Gephyrin is a multifunctional protein contributing to molybdenum cofactor (Moco) synthesis and postsynaptic clustering of glycine and GABAA receptors. It contains three major functional domains (G-C-E) and forms cytosolic aggregates and postsynaptic clusters by unknown mechanisms. Here, structural determinants of gephyrin aggregation and clustering were investigated by neuronal transfection of EGFP-tagged deletion and mutant gephyrin constructs. EGFP-gephyrin formed postsynaptic clusters containing endogenous gephyrin and GABAA-receptors. Isolated GC- or E-domains failed to aggregate and exerted dominant-negative effects on endogenous gephyrin clustering. A construct interfering with intermolecular E-domain dimerization readily auto-aggregated but showed impaired postsynaptic clustering. Finally, two mutant constructs with substitution of vertebrate-specific E-domain sequences with homologue bacterial MoeA sequences uncovered a region crucial for gephyrin clustering. One construct failed to aggregate, but retained Moco biosynthesis capacity, demonstrating the independence of gephyrin enzymatic activity and aggregation. Reinserting two vertebrate-specific residues restored gephyrin aggregation and increased formation of postsynaptic clusters containing GABAA receptors at the expense of PSD-95 clusters – a marker of glutamatergic synapses. These results underscore the key role of specific E-domain regions distinct from the known dimerization interface for controlling gephyrin aggregation and postsynaptic clustering and suggest that formation of gephyrin clusters influences the homeostatic balance between inhibitory and excitatory synapses.
Gephyrin is a ubiquitous multifunctional protein that catalyzes molybdenum cofactor (Moco) biosynthesis and contributes to postsynaptic glycine and GABAA receptor clustering in neurons (Feng et al., 1998; Kneussel and Betz, 2000a; Meier and Grantyn, 2004; Stallmeyer et al., 1999). The gephyrin G-domain and E-domain are homologous to Escherichia coli MogA and MoeA enzymes and catalyze Moco synthesis. Unlike in bacteria, they are interconnected by a central linker region, the C-domain, which harbors several binding sites for proteins interacting with gephyrin, such as GABARAP (Kneussel et al., 2000) and dynein light-chain 1 and 2 (Fuhrmann et al., 2002; Maas et al., 2006). Glycine receptors bind to identified sites on the E-domain (gephyrin-E) (Bedet et al., 2006; Kim et al., 2006; Schrader et al., 2004; Sola et al., 2004). Gephyrin also binds to collybistin, a neuronal GTP-exchange factor (Grosskreutz et al., 2001; Harvey et al., 2004; Kins et al., 2000) and cytoskeletal proteins, including tubulin (Kirsch et al., 1991) and actin, either directly or via adaptor proteins, such as profilin (Mammoto et al., 1998) or mena/VASP (Giesemann et al., 2003).
Crystal structures of individual gephyrin domains (Schrader et al., 2004; Schwarz et al., 2001; Sola et al., 2001; Sola et al., 2004) suggested the formation of multimeric, hexagonal aggregates by dimerization of gephyrin-E and trimerization of gephyrin-G (Kneussel and Betz, 2000b). Four subdomains were distinguished in gephyrin-E, with binding of the glycine receptor β subunit (Meyer et al., 1995) occurring in subdomains 3-4, in a pocket adjacent to the dimer interface (Kim et al., 2006; Sola et al., 2004). According to the model, glycine receptor binding might influence gephyrin aggregation. However, the formation of postsynaptic gephyrin scaffolds and their regulation remain to be elucidated. So far, it is not known whether gephyrin clustering occurs via dimerization of gephyrin-E as seen in crystal structures of isolated E-domains.
Comparison of gephyrin structure with its homologues revealed extensive conservation of residues crucial for the structure of the G- and E-domains, but also specific differences in both domains, such as the glycine receptor binding site (Kim et al., 2006; Sola et al., 2004), suggesting that further motifs might be important for neuron-specific functions of gephyrin. In particular, it is not known whether gephyrin enzymatic activity, oligomerization and cytoskeleton binding are interdependent properties. Furthermore, gephyrin is colocalized extensively with GABAA receptors in GABAergic synapses (Sassoè-Pognetto et al., 1995; Sassoè-Pognetto et al., 2000), where its clustering depends on the presence of the γ2 subunit (Alldred et al., 2005; Essrich et al., 1998; Li et al., 2005; Schweizer et al., 2003). However, the mechanism of this interaction is not established.
Here, we investigated the structural requirements of gephyrin aggregation and postsynaptic clustering, using acute transfection of EGFP-tagged gephyrin constructs in hippocampal or cortical neurons in culture. We focused on the role of gephyrin-E using various truncated and mutant constructs (Fig. 1A). Functional integrity and native folding of these constructs was verified by reconstitution of Moco synthesis activity. The subcellular location of recombinant gephyrin in relation to endogenous gephyrin, GABAA receptors and presynaptic terminals was determined by immunofluorescence staining and confocal laser-scanning microscopy.
Ten EGFP-tagged gephyrin constructs were used in this study (Fig. 1). They were tested for aggregation and interaction with recombinant GABAA receptors in HEK293 cells, whereas neuronal transfection served to investigate interactions with endogenous gephyrin and GABAA receptors at postsynaptic sites and possible dominant negative effects on endogenous gephyrin clusters. The subcellular distribution of transfected gephyrin constructs was visualized directly with the EGFP tag whereas gephyrin immunofluorescence staining (using mAb7a) labeled both endogenous and recombinant gephyrin clusters. Three transfection schedules were used to investigate expression before, during, and after synaptogenesis (Studler et al., 2005): transfection at 6 days in vitro (DIV) followed by analysis at either 8 DIV (denoted 6+2) or 12 DIV (6+6); transfection at 10 DIV and analysis at 12 DIV (10+2).
