Endogenous Glucuronyltransferase Activity of LARGE or LARGE2 Required for Functional Modification of α-Dystroglycan in Cells and Tissues*

Background: Glycosyltransferase activities of LARGE are required for α-dystroglycan function. Results: This study reveals that LARGE/LARGE2 glucuronyltransferase activity correlates with α-dystroglycan laminin binding activity in Largemyd and Large2−/− mice. Conclusion: Both LARGE and LARGE2 contribute to the functional modification of α-dystroglycan in vivo. Significance: Our results provide the first strong evidence that LARGE2 is involved in α-dystroglycan modification in vivo. Mutations in the LARGE gene have been identified in congenital muscular dystrophy (CMD) patients with brain abnormalities. Both LARGE and its paralog, LARGE2 (also referred to as GYLTL1B) are bifunctional glycosyltransferases with xylosyltransferase (Xyl-T) and glucuronyltransferase (GlcA-T) activities, and are capable of forming polymers consisting of [-3Xyl-α1,3GlcAβ1-] repeats. LARGE-dependent modification of α-dystroglycan (α-DG) with these polysaccharides is essential for the ability of α-DG to act as a receptor for ligands in the extracellular matrix. Here we report on the endogenous enzymatic activities of LARGE and LARGE2 in mice and humans, using a newly developed assay for GlcA-T activity. We show that normal mouse and human cultured cells have endogenous LARGE GlcA-T, and that this activity is absent in cells from the Largemyd (Large-deficient) mouse model of muscular dystrophy, as well as in cells from CMD patients with mutations in the LARGE gene. We also demonstrate that GlcA-T activity is significant in the brain, heart, and skeletal muscle of wild-type and Large2−/− mice, but negligible in the corresponding tissues of the Largemyd mice. Notably, GlcA-T activity is substantial, though reduced, in the kidneys of both the Largemyd and Large2−/− mice, consistent with the observation of α-DG/laminin binding in these contexts. This study is the first to test LARGE activity in samples as small as cryosections and, moreover, provides the first direct evidence that not only LARGE, but also LARGE2, is vital to effective functional modification of α-DG in vivo.

muscular dystrophies (CMDs) that are characterized by perturbed glycosylation and reduced ligand binding of ␣-DG, and are often associated with brain abnormalities (15). The Large myd mouse, a spontaneous model of muscular dystrophy, has a mutation in the Large gene and results in defective ␣-DG glycosylation (16).
Functional modification of, and ligand binding to, ␣-DG require the phosphorylation of O-linked mannose. LARGE, fukutin, and FKRP have all been implicated in the modification pathway that assembles the ligand-binding moiety on this phosphorylated O-mannosyl glycan (17). ␤3GnT1, which is involved in the synthesis of poly-N-acetyllactosamine (18) and whose sequence is closely related to the second domain of LARGE (19), has been shown to contribute to the synthesis of the laminin-binding glycan of ␣-DG through formation of a complex with LARGE (20). Moreover, a study using a forward genetic screen demonstrated that ␤3GnT1 is required for functional glycosylation of ␣-DG in vivo; mice expressing a mutant form of this protein exhibited axonal-guidance defects like those observed in ISPD or an epiblast-specific DG mutant mouse (21). Recently, SGK196 was found to be a protein O-mannose kinase (designated as POMK) that generates [GalNAc-␤3-GlcNAc-␤4-(phosphate-6-)Man], a modification required for the functional glycosylation of ␣-DG (22). POMK phosphorylates the O-linked mannose only when the latter is extended with GalNAc-␤3-GlcNAc-␤, a unique disaccharide structure produced by sequential actions of GTDC2 and B3GALNT2. However, the precise molecular functions of the glycosyltransferase ␤3GnT1, and the other proteins fukutin, FKRP, and ISPD, as well as that of TMEM5, in ␣-DG glycosylation are not known.
