Highly fucosylated N-glycans at the synaptic vesicle and neuronal plasma membrane

At neuronal synapses, synaptic vesicles (SVs) require glycoproteins for normal trafficking, and N-linked glycosylation is required for delivery of the major SV glycoproteins synaptophysin and SV2A to SVs. The molecular compositions of SV N-glycans, which may drive important neurobiological processes, are largely unknown. In this study, we combined organelle isolation techniques, fluorescence detection of N-glycans, and high-resolution mass spectrometry to characterize N-glycosylation at synapses and SVs from mouse brain. Detecting over 2,500 unique glycopeptides from over 550 glycoproteins, we found that abundant SV proteins harbor N-glycans with fucose on their complex antennae, and we identify a highly fucosylated N-glycan enriched in SVs as compared to synaptosomes. Antennary fucosylation was also characteristic of plasma membrane proteins and cell adhesion molecules with established roles in synaptic function and development. Our results represent the first defined N-glycoproteome of a neuronal organelle and raise new questions in the glycobiology of synaptic pruning and neuroinflammation.


INTRODUCTION
Synaptic vesicles (SVs) are small recycling organelles that store and release neurotransmitters at nerve terminals. Post-translational modifications of SV proteins may play important roles in SV formation and function (Brager et al., 2003;de Jong et al., 2016;Stewart et al., 2020;Zhang et al., 2015). Among these modifications, glycosylation of luminal asparagine residues (N-linked glycosylation) is integral to membrane protein folding and trafficking (Cherepanova et al., 2016;Reily et al., 2019), and several of the most abundant SV proteins with established roles in neurotransmission are N-glycoproteins. These include the Ca 2+ sensor synaptotagmin-1 (Bradberry et al., 2020;Geppert et al., 1994;Matthew et al., preparation containing mostly pre-and post-synaptic elements, were prepared by differential centrifugation (Fig. 1A) (Takamori et al., 2006). Alongside synaptosomes, a highly pure population of SVs was isolated by immunoprecipitation (IP) with magnetic beads conjugated in-house to a monoclonal antibody against SV2 (Buckley and Kelly, 1985), according to recently described procedures (Fig. 1A) . These preparations were subjected to proteomic analysis by trypsin digestion and nano-flow liquid chromatography-tandem mass spectrometry (nLC-MS/MS) using an Orbitrap Eclipse mass spectrometer   (Fig. 1B). The synaptosome preparation was rich in cytoskeletal and mitochondrial proteins (Fig. 1C, Supplementary Table S1), consistent with a low degree of specific enrichment for any particular cellular compartment. In contrast, SV2-IP yielded SVs of exceptionally high purity, as evinced by the nearly exclusive presence of well-established SV proteins among the top 25 most intense in this preparation (Fig. 1D, Table S1). More than 2,300 proteins were detected in this SV preparation, among which over 1700 were also identified in the synaptosome fraction ( Fig. 1E). This represents the largest number of proteins detected in a highly pure SV preparation to date Taoufiq et al., 2020). Label-free quantitation (LFQ) intensity scores of synaptosome and SV proteins were positively correlated (Fig. 1F), consistent with an expected contribution from SVs to the synaptosomal proteome. Type-A gamma-amino butyric acid (GABA A ), Nmethyl-D-aspartate (NMDA), and α -amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor subunits were present in synaptosomes but undetectable in SV samples (Fig 1G). SV samples likewise did not contain the postsynaptic scaffolding proteins PSD-95 or gephyrin, both of which were present in synaptosome samples (Fig. 1G). These results establish that our synaptosome preparation contains not just mature SVs but also a more general sample of brain proteins that represents the synaptic milieu. SVs were isolated using magnetic beads conjugated in-house to a monoclonal α -SV2 mAb, while synaptosomes were isolated by differential centrifugation. Both sample types underwent similar processing methods and identical LC-MS/MS analysis procedures. (B) Example MS 1 total ion current (TIC) chromatograms for SV and synaptosome samples. (C) The top 25 most intense proteins by labelfree quantification analysis in synaptosome samples. Cytoskeletal and mitochondrial proteins predominated, consistent with contributions from multiple subcellular components. Error bars indicate standard deviation from three biological replicates. (D) In contrast, the top 25 most intense proteins in the SV preparation are largely well-established to reside on synaptic vesicles, with only one contaminant (myelin basic protein) observed among them. (E) Venn diagram of proteins detected in synaptosome and SV samples, demonstrating the expected high degree of overlap. (F) LFQ intensities between synaptosomal and SV proteins are positively correlated (Spearman's rho = 0.28, p < 0.0001), consistent with a contribution of SVs to the synaptosomal proteome. Orange points indicate proteins listed in panel D. (G) Ionotropic glutamate and GABA receptor subunits and postsynaptic scaffolding proteins were detected in synaptosomes but largely absent from SV samples, demonstrating the purity of SVs and the inclusion of postsynaptic material in the synaptosome sample. See also Supplementary Table S1, which contains all protein-level identification and quantification data used to generate this figure.

Glycoproteomics at the synapse
With the proteomic contents of synaptosomes and SVs established, we turned our attention to the Nglycosylation of proteins in these two preparations. Several studies have established that brain N-glycans comprise mostly mannosylated glycans and fucosylated complex glycans bearing bisecting GlcNAc (Chen et al., 1998;Ji et al., 2015;Lee et al., 2020;Shimizu et al., 1993;Williams et al., 2022) (Fig. 2A).