Postsynaptic clustering of EGFP-tagged full-length gephyrin
EGFP-tagged full-length gephyrin (EGFP-geph-GCE; Fig. 1A) transfected in HEK293 cells formed large cytoplasmic aggregates, as described (Kirsch et al., 1995), which were not associated with co-transfected GABAA receptors at the cell surface (Fig. 2A). Similar cytosolic aggregates were seen in the soma and proximal dendrites of transfected neurons at all time-points examined (Fig. 2B-F). Double staining of transfected neurons for gephyrin and the α2 subunit at the cell surface revealed the presence of EGFP-geph-GCE fluorescence in postsynaptic clusters, co-localized with the α2 subunit even after expression for just 2 days (6+2 experiment; Fig. 2B). Upon expression of EGFP-geph-GCE for 6 days, the density of postsynaptic clusters was strongly increased (6+6 experiment; Fig. 2C). This was not due to overexpression since neighboring non-transfected cells exhibited a similar increase of gephyrin and GABAA receptor α2 subunit clusters. Importantly, clusters containing only endogenous gephyrin were rarely seen (Fig. 2C), suggesting that EGFP-geph-GCE was mixed with endogenous gephyrin in postsynaptic clusters associated with GABAA receptors. Depending on the level of expression, cytosolic aggregates were variable in size and number and could sometimes be seen far away from the soma (Fig. 2D).
To verify that EGFP-geph-GCE is incorporated in postsynaptic clusters also after the peak of synaptogenesis, it was transfected at 10 DIV and expressed for two days (10+2). The distribution pattern of EGFP-positive clusters and cytosolic aggregates was similar to that seen in 6+6 experiments, including numerous clusters associated with GABAA receptors (Fig. 2E). Staining for synapsin-1 was performed to distinguish EGFP-positive clusters apposed to a presynaptic terminal from cytosolic EGFP-positive aggregates. The latter were distributed at random in dendrites and were not apposed to presynaptic terminals (Fig. 2F), confirming that EGFP-geph-GCE clusters associated with GABAA receptors represent postsynaptic structures.
Isolated gephyrin domains partially disrupt endogenous gephyrin clustering
Truncated gephyrin constructs (Fig. 1A) were transfected to analyze their localization and influence on endogenous gephyrin. Constructs lacking either the entire E-domain (EGFP-geph-GC) or the G-domain and linker region (EGFP-geph-E) were reproducibly expressed in HEK293 cells (Fig. 3A,E) and in neurons (Fig. 3B-D,F-H). With both constructs, a diffuse EGFP signal was distributed homogenously in the cytosol but excluded from the nucleus. No aggregates were observed in either cell type. In 6+2 experiments, only a few endogenous gephyrin clusters were seen in either non-transfected or transfected neurons, whereas GABAA receptor α2 subunit clusters were already present (Fig. 3B,F). Immunofluorescence staining for gephyrin was restricted to clusters devoid of EGFP, indicating selective detection of endogenous gephyrin. Therefore, we could assay whether these constructs interfere with gephyrin clustering.
In 6+6 experiments, the number of gephyrin clusters was decreased in transfected cells, but the effect was variable. Some cells showed a profound reduction of endogenous gephyrin clusters compared with mock-transfected cells (Fig. 3C,G), whereas the effect appeared less pronounced in other cells (Fig. 3D,H). However, at higher magnification many gephyrin clusters were seen at cross-points with dendrites of non-transfected cells, as seen best by double labeling for MAP2, a marker of dendrites (Fig. 3I,J). In such cases, it was often difficult to determine in which dendrite gephyrin clusters were located (Fig. 3I″,J″). Therefore, for quantification, we considered the mean value calculated from the number of clusters unambiguously located on transfected dendrites (minimum) and the total number of clusters in direct contact with transfected dendrites (maximum). Distribution analysis of the number of gephyrin clusters per dendrite segment confirmed that transfection with either construct led to a distinct reduction of gephyrin clusters compared with non-transfected cells in the same culture dishes (Fig. 4A-C; Table 1). Thus, in cells expressing EGFP-geph-E, almost 55% of dendrites contained fewer than 1.2 clusters/20 μm compared with 5.5% in mock transfected cells (Fig. 4I). A similar reduction occurred upon transfection of EGFP-geph-GC (Fig. 4I). These results indicate that isolated gephyrin domains can interact with endogenous gephyrin molecules, thereby preventing the formation of postsynaptic clusters. Despite these effects, no alteration in the distribution of α2-subunit-positive clusters was discernible in either 6+2 or 6+6 experiments, suggesting that at this stage in culture, clustering of GABAA receptors can occur independently of gephyrin clustering. Similar results were obtained in 10+2 experiments (Fig. 3K-N), with a reduced number of endogenous gephyrin clusters in most transfected cells (Fig. 4D-F). Again, the reduction was quite variable (compare Fig. 3K-L with Fig. 3M-N) but the dominant-negative effect was comparable for the two constructs (Fig. 4E,F). In these experiments, α2 subunit clustering also was not visibly affected in transfected neurons.
