Introduction

Matriglycan is a linear polysaccharide composed of alternating xylose and glucuronate1 that is polymerized by LARGE12 and LARGE23 glycosyltransferases post-translationally on the heavily and heterogeneously O-glycosylated4,5,6 mucin-like domain of dystroglycan. Dystroglycan is a component of the dystrophin-glycoprotein complex[7](https://www.nature.com/articles/s41467-025-64080-z#ref-CR7 “Liu, S., Su, T., Xia, X. & Zhou, Z. H. Native DGC structure rationalizes muscular dystrophy-causing mutations. Nature https://doi.org/10.1038/s41586-024-08324-w

(2024).“),[8](https://www.nature.com/articles/s41467-025-64080-z#ref-CR8 “Wan, L. et al. Structure and assembly of the dystrophin glycoprotein complex. Nature https://doi.org/10.1038/s41586-024-08310-2

(2024)“) that connects the cytoskeleton to the extracellular matrix (ECM)9,10 by virtue of matriglycan, which binds ECM proteins that contain laminin-globular (LG) domains11, such as laminin12, agrin13,14, perlecan15,16, neurexin17,18, pikachurin19,20, EYS21 and SLIT222 as well as the Lassa Fever spike glycoprotein23,24,25. Matriglycan on dystroglycan is indispensable for neuromuscular function, brain development26,27,28, and efficient arenaviral entry29. Conversely, matriglycan of insufficient length causes muscular dystrophies that are sometimes accompanied by intellectual disability30,31. Currently, there are no treatments for matriglycan-insufficient muscular dystrophies or Lassa Fever viral infections32,33,34,35.

Gene therapies may offer the most effective solutions for genetic disorders36. However, LARGE1 gene transduction assays produce matriglycan of indiscrete lengths—a continuous smear that does not resemble physiologically observed sizes, inferred by discrete bands with a clear migration front and rear, on Western blots37. In contrast to many natural biopolymers—polynucleotides and polypeptides—whose length is typically governed by a template, LARGE glycosyltransferases polymerize matriglycan of discrete lengths without a corresponding polymeric template37. Bifunctional glycosyltransferases with tandem catalytic domains that form dimers, such as exostosins (EXT1-EXT2 and EXTL32)38,39,40,41 and chondroitin polymerase (K4CP)42, appear to synthesize their respective linear glycans distributively (enzyme dissociates from substrate between catalytic cycles) on synthetic substrates. In this study, we sought to determine the mechanism of LARGE1 matriglycan polymerization and length control. Using single particle cryo-EM reconstructions, we and others43 show that LARGE1 has a similar architecture to these bifunctional glycosyltransferases, in which the active sites on each protomer face opposite directions. However, the mechanism of matriglycan synthesis and length control is not obvious from cryo-EM reconstructions of LARGE1.

In this work, we investigated the mechanism by which LARGE1 polymerizes matriglycan on dystroglycan in vitro using soluble recombinant purified LARGE1dTM2 and an engineered substrate, prodystroglycan, in terms of processivity versus distributivity44 and the biochemical factors that regulate polymer synthesis and length. We first confirm the structural integrity of our LARGE1 construct by light scattering techniques and single particle cryo-EM reconstructions. We subsequently show that LARGE1dTM only synthesizes discrete sizes of matriglycan on prodystroglycan if (1) the mucin-like domain is contiguous with the dystroglycan N-terminal domain (DGN)45, (2) core M3 is phosphorylated on carbon-6 [N-acetylgalactosamine–β3-N-acetylglucosamine–β4-(phosphate-6-)mannose]30 and (3) a xylose-glucuronate primer is present46,47. Additionally, using a mixture of LARGE1 active site mutants, we show that LARGE1 polymerizes matriglycan processively on prodystroglycan but distributively on a synthetic substrate. And, using a neuromuscular patient mutation (T192M)31, we show that matriglycan length is controlled by the dystroglycan prodomain, DGN. Collectively, our study specifies the mechanism and factors that control matriglycan polymerization by LARGE1 on dystroglycan, which will underpin therapeutic strategies to treat neuromuscular disorders and arenaviral infections by modulating matriglycan length.

