Native DGC structure rationalizes muscular dystrophy-causing mutations

Abstract

Duchenne muscular dystrophy (DMD) is a severe X-linked recessive disorder marked by progressive muscle wasting leading to premature mortality^1,2. Discovery of the DMD gene encoding dystrophin both revealed the cause of DMD and helped identify a family of at least ten dystrophin-associated proteins at the muscle cell membrane, collectively forming the dystrophin-glycoprotein complex (DGC)^3-9. The DGC links the extracellular matrix to the cytoskeleton, but, despite its importance, its molecular architecture has remained elusive. Here we determined the native cryo-electron microscopy structure of rabbit DGC and conducted biochemical analyses to reveal its intricate molecular configuration. An unexpected β-helix comprising β-, γ- and δ-sarcoglycan forms an extracellular platform that interacts with α-dystroglycan, β-dystroglycan and α-sarcoglycan, allowing α-dystroglycan to contact the extracellular matrix. In the membrane, sarcospan anchors β-dystroglycan to the β-, γ- and δ-sarcoglycan trimer, while in the cytoplasm, β-dystroglycan's juxtamembrane fragment binds dystrophin's ZZ domain. Through these interactions, the DGC links laminin 2 to intracellular actin. Additionally, dystrophin's WW domain, along with its EF-hand 1 domain, interacts with α-dystrobrevin. A disease-causing mutation mapping to the WW domain weakens this interaction, as confirmed by deletion of the WW domain in biochemical assays. Our findings rationalize more than 110 mutations affecting single residues associated with various muscular dystrophy subtypes and contribute to ongoing therapeutic developments, including protein restoration, upregulation of compensatory genes and gene replacement.

Extended Data Fig. 1 |. Representative cryo-EM images and 2D analyses.

a,b, Negative staining (a) and drift-corrected cryo-EM (b) micrographs of DGC isolated from rabbit skeletal muscle. Representative side view and top view particles are shown in red and yellow circles, respectively. c, Representative 2D class averages of DGC isolate. Top views and side views of the DGC are labelled with red and yellow boxes, respectively.

Extended Data Fig. 2 |. Cryo-EM structural determination of DGC.

a, Data processing workflow. Binned 2 (pixel size of 2.72 Å) and binned 1 (pixel size of 1.36 Å, grey shaded) data processing is separated via dashed line. The masks used in focused 3D classification are outlined with coloured dashed lines. Data processing in RELION and cryoSPARC is denoted by a circular inscribed “R” and “C”, respectively. See Methods for more details. b, FSC as a function of spatial frequency demonstrating the resolution of the final reconstruction of DGC. c,d, View direction distribution histogram (c) and posterior precision plot (d) show view diversity of all particles used for the final map of DGC (from cryoSPARC). e, Local resolution estimation (from cryoSPARC) using a local FSC threshold of 0.5. f, FSC coefficients as a functional of spatial frequency between model and cryo-EM density maps. The generally similar appearances between the FSC curves obtained with half maps with (red) and without (blue) model refinement indicate that the refinement of the atomic coordinates did not suffer from severe over-fitting.

Extended Data Fig. 3 |. Representative cryo-EM density maps of DGC.

C-terminal cap of β-, δ-, γ-SG in panel a and dystrophin (EF-hands+WW)–dystrobrevin (EF-hands) heterodimer in panel f shown using the unsharpened 4.3 Å map. CDHL1-SEA1 of α-SG (d) displayed using a gaussian low-pass filtered map. All others are shown using sharpened densities of the 4.1-Å local-refined map or the 4.3-Å map.

Extended Data Fig. 4 |. Structure prediction of the triplex structure of β-, γ- or δ-SG using artificial intelligence (AI) programs.

Panels from left to right: cryo-EM model of the β-SG–γ-SG–δ-SG trimer (panel 1) and the predicted models of single-component trimers (panels 2–4). All predicted models were generated using AlphaFold. Regions with pLDDT > 90 are expected to be modelled to high accuracy; regions with pLDDT between 70 and 90 are expected to be modelled well; regions with pLDDT between 50 and 70 are modelled with low confidence and should be treated with caution.

Extended Data Fig. 5 |. Transfected β-, γ- and δ-SG separated by SDS-PAGE and visualized by immunoblot with Myc tag-specific antibody.

For gel source data, see Supplementary Fig. 1.

Extended Data Fig. 6 |. Identifying the potential calcium (Ca²⁺) binding site in DGC’s CDHL domains.

a, Superposition of the CDHL domains with focused views of the Ca²⁺ coordination site in endo-fucoidan hydrolase MfFcnA4 (PDB ID: 6DLH, purple), and potential Ca²⁺ binding sites in dystroglycans (salmon and grey) and α-sarcoglycan (cyan). Residues possibly involved in Ca²⁺ coordination are annotated. b, Cross-species multiple sequence alignment of the three CDHL domains in DGC. Red background depicts identical residues in all CDHL domains. Conserved residues are coloured in red and divergent residues are in black. All potential coordination residues shown in a are labelled with red circles. MfFcnA4 residues structurally homologous to CDHL2 positions are numbered; green-filled dash circles (half: not conserved; full: identical) indicate potential Ca²⁺ binding sites in CDHL2.