|.||Mock 6+6 .||Geph-GC 6+6 .||Geph-E 6+6 .||Geph-GCEE 6+6 .||Geph-L2B 6+6 .||Mock 10+2 .||Geph-E 10+2 .||Geph-GC 10+2 .|
|.||Mock 6+6 .||Geph-GC 6+6 .||Geph-E 6+6 .||Geph-GCEE 6+6 .||Geph-L2B 6+6 .||Mock 10+2 .||Geph-E 10+2 .||Geph-GC 10+2 .|
The frequency distribution of gephyrin clusters on dendritic segments was analyzed statistically pair-wise using the Kolmogorov-Smirnov test. Approximately 30-50 dendrites were sampled per construct in three independent cultures in 6+6 and 10+2 experiments. Mock transfected cells are non-transfected cells found in the same dish as cells transfected with Geph-E, Geph-GC or Geph-GCEE. NS, non-significant; *P<0.05; **P<0.01; ***P<0.001; –, comparison not done
Pair-wise comparison of gephyrin cluster distribution by cumulative distribution analysis (Table 1) confirmed that the impairment of endogenous gephyrin clustering was most severe during the phase of synaptogenesis (6+6 experiments). The effects of both constructs were milder when introduced in older cultures (10+2). In mock transfected cells, no difference in endogenous gephyrin cluster numbers was observed between 6+6 and 10+2 (Fig. 4A,D; Table 1), indicating that they do not suffer from the toxicity of the transfection reagent. In summary, isolated gephyrin domains interact with endogenous gephyrin, thereby affecting the formation or stability of postsynaptic clusters.
Constructs with partial deletions of gephyrin-E are not stable
The current model of gephyrin aggregation underscores the importance of the E-domain for gephyrin oligomerization (Kim et al., 2006; Kneussel and Betz, 2000b). To identify crucial subregions within gephyrin-E regulating this process, constructs lacking either subdomains 1-2 (EGFP-geph-GCE34) or subdomains 3-4 (EGFP-geph-GCE12) (Fig. 1A) were investigated. Based on the crystal structure of gephyrin-E, the separation of these subdomains should result in a loss of gephyrin-E dimerization. Both constructs were poorly expressed in HEK293 cells and in neurons, yielding a weak homogenous cytosolic fluorescence (not shown). Therefore, the structural integrity of the E-domain appears to be compromised in these deletion constructs, leading to their rapid degradation. Alternatively, the weak fluorescence detected might indicate interference of the mutant construct with EGFP, reduced mRNA stability or poor translation.
Interfering with gephyrin-E dimerization does not prevent gephyrin aggregation
Gephyrin oligomerization is believed to be mediated by intermolecular dimerization of gephyrin-E and trimerization of gephyrin-G, as seen in the crystal structures of isolated domains. Consistent with the fact that purified gephyrin exists as a trimer (Schrader et al., 2004; Sola et al., 2004) and with the strong dominant negative effect of isolated gephyrin-E on endogenous gephyrin clusters (Fig. 4), one can argue that this domain is crucial for regulation of gephyrin aggregation. Furthermore, biochemical data support a much stronger interaction between gephyrin-G than gephyrin-E (Schwarz et al., 2001). Consequently, intramolecular dimerization of gephyrin-E might be expected to disrupt aggregation and interfere with endogenous gephyrin clustering. A full-length gephyrin construct containing an additional E-domain (Schrader et al., 2004) spaced by a flexible linker to allow for intramolecular dimerization (EGFP-geph-GCEE; Fig. 1A) was used to test this hypothesis. Biochemical analysis of purified geph-GCEE indicated that intramolecular dimerization occurs in this construct (Schrader et al., 2004). Furthermore, we could demonstrate here that geph-GCEE is as active in Mocobiosynthesis as the wild-type variant (geph-GCE, see below and Fig. 6A).
Upon transfection in HEK293 cells, this construct was still able to form cytoplasmic aggregates (Fig. 5A), similar to those formed by EGFP-geph-GCE (Fig. 2A). Likewise, in transfected hippocampal neurons, EGFP-geph-GCEE-positive cytosolic aggregates were readily detected in the soma and proximal dendrites at 6+2 (Fig. 5B) and 6+6 (Fig. 5C-E). These aggregates were more numerous and brighter than those formed by EGFP-geph-GCE (Fig. 2B,C) and were not colocalized with the α2 subunit. Therefore, aggregation was not prevented by intramolecular E-domain duplication, but rather facilitated. Furthermore, many endogenous gephyrin clusters (detected by immunofluorescence) were not labeled for EGFP (Fig. 5D; open arrows). Therefore, EGFP-geph-GCEE rarely integrates in postsynaptic clusters associated with GABAA receptors, as confirmed by double labeling with the α2 subunit and synapsin 1 (Fig. 5E). Quantitatively, only 1.7±0.2 EGFP-geph-GCEE clusters were present per dendrite segment, of which 79.4% were postsynaptic, compared with 5.5±0.4 EGFP-geph-GCE clusters, of which 74.5% were postsynaptic (mean ± s.e.m., P<0.0001). Finally, quantification of postsynaptic gephyrin clusters (identified by colocalization with the α2 subunit) revealed a dominant-negative effect of EGFP-geph-GCEE (Fig. 4G; Table 1). However, it was less severe and qualitatively different than that caused by EGFP-geph-E (Fig. 4C,G; Table 1). EGFP-geph-GCEE produced either a complete loss of gephyrin clusters when strongly overexpressed or it was present in postsynaptic clusters, colocalized with the α2 subunit, in cells containing few cytosolic aggregates. The observation that postsynaptic clustering of EGFP-geph-GCEE only occurs when it is expressed at low level suggested that it requires interaction with endogenous gephyrin, which is available in limited amounts. To verify this possibility, transfection was performed at the time of plating by `nucleofection'. Since this method requires a large quantity of cells (5×106) for transfection, these experiments were performed with cortical neurons. The results obtained with this method were similar to those with `magnetofection'. Thus, in transfected neurons containing prominent cytosolic EGFP-geph-GCEE aggregates, indicative of a strong expression, no clusters double labeled with the α2 subunit were seen after 12 DIV (Fig. 5F). Conversely, clusters were observed in cells containing small cytoplasmic EGFP-geph-GCEE aggregates (Fig. 5G) and were predominating when expression was low (Fig. 5H). These results confirm that EGFP-geph-GCEE has a strong propensity for auto-aggregation and is not able to form postsynaptic clusters without endogenous gephyrin. Therefore, the expected intermolecular dimerization of the E-domain is not the sole mechanism for gephyrin aggregation and postsynaptic clustering.