Results

In vitro LARGE1dTM matriglycan synthesis assay on recombinant prodystroglycan

We have historically relied on synthetic small-molecule substrates, such as 4-methylumbelliferone (4MU) or biotin-conjugated sugars, to assay LARGE1 glycosyltransferase activity because the products are conveniently separated from the starting material by chromatographic techniques and easily detected. However, such substrates are unlikely to accurately recapitulate the characteristics of glycosyltransferases on their cognate endogenous targets. For example, the range of matriglycan lengths polymerized on 4MU-glucuronate-xylose by LARGE1dTM follows a Poisson distribution2 (Fig. 1a and Supplementary Fig. 1) and does not resemble the well-defined polymer lengths observed in vivo26, which are suggested by the discretely migrating band of native dystroglycan modified with matriglycan on Western blots. Because there is a paucity of appropriate substrates that reflect the configuration of dystroglycan found in the Golgi, where it would naturally encounter LARGE1 (Fig. 1b), we decided to engineer a native-like, recombinant, soluble construct, prodystroglycan. To mimic the immature, but relevant configuration of dystroglycan found in the late-ER/early Golgi stages, we mutated the furin cleavage site (R311A/R312A)48,[49](https://www.nature.com/articles/s41467-025-64080-z#ref-CR49 “Jin, M. et al. Dynamic allostery drives autocrine and paracrine TGF-β signaling. Cell https://doi.org/10.1016/j.cell.2024.08.036

(2024).“), so that dystroglycan retains the prodomain, DGN. The construct terminates prior to the transmembrane region. It preserves the autoproteolytic activity of the sea-urchin sperm protein, enterokinase and agrin (SEA) domain (Fig. 1c), which cleaves dystroglycan into alpha and beta chains[50](https://www.nature.com/articles/s41467-025-64080-z#ref-CR50 “Anderson, M. J. M. et al. Molecular basis of proteolytic cleavage regulation by the extracellular matrix receptor dystroglycan. Structure https://doi.org/10.1016/j.str.2024.08.019

(2024).“). Because alpha and beta dystroglycan remain tightly associated after cleavage, we used the engineered C-terminal hexahistidine-tag for nickel affinity purification (Fig. 1d). The recombinant protein is post-translationally modified by appropriate enzymes, including LARGE1, that are endogenous to HEK 293 Freestyle cells as it traverses the secretory pathway (Fig. 1b–d). We used batch anion exchange chromatography to recover a fraction of prodystroglycan devoid of matriglycan, synthesized by the host, for subsequent enzymatic assays (Fig. 1d). We validate this construct by showing that purified recombinant LARGE1dTM can polymerize matriglycan of defined length on prodystroglycan in vitro when provided with both UDP-xylose and UDP-glucuronic acid (Fig. 1e).

Fig. 1: Engineered recombinant prodystroglycan is a native-like substrate for LARGE1 matriglycan synthesis in vitro.

a Anion exchange chromatogram of matriglycan polymerized by LARGE1dTM on 4-methylumbelliferyl-glucuronate-xylose in vitro. Product length (enumerated peaks represent the number of monosaccharides) is inversely proportional to the pH of the reaction and product abundance follows a Poisson distribution (Supplementary Fig. 1), which is characteristic of distributive polymerization. Similar experimental results were replicated at least three times independently. Source data are provided as a Source Data file. b Schematic of the relevant core M3 glycan modifications on dystroglycan acquired in the secretory pathway. Panels from left to right: the signal peptide (residues 1-27), which is cleaved in the ER, is omitted. Dystroglycan globular N-terminal domain (DGN; orange) and the autoproteolytic extracellular C-terminal domain (eCTD; purple) flank the heavily and heterogeneously O-glycosylated (***) mucin-like domain (black line). Threonine residues 317, 319, and 379 in human dystroglycan are post-translationally modified sequentially in the ER by POMT1/2, POMGNT2, B3GALNT2, and POMK to make phosphorylated core M3 (N-acetylgalactosamine-β3-N-acetylglucosamine-β4-(phosphate-6-) mannose). In the Golgi, FKTN (FCMD) and FKRP extend core M3 by two phosphoribitol units followed by RXYLT1 (TMEM5) and B4GAT1 which add a xylose-glucuronate primer. DGN is cleaved in mature dystroglycan present on the cell membrane. O-glycans and a secondary matriglycan site on the mucin-like domain are omitted for clarity c A schematic of recombinant human prodystroglycan (DAG128-749), which is designed to retain DGN due to a mutation in the furin cleavage site (R311A/R312A; arrow) and is receptive to modification by LARGE1. d Purification of prodystroglycan using Ni-NTA resin followed by batch anion exchange chromatography, in which recombinant prodystroglycan modified by matriglycan elutes at higher ionic strength, was monitored by SDS-PAGE with Coomassie stain and Western blots for matriglycan (IIH6), dystroglycan N-terminal domain (anti-DGN; 1D9) and laminin overlay. Similar experimental results were replicated at least five times independently. e Western blots of matriglycan (IIH6 or laminin overlay) polymerized on prodystroglycan by LARGE1dTM in vitro with the addition of substrates, UDP-xylose (UDP-Xyl) and UDP-glucuronate (UDP-GlcA). Core proteins, α- and β-dystroglycan are indicated by a bracket and arrowhead, respectively. Similar experimental results were replicated at least three times independently.

