Proteomic Analysis of Glycine Receptor β Subunit (GlyRβ)-interacting Proteins

Background: Glycine receptors mediate inhibitory neurotransmission in the central nervous system. Results: Syndapin I is a new interaction partner of the glycine receptor β subunit, and its down-regulation reduces synaptic glycine receptor levels. Conclusion: Syndapin I is involved in glycine receptor trafficking to and/or anchoring at inhibitory postsynapses. Significance: Disclosing new determinants of glycine receptor clustering is crucial for understanding inhibitory synapse formation. Glycine receptors (GlyRs) mediate inhibitory neurotransmission in spinal cord and brainstem. They are clustered at inhibitory postsynapses via a tight interaction of their β subunits (GlyRβ) with the scaffolding protein gephyrin. In an attempt to isolate additional proteins interacting with GlyRβ, we performed pulldown experiments with rat brain extracts using a glutathione S-transferase fusion protein encompassing amino acids 378–455 of the large intracellular loop of GlyRβ as bait. This identified syndapin I (SdpI) as a novel interaction partner of GlyRβ that coimmunoprecipitates with native GlyRs from brainstem extracts. Both SdpI and SdpII bound efficiently to the intracellular loop of GlyRβ in vitro and colocalized with GlyRβ upon coexpression in COS-7 cells. The SdpI-binding site was mapped to a proline-rich sequence of 22 amino acids within the intracellular loop of GlyRβ. Deletion and point mutation analysis disclosed that SdpI binding to GlyRβ is Src homology 3 domain-dependent. In cultured rat spinal cord neurons, SdpI immunoreactivity was found to partially colocalize with marker proteins of inhibitory and excitatory synapses. When SdpI was acutely knocked down in cultured spinal cord neurons by viral miRNA expression, postsynaptic GlyR clusters were significantly reduced in both size and number. Similar changes in GlyR cluster properties were found in spinal cultures from SdpI-deficient mice. Our results are consistent with a role of SdpI in the trafficking and/or cytoskeletal anchoring of synaptic GlyRs.


Glycine receptors (GlyRs) mediate inhibitory neurotransmission in spinal cord and brainstem. They are clustered at inhibitory postsynapses via a tight interaction of their ␤ subunits (GlyR␤) with the scaffolding protein gephyrin. In an attempt to isolate additional proteins interacting with GlyR␤, we performed pulldown experiments with rat brain extracts using a glutathione S-transferase fusion protein encompassing amino acids 378 -455 of the large intracellular loop of GlyR␤ as bait.
This identified syndapin I (SdpI) as a novel interaction partner of GlyR␤ that coimmunoprecipitates with native GlyRs from brainstem extracts. Both SdpI and SdpII bound efficiently to the intracellular loop of GlyR␤ in vitro and colocalized with GlyR␤ upon coexpression in COS-7 cells. The SdpI-binding site was mapped to a proline-rich sequence of 22 amino acids within the intracellular loop of GlyR␤. Deletion and point mutation analysis disclosed that SdpI binding to GlyR␤ is Src homology 3 domain-dependent. In cultured rat spinal cord neurons, SdpI immunoreactivity was found to partially colocalize with marker proteins of inhibitory and excitatory synapses. When SdpI was acutely knocked down in cultured spinal cord neurons by viral miRNA expression, postsynaptic GlyR clusters were significantly reduced in both size and number. Similar changes in GlyR cluster properties were found in spinal cultures from SdpI-deficient mice. Our results are consistent with a role of SdpI in the trafficking and/or cytoskeletal anchoring of synaptic GlyRs.
Postsynaptic inhibition by glycine is mediated by strychninesensitive glycine receptors (GlyRs), 3 ligand-gated chloride channels of the Cys loop receptor family (1). GlyRs are composed of ␣ and ␤ subunits (2,3), which assemble into homomeric ␣-pentamers or ␣ 2 ␤ 3 heteromeric receptors (4). Each subunit consists of an extended N-terminal extracellular domain, four transmembrane domains, and a large intracellular loop between transmembrane domains 3 and 4. The ␤ subunit (GlyR␤) provides for synaptic localization of the GlyRs by interacting with the postsynaptic scaffolding protein gephyrin (5,6). The gephyrin-binding motif (GBM) of GlyR␤ is located in the large intracellular loop and has been shown to also be important for the anterograde and retrograde dendritic transport of GlyRs, in which gephyrin functions as an adaptor for kinesin-5 and the dynein light chains 1 and 2 (Dlc1/ 2), respectively (7)(8)(9). In addition to gephyrin, the trafficking proteins vacuolar protein sorting 35 (Vps35) and neurobeachin (Nbea) have been recently found to interact with GlyR␤ involving loop residues adjacent to the GBM (10). Notably, the GlyR␤ loop sequence is predicted to also contain Src homology 3 (SH3) domain ligand-binding motifs (SBMs) (1), suggesting that SH3 domain containing proteins might be implicated in the cell biology of synaptic GlyRs.
SH3 domain proteins are involved in endocytosis and actin remodeling (11). Syndapins (Sdps) are SH3 domain containing proteins found in all multicellular eukaryotes and belong to the family of Fes/CIP4 homology Bin-Amphiphysin-Rvs (F-BAR) proteins (12). F-BAR domains can dimerize, bind negatively charged lipids, such as phosphatidylserine and phosphatidylinositol 4,5-bisphosphate, and sculpt membranes (13). At the C terminus, Sdps contain an SH3 domain, which interacts with the GTPase dynamin and other proteins involved in vesicular transport. Of the three Sdp isoforms, SdpI is expressed in brain, SdpII ubiquitously, and SdpIII in skeletal muscle, heart, lung, and weakly in brain (14). Sdps have been implicated in vesicle formation at the plasma membrane, the trans-Golgi network, and the recycling endosome, as well as in actin polymerization (14 -19). In SdpI-deficient (SdpI Ϫ/Ϫ ) mice, synaptic vesicle endocytosis is significantly impaired, resulting in enhanced hippocampal network activity and seizures (20). Notably, in SdpI Ϫ/Ϫ neurons evoked inhibitory postsynaptic currents are more severely reduced than the excitatory ones, suggesting a specific role in inhibitory neurotransmission of this Sdp isoform (20).
