Structural Characterization of Glycoprotein Glycans and Glycosaminoglycans of Brain Tissues in Slc35a3-Knockout Mice

TL;DR

This study examines how a genetic knockout in mice affects brain glycosylation, revealing changes in complex sugar structures linked to neurological disorders.

Contribution

The study provides new insights into how SLC35A3 deficiency alters neural glycan structures, linking it to a neurological disorder in humans.

Findings

Knockout mouse brains showed reduced complex N-glycans and increased high-mannose species.

O-glycan core-2-type species decreased, while disialyl core-1 remained stable.

Total glycosaminoglycan output was preserved despite SLC35A3 deficiency.

Abstract

Glycosylation depends on luminal nucleotide sugars delivered by solute carrier 35 (SLC35) transporters. SLC35A3 is a uridine diphosphate N-acetylglucosamine (UDP-GlcNAc) transporter. In humans, biallelic mutations in SLC35A3 cause arthrogryposis, mental retardation, and seizures (AMRS). To define how loss of SLC35A3 function reshapes the neural glycome, we profiled N-, O-, and glycosaminoglycans (GAGs) in Slc35a3 knockout mouse brains. N- and O-glycans were analyzed by MALDI-TOF MS, and GAG disaccharides were quantified by anion-exchange HPLC. Knockout mouse brains exhibited attenuation of complex-type N-glycans with a reciprocal rise in high-mannose species, as revealed by MALDI-TOF MS profiling. In contrast, ConA lectin blotting showed no significant change, consistent with its preferential detection of mannose-rich glycans. Branching analysis revealed loss of tri- and tetra-antennary structures compared with biantennary species. O-glycan profiling showed core-2-type species (Hex2HexNAc2 backbone) decreased. The dominant disialyl core-1 remained stable. Total GAG output (chondroitin/dermatan sulfate, heparan sulfate, and hyaluronan) was preserved. These findings support a microdomain model in which SLC35A3 acts as a locally effective supplier of UDP-GlcNAc to MGAT4 (branching N-acetylglucosaminyltransferase that installs the β1,4-GlcNAc arm) in the brain, while alternative routes buffer UDP-GlcNAc delivery for GAG and mucin-type O-glycan biosynthesis. Accordingly, AMRS may be attributed to impaired higher-order N-glycan branching in the brain.

Figures

Full text

1. Introduction

Glycosylation depends on a continuous luminal supply of nucleotide sugars to the endoplasmic reticulum and Golgi apparatus. This transport is mediated by the solute carrier 35 (SLC35) family of nucleotide-sugar transporters (NSTs), which shuttle activated sugars from the cytosol into secretory-pathway organelles to support protein and lipid glycosylation as well as glycosaminoglycan (GAG) biosynthesis [1,2,3,4]. Among NSTs, SLC35A3 functions as an antiporter, exchanging luminal UMP for cytosolic UDP-GlcNAc, thereby supplying UDP-GlcNAc to the Golgi lumen [1,4,5]. This transport activity supports multiple GlcNAc-dependent biosynthetic processes, including N-glycan branching and GAG chain assembly [6,7,8,9]. Human SLC35A3 maps to chromosome 1p21.2 and comprises 11 exons, producing multiple Golgi-enriched isoforms [4]. In addition to simple substrate delivery, SLC35A3 participates in higher-order assemblies within the Golgi—proximity-based assays and imaging have demonstrated that SLC35A3 associates with the UDP-galactose transporter SLC35A2 and with mannoside N-acetylglucosaminyltransferases (MGAT1/2/4B/5), consistent with a transporter–glycosyltransferase organization that increases the efficiency of complex N-glycan elaboration [9,10]. Although SLC35A3 is a major portal for UDP-GlcNAc, transporter redundancy and division of labor—e.g., through SLC35D2 and the recently identified UDP-GlcNAc transporter TMEM241—suggest that Golgi UDP-GlcNAc homeostasis is regulated at multiple levels and is tissue- or pathway-specific [11,12,13].