Both LARGE and its paralog LARGE2 are bifunctional glycosyltransferases that have xylosyltransferase (Xyl-T) and glucuronyltransferase (GlcA-T) activities and the ability to polymerize Xyl-GlcA repeats (23)(24)(25). The overexpression of either of the LARGE paralogs in cultured cells enhances the functional modification of ␣-DG (26 -30). Moreover, the overexpression of LARGE circumvents defects in the modification of ␣-DG in patient cells from several distinct types of CMD (30). A recent study revealed that the ability of LARGE to catalyze the ␣-DG modification is dependent on the availability of O-mannosyl phosphate acceptor sites (9). The LARGE paralogs have different tissue expression profiles; LARGE is widely expressed, with levels highest in brain, heart, and skeletal muscle, whereas LARGE2 is highly expressed in kidney and placenta, but undetectable in brain (28,29). The ability of LARGE2 to generate polymeric Xyl-GlcA repeats and to enhance the functional modification of ␣-DG in cultured cells suggested that, like LARGE, this protein is involved in the physiological function of ␣-DG (24,25). However, there is no direct in vivo evidence for this, and LARGE2 has not been reported to be involved in dystroglycanopathies or any other diseases.
We have developed an enzyme assay that enables us to measure endogenous LARGE enzymatic activity in cultured human lymphoblasts and myoblasts, mouse fibroblasts and tissues, and even from cryosections of both human and mouse skeletal muscle. Our results demonstrate that LARGE enzymatic activity is directly involved in the functional modification of ␣-DG, and provide strong evidence that LARGE2 is indeed also involved in this modification in vivo.

EXPERIMENTAL PROCEDURES
Materials-LARGEdTM and LARGE2dTM were described previously (23,24). Mouse skin fibroblasts from wild-type (WT) C57/BL6 and Large myd mice, and human Epstein-Barr virus-immortalized lymphoblasts from the CMD patient (Patient #1) were described previously (17,31). Human control myoblasts were purchased from Cook MyoSite (Pittsburgh, Lot P201014-50M). Primary myoblasts from Patient #2 were cultured using standard method. Patient #2: 7-year-old girl, pregnancy, delivery, and neonatal period were normal. Developmental delays were suspected at 1 year. At 7 years she had minimal weakness, and was verbal with cognitive delays. MRI revealed thickened and pachygyric appearing frontal cortex and elevated creatine kinase at 2290 units/liter. Muscle biopsy revealed mild dystrophic changes and reduced staining for the glycan epitope of ␣-DG (IIH6). Control human quadriceps muscle biopsies were taken during diagnostic investigations of adult patients and were studied with patient consent. The biopsies had no histopathologic abnormalities.
Generation of Large2 (Gyltl1b) Knock-out Mice-An embryonic stem (ES) cell line (JM8A3.N1) with a targeted mutation in Large2 (Gyltl1b tm1(KOMP)Wtsi ) was obtained from the KOMP Repository at the University of California Davis. In the targeted ES cells (CloneID EPD0387_6_B09), Large2 exons 4 -11 are replaced with a LacZ/neo-containing targeting cassette. Large2 knock-out mice were generated by blastocyst injection of targeted ES cells and chimeric mice were bred with C57BL/6 to select for germline transmission. Large2 knock-out mice were viable and obtained at the expected 1:3 Mendelian ratio when heterozygous mice were bred. No breeding or fertility problems were observed in the homozygous null mice. The genotype was confirmed by PCR. A 390 bp wild-type fragment was generated using primer LG2.wtF1 (5Ј-AGCTTCTACTGCCACCCA-AGT-3Ј) in combination with LG2.wtR2 (5Ј-GGGTAATAAT-CTCTCGGCTGG-3Ј). Primer pair LG2.wtF1 and LAR3 (5Ј-CACAACGGGTTCTTCTGTTAGTCC-3Ј) amplified a 497bp mutant PCR fragment. The Large2-null mice appeared healthy and were indistinguishable from their heterozygous or control littermates with respect to body weight and life span when aged up to one year (data not shown). Also no increased spontaneous tumor load was observed in the null mice.