Structural characterization of LARGE1dTM2 used in glycosyltransferase assays

LARGE proteins consist of independent xylose and glucuronic acid transferase domains arranged in tandem (Fig. 2a). We confirmed the structural integrity of the LARGE1dTM construct2 used in enzymatic assays through light scattering experiments and single-particle cryo-EM reconstructions (Fig. 2). The tandem Rossmann-like fold of the catalytic domains of LARGE1 (Supplementary Fig. 2a) was built into the reconstructed volumes (Fig. 2b). We found that LARGE1dTM forms a dimer in which the active sites on each protomer face opposite directions (Fig. 2c)43. Our reconstructed volume of apo-LARGE1dTM does not differ globally from that of LARGE1dTM-DAG128-340 enzyme-substrate (ES) complex, except for a density for UDP-glucuronic acid (Fig. 2d and Supplementary Fig. 4) which only binds the latter (Fig. 2e). Despite collecting data at zero- and thirty-degree tilt for the ES complex, one of the xylose transferase domains in both apo-LARGE1 and the ES complex remained poorly resolved without imposed symmetry (Fig. 2f). Interestingly, the residues in LARGE1 which when mutated have the most effect on the matriglycan synthesis map to the region of the xylose transferase domain that remains unresolved[51](https://www.nature.com/articles/s41467-025-64080-z#ref-CR51 “Ma, K. et al. Saturation mutagenesis-reinforced functional assays for disease-related genes. Cell https://doi.org/10.1016/j.cell.2024.08.047

(2024).“). Although we reconstructed the volumes in both C1 and C2 symmetries, we suspect that the asymmetry of the LARGE1 homodimer is biologically relevant52,[53](https://www.nature.com/articles/s41467-025-64080-z#ref-CR53 “Basanta, B. et al. The conformational landscape of human transthyretin revealed by cryo-EM. Nat. Struct. Mol. Biol. https://doi.org/10.1101/2024.01.23.576879

(2025).“) and may hint at the stoichiometry of the LARGE1-prodystroglycan ES complex but we never observed nor could isolate the structurally symmetric dimer for comparison to determine its function. The density for prodystroglycan (DAG128-340), however, remains unresolved in the reconstruction, likely due to its dynamic nature[49](https://www.nature.com/articles/s41467-025-64080-z#ref-CR49 “Jin, M. et al. Dynamic allostery drives autocrine and paracrine TGF-β signaling. Cell https://doi.org/10.1016/j.cell.2024.08.036

(2024).“), but its presence is implied by density for UDP-glucuronate in both glucuronic acid transferase active sites (Fig. 2e). Additionally, the N-glycans (Supplementary Fig. 3), coiled-coil domain, and N-terminal tail in both reconstructions also remain unresolved, likely because they were averaged to the level of noise during particle alignment due to their dynamics relative to the globular catalytic head domains.