In an attempt to identify novel GlyR␤-binding proteins that interact with the predicted SBMs and might be implicated in the molecular organization of inhibitory glycinergic postsynapses, we performed proteomic screens using a glutathione S-transferase (GST) fusion protein containing amino acids 378 -455 of GlyR␤, i.e. the C-terminal portion of its large intracellular loop. Here, we report that SdpI and the two known splice variants of SdpII interact with the GlyR␤ loop by recognizing an SBM that lies adjacent to but does not include the GlyR␤ 378 -426 region. Mapping experiments revealed that the SBM of GlyR␤ interacts with the SH3 domain of SdpI. Upon miRNA-mediated knockdown of SdpI in cultured rat spinal cord neurons, we observed a reduction in both the size and number of postsynaptic GlyR clusters. Similar results were obtained with spinal cord cultures prepared from SdpI Ϫ/Ϫ mice. Our data indicate that SdpI participates in the trafficking and/or cytoskeletal anchoring of synaptic GlyRs.
His 6 -tagged Sdp constructs were expressed in Escherichia coli C41 DE3 (25) and all other constructs in E. coli BL21 DE3 (Merck). Expression and preparation of clear lysates were performed as described previously (26).
GST pulldowns of recombinant proteins were done as described previously (10); a mouse monoclonal ␣-His 6 antibody (1:1000, Merck) was used for Western blotting. The binding activity of GST-GlyR␤ constructs containing the GBM was monitored by pulldown of the His 6 -tagged E-domain of gephyrin. All GST pulldown experiments were repeated three times. The GST pulldowns for isolation of GlyR␤-interacting proteins from brain extracts using GlyR␤(378 -455), the preparation of tissue extracts, SDS-PAGE separation and mass spectrometry (MS) were carried out as described for GlyR␤(378 -426) (10,26).
Coimmunoprecipitation-Detergent extracts were prepared from brainstems of adult (4 months old) C57BL/6 mice by homogenizing the tissue at a w/v ratio of 1:3 in lysis buffer (20 mM HEPES, pH 7.5, 100 mM KCl, 5 mM EGTA, 5 mM MgCl 2 and 50 mM NaCl) supplemented with 1% (w/v) Triton X-100, 5 mM DTT, and complete protease inhibitor mixture (1 tablet/50 ml; Roche Diagnostics) using a Dounce homogenizer. The homogenate was left for 1 h on an overhead rotator at 4°C and centrifuged at 10,000 ϫ g for 10 min. The resulting supernatants were pre-cleared by incubation with protein A-or protein G-Sepharose for 1 h. Affinity-purified anti-SdpI from guinea pig (16) or unrelated guinea pig IgGs was immobilized on protein A-Sepharose (Sigma), and mAb4 or unrelated mouse IgGs were immobilized on protein G-Sepharose (Sigma) in the presence of 5% (w/v) bovine serum albumin. After three washes with lysis buffer containing 1% (w/v) Triton X-100, the resin was incubated overnight with 0.8 -1.0 ml of pre-cleared brainstem homogenate at a concentration of 2 mg of protein/ml. After five washes with 1% (w/v) Triton X-100-containing lysis buffer, bound proteins were eluted with SDS sample buffer and analyzed by SDS-PAGE and immunoblotting.
TAT Peptide Competition Experiments-All peptides used contained an N-terminal fluorescein conjugate and the TAT sequence as shown by underlines, and the amino acids important for SdpI binding are shown in italics as follows: TAT-GlyR␤ 22 , YGRKKRRQRRRGKPQAKNKKPPPAKPVIPTAAK; TAT-scrambled, YGRKKRRQRRRAQKVAPGPITKPKPNPA-KAPKK. The TAT-scrambled peptide contains the same amino acids as TAT-GlyR␤ 22 , but the GlyR␤ residues are randomly ordered, and no SBM is present. Peptides were purchased from Biosyntan GmbH, Berlin, Germany, and had a purity of at least 95%. GST pulldowns were performed as explained above, except that TAT-GlyR␤ 22 was added in concentrations of 1-100 M during the incubation of GST-GlyR␤(378 -455) with His 6 -SdpI. TAT-scrambled was used as control.
Molecular Modeling-The SdpI-GlyR␤ model was generated by Modeler9 version 5-1 (27) using the x-ray structure 2DRM and the nuclear magnetic resonance solution structure 1RLQ as templates for the rat SdpI SH3 domain and the rat GlyR␤ SBM, respectively. After energy minimization, the model was subjected to a 50-ns molecular dynamics simulation run. The simulation was run with Gromacs (28), version 4.6, and the following parameters: amber03 (29) all-atom force field, tip4p water model, and 0.15 M concentrations of Na ϩ and Cl Ϫ ions. From the whole trajectory an average structure was generated, which was energy-minimized (steepest descent, 800 ps) subsequently.