We reasoned that synaptosome and SV purification, combined with modern glycoproteomic techniques, might yield new insights into the relationship between N-glycosylation and SV proteins. We thus performed nLC-MS/MS experiments employing glycopeptide enrichment and Orbitrap tandem mass spectrometry with either ion-dependent stepped higher-energy collisional dissociation (sceHCD)  or activated-ion electron transfer dissociation (AI-ETD) (Peters-Clarke et al., 2020;Riley et al., 2019). Synaptosome samples were characterized by both AI-ETD and sceHCD-based methods after glycopeptide enrichment using strong anion exchange-electrostatic repulsion chromatography (SAX-ERLIC) prior to nLC-MS. SV samples were characterized using sceHCD both with and without glycopeptide enrichment (Fig. 2B-E) given the lower complexity and abundance of these samples.
Importantly, SAX-ERLIC is less specific for particular N-glycans than lectin affinity enrichment methods (Totten et al., 2017), reducing the potential for bias towards certain glyocopeptides among enriched samples. In combination with the MSFragger/FragPipe analysis pipeline (Kong et al., 2017;, these data enabled a rich glycoproteomic survey of both synaptosomes and SVs purified from whole mouse brain ( Fig. 2B-E). Glycan peptide spectral matches (GlycoPSMs) were obtained by analysis with MSFragger in glyco mode using the FragPipe analysis suite (Kong et al., 2017;Polasky et al., 2022Polasky et al., , 2020 and filtered by excluding those glycoPSMs with a calculated false discovery rate cut-off (Q-value) of < 0.025. A total of over 2,500 unique glycopeptides from over 550 glycoproteins was identified by this method (Fig. 2B, Supplementary Table S2), with the majority observed in the substantially more complex synaptosome samples (Fig. 2B). The identified glycopeptides were annotated according to their degree of mannosylation or fucosylation ( Fig. 2A-K) with guidance from recently published work describing the composition of the brain N-and O-glycomes (Trinidad et al., 2013;Williams et al., 2022) (Supplementary Table S3). GlycoPSMs containing sialic acid or unusually large numbers of sugars were omitted from annotation, as we inferred that these glycoPSMs could correspond to peptides bearing both N-and O-glycosylation on the same tryptic peptide given the low prevalence of sialylated N-linked glycans, and high prevalence of sialylated O-linked glycans, in brain (Williams et al., 2022). Despite these limitations, approximately 80% of glycoPSMs were annotated. Distributions of glycoPSMs, corresponding to major N-glycan types showin in Fig. 2A, are shown in Fig. 2F-H. In agreement with the work of Williams et al. (2022), we generally observed a predominance of oligomannose species along with fucosylated complex species (Fig. 2F-H), with a minor contribution from glycans bearing two or more fucoses. Proteins bearing three fucoses (Fuc3) were not observed in SV samples without glycopeptide enrichment (Fig. 2H), likely due in part to poorer positive-mode ionization efficiency of peptides bearing larger glycans.
While informative, the distribution of glycoPSMs may not reflect the actual composition of N-glycans in SVs or synaptosomes, as these data do not account for differences in abundance among glycoproteins giving rise to these glycopeptides. We thus analyzed these data further by considering the intensity scores obtained in our standard proteomics experiments ( Fig. 1) for each identified glycoprotein ( Fig. 2I-K).
Glycoproteins were grouped into quintiles based on their intensity values (Fig. 1), and the distribution of glycoPSMs was analyzed for each quintile in each sample type ( Fig. 2I-K). This analysis revealed a striking bias toward fucosylation in the most abundant contributors to the SV proteome ( Fig. 2J-K), which was less pronounced in the synaptosome samples (Fig. 2I). This trend was particularly evident for glycans containing two or three fucoses (i.e., highly fucosylated glycans) ( Fig. 2I-K). In glycan-enriched SV samples, nearly all the unique glycoPSMs for the most abundant SV proteins contained one or more fucoses (Fig. 2J). A bias towards fucosylated glycopeptides in the most abundant SV proteins was observed whether or not glycopeptide enrichment was used ( Fig. 2J-K), demonstrating that this observation is not an artifact of the enrichment procedure. However, a stark increase in high fucosylation in abundant SV proteins was best observed using glycopeptide enrichment ( Fig. 2J) Together, these results suggest that abundant SV proteins are rich in highly fucosylated N-glycans.

Characterization of SV and synaptosome glycans by HILIC-HPLC
While the above results provide evidence that abundant SV glycoproteins are biased toward fucosylation, additional caveats exist to the interpretation of bottom-up glycoproteomics data. For example, unexpected biases in the positive-mode ionization efficiency of certain glycopeptides could cause skewed results (Harvey, 1993). We thus employed an orthogonal, ionization-independent N-glycan analysis method involving fluorescent detection of enzymatically cleaved glycans from SV and synaptosome samples ( Fig. 3A). N-glycans were specifically and quantitatively removed from proteins with PNGase F (Supplementary Figure S1), followed by labeling with the fluorescent molecule procainamide via reductive amination (Fig. 3A). Labeled N-glycans were analyzed by amide hydrophilic interaction chromatography with fluorescence detection (HILIC HPLC, Fig. 3). Digestion with exoglycosidases specific for mannose, antennary fucose, or galactose (Fig. 3B) enabled a semi-quantitative determination of the contributions of various glycans to the SV and synaptosomal N-glycomes ( Fig. 3C-L). The identities of the mannosylated glycans were determined using a man5 standard and glucose homopolymer ladder, defining a clear sequence of glycans bearing 5-9 mannose residues (Supplementary Fig. S2).
