Glycosyltransferase POMGNT1 deficiency affects N-cadherin-mediated cell-cell adhesion

Defects in protein O-mannosylation lead to severe congenital muscular dystrophies known as α-dystroglycanopathy. A hallmark of these diseases is the loss of the O-mannose-bound matriglycan on α-dystroglycan, which leads to a reduction in cell adhesion to the extracellular matrix. Mutations in protein O-mannose β1,2-N-acetylglucosaminyltransferase 1 (POMGNT1), which is crucial for the elongation of O-mannosyl glycans, are mainly associated with muscle-eye-brain (MEB) disease. In addition to defects in cell-extracellular matrix adhesion, aberrant cell-cell adhesion has occasionally been observed in response to defects in POMGNT1. However, direct molecular mechanisms are largely unknown. We used POMGNT1 knock-out HEK293T cells and fibroblasts from a MEB patient to gain a deeper insight into the molecular changes in POMGNT1 deficiency. A combination of biochemical and molecular biological techniques with proteomics, glycoproteomics and glycomics revealed that a lack of POMGNT1 activity strengthens cell-cell adhesion. We demonstrate that the altered intrinsic adhesion properties are due to an increased abundance of N-cadherin (N-Cdh). In addition, site-specific changes in the N-glycan structures in the extracellular domain of N-Cdh were detected, which positively impact on homotypic interactions. We found that in POMGNT1 deficient cells ERK1/2 and p38 signaling pathways are activated and transcriptional changes that are comparable to the epithelial-mesenchymal transition (EMT) are triggered, defining a possible molecular mechanism underlying the observed phenotype. Our study indicates that changes in cadherin-mediated cell-cell adhesion and other EMT-related processes may contribute to the complex clinical symptoms of MEB or α-dystroglycanopathy in general, and suggests a previously underestimated impact of changes in O-mannosylation on N-glycosylation.