SdpI Knockdown-pAAV-6P-SEWB was a gift of Sebastian Kügler (Göttingen, Germany) and provides a neuron-specific synapsin promoter (31). siRNA-resistant Myc-SdpI-siR (32) was cloned via NheI-HindIII sites into pAAV-6P-SEWB. The pDP1rfps and pDP2rfps plasmids (33) were provided by Jürgen Kleinschmidt (Heidelberg, Germany). Primers TGCTGTC-CTCGTTCAGTAGGCTGTTCGTTTTGGCCACTGACTG-ACGAACAGCCCTGAACGAGGA (sense) and CCTGTC-CTCGTTCAGGGCTGTTCGTCAGTCAGTGGCCAAAA-CGAACAGCCTACTGAACGAGGAC (antisense) were designed for shRNAmiR according to Invitrogen's instructions based on the SdpI-siRNA sequence (32) GAACAGCCTACT-GAACGAG (underlined sequences of sense and antisense primers are derived from the siRNA sequence). Primers were annealed and cloned into pcDNA6.2-GW-EmGFP-miR in line with the manufacturer's instructions (Invitrogen). A control miRNA construct was provided by Invitrogen. The GFP-shRN-AmiR regions of both constructs were cloned into pAAV-6P-attR (gift from Sebastian Kügler (Göttingen, Germany)) containing a synapsin promoter by recombination using BP-clonase II, pDONR221, and LR-clonase II (Invitrogen) according to the manufacturer's protocol. Plasmids control-miR-pAAV-6P-attR (control-miR), SdpI-miR-pAAV-6P-attR (SdpI-miR), and rescue construct Myc-SdpI-siR-pAAV-6P (SdpI-siR) were individually cotransfected with pDP helper plasmids (pDP1 and pDP2) in equimolar amounts. Transfection into human embryonic kidney 293T cells was achieved by calcium phosphate transfection as described previously (33). Forty eight hours after transfection, the cells were collected, and packaged viruses were released into 50 mM Tris, pH 8.0, 150 mM NaCl by three cycles of freeze-thawing on dry ice/ethanol. Viruses were purified by iodixanol density centrifugation as described previously (34) and concentrated using a 100-kDa cutoff Amicon centrifugal filter unit (Millipore, Schwalbach, Germany). Infectious unit titers were determined in rat primary neuron cultures for each recombinant adeno-associated virus preparation. Equivalent infectious units of the control-miR and SdpI-miR were diluted in 10 l of PBS and added to the medium of spinal cord neuron cultures at day in vitro (DIV) 7. For rescue experiments, cultures were infected with a mixture of 10 l of SdpI-miR and 10 l of Myc-SdpI-siR. By DIV21, neurons were washed with PBS and either harvested in PBS for Western blot analysis or fixed for immunostaining.
Immunofluorescence Staining-Cultures of dissociated mouse (C57BL/6) and rat (Wistar; Charles River, Sulzfeld, Germany) spinal cord neurons were prepared from embryonic day 14.5 embryos and maintained as described previously (35). SdpI Ϫ/Ϫ mice were bred in the C57BL/6 background and genotyped as detailed previously (20). When comparing results obtained with WT and SdpI Ϫ/Ϫ embryos or mice, both genotypes were always derived from the same litter. The protocols used for culturing and immunostaining dissociated spinal cord neurons from rat have been published previously (10). Immunohistochemistry on brainstem slices prepared from adult mice was performed as described previously (36). Quantifications of cluster densities and sizes were performed on 140 -400 clusters per condition/genotype in spinal cord cultures and 6000 -10,000 clusters per genotype in brain slices, respectively.
Immunofluorescence and Image Analysis-GFP, DsRed2, and mRFP were visualized by endogenous fluorescence. For colocalization analysis, images were obtained using an AxioImager microscope equipped with an ApoTome (Zeiss, Göttingen, Germany). In the ER retention assay, the areas colocalizing in merged images were normalized to the areas in the red channel. Co-localization experiments of SdpI puncta with synaptic markers in rat spinal cord neurons were performed as described previously (10). For the quantification of punctate immunofluorescence (density and average size of synaptic clusters), images of immunostained spinal cord neurons were captured as described previously (39), and 50 m neuritic segments emanating from the somata were marked and evaluated. Binarization, thresholding, and quantification were performed with the ImageJ 1.42q software (National Institutes of Health, Bethesda). If not indicated otherwise, quantitative data repre-sent means Ϯ S.E. from 30 randomly selected cells (10 per independent experiment). Statistical significance was evaluated with two-tailed Student's t test for paired comparisons and analysis of variance with post hoc Bonferroni correction for multiple comparisons.

RESULTS
Sdps Interact with GlyR␤-In an attempt to search for novel GlyR␤-binding proteins, we employed a fragment of the GlyR␤ loop that included its predicted SBMs (residues 378 -455) in GST pulldown experiments with rat brain detergent extracts, as used previously for identifying the GlyR␤ interaction partners Vps35 and Nbea (10). Fig. 1A shows that GST-GlyR␤(378 -455) but not GST reproducibly isolated a protein band with an apparent molecular mass of about 52 kDa that had not been detected in previous screens with GST-GlyR␤(378 -426) (10,26). By matrix-assisted laser desorption/ionization-time of flight spectrometry, this protein was identified as SdpI (Table  1). A representative mass spectrum in its deisotopized form is shown in Fig. 1B.
To examine whether the interaction also occurs between native GlyRs and SdpI, we performed coimmunoprecipitation experiments with detergent extracts from adult mouse brainstem. Antibodies against SdpI precipitated not only their native antigen (data not shown) but also GlyR␣ immunoreactivity (Fig. 1E). Inversely, a monoclonal antibody against GlyR␣ (mAb4a) coisolated SdpI (Fig. 1F). These results indicate that native GlyRs present in brainstem extracts interact with SdpI.
