A balance between excitatory and inhibitory synapses is controlled by PSD-95 and neuroligin

Abstract

Factors that control differentiation of presynaptic and postsynaptic elements into excitatory or inhibitory synapses are poorly defined. Here we show that the postsynaptic density (PSD) proteins PSD-95 and neuroligin-1 (NLG) are critical for dictating the ratio of excitatory-to-inhibitory synaptic contacts. Exogenous NLG increased both excitatory and inhibitory presynaptic contacts and the frequency of miniature excitatory and inhibitory synaptic currents. In contrast, PSD-95 overexpression enhanced excitatory synapse size and miniature frequency, but reduced the number of inhibitory synaptic contacts. Introduction of PSD-95 with NLG augmented synaptic clustering of NLG and abolished NLG effects on inhibitory synapses. Interfering with endogenous PSD-95 expression alone was sufficient to reduce the ratio of excitatory-to-inhibitory synapses. These findings elucidate a mechanism by which the amounts of specific elements critical for synapse formation control the ratio of excitatory-to-inhibitory synaptic input.

Synapse formation involves stabilization of initial sites of contact between axons and dendrites, followed by recruitment of specific protein complexes to newly formed presynaptic and postsynaptic structures (1–4). Neuronal contact formation is spatially and temporally controlled by changes in protein content and shape at areas of contact (5, 6). The total number of synapses formed and ratio of excitatory-to-inhibitory synaptic inputs a neuron receives are factors critical for determining neuronal excitability. Appropriate synthesis and recruitment of specific factors important for building synaptic contacts are thought to power this process. However, the identity of molecules that dictate the balance between excitatory and inhibitory synaptic contacts remains elusive. Several lines of evidence indicate that the scaffolding postsynaptic density (PSD) protein, PSD-95, is involved in orchestrating excitatory synapse maturation and specificity (7). PSD-95 is exclusively localized to glutamatergic synapses (8). Moreover, PSD-95 expression correlates with the period of excitatory synapse maturation (7, 9–11). Augmentation of excitatory synapse activity and ion channel clustering is driven by PSD-95 but not by related proteins, including synapse-associated protein (SAP)-102 and SAP-97 (12–14), and PSD-95 regulates clustering and activity of the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid-type glutamate receptors through a direct interaction with stargazin (7, 13–20). However, it is unknown how PSD-95 effects are translated into changes in apposing presynaptic terminals. A candidate molecule for mediating PSD-95 effects on presynaptic maturation is the cell adhesion molecule neuroligin (NLG). NLG is present at excitatory postsynaptic sites and associates with the third PSD-95/Dlg/ZO-1 homology (PDZ) domain of PSD-95 through its C-terminal PDZ-binding site (2, 21, 22).

The postsynaptic PDZ protein S-SCAM, another known binding partner of NLG, is also possibly involved in modulating NLG effects on synapse formation (23). Transsynaptically, NLG binds with high affinity to β-neurexin, and this interaction is thought to induce synapse formation (2, 21, 24–26). Indeed, presynaptic contact formation is induced in axons contacting nonneuronal cells expressing NLG, and these effects are abolished by disruption of NLG–neurexin protein interaction (26, 27). Moreover, coexpression of PSD-95 with NLG in heterologous cells potentiated NLG effects on N-methyl-d-aspartic acid currents (28). These results support a role for PSD-95 in the regulation of NLG-mediated effects in establishing synaptic contacts. However, it remains unclear whether coordinated actions of NLG and PSD-95 dictate synapse specificity.

Our data provide evidence that assembly of specific postsynaptic elements may not simply serve to form contacts between presynaptic and postsynaptic compartments but can determine synapse specificity. We find that the amounts of PSD-95 control proper localization and/or retention of NLG at the synapse, and that coordinated actions of NLG and PSD-95 regulate type (excitatory vs. inhibitory), number, and morphology of newly formed synaptic contacts. These effects are manifested in an overall change in the ratio of excitatory/inhibitory synaptic contact number and activity.