Fig. 2: Reconstructed volume of LARGE1dTM glycosyltransferase.

a LARGE protein constructs, LARGE1dTM and LARGE2dTM (right panel), used for structural and biochemical assays carry an N-terminal FLAGx3 tag, C-terminal hexahistidine tag, and lack the transmembrane domain. b Two molecules of LARGE1 (grey cartoon) were refined into the 3.7-Å map of LARGE1dTM; protomers are displayed as cyan and purple transparent surfaces. c Orthogonal active sites each coordinating a manganese ion (magenta spheres) on alternate protomers face the same direction. A groove (highlighted surface) connects the xylose (cyan shading) and glucuronate (pink shading) transferase domains. The quaternary structure implies that matriglycan is polymerized using orthogonal active sites on alternate protomers of LARGE1 (yellow double-headed arrow). d Magnified view of a glucuronic acid transferase active site showing residues (sticks, labelled in blue text) that coordinate manganese (violet sphere) and UDP-glucuronate (sticks labelled “Ura” and “GlcA”) in the LARGE1dTM-prodystroglycan (DAG128-340) reconstruction. Coordination distances (magenta dotted lines, Å) are in black text. Other residues that compose the active site are shown as lines with grey labels. e Titration of UDP-GlcA into NHS-red labeled LARGE1dTM alone (open circles) or saturated with prodystroglycan (DAG128-749, closed squares) observed by microscale thermophoresis. Data points and error bars represent the averages and standard error for n = 3 replicates, respectively. Similar experimental results were obtained independently four times. Source data are provided as a Source Data file. f Partial absence of density in the xylose transferase domain (red circle) when reconstructed with no symmetry imposed (purple, C1) aligned with a volume refined with axial symmetry (transparent cyan, C2).

We confirmed that LARGE1dTM dimerized in solution independently of its coiled-coil domain (Supplementary Fig. 2b) and N-glycans, by comparing it to LARGE2dTM, which lacks the coiled-coil domain (Supplementary Fig. 4), and digestion with PNGase F, respectively, using on-line multi-angle light scattering (SEC-MALS; Supplementary Fig. 2b). LARGE2dTM was not analyzed further in this study because deletion of LARGE1, but not LARGE2, causes neuromuscular pathology. We used small-angle X-ray scattering (SAXS) to generate a low-resolution molecular envelope of apo-LARGE1dTM (Supplementary Figs. 2c, 6-11 and Supplementary Tables 1, 3, 4) to show a stem domain attached to a globular head (Supplementary Fig. 2c), which we conclude represents the coiled-coil domain attached to the catalytic domains and is consistent with the expected topology of LARGE1 in the Golgi membrane (Fig. 2c). Conserved surface residues are primarily localized to the dimer interface (Supplementary Fig. 2d). The xylose and glucuronate transferase active sites coordinate a cation in apo-LARGE1dTM (Supplementary Fig. 2e). LARGE1dTM coordinates and maps were deposited with the following PDB and EMDB accession codes: 7UI6, EMD-26540 (C1 symmetry) and 7UI7, EMD-26541 (C2 symmetry). LARGE1dTM-DAG128-340 coordinates and maps were deposited with PDB ID 9E1T and EMD-47420.

We expect highly active enzymes from the structural integrity observed in both reconstructions of recombinant LARGE1dTM. However, despite elucidating the relative positions of the active sites in the reconstructed volumes, which imply that matriglycan is likely polymerized by switching between the orthogonal active sites on different protomers (Fig. 2c, arrow), we were unable to deduce whether matriglycan is synthesized processively or distributively on dystroglycan and how polymer length is controlled from the LARGE1 cryo-EM reconstructions.

The mucin-like domain of dystroglycan must be contiguous with DGN and additionally carry a xylose-glucuronate (Xyl-β1,4-GlcA) primer for matriglycan synthesis by LARGE1

To determine the biochemical factors necessary for matriglycan synthesis by LARGE1 on dystroglycan we assayed in vitro matriglycan synthesis activity of recombinant LARGE1dTM on constructs of dystroglycan, (Fig. 3). To test whether LARGE1dTM can polymerize matriglycan on mature dystroglycan, in which DGN is absent, or whether DGN must remain contiguous with the mucin-like domain of dystroglycan, we expressed and purified recombinant constructs of wild-type (rDystroglycan, ΔDGN313-749) or dystroglycan with a furin site mutation (rProdystroglycan, +DGN28-749) for in vitro matriglycan synthesis assays (Fig. 3a). LARGE1dTM can polymerize matriglycan on prodystroglycan but not on mature dystroglycan (Fig. 3b), suggesting that the mucin-like domain of dystroglycan must be contiguous with DGN (Fig. 3a, b) for matriglycan synthesis by LARGE1.