Clinically and biologically, SLC35A3 first gained prominence through complex vertebral malformation—an autosomal recessive skeletal dysplasia caused by a missense mutation in bovine SLC35A3—that leads to severe vertebral or limb anomalies and high perinatal mortality in Holstein cattle [14,15,16,17]. Consistently, Slc35a3 knockout (KO) mice recapitulated the CMV-like phenotypes [18]. In humans, biallelic SLC35A3 variants underlie arthrogryposis, mental retardation, and seizures (AMRS) (also termed SLC35A3–congenital disorder of glycosylation; CDG). Taken together, these observations point to species-dependent liabilities associated with SLC35A3 deficiency.

Vertebral patterning defects may reflect impaired Notch receptor O-fucose elongation (Fuc–GlcNAc–Gal–Sia), which depends on the addition of UDP-GlcNAc by Fringe in the Golgi [19,20], whereas dwarfism or skeletal dysplasia likely involves reduced chondroitin sulfate on aggrecan in cartilage [21,22]. Epilepsy is a recurrent feature in SLC35A3-CDG [23]. N-glycan branching stabilizes neuronal receptors through galectin-mediated lattices, whereas core-2 O-glycans support glycan-dependent interactions of mucin-domain proteins and adhesion molecules [24,25]. Based on these observations, we hypothesize that alterations in brain glycosylation caused by SLC35A3 deficiency contribute to seizure susceptibility.

At the cellular level, perturbing SLC35A3 limits complex N-glycan branching—particularly tri- and tetra-antennary structures—consistent with restricted UDP-GlcNAc availability to MGAT enzymes and/or physical coupling between SLC35A3 and branching glycosyltransferases [7,26]. However, context-dependent compensation has been reported, highlighting the possibility of alternative UDP-GlcNAc routes in certain cell types [27]. We previously generated Slc35a3-KO mice using CRISPR/Cas9 to investigate SLC35A3 function in vivo [18]. The mutants were perinatally lethal and exhibited chondrodysplasia with complex vertebral malformation-like vertebral abnormalities; in the spine and limb, levels of heparan sulfate (HS), keratan sulfate, and chondroitin/dermatan sulfate (CS/DS) were reduced, implicating SLC35A3 in GAG biosynthesis of skeletal tissues [18]. These observations motivate a focused analysis of the brain glycome, where N-/O-glycans and proteoglycans robustly regulate neural development, excitability, and circuit function. Here, we systematically profile N-glycans, mucin-type O-glycans, and GAGs in brain tissue from Slc35a3-KO mice to define the glycan classes most sensitive to SLC35A3 loss, assess how transporter–enzyme coupling and transporter redundancy shape UDP-GlcNAc partitioning in neural Golgi, and link altered glycosylation to SLC35A3-related neuroskeletal phenotypes.

2. Results

2.1. N-Glycan Profiling

Comparative MALDI-TOF analysis of brain N-glycans showed a broad reduction of complex-type signals in whole brain tissue from Slc35a3-KO mice, with a relative increase in high-mannose peaks (Figure 1a). Within the highlighted m/z window, complex-type peaks were markedly attenuated in Slc35a3-KO mice, whereas heterozygotes displayed intermediate profiles. Classwise quantification corroborated these trends, revealing decreases in neutral complex glycans in Slc35a3-KO mice, accompanied by increases in high-mannose species while showing no appreciable changes in sialylated complex glycans and paucimannose species (Figure 1b, Supplementary Table S1). Branching analysis demonstrated a marked reduction in higher-branched structures (Core + HexNAc3/4; tri- and tetra-antennary) in Slc35a3-KO mice (Figure 1c). In parallel, species corresponding to Core + HexNAc0/1, which largely represent high-mannose, paucimannose, and hybrid glycans, were relatively increased. Notably, the effect observed in Slc35a3-KO mice was greater for tri- and tetra-antennary than for biantennary structures in the brain, indicating a preferential loss of higher-order branching.