Methods

cDNA Cloning and Mutagenesis. Hemagglutinin (HA)-tagged WT NLG (1ab splice variant) amplified from mouse cerebellum was a gift from Peter Scheiffele (Columbia University, New York). HA-NLG was subcloned into pNice vector (Clontech), and the HA tag was inserted between amino acids 45 and 46 as described by Scheiffele et al. (27). Generation of GW1 PSD-95 fused to GFP and PSD-95 mutants has been described (15, 29). A NLG construct with a deletion of the last 4 aa (NLGΔPDZb) and the other truncated forms of NLG lacking the C-terminal cytosolic domain (containing amino acids 1–730; NLGΔCT GFP) or N-terminal extracellular sequences (containing amino acids 671–843; GFP NLGΔNT) were constructed by PCR and subcloned into an enhanced GFP vector (Clontech). NLGΔCT and NLGΔNT constructs were tagged with GFP to allow detection of these proteins. The signal peptide of NLG (amino acids 1–46) was added to the N terminus of NLGΔNT to maintain proper protein sorting.

Neuronal Cell Culture and Transfections. Dissociated primary neuronal cultures were prepared from hippocampi of embryonic day (E)18/E19 Wistar rats as described (30). Cultures were maintained in Neurobasal media (GIBCO-Invitrogen) supplemented with B27, penicillin, streptomycin, and l-glutamine. Hippocampal cultures were transfected by lipid-mediated gene transfer 3 days before immunostaining by using the transfection agent Effectene (GIBCO-Invitrogen) as described (31) and according to the manufacturer's protocol.

Immunocytochemistry. Coverslips were removed from culture wells, fixed in –20°C methanol, and immunolabeled for synaptic protein as described (15, 29). The following primary antibodies were used: PSD-95 (mouse, 1:500; Affinity BioReagents, Golden, CO), GFP (mouse, 1:1,000; Clontech), NLG (mouse, 1:1,000; Synaptic Systems, Goettingen, Germany), synaptophysin (rabbit, 1:1,000; Pharmingen), vesicular γ-aminobutyric acid (GABA) transporter (VGAT) (rabbit, 1:1,000; Synaptic Systems), vesicular glutamate transporter (VGLUT)-1 (rabbit, 1:1,000; Synaptic Systems), GluR1 (rabbit, 1:500; Upstate Biotechnology, Lake Placid, NY), and the N-methyl-d-aspartic acid receptor subunit 1 (mouse; 1:200; Synaptic Systems). The GFP antibody (guinea pig, 1:300) was generously provided by David Bredt (University of California, San Francisco). All antibody reactions were performed in blocking solution (2% normal goat serum) for 1 h at room temperature.

Imaging and Analysis. Images were acquired on a Zeiss Axiovert M200 motorized microscope by using a monochrome 14-bit Zeiss Axiocam HR charge-coupled device camera at 1,300 × 1,030 pixels. Exposure times were individually adjusted to yield an optimum brightness of immunofluorescent clusters without saturation. To correct for potential out-of-focus clusters within the field of view, focal plane (z-) stacks were acquired and a maximum intensity projection was performed offline. Images were scaled to 16 bits and analyzed in northern eclipse (Empix Imaging, Missasauga, Canada), by using custom-written software routines. Briefly, images were processed at a constant threshold level (of 32,000 pixel values) to create a binary (“mask”) image, which was multiplied with the original image by using Boolean image arithmetics. The resulting image contained a discrete number of clusters with pixel values of the original image. Dendrites of the cell of interest and untransfected control cells were outlined by using a combination of the GFP fluorescence signal and differential interference contrast bright-field images. Only clusters with average pixel values 2 times greater than corresponding background pixel values were used for analysis. Unless otherwise indicated, the size of stained clusters was measured, which represents the total (integral) pixel value of each object (see Supporting Text, which is published as supporting information on the PNAS web site, for more details).