Fig. 3: Biochemical factors that promote matriglycan polymerization by LARGE1 on dystroglycan.

a Schematics of mature recombinant dystroglycan (rDystroglycan) in which DGN is cleaved (ΔDGN; DAG1313-749) and recombinant prodystroglycan (rProdystroglycan) in which DGN remains covalently attached to the mucin-like domain through a peptide bond (+DGN; DAG128-749). b Western blot of matriglycan polymerized on mature dystroglycan (ΔDGN; DAG1313-749) and prodystroglycan (+DGN; DAG128-749). c Schematic showing digestion of prodystroglycan with β-glucuronidase. Similar experimental results were obtained independently at least three times. d Western blot of matriglycan synthesized in vitro by LARGE1dTM on prodystroglycan digested with β-glucuronidase (+) or control (-). Recombinant α- and β-dystroglycan fragments are marked. Similar experimental results were generated independently at least three times. e Western blot of matriglycan synthesized by LARGE1dTM in vitro on prodystroglycan. Wild-type or POMK KO HEK 293T cells were transfected with an adenovirus to express prodystroglycan (Ad-DAG128-749). Similar experimental results were obtained independently at least seven times.

Our previous work showed that LARGE1dTM was able to transfer glucuronic acid to xylose-α-pNP2. To test whether LARGE1 requires a xylose-glucuronate primer or whether a terminal xylose, naturally appended by RXylT1 (TMEM5), is sufficient for matriglycan synthesis, we digested prodystroglycan with β-glucuronidase54 (Fig. 3c). LARGE1dTM could not transfer glucuronic acid to a lone, terminal xylose on prodystroglycan—a function that is usually performed by B4GAT1—and could not polymerize matriglycan in the absence of a xylose-glucuronate primer on prodystroglycan (Fig. 3d) consistent with previous findings46,47.

Phosphorylated core M3 is required for LARGE1 to efficiently synthesize matriglycan on dystroglycan

Phosphorylation of mannose at carbon-6 on core M3 by Protein O-mannose Kinase (POMK) is critical for efficient matriglycan elongation by LARGE130,45 and subsequent laminin binding55,56. Mutations in POMK that decrease mannose phosphorylation in dystroglycanopathy patients produce short matriglycan30,45. We previously showed that LARGE1dTM binds phosphorylated core M3 (KD = 11.5 µM) better than its non-phosphorylated counterpart (KD > 90 µM)30. To test the effect of core M3 phosphorylation on LARGE1 matriglycan polymerization in vitro, we transduced wild-type and POMK KO HEK 293T cells with adenovirus encoding prodystroglycan (DAG128-749) and purified the secreted protein from the media for use in in vitro matriglycan synthesis assays. LARGE1dTM was unable to synthesize matriglycan efficiently on prodystroglycan purified from POMK KO cells, which lacks the phosphate modification on mannose-C6 of core M3, within the same timeframe as proprotein from wild-type cells (Fig. 3e). However, given longer reaction times (16 h), LARGE1dTM could polymerize matriglycan on prodystroglycan from POMK KO cells (Supplementary Fig. 12). The slow rate of matriglycan synthesis on prodystroglycan lacking mannose-6-phosphate suggests that phosphorylated core M3 accelerates matriglycan synthesis.

Elucidating the mechanism of matriglycan synthesis using active site mutants of LARGE1

Processive polymerization refers to the consecutive additions to a chain without dissociation of the substrate and the enzyme between catalytic cycles—these phenomena are typically observed in polynucleotide (DNA/RNA polymerases), polypeptide (ribosome) or polysaccharide syntheses. LARGE1 is a bifunctional glycosyltransferase with separate xylose and glucuronate transferase domains2. To determine whether LARGE1 polymerizes matriglycan distributively or processively on dystroglycan, we exploited the independent glycosyltransferase activities of LARGE1 in both cell-based and in vitro assays (Fig. 4). Mutating the active sites in isolation results in constructs that either lack xylose (DXD1; D242N/D244N) or glucuronate (DXD3; D563N/D565N) transferase activity but retains the activity of the auxiliary domain (Fig. 4a).