Introduction
The modification of proteins by glycosylation is an ubiquitous feature of all living organisms (1).
Protein-linked glycans are involved in a multitude of cellular processes ranging from monitoring the folding state of glycoproteins to cell adhesion and migration (2). Among the different types of glycosylation, Nglycosylation and O-mannosylation are evolutionary conserved from bacteria to mammals. In humans, changes in those essential protein modifications inter alia can modulate immune responses, promote cancer cell metastasis and underlie the pathophysiology of severe congenital disorders (3)(4)(5). Both modifications initiate at the endoplasmic reticulum (ER), where the target polypeptides and the donor saccharides are synthesized and eventually covalently linked (2). Only properly glycosylated and folded proteins can leave the ER and travel through the Golgi apparatus to reach their final cellular destinations. On their way, Nlinked and O-mannosyl glycans can be further modified which leads to diverse species-or even cell typespecific glycans (2).
In the case of N-glycosylation, the dolichol-pyrophosphate-linked oligosaccharide Glc3Man9GlcNAc2 is assembled at the ER membrane and the glycan moiety is transferred en bloc to Asn residues of the consensus sequon Asn-Xaa-Ser/Thr/Cys (X: proline is excluded). This way, the vast majority of proteins that enter the secretory pathway are N-glycosylated including many cell surface receptors and cell adhesion molecules (6). Protein-linked carbohydrate moieties are then further processed, and finally extended in the Golgi through the concerted action of diverse specific glycosyltransferases resulting in three distinct types of N-glycans: high-mannose, complex-and hybrid-type, which contain the common core Man3GlcNAc2-Asn (2). The diverse glycan structures and glycosylation patterns on cell surface molecules are highly dynamic and can be differentially regulated both during development and in certain pathological conditions, often associated with the acquisition of altered cellular properties (4).
Classically, O-mannosylation is initiated by the conserved PMT family of protein Omannosyltransferases (POMT1 and POMT2 in mammals), which catalyse the transfer of mannose from dolichol-phosphate-linked mannose to Ser and Thr residues of nascent proteins (7). Three different core structures can be built on the protein-linked mannose (8). Linear core m1 and branched core m2 glycans, which share the common inner core GlcNAc-β1,2-Man-Ser/Thr, are initiated in the cis Golgi by the addition of a GlcNAc residue by the protein O-mannose β1,2-N-acetylglucosaminyltransferase 1 (POMGNT1), and are further extended while proteins travel through the Golgi to the cell surface. In contrast, core m3 glycans are already elongated in the ER (GalNAc-β1,3-GlcNAc-β1,4-(phosphate-6)-Man-Ser/Thr) and then further modified in the Golgi by the sequential action of numerous glycosyltransferases including the ribulose-5phosphate transferase fukutin (FKTN). The resulting complex polysaccharide structure, known as "matriglycan" is so far only found on α-dystroglycan (α-DG), a central member of the dystrophin glycoprotein complex family in peripheral membranes, and enables its interaction with extracellular matrix (ECM) components such as laminin (9). Defects in this complex biosynthetic pathway lead to the loss of the matriglycan on α-DG, and consequently impair interactions between α-DG and e.g. laminin, which interferes with the formation of basement membranes (9). This defect has been recognized as a major pathomechanism of severe congenital muscular dystrophies with neuronal migration defects, known as αdystroglycanopathy (OMIM 236670; 253280; 253800; 606612; 607155; 608840) (10).
The glycosyltransferase POMGNT1 has a key role in the elongation of O-mannosyl glycans (11). In its absence not only core m1 and m2 structures are missing, also formation of the matriglycan fails, since POMGNT1 recruits FKTN to maturing core m3 structures (12,13). The great majority of mutations in POMGNT1 have been linked to muscle-eye-brain disease (MEB; OMIM 253280), a congenital muscular dystrophy in humans, which is characterized by additional brain malformations and structural anomalies in the eye (11). In the murine model, knock-out of POMGNT1 is viable with multiple developmental defects, similar to the clinical picture of human MEB patients (14,15). The pathology of MEB suggests a functional role for POMGNT1 in control of cell adhesion and migration. For example, in the transgenic POMGNT1based MEB mouse model impaired cell-ECM adhesion results in disruption of basement membranes and over migration of neurons during development of the cerebral cortex (15). However, also clusters of granule cells which failed to migrate have been frequently observed (15). In addition to its important role during mammalian development, POMGNT1 has recently been linked to the progression of glioblastoma, fatal primary brain tumors with survival time of 12-15 months, as well as the resistance of glioblastoma cells to the chemotherapeutic agent temozolomide (16,17). Strikingly, in glioblastoma models increased cell-cell adhesion has been observed when POMGNT1 is missing (16). However, molecular reasons for the different consequences of POMGNT1 deficiency are just emerging.
Very recently, glyco-engineered human embryonic kidney (HEK) 293 cells turned out to be especially useful for the characterization of known, as well as the identification of new glycosylation pathways (18,19). In the present work, we took advantage of a gene-targeted POMGNT1 knock-out in HEK293T cells to study the consequences of POMGNT1 deficiency. The combination of glyco(proteo)mics with classic biochemistry, molecular and cell biology resulted in the discovery that cell-cell adhesion mediated by neuronal cadherin (N-Cdh) is affected and defined a possible molecular mechanism underlying the observed phenotype. Similar effects in MEB patient-derived fibroblasts confirmed the validity of the HEK293T model to study molecular effects of O-mannosylation deficiencies.

POMGNT1 deficiency impairs cell-matrix and reinforces cell-cell interactions in a HEK293T cell model
To gain insight into functional implications of POMGNT1 deficiency, we generated a gene-targeted knock-out in HEK293T cells (ΔPOMGnT1) as detailed in Experimental procedures and Fig. S1. The loss of POMGNT1 activity was confirmed by the O-mannosylation status of the endogenous substrate α-DG using the matriglycan-directed antibody IIH6. Whereas the O-linked matriglycan is absent in POMGNT1depleted cells, reintroduction of human POMGNT1 rescued O-mannosylation of α-DG verifying the specificity of the system (Fig. 1A). General characterization of the morphology of POMGNT1 knock-out cells by confocal microscopy revealed that POMGNT1-deficient cells appear more rounded and stronger aggregated compared to wild type cells which show extensive spreading and even distribution. This phenotype is also reverted upon reintroduction of POMGNT1 (Fig. 1B).
To further characterize molecular events responsible for the morphological differences, we analyzed cell-matrix and cell-cell adhesion. As expected, POMGNT1-deficient cells adhere to laminin, a major ECM component and interactor of the α-DG matriglycan, to a significantly lower extent when compared to wild type cells ( Fig. 2A). Intriguingly, when confluent monolayers of wild type cells were incubated with wild type and knock-out cells, respectively, cell-cell adhesion of ΔPOMGnT1 cells turned out to be significantly increased. The same result is observed using a monolayer of ΔPOMGnT1 cells (Fig. 2B). Since cell-matrix and cell-cell interactions are major opposing forces balancing cellular migration, we further took advantage of xCELLigence real-time cell analysis that allows live monitoring of cell proliferation and cell migration.
ΔPOMGnT1 cells proliferate slower than wild type cells with slopes of 0.07 and 0.09, respectively (Fig. S2,   A and B). In agreement with increased cell-cell adhesion, the migration rate of ΔPOMGnT1 cells is reduced by a factor of three ( Fig. 2C and Fig. S2D).
Taken together, the POMGNT1 knock-out HEK293T cell model revealed that cell-cell adhesion increases, whereas cell-matrix interactions and cell migration are negatively affected when O-mannosyl glycans are not further elongated.