Identification of a motif in gephyrin-E that regulate aggregation and clustering
To gain further insight into the mechanism of gephyrin aggregation, constructs carrying mutations in vertebrate-specific sequences were tested. In particular, beside residues contributing to the high-affinity glycine receptor binding site, subdomain 3 of gephyrin-E harbors two vertebrate-specific loops (L1 and L2) that are surface exposed (Fig. 1B-D) and might be involved in a specialized function such as postsynaptic clustering (Schrader et al., 2004). To test the functional relevance of these loops, mutant EGFP-tagged constructs were expressed, in which the vertebrate-specific sequences were swapped with the corresponding sequences of E. coli MoeA (L1, L2A, L2B, L2C; Fig. 1B-D).
First, we investigated the functional integrity of these L-variants by functional reconstitution of Moco biosynthesis in the E-domain-homologous E. coli moeA mutant. This mutant is not able to form active Moco (Fig. 6A, pQE30 line) but upon expression of geph-GCE, Moco was formed (Fig. 6A, GCE, black bars), which was determined by the reconstitution of nit-1 apo-nitrate reductase, an assay that is commonly used to quantify active Moco (Schwarz, 2005). The same activity was also observed for the geph-P713E variant, which is unable to bind the glycine receptor β-loop (Kim et al., 2006). A somewhat lower activity was seen in geph-GCEE and a construct expressing the C- and E-domain (CE), which was reflected by lower levels of their expression as monitored by western blots (Fig. 6B). Normalizing these activities to the expression levels of geph-GCE, one can observe a complete reconstitution (Fig. 6A, white bars). No activity was seen in cells expressing only the G-domain. Also the L1, L2B, and L2C variants exhibited 15-20% of geph-GCE activity but expression of these proteins was approximately 5-20 times lower, resulting in a GCE-like normalized activity (Fig. 6A, white bars). These data demonstrate that expression of all L-variants is significantly lower than that of gephyrin but, relative to the amount of protein present in the cells, they exhibit similar activities in Moco synthesis. This finding is consistent with the fact that the site of Moco synthesis is clearly separated from the region that has been modified here (Fig. 1C).
Upon transfection in HEK293 cells (not shown) and in neurons (Fig. 7A), EGFP-geph-L1 formed aggregates and postsynaptic clusters, respectively, similar to those observed with EGFP-geph-GCE. Thus, this region of the E-domain is not essential for aggregation. By contrast, EGFP-geph-L2A and EGFP-geph-L2B failed to aggregate and produced a diffuse cytosolic fluorescence. The signal for EGFP-geph-L2B in HEK293 cells (Fig. 7B) was comparable to EGFP-geph-E, whereas EGFP-geph-L2A was very weak (not shown), suggesting that this mutant construct was unstable and/or poorly expressed. Unexpectedly, EGFP-geph-L2C, which contains two more gephyrin-specific residues than EGFP-geph-L2B (Fig. 1B), formed distinct aggregates in HEK293 cells (Fig. 7C). Therefore, this region of subdomain 3, which is located away from the postulated dimerization interface of gephyrin-E, should be involved in gephyrin aggregation. The fact that Moco synthesis activity was not affected can be taken as a measure for proper folding of both protein variants. Thus, residues D613 and/or I620, which are not substituted in L2C, appear to be crucial for gephyrin aggregation.
The differential aggregation capacity of EGFP-geph-L2B and L2C was replicated in hippocampal neurons. Thus, at all time-points tested (6+2, 6+6 and 10+2), transfection of EGFP-geph-L2B resulted in a diffuse, grainy fluorescence (Fig. 7D,E) of comparable intensity as EGFP-geph-E or EGFP-geph-GC. Furthermore, as observed with these truncated constructs, EGFP-geph-L2B produced a dominant negative effect on gephyrin clusters (Fig. 4H and Table 1), indicating that this construct interacts with endogenous gephyrin and prevents its clustering.
In striking contrast, in neurons transfected with EGFP-geph-L2C, numerous gephyrin clusters appeared along dendrites and the soma, in addition to cytosolic aggregates (Fig. 7F-G). As observed for EGFP-geph-GCE, most EGFP-positive clusters were colocalized with α2 subunit immunoreactivity (Fig. 7F). EGFP-geph-L2C therefore appeared to readily associate with endogenous gephyrin at postsynaptic sites. Quantification of transfected cells revealed that EGFP-geph-L2C-positive clusters were slightly smaller but more numerous (Fig. 7J-L) than those formed by EGFP-geph-GCE (190.2±8.8 versus 107.2±7.8 dendritic clusters per cell; mean ± s.e.m.; P<0.001).