Among mannosylated glycans, we observed a trend towards predominance of Man5 in both synaptosomes and SVs (Fig. 3I), but we did not observe significant differences between the two sample types either in the distribution of mannosylated glycans (Fig. 3I) or the total contribution from mannosylated glycans (Fig. 3J).
In accordance with our LC-MS glycoproteomics data (Fig. 2), both SV and synaptosomal samples contained a major contribution from antennary fucosylated glycans (Fig. 3D, G, K), equal to ~25% of all N-linked glycans. Strikingly, a late-eluting antennary fucosylated species bearing two antennary fucose residues, denoted φ , was significantly enriched in the SV samples (Fig. 3D,G; Fig. 3L). This peak was susceptible to further cleavage when incubated with α (1-2,4,6) fucosidase ( Supplementary Fig. S3), demonstrating that this species is also core fucosylated and thus contains 3 total fucoses. This peak was also susceptible to cleavage with β (1-3,4) galactosidase, which caused another ~2 glucose unit shift in the elution time of this peak (Fig. 3H). The resulting peak was sensitive to cleavage with β -GlcNAcase (Fig.   S3). Examination of our glycoproteomics data (Fig. 2, Table S2) demonstrated a predominant Fuc3 species, HexNAc 5 Hex 5 Fuc 3 , on the abundant SV glycoproteins SV2B and synaptophysin in glycopeptideenriched SV samples. Several potential structures for φ , informed by our mass spectrometry data, HPLC data with enzyme digestion, and previous characterizations of the mouse brain N-glycome (Helm et al., 2021;Williams et al., 2022), are given in Fig. 3M. These structures represent complex N-glycans bearing up to two Lewis X moieties on their antennae. We did not observe sensitivity to α (1-2) fucosidase among synaptosome or SV glycans (Supplementary Fig. S4), consistent with the absence of the corresponding fucosyltransferase enzymes Fut1 and Fut2 in mouse brain (Williams et al., 2022). Finally, we ensured that our findings were not due to elution of SV2 mAb from the magnetic beads, as these SV2 mAb glycans eluted at different times and were not sensitive to α (1-3,4) fucosidase or α (1-2,4,6) mannosidase (Fig.   S4). These results confirm that both SVs and synaptosomes are rich in antennary-fucosylated glycans, with specific enrichment of at least one highly fucosylated species in SVs. φ was also subject to further cleavage by α (1-2,4,6) fucosidase, demonstrating core fucosylation (Supplementary Figure 3) (I-K) The distribution of peak areas among mannosylated glycans, as well as the total contribution from mannosylated and antennary fucosylated glycans, was equivalent in synaptosomes and SVs. (L) The peak corresponding to highly fucosylated glycan φ made a significantly larger contribution to SV versus synaptosome glycans (p = 0.02, Welch's test), demonstrating enrichment of this highly fucosylated glycan in SVs. (M) Possible structures for glycan φ compatible with HPLC, glycoproteomics, and published brain N-glycome studies (Williams et al., 2022). Dashed box indicates a species previously predicted by Williams et al. (2022). See also Supplementary Figures S2-S4, which contain additional raw HPLC data used to establish identity of Man5-9 and φ .

Deep characterization of the synaptic vesicle glycoproteome
Previous work has shown that N-glycosylation can be remarkably heterogeneous not only across glycosites for a given protein, but also at any given glycosite (Riley et al., 2019). Moreover, while the above results support the enrichment of at least one glycan with antennary fucosylation on SV proteins, SVs still contain a substantial proportion of oligomannose glycans (Fig. 2, 3). A deeper examination of the glycosylation sites on each major SV glycoprotein would yield additional insights into the nature and distribution of protein glycosylation in SVs. We thus immunoprecipitated SV2 and syt1 from detergentsolubilized synaptosomes and analyzed these samples by glycan-targeted LC-MS to characterize Nglycosylation more thoroughly for these major SV glycoproteins ( Table S2). The resulting data were combined with our SV glycoproteomics data (Fig. 2, Table S2) to define the set of unique glycopeptides present at each glycosite in each major SV glycoprotein (Fig. 4). In accordance with the work of Riley et al. (2019), we found that each glycosite could carry any of several unique glycans, and heterogeneity across glycosites was observed for each protein (Fig. 4). Strikingly, while high fucosylation was observed on each protein except for syt1, high fucosylation was usually observed on only one glycosylation site per protein (Fig. 4). In the case of synaptophysin, which harbors a single glycosylation site, only fucosylated glycans were detected. In SV proteins with multiple glycosylation sites, especially SV2B, nonfucosylated glycosylation sites harbored oligomannose or high-mannose glycans (Fig. 4). We did not observe high molecular weight forms of SV2 previously characterized by immunoblotting as keratan sulfate proteoglycans (Scranton et al., 1993) (Supplementary Fig. S5). Rather, this earlier result may have represented an experimental artifact, because we found that boiling SV or synaptosome samples caused an apparent increase in the molecular weight of SV2 (Fig. S5). Thy1, which is among the most abundant neuronal plasma membrane glycoproteins and is also found in SVs Jeng et al., 1998;Morris, 2018), contained fucosylation at all three glycosites but was likewise biased toward fucosylation at a single site, N94 (Fig. 4). By contrast, syt1 was unique in its lack of fucosylation (Fig. 4).
