Abstract
Synaptic diversity is a key feature of neural circuits. Its underlying molecular basis is largely unknown, due to the challenge of analyzing the protein composition of specific synapse types. Here, we isolate the hippocampal mossy fiber (MF) synapse, taking advantage of its unique size and architecture, and dissect its proteome. We identify a rich cell-surface repertoire that includes adhesion proteins, guidance cue receptors, extracellular matrix (ECM) proteins, and proteins of unknown function. Among the latter, we find IgSF8, a previously uncharacterized neuronal receptor, and uncover its role in regulating MF synapse architecture and feedforward inhibition on CA3 pyramidal neurons. Our findings reveal a diverse MF synapse surface proteome and highlight the role of neuronal surface-ECM interactions in the specification of synapse identity and circuit formation.
One Sentence Summary Proteomic dissection of a specific synapse
Main Text
Neural circuits are composed of distinct neuronal cell types connected in highly specific patterns. Establishing these patterns of connectivity critically relies on cell-surface proteins (CSPs) expressed in cell type-specific combinations. CSPs, including transmembrane, membrane-anchored, and secreted proteins, engage in networks of interactions that control neurite guidance, target selection, and synapse development required for the formation of functional circuits (1). Single-cell RNA sequencing has enabled the characterization of cell type-specific CSP repertoires (2–6), but determining how these dictate complex patterns of connectivity (7, 8) poses a major challenge.
This challenge is exemplified by pyramidal neurons, which receive different types of synapses on their dendritic arbor, each with a distinct architecture, subcellular location, and functional properties. This synaptic diversity is essential for information processing in pyramidal neurons (9). Recent studies reveal a synapse type-specific localization and function of postsynaptic adhesion molecules in hippocampal pyramidal neuron dendrites (10–12), suggesting that compartmentalized distributions of CSPs contribute to the specification of synaptic structure and function. Analogous to single-cell sequencing, probing the mechanisms underlying synaptic diversity requires dissecting the molecular composition of specific synapse types. This has remained challenging, as microdissection or chemical labeling strategies combined with mass spectrometry (MS) (13–15) average different synapse types, and affinity purification of synapse type-specific protein complexes (16) requires genetically engineered mice. Here, we isolate the hippocampal mossy fiber (MF) synapse, a large and morphologically complex excitatory synapse (17) connecting dentate granule cell axons (mossy fibers) and CA3 pyramidal neuron dendrites in stratum lucidum (SL) (Fig. 1A, B), from wild-type (WT) tissue and map its CSP landscape.
To isolate a specific synapse type from the hippocampus, we started with a previously published approach (18) that takes advantage of the MF-CA3 synapse’s large size. We verified that this method enriches for MF-CA3 synaptosomes (fig. S1) and subsequently adapted the procedure to increase efficiency and improve enrichment of synaptic material (Fig. 1C). We accelerated the procedure by omitting gradient centrifugation and depleting myelin from the sample (fig. S2A). To improve enrichment of synaptic material, we utilized a fluorescence-activated synaptosome sorting (19) approach in which we labeled MF-CA3 synaptosomes with a fluorophore-conjugated monoclonal antibody against the extracellular domain of Nectin-3 (fig. S2B), a CA3-expressed adhesion protein that localizes to puncta adherentia junctions in mature MF-CA3 synapses (20), and FM4-64 membrane dye (fig. S2C, D). Western blot (WB) analysis of sorted MF-CA3 synaptosomes revealed a robust enrichment of MF-CA3 synapse markers Synaptoporin (Synpr) and Nectin-3, and depletion of myelin and nuclear markers (fig. S2E). Immunofluorescence analysis confirmed the presence of Synpr-, Nectin-3-, and vesicular glutamate transporter 1 (VGluT1)-positive large synaptosomes and depletion of Hoechst-positive nuclei in the sorted sample (fig. S2F and G), demonstrating feasibility of our approach to isolate MF-CA3 synaptosomes from hippocampal tissue.
