Summary
At excitatory synapses, the choice between recycling or degradation of glutamate AMPA receptors controls the direction of synaptic plasticity. In this context, how the degradation machinery is targeted to specific synaptic substrates in an activity-dependent manner is not understood. Here we show that SUSD4, a complement-related transmembrane protein, is a tether for HECT ubiquitin ligases of the NEDD4 subfamily, which promote the degradation of a large number of cellular substrates. SUSD4 is expressed by many neuronal populations starting at the time of synapse formation. Loss-of-function of Susd4 in the mouse prevents activity-dependent degradation of the GLUA2 AMPA receptor subunit and long-term depression at cerebellar synapses, and leads to impairment in motor coordination adaptation and learning. SUSD4 could thus act as an adaptor targeting NEDD4 ubiquitin ligases to AMPA receptors during long-term synaptic plasticity. These findings shed light on the potential contribution of SUSD4 mutations to the etiology of neurodevelopmental diseases.
Introduction
AMPA-type glutamate receptors are responsible for fast excitatory transmission in the central nervous system. Fine-tuning of their properties and their synaptic numbers underlies synapse maturation and synapse plasticity and is central to proper circuit function and behavior (Anggono and Huganir, 2012; Diering and Huganir, 2018; Kessels and Malinow, 2009). Long-term synaptic plasticity has been proposed to underlie learning and memory (Collingridge et al., 2010; Nicoll, 2017). Synaptic strength can be either decreased (long-term depression, LTD) or increased (long-term potentiation, LTP) depending on the timing and pattern of neuronal activity in a network. Defects in the regulation of long-term synaptic plasticity and AMPA receptors could contribute to a range of brain diseases such as epilepsy, intellectual deficiencies and autism spectrum disorders (ASDs). Long-term synaptic plasticity deficits are a common phenotype of many mice models for intellectual disabilities and ASDs (Baudouin et al., 2012; Piochon et al., 2014; Volk et al., 2015). Mutations in genes coding for AMPA receptor subunits (Davies et al., 2017; Martin et al., 2017; Salpietro et al., 2019; Satterstrom et al., 2018) or their regulatory subunits (Brechet et al., 2017; Chiu et al., 2017) have been found in patients with neurodevelopmental disorders.
The modification of AMPA receptor numbers at synapses is a highly dynamic process, involving regulation of receptor diffusion (Choquet and Triller, 2013; Penn et al., 2017), their insertion in the plasma membrane, anchoring at the postsynaptic density and endocytosis (Anggono and Huganir, 2012). After activity-dependent endocytosis, AMPA receptors are either recycled to the plasma membrane or targeted to the endolysosomal compartment for degradation (Ehlers, 2000; Lee et al., 2004; Park et al., 2004). The decision between these two fates regulates the direction of synaptic plasticity. Recycling promotes LTP and relies on many molecules, such as GRASP1, GRIP1, PICK1 and NSF (Anggono and Huganir, 2012). Targeting to the endolysosomal compartment and degradation promote LTD (Fernandez-Monreal et al., 2012; Kim et al., 2017; Matsuda et al., 2013), but the regulation of the targeting and degradation process remains poorly understood. One mechanism that regulates activity-dependent degradation of synaptic proteins is ubiquitination, a major post-translational modification (Cajigas et al., 2010). Both GLUA1 and GLUA2 AMPA receptor subunits have been shown to be ubiquitinated in an activity-dependent manner (Lussier et al., 2011; Schwarz et al., 2010; Widagdo et al., 2015). A major remaining question however is how AMPA receptor degradation is specifically controlled given the large number of substrates for E3 ubiquitin ligases in cells, and the spatio-temporal specificity required for proper synaptic plasticity in neurons.