Indeed, the single N-glycosylation site on syt1 was observed to contain at least six mannose residues, even though Man5 was the most common mannosylated glycan observed in each sample (Fig. 2, Fig. 3) and in the brain generally (Williams et al., 2022). Syt1 is also unique among these proteins in lacking a predicted globular luminal domain. Together, these results paint a highly resolved picture of the SV glycoproteome, raise new questions about the trafficking itineraries of SV proteins, and suggest important biochemical constraints on protein glycosylation in Golgi apparatus.  (Jumper et al., 2021) and symbols corresponding to unique Nglycopeptides detected for each N-linked glycosylation site. Sites are noted by their amino acid sequence number for each glycosylated asparagine. High fucosylation was observed for all SV glycoproteins shown here except for syt1. SV glycoproteins with multiple glycosylation sites demonstrated a bias towards fucosylation at a single site, while other sites were more likely to retain less mature mannosylated Nglycans. No glycoPSMs were observed for peptides containing N548 on SV2A. See also Supplementary  Table S2, which contains the glycoPSM data used to generate this figure; and Supplementary Figure S5, which addresses prior studies on the glycosylation of SV2.

Fucosylated N-glycans are characteristic of synaptic vesicle and plasma membrane proteins
We conducted further analyses of our glycoproteomics data to better define the biological context for antennary fucosylation at central synapses. Among unique glycopeptides found in SVs, the majority (164/235) were also found in synaptosome samples (Fig. 5A), consistent with the substantial observed overlap between the contents of these samples (Fig. 1E-F). A Venn diagram describing the distribution of all proteins with detected N-glycans is shown in Fig. 5B. As expected, the majority of the glycoproteins with annotated glycoPSMs in this study were mannosylated (Fig. 2), with a relatively smaller proportion of proteins observed with singly fucosylated and highly fucosylated complex N-glycans (Fig. 5B). In agreement with previous large-scale glycoproteomic studies of mouse brain (Riley et al., 2019), we observed that many proteins contained both mannosylated and fucosylated N-glycans (Fig. 5B).
Gene ontology (GO) analysis demonstrated that cell adhesion was among the most enriched biological process terms in highly fucosylated proteins (Fig. 5D). Other highly enriched processes included several that are essential for synaptic development, including neurogenesis, axonogenesis, and chemotaxis (Fig.   5D). There was substantial overlap in top-ranked GO enrichment terms between singly fucosylated and highly fucosylated proteins (Fig. 5D), while proteins that were not fucosylated demonstrated substantially lower enrichment of cell adhesion-related processes (Fig. 5D). All ranked GO terms are shown in Supplementary Table S4. While a previous study using lectin affinity enrichment implied a connection between fucose-α(1,2)-galactose and a similar set of protein functions in mouse olfactory bulb (Murrey et al., 2009), we are unsure how to interpret those results given evidence that the mouse brain is largely devoid of fucose-α(1,2)-galactose and the enzymes that catalyze its addition (Murrey et al., 2009;Williams et al., 2022) (Fig. S4). Thus, while roles for fucosylated glycans in neurodevelopment and cell migration have previously been proposed (Gouveia et al., 2012;Kudo et al., 2007;Murrey et al., 2009), our work provides foundational evidence of links among protein fucosylation, trafficking to SVs and the plasma membrane, and cell adhesion-related processes in the mammalian brain. While some proteins were found with only one subtype of N-glycosylation, many proteins contained both mannosylated and fucosylated N-glycans. (C) The top 11 proteins by LFQ intensity with Uniprot-annotated N-glycosylation sites in synaptosomes and SVs that were found to bear antennary fucose, as defined by the presence of at least 2 fucoses. Antennary fucosylation was observed on almost all of the most abundant SV proteins along with a number of cell adhesion proteins with roles in synaptic development. (D) Enrichment for gene ontology (GO) biological process terms in nonfucosylated, singly fucosylated, and highly fucosylated proteins. GO terms were ranked by inverse pvalue for enrichment, and the top 10 terms enriched in highly fucosylated proteins are shown. While substantial agreement is observed between singly fucosylated and highly fucosylated proteins, mannosylated proteins demonstrate substantially less enrichment of cell adhesion processes. See also Supplementary Table S4, containing all glycoprotein glycosylation assignments and GO biological process enrichment data used to generate this figure.