Enzymatic Assay-The HPLC-based enzymatic assay for LARGEdTM and LARGE2dTM was performed using GlcA-␤-MU as the acceptor for Xyl-T activity, and Xyl-␣1,3-GlcA-␤-MU for GlcA-T activity as described previously (23,24). For the assessment of metal dependence, a buffer at an optimal pH for each activity was used: LARGEdTM Xyl-T with 0.1 M sodium acetate, pH 5.0, LARGEdTM GlcA-T with 0.1 M MOPS, pH 6.5, LARGE2dTM Xyl-T with 0.1 M MOPS, pH 7.5, and LARGE2dTM GlcA-T with 0.1 M sodium acetate, pH 5.5. For the assessment of endogenous LARGE GlcA-T activity in cells or tissues, solubilized microsomes were incubated in a volume of 25 l for 3 h at 37°C, with 0.4 mM Xyl-␣1,3-GlcA-␤-MU and 10 mM UDP-GlcA in 0.1 M MES buffer, pH 6.0, at 5 mM MnCl 2 , 5 mM MgCl 2 , and 0.5% Triton X-100. The reaction was terminated by adding 25 l of 0.1 M EDTA and boiling for 5 min, and the supernatant was analyzed using an LC-18 column. The GlcA-T activity was assessed by subtracting the background observed in the negative control sample without donor sugar, and normalized against the amount of protein measured using the DC Protein Assay (Bio-Rad). For the GlcA-T activity in cryosections of mouse or human quadriceps muscles, wheatgerm agglutinin (WGA)-agarose beads incubated with the Triton X-100 extract of proteins from the sections were directly used in the assay. The activity was assessed by subtracting the nonspecific background observed in the WGA-agarose incubated with the Large myd muscle sections, and normalized to tissue volume (cross sectional area x number of sections x thickness of sections).
Mass Spectrometry (MS) Analysis-MS analysis of the enzymatic product was conducted on an Agilent 6520A Quadrupole Time of Flight mass spectrometer as described previously (24).

Metal Dependence of Xyl-T and GlcA-T Activities of LARGE and LARGE2-In a previous study of LARGE and LARGE2
(24), using recombinant enzymes that lack their transmembrane domains (LARGEdTM and LARGE2dTM), we found that these proteins have distinct pH optima for both their Xyl-T and GlcA-T activities. To further optimize the conditions for assaying the endogenous enzymes, we tested the metal dependence of each activity, using Xyl-␣1,3-GlcA-␤-MU and GlcA-␤-MU as acceptors for the GlcA-T and Xyl-T assays, respectively (24). The DXD motifs (Asp-X-Asp; X, any amino acid) required for metal-ion binding and catalysis are conserved in the LARGE paralogs, and many glycosyltransferases with this motif require Mn 2ϩ as a cofactor (34,35). We assayed LARGE and LARGE2 using a buffer containing either a metal ion or EDTA at 5 mM, at the pH that is optimal for each activity. For the Xyl-T activities ( Fig. 1 left), both LARGE paralogs require a metal ion, with LARGE activity dependent on specifically Mn 2ϩ ; notably, LARGE2 activity was lower yet substantial in the presence of Mg 2ϩ or Ca 2ϩ . In contrast, the GlcA-T activities ( Fig. 1, right) of both the LARGE paralogs were substantial in the absence of any metal ion (enzyme preparations were subjected to extensive buffer exchange to remove residual metal ions before the assay was carried out), yet both activities were completely inhibited by addition of EDTA. Mn 2ϩ enhanced the activities of both LARGE and LARGE2 substantially. However, whereas Mg 2ϩ and Ca 2ϩ were as effective as Mn 2ϩ for LARGE activity, they had no effect on LARGE2 activity.
Endogenous LARGE GlcA-T Activity in Cultured Cells-Using the disaccharide acceptor Xyl-␣1,3-GlcA-␤-MU, we tested for endogenous LARGE GlcA-T activity ( Fig. 2A) in cultured cells. First, we used fibroblasts isolated from skeletal muscle of WT or Large myd mice (in the latter, a mutation in the Large gene causes defects in ␣-DG glycosylation) (16). The microsomal membranes were extracted from these fibroblasts and solubilized in 1% Triton X-100, conditions compatible with the GlcA-T assay for LARGEdTM (data not shown). When these solubilized samples were incubated with Xyl-␣1,3-GlcA-␤-MU and UDP-GlcA in a buffer at pH 6.5 (the optimum for LARGE GlcA-T activity (24)), a specific product peak (P) was observed; its retention time was the same as that of its counterpart in the control reaction with LARGEdTM, on both an amide column (Fig. 3A, gradient elution) and an LC-18 column (Fig. 3B, isocratic elution). The elution profiles are consistent with the reaction of LARGEdTM, whose product peak was confirmed as GlcA-added Xyl-␣1,3-GlcA-␤-MU by MS (Fig. 2, B-D). The reaction was specific because the product peak required the presence of both UDP-GlcA (ϩdonor, Fig. 3, A and B) and LARGE (no such peak was detected using Large myd fibroblasts). Product detection was more sensitive when the sample was purified on the LC-18 column (Fig. 3B) versus amide column (Fig. 3A). We were able to detect specific GlcA-T activity in the control fibroblasts with as few as 250,000 cells.