2.2. O-Glycan Profiling

The most abundant O-glycan, Hex_1_HexNAc_1_NeuAc_2_ (e.g., disialyl core-1, NeuAcα2-3Galβ1-3(NeuAcα2-6)GalNAc), showed no material difference among groups (Figure 2). However, glycans containing a Hex_2_HexNAc_2_ backbone—indicative of core-2-type structures (e.g., Galβ1-3(GlcNAcβ1-6)GalNAc and their sialylated extensions, such as NeuAc_1_Hex_2_HexNAc_2_ and NeuAc_2_Hex_2_HexNAc_2_)—decreased in whole brain tissue from Slc35a3-KO mice, whereas other major O-glycan species did not exhibit a systematic shift (Figure 2c, Supplementary Table S2). The profiles of O-glycan in whole brain tissue from heterozygote mice were generally intermediate or overlapped with those from wild-type mice.

2.3. Lectin Blotting Analysis of N- and O-Glycans in Brain Tissue Lysates

To validate the N- and O-glycan structures revealed by MALDI profiling, lectin blotting was performed using brain tissue lysates from wild-type and Slc35a3-KO mice. For N-glycan analysis, concanavalin A (ConA), which preferentially recognizes high-mannose-type N-glycans rather than complex-type N-glycans [28,29], was used. ConA signals tended to be higher in Slc35a3-KO brain lysates than in WT controls. Treatment of the lysates with PNGase F and Endo H abolished these signals (Figure 3), confirming that the observed lectin reactivity was derived from N-linked glycans. These results were in good agreement with the MALDI profiling data (Figure 1), supporting a relative increase in high-mannose-type N-glycans in the Slc35a3-KO brain.

In contrast, O-glycan analysis using peanut agglutinin (PNA), which recognizes Galβ1–3GalNAc structures, revealed no apparent difference in signal intensity between wild-type and KO brains under untreated conditions. Following sialidase treatment, PNA signals were markedly increased in both wild-type and Slc35a3-KO brain lysates (Figure 3c,d). These findings are consistent with the MALDI profiling data (Figure 2), indicating that a large proportion of O-glycans possess a sialylated Hex–HexNAc backbone in both genotypes.

2.4. GAG Disaccharide Quantification

The Slc35a3-KO brain contained 85% of the total CS/DS disaccharides found in the heterozygote brain (Table 1). Hyaluronan (HA) was at 101% (Table 2) and HS was at 94% (Table 3). However, these differences were insignificant (Figure 4). Thus, the loss of Slc35a3 did not measurably reduce bulk GAG output in the brain, in contrast to the clear reduction in N-glycan branching observed in the same cohort (Figure 1), highlighting class-specific sensitivity to UDP-GlcNAc availability.

3. Discussion

In brain tissue, loss of Slc35a3 selectively affected specific classes of glycans. Bulk GAG outputs (CS/DS, HA, and HS) and core-1 O-glycans remained largely unchanged, whereas tri- and tetra-antennary N-glycans and core-2 O-glycans were significantly reduced. These findings are consistent with in vitro studies indicating that SLC35A3 is not the sole route for UDP-GlcNAc import and that additional UDP-GlcNAc delivery mechanisms operate within the Golgi [7,8,27]. SLC35A3, SLC35D2, and TMEM241 have emerged as SLC35 family members that transport UDP-GlcNAc [4,7,11,12]. TMEM241 mediates mannose-6-phosphate tagging of lysosomal proteins. Notably, Tmem241-KO mice develop cholesterol accumulation in pulmonary cells and pulmonary injury accompanied by reduced locomotor activity—phenotypes clearly distinct from those caused by SLC35A3 deficiency [6]. These contrasts suggest cell- and tissue-specific specialization across UDP-GlcNAc transporters. To clarify the role of TMEM241 in the nervous system, glycomic profiling of Tmem241-KO brain will be informative. By comparison, a comprehensive phenotype for Slc35d2-KO mice has not been reported [1,2,4]. Given that many SLC35 family members are “orphans” with unidentified substrates, additional UDP-GlcNAc transporters probably remain undiscovered. A full accounting of UDP-GlcNAc transport in vivo will therefore require identification of all transporters that contribute to luminal supply, together with a systematic analysis of single- and compound-KO mice. In this context, interpretation of protein-level data for individual transporters requires careful consideration. It should be noted that SLC35A3 protein was not detectable by Western blot analysis in embryonic mouse brain samples under the experimental conditions used in this study. Importantly, the Slc35a3-KO mouse line carries a premature stop codon within the Slc35a3 coding sequence, which is expected to prevent production of the full-length SLC35A3 protein [18].