Triple staining for the α2 subunit and synapsin-1 showed that EGFP-geph-L2C-positive clusters were almost exclusively colocalized with α2 and frequently apposed to synapsin-1-positive terminals (Fig. 7G). Compared with cells transfected with EGFP-geph-GCE, the ratio of synaptic and extrasynaptic EGFP-positive clusters was exactly the same (60.4±8.9% for EGFP-geph-L2C versus 61.8±14.1% for EGFP-geph-GCE; mean ± s.d.) (Fig. 7K). This finding implies that this construct induces formation of supernumerary GABAergic postsynaptic clusters. To determine whether this `synaptogenic' effect is specific, or whether neurons transfected with EGFP-geph-L2C make more synapses in general, we quantified the number of EGFP clusters and PSD-95 clusters, representing glutamatergic synapses, compared with control (transfection with EGFP-geph-GCE) (Fig. 7H-I). The ratio of EGFP/PSD95 clusters was significantly higher in EGFP-geph-L2C-transfected dendrites (0.65±0.31 versus 0.44±0.20, P<0.0001) (Fig. 7L), confirming that the increase of EGFP-geph-L2C clusters was specific for inhibitory synapses. Furthermore, the total number of synapses remained unchanged (Fig. 7L), indicating that the increased number of EGFP-geph-L2C clusters occurred at the expense of PSD-95-positive clusters. These findings suggest that gephyrin clustering contributes to the formation of GABAergic synapses during development.
Postsynaptic clustering of gephyrin is essential for proper function of inhibitory synapses in the CNS. Our results demonstrate that modifications of gephyrin-E that do not affect its enzymatic activity differentially influence its cytosolic aggregation and postsynaptic clustering with GABAA receptors, pointing to the latter involving specific residues at the surface of the protein. First, proper folding and oligomerization of gephyrin requiring both G and E domains is essential for aggregation in HEK293 cells and in neurons, as well as clustering with endogenous gephyrin at postsynaptic sites. Deletion constructs containing isolated GC- or E-domains interact with endogenous gephyrin, producing dominant-negative effects on postsynaptic clusters. Second, Moco synthesis activity, which confirms the structural integrity of gephyrin constructs, is independent of gephyrin aggregation or binding to glycine receptors. Third, intermolecular interactions between gephyrin-E are not dependent solely on the dimerization interface identified in structural studies. Intramolecular dimerization of gephyrin-E increases the propensity for autoaggregation but impairs postsynaptic clustering. Furthermore, residues localized in subdomain 3 of gephyrin-E (L2) act as an all-or-none switch for both gephyrin aggregation and postsynaptic clustering. Most strikingly, a construct facilitating the formation of postsynaptic clusters, EGFP-geph-L2C, induces supernumerary GABAergic postsynaptic clusters at the expense of PSD-95 clusters, suggesting that gephyrin clustering influences the homeostatic balance between inhibitory and excitatory synapses. The differential regulation of gephyrin aggregation and postsynaptic clustering points to posttranslational modifications and/or specific protein-protein interactions in these processes.
Expression of EGFP-tagged full-length gephyrin (EGFP-geph-GCE) confirmed that EGFP does not interfere with any of the functions of gephyrin investigated here (Fuhrmann et al., 2002; Harvey et al., 2004). Overexpression resulted in the formation of intracellular aggregates, but the presence of brightly stained postsynaptic clusters associated with GABAA receptors indicated that synaptic targeting and clustering of recombinant gephyrin were not affected by the presence of EGFP. The absence of clusters containing only endogenous gephyrin was taken as evidence for the association of both forms of gephyrin within the vast majority of postsynaptic clusters. The fact that this association occurred within 48 hours of transfection points to a constant turnover of gephyrin in postsynaptic clusters, in line with recent tracking studies (Hanus et al., 2006).
GABAA receptor staining was applied to living cells in order to label only cell surface and postsynaptic receptors (van Rijnsoever et al., 2005). Therefore, postsynaptic gephyrin clusters were distinguished from cytosolic aggregates based on size, location relative to presynaptic terminals and colocalization with GABAA receptors. Postsynaptic clustering of EGFP-geph-GCE implies that the protein is transported to the cell surface, targeted to postsynaptic sites and associated with GABAA receptors by the same mechanisms as endogenous gephyrin. Conversely, cytosolic aggregates were characterized by a larger size and absence of GABAA receptor staining, confirming their intracellular localization. These aggregates could represent macromolecular gephyrin complexes, as suggested by the spontaneous aggregation of purified gephyrin (Schrader et al., 2004; Sola et al., 2004). It cannot be excluded, however, that they are associated with specific proteins or localized in a particular organelle (Maas et al., 2006). This distinction underscores that postsynaptic clustering is a highly regulated process, leading to multimeric scaffolds of defined size, irrespective of the amount of gephyrin available in the cell. One might surmise that mechanisms of gephyrin transport to synaptic sites are saturable and excessive gephyrin forms intracellular aggregates owing to a non-synaptic type multimerization. In vivo, postsynaptic clustering of gephyrin is disrupted in the absence of GABAA receptors, leading to the formation of cytosolic aggregates (Kralic et al., 2006; Studer et al., 2006), demonstrating the importance of its interaction with this major partner for postsynaptic targeting and clustering.