We also applied our assay to microsomes prepared from human Epstein-Barr virus-transformed lymphoblasts from a control individual and a CMD patient (Patient #1) in whom abnormal splicing of the LARGE transcript is accompanied by markedly reduced ␣-DG glycosylation (31). We detected specific GlcA-T activity in the control lymphoblasts but no or very little activity in the patient cells (Fig. 4, A and B). In addition, we examined human myoblasts from a control individual and the other CMD patient (Patient #2) who has compound heterozygous mutations in the LARGE gene. The patient has a novel 74 kb deletion including exons 4 -6 in one allele, and a known missense G1525A (E509K) mutation in exon 13 in the other allele (2). We were able to detect specific GlcA-T activity in the control myoblasts and not in the patient cells (Fig. 4C). Thus, the assay detects specifically the endogenous GlcA-T activity of LARGE in cultured cells. LARGE GlcA-T Activity in Mouse Tissues-We next tested for measurable endogenous LARGE GlcA-T activity in mouse tissues. First, we examined microsomes prepared from the brains of WT mice. The GlcA-T activity was linearly dependent on time, the amount of protein, and the presence of acceptor (Xyl-␣1,3-GlcA-␤-MU) (Fig. 5, A-C) and showed a typical saturation curve for the donor (UDP-GlcA) (Fig. 5D). The activity was specific to LARGE, as it was reduced to 43.5% in the Large myd heterozygous mouse (ϩ/Ϫ) versus WT (ϩ/ϩ) brain sample, and was undetectable in the homozygous Large myd mouse (Ϫ/Ϫ) brain sample (Fig. 5E). Next we assessed the total LARGE-and LARGE2-derived GlcA-T activity in the tissues of WT, Large myd , and Large2 Ϫ/Ϫ mice, using a buffer at pH 6.0 (at which both LARGE paralogs show substantial GlcA-T activity (24)). We observed specific GlcA-T activity in both the WT and Large Ϫ/Ϫ brains but none in the Large myd brain (Fig. 6, A and B). Results for heart and skeletal muscle were similar, showing that LARGE2 is not substantially active in these tissues. In the case of kidney, in contrast, both the Large myd and the Large2 Ϫ/Ϫ mice had reduced but some levels of enzyme activity (Fig. 6B). These results are consistent with the expression profiles of the LARGE paralogs in these tissues (28,29).
Physical Interaction between LARGE2 and ␣-DG-The N-terminal domain of ␣-DG must be present for LARGE to functionally modify the ␣-DG mucin-like domain (26). Recently, a limb-girdle muscular dystrophy patient was found to have a missense mutation that results in substitution T192M (Thr-1923 Met) within the ␣-DG N terminus, and it was sug-gested that the accompanying failure in functional modification of ␣-DG is a consequence of disruption of the molecular interaction between ␣-DG and LARGE (32). Using ␣-DG-Fc fusion proteins with or without the T192M substitution, we tested the interaction with LARGE2 by a pull-down assay, as examined with LARGE in the previous study. While both Fc5 (␣-DG-Fc) and Fc2 (␣-DG N terminus-Fc) bound to LARGE2, Fc proteins harboring the T192M substitution bound much less efficiently (Fig. 7). These data support the notion that, like LARGE, LARGE2 is involved in the functional modification of ␣-DG, and that the T192M substitution also prevents this modification.