To dissect the MF-CA3 synapse proteome, we compared fluorescently sorted MF-CA3 synaptosomes to P2 synaptosomes isolated in parallel from the same homogenate (fig. S2A). P2 synaptosomes represent a mixed population of small hippocampal synapses and serve as a background reference in our analysis. We analyzed sorted MF-CA3 synaptosomes and P2 synaptosomes from three independent experiments (10-12 mice per experiment) by LC-MS/MS. We identified 3592 proteins with at least 3 peptide identifications among replicates, and 11,6% and 7,9% of these proteins were exclusive to sorted MF-CA3 synaptosomes or P2 synaptosomes, respectively (Fig. 1D and table S1). Gene ontology (GO) analysis showed a similar enrichment for synaptic terms in sorted MF-CA3 and P2 synaptosomes (fig. S3A). We calculated log2 fold-change (FC) enrichment in sorted MF-CA3 versus P2 synaptosomes using a label-free semi-quantitative approach based on the normalized spectral abundance factor (NSAF) (Fig. 1E; fig. S3B and table S1). Our analysis revealed 605 significant proteins with positive MF-CA3/P2 FC, and 138 significant proteins exclusively detected in sorted MF-CA3 synaptosomes (Fig. 1E; table S1). This MF-CA3 synaptic proteome (fig. S3C) comprises multiple proteins previously reported to be strongly enriched at MF-CA3 synapses, including the synaptic vesicle-associated proteins Synpr, Syn3, Rabphilin-3A, and ZnT3; the glutamate receptors GluK2 and mGluR3; the presynaptic scaffold protein liprin-α2; the dense core vesicle secretion-related protein CAPS2; and the puncta adherentia junction-associated proteins Nectin-3 and Af-6 (Fig. 1E; fig. S3D). We confirmed Syn3, Af-6, mGluR2/3, and ZnT3 localization to SL using immunohistochemistry (IHC) (Fig. 1F). These validations indicate that our approach confidently identifies MF-CA3 synaptic proteins.
Using the UniProt database to query our results for transmembrane, membrane-anchored, and secreted proteins among the MF-CA3 synaptic proteome, we identified a rich repertoire of CSPs that includes adhesion proteins, receptors, secreted glycoproteins, receptor protein tyrosine phosphatases and tyrosine kinases (Fig. 2A and table S2). Most major CSP protein families, such as the immunoglobulin (Ig) superfamily (IgSF), fibronectin type-III (FN3), and leucine-rich repeat (LRR) family, are represented (Fig. 2B). Only a small proportion (20,8%) of these CSPs has previously been reported to localize or function at MF-CA3 synapses (table S2). Using the synapse biology GO database SynGO (21), we determined that 48% of the identified CSPs genes lack a SynGO annotation (Fig. 2C), suggesting that these are novel candidate synaptic CSPs. The secreted protein BRINP2 and the receptors FAM171A2, APMAP, and IgSF8 lack a function in the brain altogether (Fig. 2C).
To validate these findings, we tested a large panel of antibodies for detection of CSPs by WB and IHC (table S2). We validated the presence of 15 CSPs in sorted MF-CA3 synaptosomes by WB, including BRINP2 and IgSF8 (Fig. 2D). Using IHC on postnatal day (P) 28 mouse hippocampal sections, we confirmed localization to the MF-CA3 pathway for 21 CSPs (Fig. 2E; fig. S4). Of the CSPs lacking synaptic function, the axon guidance-related receptors ISLR2, Neogenin 1, ROBO2, Plexin-A3, and secreted protein CRTAC1 (22), displayed strongly enriched immunoreactivity in SL (Fig. 2E). We further confirmed robust localization to SL of the GPI-anchored protein Negr1 (Fig. 2E). Of the CSPs without a known function in the brain, we validated the presence of FAM171A2 in SL and observed strong labeling for IgSF8 in the MF-CA3 pathway (Fig. 2E). Together, these results confirm the presence of a diverse repertoire of CSPs at MF-CA3 synapses.
IgSF8 was of particular interest due to its strong localization to the MF-CA3 tract (Fig. 2E) and lack of known brain function. IgSF8 (Fig. 3A) has previously been detected in excitatory postsynaptic densities (PSDs) (14) and active zone (AZ) membrane preparations (23), but not in synaptic cleft proteomes (24, 25). IgSF8 is highly expressed in the nervous system (26) and localizes to axons and presynaptic terminals of olfactory sensory neurons during synaptogenesis (27). We analyzed IgSF8 expression in hippocampal lysates and found that IgSF8 protein levels mildly increased during postnatal development (fig. S5A). To determine the synaptic localization of IgSF8, we performed subcellular fractionation, as the conditions required for IgSF8 IHC proved unsuitable for high-resolution imaging. IgSF8 mainly distributed to the Triton-soluble fraction of synaptosomes containing the presynaptic protein Synaptophysin (fig. S5B), supporting a presynaptic localization.