For colocalization analysis, a number of equally sized measurement boxes was placed over stained clusters on the PSD-95 GFP (green) channel, and background-subtracted immunofluorescence values for all imaging channels (red, green, and blue) were correlated within each box. Colocalization was scored if clusters in two color channels were overlapping by at least 1 pixel, and the average staining intensity was >2 background staining. HA-NLG overexpression levels were calculated by calculating the ratio of NLG immunofluorescence over the soma of HA-NLG transfected with untransfected cells within the same field of view. The two-tailed nonparametric Mann–Whitney U test was used to compare clusters between experimental groups. For correlation analysis all data were rank-ordered and the Spearman correlation coefficient was calculated.

Electrophysiology. Day in vitro (DIV) 9 neurons were transfected with different cDNA combinations by using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol and expressed for 48 h before recording. For recording of miniature postsynaptic currents (PSCs) neurons were continuously perfused with extracellular solution [pH 7.4; 320–330 milliosmolar (mosM)] containing 140 mM NaCl, 5.4 mM KCl, 1 mM MgCl₂, 1.3 mM CaCl₂, 25 mM Hepes, 33 mM glucose, and 0.0005 mM tetrodotoxin (Alomone, Jerusalem) to block voltage-gated sodium channels and isolate action potential-independent miniature PSCs. Transfected cells with GFP signal were identified under a fluorescent upright microscope. Intracellular solution (pH 7.2; 300–310 mOsm) was composed of 115 mM Cs gluconate, 17.5 mM CsCl, 10 mM Hepes, 2 mM MgCl₂, 10 mM EGTA, 4 mM ATP, 0.4 mM GTP, and 0.1% Lucifer yellow (Sigma-Aldrich). A MultiClamp 700A amplifier (Axon Instruments, Foster City, CA) was used for recording. Access resistance was monitored, and recordings where series resistance varied by >10% were rejected. No electronic compensation for series resistance was used. Whole-cell patch-clamp recordings were performed in voltage-clamp mode while maintaining the membrane potential either at the reversal potential for GABA receptor-mediated miniature PSCs (–60 mV) to isolate miniature excitatory PSCs (mEPSCs) or at the reversal potential for ionotropic glutamate receptor-mediated miniature PSCs (+10 mV) to isolate miniature inhibitory PSCs (mIPSCs). Recorded mEPSCs and mIPSCs were antagonized completely by the ionotropic glutamate receptor antagonist cyano-7-nitroquinoxaline-2,3-dione (Sigma-Aldrich) and the GABA type A receptor antagonist bicuculline (Sigma-Aldrich), respectively (data not shown). Recordings were low-pass-filtered at 2 kHz, sampled at 10 kHz, and stored in a computer by using clampex 8.0 (Axon Instruments).

Small Interference RNA (siRNA) Experiments. We used annealed and HPLC-purified predesigned siRNA (Ambion, Austin, TX) to specifically interfere with PSD-95 expression levels in hippocampal cells. Sense and antisense sequences of siRNA were GGAGAUCACAUUGGAAAGGtt and CCUUUCCAAUGUGAUCUCCtc, respectively. DIV 7 hippocampal neurons were transfected by using Lipofectamine 2000 (Invitrogen) and immunostained 3 days later as described. For each well of a 24-well plate, siRNA (0.2 μg) and a transfection reporter (0.5 μg eGFP-C1) (Clontech) were cotransfected with 1 μl of Lipofectamine 2000.