Functional Modification of ␣-DG in Mouse Tissues-We examined the functional modification of ␣-DG in mouse tissues by immunoblotting WGA-enriched glycoproteins from the WT, Large myd , and Large2 Ϫ/Ϫ mice using antibodies to the ␣-DG glycan epitopes (IIH6 and VIA4 -1), and by carrying out a laminin overlay assay. In the cases of skeletal muscle, heart, and brain, the functional modification of ␣-DG in the Large2 Ϫ/Ϫ mouse was comparable to that in the WT mouse (Fig. 8). The samples from the corresponding Large myd tissues showed no reactivity to IIH6 or VIA4 -1, or laminin binding capacity, even though the ␣-DG core protein was expressed at high levels (Fig. 8). In the kidneys, in contrast, the functional modification was robust only in the WT samples; it was diminished in both the Large myd and the Large2 Ϫ/Ϫ samples, with respect to both intensity of staining and molecular weight (Fig.  8). Thus, like LARGE, LARGE2 contributes to the functional modification of ␣-DG in the kidney.
LARGE GlcA-T Activity in Mouse and Human Skeletal Muscle Sections-In the course of carrying out this study, we found that we could enrich specifically the GlcA-T activity of endogenous LARGE by purification with WGA-agarose. When solu-

Endogenous LARGE Enzymatic Activity in Cells and Tissues
OCTOBER 10, 2014 • VOLUME 289 • NUMBER 41 bilized membranes from rabbit brain were incubated with WGA-agarose, almost all of the LARGE GlcA-T activity bound to the beads, and the activity was specifically eluted from the beads by N-acetylglucosamine (Fig. 9A). Given the potential usefulness of assessing LARGE GlcA-T activity from biopsy sections, we took advantage of this enrichment to adapt our assay to measuring LARGE GlcA-T activity from relatively small amounts of mouse and human skeletal muscle. We first subjected extracts of 30 quadriceps muscle cryosections of 10-m thickness (section area; 18ϳ42 mm 2 ) to WGA enrichment. Subsequent measurement of the specific GlcA-T activity as for the experiments above revealed that the activity in the mouse and human skeletal muscle sections was comparable, and that no activity was present in the sections from Large myd mice (Fig. 9B). Similar results were obtained when the assay was repeated with the other sets of extracts from human and mouse quadriceps muscle cryosections.

DISCUSSION
Given that the expression levels of endogenous glycosyltransferases in the Golgi apparatus are generally low, it has been a challenge to detect the endogenous LARGE proteins in cells and tissues. Our modified assay, which enables us to measure the endogenous enzymatic activities of LARGE and LARGE2 in cultured cells and tissues, revealed that LARGE2 has the capacity to functionally modify ␣-DG in a physiologically relevant setting.
Our initial characterization of the metal dependence of the enzymatic activities of LARGE and LARGE2 indicates that, as is the case for the pH optima of the Xyl-T and GlcA-T activities (24), their metal binding requirements also differ. This is consistent with what we know about the other glycosyltransferases that contribute to the functional modification of ␣-DG. For example, POMT activity does not require metal cations whereas POMGnT1 does and, moreover, has a strong preference for Mn 2ϩ (36 -38). In the cases of LARGE and LARGE2, both the Xyl-T and GlcA-T activities have a preference for Mn 2ϩ . Notably, although the Xyl-T activities showed an almost complete dependence on the cation, as did the GlcA-T activity of LARGE2, the GlcA-T activity of LARGE was substantial in the presence of either Mg 2ϩ or Ca 2ϩ . The physiological underpinnings of this difference remain to be determined.
For measuring the endogenous activity, we used Xyl-␣1,3-GlcA-␤-MU as the acceptor in the GlcA-T assay for the LARGE proteins and found it to be a good acceptor substrate. The disaccharide acceptor probably has an advantage over the simpler monosaccharide acceptors (such as the Xyl-␣-pNP and GlcA-␤-MU used in the initial characterization of enzymatic activities of LARGE) with respect to the specificity and sensitivity of the detection in contexts where enzyme is limiting. This is in contrast to our initial study in which abundant recombinant enzyme was used for the assay (23), and was crucial for developing an assay that could be carried out on small samples such as those obtained from patient biopsies.