Increased cell-cell adhesion of POMGNT1-deficient cells is caused by changes in N-Cdh abundance
In order to identify determinants which underlie the observed phenotype in ΔPOMGnT1 cells, we performed label-free quantitative proteomics of whole cell lysates from wild type and ΔPOMGnT1 HEK293T cells. Five independent replicates were analyzed and homoscedasticity and normal distribution were confirmed (Fig. S3, A and B). Altogether, 86 out of 437 proteins with differential abundance in POMGNT1-deficient cells could be identified ( Fig. 3A and Fig. S3D, Table S4). Interestingly, gene ontology term functional annotation of proteins with a significant regulation revealed enrichment for proteins under the molecular function term of "cadherin binding involved in cell-cell adhesion" (Fig. 3A and S3D, Table S4), pointing to an impact of POMGNT1 deficiency on cadherin-mediated cell-cell adhesion. In addition, the protein N-Cdh was found to be increased by a factor of three in ΔPOMGnT1 cells (Fig. 3A). This result was also confirmed by Western blot (Fig. 3, B and C) and correlated well with increased mRNA levels of N-Cdh (Fig. 3D). To investigate the general validity of our findings, we took advantage of skin fibroblasts derived from an MEB patient who presented characteristic symptoms such as mental retardation and blindness due to variant c.535_751del (p.Asp179Argfs*11) in the POMGNT1 gene (NM_017739.4). In accordance with our HEK293T model, protein and mRNA abundance of N-Cdh in MEB patient-derived fibroblasts showed increased values compared to fibroblasts from two healthy donors (Fig.

3, E-G).
Cadherins are major players in the formation of cellular junctions (20). These membrane-anchored cell surface glycoproteins mediate cell-cell adhesion through homotypic interactions of their conserved extracellular domains. Thus, to determine whether elevated cell-cell adhesion observed in POMGNT1deficient cells is directly linked to the increased abundance of N-Cdh, we performed cell-cell adhesion assays in presence of either N-Cdh blocking or IgG-directed antibodies. As shown in Fig. 4, ΔPOMGnT1 cells show increased adhesion to wild type and other ΔPOMGnT1 cells in the IgG-treated controls. This increase, however, is diminished upon incubation with an N-Cdh targeting antibody, demonstrating N-Cdh as a key molecular driver for increased cell-cell adhesion in the established ΔPOMGnT1 HEK293T cell model.