Results and Discussion

NLG Drives the Formation of Excitatory and Inhibitory Presynaptic Contacts. To evaluate the role of NLG in synapse development, we assessed changes in synaptic contacts with expression of an HA-tagged NLG isoform-1 (HA-NLG) in hippocampal neurons. Both endogenous and overexpressed NLG were inefficiently clustered in developing neurons (Fig. 7, which is published as supporting information on the PNAS web site). Despite the lack of discrete NLG clusters, total number and size of presynaptic terminals contacting dendrites from cells expressing HA-NLG were significantly increased at DIV 12 (4.7- ± 0.3-fold and 2.4- ± 0.3-fold, respectively) and DIV 15 (4.1- ± 0.3-fold and 2.5- ± 0.3-fold, respectively) when compared to terminals on dendrites from cells transfected with GFP (Fig. 1 A, B, and E). In contrast, HA-NLG expression did not significantly change the average size of PSD-95 puncta and only resulted in a modest increase (1.5- ± 0.1-fold) in the number of PSD-95 clusters (Fig. 1 C, D, and F). Similarly, no significant change in size of either GluR1 or N-methyl-d-aspartic acid receptor subunit 1 clusters was detected (data not shown). These results demonstrate that NLG exerts differential effects on presynaptic and postsynaptic compartments. Moreover, unlike PSD-95 (7), NLG does not potentiate the size of the PSD and assembly of postsynaptic receptor complexes.

Fig. 1.

NLG induces excitatory and inhibitory presynaptic contact formation. Hippocampal neurons were transfected with HA-tagged NLG (HA-NLG) or GFP and stained for synaptophysin (Syn), PSD-95, VGAT, or VGLUT. (A) Size and number of Syn-positive terminals (white arrowheads; n = 10 and 11 cells) were enhanced on dendrites from neurons expressing HA-NLG when compared with GFP-expressing cells (B)(n = 12 and 9 cells). (C and D) A modest increase in the number of PSD-95 puncta was observed. (E) Similar changes in Syn cluster size and number were seen at both DIV 12 and 15. (F) Summary of changes in PSD-95 cluster number and size in cells expressing HA-NLG. (G–I) Size and number of VGAT-positive (n = 11 cells) and VGLUT-positive (n = 8 cells) contacts (white arrowheads) increased on dendrites from HA-NLG-expressing cells when compared with GFP-expressing cells (n = 12 cells, 242 terminals). (J) Summary of changes in VGAT and VGLUT cluster size and number in HA-NLG-expressing cells. In A–D and G–I, black arrowheads depict synaptic puncta from neighboring untransfected cells. *, P < 0.05; ***, P < 0.001 (Mann–Whitney U test). (Scale bars: 10 μm, A, C, and G; 1 μm, A Inset, B Inset, D, H, and I.)

A small increase in the number of PSD-95 clusters was observed when compared to the total number of synapses induced by HA-NLG overexpression. This finding suggests that HA-NLG may have induced a heterogeneous population of synaptic contacts. To evaluate the identity of synapses formed, we immunolabeled HA-NLG-transfected cells with the inhibitory presynaptic marker VGAT and the excitatory presynaptic marker VGLUT. We found that HA-NLG expression enhanced the size (1.9- ± 0.1-fold) and number (3.4- ± 0.2-fold) of both VGAT-positive presynaptic contacts and the size (2.2- ± 0.2-fold) and number (1.4- ± 0.1-fold) of VGLUT-positive contacts (Fig. 1 G–J). These results demonstrate that NLG can drive the formation of excitatory and inhibitory synapses.

Regulation of NLG Clustering by a PDZ-Dependent Pathway. The lack of distinct clusters of HA-NLG may have resulted from insufficient amounts of endogenous PSD-95. To test this, we first examined whether PSD-95 GFP could drive clustering of endogenous NLG. Using antibodies specific to NLG subtypes 1 and 3, we found that the size of NLG clusters was significantly enhanced (2.6- ± 0.2-fold) at sites containing PSD-95 GFP (Fig. 2 A and B). The size of presynaptic contacts (labeled with synaptophysin) was also enhanced (2.6- ± 0.2-fold). Moreover, a linear correlation (P < 0.0001) between the size of PSD-95 GFP, NLG, and Syn clusters was seen at single synapses (n = 10 cells, 200 synapses). This analysis revealed that PSD-95-mediated clustering of NLG and presynaptic enhancement are tightly regulated at individual excitatory synapses.