Identification of SdpI-and GlyR␤-binding Sites-SdpI consists of an N-terminal F-BAR domain, followed by a region containing NPF motifs, and a C-terminal SH3 domain ( Fig. 2A). To delineate whether the SH3 domain of SdpI is involved in GlyR␤ binding, we introduced a point mutation known to interfere with binding of the SH3 domain to its ligands, SdpI P434L (24), and also deleted the SH3 domain in the SdpI ⌬SH3 construct ( Fig. 2A). GST pulldowns disclosed that the binding of both SdpI P434L and SdpI ⌬SH3 to GST-GlyR␤(378 -455) was almost completely abrogated as compared with wild type (WT) SdpI (Fig. 2B). This indicates that an intact SH3 domain of SdpI is essential for tight interaction. To confirm this result, we used an inverted GST pulldown with GST-tagged SdpI constructs and maltose-binding protein (MBP)-tagged GlyR␤ constructs. We found that both full-length SdpI and the isolated SH3 domain of SdpI, but neither the SH3 domain with the P434L point mutation, the full-length SdpI with this mutation, nor SdpI without the SH3 domain, were able to bind MBP-GlyR␤(378 -455) (Fig.  2C). To exclude false positive results due to the presence of the MBP tag, binding of MBP-GlyR␤(378 -426) was also tested. We detected a strong binding of MBP-GlyR␤(378 -426) to gephyrin, showing that the protein fragment is properly folded Upon Western blotting and subsequent immunodetection with the anti-His 6 antibody, weak binding to GST was observed in this experiment (bottom panel, right lane). D, immunodetection of GST pulldowns of His 6 -tagged SdpII-s (first three lanes) and SdpII-l (last three lanes) obtained with the indicated GST fusion proteins. Both the long and the short splice variant of SdpII interact with GlyR␤(378 -455) but not with the shorter fragment GST-GlyR␤(378 -426) and the negative control GST. *, GST. E, coimmunoprecipitation of GlyR␣ immunoreactivity from brainstem detergent extract by anti-SdpI; for immunodetection, mAb4 was used. Left lane, eluate from anti-SdpI beads; middle lane, eluate from control IgG beads; right lane, input. Note that GlyR was hardly detectable in the input fraction and control IgG eluate but strongly detected in the anti-SdpI eluate. F, coimmunoprecipitation of SdpI immunoreactivity from brainstem detergent extract by anti-GlyR (mAb4); for immunodetection, anti-SdpI was used. Note prominent SdpI band in the eluate from the anti-GlyR beads.
but not to any of the SdpI constructs (Fig. 2C). Together, these results indicate that an intact SH3 domain is both necessary and sufficient for the binding of SdpI to GlyR␤(378 -455).
The binding site of SdpI within GST-GlyR␤(378 -455) was mapped using a deletion approach. Stepwise removal of the first 50 amino acids, including the GBM (GST-GlyR␤(403-455) and GST-GlyR␤(427-455)), did not affect SdpI binding (Fig.  2D). A construct lacking the last 15 amino acids of GlyR␤(378 -455), including a part of the SBM (GST-GlyR␤(378 -440)), exhibited a slightly reduced binding. We observed that 22 amino acids from position 427 to 448 containing the entire SBM were sufficient for a strong interaction (Fig. 2D). This result was confirmed by peptide competition (Fig. 2E). A TAT peptide corresponding to sequence 427-448 of GlyR␤, TAT-GlyR␤ 22 , but not a TAT peptide containing the same amino acids but randomly ordered (TAT scrambled), inhibited SdpI binding to GST-GlyR␤(378 -455). This result shows that the interaction between the cytosolic loop of GlyR␤ and the SH3 domain of SdpI is sequence-specific.
In a typical (R/K)XXPXXP SH3 ligand-binding motif, Arg/ Lys provides additional binding energy through electrostatic interactions (40). Amino acid sequence analysis revealed the presence of such a motif in the SBM of GlyR␤, i.e. residues Lys-435, Pro-438, and Pro-441. Therefore, we designed dou-ble point mutations to examine its importance. The GST-GlyR␤ P438A/P441A (PPAA) double substitution led to a dramatic decrease in SdpI binding. For Lys-435, a double mutant K434A/ K435A (KKAA) was created to avoid compensatory effects due to the neighboring Lys-434 (Fig. 3A); this led to a complete loss of SdpI binding (Fig. 3B). To demonstrate that this result was specifically due to the substitutions introduced and not caused by more general conformational effects, all aforementioned deletion and mutant constructs were examined for gephyrin binding. Fig. 3B (lower panels) shows that gephyrin interacted with all fusion proteins examined to a similar extent as found with GST-GlyR␤(378 -455). Taken together, these results indicate that the KXXPXXP motif is indeed the SdpI-binding site.
To better understand how GlyR␤ interacts with SdpI, we made an attempt to model the regions involved based on known

Identification of SdpI by mass spectrometry
Proteins with an apparent molecular mass of 52 kDa from three independent experiments were analyzed by MALDI-TOF mass spectrometry. Sequence coverage at the amino acid level and accuracy of the measurements are indicated (ppm, parts per million). p, the probability that the observed match of the mass spectrum with SdpI from the database is a random event, is also given. Additionally, the number of peptides matching SdpI is indicated. Of the different Sdp isoforms detected, SdpI is the only isoform detected in all three pulldown experiments analyzed. Of the few peptides corresponding to SdpII and SdpIII, only two are clearly not derived from SdpI.  crystal structures. A three-dimensional structure of SdpI has been solved more recently (42). However, in this structure the SH3 domain interacts with two right-handed ␣-helices of the F-BAR domain, whereas molecular dynamics simulations predict the SBM of GlyR␤ to contain a left-handed polyproline type II helix (Fig. 3C)   In summary, the model shown in Fig. 3C supports our biochemical results.