POMGNT1 deficiency affects N-Cdh N-glycosylation
N-Cdh is highly N-glycosylated and N-linked glycans affect its adhesive properties by modulating its homomeric interactions in cis and in trans (21,22). We therefore asked whether an altered N-glycosylation, as an indirect effect of POMGNT1 deficiency, could contribute to increased N-Cdh-mediated cell-cell adhesion. We performed comprehensive N-glycomics and N-glycoproteomics on N-Cdh. For that purpose, the extracellular domain (EC) of N-Cdh was recombinantly expressed and purified from wild type and ΔPOMGnT1 cells (detailed in Experimental procedures; Fig. S4). N-glycomics analysis by multiplexed capillary gel electrophoresis with laser induced fluorescence detection (xCGE-LIF) revealed that total abundance of fully galactosylated and sialylated N-glycan structures is remarkably decreased on N-Cdh derived from ΔPOMGnT1 cells when compared to wild type (Fig. 5A, normalized intensity of peaks 2, 4, 9 and 10). In accordance, the abundance of non-galactosylated N-glycan structures is increased (Fig. 5A, normalized intensity of peaks 6 and 7). Moreover, fully galactosylated multiantennary N-glycans (Fig. 5A, A3+F) are reduced on ΔPOMGnT1-derived N-Cdh and a non-galactosylated complex-type N-glycan with a bisecting GlcNAc (GlcNAc(5)Man(3)+Fuc(1)) represents the dominating N-glycan structure (detailed in Supporting information SI5).
In order to gain even deeper insights, an exploratory site-specific approach was taken to Characterization of the N-glycan microheterogeneity based on intact N-glycopeptides revealed differences in relative abundance of the major N-glycoform at each site on N-Cdh derived from ΔPOMGnT1 compared to wild type cells (Fig. 5C). All N-glycosylation sites (except for Asn651 and Asn692 which are both located at the stem region of the molecule) feature non-galactosylated complex-type N-glycans with a bisecting GlcNAc (GlcNAc(5)Man(3)±Fuc(1)) as the dominating composition on N-Cdh from ΔPOMGnT1 cells.
Comparative and site-specific N-glycoproteomics of quantitative changes in the N-glycan microheterogeneity revealed a significant decrease in galactosylation (-27% on average) and sialylation (-16% on average) throughout the vast majority of N-glycosylation sites (only exception are Asn190 and Asn651) (Fig. 5C). This decrease in galactosylation and sialylation is associated with a significant increase of non-galactosylated N-glycan compositions that feature a bisecting GlcNAc (Fig. 5B). The site-specific relative changes of N-glycan traits on N-Cdh are shown in detail in Supporting information SI1-SI3. Overall, N-glycosylation of N-Cdh in ΔPOMGnT1 cells exhibits a reduction in the degree of galactosylation and sialylation with non-galactosylated complex-type N-glycans with a bisecting GlcNAc dominating. We next measured the transcript levels of genes encoding for major glycosyltransferases of N-glycan processing that could be responsible for the observed changes in the N-glycan profile of N-Cdh using the nCounter technology (23). A decrease by 27% of β-1,4-galactosyltransferase 1 B4GALT1 mRNA as well as a reduction by 83 % in the transcript level of sialyltransferase ST6GAL1 was detected, explaining the limited occurrence of respective N-glycans on N-Cdh in ΔPOMGnT1 cells (Table S6) and wild type HEK293T cells was detected, excluding indirect effects on the activity of TMTCs.

Changes in N-Cdh N-glycosylation impact on its homotypic interactions in vitro
Hypo-N-glycosylation of N-Cdh increases the prevalence of cis N-Cdh dimers on the cell membrane, thereby stabilization of cell-cell contacts (21,22). To determine whether the observed changes in N-glycosylation of N-Cdh also favor homotypic interactions, the adhesive properties of the recombinant extracellular domain of N-Cdh (see above) were analyzed using an in vitro bead aggregation assay (26).
Dynabeads coated with the protein isolated from wild type or ΔPOMGnT1 HEK293T cells were incubated with either CaCl2 or EDTA as a control. Aggregation of beads, reflecting Ca 2+ -dependent homotypic interactions of the extracellular domain of N-Cdh, was recorded and quantified at specific time intervals over the course of 1 hour. As shown in Fig. 6A, ΔPOMGnT1-derived N-Cdh triggers the formation of larger aggregates in comparison to wild type-derived N-Cdh, demonstrating that the adhesive properties of N-Cdh from ΔPOMGnT1 cells are enhanced. Taken together, changes in cell-cell adhesion of ΔPOMGnT1 cells are due to a higher frequency of N-Cdh with less complex N-linked glycan structures that facilitate stronger homotypic interactions.

Loss of function of POMGNT1 activates ERK and p38 signaling
To further investigate how POMGNT1 deficiency modulates N-Cdh-mediated cell-cell adhesion, we explored possible changes in cellular signaling using a human phosphokinase array with 46 different kinases. Of all kinases examined, the extracellular signal-regulated kinase ERK1/2 and the mitogenactivated protein kinase p38 showed significantly enhanced phosphorylation in ΔPOMGnT1 HEK293T cells when compared to wild type cells (Fig. 7, A and B). Since dysregulation of DG and matriglycan biosynthesis has been associated with modulation of the ERK-MAPK pathway (27,28), the identified changes were further validated by Western Blot. Consistent with the phospho-kinase array data, ΔPOMGnT1 cells showed a significant increase in the phosphorylation levels of ERK1/2. Additionally, phosphorylation of the upstream MAP kinase kinase MEK1/2 (29) was found to be increased (Fig. 7, C and D) further corroborating that the ERK-MAPK pathway is activated in our ΔPOMGnT1 HEK293T cell model.