The SUSD4 gene is located in a genomic region deleted in patients with the 1q41q42 syndrome that includes developmental delays and intellectual deficiency (Rosenfeld et al., 2011). SUSD4 has been identified recently as one of the 124 genes enriched in de novo missense mutations in a large cohort of individuals with ASDs or intellectual deficiencies (Coe et al., 2019). A copy number variation and several de novo mutations with a high CADD score, which indicates the deleteriousness of the mutations, have been described in the SUSD4 gene in patients with ASDs ((Cuscó et al., 2009); denovo-db, Seattle, WA (denovo-db.gs.washington.edu) 10, 2019). The mammalian Susd4 gene codes for a transmembrane protein with four extracellular CCP (complement control protein) domains (Figure 1A) and is highly expressed in the central nervous system (Holmquist et al., 2013). The CCP domain, also known as Sushi domain, is evolutionarily conserved and found in several proteins with synaptic function. Acetylcholine receptor clustering is regulated by CCP-containing proteins in Caenorhabditis elegans (Gendrel et al., 2009) and in Drosophila melanogaster (Nakayama et al., 2016). In humans, mutations in the CCP domain-containing secreted protein SRPX2 are associated with epilepsy and speech dysfunction, and SRPX2 knockdown leads to decreased synapse number and vocalization in mice (Sia et al., 2013). Despite the increase in the diversity of CCP domain-containing proteins in evolution (18 in C. elegans and 198 in humans; smart.embl.de), the function of many of the CCP domain-containing proteins remains unknown. The only described role for the SUSD4 protein is its ability to regulate complement system activation in erythrocytes by binding the C1Q domain (Holmquist et al., 2013). Interestingly, the C1Q globular domain is found in major synaptic regulators such as C1QA (Stevens et al., 2007), CBLNs (Matsuda et al., 2010; Uemura et al., 2010) and C1Q-like proteins (Bolliger et al., 2011; Kakegawa et al., 2015; Sigoillot et al., 2015). Altogether these studies point to a potential role of SUSD4 in synapse formation and/or function and in the etiology of several neurodevelopmental disorders.
Proper development and function of the cerebellar circuitry is central for motor coordination and adaptation, and a range of cognitive tasks (Badura et al., 2018; Hirai et al., 2005; Ichise et al., 2000; Lefort et al., 2019; Rochefort et al., 2011; Tsai et al., 2012), and cerebellar dysfunction is associated with several neurodevelopmental disorders including ASDs (Stoodley, 2016; Stoodley et al., 2018; Wang et al., 2014). Cerebellar Purkinje cells (PCs) receive more than a hundred thousand parallel fiber (PF) synapses whose formation, maintenance and plasticity are essential for cerebellar-dependent learning (Gutierrez-Castellanos et al., 2017; Hirai et al., 2005; Ito, 2006; Kashiwabuchi et al., 1995). Postsynaptic LTD was first described at synapses between PFs and cerebellar PCs (Gao et al., 2012; Hirano, 2018; Ito, 2001; Ito and Kano, 1982), where it can be induced by conjunctive stimulation of PFs with the other excitatory input received by PCs, the climbing fiber (CF) (Coesmans et al., 2004; Ito, 2001; Suvrathan et al., 2016). The function of members of the C1Q family, such as CBLN1 and C1QL1, is essential for excitatory synapse formation and LTD in cerebellar PCs (Hirai et al., 2005; Kakegawa et al., 2015; Matsuda et al., 2010; Sigoillot et al., 2015; Uemura et al., 2010), suggesting that proteins, such as SUSD4, that interact with the C1Q globular domain could regulate these processes. PF LTD is dependent on AMPA receptor endocytosis, in particular the GLUA2 subunit (Chung et al., 2003; Xia et al., 2000), and their targeting to RAB7 compartments (Kim et al., 2017). However, the mechanism regulating AMPA receptor targeting to the endolysosomal compartment at PF synapses, as at other synapses, remains unknown.
To identify the role of SUSD4 in brain development and function, we thus analyzed the phenotype of a Susd4 loss-of-function mouse model in the cerebellum. We found that knockout of the Susd4 gene leads to impaired synaptic LTD in cerebellar PCs and deficits in motor coordination and adaptation learning. Using affinity-purification followed by proteomics analysis and biochemical assays, we show that the SUSD4 protein binds E3 ubiquitin ligases of the NEDD4 family and regulates activity-dependent degradation of GLUA2 receptors. Because of the domain structure of SUSD4, our results suggest a new regulatory mechanism that can bring spatio-temporal specificity to the degradation machinery in neurons, allowing proper synaptic plasticity and learning.
Results
Susd4 is broadly expressed in neurons during postnatal development
Given the potential synaptic role for SUSD4, its pattern of expression should correlate with the timing of synapse formation and/or maturation during postnatal development. In situ hybridization experiments using mouse brain sections showed high expression of Susd4 mRNA in neurons in many regions of the central nervous system, including the cerebral cortex, the hippocampus, the cerebellum and the brainstem (Figure 1 and S1). Susd4 expression was already detected by postnatal day 0 (P0) in some regions, but increased with brain maturation (Figure S1). In the cerebellum, a structure where the developmental sequence leading to circuit formation and maturation is well described (Sotelo, 2004), quantitative RT-PCR showed that Susd4 mRNA levels start increasing at P7 and by P21 reach about 15 times the levels detected at birth (Figure 1B). At P7, a major increase in synaptogenesis is observed in the cerebellum, due to the formation of hundreds of thousands of PF excitatory synapses on the distal dendritic spines of each PC, and the translocation of a single CF arising from an inferior olivary neuron that forms about 300 excitatory synapses on proximal PC dendrites (Leto et al., 2016). In the brainstem, where cell bodies of inferior olivary neurons are located, the increase in Susd4 mRNA occurs earlier, already by P3, in parallel with the rate of synaptogenesis increasing during the first postnatal week in the inferior olive (Gotow and Sotelo, 1987). To identify the subcellular localization of the SUSD4 protein and because of the lack of suitable antibodies for immunolabeling, viral particles enabling CRE-dependent coexpression of HA-tagged SUSD4 and GFP in neurons were injected in the cerebellum of adult mice expressing the CRE recombinase specifically in cerebellar PCs. Immunofluorescent labeling against the HA tag demonstrated the presence of tagged SUSD4 in dendrites and distal dendritic spines that are the postsynaptic compartments of PF synapses in PCs (Figure 1C). Thus, its timing of expression during postnatal development and its subcellular localization in cerebellar PCs support a role for SUSD4 in excitatory synapse formation and/or function.
