The function of glutamatergic synapses is not perturbed by severe knockdown of 4.1N and 4.1G expression

AMPA-type glutamate receptors mediate fast excitatory synaptic transmission in the vertebrate brain. Their surface expression at synapses between neurons is regulated in an activity-dependent and activity-independent manner. The protein machinery that regulates synaptic targeting, anchoring and turnover of AMPA receptors consists of several types of specialized scaffolding proteins. The FERM domain scaffolding proteins 4.1G and 4.1N were previously suggested to act jointly in binding and regulating synaptic trafficking of the AMPA receptor subunits GluR1 and GluR4. To determine the functions of 4.1G and 4.1N in vivo, we generated a mutant mouse line that lacks 4.1G entirely and expresses 4.1N at 22% of wild-type levels. These mice had combined 4.1G and 4.1N protein expression in the hippocampus at 12% of wild-type levels (equivalent to 8-10% of combined GluR1 and GluR4 expression levels). They show a moderate reduction in synaptosomal expression levels of the AMPA receptor subunit GluR1 at 3 weeks of age, but no change in basic glutamatergic synaptic transmission and long-term potentiation in the hippocampus. Our study indicates that 4.1G and 4.1N do not have a crucial role in glutamatergic synaptic transmission and the induction and maintenance of long-term plastic changes in synaptic efficacy.


Introduction
The excitatory neurotransmitter glutamate activates α-amino-3hydroxy-5-methylisoxazole-4-propionic acid (AMPA) receptors, Nmethyl-D-aspartate (NMDA) receptors and kainate receptors. Fast excitatory transmission in the brain is mainly mediated by AMPA receptors (AMPARs), and its dynamic alteration is thought to be involved in several cognitive processes, including learning and memory.
One of the most extensively studied models of activity-dependent plastic changes in synaptic strength is long-term potentiation (LTP), which is thought to be crucial for learning and memory. NMDAtype receptors (NMDARs) are important for the induction of LTP, which involves long-lasting changes in synaptic strength caused by activity-dependent modulation of AMPAR-mediated transmission (Malenka and Bear, 2004). In the course of LTP generation, highfrequency stimuli activate AMPARs and NMDARs, leading to subsequent changes in trafficking of AMPARs to the synapse (Barry and Ziff, 2002;Bredt and Nicoll, 2003;Collingridge et al., 2004).
The founding member of the protein 4.1 family, 4.1R (official mouse protein symbol, Epb4.1), was originally identified as a stabilizer of erythrocyte shape, connecting the transmembrane protein glycophorin C (GLPC) and the MAGUK p55 with the spectrin cytoskeleton (Marfatia et al., 1995). In addition to 4.1R, AMPA-type glutamate receptors mediate fast excitatory synaptic transmission in the vertebrate brain. Their surface expression at synapses between neurons is regulated in an activitydependent and activity-independent manner. The protein machinery that regulates synaptic targeting, anchoring and turnover of AMPA receptors consists of several types of specialized scaffolding proteins. The FERM domain scaffolding proteins 4.1G and 4.1N were previously suggested to act jointly in binding and regulating synaptic trafficking of the AMPA receptor subunits GluR1 and GluR4. To determine the functions of 4.1G and 4.1N in vivo, we generated a mutant mouse line that lacks 4.1G entirely and expresses 4.1N at 22% of wild-type levels. These mice had combined 4.1G and 4.1N protein expression in the hippocampus at 12% of wild-type levels (equivalent to 8-

10% of combined GluR1 and GluR4 expression levels). They show a moderate reduction in synaptosomal expression levels of the AMPA receptor subunit GluR1 at 3 weeks of age, but no change in basic glutamatergic synaptic transmission and longterm potentiation in the hippocampus. Our study indicates that 4.1G and 4.1N do not have a crucial role in glutamatergic synaptic transmission and the induction and maintenance of long-term plastic changes in synaptic efficacy.
the 4.1 family includes the members 4.1B (Epb4.1L3), 4.1G (Epb4.1L2) and 4.1N (Epb4.1L1). All four 4.1 proteins share a highly conserved domain structure characterized by the presence of an N-terminal FERM (four point one, ezrin, radixin, moesin) domain, a spectrin-actin-binding domain and a C-terminal domain, all of which interact with multiple transmembrane proteins. The ability of 4.1 proteins to bind very structurally diverse interaction partners via their different protein domains enables them to participate in many different physiological processes in a variety of cell types and tissues.
Proteins 4.1G and 4.1N have been shown to be expressed in neuronal and non-neuronal cells in the brain (Lu et al., 2004;Ohara et al., 2000;Ohno et al., 2005). Protein interactions with the GluR1 and GluR4 subunits of AMPARs were described for the MAGUK SAP97 and the 4.1 proteins 4.1N and 4.1G (Cai et al., 2002;Coleman et al., 2003;Rumbaugh et al., 2003;Shen et al., 2000). The Drosophila 4.1 protein homologue coracle interacts with GluRIIA (Chen et al., 2005) and binding of 4.1N and 4.1G is suggested to regulate surface expression of GluR1 and GluR4 (Coleman et al., 2003;Shen et al., 2000).
In the present study, we examined the role of the 4.1 paralogues 4.1G and 4.1N in hippocampal neurotransmission and synaptic plasticity using mutant mice in which combined 4.1N and 4.1G protein expression in the hippocampus is reduced to 12% of wildtype levels, which is equivalent to 8-10% of combined GluR1 and GluR4 expression levels.