GlyR␤ and Sdps Colocalize in Mammalian Cells-The results presented above show that Sdps and GlyR␤ interact in vitro.
To demonstrate that this interaction also occurs in mammalian cells, we coexpressed these proteins in COS-7 cells. In the absence of GlyR␣ subunits, GlyR␤ is known to be retained in the ER (6). Based on this observation, we used an ER retention assay for monitoring Sdp interactions with GlyR␤. First, we confirmed the validity of this assay by coexpressing mRFP-GlyR␤ in COS-7 cells together with GFP-gephyrin. In agree-ment with earlier observations (6), this resulted in a significant colocalization of both proteins in the ER, whereas a previously described gephyrin mutant unable to interact with GlyR␤ (23) formed aggregates in the cytosol (data not shown). Upon coexpression of Myc-tagged SdpI with mRFP-GlyR␤, Myc-SdpI was found to colocalize with mRFP-GlyR␤ to a much higher extent (32.3 Ϯ 2.8% area colocalizing) than obtained upon coexpression of Myc-SdpI with a vector staining the ER (DsRed ER; 11.4 Ϯ 1.7%, p Ͻ 0.01; Fig. 4B) or compared with the colocalization of Dlc1, a Myc-tagged gephyrin-binding protein (7), with mRFP-GlyR␤ (9.5 Ϯ 3.8%, p Ͻ 0.05) (Fig. 4, A and B, and data not shown). Similarly, significant colocalization with mRFP-GlyR␤ was observed for both SdpII-l (33.5 Ϯ 4.3% versus 6.2 Ϯ 2.9% for SdpII-l with DsRed ER, p Ͻ 0.01; Fig. 4C) and SdpII-s (29.2 Ϯ 2.9% as compared with 6.0 Ϯ 0.8% for SdpII-s with DsRed ER ; Fig. 4D). These results demonstrate that Sdps can interact with GlyR␤ in mammalian cells.

Synaptic Localization of SdpI in Cultured Rat Spinal Cord
Neurons-A prerequisite for a physiological role of SdpI binding to GlyR␤ is the colocalization of both proteins in neurons. Initial Western blot experiments had revealed that SdpI is expressed not only in brain but also in spinal cord, where GlyRs are abundant (data not shown). We hence double-immunostained cultured embryonic rat spinal cord neurons at DIV11-13, i.e. a time point when many glycinergic synapses are known to be formed (43), with antibodies specific for SdpI and different synaptic marker proteins. This revealed punctate SdpI immunoreactivity in the soma and along the dendrites (Fig. 5A). A considerable fraction of the dendritic SdpI puncta colocalized or overlapped with glycinergic postsynapses (27 Ϯ 2%; Fig.   5B), as identified by immunostaining with a GlyR-specific monoclonal antibody. Notably, the fraction of SdpI puncta apposed to VIAAT immunoreactive puncta, i.e. inhibitory nerve terminals visualized by an antibody specific for this synaptic vesicle protein ("VIAAT boutons"), was even higher (40 Ϯ 3%, p Յ 0.05) and identical to the extent of colocalization found with gephyrin clusters (40 Ϯ 9%, p Ͼ 0.05). This suggests that SdpI is found not only at glycinergic but also at GABAergic and mixed glycinergic/GABAergic (44) synapses. To unravel whether the occurrence of numerous nonapposed SdpI puncta in these stainings might reflect the presence of SdpI at excitatory synapses, we repeated these double-immunolabeling experiments at a later developmental stage (DIV20 -22) and for a larger number of marker proteins (Fig. 5C). This confirmed the partial colocalization of SdpI punctate staining with GlyR (28 Ϯ 3%), gephyrin (34 Ϯ 2%), and VIAAT (39 Ϯ 4%) immunoreactivities. Additionally, we found appositions to other established markers of inhibitory and excitatory synapses, e.g. the 67-kDa isoform of the GABA-synthesizing enzyme, GAD67, that stains presynaptic inhibitory terminals (32 Ϯ 5%) and the postsynaptic scaffolding protein PSD-95 localized in the postsynaptic densities of excitatory synapses (36 Ϯ 3%). These results are consistent with SdpI being present at glycinergic, GABAergic, and glutamatergic synapses.
SdpI Knockdown Reduces GlyR Cluster Density and Size in Cultured Spinal Neurons-The results described above indicated that in spinal cord neurons SdpI is present at a significant fraction of GlyR-containing inhibitory synapses. To determine whether SdpI might contribute to GlyR synaptic trafficking and/or localization, we performed acute SdpI knockdown experiments in cultured spinal cord neurons using a vectorbased microRNA-adapted short hairpin approach. The recombinant adeno-associated virus system (45) was employed for efficient delivery of the knockdown vectors at DIV7. Western blotting at DIV20 showed that infection with the SdpI-miR virus reduced SdpI expression to 39 Ϯ 16% of the level found in sister cultures infected with a control-miR virus (Fig. 6, A and B). This down-regulation could be overcome by a siRNA-resistant Myc-tagged SdpI construct (SdpI-siR), which produced strong SdpI overexpression when coinfected with the SdpI-miR virus (812 Ϯ 469% of control) but also cell death due to viral overload, as indicated by a major loss of anti-␤3-tubulin (Fig. 6A).