Fig. 2.

PSD-95 regulates NLG clustering and presynaptic maturation. Hippocampal neurons were transfected with either PSD-95 GFP or a mutant form of PSD-95 containing PDZ domains 1 and 2 (PSD-95 GFP 1-PDZ2). (A–C) Cells were stained with antibodies specific to NLG1 and NLG3 and the presynaptic marker synaptophysin (Syn). Dendritic clusters of endogenous NLG were compared between cells expressing PSD-95 GFP (A) (white arrowheads; n = 11 cells) and neighboring untransfected cells (B) or cells expressing PSD-95 GFP 1-PDZ2 (C). (D) Summary of changes of NLG and Syn in these neurons. PSD-95 GFP expression increased NLG and Syn cluster size. No significant change (n.s.) in NLG and Syn cluster size was observed in neurons expressing PSD-95 GFP 1-PDZ2. (E–H) Expression of PSD-95 GFP with HA-NLG enhanced the size of NLG clusters but reduced Syn-positive contact number when compared with neurons expressing HA-NLG alone. (E and F Right) Higher magnification micrographs of boxed regions overview images are shown. **, P < 0.01; ***, P < 0.001 (Mann–Whitney U test). (Scale bars: 10 μm, A Left, C Left, E Left, and F Left; 1 μm, A Right, B, C Right, E Right, and F Right.)

Previous findings showed that NLG associates with the third PDZ domain of PSD-95 through its C-terminal PDZ binding site (22, 32). To examine whether a PDZ-dependent interaction is involved in PSD-95-mediated clustering of NLG, we expressed a truncated form of PSD-95 that contains only the first two PDZ domains (PSD-95 1-PDZ2). This mutant form was previously shown to enhance GluR1 clustering (14). However, we found expression of PSD-95 1-PDZ2 did not significantly enhance either NLG clustering or the size of opposed presynaptic sites (Fig. 2 C and D). These results indicate that PSD-95-mediated clustering of GluR1 and NLG involves two independent processes.

Similar to the enhanced clustering of endogenous NLG, we also found that coexpression of HA-NLG with PSD-95 increased the average size of HA-NLG clusters (1.5- ± 0.1-fold), which correlated with a reduction in the total number of presynaptic terminals (2.5- ± 0-fold) (Fig. 2 E–H). Notably, 85 ± 3% of HA-NLG and PSD-95 GFP coclusters were present at sites apposed to excitatory presynaptic terminals (VGLUT-labeled) (Fig. 3 A–C). These sites were also positive for the excitatory postsynaptic receptor GluR1 (Fig. 3D). Consistent with the modest effect of NLG on postsynaptic elements, HA-NLG did not potentiate PSD-95-mediated clustering of GluR1 (Fig. 3E). Taken together, these results demonstrate that PSD-95 controls NLG and GluR1 accumulation at postsynaptic sites and restricts NLG clusters to excitatory synapses.

Fig. 3.

PSD-95 restricts NLG localization to excitatory synapses. Cells transfected with HA-NLG and PSD-95 GFP and stained with VGAT or VGLUT. (A Right) Magnifications of Inset Left.(A and B) PSD-95 GFP recruited HA-NLG to sites positive for VGLUT (n = 7 cells) but not VGAT (n = 10 cells). (C) Graph summarizing these results. (D and E) GluR1 cluster size was enhanced in dendrites from cells expressing PSD-95 GFP alone (n = 9 cells) or HA-NLG (n = 8 cells; white arrowheads) when compared with GluR1 clusters from untransfected controls (n = 10 cells). Expression of HA-NLG alone (n = 8 cells) did not enhance GluR1 clustering. **, P < 0.01; ***, P < 0.001 (Mann–Whitney U test). (Scale bars: 10 μm, A Left; 1 μm, A Right, B, and D.)