Susd4 loss-of-function leads to deficits in motor coordination and learning
To determine the synaptic function of SUSD4, we analyzed the phenotype of constitutive knockout (KO) mice completely lacking expression of Susd4 mRNA due to deletion of exon 1 (Figure 1D, 1E and S2). No obvious alterations of mouse development and behavior were detected in those mutants, an observation that was confirmed by assessment of their physical characteristics (weight, piloerection, cf. Table S1), basic behavioral abilities such as sensorimotor reflexes (whisker responses, eye blinking, cf. Table S1) and motor responses (open field locomotion, cf. Table S1). Because of the high Susd4 expression in the olivocerebellar system (Figure 1 and S1), we further assessed the behavior of Susd4 KO mice for two abilities well known to depend on normal function of this network, motor coordination and motor learning (Kayakabe et al., 2014; Lalonde and Strazielle, 2001; Rondi-Reig et al., 1997). Using a footprint test, a slightly larger print separation of the front and hind paws in the Susd4 KO mice was detected but no differences in the stride length and stance width were found (Figure S3). In the accelerated rotarod assay, a classical test of motor adaptation and learning (Buitrago et al., 2004), the mice were tested three times per day at one hour interval during five consecutive days. The Susd4 KO mice performed as well as the wild-type (WT) controls on the first trial, indicating that there is no deficit in their balance function (Figure 1F), despite the slight change in fine motor coordination found in the footprint test. However, while the WT mice improved their performance as early as the third trial on the first day, and further improved with several days of training, no learning could be observed for the Susd4 KO mice either during the first day, or in the following days. These results show that Susd4 loss-of-function leads to impaired motor coordination and learning in adult mice.
Susd4 loss-of-function prevents long-term depression at cerebellar parallel fiber/Purkinje cell synapses
Motor coordination and learning are deficient when cerebellar development is impaired (Hirai et al., 2005; Ichise et al., 2000; Tsai et al., 2012; Zuo et al., 1997). No deficits in the global cytoarchitecture of the cerebellum and morphology of PCs were found in Susd4 KO mice (Figure S4). Using high density microelectrode array, we assessed the spontaneous activity of PCs in acute cerebellar slices from Susd4 KO compared to WT (Figure S5). No differences were detected in either the mean spiking frequency, the coefficient of variation of interspike intervals (CV) and the intrinsic variability of spike trains (CV2, (Holt and Douglas, 1996) showing that the firing properties of PCs are not affected by Susd4 loss-of-function.
Co-immunolabeling of parallel fiber presynaptic boutons using an anti-VGLUT1 antibody and PCs using an anti-calbindin antibody in cerebellar sections from juvenile WT mice revealed an extremely dense staining in the molecular layer corresponding to the highly numerous parallel fibers contacting PC distal dendritic spines (Figure 2A). The labeling pattern was similar in Susd4 KO and high-resolution microscopy and quantitative analysis confirmed that there are no significant changes in the mean density and volume of VGLUT1 clusters following Susd4 loss-of-function (Figure 2A). Electric stimulation of increasing intensity in the molecular layer allows the progressive recruitment of the PFs (Konnerth et al., 1990), and can be used to assess the number of synapses and the basic PF/PC transmission using whole-cell patch-clamp recordings of PCs on acute cerebellar slices (Figure 2B). No difference was observed in the amplitude and the kinetics of the responses to PF stimulation in PCs from Susd4 KO and control mice (Figure 2C and Figure S6). Furthermore, the probability of vesicular release in the presynaptic PF bouton, as assessed by measurements of paired pulse facilitation (Atluri and Regehr, 1996; Konnerth et al., 1990; Valera et al., 2012), was not changed at PF/PC synapses (Figure 2C). Finally, no differences in the frequency and amplitude of evoked quantal events were detected (Figure S6). Thus, in accordance with the morphological analysis, there is no major effect of Susd4 invalidation on the number and basal transmission of PF/PC synapses.