To examine whether the viral SdpI-miR knockdown specifically affects GlyRs, we also performed SdpI knockdown experiments, in which other marker proteins of inhibitory synapses were analyzed. However, in contrast to the changes in GlyR Intense SdpI staining is seen ubiquitously in the soma and enriched at some locations near the plasma membrane. In addition, SdpI immunoreactive puncta are found along neuronal processes, often apposed to gephyrin clusters. B, right panel, spinal cord neurons were stained for SdpI (red) and GlyR (top row, green), gephyrin (middle row, green), and VIAAT (bottom row, green) at DIV11-13. Left panel, quantification of the colocalizations shown exemplarily in the right panel. The colocalization of VIAAT with SdpI was significantly higher than that with the GlyR. *, p Յ 0.05. C, spinal cord neurons analyzed at a later stage (DIV20 -22) and for even more synaptic marker proteins. Again, SdpI (red) exhibited clear appositions to VIAAT (green), gephyrin (green), and GlyR (green) immunoreactive puncta. In addition, comparable fractions of the structures positive for GAD67 (green), a presynaptic marker for GABAergic terminals, and PSD-95 (green), a postsynaptic protein of excitatory synapses, were apposed to SdpI. Data represent means Ϯ S.E. from three independent experiments with n ϭ 10 neurons per condition assay. Scale bars, 10 m.
cluster properties observed upon SdpI-miR infection, neither significant reductions in the relative densities of gephyrin clusters (89 Ϯ 18% of control; p Ͼ 0.05) and VIAAT immunoreactive boutons (97 Ϯ 14% of control; p Ͼ 0.05) nor in their average sizes (73 Ϯ 21% of control for gephyrin clusters; 74 Ϯ 13% of control for VIAAT boutons; both p Ͼ 0.05) were found (data not shown). Thus, SdpI deficiency appears to specifically affect GlyR cluster properties. Notably, SDS-PAGE followed by immunoblotting with mAb4a failed to reveal reductions in GlyR␣ protein levels in the SdpI-miR-infected cultures (data not shown). Hence, the reduced density and size of GlyR clusters observed upon SdpI knockdown are not due to enhanced receptor degradation.
GlyR Cluster Sizes and Densities In Brainstem Sections and Cultured Neurons Derived from SdpI Ϫ/Ϫ Mice-To further corroborate a role of SdpI in synaptic GlyR trafficking and/or localization, we also analyzed the densities and sizes of GlyR clusters and VIAAT boutons in brainstem sections prepared from adult SdpI Ϫ/Ϫ and WT littermates. However, no significant differences between genotypes could be detected (GlyR clusters/ 10,000 m 2 , 738 Ϯ 128 in WT versus 609 Ϯ 57 in SdpI Ϫ/Ϫ ; GlyR cluster size, 0.355 Ϯ 0.019 m 2 in WT versus 0.320 Ϯ 0.021 m 2 in SdpI Ϫ/Ϫ ; VIAAT boutons/10,000 m 2 , 851 Ϯ 39 in WT versus 727 Ϯ 89 in SdpI Ϫ/Ϫ ; VIAAT bouton size, 0.520 Ϯ 0.053 m 2 in WT versus 0.509 Ϯ 0.027 m 2 in SdpI Ϫ/Ϫ ; all p Ͼ 0.05; see Fig. 7, A and B). This could not be attributed to an up-regulation of any of the other less abundant Sdp isoforms, SdpII and SdpIII, in the mutant animals. A semi-quantitative analysis of immunoblots of brainstem detergent extracts prepared from adult SdpI Ϫ/Ϫ (n ϭ 5) and WT (n ϭ 6) animals revealed relative expression levels (normalized to actin) of 96 Ϯ 8% for SdpII and of 97 Ϯ 8% for SdpIII, as compared with WT (data not shown).
Because GlyRs accumulate at inhibitory synapses primarily during early developmental stages, we examined GlyR cluster formation in spinal cord cultures prepared from SdpI Ϫ/Ϫ mice (20) at embryonic day 14.5. After 21 days of in vitro differentiation (43), the average size of GlyR clusters was found to be significantly lower in SdpI Ϫ/Ϫ neurons as compared with that in WT cells (0.340 Ϯ 0.021 m 2 in WT versus 0.240 Ϯ 0.018 m 2 in SdpI Ϫ/Ϫ ; p Յ 0.05; see Fig. 7, C and D). Furthermore, the average number of GlyR clusters per 50-m dendrite was lower in SdpI Ϫ/Ϫ as compared with WT neurites (32 Ϯ 1 in WT versus 24 Ϯ 2 in SdpI Ϫ/Ϫ ; p Յ 0.05). In contrast to these alterations seen for synaptic GlyRs, no significant differences were observed in the average sizes of both postsynaptic gephyrin clusters (0.537 Ϯ 0.025 m 2 in WT versus 0.518 Ϯ 0.017 m 2 in SdpI Ϫ/Ϫ ) and VIAAT-positive presynaptic terminals (0.450 Ϯ 0.018 m 2 in WT versus 0.431 Ϯ 0.069 m 2 in SdpI Ϫ/Ϫ ) and their corresponding densities (29 Ϯ 1 per 50-m dendrite in WT versus 31 Ϯ 1 in SdpI Ϫ/Ϫ for gephyrin clusters; and 27 Ϯ 3 in WT versus 27 Ϯ 5 in SdpI Ϫ/Ϫ for VIAAT boutons; both p Ͼ 0.05) (Fig. 7, C and D). Together, the results obtained from the SdpI Ϫ/Ϫ brainstem sections and cultured spinal neurons are consistent with SdpI contributing primarily to the formation of synaptic GlyR clusters at early stages of neuronal differentiation and synaptogenesis.

DISCUSSION
In this study, we identified SdpI by affinity binding and MS as a novel binding protein of GlyR␤. SdpI coimmunoprecipitates with native GlyR from brainstem detergent extracts. Using deletion approaches and point mutation analysis, SdpI binding to GlyR␤ was found to be mediated by a classical SH3 domainligand interaction and thus defines the first intracellular protein binding to the previously predicted SBMs of GlyRs (1). Upon coexpression in COS-7 cells, SdpI and both splice variants of SdpII colocalized with GlyR␤. In cultured spinal cord neurons containing reduced levels of or lacking SdpI, the mean sizes and densities of postsynaptic GlyR clusters were decreased, whereas in adult mice SdpI deficiency did not affect GlyR clusters. Our results are consistent with a function of SdpI in the trafficking and/or cytoskeletal anchoring of synaptic GlyRs that is redundant at adult stages.