DISCUSSION
The present study demonstrates the power of combining stringent organelle purification techniques and analytical chemical methods to answer-and generate-specific questions in neurobiology. Our proteomics results (Fig. 1) represent coverage of the SV proteome at unprecedented depth, providing a valuable resource for investigators and extending the body of evidence supporting the use of modern IP techniques for SV purification . The combination of SV IP, glycan-focused mass spectrometry, and fluorescence methods enabled the first direct connection, to our knowledge, between a specific N-glycan and a brain-specific organelle (Figs. 2-4). We note that a prior study (Riley et al., 2019) reported a larger number of glycopeptides using whole mouse brain, lectin affinity chromatography, offline fractionation, and activated-ion electron transfer dissociation (AI-ETD). However, a comparison between the present study and the work of Riley et al. (2019) demonstrates the utility of our focused approach using SV protein purification and SAX-ERLIC, which represents a compromise of breadth for depth at the synapse. For example, MSFragger-Glyco analysis of the dataset from Riley et al. yields no glycoPSMs for synaptophysin or synaptoporin, and the glycoPSMs found for SV2 isoforms contain at most one fucose (Supplementary Table S5). By contrast, we found 15 unique glycoforms for synaptophysin alone, many of which were highly fucosylated (Fig. 4). Moreover, while previous studies have identified brain glycans with high degrees of antennary fucosylation corresponding to Lewis X or possibly fucose-α(1,2)-galactose (Chen et al., 1998;Kudo et al., 2007;Murrey et al., 2009;Trinidad et al., 2013;Williams et al., 2022), this study represents the first effort to systematically define the subcellular context for proteins bearing antennary fucosylation across the mouse brain. As with other LC-MS glycoproteomic studies (Riley et al., 2019;Trinidad et al., 2013), we do not address the large, polyanionic proteoglycans associated with the extracellular matrix and perineuronal nets (Dani and Broadie, 2012;Testa et al., 2019), which are comparatively poor analytes for positive-mode mass spectrometry.
Future application of negative ionization and fragmentation approaches, such as negative electron transfer dissociation (e.g., AI-NETD) (Riley et al., 2015), may enable the characterization of glycans not detected in the present study.
Our deep profiling of the SV N-glycoproteome raises several questions about the processing of N-glycans on proteins that undergo antennary fucosylation. Strikingly, SV proteins contained a single preferred site for fucosylation, while other sites were dominated by less mature mannosylated glycoforms (Fig. 4).
These results are consistent with our HPLC results, which demonstrate that SVs still contain a substantial proportion of mannosylated glycans despite being enriched for antennary fucosylated structures (Fig. 3).
Cleavage of mannose residues, and subsequent addition of GlcNAc and antennary fucose, thus appears to be site-specific. The basis for this site selectivity is unclear, though differential accessibility of glycan sites to glycan processing enzymes may play a role (Lee et al., 2014). Because only a single glycosylation site tends to undergo maturation in each SV protein (Fig. 4), it is tempting to speculate that glycan processing may be a rate-limiting step in the progress of some SV proteins through the Golgi. Examples of such glycan-dependent sorting pathways exist, most notably involving ER quality control mechanisms (Cherepanova et al., 2016) and mannose-6-phosphate receptors that recognize this specific glycan to recruit lysosomal proteins (Ghosh et al., 2003). While an oligomannose-, GlcNAc-or fucose-dependent Golgi transport lectin would provide an explanation for the limited maturation of all but one glycosite, more work is needed to address this hypothesis. We note that this site-specific maturation bias was less pronounced on highly abundant plasma membrane proteins such as Thy1 (Fig. 4) and Na + /K + -ATPase subunit β 1 ( Table S2) (Riley et al., 2019). Top-down analyses of glycosylation across multiple sites on intact proteins may clarify the nature of the heterogeneity observed in our bottom-up studies.
The absence of fucosylation on syt1 (Fig. 4), a highly abundant SV glycoprotein, raises yet more questions. At least two potential explanations exist for this divergence from other SV glycoproteins: the luminal sequence of syt1, which lacks a globular domain, may disfavor certain glycan processing steps; or syt1 may not spend enough time in the Golgi compartments containing the requisite enzymes and metabolites. We note that, unlike the case for synaptophysin and SV2, trafficking of syt1 does not require N-glycosylation (Kwon and Chapman, 2012) but does require at least one C2 domain (Courtney et al., 2019), suggesting that syt1 sorts to SVs via a distinct molecular recognition process. Further experiments combining glycosylation site mutagenesis, enzyme manipulation, microscopy, and glycoproteomics may better define the site-specific determinants of SV protein glycosylation and their role in protein trafficking.
The enrichment of antennary fucosylated glycans on SV proteins, plasma membrane transporters and cell adhesion proteins (Figs. 2-5) is a key finding that merits further investigation. While antennary fucosylated glycans including Lewis X have long been described as neuronal surface markers (Capela and Temple, 2002;Götz et al., 1996;Nishihara, 2003), specific roles for these glycans have been more elusive. In the brain, Fut9 is responsible for antennary fucosylation (Kudo et al., 2007;Williams et al., 2022), and deletion of Fut9 may drive mouse behavioral abnormalities (Kudo et al., 2007), cellular migration deficits (Kudo et al., 2007), and impaired neurite outgrowth (Gouveia et al., 2012). The abundance of antennary fucosylation on proteins important for cell adhesion and synaptogenesis (Fig. 5) suggests that this glycosylation may impact the biosynthesis, trafficking, or function of some of these proteins. The commonality of antennary fucosylation among SV and plasma membrane proteins (Fig. 5) is consistent with the notion that SV proteins first traffic to the plasma membrane prior to sorting to SVs (Régnier-Vigouroux et al., 1991). While high fucosylation is neither necessary nor sufficient for protein trafficking to SVs (Fig. 4, Fig. 5), the enrichment of the Fuc3 glycan φ on SVs (Fig. 3) suggests that SV recycling is associated with the presentation of fucosylated glycans at high density on the presynaptic plasma membrane. At present, the endogenous binding sites in brain for antennary fucose and Lewis X are to our knowledge unknown. Further studies examining specific links among antennary fucosylation, protein trafficking, neuronal circuits, and synaptic physiology are needed to clarify the functional roles of these glycoforms.