Extended Data Fig. 7 |. Divergent domain orientations between CDHL and SEA in the three CDHL-SEA modules of DGC.

a, Structural superposition of the three CDHL-SEA modules within DGC. SEA-CDHL modules are aligned against the SEA domain. b, Unique orientations are essential for the TM-proximal loop of β-DG (left), α-DG (top right) and α-SG (bottom right) to interact with the β-helix. The circle slash denotes interaction disruption by detachment, while the circle cross indicates interaction disruption via atomic clashes.

Extended Data Fig. 8 |. Comparison of β-DG binding sarcospan with both partner-binding and apo states of the canonical tetraspanin CD81.

Sarcospan and tetraspanin CD81 are depicted as ribbons, while β-DG and CD81’s binding partner are shown as surfaces. CD81’s unmodelled regions are represented by dashed lines.

Extended Data Fig. 9 |. Subcellular localization of dystrophin cysteine-rich region and its WW deletion construct.

The fragments used are residues 3042–3426 and 3077–3426 for dystrophin and dystrophin (ΔWW), respectively.

Fig. 1 |. Enrichment and overall architecture of the native DGC from rabbit skeletal muscle.

a,b, SEC (a) and SDS–PAGE (b) of native DGC. Peak 1 corresponds to isolated DGC. Lanes for peak 1 in SDS–PAGE correspond to sequentially fractionated volumes in gel filtration; E is the elution of affinity isolation. Protein bands are annotated with their respective reported molecular weight. A₂₈₀, absorbance at 280 nm; mAu, milli-absorbance units; MW, molecular weight. c, Representative key-shaped two-dimensional (2D) class average of the DGC with section names labelled. d, Domain arrangement of all subunits in the complex. Residue numbers at domain boundaries are indicated. Modelled N-glycan sites are shown as magenta dots. Unresolved domains or residues are indicated by dashed lines for SGs and DGs and are omitted for dystrophin and α-dystrobrevin. e, Perpendicular views of the cryo-EM density map and atomic models (cartoon representation) of the DGC. N-glycan densities are coloured in magenta with their linked asparagine residues shown as magenta spheres. Unresolved domains or residues are indicated by dashed lines.

Fig. 2 |. Triple β-helix structure of β-SG–γ-SG–δ-SG and its interaction with α-SG.

a, A boomerang-like β-helix formed by β-SG–γ-SG–δ-SG (left) and its constituent units (left and right). Left, the section names of the boomerang-like β-helix are labelled in red, the curvature site is highlighted with a dashed circle and the number of β-strands in each structural unit is indicated in brackets. Right, cysteines are visualized as orange spheres, with potential disulfide linkages illustrated by orange dashed lines. A pseudo-three-fold symmetry symbol is at the centre of each panel. b, Structural superposition of the β-, γ- and δ-SG proteins on their C-terminal (left) and N-terminal (right) halves, with deviation sites marked by black arrows. c, Structure-based sequence alignment of β-, γ- and δ-SG proteins (top) and cross-species multiple-sequence alignment of β-SG (bottom). A red background depicts identical residues; conserved residues are shown in red whereas divergent residues appear in black. Sequence gaps in γ- and δ-SG compared to β-SG are represented by black dots. Oc, Oryctolagus cuniculus; Hs, Homo sapiens; Xt, Xenopus tropicalis; Dr, Danio rerio; Bl, Branchiostoma lanceolatum. d, Pathogenic missense substitutions in β-, γ- and δ-SG likely to interfere with β-helix formation. Residues are numbered according to the rabbit proteins. Residue changes disrupting disulfide bonds are marked with a superscript ‘SS’, while those potentially affecting the hydrophobic core of the β-helix are indicated with a superscript ‘core’ label. e, α-SG structure, including its domain configuration (left) and interactions with the β-helix (middle and right). The individual CDHL and SEA domains on the left are viewed from the same perspective as in the middle panel. α-SG is shown as a ribbon, while the β-helix is shown in surface representation. Possible hydrophobic interactions and main chain hydrogen bonds are denoted by curved black and magenta dashed lines, respectively. f, Pathogenic missense to the rabbit proteins. g, A co-immunoprecipitation (IP) assay demonstrating GST, glutathione S-transferase.