Lack of POMGNT1 activity induces an EMT-related transcriptional response
Increased expression of N-Cdh (30), transcriptional modulation of N-glycan modifying enzymes (31) and activation of the ERK-MAPK pathway (32) have been linked to epithelial-mesenchymal transition (EMT), a process which is of major importance during development, wound healing as well as tumor progression and metastasis (33). Therefore, the impact of POMGNT1 deficiency on the expression of EMT marker genes was investigated. As shown in EMT processes are associated with the upregulation of matrix metallopeptidases (MMPs) which represent a family of zinc-dependent endoproteases involved in degrading ECM components (34). A strong and specific increase of MMP1, MMP3, MMP8 and MMP17 mRNA is observed in ΔPOMGnT1 cells (Fig.   8A). Also expression of metallopeptidase inhibitors (TIMPs; (35)) is affected. Likewise, analysis of the abundance of MMP proteins using a human MMP antibody array showed that MMP1, MMP3 and MMP8 levels are increased (Fig. 8, B and C).
To determine whether decreased POMGNT1 activity also triggers an EMT-like transcriptional response in MEB patient-derived fibroblasts, mRNA levels of the above genes were assessed. Indeed, in comparison to control fibroblasts, epithelial markers (LAMA2 and CLDN6) were found to be significantly reduced, whereas mesenchymal markers (ACTA2, VIM encoding vimentin and FN encoding fibronectin) were significantly elevated in MEB patients (Fig. 9A). In addition, the transcriptional modulation of MMPs and TIMPs was also comparable to the changes observed in ΔPOMGnT1 HEK293T cells with particularly large increase in transcript levels of MMP1 and MMP17 (Fig. 9B).
Altogether, our data demonstrate that lack of POMGNT1 activity induces an EMT-like transcriptional response resulting, inter alia, in elevated levels of ECM degrading enzymes.