Results
Targeting of the mouse 4.1G and 4.1N genes The 4.1G and 4.1N double-mutant (4.1G/N) mice (abbreviated as DKO in figures) were generated by a conditional mutagenesis strategy. In both cases, two LoxP sites flanking the first coding exon and a neomycin-resistance gene flanked by frt sites were introduced by homologous recombination in embryonic stem cells (129/ola) (Fig.  1A,D). A fragment representing bp 25,131,023-25,131,494 on band A3 of Chromosome 10 and bp 1-471 in the 4.1G mRNA (AJ542537) was floxed in the mouse 4.1G gene Epb41l2 (GeneID 13822). The 4.1N gene Epb41l1 (GeneID 13821) is located on band H2 of Chromosome 2. The region representing bp 156,185,185,551 in the 4.1N gene and bp 206-396 in 4.1N mRNA (AF061283) was floxed. Of 192 screened embryonic stem (ES) cell clones in both cases, nine clones with the desired 4.1G recombination and 15 clones with the desired 4.1N recombination were identified by Southern blot analysis with a 5Ј external probe (probe 1 for 4.1G, Fig. 1A; probe 2 for 4.1N, Fig. 1D). Two ES cell clones of each gene were injected into C57BL/6 blastocysts. The offspring of chimeric males and subsequent generations were genotyped by PCR (Fig. 1C,F,G). The resulting recombinant single 4.1G +/and 4.1N +/mice were bred with EIIaCre deleter mice (Lakso et al., 1996), which resulted in the loss of the first coding exon in both cases (Fig. 1G). The 4.1G-knockout mice were then crossed with 4.1N mutant mice to breed 4.1G/N double-mutant animals. The offspring were bred to homozygosity and selected for the absence of the Cre transgene.
Journal of Cell Science 122 (5) Neither single mutants nor homozygous 4.1G/N double mutants showed obvious phenotypic alterations. The 4.1G/N double-mutant mice were born at the expected mendelian frequency, and no gross behavioural abnormalities were apparent. The life spans were similar to those of control animals (data not shown).
Examination of the double-mutant animals by MRI using T1and T2-weighted images showed only slight changes in the gross anatomy of the brain in 3-week-old animals ( Fig. 2A). Quantitative volumetric analyses revealed mild, although significant, reductions in whole brain volume and cerebellar volume in young 4.1G/N double-mutant mice (Fig. 2C,D). These changes seem to be compensated during development, as they were absent in 7-monthold double-mutant mice ( Fig. 2B-E).
Immunohistochemical staining was performed to determine the number of matched and mismatched pre-and postsynaptic components of glutamatergic synapses in the stratum radiatum of hippocampal CA1 area of 4.1G/N double mutants and wild-type (WT) controls at the age of 3 weeks. Neither the spatial distribution nor the densities of puncta positive for VGLUT1/VGLUT2, ProSAP1 or both were altered significantly in 4.1G/N double-mutant hippocampus (Fig. 3A,B). Moreover, the ultrastructure of synapses formed in double-mutant animals was indistinguishable from that of WT controls (Fig. 3C). Mature synaptic specializations were observed in the CA1 region of 3-week-old mice in both the mutant and WT animals. Synaptic parameters, such as the number of synapses per 100 μm 2 , the length of the postsynaptic density and the width of the postsynaptic density, were similar in both groups ( Fig. 3D) (WT,n=3;DKO,n=3). Taken together, these data indicate that the expression of proteins 4.1G and 4.1N is not critical for proper formation of glutamatergic synapses.