Long-term synaptic plasticity of PF/PC synapses is involved in proper motor coordination and adaptation learning (Gutierrez-Castellanos et al., 2017; Hirano, 2018; Kakegawa et al., 2018). We first assessed LTD in PF/PC synapses using conjunctive stimulation of PFs and CFs and whole-cell patch-clamp recordings of PCs in acute cerebellar slices from juvenile mice. Our protocol induced a 42% average decrease in the amplitude of PF excitatory postsynaptic currents (EPSCs) in PCs from WT mice while the paired pulse facilitation ratio was not changed during the course of our recordings (Figure 2D and Figure S6). In Susd4 KO PCs, the same LTD induction protocol did not induce any significant change in PF EPSCs during the 30 minutes recording period, showing that LTD is greatly impaired in the absence of SUSD4 (Figure 2D). LTP can be induced by high frequency stimulation of parallel fibers only, and is also involved in cerebellar dependent-learning (Binda et al., 2016; Gutierrez-Castellanos et al., 2017). In control mice, tetanic stimulation during 5 minutes induced a transient increase in transmission of about 20% and the amplitude of the response returned to baseline after only 15 minutes (Figure 2E and Figure S6). However, in the case of Susd4 KO PCs, the same protocol induced a 27% increase in transmission that was maintained after 35 minutes (Figure 2E), indicating a facilitation of LTP of PF/PC synapses in the absence of Susd4 expression.
Lack of LTD of PF/PC synapses could arise from deficient CF transmission. To test this possibility, we first crossed the Susd4 knockout mice with the Htr5b-GFP BAC transgenic line (www.gensat.org) expressing soluble GFP specifically in inferior olivary neurons in the olivocerebellar system to visualize CFs. We found that CFs had a normal morphology and translocated along the proximal dendrites of their PC target in Susd4 KO mice (Figure 3A). We then assessed whether developmental elimination of supernumerary CFs was affected by Susd4 invalidation using whole-cell patch-clamp recordings of PCs on cerebellar acute slices (Crepel et al., 1976; Hashimoto and Kano, 2003). No difference was found in the percentage of remaining multiply-innervated PCs in the absence of Susd4 (Figure 3B). We next used VGLUT2 immunostaining to label CF presynaptic boutons and analyze their morphology using high resolution confocal microscopy and quantitative image analysis. VGLUT2 immunostaining revealed the typical CF innervation territory on PC proximal dendrites, extending up to about 80% of the molecular layer height both in control and Susd4 KO mice (Figure 3C). Furthermore, the number and density of VGLUT2 clusters were not significantly different between WT and Susd4 KO mice. To test whether the lack of PF LTD was due to deficient CF transmission, we used whole-cell patch-clamp recordings of PCs in acute cerebellar slices. Contrary to what could have been expected, the typical all-or-none climbing fiber evoked EPSC was detected in PCs from Susd4 KO mice with similar kinetics (Figure S7), but with increased amplitude when compared to WT PCs (Figure 3D). Thus, the lack of CF-dependent PF/PC synapse LTD in Susd4 KO mice is not due to reduced CF/PC synapse formation or transmission. Measurements of evoked quantal events revealed an increase in the amplitude of the quantal EPSCs at CF/PC synapses from juvenile mice (Figure 3E and S7). In Susd4 KO, paired-pulse facilitation and depression at PF/PC and CF/PC synapses, respectively, are normal both in basal conditions and after plasticity induction (Figures 2C, 3D, S6D and S6F) suggesting that the changes in PF/PC synaptic plasticity and in CF/PC transmission in Susd4 KO PCs have a postsynaptic origin. Overall our results show that Susd4 loss-of-function in mice leads to a highly specific phenotype characterized by lack of postsynaptic LTD in the absence of defects in synaptogenesis and in basal transmission in PCs.