As a presynaptic receptor, IgSF8 might interact with other synaptic CSPs. To identify synaptic ligands for IgSF8, we performed affinity chromatography using recombinant IgSF8-ecto-Fc as bait on whole brain synaptosome extract (fig. S6A and table S3) and MF-CA3 synaptosome extract (fig. S6B and table S3) and analyzed bound proteins by LC-MS/MS. We identified Tenascin-R (TenR), an ECM protein with roles in cell adhesion, synaptic transmission and plasticity (28), as the main IgSF8 binding partner in both experiments. We validated the IgSF8-TenR interaction using cell-surface binding assays and pull-down assays in HEK293T cells (fig. S6C and D). To test whether IgSF8 and TenR interact directly, we mixed equimolar amounts of Fc control protein, TenR-Fc, or Brevican-Fc, another brain ECM protein, with His-IgSF8 recombinant protein and precipitated Fc proteins. We found that the IgSF8-TenR interaction is direct and specific (fig. S6E). TenR is among the CSPs we identified in the MF-CA3 synapse proteome (Fig. 2C and table S2), indicating that our approach is capable of isolating both receptor and ligand at MF-CA3 synapses.
To determine the functional significance of IgSF8 for MF-CA3 synapse development, we removed IgSF8 presynaptically by crossing Igsf8 conditional knockout (cKO) mice with the dentate granule cell-specific Rbp4-Cre line. This selectively abolished IgSF8 immunoreactivity in CA3 SL, without affecting gross MF-CA3 tract morphology (Fig. 3B). Infecting cultured Igsf8 cKO hippocampal neurons with a lentiviral (LV) vector harboring Cre recombinase robustly reduced IgSF8 protein levels (fig. S7), validating the utility of the Igsf8 cKO line for assessing IgSF8 function. We prepared hippocampal sections from P30 Rbp4-Cre:Igsf8 cKO and control littermates and imaged MF-CA3 synapses by transmission electron microscopy (TEM) (Fig. 3C). Bouton number and area were not affected in Igsf8 cKO (fig. S8A and B), but we observed a clear reduction in AZ number and length, and a corresponding decrease in PSD number and length (Fig. 3D). To analyze MF bouton morphology, we injected adeno-associated viral (AAV) vectors expressing membrane-GFP (mGFP) and Cre recombinase in Igsf8 cKO mice to remove IgSF8 expression (Fig. 3E). AAV-mGFP was used as a control. We observed a mild reduction in MF bouton volume, but a dramatic decrease in the number of filopodia emerging from the main bouton in Igsf8 cKO (Fig. 3F and G). These results indicate that IgSF8 plays a role in shaping MF-CA3 synapse architecture and morphology.
To analyze functional changes at MF-CA3 synapses in the absence of IgSF8, we performed whole-cell voltage-clamp recordings of CA3 neurons in acute hippocampal slices from P27-35 Rbp4-Cre:Igsf8 cKO and control littermates. Frequency and amplitude of spontaneous excitatory postsynaptic currents (sEPSCs) were reduced in Igsf8 cKO mice (Fig. 4A to E), whereas decay time was not affected (fig. S9A). Histogram analysis showed a loss of large-amplitude sEPSCs (fig. S9B and C), which originate from MF-CA3 synapses (29), consistent with the decrease in number and length of synaptic junctions observed in Igsf8 cKO mice. To specifically assess transmission at MF-CA3 synapses, we expressed Cre-dependent Channelrhodopsin-2 (DIO-ChR2) in Rbp4-Cre:Igsf8 cKO and control littermates and optically stimulated MF axons while recording from CA3 neurons (Fig. 4F). We analyzed paired-pulse facilitation, a form of short-term plasticity, to assess MF-CA3 presynaptic properties, but found no differences in the amplitude of the first evoked EPSC or paired-pulse ratio between Igsf8 cKO and control mice (fig. S9D to F), suggesting that evoked transmission at MF-CA3 synapses is not altered in the absence of IgSF8.