Expression of 4.1 proteins in 4.1G/N double-mutant mice
Western blot analyses confirmed that 4.1G was completely absent in 4.1G/N double-mutant mice (Fig. 4A), whereas the short variant of two 4.1N isoforms was still present ( Fig. 4B; Fig. 5B,C), albeit reduced to a level of 30% of that in control preparations (Fig. 5D). The level of 4.1B protein was unchanged in the 4.1G/N doublemutant animals (supplementary material Fig. S1). The ratio between the short and the long variant of 4.1N in the hippocampus is 2.6±0.7 (n=6, calculated based on a densitometric analysis of the data shown in Fig. 5B,C). Thus, a 70% reduction of the short form of 4.1N and a complete elimination of the long 4.1N variant in our mutants is equivalent to a 78% reduction of total 4.1N expression in the hippocampus of 4.1N/G double-mutant mice. The ratio between 4.1N and 4.1G protein expression is 1:0.78 in rat total hippocampus  and 1:0.94 in mouse hippocampus postsynaptic density preparations (Trinidad et al., 2008). Thus, the reduction of combined 4.1G/N protein expression in our homozygous double-mutant hippocampus is estimated to be 88-89% compared with WT controls.
To deduce the primary structure of the residually expressed 4.1N isoform, we performed western blot analyses with three domainspecific antibodies: anti-4.1N exon 2 directed against the first 15 amino acids encoded by exon 2, anti-4.1N monoclonal antibody (BD Biosciences), which is specific for a region of the spectrinactin-binding domain, and an anti-CTD antibody against the Cterminal domain of 4.1N (Fig. 5E). Fig. 5A shows that the anti- To characterize the residual 4.1N splicing variant in the 4.1G/N double mutants, we combined RACE (rapid amplification of cDNA ends) experiments and database analyses. The disappearance of the lower WT 4.1N band in western blots stained with the anti-4.1N exon 2 antibody (Fig. 5A, left panel) indicated an alternative use of the first coding exon, as previously reported for 4.1R and 4.1B (Conboy, 1991;Gascard, 2004). There are indications of alternative start codon use in human (NM_177966) and equine (XM_001501911) 4.1N mRNA, but no corresponding 4.1N mRNA or ESTs could be identified for the mouse.
To define the full brain-specific use of 4.1N exons in WT and 4.1G/N double mutants, we generated 5ЈRACE and 3ЈRACE transcript sequences (Fig. 5G). All transcripts cloned experimentally by 5ЈRACE from WT (12 clones) included exon 2. Exon 3 never contributed to the 5ЈRACE sequences in the WT.
The 3Ј RACE-PCRs generated two different populations of transcripts. Four of nine clones encoded the high molecular weight class of 4.1N including exon 17B. The other five clones corresponded to the lower 4.1N form. To our surprise, in the 4.1G/N double mutants all 5ЈRACE-PCR transcripts (nine clones) started with exon 3, which is usually removed from mature mRNA, and the first initiation codon was localized in exon 4. All 3ЈRACE PCR clones from the 4.1G/N double mutants lacked exon 17A and exon 17B. As the anti-4.1N exon 2 peptide antibody is directed against the first 15 amino acids of the N-terminus, it is possible that this epitope is not accessible for the antibody in the WT 4.1N short variant. The absence of a band in the 4.1G/N double mutants can be explained by the genetic knockout of exon 2, which includes the epitope of the anti-4.1N exon 2 antibody. We conclude from the results obtained by western blots and RACE experiments that the residually expressed protein in 4.1G/N double mutants is similar to the short form expressed in the WT, but differs in its N-terminal sequence.
Next, we determined expression levels of GluR1, GluR2/3, GluR4 and NR1 in young adult 4.1G/N double-mutant mice by quantitative western blot analyses of synaptosome preparations from Journal of Cell Science 122 (5) whole-brain homogenates (Fig. 6A). Here, we found significantly reduced synaptic GluR1 protein levels in synaptosomes from 4.1G/N double-mutant mice (WT, n=6; DKO, n=6; **P<0.01). Furthermore, we measured GluR1, GluR2/3 and GluR4 concentrations in Triton-X-100-insoluble postsynaptic density (PSD) preparations of hippocampi obtained from young adult 4.1G/N double-mutant mice and WT controls. Supporting our findings in synaptosomes, we found significant changes in the levels of the GluR1 subunit ( Fig. 6B) (WT, n=7; DKO, n=7; *P<0.05). GluR2/3 levels were also significantly decreased in PSD preparations, although a direct interaction between 4.1G or 4.1N and GluR2/3 has not been described. In vivo, AMPARs usually form heteromers composed of different subunits. A reduction in the GluR1 subunit might therefore be accompanied by a reduction of other GluR subunits such as GluR2/3.