Discussion
-Dystroglycanopathy is associated with the development of a variety of medical conditions, including muscular dystrophy and mild to severe changes in the central nervous system and the eyes. So far, 18 causative genes for α-dystroglycanopathy have been identified that are associated with the O-mannose glycosylation pathway of α-DG (10). Among those are POMT1/POMT2 and POMGNT1, the major disease causing factors of the Walker Warburg syndrome (WWS; lacking classic O-mannose glycans), and MEB (no elongation of O-linked mannoses), respectively. Current knowledge regarding the predominant disease mechanism suggests that incomplete matriglycan biosynthesis impairs the binding of α-DG to extracellular matrix proteins such as laminin, resulting in damage to cell membrane integration and defective basement membranes (9). In recent years, it turned out that changes in classic O-mannosylation due to altered expression of POMT2 impact also on epithelial cadherin (E-Cdh)-mediated cell-cell adhesion during murine embryonic development and in human gastric carcinoma (36,37). However, whether the observed changes only apply to E-Cdh and POMT2 is unclear, as are the underlying molecular mechanisms. Taking advantage of POMGNT1-deficient HEK293T cells and MEB patient-derived fibroblasts, we now show that, in general, aberrant classic O-mannosylation impacts on cadherin-mediated cell-cell adhesion. We demonstrate that loss of POMGNT1 function results in (1) reinforced N-Cdh-mediated cell-cell adhesion and impaired cell migration potential; (2) increased levels of N-Cdh and changed N-glycan structures on its extracellular domain, which in turn enhance its intrinsic adhesive properties; (3) activation of ERK1/2 and p38 signaling pathways and induction of transcriptional modulations which are comparable to EMT-like events.
Changing cell migration behavior is a (patho)physiological process determined by the opposing forces that define cell-ECM and cell-cell interactions. In invasive cancers, decreased levels of POMT1, POMT2 and other enzymes of the classic O-mannosylation pathway correlate with high cell migration and invasion (28). But, low POMGNT1 levels can also hinder cell migration. A POMGNT1-based MEB mouse model revealed clusters of granule cells within the cerebellum, which have failed to migrate during development (14). Further, Abbott and coworkers reported that knock-down of POMGNT1 and MGAT5B impairs neuronal cell migration (38), and Lan and coworkers found that silencing of POMGNT1 decreases cell proliferation and invasion in glioblastoma (16). Therefore, POMGNT1 deletion may also act as a break on invasion and migration, however, the direct molecular mechanism for the observed phenotypes are not fully understood. Our HEK293T cell model recapitulates the observed impact of POMGNT1 on cell adhesion and migration observed in neuronal and glioblastoma cells, and identifies increased N-Cdh-mediated cellcell adhesion as one of the reasons that influences the cellular migration potential when POMGNT1 activity is reduced. In addition to the increased amount of N-Cdh, its N-glycosylation pattern is changed to further promote N-Cdh homotypic interactions. In general, on N-Cdh from POMGnT1 N-glycans with a lower degree of galactosylation and sialylation, along with a slight increase in the abundance of bisecting GlcNAc are observed. These changes are driven by the transcriptional regulation of specific glycosyltransferases, such as B4GALT1 and ST6GAL1, which act as heteromeric complex in the successive addition of terminal ß1,4-linked galactose and -2,6-linked sialic acid to N-glycans (39). Most interestingly, the changes in Nglycosylation are site specific and especially site N402 in the EC3 domain contains less branched complex glycans, which have been reported to favor cell-cell adhesion. N402, along with N273 and N325, has previously been identified as one of the most relevant N-glycosylation sites that strongly increase cisdimerization of N-Cdh when complexity of N-glycans is decreased (21). In line with previous findings of the group of S. Pinho (in collaboration with our group) that in human gastric carcinomas expression levels of POMT2 correlate with altered structures of N-glycans on E-cadherin (37), we now demonstrate a link between classic O-mannosylation and N-glycosylation of cadherins which was for the first time analyzed in such depth.
Very recently it turned out that cadherins are also major targets of TMTC1-3 which add single nonextended O-linked mannoses to EC domains (18) that impact on cellular adherence (40). Indirect effects on TMTC-based O-mannosylation of N-Cdh due to lack of POMGNT1 could be excluded, as its Omannosylation is not altered in our HEK293T cell model.
Increased expression of N-Cdh is part of a transformation process that mostly epithelial cells undergo known as epithelial-to-mesenchymal transition (30). EMT is inherent to physiological processes such as embryonic stem cell differentiation and development (41) and has further been associated with pathological conditions such as wound healing, fibrosis and cancer stemness and progression (42)(43)(44)(45). Key events of EMT are dissolution of epithelial cell-cell junctions, loss of the apical-basal polarity and reorganization of the cytoskeleton as well as downregulation of epithelial gene expression in favor of genes that establish the more motile mesenchymal phenotype (33). Long regarded as an all-or-nothing event, EMT is now considered a transition suggesting a gradual and reversible process that does not exclusively concern epithelial cells (46). EMT does not necessarily include increased cell motility but can even lead to the opposite migration behavior (47,48). Our results on POMGNT1-deficient HEK293T and fibroblast cells point to a partial EMT event that is accompanied by reduced expression of epithelial and increased expression for most of the mesenchymal markers tested. Very recently differential expression of POMGNT1 in human glioma cell lines has been reported to impact on the expression of some EMT marker proteins (17) further corroborating the general relevance and validity of our HEK293T cell model. We also observe a predominant induction of MMPs that is indicative of an EMT-like transition. Increased level of MMPs have been shown to be involved in degradation of the basement membrane in developmental processes (49,50), wound healing (51) and cancer progression (34). Since breakdown of basement membranes is also one of the hallmark events in MEB disease (11) or -dystroglycanopathy in general, elevated levels of MMPs could increase ECM protein degradation and impede cell-ECM interactions further deteriorating the clinical outcome of these patients. A similar role has been suggested for MMP2 and MMP9 in physiological and pathological conditions involving members of the dystrophin glycoprotein complex (52).
In POMGnT1 cells partial EMT correlates with sustained activation of ERK1/2 and p38 MAPK signaling which have been associated with EMT induction (53). Whereas these pathways seem to be decisive for the EMT response observed, the source of its activation remains unclear. Although other factors cannot be excluded completely, the most likely candidate that induces ERK signaling is the dystrophin glycoprotein complex and DG in particular, that was suggested as a multifunctional adaptor, capable of interacting with components of the MAPK cascade including MEK and ERK (27). The role of DG on developmental and pathological EMT processes is well documented (50,54). In line with our data, α-DG glycosylation and its interaction with laminin has been suggested to enable β-DG to sequester ERK preventing it from translocating into the nucleus and promoting differentiation of a mesenchymal phenotype (55). To our best knowledge, so far no link between p38 and DG has been reported. How p38 MAPK pathway is linked to POMGNT1 deficiency will be an interesting future question. In many cancers, heterogeneity in protein structures due to aberrant glycosylation results in induction of EMT (31). Thus, the observed modulation of N-glycan modifying enzymes in ∆POMGnT1 cells also might contribute to some extend to the induction of EMT-linked signaling pathways.
Our study proved ∆POMGnT1 HEK293T cells as an excellent system to study molecular details underlying -dystroglycanopathy. Our data suggest a model (Fig. 10) in which due to the loss of O-linked matriglycan on α-DG, ERK and p38 signaling cascades are activated in POMGNT1-deficient cells. As a consequence, EMT-like transcriptional events result inter alia in the induction of N-Cdh (CDH2) and MMPs. In addition, transcriptional modulation of N-glycan modifying enzymes contributes to increased N-Cdh homotypic interactions. In combination, these events result in enhanced cell-cell adhesion and disordered basement membranes thereby contributing to the molecular pathogenesis of MEB disease.