In addition to these global trends, O-glycomics revealed that core-2-type species (Hex_2_HexNAc_2_ backbone; e.g., Galβ1-3(GlcNAcβ1-6)GalNAc and sialylated extensions) decreased in Slc35a3-KO brain, whereas the most abundant disialyl core-1 species (Hex_1_HexNAc_1_NeuAc_2_; e.g., NeuAcα2-3Galβ1-3(NeuAcα2-6)GalNAc) were unchanged. Because core-2 branching is catalyzed by C2GnT/GCNT1 (addition of GlcNAcβ1-6 to core-1 using UDP-GlcNAc), this pattern suggests local substrate limitation in select Golgi microdomains [24,25]. The higher-order N-glycan branching appears preferentially vulnerable in the brain. Studies have reported transporter–enzyme microdomains in which NSTs, including SLC35A3, colocalize or associate with MGATs within the Golgi, thereby enabling local UDP-GlcNAc delivery to branching enzymes [9,10]. In particular, SLC35A3 was shown to interact with MGAT4 (GnT-IV) but not MGAT5 (GnT-V) by co-immunoprecipitation, consistent with our finding that tri- and tetra-antennary structures are most compromised in Slc35a3-KO brain [8]. Within this framework, removing SLC35A3 would constrain local substrate availability to MGAT4 even when global UDP-GlcNAc remains adequate for earlier steps (MGAT1/2), producing the selective depletion of higher-branch structures we observed [8,9,10]. Reduced higher-order branching is predicted to weaken the galectin lattice, shortening receptor surface half-life and shifting neuronal signaling thresholds, whereas diminished core-2 structures may compromise mucin-domain scaffolds and synaptic adhesion—together providing a plausible route to network hyperexcitability and seizures in AMRS/SLC35A3-CDG [23,24,25,30,31].

In the skeletal tissues of Slc35a3-KO mice, GAG output decreased [18], whereas in the brain, it remains near normal despite selective loss of tri- and tetra-antennary N-glycans. This divergence can be rationalized by two nonmutually exclusive explanations. First, compensatory UDP-GlcNAc transport appears to be more effective in the brain than in cartilage, such that alternative transporters can sustain GAG biosynthesis when SLC35A3 is absent, whereas redundancy is insufficient in the skeletal system. Second, tissue-specific transporter–enzyme organization may differentially prioritize UDP-GlcNAc delivery as follows: in the cartilage, SLC35A3-containing assemblies could preferentially channel substrate to GAG-synthetic machinery, making GAG output acutely sensitive to SLC35A3 loss, whereas in the brain, such coupling is less dominant and other transporters can maintain GAG backbones. These hypotheses highlight the possibility that SLC35A3–glycosyltransferase complexes are tissue-restricted and underlie differences in glycan vulnerability across organs.