To further characterize the relationship between NLG and PSD-95 in regulating synapse morphology, we compared correlation of the presynaptic and postsynaptic contacts in neurons transfected with PSD-95 and WT NLG or a mutant form of NLG lacking the PDZ-binding domain (HA-NLGΔPDZb) (Fig. 4). This analysis revealed that the shape of presynaptic and postsynaptic contacts was highly correlated only at sites coexpressing PSD-95 GFP and WT HA-NLG (Fig. 4C). This finding supports the notion that both proteins influence presynaptic contact morphology.

Fig. 4.

PDZ-dependent interactions regulate the morphology of synapses induced by PSD-95 and NLG. (A and B) Neurons were transfected with PSD-95 GFP and either WT HA-NLG or a mutant HA-NLG lacking the PDZ-binding site (HA-NLGΔPDZb) and immunostained as indicated. (A) Colocalization (white arrowheads) of HA-NLG with PSD-95 GFP and apposed synaptophysin (Syn)-labeled presynaptic terminals. (A1 and A2) Localization of HA-NLG clusters and apposed presynaptic terminal to PSD-95 GFP. (B) HA-NLGΔPDZb is not recruited to PSD-95 GFP clusters (white arrowheads). (C) The correlation between the size of PSD-95 GFP, HA-NLG, and Syn clusters at single synaptic sites (linear fit; red line) was stronger in cells coexpressing PSD-95 GFP and WT HA-NLG (P < 0.0001; n = 10 cells, 219 clusters) when compared to cells expressing PSD-95 GFP and HA-NLGΔPDZb (P < 0.001; n = 10 cells, 225 clusters). (Scale bars: 10 μm, A and B;1 μm, A1, A2, B1, and B2.)

Next, to prevent binding to extracellular partners, we replaced the N-terminal domain (amino acids 46–694) of NLG with a sequence coding for GFP (GFP NLGΔNT) (Fig. 8, which is published as supporting information on the PNAS web site). Clusters of GFP NLGΔNT partially colocalized (to 43 ± 5%) with PSD-95. At these sites, the number of synaptophysin-positive contacts was reduced (by 86 ± 4%). This finding indicates that coupling of these molecules may regulate synapse formation at sites containing PSD-95.

PSD-95 Modulates the Specificity of NLG-Induced Synapses. To test whether PSD-95 modulates the specificity of HA-NLG-induced synapses, we evaluated the ratios of inhibitory and excitatory synaptic contacts by comparing the number of VGAT- and VGLUT-positive puncta relative to the total number of synapses (Fig. 5). Expression of PSD-95 GFP resulted in a significant reduction in number (2.5- ± 0.1-fold) and size (1.5- ± 0.1-fold) of HA-NLG-induced VGAT-positive contacts, to levels similar to those of GFP-expressing controls (Fig. 5A Upper). In contrast, PSD-95 GFP did not alter the number or size of VGLUT-positive terminals induced by HA-NLG (Fig. 5A Lower).

Fig. 5.

Manipulation of NLG and PSD-95 expression alters the ratio of excitatory to inhibitory synaptic contacts. (A Upper) The number and size of VGAT-labeled contacts significantly decreased in cells coexpressing HA-NLG and PSD-95 GFP (n = 10 cells, 701 terminals) compared with cells expressing HA-NLG alone (n = 11 cells, 2,145 terminals). (A Lower) No change in number and size of VGLUT-labeled excitatory presynaptic contacts in cells coexpressing HA-NLG and PSD-95 GFP (n = 7 cells, 572 terminals) compared with cells expressing HA-NLG alone (n = 8 cells, 705 terminals). (B and C) Electrophysiological recordings from hippocampal neurons transfected with HA-NLG and GFP (to visualize cells) (n = 26), PSD-95 GFP (n = 8), HA-NLG and PSD-95 GFP (n = 9), or GFP alone (controls, n = 16). Spontaneous mEPSCs (B) and mIPSCs (C) were measured at holding potentials of –60 mV and +10mV, respectively. (B) Enhanced mEPSC frequency and amplitude in cells expressing PSD-95 GFP and HA-NLG with PSD-95 GFP. Enhancement of mEPSC frequency, but not of amplitude, in cells expressing HA-NLG with GFP is shown. (C) mIPSC frequency was enhanced in cells expressing HA-NLG with GFP, reduced in cells expressing PSD-95 GFP, and unchanged in cells expressing HA-NLG with PSD-95 GFP. No difference in the average mIPSC amplitude between cell groups was found. *, P < 0.05; **, P < 0.01; ***, P < 0.001 (Mann–Whitney U test).