Several constructs, EGFP-geph-GC, EGFP-geph-E and EGFP-geph-L2B, produced a diffuse fluorescence in the cytoplasm, indicative of failure to aggregate when the G- and E-domains are separated or when specific residues in L2 are mutated. However, oligomerization of isolated domains with endogenous gephyrin still occurred, as seen by the rapid and profound reduction of postsynaptic gephyrin clusters. Such dominant-negative effects were most pronounced when the truncated constructs were introduced during synapse formation and expressed for a longer time period, with the strongest effects being produced by the single domain constructs (Table 1). However, no dominant-negative effects were observed on GABAA receptor clustering up to 12 DIV, indicating that, unlike glycine receptors (Bedet et al., 2006), their surface expression and initial clustering are not critically regulated by gephyrin.
Reconstitution of Moco biosynthetic activity provides a sensitive test to probe for proper folding of gephyrin constructs and retention of their functional properties. In particular, geph-L2B exhibits wild-type-like Moco activity despite its failure to aggregate whereas geph-L1 formed aggregates but had reduced Moco activity. Therefore, these constructs allow dissociating gephyrin aggregation or clustering from its enzymatic activity. Moco synthesis was also observed with the geph-P713E variant, which is unable to bind the glycine receptor β-loop (Kim et al., 2006), demonstrating that Moco-synthetic activity and glycine-receptor binding are not dependent on and do not exclude each other.
Enhanced aggregation but defective clustering upon intramolecular E-domain dimerization within gephyrin
Transfection of EGFP-geph-GCEE in HEK293 cells showed that aggregation still occurs despite the E-domain duplication. In neurons, EGFP-geph-GCEE formed more cytosolic aggregates than EGFP-geph-GCE and its postsynaptic clustering was markedly impaired upon overexpression, suggesting that this variant has an increased propensity for auto-aggregation. These observations raise the question whether the E-domains are dimerized in EGFP-geph-GCEE as seen in structural studies of isolated gephyrin-E. It should be noted that purified gephyrin exists mainly as a trimer with monomeric E-domains (Schrader et al., 2004; Sola et al., 2004), suggesting that E-domain dimerization is regulated and is dependent on other cellular factors. Biochemical evidence indicates that intramolecular dimerization occurs in geph-GCEE: size exclusion chromatography and dynamic light scattering demonstrated a molecular mass of geph-GCEE of 380-450 kDa, which is very close to the theoretical mass of the trimer (390 kDa), whereas geph-GCE was estimated to be significantly above the calculated value because of a non-globular hydrodynamic radius (Schrader et al., 2004). These data were interpreted as evidence that the additional E-domain efficiently occupies the remaining space within the molecule upon intramolecular dimerization. In addition, the present functional analysis demonstrates that this mutant is enzymatically as active as gephyrin (Fig. 6A).
The increased propensity for auto-aggregation of EGFP-geph-GCEE suggests that E-domain duplication increases the number of interaction domains with other gephyrin molecules, leading to uncontrolled multimerization. As a consequence, endogenous gephyrin might become titrated out upon overexpression of EGFP-geph-GCEE, which would explain the dominant-negative effect and defective postsynaptic clustering of this gephyrin variant. Assuming that intramolecular dimerization occurs in EGFP-geph-GCEE as indicated biochemically, the oligomerization would involve other regions of gephyrin-E than the known dimerization interface.
Specific gephyrin-E Loop 2 residues regulate aggregation and postsynaptic clustering
An intriguing result of this study is the observation that re-introducing two gephyrin-specific residues in EGFP-geph-L2B (Fig. 1B) restored its capacity for aggregation in HEK293 cells and postsynaptic clustering in neurons. The all-or-none aggregation of gephyrin constructs uncovered by manipulating sites located away from the E-domain dimerization interface raises the question whether this region is directly involved in forming higher ordered gephyrin oligomers. It is not known which of the two pairs of residues differing between L2C and L2B (D615K and I620W) are crucial for clustering. However, I620W exchange is somewhat important because I620 points towards an adjacent helix that might move and could subsequently cause conformational changes at neighboring surface residues. Although known gephyrin interactors, such as dynein light-chain, Mena-Vasp and collybistin, have binding sites that are located outside the L2 loop region (Bausen et al., 2006; Giesemann et al., 2003; Harvey et al., 2004), it is conceivable that subdomain 3, in which L2 is localized, is part of the interface involved in gephyrin clustering. Alternatively, surface-exposed residues of L2 might be crucial for specific protein-protein interactions or posttranslational modifications regulating gephyrin postsynaptic clustering.
Gephyrin-E Loop 2 regulates postsynaptic clustering and synapse formation
Compared with the control, the number of gephyrin clusters was increased significantly in neurons transfected with EGFP-geph-L2C (Fig. 7K). This effect might be attributed to the incorporation of new residues derived from the homologous MoeA sequence that cause increased gephyrin clustering. Furthermore, the proportion of postsynaptic clusters was unchanged, indicating that this construct is targeted properly to synaptic sites. Strikingly, the increased number of EGFP-geph-L2C clusters led to a decrease of PSD-95 clusters without affecting the total number of synapses. This observation suggests a mechanism by which presynaptic terminals are targeted towards gephyrin-rich microdomains on the dendrite. A similar observation for excitatory synapses has been made by Prange et al. (Prange et al., 2004) upon overexpression of PSD-95, which led to an increase of glutamatergic synapses with a parallel decrease of GABAergic synapses. Therefore, scaffolding proteins may stabilize specific molecules that are able to induce synapse formation. A limited amount of such synaptogenic factors might be distributed evenly among inhibitory and excitatory postsynaptic scaffolds. After overexpression of a scaffolding protein, such a factor might be trapped at the corresponding scaffold and induce the formation of the corresponding type of synapse, thereby shifting the ratio of excitatory and inhibitory synapses. Possible candidates for such interacting partners are the neuroligins that can induce synapse formation in vitro (Scheiffele et al., 2000) and influence the balance of excitatory and inhibitory synapses in vitro (Levinson and El-Husseini, 2005; Prange et al., 2004) and in vivo (Varoqueaux et al., 2006). By stabilizing such synaptogenic cell-adhesion molecules, gephyrin might exert an indirect synaptogenic effect. Altogether, these findings suggest that gephyrin might play an essential role in controlling the balance between excitatory and inhibitory synapse formation. The possibility that its postsynaptic clustering is regulated dynamically by mechanisms involving sites on gephyrin-E opens a new avenue for investigating inhibitory synapse formation and plasticity.