Generation of TALEN-mediated POMGNT1 knock-out cells
Transcription Activator-Like Effector Nuclease (TALEN)-based constructs were designed utilizing the TAL Effector Nucleotide Targeter 2.0 online tool (56). Three TALEN constructs were selected which target the ATG start codon of POMGNT1, the coding sequence of the D-X-D motif and an intron-exon boundary upstream of the exon containing the D-X-D motif. All constructs present a restriction site close to the predicted TALEN cleavage site to allow screening of induced mutations by restriction fragment length polymorphism (RFLP) analysis. TALENs were assembled using the Golden Gate TALEN and TAL Effector Kit 2.0 (57) and employing pC-GoldyTALEN as the final expression vector (#1000000024 and #38143 respectively; both from Addgene). Correct assembly of TALEN plasmid DNA was verified by restriction site analysis and sequencing and TALEN plasmids were transfected into wild type HEK293T cells. Next, genomic DNA was isolated from transfected cells using the DNeasy Blood and Tissue Kit (Qiagen) according to manufacturer's instructions and POMGNT1 regions were amplified by PCR and digested using restriction sites close to the predicted TALEN cutting site. Detection of PCR products resistant to restriction, indicated efficient cleavage by TALENs and was followed by dilutional cloning (1 cell/100 μl/96-well) and further genomic DNA isolation, PCR and RFLP analysis to obtain single cell knock-outs. From the selected TALEN construct combinations, the one targeting the ATG start codon of POMGNT1 (close to a HpaII restriction site; forward TALEN construct: TGGTGACCCGCCAAT, reverse TALEN construct: GGAAGCCCAGCCCCCTC) was the most efficient in inducing mutations and was therefore used for subsequent dilutional cloning, PCR and RFLP. PCR analysis and sequencing of genomic DNA confirmed the knock-out of POMGNT1 (Fig. S1). POMGNT1 knock-out clones were furthermore evaluated by Wheat Germ Agglutinin (WGA) enrichment for absence of reactivity towards matriglycan on α-DG recognized by IIH6 antibodies (58) (see below). For POMGNT1 complementation, pMBA40 was stably transfected into POMGNT1 knock-out HEK293T cells.

Generation of stable cell lines
Transfections were performed with the calcium phosphate method (59). In short, 5 µg of plasmid DNA in 250 mM CaCl2 were added dropwise to double strength HBSS (280 mM NaCl, 2.8 mM Na2HPO4, 50 mM HEPES, pH 7.2) while mildly vortexing and the final transfection solution was applied drop wise to cells with 40-50% confluency in 15 cm dishes. After 6 h the medium was replaced, and selection was performed with 500 μg/ml Zeocin (Invitrogen) for 4 weeks.

Patient material
The study was performed in accordance with the declaration of Helsinki and approved by the Ethics Human full-length POMGNT1 was amplified by PCR from cDNA extracted from HEK293T cells using

Heterologous protein expression and purification
For expression of EC-N-Cdh, HEK293T cells stably transfected with pMLHD7 were cultured for 72 h in serum-free medium. Upon initial centrifugation for removal of dead cells (1,000 x g, 10 min, 4 °C) culture medium from two 90% confluent 15 cm dishes was filtered through a 1.2 μm filter (Sartorius Stedim Biotech) and applied to a HisTrap column (1 ml bed volume, GE Healthcare) at a drop rate of 500 µl/min at 4 °C for o/n. The column was washed with 10 ml of wash buffer (20 mM Na3PO4, 500 mM NaCl, 5 mM imidazole, pH 7.4) and elution of bound proteins was performed in 10 fractions of 500 µl elution buffer (20 mM Na3PO4, 500 mM NaCl, 500 mM imidazole, pH 7.4). Protein concentration of eluates was determined by BCA protein assay (Pierce) and analyzed by SDS-PAGE followed by Coomassie blue staining and Western Blot.