We next sought to examine functional correlates of these synaptic changes by using an electrophysiological approach (Fig. 5 B and C). We compared changes in excitatory and inhibitory currents in hippocampal neurons transfected with either GFP alone, HA-NLG and GFP, PSD-95 GFP, or PSD-95 GFP and HA-NLG by using whole-cell voltage-clamp recordings of mEPSCs and mIPSCs. Miniature frequency reflects both the size and number of presynaptic contacts, whereas the average amplitude of miniature currents correlates with postsynaptic parameters (33). Ectopic expression of HA-NLG significantly increased the frequency of mEPSCs (2.1- ± 0.3-fold) and IPSCs (1.9- ± 0.4-fold). A similar increase in mEPSC frequency was observed in neurons expressing PSD-95 GFP with (1.8- ± 0.3-fold) or without HA-NLG (2- ± 0.4-fold). Moreover, HA-NLG enhanced the average amplitude of mEPSCs only when coexpressed with PSD-95. This finding is consistent with the lack of changes in GluR1 clustering upon overexpression of NLG. In addition, HA-NLG induced an increase in mIPSC frequency that was abolished upon coexpression with PSD-95 GFP, a result paralleling our immunocytochemical data. Strikingly, expression of PSD-95 GFP alone resulted in a significant decrease (by 64 ± 11%) in mIPSC frequency (Fig. 5C). However, none of the tested constructs affected the average amplitude of mIPSCs analyzed.

Manipulations of the Levels of Endogenous PSD-95 Alters the Ratio of Excitatory-to-Inhibitory Synaptic Contacts. The opposite effects exerted by PSD-95 GFP on mEPSCs and mIPSCs suggests that PSD-95 may have sequestered cell adhesion molecules involved in excitatory and inhibitory synapse formation to specifically drive the development of excitatory synapses. Indeed, overexpression of PSD-95 GFP significantly decreased the percentage of inhibitory (VGAT-positive) contacts (by 53 ± 5%), without altering the total number of puncta positive for PSD-95 (Fig. 6A). These results mirror the physiological changes in the ratio of excitatory/inhibitory synaptic currents we observed (see Fig. 5).

Fig. 6.

Altered expression of PSD-95 influences the ratio of excitatory-to-inhibitory presynaptic contacts. (A) Hippocampal cells were transfected with either GFP (controls, Left) or PSD-95 GFP (Center) and immunostained at DIV 12 for GFP, synaptophysin (Syn), and the GABA inhibitory presynaptic marker VGAT. PSD-95 GFP expression resulted in a significant decrease in the percentage of VGAT-positive inhibitory contacts (Upper Right) but did not alter the number of PSD-95 clusters (Lower Right) (n = 15 cells, 288 VGAT+, 778 VGAT–) compared with controls (n = 12 cells, 242 VGAT+, 272 VGAT–). (B and C) Introduction of GFP with either a scrambled siRNA (control siRNA; Left) or PSD-95 siRNA (Center). (B Right) A reduction in the number of PSD-95-positive puncta in cells cotransfected with GFP and PSD-95 siRNA (n = 11 cells, 301 puncta) compared with controls (n = 7 cells, 406 puncta) was found. (C Right) A significant increase in the percentage of inhibitory contacts (VGAT+) and a concurrent decrease in the percentage of excitatory contacts (VGAT–)(Upper) was found, but there was no change in the total number of synapses in cells expressing GFP and PSD-95 siRNA (n = 11 cells, 322 VGAT+, 282 VGAT–) compared with controls (n = 12 cells, 230 VGAT+, 494 VGAT–)(Lower). **, P < 0.01; ***, P < 0.001 (Mann–Whitney U test). (Scale bars: = 1 μm.)