SUSD4 binds to HECT E3 ubiquitin ligases
What are the mechanisms that could allow control of long-term synaptic plasticity by SUSD4? We searched for SUSD4 molecular partners by affinity-purification of cerebellar synaptosome extracts using GFP-tagged SUSD4 as a bait (Figure 4A), followed by proteomic identification of interacting partners by liquid chromatography with tandem mass spectrometry (LC-MS/MS; Savas et al., 2014). Twenty-eight candidates were identified, several of which were functionally linked to ubiquitin ligase activity by gene ontology term analysis (Figure 4A and Table 1). Immunoblot analysis of affinity-purified extracts confirmed the interaction of SUSD4 with several members of the HECT E3 ubiquitin ligases of the NEDD4 subfamily (Figure 4C). Removal of the intracellular domain of SUSD4 (SUSD4-ΔCT mutant) prevented this interaction proving the specificity of SUSD4 binding (Figure 4B and C). Ubiquitin ligases of the NEDD4 family bind variants of PY motifs on target substrates and adaptors (Chen et al., 2017). Two potential PY binding sites are present in the intracellular tail of SUSD4 (Figure 4B). To test whether these sites mediate interaction between SUSD4 and NEDD4 ubiquitin ligases, we generated single- and double-point mutants of SUSD4 and tested their ability to interact with NEDD4 ligases in transfected HEK293 cells. While the mutation of the PPxY site (SUSD4-ΔPY mutant) abrogated binding of SUSD4 only partially, mutation of only the LPxY site (SUSD4-ΔLY mutant) or of both sites (SUSD4-ΔPY/LY mutant) completely prevented the binding to NEDD4 ubiquitin ligases (Figure 4D). These mutations did not change the level of SUSD4 expression in transfected HEK293 cells suggesting that SUSD4 itself is not a substrate for NEDD4 ubiquitin ligases. Interestingly, NEDD4 ubiquitin ligases are autoinhibited by binding of the HECT domain to their WW domains. Interaction of regulatory proteins with these WW domains has been shown to release autoinhibition and allow ubiquitination of substrates by the NEDD4 ubiquitin ligases (Wang et al., 2019). Thus, our results suggest that while the LPxY site in SUSD4 is essential for binding to NEDD4 ubiquitin ligases, the PPxY might have a regulatory role in controlling the ubiquitin ligase activity. A survey of the expression of HECT-ubiquitin ligases shows that different members of the NEDD4 subfamily are broadly expressed in the mouse brain, however with only partially overlapping patterns (Figure S8, http://mouse.brain-map.org, Allen Brain Atlas). Nedd4 and Wwp1 are the most broadly expressed and are found in neurons that also express Susd4, such as hippocampal neurons, inferior olivary neurons in the brainstem and cerebellar PCs. Thus, SUSD4 is a tether for ubiquitin ligases of the NEDD4 family and might modulate their activity in neurons.
SUSD4 regulates activity-dependent degradation of GLUA2
Ubiquitination is a post-translational modification essential for the regulation of protein turnover and trafficking in cells (Tai and Schuman, 2008). The NEDD4 family of ubiquitin ligases are known to ubiquitinate and target for degradation many key signaling molecules, including GLUA1- and GLUA2-containing AMPA receptors (Widagdo et al., 2017). SUSD4 binding to NEDD4 ubiquitin ligases, the lack of LTD at PF/PC synapses and our analysis of evoked quantal events thus suggested the involvement of SUSD4 in the regulation of postsynaptic receptor numbers, in particular, GLUA2-containing receptor numbers. Further evidence came from co-immunolabeling experiments using an anti-GLUA2 and an anti-VGLUT2 antibody on cerebellar sections followed by high-resolution microscopy. Several GLUA2 clusters of varying sizes were detected in close association with each VGLUT2 presynaptic bouton corresponding to CF synapses, while very small and dense GLUA2 clusters were found in the rest of the molecular layer corresponding mostly to GLUA2 clusters at the PF/PC synapses (Figure 5A). No obvious change in GLUA2 distribution in the molecular layer in Susd4 KO mice was found when compared to WT controls, in accordance with normal basal transmission in PF/PC synapses. Quantitative analysis of the GLUA2 clusters associated with VGLUT2 labelled CF presynaptic boutons did not reveal a significant change in the total mean intensity of GLUA2 clusters per CF presynaptic bouton (Figure 5A). However, the proportion of CF presynaptic boutons with no GLUA2 cluster was smaller in juvenile Susd4 KO mice than in WT mice (Figure 5A), partially explaining the increase in the amplitude of quantal EPSCs and CF transmission (Figure 3E).
Our model implied that a tripartite complex could be formed between SUSD4, NEDD4 ubiquitin ligases and GLUA2-containing AMPA receptors. To test this, we first co-expressed GLUA2 with various HA-tagged SUSD4 constructs (Figure 5B) in heterologous HEK293 cells and performed co-immunoprecipitation experiments (Figure 5C). GLUA2 and the ubiquitin ligase NEDD4 could both be found in extracts obtained after affinity-purification of the HA-tagged full length (FL-) SUSD4, while both were absent if HA-tagged FL-SUSD4 was replaced by a control transmembrane protein, PVRL3α (Figure 5C). NEDD4 interaction was absent when the SUSD4ΔCT construct lacking the intracellular domain was used but not with the SUSD4ΔNT construct lacking the extracellular domain, further confirming the binding of NEDD4 ubiquitin ligases to the intracellular domain of SUSD4. Strong co-immunoprecipitation of GLUA2 was detected in affinity-purified extracts from cells expressing the HA-tagged extracellular domain of SUSD4 alone (SUSD4NT construct), showing that this domain is sufficient for GLUA2 interaction. Conversely, binding to GLUA2 of a SUSD4 construct lacking the extracellular domain (SUSD4ΔNT) was lower than the binding of the extracellular domain alone (SUSD4NT), but was not completely abrogated, showing the cooperation of several domains of SUSD4 for the binding to GLUA2 (Figure 5C). In transfected cultured hippocampal neurons, clusters of GFP-tagged SUSD4 partially colocalize with GLUA2 clusters in spines (Figure 5D). Altogether our results suggest that SUSD4 acts as an adaptor protein that targets NEDD4 ubiquitin ligases to GLUA2-containing AMPA receptors to regulate their synaptic numbers.