Materials and Methods
Animals and reagents
Rat embryos were obtained at embryonic day 18 from time mated pregnant Wistar rats (RCC, Basel, Switzerland). All experiments were performed in accordance with international guidelines on animal care and use and were approved by the cantonal veterinary office of Zurich. Reagents were from Sigma (St Louis, MO) or Merck (Darmstadt, Germany), except where indicated.
Gephyrin constructs (Fig. 1A) derived from pcDNA3-based constructs (Schrader et al., 2004) were PCR-cloned into the EcoRI/SalI sites or XhoI/HindIII sites of pEGFP-C2 (Clontech, Mountain View, CA). The gephyrin-E subdomain 3 loop variants (Fig. 1B) were generated based on pQE-gephP2 (Schrader et al., 2004) by fusion PCR and subsequently cloned into pQE-80 or pcDNA3 via KpnI. The corresponding pEGFP-constructs were generated as described above. Residues exchanged with the corresponding sequences present in E. coli MoeA are indicated in Fig. 1B. cDNAs were purified using an endofree Maxiprep (Qiagen, Hombrechtikon, Switzerland). Bacterial expression constructs pQE30-geph-GCE (=PQE30-P2), pQE30-geph-G, pQE30-geph-CE, pQE30-geph-GCEE were described earlier (Schrader et al., 2004). Construction of pQE-geph-P713E and biochemical characterization of the resulting protein was shown by Kim et al. (Kim et al., 2006). cDNAs of rat GABAA receptor α1, β2 and γ2 subunits (Benson et al., 1998) were obtained from P. Seeburg (Heidelberg) and P. Malherbe (Roche).
HEK293 cells (American Type culture collection, Manassas, VA) were cultured at 37°C and 5% CO2 in MEM (Gibco) containing 1% glutamine, 10% FCS and 50 μg/ml gentamycin. They were transfected by calcium phosphate precipitation with 2-5 μg cDNA dissolved in 225 μl sterile water containing 25 μl CaCl2 (2.5 M) and 250 μl BES buffer, incubated for 6-8 hours at 35°C and 3% CO2, washed with PBS and kept at 37°C and 5% CO2 in fresh medium for two days before immunofluorescence staining.
Primary hippocampal and cortical neuron cultures were prepared as described (Brünig et al., 2002). Hippocampal cells to be transfected by magnetofection (Plank et al., 2003) were plated on poly-L-lysine-coated coverslips (Ø15 mm; 40,000 cells) and cultivated at 37°C and 5% CO2 with MEM (Gibco) supplemented with 1 mM sodium pyruvate, 0.45% glucose, 15 mM HEPES, antibioticum/antimycoticum® (Gibco) and 12% NU serum (BD Biosciences, San Jose, CA) (Chudotvorova et al., 2005). Cells were transfected after 6 or 10 DIV and processed for immunofluorescence between 2 and 6 days later (referred to as 6+2, 6+6, 10+2 experiments). Magnetofection (OZ Biosciences, Marseille, France) was performed according to the manufacturer's protocol, using 0.8 μg cDNA and 7 μl lipofectamine for 300 μl transfection mix. For transfection at 10 DIV, cells were conditioned the day before with MEM supplemented with B27, 1 mM sodium pyruvate, 200 mM L-glutamine, 0.23% glucose and 15 mM HEPES. After transfection, the medium was exchanged with serum-free medium supplemented with B27. No conditioning was required for cells transfected at 6 DIV, which were kept in transfection medium. For 6+6 experiments, the culture conditions of 10+2 experiments were replicated.
Transfection by nucleofection (Amaxa Biosystems, Cologne, Germany) (Gartner et al., 2006) was performed with cortical neurons, owing to the high cell density required for this protocol (5×106 cells/100 μl). Acutely dissociated neurons were transfected by using 2 μg cDNA and electroporation program O-03. They were plated at a density of 15,000 cells per coverslip (Ø18 mm) and grown at 37°C and 10% CO2 in serum-free neurobasal medium (Gibco) supplemented with B27, L-glutamine and antibioticum/antimycoticum© up to 21 DIV. Coverslips were placed up-side down above a feeder layer of astrocytes.