Our findings are particularly striking given the recent identification of Fut9 as a gene locus linked to schizophrenia (Schizophrenia Working Group of the Psychiatric Genomics Consortium, 2014). Abnormal glycosylation of the antennary fucose-bearing proteins EAAT2/GLT-1 (Bauer et al., 2010) and GABA A receptor subunits (Supplementary Table S4) (Mueller et al., 2014) has been described in schizophrenia postmortem brain studies. Altered fucosyltransferase enzyme levels have also been described in similar work (Mueller et al., 2017). Given that the specific brain functions of antennary fucosylation remain undefined, the potential links between this molecular feature and schizophrenia are numerous. For example, inefficient trafficking to axonal projections would broadly inhibit neurite outgrowth and neurotransmission, while reduced efficacy of cell-cell adhesion might reduce the stability of neuronal circuits. We note that fucose and fucosylated glycans are reportedly detectable in brain by MRI spectroscopy (Mountford et al., 2015), and their study may thus yield insights into the biology of schizophrenia and other developmental processes across the human life span.
Finally, we emphasize the need for further investigation into potential links among protein glycosylation, neuroinflammation, and synaptic pruning. Several studies have demonstrated a role for complement proteins in both synaptic pruning (Schafer et al., 2012;Stevens et al., 2007) and schizophrenia (Sekar et al., 2016), and lectins can directly activate the complement cascade (Garred et al., 2016). Among the bestknown complement-activating lectins is mannose-binding lectin, which is activated by mannosylated glycans typically found on microorganisms but not on circulating glycoproteins (Garred et al., 2016).
Mannose-binding lectin is a well-known mediator of inflammatory brain injury after ischemia or trauma (Neglia et al., 2020;Orsini et al., 2012) Strikingly, mannosylated glycans are common on neuronal glycoproteins (Williams et al., 2022) (Fig. 2) including at nerve terminals and SVs (Figs. 2-3). Abnormal expression or trafficking of neuronal N-glycans might thus modulate complement activation and synaptic pruning via the lectin pathway, particularly during periods of neuroinflammatory stress (Calcia et al., 2016;Weinhard et al., 2018). Future studies may define glycan-dependent processes in neuroinflammation-related brain injury, with a view toward new approaches for its treatment and prevention.

Materials and Methods
Animals C57B/6J mice of either sex between 15 and 20 days of age were used for all experiments. All work was conducted according to protocols approved by the University of Wisconsin Institutional Animal Care and Use Committee.
Antibody and bead preparation. Anti-SV2 mAb (SV2, DSHB) and anti-syt1 mAb (mAb 48, DSHB) were purified from ascites stocks generated prior to 2010 using protein G chromatography (Protein G Sepharose Fast Flow, Cytiva) and dialyzed extensively against phosphate-buffered saline (140 mM NaCl, 10 mM sodium phosphate buffer, pH 7.4). Single-use aliquots (200 µl, 300 µg) of mAb in PBS were kept frozen at -80 ºC. Dynabeads M-270 Epoxy (14302D, Thermo Fisher) were coupled to mAb according to published procedures . 10 mg of Dynabeads stored in DMF were collected with a magnetic stand and resuspended in 200 µl borate buffer (100 mM boric acid-NaOH, pH 8.5). To this suspension was added 200 µl mAb solution, followed by 200 µl 3 M ammonium sulfate in borate buffer, with mixing by pipetting up and down after each addition. This coupling mixture was incubated with rotation at 37 ºC overnight. The beads were then collected with a magnetic stand, the supernatant was discarded, and the beads were washed six times by trituration in 1 ml buffer followed by collection with the magnetic stand. The wash buffers were 500 mM NaCl, 50 mM ammonium acetate pH 4.5 and 500 mM NaCl, 50 mM Tris-HCl pH 8.0, used in an alternating manner (i.e., one buffer followed by the other, for three cycles total). The beads were then resuspended in 1 ml KPBS (145 mM KCl, 10 mM potassium phosphate buffer pH 7.4), transferred to a fresh tube, and collected on a magnetic stand. The supernatant was removed and the beads were resuspended at 30 mg/ml using 300 µl KPBS. Beads stored in this manner at 4 ºC remained effective for at least 2 months. SV preparation All buffers, tubes, and centrifuge rotors were cooled to 0-2 ºC prior to use. One or two mice were anesthetized with isoflurane, euthanized, and the brains including cerebellum and brain stem were rapidly removed. Each brain was placed in a tight-fitting Teflon-glass Dounce homogenizer with 3.8 ml ice-cold potassium homogenization buffer (125 mM KCl, 25 mM potassium phosphate buffer, 5 mM EGTA, pH 7.4) containing protease inhibitors (cOmplete mini EDTA-free, Roche, 1 tablet / 10 ml buffer) and homogenized with ten strokes using an overhead mixer rotating at 900 RPM. The homogenate was centrifuged at 35,000 x g for 20 minutes at 2 ºC. During centrifugation, 5 mg α -SV2 Dynabeads per brain were washed in KPBS and resuspended in two 2-ml microcentrifuge tubes (2.5 mg beads in 100 µl in each tube). The supernatant from each brain homogenate was added to two tubes containing Dynabeads (1.9 ml/tube), and the tubes were placed inside 50-ml conical tubes packed with ice and incubated with rotation for 25 minutes in a cold room. The supernatants were then discarded and each 2.5 mg portion of beads, corresponding to SVs from ½ mouse brain, was washed 3 times by gentle trituration in 1 ml icecold KPBS followed by collection on a magnetic stand. The beads were then resuspended and transferred to a fresh tube using KPBS, with beads bearing SVs from the same brain combined into the same tube (5 mg/tube), and the supernatant was removed. For proteomics and glycoproteomics studies, the beads were eluted using 50 µl 2% SDS containing 25 mM Tris-HCl pH 8.0 with heating to 50 ºC for 5 minutes, and the eluates were frozen at -80 ºC prior to use. For HPLC studies, the beads were eluted using 45 ul 0.5% SDS with heating to 50 ºC for 5 minutes, followed by a second elution with 45 µl 2% n-βdodecylmaltoside (DDM) (Gold Biotechnologies) for 5 minutes at room temperature. These eluates were combined and 2 µl 1 M triethylammonium bicarbonate (TEAB) was added prior to storage at -80 ºC.