Synaptic and extrasynaptic AMPAR responses are normal in young 4.1G/N double-mutant mice
The Schaffer collateral synapses in the hippocampus are very well characterized with respect to their short-and long-term synaptic plasticity. It is generally accepted that the induction of long-term potentiation (LTP) in area CA1 is NMDAR dependent (Malenka and Bear, 2004). A leading hypothesis states that LTP in area CA1 is activity dependent and relies on insertion of AMPARs as GluR1/GluR2-containing receptors are driven into the synapses after the induction of LTP (Hayashi et al., 2000).  gene in brain. Numbering is derived from mouse Epb41l1 as described (Ramez et al., 2003). The colour code is modelled on the domain structure shown in E. Dotted boxes denote nontranscribed exons. The positions of the specific 5ЈRACE primers and the specific 3ЈRACE primers are indicated by arrows. (G) Schematic representation of the obtained RACE transcripts from WT and 4.1G/N double mutants.
As 4.1 proteins were shown to interact with the GluR1 subunit and because we found a reduction of GluR1 expression in synaptosomes of 4.1G/N double-mutant mice (Fig. 6A), we tested whether the loss of protein 4.1G together with the strong reduction of protein 4.1N leads to alterations in synaptic transmission and changes in LTP in young adult 4.1G/N double mutants. To measure synaptic responses in hippocampal brain slices of 4.1G/N doublemutant mice, we stimulated the Schaffer collaterals in area CA1 and recorded postsynaptic field potentials (fEPSPs) in the stratum radiatum of area CA1. Comparing the size of the presynaptic afferent volley (representative of presynaptic excitability), with the slope of the fEPSP (representing postsynaptic responsiveness), we found no significant differences in the fEPSP slopes at various afferent volley amplitudes (WT, n=10; DKO, n=13) (Fig. 7A). Using whole-cell patch-clamp recordings, we found that the ratio of AMPAR-to NMDAR-mediated excitatory postsynaptic current (EPSC) is not altered in 4.1G/N double mutants (AMPA/NMDA ratio: WT, 4.8±0.9, n=13; DKO, 6.1±1.1, n=11; P=0.39; AMPARmediated EPSC: WT, 241±20 pA; DKO, 281±26 pA, P=0.23; NMDAR-mediated component: WT, 66±10 pA; DKO, 54±9 pA, P=0.35) (Fig. 7B).
Paired-pulse facilitation (PPF) is a form of short-term plasticity and generally assumed to be of presynaptic origin. Using a pairedpulse stimulation protocol, the second response is facilitated compared with the first response within a short time-window. Using different stimulation intervals varying from 50 mseconds to 500 mseconds, we found that PPF in field recordings is not significantly changed in 4.1G/N double mutants (WT, 1.44±0.06, n=7; DKO, 1.39±0.03, n=10; P=0.4). Whole-cell recordings confirmed these data at an interstimulus interval of 50 mseconds (WT PPF, 1.47±0.07, n=15; DKO PPF, 1.47±0.04, n=15; P=0.97) (Fig. 7C,D). We also recorded miniature EPSCs (mEPSCs) in the presence of the sodium channel blocker tetrodotoxin (TTX, 1 μM) and the AMPAR desensitization inhibitor cyclothiazide (CTZ, 100 μM). AMPAR-mediated mEPSCs showed neither a difference in frequency nor a difference in amplitude recorded at -60 mV between WT and 4.1G/N double-mutant animals. The mean mEPSC frequency was 0.9±0.1 Hz in WT mice and 1.2±0.1 Hz in 4.1G/N double mutants (WT, n=5; DKO, n=9; P=0.11) and the mean amplitude of the AMPAR-mediated mEPSCs was 18±2 pA in WT animals and 16±1 pA in 4.1G/N double-mutant mice (Fig. 8)   Next, we tested for alterations in the biophysical properties of AMPAR-mediated currents by measuring the rectification index, which was not significantly different between WT and 4.1G/N double-mutant neurons (2.8±0.4, n=6 and 2.7±0.4, n=6, respectively, P=0.8) (see supplementary material Fig. S2). Also, pharmacological challenging of AMPARs with the relatively selective GluR2 antagonist Naphthyl-acetyl-spermine (100 μM) revealed no differences in the slope of fEPSP recordings between the two genotypes (WT slope, 97.7±0.5% of control; DKO slope, 99.0±3% of control, n=4). The effectivity of the drug was confirmed by EPSC recordings from interneurons, where a substantial amplitude reduction was observed (see supplementary material Fig. S3). We thus did not find any indication for a change in AMPAR subunit composition in 4.1G/N double-mutant neurons.
As it is known that 4.1 proteins interact with AMPARs, we tested whether extrasynaptically located AMPAR are altered in 4.1G/N double-mutant mice. In the presence of TTX and CTZ, we bathapplied S-AMPA (100 nM) for 5 minutes and recorded the wholecell current. By this procedure AMPARs are activated in the synaptic cleft, on the dendrites and on the somata. Application of S-AMPA led to a robust inward current of approximately 400-500 pA in WT and 4.1G/N double-mutant cells, with no significant differences between the tested genotypes (WT, 510±47 pA, n=4; DKO, 462±51 pA, n=5; P=0.52) (Fig. 9). Taken together, these results show that neither synaptic nor extrasynaptic AMPAR function is affected in 4.1G/N double mutants.