If SUSD4 is involved in the targeting of AMPA receptors to RAB7-containing endolysosomal compartments for degradation, then it should also be found in RAB7-containing endolysosomal compartments. Co-transfection experiments in HeLA cells followed by immunolabeling showed that indeed SUSD4 is found in several intracellular compartments including the RAB7-positive endolysosomal compartments. Mutation of the PY binding site in SUSD4 decreases the colocalization of SUSD4 with RAB7 further suggesting a link between binding to NEDD4 ubiquitin ligases and SUSD4 localization in the degradation compartments (Figure S9). Given the role of NEDD4 ubiquitin ligases in the regulation of substrate degradation, SUSD4 could thus regulate GLUA2-containing AMPA receptor degradation. Indeed, total levels of GLUA2 are greatly reduced in HEK293 cells overexpressing SUSD4 when compared to control cells overexpressing a control transmembrane protein PVRL3α (Figure 5 and S10). Lack of LTD in Susd4 KO mice suggests that SUSD4 regulation of GLUA2 degradation is activity-dependent in neurons. To assess this, we set up a biochemical assay in which we induced chemical LTD (cLTD) in acute cerebellar slices (Kim et al., 2017) and measured total protein levels either in control conditions or with inhibitors of the proteasome and lysosomes (to estimate the GLUA2 degraded pool, Figure 5E). Total GLUA2 levels were not significantly different on average between WT and Susd4 KO cerebellar slices in basal conditions, in accordance with our morphological and electrophysiological analysis of PF/PC synapses (Figure S11). In slices from WT mice, chemical induction of cLTD induced a significant reduction in total GLUA2 protein levels (Figure 5E). This reduction was prevented by incubation with the proteasome inhibitor MG132 and the lysosomal inhibitor leupeptin, showing that it corresponds to the pool of GLUA2 degraded in an activity-dependent manner (Figure 5E). In slices from Susd4 KO mice, this activity-dependent degradation of GLUA2 was completely absent. Furthermore, the chemical induction of LTD did not change the total protein levels of another synaptic receptor highly present at PF/PC postsynaptic densities, GLURδ2 (GRID2), either in slices from WT or from Susd4 mice (Figure 5E and S11). Thus, SUSD4 specifically controls the activity-dependent degradation of GLUA2 subunits during synaptic LTD.
Discussion
Our study shows that the CCP domain containing protein SUSD4 starts to be expressed in various neurons of the mammalian central nervous system when synapses are formed and mature. SUSD4 can bind both ubiquitin ligases of the NEDD4 family through its C-terminus domain and the GLUA2 AMPA receptor subunit, thus allowing the formation of a tripartite complex. Susd4 loss-of-function in mice leads to deficient activity-dependent degradation of GLUA2 AMPA receptor subunits, lack of synaptic LTD in cerebellar PCs and impaired motor coordination and adaptation learning.
SUSD4 promotes long-term synaptic depression
The choice between recycling of AMPA receptors to the membrane or targeting to the endolysosomal compartment for degradation is key for the regulation of the number of AMPA receptors at synapses, as well as for the direction and degree of activity-dependent synaptic plasticity (Ehlers, 2000; Lee et al., 2002). Blocking trafficking of AMPA receptors through recycling endosomes, for example using a RAB11 mutant, prevents LTP in neurons (Park et al., 2004). Conversely, blocking the sorting of AMPA receptors to the endolysosomal compartment, for example using a RAB7 mutant, impairs LTD in hippocampal CA1 pyramidal neurons and cerebellar PCs (Fernandez-Monreal et al., 2012; Kim et al., 2017). Further support for the role of receptor degradation comes from mathematical modeling showing that in cerebellar PCs LTD depends on the regulation of the total pool of glutamate receptors (Kim et al., 2017). The GLUA2 AMPA receptor subunit, and its regulation, is of particular importance for LTD (Diering and Huganir, 2018). Phosphorylation in its C-terminal tail and the binding of molecular partners such as PICK1 and GRIP1/2 is known to regulate endocytosis and recycling (Bassani et al., 2012; Chiu et al., 2017; Fiuza et al., 2017), and mutations in some of the phosphorylation sites leads to impaired LTD (Chung et al., 2003). The molecular partners regulating the targeting for degradation of GLUA2 subunits in an activity-dependent manner during LTD remain to be identified. Our study shows that loss-of-function of Susd4 leads both to loss of LTD and of activity-dependent degradation of GLUA2 subunits. Furthermore, loss-of-function of Susd4 facilitates LTP of PF/PC synapses, and SUSD4 localization to RAB7 compartments is partly dependent on its binding to NEDD4 ubiquitin ligases. Overall our results support a model in which, during LTD, a specific molecular machinery containing SUSD4 promotes targeting of GLUA2-containing AMPA receptors to the degradation compartment in an activity-dependent manner, away from the recycling pathway that promotes LTP.