Fig. 4: Mixtures of LARGE1(dTM) active site mutants can polymerize matriglycan.

a Schematic of wild-type and active site mutants of LARGE1 that either lack xylose transferase activity (DXD1) or glucuronate transferase activity (DXD3). b Western blot of LARGE1 KO and wild-type Hap1 cell lines transduced with adenovirus encoding LARGE1 (20 MOI): wild-type (WT), xylose transfer mutant (DXD1; D242N/D244N), glucuronate transfer mutant (DXD3A/B; D563N or D565N, respectively) or a combination of both (DXD1 + DXD3A/B). Matriglycan was detected using an anti-matriglycan antibody (IIH6) or laminin overlay; core dystroglycan protein was detected using the AF6868 antibody. Pink arrow indicates a faint matriglycan-positive band in LARGE1 KO Hap1 cells transduced with DXD3B only. Green arrow indicates a perturbation in the migration of dystroglycan in wild-type Hap1 cells transduced with DXD1. Similar experimental results were obtained independently at least three times. c Anion exchange chromatogram of matriglycan polymerized on 4-methylumbelliferone-glucuronate-xylose by LARGE1dTM or active site mutants DXD1, DXD3, or DXD1 + DXD3. Similar experimental results were obtained independently at least three times. Source data are provided as a Source Data file. d Western blot of matriglycan polymerized on prodystroglycan constructs (as indicated) by a combination of LARGE1dTM active site mutants in the presence of UDP-sugars over time (left panel). Western blot of matriglycan polymerized on prodystroglycan (DAG128-398) after 24 h by wild-type LARGE1dTM, DXD1:DXD3 heterodimers, and a mixture of DXD12 and DXD32 mutants. Similar experimental results were obtained independently at least seven times.

We used a commercially available Hap1 LARGE1 KO cell line and its wild-type counterpart to evaluate the matriglycan synthesis activity of a combination of LARGE1 active site mutants. We found that adenoviral transduction of Hap1 cells with wild-type LARGE1 resulted in matriglycan polymers of dispersed lengths on endogenous dystroglycan and did not recapitulate a discretely migrating band observed in the Hap1 parental control cell line, C631 (Fig. 4b). This suggests that overexpressing LARGE1 overrides the control of matriglycan length observed in wild-type cells consistent with LARGE1 gene therapy studies36,37. We nonetheless evaluated the effects of combining LARGE1 mutants that lacked either xylose transferase activity (DXD1) or glucuronate transferase activity (DXD3A/B) in LARGE1 KO Hap1 cells. We show that co-transduction with a mix of LARGE1 mutants could restore matriglycan synthesis on endogenously expressed dystroglycan (Fig. 4b). The reason we used single mutants DXD3A (D563N) and DXD3B (D565N) for adenoviral expression is that two independent vendors found the production of LARGE1 encoding D563N/D565N resulted in empty virions. Co-transduction of LARGE1 mutants produced combinations of DXD12 and DXD3[A/B]2 homodimers as well as DXD1:DXD3[A/B] heterodimers, which were confounding to interpret. However, we include these data because of two interesting phenomena: (1) Transduction of LARGE1 KO Hap1 cells with the glucuronate transferase inactive mutant, DXD3B (D565N), results in a faint matriglycan-positive band, which may represent the addition of a single xylose to the xylose-glucuronate primer (Fig. 4b, pink arrow). This band is less pronounced when cells are transduced with DXD3A. And (2) Transduction of wild-type Hap1 cells with adenovirus encoding the xylosyltransferase-inactivated mutant (DXD1), but not DXD3 mutants, did not produce a high molecular weight smear and maintained a discretely migrating matriglycan-positive band of reduced size (Fig. 4b, green arrow). Although we do not have a mechanistic explanation for why the singly transduced DXD1 variant decreases matriglycan length, it suggests that the glucuronate transferase domain partly controls the length of matriglycan.

Because we could not conclude whether LARGE1 polymerized matriglycan distributively or processively from adenoviral co-transduction experiments in cells, we compared matriglycan synthesized in vitro by a mixture of mutants that lacked either xylose or glucuronate transferase activity to wild-type LARGE1dTM using 4-methylumbelliferone-glucuronate-xylose (MU-GX) and prodystroglycan as substrates. Unlike heterodimers of mutant LARGE1 that might be formed in co-transduction assays, we noted that recombinant homodimers (DXD12 and DXD32) were unlikely to exchange to form DXD1:DXD3 heterodimers in solution because LARGE1dTM dimers are SDS-resistant in the absence of reducing agent (Supplementary Fig. 13). Additionally, using co-expression and subsequent tandem affinity purification of FLAG-tagged DXD1 and his-tagged DXD3, we show that heterodimers of DXD1:DXD3 and pseudomonomeric LARGE1dTM heterodimers, in which the active sites on one protomer is mutated, cannot polymerize matriglycan (Supplementary Fig. 14 and Fig. 4d).