We used the GlcA-T assay to compare activities among control and Large myd mouse fibroblasts, as well as human lymphoblasts from a control and a CMD patient (Patient #1) who has an abnormal transcript of LARGE. The Large gene in the  Large myd mouse has an intragenic deletion of exons 4 -7 which causes a frameshift and premature stop codon before the first catalytic domain in the resultant mRNA (16). The CMD patient is homozygous for a LARGE allele in which an intra-chromosomal duplication is inserted into intron 10, and thus most of the transcripts are truncated immediately after exon 10 (31) and produce a protein that does not contain the second catalytic domain, i.e. the GlcA-T domain (23). The finding of no or only very low GlcA-T activity in these cells accounts for the defective glycosylation of ␣-DG in the mutant mouse and patient. In the case of the other CMD patient (Patient #2), one of the alleles has a novel 74 kb deletion including exons 4 -6, and the other allele has a known missense G1525A (Glu509Lys) mutation (2). When LARGE was coexpressed with ␣-DG in cultured cells, the Glu509Lys mutant failed to enhance the functional modification (28). Our assay data is consistent with those previous reports; the patient cells had no or very low GlcA-T activity with the mutations. Dystroglycanopathies can be caused by mutations in multiple genes related to O-mannosyl glycan synthesis, and these lead to an extremely broad clinical spectrum and relatively poor correlations between genotype and phenotype (15). An enzy-matic assay of POMGnT1 activity has been carried out on extracts of muscle biopsies from Muscle-Eye-Brain disease patients and various controls, as well as on fibroblasts and lymphoblasts from the patients and carriers (39,40). Similarly, assays have been developed for both the POMGnT1 and POMT activities, and have been used on lymphoblasts and fibroblasts from dystroglycanopathy patients, including some that have not been genetically characterized (41,42). These reports indicate that enzymatic assays of known glycosyltransferases are viable tools for rapid and sensitive screening for dystroglycanopathies. Our enzymatic assay for LARGE activity, which can be applied to cryosections of tissue biopsies as well as fibroblasts, immortalized lymphoblasts and myoblasts, is expected to likewise be a powerful and complementary tool that will identify patients with pathogenic mutations of LARGE.
Our analysis of tissues from WT, Large myd , and Large2 Ϫ/Ϫ mice using the LARGE GlcA-T assay demonstrated a good correlation between enzymatic activity and the tissue expression profiles of the Large transcripts (i.e. robust expression of Large but not Large2 transcript in mouse brain, heart, and skeletal muscle, and substantial expression of both transcripts in kidney (28,29)). Since Large2 expression is highest in kidney we ana-

Endogenous LARGE Enzymatic Activity in Cells and Tissues
OCTOBER 10, 2014 • VOLUME 289 • NUMBER 41 lyzed the null mice for kidney disease. Our tests for proteinuria as an indicator of kidney disease revealed no overt abnormalities in the null mice (data not shown). Considering the overlap in function and tissue-specific expression of Large and Large2, it does not come as a surprise that a lack of Large2 expression can be tolerated and functionally compensated by the wide expression of the paralog Large. Such genetic redundancy has been widely accepted to account for the lack of phenotypes in many knockouts (43). Consistent with redundancy as an explanation for the lack of an apparent phenotype in the Large2 Ϫ/Ϫ mice, a study of a mouse model of Fukuyama-type congenital muscular dystrophy, which is caused by a mutation in the fukutin gene, has indicated that even a small amount of IIH6-reactive ␣-DG is sufficient to maintain integrity of the skeletal muscle (44). On the other hand, it must also be mentioned that in contrast to previous reports suggesting functions for DG in the kidney (45,46), a study using a conditional DG allele with a variety of Cre promoters suggests that DG does not contribute significantly to kidney development or function (47). Given that DG is the only known substrate for modification by LARGE or LARGE2, these findings raise questions about the physiological significance of the LARGE paralogs in the kidney. An alternative possibility is that LARGE or LARGE2 adds the lamininbinding glycan to an as yet unknown substrate(s), as suggested by a study based on adenovirus-mediated LARGE overexpression in DG-deficient cells (48), and that this aspect of LARGE activity may account for their physiological relevance in the kidney.
In conclusion, our results from immunoblotting kidney extracts from WT and mutant mice, as well as our pull-down assay for the interaction between the ␣-DG N terminus and LARGE2, provide strong evidence that LARGE2 contributes to the functional modification of ␣-DG in vivo.