SUSD4 is a tether for NEDD4 E3 ubiquitin ligases
The degradation of specific targets such as neurotransmitter receptors must be regulated in a stimulus-dependent and synapse-specific manner in neurons, to ensure proper long-term synaptic plasticity, learning and memory (Tai and Schuman, 2008). How is this level of specificity achieved? Ubiquitination is a post-translational modification that has been shown to regulate protein degradation or signaling depending on the type of the ubiquitin chain added to lysines on the substrate protein. This modification is catalyzed by hundreds of different E3 ubiquitin ligases that belong to several families distinguished by their structure, mode of action and substrate selectivity (Zheng and Shabek, 2017). The family of HECT E3 ubiquitin ligases contains 28 enzymes including the NEDD4 subgroup that is characterized by an N-terminal C2 domain, several WW domains and the catalytic HECT domain. This subgroup of E3 ligases adds K63 ubiquitin chains to their substrate, a modification that promotes sorting to the endolysosomal compartment for degradation (Boase and Kumar, 2015). Our affinity-purification experiments identified SUSD4 as a binding partner for HECT ubiquitin ligases of the NEDD4 family and the mutation of the binding motifs in SUSD4 do not increase the amount of SUSD4 protein in cells suggesting that SUSD4 degradation itself is not regulated by NEDD4 ligases. Adaptor proteins that regulate both activation and substrate binding have been described for several E3 ligases. The activity of NEDD4 ligases depends on the opening of a closed conformation that can be controlled by the binding of several WW domains to PY motifs (Chen et al., 2017). Interestingly, mutation of only one of the two PY motifs of SUSD4 completely abrogates binding to NEDD4 while mutation on the other one only partially impairs it. Thus, SUSD4 could act as an adaptor protein that regulates NEDD4 E3 ligases activity by changing their conformation.
Adaptor proteins can also regulate substrate binding and thus participate in the specificity of signaling through E3 ubiquitin ligases. NEDD4 E3 ligases are highly expressed in neurons in the mammalian brain and have many known substrates, including ion channels and the GLUA1 AMPA receptor subunit. Their activity and substrate selectivity thus need to be finely tuned. Both GLUA1 and GLUA2 AMPA receptor subunits are ubiquitinated on lysine residues in their intracellular tails in an activity-dependent manner (Lin et al., 2011; Lussier et al., 2011; Schwarz et al., 2010; Widagdo et al., 2015). Mutation of these lysine residues decreases localization of GLUA1 and GLUA2 AMPA receptor subunits in the endolysosomal compartment in neurons (Widagdo et al., 2015). On the contrary, GLUA1 and GLUA2 subunits lack any obvious intracellular direct binding motif to the WW domain of NEDD4 ubiquitin ligases and thus an adaptor molecule would be necessary for ubiquitination by these enzymes. SUSD4 could act as a chaperone for NEDD4 binding to the GLUA2 subunit, its ubiquitination and subsequent targeting to the endolysosomal compartment for degradation. While direct modulation of GLUA2 ubiquitination by SUSD4 remains to be demonstrated, the formation of this complex could provide a simple mechanism to regulate the specificity of action of NEDD4 ubiquitin ligases in neurons and proper control of LTD. Indeed, loss-of-function of Susd4 leads to misregulation of GLUA2 activity-dependent degradation, but does not affect regulation of another postsynaptic receptor, GRID2. The SUSD4-dependent complex could be used to increase the efficiency of GLUA2 ubiquitination since it has been shown recently that membrane tethering and polymerization increases activity of NEDD4 ubiquitin ligases (Mund and Pelham, 2018). The presence of four CCP domains in the extracellular domain of SUSD4 also suggests that it could bind other proteins extracellularly. In particular, it was previously shown that SUSD4 can bind the C1Q globular domain of the complement protein C1 (Holmquist et al., 2013), a domain that is also found in presynaptic proteins of the C1Q family known for their role in synapse formation and function (Sigoillot et al., 2015; Südhof, 2018; Yuzaki, 2011). SUSD4 could thus enable fine spatiotemporal regulation of the activity of HECT ubiquitin ligases and the degradation of GLUA2-containing AMPA receptors in a trans-synaptic manner.