Collectively, our findings support a model in which SLC35A3 is not a universal gatekeeper of UDP-GlcNAc entry into the secretory pathway, but rather a locally acting supplier that preferentially fuels MGAT4-driven N-glycan branching and, by extension, GCNT1-dependent core-2 O-glycan branching in the brain. Such localized substrate delivery is well placed to stabilize receptor glycophenotypes that influence the excitatory–inhibitory balance. Thus, reductions in higher-order N-glycan branching and core-2 O-glycan branching may jointly contribute to the neurological features of SLC35A3-CDG. Because germline Slc35a3-KO mice are perinatally lethal, testing this link will require brain-specific conditional KO mice coupled to region-resolved glycomics and physiology.

SLC35A3 associates with the UDP-galactose transporter SLC35A2, and cooperative interactions between NSTs have been proposed [10,27,32,33]. Understanding tissue-specific SLC35A3 function will therefore benefit from mapping the composition of SLC35A3-containing complexes across organs and from systematic analysis of single- and compound-NST KO mice to delineate its complementary and compensatory roles in vivo.

4. Materials and Methods

4.1. Mice and Ethical Statement

Slc35a3-KO mice have been developed previously [18]. The mice were housed in a temperature-controlled room under a 12/12-h light/dark cycle. They were fed with standard mouse laboratory chow and had free access to water. The mice were euthanized using an overdose of pentobarbital or cervical dislocation under isoflurane anesthesia. This study was conducted in strict accordance with the guidelines for the Proper Conduct of Animal Experiments (Science Council of Japan). The mouse embryos were euthanized by decapitation.

Brain samples used in this study were collected at the embryonic stage. At this developmental stage, the sex of the embryos could not be reliably determined and was not recorded at the time of collection. Therefore, the samples were not stratified by sex in the subsequent analyses, and all data represent pooled embryonic samples irrespective of sex.

All animal experiments were approved by the Animal Experimentation Committee at Iwate University (approval nos. A202047 and A202328). This study was conducted in strict accordance with the guidelines for the Proper Conduct of Animal Experiments (Science Council of Japan). All efforts were made to minimize animal suffering.

4.2. N-Glycosylation Analyses by MALDI-TOF-MS

Brain tissue lysates (corresponding to 100 μg of protein) from embryonic day 18.5 (E18.5) mouse embryos were subjected to reductive alkylation using TCEP and iodoacetamide, as described [34]. The protein precipitates were digested first with trypsin and then with PNGase F (Roche, Basel, Switzerland). The released N-glycans were captured and purified using BlotGlyco beads (Sumitomo Bakelite, Tokyo, Japan), followed by sialic acid linkage-specific ester-to-amide derivatization using the LEAD-SALSA method [35]. Briefly, N-glycans with MTT-derivatized sialic acids were converted using propylamine. Excess reagents were removed by HILIC. The purified N-glycans were measured using Autflex speed and UltrafleXtreme MALDI-TOF MS (Bruker, Billerica, MA, USA).

4.3. O-Glycosylation Analyses by MALDI-TOF-MS

For O-glycan analysis, the protein precipitates were directly subjected to in-solution SALSA for sialic acid linkage-specific derivatization. SALSA derivatized O-glycans were cleaved using mild β-elimination in the presence of the pyrazolone analogs (BEP) method, as reported [36]. The released glycans were then derivatized with 1-phenyl-3-methyl-5-pyrazolone (PMP). PMP-labeled O-glycans were purified by sequential solid-phase extraction with C18 and silica and subjected to MALDI-TOF MS analysis.

4.4. Normalization and Quantification of Glycan Signals

For comparative analyses of N- and O-glycan profiles, all MALDI-TOF MS signal intensities were normalized to an exogenously added internal standard, maltotetraose (Hex5) for N-glycan and chitotetraose (GN4) for O-glycan, prior to comparison. The internal standard was added at a fixed amount to each sample before MS analysis. Quantitative comparisons were therefore based on normalized peak intensities rather than raw signal intensities, consistent with previous studies demonstrating a linear relationship between MALDI-TOF peak intensity and glycan abundance under standardized conditions [36,37].

