of the specialized ribbon synapse between photoreceptor and bipolar cells. There, it binds the matriglycans on αDG to link presynaptic rod/cone photoreceptors to the postsynaptic bipolar cell dendrite, allowing for rapid communication across the first synapse in the visual system (Kanagawa et al., 2010; Hu et al., 2011b). A recent analysis demonstrates that pikachurin acts as a trans-synaptic bridge, physically joining presynaptic αDG to the postsynaptic orphan receptor GPR179 (Orlandi et al., 2018).

Pikachurin and αDG are mutually required for proper localization, because ablation of either causes a significant reduction to the other (Omori et al., 2012). Ultrastructurally, loss of the photoreceptor αDG-pikachurin complex leads to abnormal synapse formation between photoreceptors and bipolar cells (Sato et al., 2008; Omori et al., 2012; Rubio-Fernández et al., 2018). This uncoupling of the photoreceptor from the bipolar cell was confirmed by ERG in both mice and patients, where the a-wave is unimpaired but the b-wave exhibits a significantly attenuated amplitude (Santavuori et al., 1989; Pihko et al., 1995; Longman et al., 2003; Sato et al., 2008; Omori et al., 2012; Rubio-Fernández et al., 2018).

Based on the wealth of data now available from various experimental animal models, we propose at least two broad, but distinct, categories of nervous system pathology in α-dystroglycanopathies: (1) a histological abnormality arising from dysfunction of glial αDG, largely responsible for the distinctive migration abnormalities during brain development and white matter changes; and (2) a synaptic defect from dysfunction of neuronal αDG. α-Dystroglycanopathy patients often present with intellectual deficits presumably caused by developmental brain malformation. However, an increasingly recognized subset of patients present with significant cognitive delays but normal or near-normal brain structure as determined by MRI (Godfrey et al., 2007; Clement et al., 2008; Jimenez-Mallebrera et al., 2009). We hypothesize that these individuals may harbor a selective defect at the synapse level that contributes to their cognitive dysfunction. This will be important to identify, as synaptic pathology could potentially be amenable to treatments that are targeted at improved αDG function (Box 4).

Few studies have comprehensively assessed cognitive ability in α-dystroglycanopathies, but some reports indicate impaired executive control, memory and visuospatial attention in patients with macroscopically normal brain structure or minimal brain malformation (Palmieri et al., 2011; Meilleur et al., 2014). The Large^myd and Emx1-Cre/DG-null mouse models of α-dystroglycanopathy similarly show a memory consolidation deficit (Comim et al., 2013, 2014; Hu et al., 2016). Attention and memory are correlated with gamma oscillations of 30-80 Hz in the cortex – a frequency of electrical activity synchronized partly by CCK interneurons (Engel et al., 2001; Tukker et al., 2007). Interestingly, as discussed earlier, the NEX-Cre/DG-null mouse displays a selective loss of CCK GABAergic synaptic inputs (Fruh et al., 2016). Considering these results, gamma oscillations could be studied in α-dystroglycanopathies, especially given that abnormal gamma oscillations have already been reported in other brain disorders, including autism, epilepsy and schizophrenia (Uhlhaas and Singer, 2006). The NEX-Cre/DG-null line is particularly suited to these investigations, as it avoids the confounding factor of brain malformation by selective knockout of neuronal Dag1.

Much is still unknown about the role of dystroglycan in the nervous system, particularly regarding its putative functions in neurons of the olfactory bulb, thalamus, hypothalamus and brainstem (Górecki et al., 1994; Zaccaria et al., 2001). Further use of inducible Dag1 knockout models and precise domain deletions is crucial to expand our understanding of the specific roles for αDG and βDG in both developing and mature cell types. Additional factors are possibly at play in the dystroglycan-deficient brain that are not discussed in this Review, including aberrant energy metabolism and cholinergic signaling (Rae et al., 1998, 2002; Takeuchi et al., 2009; Tuon et al., 2010; Parames et al., 2014). In future work, parsing the relationship between these pathologic elements should clarify the functional roles of dystroglycan and elucidate the order of events in its dysfunction. Continued study and development of clinically relevant animal models will be a key endeavor in understanding the basic functions of αDG in the nervous system and in designing rational therapies for α-dystroglycanopathies.

We thank Kristen Zukosky and Michelle Lee for their helpful discussion and insight.