We next tested whether blocking expression of endogenous PSD-95 has opposite effects on the ratio of excitatory-to-inhibitory synaptic input. Therefore, we generated an siRNA that was designed to specifically decrease endogenous levels of PSD-95 (see Methods) (Fig. 6 B and C). Cotransfection of PSD-95 siRNA together with GFP reduced PSD-95 clusters (by 56 ± 5%). In contrast, no change in the number of PSD-95 clusters was observed in neurons expressing a scrambled siRNA (Fig. 6B). Expression of PSD-95 siRNA also resulted in an increase (1.5- ± 0.1-fold) in inhibitory (VGAT-positive) and a decrease (by 28 ± 7%) in excitatory (VGAT-negative) synaptic contacts. However, no change in total number of synaptic contacts was observed (Fig. 6C). These results demonstrate that the amounts of PSD-95 available can dictate the balance between excitatory and inhibitory presynaptic inputs.

In conclusion, our results show a striking difference between the effects of PSD-95 and NLG in dictating synaptic identity. Whereas NLG expression results in a dramatic increase of both excitatory and inhibitory presynaptic contacts, PSD-95 selectively induces maturation of excitatory presynaptic and postsynaptic elements. PSD-95, by recruiting NLG at sites of contact, plays a decisive role in controlling synaptic contact number, morphology, and specificity. In contrast, the lack of specificity of synapses induced by NLG suggests that NLG is strictly involved in establishing initial synaptic contacts regardless of their phenotype. The identity of these contacts may be determined by events that require the recruitment of additional factors. This study reveals that PSD-95 is one of the potential factors involved in this process. By assembling a core postsynaptic complex containing stargazin, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor, and GKAP, PSD-95 may determine synaptic identity (34).

How does NLG communicate with excitatory and inhibitory presynaptic elements? A known binding partner of NLG is β-neurexin, a neuron-specific cell surface protein that exists in several alternatively spliced isoforms (22, 27, 35, 36). Similar to NLG, neurexins also contain PDZ-binding motifs that bind to type II PDZ domains present in CASK and syntenin but not PSD-95. The association of β-neurexin to a tripartite protein complex formed of CASK, Mint-1, and Veli has been proposed to act as a nucleation site for coupling cell adhesion molecules to synaptic vesicle exocytosis (36). Also, neurexins are directly coupled to synaptotagmins, core molecules of the synaptic vesicle machinery that regulate neurotransmitter exocytosis (37). Therefore, binding of NLG to β-neurexin may activate an array of presynaptic molecular responses, leading to the structural reorganization of the presynaptic compartment. However, it remains unclear whether specific neurexin isoforms are present at GABAergic presynaptic terminals and whether an interaction between NLG and neurexin mediates the effects we observed on inhibitory presynaptic sites.

The importance of our findings is emphasized by the recent discovery that frameshift mutations in NLG3 and NLG4 genes, which result in early protein truncation and misfolding, are associated with autism (38–41). Moreover, chromosomal rearrangements that harbor NLG1 and NLG2 and PSD-95 genes have been associated with autism (42–44). Abnormal expression of PSD-95 is also altered in fragile-X syndrome (45). We propose a model in which improper expression and/or targeting of molecules that control synaptic specificity, such as PSD-95 and NLG, may trigger an imbalance in neuronal excitability. This model is supported by our finding that manipulation of endogenous PSD-95 alone was sufficient to alter the ratio of excitatory/inhibitory synaptic contacts. This notion is consistent with a recent model suggesting that the ratio of neuronal excitation/inhibition is altered in psychiatric disorders, specifically in autism and mental retardation (46).