Molecular Basis of Sdp Binding to GlyR␤-The SH3 domains are highly conserved between SdpI, SdpII, and SdpIII (15,46). This explains why we observed binding not only of SdpI but also of SdpII to GlyR␤ and suggests that SdpIII may also interact with GlyR␤.
The binding region of GlyR␤ contains a KKXXPXXP motif. The proline residues of this motif are likely to form a lefthanded polyproline type II helix, whereas its lysine residues are thought to provide binding specificity by electrostatically inter-acting with residues of the SH3 domain (40). Indeed, our modeling suggests that lysine 435 interacts with glutamate residues 397 and 400 of the SH3 domain of SdpI; this might help to orient the ligand motif within the binding groove of the SH3 domain (40). The KKXXPXXP motif is highly conserved and found not only in rat and human but also in chick, fish, and frog GlyR␤ sequences. This conservation underlines its importance and suggests that the interaction of Sdps with GlyR␤ occurs in all vertebrates.
Polyproline type II helices provide the structural scaffold for SBM-SH3 domain interactions. Molecular dynamics simulations suggested that, similar to the corresponding proline of the template structure 1RLQ (47), Pro-441 of GlyR␤ is directly involved in interactions with the SH3 domain. However, only FIGURE 7. GlyR clusters in adult brainstem and cultured spinal cord neurons from SdpI ؊/؊ mice. A, immunostainings for GlyR (green) and VIAAT (red) in brainstem sections from adult WT and SdpI Ϫ/Ϫ mice. GlyR and VIAAT cluster densities and sizes appeared comparable in SdpI Ϫ/Ϫ and WT brainstems. Scale bar, 10 m. B, quantification of experiments as shown in A. No significant differences in GlyR cluster and VIAAT bouton densities and sizes were found between WT and SdpI Ϫ/Ϫ sections (all p Ͼ 0.05; data represent means Ϯ S.E. from three animals analyzed per genotype). C, immunostainings showing reduced mean GlyR cluster size and density in SdpI Ϫ/Ϫ spinal cord neurons. Spinal cord neurons were prepared from WT or SdpI Ϫ/Ϫ mouse embryos and stained at DIV21 for GlyR (green channel in upper panels), VIAAT (red channel in upper and lower panels), and gephyrin (green channel in lower panels). Scale bars, 5 m. D, quantifications of densities and sizes of GlyR and gephyrin clusters and of VIAAT boutons, determined from immunostained neurons as shown in C. Note significant reductions in GlyR cluster density and size in SdpI Ϫ/Ϫ as compared with WT spinal neurons (*, p Յ 0.05). In contrast, gephyrin cluster and VIAAT bouton densities and sizes were not significantly different between genotypes (all p Ͼ 0.05; 20 cells each analyzed per genotype in three independent experiments). n.s., not significant. minor effects were seen upon its deletion or point mutation; this likely reflects an only partial contribution of Pro-441 to SdpI binding in addition to the side chain interactions mediated by Pro-438 and lysines Lys-434 and Lys-435. Notably, the binding of SdpI to the GTPase dynamin depends on a similar motif, RRXPXXP, in which arginine residues provide positive charges for SdpI association (18). Recently, KXRAPXPP motifs have been implicated in SdpI binding to the actin nucleator cordon bleu (48). A comparison of all these SdpI-binding motifs reveals that dynamin and GlyR␤ and two of the three cordon bleu-binding sites all share the well known motif (R/K)XXPXXP involved in SH3 domain binding of polyproline type II helices (40).
It is noteworthy that the SBM of GlyR␤ as identified here (residues 427-448) is distinct from, but located adjacent to, the other known intracellular protein binding regions, i.e. the GBM and the Vps35 and Nbea interaction sites, which all lie within GlyR␤(378 -426) (5, 6, 10). Thus, SdpI and these other proteins might bind simultaneously to a single GlyR␤ subunit. This may be of physiological relevance, as GlyRs are known to interact with gephyrin not only at synapses but also at different stages of intracellular transport (8,9). Moreover, pentameric synaptic GlyRs are thought to contain three copies of GlyR␤ (4). Hence, simultaneous binding of different interacting proteins by a single pentameric GlyR may occur at variable stoichiometries.
Localization of SdpI in Spinal Cord Neurons-So far, SdpI has been studied pre-and postsynaptically at excitatory synapses in rat hippocampus (24,48,49) and presynaptically at the reticulospinal synapse in lamprey (19). Additionally, SdpI has been found in parvalbumin-positive inhibitory interneurons of the mouse brain (20). Here, we confirmed the presence of SdpI at excitatory synapses, as revealed by apposition to PSD-95 (Fig.  5C). Additionally we found SdpI immunoreactivity at inhibitory synapses as indicated by its vicinity to GlyR clusters (Fig.  5B). Notably, at DIV11-13 colocalization with VIAAT boutons was significantly higher than with GlyR clusters (Fig. 5B). This finding suggested that SdpI is present not only at glycinergic but also at other inhibitory synapses. Indeed, at DIV20 -22 we saw colocalization of SdpI with GAD67, the presynaptically localized GABA-synthesizing enzyme, to a similar extent as compared with VIAAT, gephyrin, and the GlyR (Fig. 5C). Our results are consistent with SdpI being present at glutamatergic, glycinergic, and GABAergic synapses of spinal cord neurons.