Synaptosome preparation Synaptosomes were prepared by removing and homogenizing brains as above except that the homogenization buffer contained 125 mM NaCl, 25 mM HEPES-NaOH, and 5 mM EGTA pH 7.4. The brain homogenate was centrifuged at 3,500 x g for 2 minutes, the pellet was discarded, and the supernatant transferred into two 2-ml microcentrifuge tubes and centrifuged for 12 minutes at 14,000 x g at 4 ºC. The supernatant was discarded and the pellet was resuspended in 1530 µl 50 mM TEAB per 2-ml tube. 90 µl of 10% SDS and 180 µl 10% DDM (0.5% SDS, 1% DDM final) were added to each tube, which was incubated for 1 hour with rotation at 4 ºC prior to aliquoting and freezing at -80 ºC.
Immunopurification of SV2 and syt1 SV2 and syt1 were immunoprecipitated from synaptosomes prepared as above but resuspended in 1.6 ml synaptosome homogenization buffer followed by the addition of 180 µl 10% DDM (~1% DDM final) per one-half brain. The samples were incubated with rotation for 1 hour at 4 ºC followed by pelleting of insoluble material by centrifugation at 20,000 x g for 20 minutes at 4 ºC. The supernatants were transferred to new tubes, 3 mg α -syt1 or α -SV2 Dynabeads were added, and the tubes were incubated with rotation for 30 minutes. The beads were then washed with cold synaptosome resuspension buffer (4 x 1 ml) and eluted with 50 µl 2% SDS containing 25 mM Tris-HCl pH 8.0 with heating to 50 ºC for 5 minutes. Eluates were stored frozen at -80 ºC prior to use. and the tubes were mixed well, followed by the addition of 1 volume of absolute ethanol and brief incubation on a thermomixer (5', 1000 RPM, 23 ºC) to drive protein adsorption to the beads. The beads were washed three times with 200 µl 80% ethanol and transferred to a fresh tube with the final wash. The supernatant was removed and tryptic peptides were eluted from the beads by overnight digestion in 50 µl trypsin solution (V5111, Promega, 0.01 µg/µl in 100 mM ammonium bicarbonate) with shaking in a thermomixer (1000 RPM, 37 ºC). Eluates from this step were used directly for LC-MS or subject to glycopeptide enrichment (vide infra). For synaptosomes, a similar procedure was followed but scaled up 10-fold, using 500 µl synaptosomal lysate as input, washes of 3 x 1 ml 80% ethanol, and elution in 250 µl trypsin solution.

LC-MS/MS Data Analysis
Raw files from LC-MS runs were analyzed using the FragPipe software suite (v.17.1) with further processing of FragPipe output performed in R. All R scripts used are available via Github at https://github.com/mazbradberry/public/tree/glycoproteomics. For proteomics experiments with label-free quantification (LFQ), SV and synaptosome experiments were analyzed separately. Spectra were searched with MSFragger (v.3.4) (Kong et al., 2017) using a mouse proteome database downloaded from Uniprot on 11 October 2021 and the following settings: precursor mass tolerance, ± 20 ppm; fragment mass tolerance, ± 20 ppm, mass calibration and parameter optimization enabled; isotope error, 0/1/2; enzymatic cleavage, strict trypsin with up to 2 missed cleavages; peptide length, 7-50; peptide mass range, 500-5000 Da. Methionine oxidation and N-terminal acetylation were allowed as variable modifications and cysteine carbamidomethylation was included as a fixed modification. Validation was performed with PeptideProphet using closed search defaults for peptides and ProteinProphet for proteins.