4.5. Lectin Blotting of Brain Tissue Lysates

Brain tissue lysates were prepared using standard RIPA buffer (Thermo Scientific, Waltham, MA, USA) supplemented with protease inhibitors (Roche). Protein concentrations were determined by a BCA assay, and equal amounts of protein (20 µg for ConA blotting and 15 µg for PNA blotting per lane) were separated by SDS–PAGE and transferred onto PVDF membranes.

For analysis of N-glycan structures, membranes were probed with horseradish peroxidase (HRP)-conjugated ConA (Vector Laboratories, Newark, CA, USA), which preferentially recognizes high-mannose-type N-glycans. To verify N-glycan dependence of lectin binding, lysates were treated with PNGase F (500 U per 20 µg protein; New England Biolabs, Ipswich, MA, USA) to remove N-glycans, or with Endoglycosidase H (Endo H; 500 U per 20 µg protein; Seikagaku Kogyo Co., Ltd., Tokyo, Japan) to selectively cleave high-mannose and hybrid-type N-glycans, prior to electrophoresis.

For analysis of O-glycan structures, membranes were probed with HRP-conjugated PNA (Vector Laboratories, Newark, CA, USA), which recognizes Galβ1–3GalNAc structures. To assess the contribution of sialylation and core O-glycan structures, lysates were treated with neuraminidase (0.1 U per 20 µg protein, Nakarai Tesqe, Kyoto, Japan) to remove terminal sialic acids and/or with O-Glycosidase (5 mU per 20 µg protein; New England Biolabs, Ipswich, MA, USA) prior to SDS–PAGE and blotting. Lectin binding was detected by chemiluminescence using the HRP signal.

Signal intensities of the entire lane above 48 kDa for both ConA and PNA lectin blots were quantified using ImageJ (version 1.54k; National Institutes of Health, Bethesda, MD, USA) and normalized to the corresponding β-actin signals. Both lectin and β-actin signals were first normalized to the WT control prior to calculating relative intensities. Statistical analysis was performed using two-way ANOVA with GraphPad Prism software (version 9).

4.6. Disaccharide Composition Analysis of CS/DS, HS, and HA Isolated from Various Tissues

The levels of total disaccharides of CS/DS, HS, and hyaluronan (HA) in various tissues from Slc35a3-KO mice were determined, as described [38]. Briefly, each brain was homogenized, sonicated, and treated exhaustively with actinase E (Kaken Pharm. Kyoto, Japan) to degrade proteins. Crude GAG-peptide fractions were obtained by precipitation with 5% tricarboxylic acid. Excess reagent was removed using ether. The GAG-peptides were desalted using Amicon Ultra-4 (3 K, Millipore, Burlington, MA, USA). The samples were treated individually with a mixture of chondroitinases ABC and AC-II (Seikagaku Corp., Tokyo, Japan) or a mixture of heparinase-I, -III (IBEX Pharmaceuticals, Montreal, QC, Canada) and heparitinase-II (R&D Systems, Minneapolis, MN, USA) for analysis of the disaccharide composition of CS/DS/HA or HS, respectively. The digests were labeled with the fluorophore 2-aminobenzamide (2AB) and aliquots of the 2AB-derivatives of GAG disaccharides were analyzed by anion-exchange HPLC on a PA-G column (YMC Co., Kyoto, Japan), as described [38,39]. Unsaturated disaccharides detected in the digests were identified by comparison with the elution positions of authentic 2AB-labeled disaccharide standards.

In addition, authentic 2AB-labeled disaccharide standards were analyzed under identical HPLC conditions to generate standard curves, which were used to validate peak identification and to confirm the linearity between fluorescence intensity and disaccharide amount. The chromatograms and standard curves for the authentic disaccharides are provided in the Supplementary Figure S1.