Long-term potentiation is normal in young 4.1G/N doublemutant mice
We studied long-term synaptic plasticity in WT and 4.1G/N doublemutant mice using field potential recordings. After recording stable fEPSPs in area CA1 for 10-20 minutes, the Schaffer collaterals were stimulated tetanically (four 1-second trains of 100 Hz, separated by 20 seconds). LTP was stable for 30-40 minutes after the tetanic stimulation in WT and 4.1G/N double mutants. No significant differences between WT and 4.1G/N double-mutant mice were found in the fEPSP slope 30-40 minutes after induction of LTP ( Fig.  10) (WT, 137±6%, n=10; DKO, 135±7%, n=11; P=0.8). These data show that neither basal synaptic transmission nor short-term or longterm plasticity are altered in 4.1G/N double-mutant mice.

Discussion
In the present study, we used 4.1G/N double-mutant mice to assess the role of proteins 4.1G and 4.1N in glutamate receptor trafficking and synaptic anchoring, in hippocampal synaptic transmission and in synaptic plasticity. We found that (1) in young adult mice the GluR1 expression levels in 4.1G/N double mutants are reduced in synaptosomal and PSD fractions (Fig. 6); (2) GluR2/3 expression levels in 4.1G/N double mutants are reduced in PSD fractions (Fig.  6); (3) AMPAR-mediated glutamatergic synaptic transmission in  (Figs 7-9), which is not consistent with the previously published notion that proteins 4.1G and 4.1N bind GluR1 and tether it to the postsynaptic cytoskeleton (Shen et al., 2000); and (4) LTP at CA3-CA1 Schaffer collaterals in 4.1G/N double mutants is not altered (Fig. 10), which argues against a crucial role of proteins 4.1G and 4.1N in the activitydependent insertion of GluR1-containing AMPARs into potentiated synapses (Hayashi et al., 2000). The two AMPA receptor subunits that were shown to bind 4.1 proteins, GluR1 and GluR4, bind to both 4.1N and 4.1G (Shen et al., 2000;Coleman et al., 2003). Thus, 4.1N and 4.1G can be viewed to have a joint role in GluR1 and GluR4 trafficking. Our knockout strategy eliminated 4.1G expression entirely and 4.1N expression very strongly. The combined 4.1N/G protein expression in double-mutant hippocampus was reduced to 11-12% of WT levels. If 4.1G and 4.1N acted jointly as crucial stoichiometric interaction partners and scaffold proteins of GluR1 and GluR4 under normal circumstances, the residual expression of 4.1N in the 4.1G/N double-mutant mice is unlikely to be sufficient for the maintenance of normal function, because the relative expression levels of 4.1G, 4.1N, GluR1 and GluR4 in mouse hippocampus are 0.89-0.99, 0.95-1.05, 1.48-1.61, and 0.87-0.95, respectively (Trinidad et al., 2008). Thus, total deletion of 4.1G and a reduction of 4.1N expression by 78% would lead to a massively substoichiometric ratio between combined 4.1G and 4.1N, and combined GluR1 and GluR4 expression in mouse hippocampus of at least 0.08:1, and 0.1:1, at most. Such a massive deficiency of 4.1G and 4.1N expression compared with GluR1 and GluR4 expression in the double-mutant hippocampus would be expected to have severe functional consequences on glutamatergic transmission if 4.1G and 4.1N had a crucial role in GluR1 and GluR4 trafficking and anchoring, but our data show that this is not the case.