The filopodia emerging from MF boutons synapse onto interneurons in SL that mediate feedforward inhibition (FFI) on CA3 neurons (30, 31). Given the dramatic decrease in the number of filopodia in boutons of Igsf8 cKO mice (Fig. 3F and G), we hypothesized that FFI would be reduced. Indeed, we observed a robust increase in the excitation-inhibition ratio when measuring light-evoked EPSCs and IPSCs from the same CA3 neuron (Fig. 4G to I). To determine the consequences of reduced FFI, we optically stimulated MF axons with a 10 Hz train of 20 stimuli and observed that CA3 neurons in Igsf8 cKO mice fired action potentials earlier in the train than in controls (Fig. 4J to L), supporting the notion that reduced FFI in the absence of IgSF8 leads to increased excitability of CA3 neurons. Together, these results indicate that IgSF8 promotes MF bouton filopodia density to control microcircuit development and CA3 neuron excitability.
Our approach to isolate MF-CA3 synaptosomes and map their surface proteome provides the first insight into the CSP landscape of a specific excitatory synapse. We identify a rich CSP repertoire at MF-CA3 synapses. Our data suggest that multiple CSP complexes, likely acting in parallel, contribute to the specification of MF-CA3 synapse identity (fig. S10). The diverse CSP repertoire may also reflect heterogeneity among MF-CA3 synapses. MF synapse maturation varies due to the continuous integration of newborn granule cells into the hippocampal circuit (32), and distinct histories of synaptic activity may further diversify CSP composition. One of the CSPs without a previously characterized brain function we uncover at MF-CA3 synapses is the receptor IgSF8, which regulates synapse architecture, MF filopodia density, and FFI in the MF-CA3 microcircuit. The cytoplasmic tail of IgSF8 binds Ezrin-Radixin-Moesin (ERM) (33) and alpha-actinin (34), both of which are linked to filopodia formation. We identify the ECM protein TenR as a novel ligand of IgSF8. TenR induces actin-rich protrusions in cultured neurons (35), suggesting that the IgSF8-TenR interaction may promote formation or stabilization of MF filopodia. TenR is one of multiple ECM proteins in the MF-CA3 synaptic proteome, which also includes the largely uncharacterized ECM proteins Crtac1 and Brinp2 (36). ECM proteins play important roles in the development of synaptic connectivity (37) but remain poorly understood. Our findings emphasize the role of neuronal surface-ECM interactions at MF-CA3 synapses (fig. S10).
The approach taken here may be applicable to other types of synapses. Our data can be used as a resource to explore other aspects of the MF-CA3 synapse, including molecular mechanisms of synaptic transmission, and may be of use in efforts to model synaptic connectivity in the dentate gyrus (38). Dissecting the molecular composition of specific synapse types will be an essential step toward molecular connectomics (39).
Funding
Leica SP8x confocal microscope provided by InfraMouse (KU Leuven-VIB) through a Hercules type 3 project (ZW09-03). N.A. is supported by the Fundação para a Ciência e a Tecnologia (FCT, grant number SFRH/BD/128869/2017). J.N.S. is supported by R01AG061787. J.d.W is supported by ERC Starting Grant (#311083), FWO Odysseus Grant, FWO Project grant G0C4518N, FWO EOS grant G0H2818N, and Methusalem grant of KU Leuven/Flemish Government.
Author contributions
N.A., J.N.S., and J.d.W. conceived the study and designed experiments. N.A., S.N.S, J.V., G.C., V.R., J.t.B., S.P., K.M.V., N.V.G., and K.D.W. performed experiments and analyzed data. N.A. and J.d.W. wrote the paper with input from all authors. All authors contributed to and approved the final version.
Competing interests
The authors declare no competing interests.
Data and materials availability
MS raw files and search results can be accessed via Proteome Exchange at PXD013492.
Supplementary Materials
Materials and Methods
Figures S1-S10
Tables S1-S4
Table legends S1 to S4
Acknowledgments
We thank Franck Polleux, Dietmar Schmucker, Pierre Vanderhaeghen, Dan Dascenco, Luís Ribeiro, and Sara Calafate for critical reading of the manuscript and De Wit lab members for discussion and comments. We thank Pier Andrée Penttila and Christèle Nkama (VIB-KU Leuven FACS Core), the VIB-KU Leuven BioImaging Core, and Joris Vandenbempt for experimental help; Nicola Fattorelli for data visualization advice; Etienne Herzog and Matthew Holt for experimental advice.