Fig. 3 |. Architecture and interactions of the DG complex with SGs.

a, Overall structure of the DG complex (ribbon) on the molecular surface of the SG complex. b,c, Structural superposition of the three SEA domains (b) and CDHL domains (c) in the DGC. The three numbers in brackets indicate, from left to right, the number of aligned residues, Cα r.m.s.d. and sequence identity. The intervening loops in α-DG’s SEA1 and α-SG’s SEA domains are marked with blue stars. A calcium ion (Ca²⁺), potentially coordinated in β-DG’s CDHL2 domain, is identified through alignment with the Ca²⁺-binding site of the endo-fucoidan hydrolase MfFcnA4. Ca²⁺ in MfFcnA4 is shown as a green sphere. d, Asparagine (N661)-linked glycan of β-DG wedged between α-DG and β-DG. e, Structure-based sequence alignment of the three SEA domains, with shared secondary structures (E, β-stands; H, α-helix) indicated above the sequences and specific residue numbers below. Intervening loops in SEA1 and SEA, marked in b, are enclosed in blue boxes, the autoproteolytic cleavage sequence and cleavage site in SEA2 are marked with a black box and a scissors icon, respectively, the N-glycosylation sites in all SEA domains are marked in magenta and the disulfide bond involving the intervening loop in SG’s SEA domain is marked with a bracket. f, Multiple-sequence alignment of the three CDHL domains. Top, MfFcnA4 residues structurally homologous to CDHL2 positions are numbered; bottom, residues of CDHL2, CDHL1 and CDHL are numbered sequentially at five sites. Green-filled dashed circles (half, not conserved; full, identical) indicate potential Ca²⁺-binding sites in CDHL2. g, Divergent domain orientations in the CDHL–SEA modules of the DGC (left panel), each featuring unique hydrophobic interactions between the CDHL and SEA domains (right panels). h,i, Interaction of β-DG’s CDHL2 domain with the β-SG–γ-SG–δ-SG trimer (h) and α-SG (i), specifying involved residues. In the left panels, interacting regions of β-DG and α-SG are highlighted in red (h) and magenta (i), respectively; V^a, V^h and Vⁱ represent the structural views shown in panels a,h and i, respectively; right panels show co-immunoprecipitation assays confirming direct interactions of DG’s CDHL2–SEA2 module with both β-DG–γ-SG–δ-SG trimer and α-SG, persisting even with mutations.

Fig. 4 |. Sarcospan-mediated linkage of β-DG to SGs.

a, Orthogonal views of the tetra-spanning sarcospan in the sarcolemma, with an inset (top right) showing the β-meander structure of the LEL. Cysteine residues are visualized as yellow spheres, with potential disulfide linkages illustrated by yellow dashed lines. b,c, Insets illustrating sarcospan’s engagement with β-DG (b) and the β-SG–γ-SG–δ-SG trimer (c).

Fig. 5 |. Dystrophin and α-dystrobrevin dimerize through their cysteine-rich regions.

a, Sarcolemma-adjacent dystrophin–dystrobrevin heterodimer. Interaction sites of dystrophin with β-DG and α-dystrobrevin are indicated by magenta and blue stars, respectively. b, Domain architecture of the cysteine-rich region in dystrophin. In the ribbon structure, the four domains are labelled from N to C terminus: WW (magenta), EF1 (blue), EF2 (purple) and ZZ (green). The inset shows the interaction between the EF1 and ZZ domains of dystrophin. c, Structural superposition of the cysteine-rich regions in dystrophin and α-dystrobrevin. d, Inset illustrating the interaction between dystrophin and α-dystrobrevin. The three contact sites are labelled with numbers in blue circles: (1) interaction between α-dystrobrevin’s EF1 domain and dystrophin’s WW domain; (2) interaction between α-dystrobrevin’s EF1 domain and dystrophin’s EF2 domain; and (3) interaction between α-dystrobrevin’s EF1 domain and dystrophin’s EF1 domain. e,f, Co-immunoprecipitation assay (e) and confocal images (f) demonstrating the direct interaction between dystrophin (Dp)’s WW domain and α-dystrobrevin (Db). The fragments used were residues 1–294, 3042–3426, 3042–3426 and 3077–3426 for α-dystrobrevin, dystrophin, dystrophin-G3052D and dystrophin-ΔWW, respectively. For gel source data, see Supplementary Fig. 1. g, Superposition of the dystrophin–dystrobrevin complex onto the reported dystrophin crystal structure (dystrophin’s WW domain + EF-hand domains with the PPXY peptide of β-DG; PDB ID: 1EG4). h, Inset showing the interaction between dystrophin and β-DG. i, Pathogenic missense alterations mapped on our dystrophin–dystrobrevin structure. Residues are numbered according to the -human protein sequences.

Fig. 6 |. Model of the DGC on the sarcolemma.

Domains within DGC components that remain uncharacterized in the cryo-EM model are rendered as dashed lines or shapes. Actin filaments and laminin 2, integral to the DGC interface, are shaded in grey.