4.1G/N double mutants is not changed
The fact that GluR1 levels are reduced in 4.1G/N double mutants supports the notion that proteins 4.1G and 4.1N bind GluR1 and stabilize it (Shen et al., 2000). However, the finding that neither basal AMPAR-mediated synaptic transmission nor synaptic plasticity in the hippocampus are altered in 4.1G/N double mutants makes an essential role of proteins 4.1N and 4.1G in the trafficking, function, and dynamics of AMPARs in vivo rather unlikely. This lack of phenotypic changes in 4.1G/N double mutants is either due to the fact that previously published in vitro experiments on the role of 4.1G and 4.1N in GluR1 trafficking and anchoring detected cell-culture specific phenotypes that are not relevant in vivo, or is caused by the fact that multiple AMPAR scaffold proteins operate in parallel at synapses and can substitute for the loss of the two scaffold proteins studied here. We favour the latter notion, as outlined below.
Striking functional redundancy has been reported for other postsynaptic AMPAR scaffold protein families that have been studied in detail. Targeted truncation of PSD95 in mutant mice, for example, does not affect AMPAR-mediated synaptic transmission in the hippocampus (Migaud et al., 1998). Similarly, single deletions of PSD95 or PSD93 do not cause functional changes in basal synaptic transmission (Elias et al., 2006). Only the simultaneous deletion of both PSD93 and PSD95 leads to marked reductions in glutamatergic synaptic transmission (Elias et al., 2006). By contrast, individual knockdown of either PSD93 or PSD95 using an shRNA strategy causes impaired glutamatergic transmission in hippocampal slice cultures (Elias et al., 2006). These findings indicate that PSD93 and PSD95 are functionally redundant and can compensate for the loss of the other MAGUK if sufficient time is available for these compensatory mechanisms to take place, as is the case in the deletion mutant mice, but apparently not when the proteins are knocked down by shRNAs.
Two recent studies found the interaction of stargazin and other TARPs with AMPARs to be the most robust and least transient compared with other putative interactions, including those with 4.1 proteins (Fukata et al., 2005;Vandenberghe et al., 2005). Indeed, TARPs function as auxiliary subunits of AMPARs and are important determinants of AMPAR trafficking and synaptic anchoring. An 80-90% reduction in hippocampal AMPAR protein levels and a 30-40% reduction in synaptic transmission are observed in γ-8knockout mice, and extrasynaptic AMPARs were also reduced by 80-90% (Rouach et al., 2005). Similarly, cerebellar granule cells from mice lacking Stargazin lack synaptic and extrasynaptic AMPARs (Chen et al., 2000). These data show that γ-8 and Stargazin determine AMPAR stability and synaptic recruitment. Furthermore, the data obtained from γ-8-deficient mice indicate that even strong reductions in overall AMPAR levels do not cause similarly severe reductions in synaptic AMPAR function. This apparent lack of a linear correlation between overall AMPAR levels and AMPAR function at synapses might explain why, in the case of the 4.1G/N double-mutant mice described here, a 20-30% reduction of GluR1 levels at postsynapses does not cause a concomitant reduction of synaptic transmission in the hippocampus.
In summary, the present study indicates that proteins 4.1G and 4.1N are generally dispensable for proper function of glutamatergic Journal of Cell Science 122 (5) synapses in vivo. Data obtained in in vitro studies on these proteins can therefore not be directly extrapolated to the in vivo situation. It is likely that the lack of phenotypic changes in glutamatergic hippocampal synapses of 4.1G/N double mutants is due to a redundancy of these proteins with other 4.1 isoforms or with other postsynaptic scaffold proteins. This notion is supported by the fact that deletion of the only Drosophila 4.1 homologue (coracle), causes the loss of coracle-binding GluRIIA receptors from postsynaptic sites, whereas GluRIIB receptors, whose C-terminus does not bind coracle, are not affected, indicating an important role of 4.1 proteins in glutamate receptor anchoring at invertebrate synapses (Chen et al., 2005). However, whether the same is true for mammalian 4.1 proteins will have to be determined in future genetic studies on mice lacking all 4.1 proteins.

Generation of 4.1G/N double-mutant mice
For the generation of the 4.1N and the 4.1G single mutant mouse lines (Thomas and Capecchi, 1987) we used genomic clones that had been isolated from a 129SV mouse genomic λ FIXII library (Stratagene). DNA fragments of these clones were used to construct the targeting vectors (Fig. 1A,D). The linearized DNAs of the 4.1N and 4.1G targeting vectors were electroporated into embryonic stem cells (E14), and colonies were selected with G418 and Ganciclovir. Double-resistant clones were analyzed by Southern blotting (Fig. 1B,E). Clones containing the recombined genes were expanded and injected into mouse blastocysts to obtain highly chimeric mice that transmitted the mutation through the germ line. Germ-line transmission was confirmed by Southern blotting of tail-biopsy genomic DNA. Subsequently, routine genotyping of DNA from tail biopsies was performed by PCR (Fig. 1C,F,G).