SUSD4 and neurodevelopmental disorders
The 1q41-42 deletion syndrome is characterized by many symptoms including intellectual deficiencies and seizures, and in a high majority of the cases the microdeletion encompasses the Susd4 gene (Rosenfeld et al., 2011). A Susd4 copy number variation has been identified in a patient with ASD (Cuscó et al., 2009). Susd4 was recently identified amongst the 124 genes with genome wide significance for de novo mutations in a cohort of more than 10,000 patients with ASD or intellectual deficiencies (Coe et al., 2019). The GRIA2 gene (coding for the GLUA2 subunit) has been found as an ASD susceptibility gene (Satterstrom et al., 2018) and mutations or misregulation of ubiquitin ligases have been found in many models of ASDs or intellectual deficiencies (Cheon et al., 2018; Lee et al., 2018; Satterstrom et al., 2018). For example, ubiquitination of GLUA1 by NEDD4-2 is impaired in neurons from a model of Fragile X syndrome (Lee et al., 2018). Understanding the precise molecular mechanisms underlying the activity-dependent degradation of GLUA2 by the SUSD4/NEDD4 complex will thus be of particular importance for our understanding of the etiology of these neurodevelopmental disorders.
Mutations in the Susd4 gene could contribute to the etiology of neurodevelopmental disorders by impairing synaptic plasticity. Deficits in LTD such as the one found in the Susd4 KO mice are a common feature of several mouse models of ASDs (Auerbach et al., 2011; Baudouin et al., 2012; Piochon et al., 2014). Susd4 loss-of-function leads to motor impairments, a symptom that is also found in ASD patients (Fournier et al., 2010). Long-term synaptic plasticity has been proposed as a mechanism for learning and memory. While in vivo evidence for the role of LTP in these processes has accumulated, the role of LTD is still discussed (Andersen et al., 2017; Kakegawa et al., 2018; Raymond and Medina, 2018; Schonewille et al., 2011). Because of the broad expression of SUSD4 and of ubiquitin ligases of the NEDD4 subfamily in the mammalian central nervous system, this complex is likely to control LTD at many synapse types and its misregulation might lead to impairments in many learning and memory paradigms. Manipulation of SUSD4’s function will thus now provide the means to test the relevance of LTD in many behavioral paradigms and its implication in the etiology of neurodevelopmental disorders.
Funding
his work was supported by funding from: ATIP-AVENIR program (RSE11005JSA to FS), Idex PSL ANR-10-IDEX-0001-02 PSL*(FS), ANR 9139SAMA90010901 (to FS and PI), ANR-15-CE37-0001-01 CeMod (to PI and FS), Fondation pour la Recherche Médicale Equipe FRM DEQ20150331748 (FS) and DEQ20140329514 (PI), and European Research Council ERC consolidator grant SynID 724601 (to FS). KI was supported by a PhD grant from the Ecole des Neurosciences de Paris (ENP) and the ENS Labex MemoLife (ANR-10-LABX-54 MEMO LIFE). JV and LR-R were supported by ANR-10-LABX-BioPsy (ANR-11-IDEX-0004-02) and ENP Foundation.
Contributions
FS, PI, LR-R, IG-C and KI designed the study and the experiments. IG-C, KI, MC, AK, FAG, JV, MS, ST, MA, OV, YN, AD, FS, SMS and MV performed the experiments, and collected the data; AT and J-LB provided the SUSD4 knockout mice and conceptual advice; FS and IG-C wrote the first draft of the manuscript; all authors read the manuscript and KI, MC, AK, MS, SMS, MV, AD, J-LB, PI and LR-R revised the manuscript.
Competing interests
Authors declare no competing interests.
Data and materials availability
All data is available in the main text or the supplementary materials.
Supplementary Information
Materials and Methods
Figs. S1 to S11
Tables 1 and S1
Acknowledgments
The authors wish to thank all the personnel from the CIRB Imaging facility, in particular P. Mailly for help with the design of the macro for GLUA2 quantification and Estelle Anceaume for image acquisition (panel 1C). We also thank the personnel from the CIRB, INCI and chronobiotron CNRS UMS 3415, IBPS and IBENS animal facilities. We would like to thank Philippe Marin for advice on proteomics analysis. Mass spectrometry experiments were carried out using facilities of the Functional Proteomics Platform of Montpellier.