Antibodies used were: mouse anti-gephyrin (1:1000, mAb7a, Synaptic Systems, Göttingen, Germany), rabbit anti-gephyrin (1:2000) (Giesemann et al., 2003), anti-synapsin-1 (1:400, Molecular Probes, Eugene, OR), anti-MAP2 (1:2000, Chemicon, Temecula, CA), anti-PSD-95 (1:1000, Affinity BioReagents, Golden, CO), guinea pig anti-GABAA receptor α1 and α2 subunit (Fritschy and Mohler, 1995). The α1 and α2 subunit antibodies were incubated in living cultures (HEK293 cells and neurons, respectively) as described (Brünig et al., 2002). Cells were then rinsed and fixed for 10 minutes with methanol at –20°C. Detection of intracellular proteins was achieved by incubation for 90 minutes at room temperature with primary antibodies diluted in PBS containing 10% normal serum, followed by secondary antibodies coupled to Cy3 (1:500, Jackson ImmunoResearch, West Grove, PA), Cy5 (1:100) or Alexa Fluor 488 (1:1000; Molecular Probes). Finally, coverslips were mounted with fluorescent mounting medium (Dako Cytomation, Carpinteria, CA).
For visualization of dendrites, we either used immunostaining for MAP2 or Vybrant™ DiD cell-labeling solution (1:200, Molecular Probes). This marker was applied to living cells cultures for 20 minutes using the buffer described in Brünig et al. (Brünig et al., 2002). Cultures were subsequently washed three times for 5 minutes, fixed with methanol and processed for further antibody staining as described above.
Image analysis and quantification
Specimens were analyzed by confocal laser-scanning microscopy (LSM 510 Meta, Carl Zeiss, Jena, Germany), using a 100× lens (NA 1.4). Images from each fluorochrome were acquired sequentially using the full dynamic range of the photodetectors and processed with the software Imaris (Bitplane, Zurich, Switzerland). For display, only minimal contrast adjustments were made.
All quantifications were performed on raw images from single confocal sections (pixel size 0.09-0.18 μm). Dominant-negative effects of non-aggregated EGFP-constructs on endogenous gephyrin clustering were determined by counting manually gephyrin-positive clusters (defined by intensity above 30-40% of maximal values) in 30-50 dendritic segments from 8-20 cells per experiment and construct. Dendrites from non-transfected cells were used as a control. In cases where gephyrin clusters were located at cross points between dendrites of transfected and non-transfected cells, we used the mean value calculated from the maximal and minimal possible number of gephyrin clusters located in EGFP-positive dendrites. All counts were performed in three independent experiments after long-term (6+6) and short-term (10+2) expression. For statistical analysis, normalized numerical densities of gephyrin clusters were compared pair-wise between constructs by cumulative distribution analysis (see Table 1; Kolmogorov-Smirnov test).
After transfection of constructs that formed postsynaptic clusters, the size of EGFP clusters was measured in 6+6 experiments using a threshold segmentation algorithm based on cluster size (0.07-0.5 μm2) and intensity (30-40% of maximum value) (MCID M5; Imaging Research, St Catharines, ON, Canada). Statistical comparison was done by cumulative distribution analysis. Postsynaptic clusters identified by their apposition to synapsin-1-positive terminals were quantified with the software Volocity (Improvision, Coventry, England) in dendrites of ten cells from three independent experiments per construct. For EGFP-geph-GCEE (Fig. 1A), postsynaptic clusters were identified by colocalization with the α2 subunit and were counted manually in electronically zoomed images. For statistical analysis (Instat; Graph Pad Software), transfection with EGFP-geph-GCE was taken as control.
The number of EGFP and PSD-95 clusters in cells transfected for EGFP-geph-GCE and EGFP-geph-L2C (Fig. 1) in 6+6 experiments was counted manually in electrically zoomed images in dendrite segments of 20-25 μm in three independent experiments. The analysis was performed blindly using at least 15 cells per experiment. Statistical analysis was performed on pooled data.
Moco synthesis activity
All gephyrin constructs (pQE30-geph-GCE, pQE30-geph-G, pQE30-geph-CE, pQE30-geph-GCEE, pQE30-P713E, pQE80-geph-L1, pQE80-geph-L2B, pQE80-geph-L2C) and empty vector (pQE30) were expressed in E. coli MoeA mutant SE1588 cells induced with 50 μM isopropyl-1-thio-β-D-galactoside at 25°C overnight. Each sample was harvested, washed with nit-1 buffer (50 mM sodium phosphate pH 7.2, 200 mM NaCl, 5 mM EDTA), solubilized in 1 ml nit-1 buffer and homogenized by sonication. After centrifugation, 5-15 μl of protein extract were transferred to 20 μl of desalted Neurospora crassa nit-1 extract, as described (Nason et al., 1971) and supplemented with 5 mM reduced gluthathione. Protein extracts were diluted according to the linear range of the reconstitution. The reaction was incubated anaerobically for 2 hours at room temperature. After addition of 35 μl of 0.1 mM FAD and 18 μl of 0.1 M KNO3 and 0.5 mM NADPH for 20 minutes, reconstituted nitrate reductase activity was determined. One unit Moco activity is defined as reconstituted nit-1 nitrate reductase sufficient to produce an increase at 540 nm of 1.0 absorbance unit per 20-minute reaction time. Activity was either expressed as units per mg crude extract protein or correlated with the activity of geph-GCE by normalizing the activity to the expression level of the investigated variants.
This study was supported by the Swiss National Science Foundation grant Nr. 3100A0-108260 to J.-M.F. and the Deutsche Forschungsgemeinschaft grant Nr. Schw 759/2-4 to G.S., ME 2075/3-1 to J.C.M. and Helmholtz association grant VH-NG-246 to J.C.M. We thank Ela Balic and Thomas Bürli for help with various experiments, Igor Medina and Christophe Pellegrino (INMED, Marseille) for advice with magnetofection, Thomas Grampp and Ruth Keist for excellent technical support and Katrin Fischer for preparing figures.