In vivo 3D MRI
WT and 4.1G/N double-mutant mice underwent MRI measurements at the age of 3 weeks and 7 months. All measurements were obtained at 2.35 T (Bruker BioSpin). In vivo 3D T1-weighted (FLASH, TR/TE=17/7.58 mseconds, flip angle 25°) and T2-weighted images (FSE, TR/TE=3000/60.95 mseconds, 8 echoes, inter-echo spacing=14.4 mseconds) were acquired with an isotropic resolution of 117 μm and a total measurement time of about 195 minutes per animal. The whole brain volume (excluding brainstem), size of cerebellum and size of lateral ventricles were determined in T1-weighted images by manually drawing respective regions of interests on up to 50 contiguous horizontal MRI sections.
Immunostaining and light microscopic quantification 4-week-old wild-type and 4.1G/N double-mutant mice were deeply anesthetized with tribromoethanol and perfused transcardially with 4% paraformaldehyde in 0.1 M PB (sodium phosphate buffer, pH 7.4). Brains were removed and fixed overnight in 4% paraformaldehyde, cryoprotected in sucrose and frozen on dry ice. Serial sagittal cryostat sections (18 μm thick) were collected on slides and stored at -80°C until used. After blocking in PB containing 5% normal goat serum and 0.25% Triton X-100, the sections were stained using antibodies against VGLUT1/2 (both guinea-pig, 1:10,000, Chemicon) and ProSAP1 (Boeckers et al., 1999(Boeckers et al., ) (rabbit, 1:1000. The sections were washed three times in PB and then incubated with Alexa Fluor 488or Alexa Fluor 555-labelled goat anti-guinea pig or goat anti-rabbit IgG secondary antibodies (Molecular Probes) at room temperature. Coverslips were mounted with Mowiol (Calbiochem). All images were acquired as single layers at zoom factor 2 using a Leica inverted confocal laser scanning microscope (DM IRE 2) with a ϫ63 oil-immersion lens. To allow for intensity comparisons, the gain and offset were held constant across images. Images were imported into the AnalySIS software (Soft-Imaging Systems) and puncta within the hippocampal CA1 stratum radiatum were quantified by touch-count analysis.

Ultrastructural analyses and quantification
Brains of 3-week-old mice were processed according to standard procedures. In brief, mice were perfused with 4% paraformaldehyde and 0.25% glutaraldehyde in PBS. Sections of hippocampus (2 mm thick) were washed, osmicated for 4 hours (1% OsO 4 in PBS), dehydrated through a graded series of ethanol and propylene oxide, and embedded in Epon by polymerization for 12 hours at 60°C. Ultrathin sections of the CA1 region were cut, contrasted with uranyl acetate and lead citrate, and observed in a LEO 912AB transmission electron microscope (Zeiss). Digital images were taken with a ProScan CCD camera and analyzed with CellP analysis software (Olympus).

RNA isolation, RT-PCR and RNA-ligase-mediated (RLM)-RACE
Total RNA was isolated from WT and 4.1G/N double-mutant whole brains using the RNeasy Lipid Tissue Kit (Qiagen). First-strand cDNA was synthesized with Superscript III (Invitrogen) from adaptor-ligated RNA, which was obtained according to the manual of the FirstChoice RLM-RACE Kit (Ambion). A gene-specific primer or an oligo dT primer was utilized for this purpose. 5ЈRACE of 4.1N was accomplished according to the manual of the FirstChoice RLM-RACE Kit (Ambion). The 5ЈRACE was performed using two gene-specific nested primers as the reverse primers and the kit provided nested primers (outer primer and inner primer) as the forward primers. Both PCR reactions (outer and inner PCR) were performed at an annealing temperature of 59°C. The amplified products of two different 5ЈRACE gene-specific primer combinations were purified using QIAquick Gel Extraction Kit (Qiagen), cloned into pCR2.1TOPO (Invitrogen) and sequenced. The 3ЈRACE of 4.1N was carried out with two gene-specific nested primers as the forward primers and the nested primers provided with the kit as the reverse primers. 3ЈRACE PCR products were analyzed as described above for the 5ЈRACE. The sequences of primers are: first strand synthesis primer, 5Ј-CGTGGGCGTGGTGGTTTGA-3Ј; gene-specific 5Ј RACE primers, 5Ј-TCGGCATCATAGTCACCCAGTTCA-3Ј; 5Ј-CCAGAAGGGCATGCGTGA -CAAAG-3Ј; 5Ј-TGCGCTCCCGATCCCAGTCT-3Ј; 5Ј-GATGCTGGCCTGGTGCT -TC AAC-3Ј; 5ЈRACE outer primer, 5Ј-GCT GAT GGC GAT GAA TGA ACA CTG-3Ј; 5ЈRACE inner primer, 5Ј-CGC GGA TCC GAA CAC TGC GTT TGC TGG CTT TGA TG-3Ј; gene-specific 3ЈRACE primers, 5Ј-CCTCAGTCAGTGAGA -ATCACGATGC-3Ј; 5Ј-TGAAGGAGCCCAACAGCAAACT-3Ј; 3ЈRACE outer primer, 5Ј-GCG AGC ACA GAA TTA ATA CGA CT-3Ј; 3ЈRACE inner primer: 5Ј-CGC GGA TCC GAA TTA ATA CGA CTC ACT ATA GG-3Ј.