LFQ was performed with IonQuant  with match-between-runs, normalization, and MaxLFQ enabled. The following settings were used: feature detection m/z tolerance, ± 10 ppm; feature detection RT tolerance, 0.4 minutes; match between runs (MBR) tolerance 5 minutes, MBR ion FDR, 0.01; MBR peptide and protein FDR, 1; top 3 ions used for quantification with a minimum frequency of 0.5 and detection in at least 1 experiment. For glycoproteomics experiments, runs were grouped by sample type (SV, SV with enrichment, or synaptosomes with enrichment) prior to analysis. The same mouse protein database was used and spectra were searched with MSFragger (v.3.4) using glyco mode (Polasky et al., 2022 with the following settings: precursor mass tolerance, ± 20 ppm; fragment mass tolerance, ± 20 ppm, mass calibration and parameter optimization enabled; isotope error, 0/1/2; enzymatic cleavage, strict trypsin with up to 2 missed cleavages; peptide length, 7-50; peptide mass range, 400-5000 Da. Methionine oxidation and N-terminal acetylation were allowed as variable modifications and cysteine carbamidomethylation included as a fixed modification. The default 183 mass offsets corresponding to possible N-glycan compositions were included and restricted to asparagine residues. Labile modification search mode was set to nglycan and diagnostic Y ion and fragment masses were left as defaults. PTM-Shepherd was enabled, diagnostic ion search was enabled with default settings, and glycans were assigned in N-glycan mode with an FDR of 0.025, mass tolerance of ± 50 ppm, and isotope error range of -1 to 3. PeptideProphet and ProteinProphet were used for peptide and protein validation, respectively, with default settings. The same parameters were used for glycoproteomic studies of immunoprecipitated SV2 and syt1. Glycopeptide spectral matches (GlycoPSMs) were obtained from the psm output file, filtered for Q-values of ≤ 0.025, and annotated using the table shown in Table S3, with sialic acid-containing compositions omitted from annotation based on the likelihood that they represent combinations of N-and O-glycosylation on the same peptide (Williams et al., 2022). Tetrafucosylated glycans and compositions tentatively identified as singly fucosylated oligomannose/highmannose glycans were also detected but were omitted from analysis given their relatively minor contribution to the total glycan pool. Relative frequencies of unique annotated N-glycans (Fig. 2) were determined using R. For weighting based on LFQ intensity from proteomics experiments (Fig. 2), each unique glycoPSM was assigned a weight equal to the square root of the LFQ intensity for the corresponding protein and sample type divided by the number of unique glycoPSMs attributed to that protein. For GO analysis (Fig. 5), gene lists were extracted from annotated glycoPSM tables and subjected to biological process GO term search (geneontology.org, accessed 25 April 2022) (Ashburner et al., 2000;The Gene Ontology Consortium et al., 2021).

N-glycan release and labeling
For synaptosome samples, 50 µl synaptosomal lysate in 1% DDM and 0.5% SDS was combined with 50 µl 2% DDM and 5 µl 1 M DTT and heated to 50 ºC for 15 minutes. For SV samples, 100 µl of SDS-DDM eluate (0.25% SDS, 1% DDM final) was combined with 5 µl 1M DTT and heated to 50 ºC for 15 minutes. For each sample, 1 µl PNGase F (P0708, NEB) was then added, and the mixture was incubated at 37 ºC for two hours. Each sample was allowed to cool to room temperature, combined with 50 µl of freshly prepared procainamide solution (40 mg/ml in 70:30 DMSO:acetic acid) and incubated on ice for 5 minutes. Samples were then centrifuged (20,000 x g, 10 minutes, 4 ºC) and the supernatants (150 µl) transferred to fresh PCR tubes. 10 µl of sodium cyanoborohydride solution (5 M in 1 M NaOH, Sigma 296945) was added to each reaction, which was then incubated at 65 ºC for two hours in a fume hood using a miniature PCR block. All steps involving sodium cyanoborohydride, including sample cleanup, were carried out in a fume hood. The samples were then combined with 20 volumes of MeCN (i.e., 75 µl sample was added to 1.5 ml MeCN) in 2-ml microcentrifuge tubes and subjected to cleanup by solid-phase extraction using a vacuum manifold. For each sample, a solid-phase extraction cartridge (OASIS HLB 30 mg, Waters) was equilibrated with 1 ml 95:5 MeCN:H 2 O, and the sample (~3.2 ml) was applied. The cartridge was washed (2 x 1 ml 95:5 MeCN:H 2 O) and procainamide-labeled glycans eluted with 500 µl 50:50 MeCN:H 2 O. The eluates were dried using a speedvac, resuspended in Milli-Q water (200 µl for synaptosome samples, 40 µl for SV samples), and centrifuged to remove insoluble material.

SDS-PAGE and immunoblot
Synaptosome and SV samples were combined with 4X SDS sample buffer containing DTT and heated to 50 ºC for 15 minutes except as shown in Supplementary Fig. S5. 10 µl of prepared sample containing 6.7 µl synaptosome lysate or 1-1.5 µl SV eluate was subjected to SDS-PAGE on 4-20% gradient gels (Criterion TGX, Bio-Rad) and transferred to a PVDF membrane using a semi-dry blotting apparatus. Blots were blocked using TBS-T (150 mM NaCl, 10 mM Tris-HCl pH 7.4, 0.1% Tween-20) containing 5% nonfat dry milk and incubated overnight at 4 ºC with primary antibody in TBS-T containing 1% nonfat dry milk. Antibodies for immunoblot included guinea pig anti-synaptophysin (104 211, Synaptic Systems, 1:1000 dilution of a 0.5 mg/ml stock) or mouse monoclonal anti-SV2 (SV2, DSHB, 1:1,000 dilution of a 1.2 mg/ml stock purified from ascites). Blots were washed in TBS-T, and HRP-labeled secondary antibodies were used for detection.