Both wild-type LARGE1dTM and a mixture of active site mutants could polymerize matriglycan on 4-methylumbelliferone-glucuronate-xylose (Fig. 4c) and prodystroglycan (Fig. 4d). We found that the wild-type enzyme produced a discretely migrating band of matriglycan-modified prodystroglycan, as observed in vivo, in contrast to relaxed length control over short periods of time and high molecular weight smear given ample time observed for a mixture of LARGE1dTM active site mutants (DXD12 and DXD32; Fig. 4d), which are forced to polymerize matriglycan distributively. This suggests that LARGE1 polymerizes matriglycan processively on its native-like substrate, prodystroglycan, but acts distributively on synthetic substrates. Moreover, in the context of the LARGE1-prodystroglycan ES complex, length control is linked to processive matriglycan polymerization.

An enzyme-substrate complex between LARGE1 and DGN controls matriglycan length

To investigate how LARGE1 controls matriglycan length synthesized on prodystroglycan, we leveraged a patient mutation in DGNT192M which produces dystroglycan modified with short matriglycan compared to wild-type, that also migrates as a discrete band on Western blot31. The peptide backbone of the crystal structures for mouse DGNT190M aligns well with wild-type protein (Fig. 5a). The threonine to methionine mutation obstructs a cleft on DGN that would otherwise be of a greater length (Fig. 5b). Human recombinant prodystroglycanT192M carries short matriglycan (Fig. 5c, d), suggesting that matriglycan length control is intrinsic to DGN. Western blot analysis shows that although there are discrete matriglycan-positive bands from the analogous T190M mutant mouse tissues, the mutation results in matriglycan that migrates at correspondingly lower molecular weight compared to wild-type dystroglycan (Fig. 5e). Taken together, these results imply that DGN controls matriglycan length in the context of the LARGE1-prodystroglycan ES complex.

Fig. 5: LARGE1dTM recapitulates pathologically short matriglycan on prodystroglycanT192M.

a Protein backbone of wild-type (magenta; PDB ID:1U2C) and T190M (teal; PDB ID: 4WIQ) mouse DGN are overlaid. b The volume of clefts in both crystal structures was calculated using Caver Analyst86. c Schematic of prodystroglycan representing human wild-type and T192M mutant constructs. d Western blots of wild-type and T192M recombinant prodystroglycan purified from HEK 293 Freestyle cells. Matriglycan was detected using an anti-matriglycan antibody (IIH6); core dystroglycan protein was detected using the AF6868 antibody. Results are representative of at least three independent experiments. e Laminin overlay of dystroglycan in the indicated tissues from wild-type and T190M mice. Results are representative of at least three independent experiments. f AlphaFold 3 model of DGN (white surface) bound to LARGE1 dimer (backbones of protomers are shown in orange and blue cartoon) with matriglycan (pink sticks) modeled into a cleft (dotted lines) that acts as a ruler to measure length. Threonine-192 is shown in red surface (black arrow). g Confidence of the AlphaFold 3 prediction is color-coded onto the cartoon backbone of the complex. Glycans have been removed from predicted models for clarity and are shown in Supplementary Fig. 16.

We were unable to polymerize matriglycan on prodystroglycanT192M in vitro under the same conditions used for the wild-type proprotein (Supplementary Fig. 15a-c), suggesting that other factors may aid matriglycan polymerization in vivo. To determine whether prodystroglycanT192M can form a competent ES complex with LARGE1, like wild-type proprotein, we showed that both wild-type and T192M prodystroglycan can bind LARGE1dTM with similar affinity (Supplementary Fig. 15d). We further showed that, whereas LARGE1dTM in isolation cannot bind UDP-glucuronate, the LARGE-prodystroglycan ES complex formed using either wild-type or a T192M variant bound UDP-glucuronic acid (Supplementary Fig. 15e). Because prodystroglycanT192M retains its backbone structure[57](https://www.nature.com/articles/s41467-025-64080-z#ref-CR57 “Bozic, D., Sciandra, F., Lamba, D. & Brancaccio, A. The structure of the N-terminal region of murine skeletal muscle α-dystroglycan discloses a modular architecture. J. Biol. Chem. 279, 44812–44816