PSD preparation
3-week-old 4.1G/N double-mutant mice and WT control animals were sacrificed and the hippocampi were isolated. The hippocampi of each animal were homogenized in ten volumes of homogenisation buffer (320 mM sucrose, 20 mM HEPES-KOH (pH 7.4), 5 mM EDTA, 0.1 mM PMSF). The homogenate was centrifuged for 10 min at 1000 g. Supernatants S1 were recentrifuged for 10 minutes at 15,000 g, and the resulting P2 pellets were resuspended in resuspension buffer (20 mM Tris-HCl pH 7.4, 2 mM EGTA, 1 mM PMSF). Protein concentration was determined by Bradford assay. For solubilization, the protein concentration was set to 2 mg/ml, and Triton X-100 was added to a final concentration of 1%. The samples were sonicated, followed by agitation for 1 hour at 4°C. Extracts were centrifuged at 100,000 g for 1 hour. The supernatant is the Triton-soluble fraction. The PSD-pellet (Triton-insoluble fraction) was resuspended in 0.1 volume of 1% SDS, 2 mM EDTA, and 20 mM Tris-HCl (pH 7.4). Then 0.9 volume of resuspension buffer was added and the samples were agitated for 1 hour at 4°C. Protein concentration was determined by BCA assay. 25 μg protein was loaded per lane and analyzed by SDS-PAGE followed by western blotting. Densitometric analysis was carried out using ImageJ software, and the signals were normalized to β-tubulin signals. Significance was determined by Student's ttest.

Electrophysiology
WT and 4.1G/N double-mutant mice (3-to 5-weeks old) of both genders were decapitated under anesthesia and the brains were quickly removed. Horizontal hippocampal mouse brain slices (300 μm thick) were prepared in saccharose-based ACSF containing 87 mM NaCl, 1.25 mM NaH 2 PO 4 , 2.5 mM KCl, 26 mM NaHCO 3 , 7 mM MgCl 2 , 0.5 mM CaCl 2 , 75 mM saccharose and 25 mM glucose. After incubation for 30 minutes at 36°C, slices were transferred to physiological ACSF solution (119 mM NaCl, 1 mM NaH 2 PO 4 , 2.5 mM KCl, 1.3 mM MgSO 4 , 2.5 mM CaCl 2 , 26 mM NaHCO 3 , 10 mM glucose, saturated with 95% O 2 and 5% CO 2 at pH 7.4) and kept at room temperature. Field potential recordings were performed with low-resistance patch-clamp electrodes filled with ACSF. Field EPSPs were recorded in stratum radiatum in area CA1. Schaffer collaterals were stimulated with a frequency of 0.05 Hz. Single cell recordings were performed in whole cell patch-clamp mode. Patchclamp electrodes (electrode resistance 2-5 MΩ) were filled with 117.5 mM Csgluconate, 2.5 mM CsCl, 8 mM NaCl, 10 mM HEPES, 10.0 mM TEA, 0.2 mM EGTA, 4 mM Mg-ATP, 0.3 mM Na 3 GTP and QX-314 2; pH was adjusted to 7.2 with CsOH. Access resistance ranged from 6 to 20 MΩ and was continuously monitored throughout the experiment. Recordings were discarded when the series resistance changed more than 20%. All patch-clamp experiments were done in the presence of the GABA A receptor-antagonist gabazine (SR 95531, 1 μM; Sigma) and 4 mM MgSO 4 and CaCl 2 . AMPA/NMDA ratios were obtained by evoking singlecomponent EPSCs at -60 mV and a dual-component EPSCs at +40 mV. The NMDAmediated portion of the dual-component current at +40 mV was measured 100 mseconds after the stimulation artefact. AMPAR-rectification index was determined (in the presence of 40 μM APV) as current amplitude at a holding potential of -60 mV divided by the amplitude at + 40 mV. Miniature EPSCs were recorded in the presence of tetrodotoxin (TTX, 1 μM) and cyclothiazide (100 μM). AMPA-mediated whole-cell currents were obtained by bath application of 100 nM S-AMPA for five minutes in the presence of TTX and cyclothiazide. In field recordings, LTP was induced by four tetani of high-frequency stimulation at 100 Hz for 1 second with 20 second intertrain intervals. The magnitude of LTP was determined by averaging the responses collected during the last 5 minutes of each experiment. Paired-pulse facilitation (P2/P1) in the field and whole-cell recordings was investigated by analysing the ratio of second the first synaptic response. Data are expressed as mean ± s.e.m., and statistical comparison was done by applying Student's t-test (Excel, Microsoft). The significance level was set to P<0.05.