Western Blot
Protein samples were resolved on 10% SDS PAA gels and transferred on nitrocellulose membranes according to standard protocols. After blocking and incubation with respective antibodies, proteins were detected using ECL Prime Western Blotting Detection Reagent (GE Healthcare) and imager ImageQuant LAS 500 (GE Healthcare). Protein levels were normalized to the loading control using ImageJ software

Cell proliferation and migration assay
Cell proliferation and migration were monitored by the xCELLigence system (ACEA Biosciences) according to manufacturer's instructions. The xCELLigence system is a real-time, non-labeled, impedancebased cell analysis system that allows cell proliferation and migration to be monitored in a continuous and quantitative manner.
To monitor proliferation, cells from exponential phase cultures were detached using 0.

N-and O-mannose glycoproteomics of N-cadherin
Sample preparation, measurement and glycoproteomic analysis was conducted as previously described Glycopeptide mass spectra were analyzed using glyXtoolMS (in-house developed software) (64), as well as Byonic and Byologic (both from Protein Metrics). All MS glycoproteomics raw data have been deposited to the MassIVE repository under the dataset identifier MSV000085243.

N-glycome analysis of N-cadherin
N-glycan structures were analyzed by xCGE-LIF, as previously described with slight modifications (65)(66)(67). Briefly, proteins were linearized and disulfide bonds were reduced using SDS and DTT for 10 min at 60 °C. N-glycans were released from protein backbone by PNGase F incubation (Sigma Aldrich) for 12 h at 37 °C. Released N-glycans were labeled with 8-aminopyrene-1,3,6-trisulfonic acid (APTS, Sigma Aldrich), followed by a sample clean-up step (HILIC SPE) using glyXbeads™ (glyXera) to remove excess label, salt and other impurities. The purified labeled N-glycans were analyzed by xCGE-LIF. Using the glycoanalysis system glyXbox™ (incl. kits, software (glyXtool™ v5.3.1) and database (glyXbase™); glyXera), the generated N-glycan electropherogram data (migration times) were normalized to an internal standard, resulting in a so-called "N-glycan fingerprint" with highly reproducible standardized migration time units. Subsequently, glyXtool™ was utilized for automated peak picking, integration and relative Only qRT-PCR reactions with efficiencies ranging from 0.9 to 1.1 were further analyzed. Gene expression was normalized to expression of the housekeeping gene hypoxanthine-guanine phosphoribosyltransferase (HPRT). For calculation of relative gene expression, the standard curve-based method was used. The primers are provided in Table S1.

nCounter gene expression profiling
Total RNA was isolated from 1.0 x 10 6 HEK293T cells by using the RNeasy Mini Kit (Qiagen) in combination with the QIAshredder system (Qiagen) according to the manufacturer's protocol. 50 ng of total RNA were used per hybridization reaction to determine the transcript levels of Golgi glycosyltransferases.
Further processing was conducted at the nCounter Core Facility Heidelberg using the nCounter SPRINT system and nCounter Elements chemistry as described (68). Probe design for the glycosyltransferase genes and reference genes are given on request. Data normalization was performed with the nSolver Analysis Software 3.0 (nanoString Technologies).

Human phospho-kinase and MMP array
Phosphorylation      WT or ∆POMGnT1 cells that were deprived of FBS, were incubated with an anti-N-Cdh antibody and subsequently seeded on an anti-N-Cdh antibody-treated confluent monolayer of cells. Anti-human IgG antibody served as a control. Cells were allowed to adhere for 20 min at 37 °C. Adherent cells were stained with crystal violet and quantified at OD600 after stain extraction. The absorbance of each well to which cells were added was normalized against the mean absorbance of wells, where no cells were added. Respective cell-cell adhesion is represented as relative adhesion in %, considering relative adhesion of WT to WT cells pretreated with anti-human IgG antibody as 100%. Assays were performed in triplicate from two independent experiments. All data are presented as means ± SD. Asterisks denote statistical significance in comparison to WT cells: * p ≤ 0.05, *** p ≤ 0.001, n.s. not significant.