Putative Functions of SdpI in the Trafficking and Cytoskeletal Anchoring of Synaptic GlyRs-Sdps have been implicated in vesicle formation at the plasma membrane and the trans-Golgi network and in endocytotic receptor recycling (15)(16)(17)(18)(19)(20). Additionally, SdpI has been reported to mediate endocytotic removal of NR3A containing N-methyl-D-aspartate (NMDA) receptors from excitatory postsynapses (49). These reports are consistent with SdpI regulating dynamin-driven fission of endocytotic and transport vesicles from their donor membranes (15,17,18,20). In this study, we observed decreases in GlyR cluster size and density upon SdpI knockdown or gene inactivation in cultured neurons, whereas GlyR expression levels as detected by immunoblotting were unchanged under knockdown conditions. Our results are difficult to reconcile with a function of SdpI in GlyR internalization or degradation but are consistent with its contribution to the formation of GlyR containing transport vesicles or other membrane fissiondependent trafficking steps occurring after the assembly of the receptor pentamer required for GlyR exit from the ER (50). As schematically depicted in Fig. 8, right, SdpI might contribute to the budding of GlyR transport vesicles from the Golgi apparatus. Alternatively, local recycling mechanisms for synaptic GlyRs might exist in dendrites which, in analogy to the synaptic re-insertion of excitatory 2-amino-3-(3-hydroxy-5-methylisoxazol-4-yl)propanoic acid (AMPA) receptors from an intracellular pool localized in dendritic spines (51), allow for the adjustment of synaptic strength of individual inhibitory synapses to changes in neuronal activity (Fig. 8, center). Such receptor recycling processes are likely to involve an endosome-derived recycling vesicle pool, from which GlyR containing vesicles are generated for re-incorporation into the plasma membrane in a SdpI-dependent membrane fission reaction. Consistent with this proposal, previous reports have shown that SdpI is required for synaptic vesicle recycling in presynaptic nerve terminals (19,20).
In addition to eliminating functions of SdpI in the intracellular trafficking and/or recycling of GlyRs as proposed above, SdpI deficiency may directly affect the organization or stability of GlyR clusters at inhibitory postsynapses. Previous studies have shown that the pharmacological disruption of the actin cytoskeleton leads to a loss of gephyrin and GlyRs from synaptic sites (52,53). Moreover, the actin regulatory proteins profilin and ena/VASP interact with gephyrin and have been implicated in the regulation of synaptic GlyR densities by microfilaments (39,54,55). SdpI is known to bind the Arp2/3 complex activator neural Wiskott-Aldrich syndrome protein (N-WASP), thereby releasing N-WASP's auto-inhibition and thus enhancing actin polymerization (32). Consequently, SdpI knockdown or deficiency should reduce N-WASP activity and thereby result in a weakening of the local F-actin meshworks present at postsynaptic sites, which in turn should decrease GlyR cluster density and/or size (see Fig. 8, center top). This local destabilization of synaptic F-actin might be further enhanced by additional effects of reduced SdpI levels on the actin nucleator cordon bleu (48). In conclusion, SdpI could contribute to synaptic GlyR anchoring by regulating receptor immobilization via the actin cytoskeleton. It should, however, be noted that we found no significant change in gephyrin cluster properties in our SdpI loss-of-function experiments. This suggests that the consequences of SdpI deficiency are gephyrin-independent and thus likely to not target GlyRs already localized at synapses. In agreement with this view, previous analyses of gephyrin-deficient mice have shown that gephyrin is not required for the insertion of GlyRs and the related GABA A receptors into the neuronal plasma membrane (56,57). We therefore propose that binding of SdpI to GlyR␤ primarily regulates the intracellular trafficking of synaptic GlyRs.
A puzzling finding of this study is the lack of detectable changes in GlyR cluster number and size in brainstem sections derived from adult SdpI Ϫ/Ϫ mice. This result could reflect limitations in the sensitivity of our immunostaining procedure, but it might also be due to a compensation of the SdpI deficiency by either of the less abundant SdpI isoforms, e.g. SdpII and SdpIII, or by other functionally related trafficking proteins that are expressed only at later developmental stages. Here, immunoblotting experiments confirmed the expression of SdpII and SdpIII in brainstem but failed to reveal an up-regulation of these proteins in SdpI Ϫ/Ϫ mice, which are known to suffer from impaired inhibitory neurotransmission (20). Clearly, a more detailed investigation of the consequences of SdpI deficiency on the formation and function of glycinergic synapses is warranted.
The GlyR␤ Loop, a Key Determinant of Synaptic GlyR Trafficking and Immobilization-The cytoplasmic loop of GlyR␤ is known to be essential for the synaptic localization of GlyRs due to its tight interaction with gephyrin (5,6), and immunostainings with a monoclonal antibody specific for this GlyR subunit have confirmed its presence at the vast majority of glycinergic synapses in the rodent central nervous system (58). Recently, we identified the trafficking proteins Vps35 and Nbea as additional interaction partners of the GlyR␤ loop (10). Here, we further extend the number of GlyR␤-binding proteins by adding SdpI, which interacts via its SH3 domain. The underlying SH3 domain-SBM interaction appears to be of high affinity, because our proteomic search employed stringent biochemical conditions to select for robust interactions. Hence, less abundant proteins may have escaped detection, a view that is supported by the observation that only SdpI but not SdpII, which is also expressed in brain (14) and clearly binds GlyR␤, was consistently detected by MS in our pulldown experiments. Thus, additional binding partners of GlyR␤ might exist that remain to be identified. The SBMs of GlyRs have been recognized some time ago (1); our results provide a first indication for a role of these motifs in GlyR regulation.