Synaptic Scaffolding Proteins in Rat Brain ANKYRIN REPEATS OF THE MULTIDOMAIN Shank PROTEIN FAMILY INTERACT WITH THE CYTOSKELETAL PROTEIN (cid:1) -FODRIN*

,

The postsynaptic density is the ultrastructural entity containing the neurotransmitter reception apparatus of excitatory synapses in the brain. A recently identified family of multidomain proteins termed Src homology 3 domain and ankyrin repeat-containing (Shank), also known as proline-rich synapse-associated protein/somatostatin receptor-interacting protein, plays a central role in organizing the subsynaptic scaffold by interacting with several synaptic proteins including the glutamate receptors. We used the Nterminal ankyrin repeats of Shank1 and -3 to search for interacting proteins by yeast two-hybrid screening and by affinity chromatography. By cDNA sequencing and mass spectrometry the cytoskeletal protein ␣-fodrin was identified as an interacting molecule. The interaction was verified by pull-down assays and by coimmunoprecipitation experiments from transfected cells and brain extracts. Mapping of the interacting domains of ␣-fodrin revealed that the highly conserved spectrin repeat 21 is sufficient to bind to the ankyrin repeats. Both interacting partners are coexpressed widely in the rat brain and are colocalized in synapses of hippocampal cultures. Our data indicate that the Shank1 and -3 family members provide multiple independent connections between synaptic glutamate receptor complexes and the cytoskeleton.
The search for proteins that are localized at the PSD revealed a family of multidomain proteins termed proline-rich synapse-associated proteins 1 and 2 or somatostatin receptorinteracting protein. For convenience and in accord with the most frequently used nomenclature, we will refer to these proteins as Src homology 3 domain and ankyrin repeat-containing (Shank) proteins (see Scheme 1). These proteins may act as a molecular interface between neurotransmitter receptors and cell adhesion molecules and the actin-based cytoskeleton (5,7,9). To date three members of this family, Shank1-3, have been identified (4 -11), containing multiple protein-protein interaction domains including ankyrin repeats, an Src homology 3 (SH3) domain, a PSD-95/discs large/ZO-1 (PDZ) domain, several proline-rich regions, and a C-terminal sterile ␣-motif domain (Scheme 1). The highly conserved PDZ domain has been shown to interact with the C terminus of several different proteins such as the postsynaptic density protein synapseassociated protein-associated protein/guanylate kinase-associated protein (4,6,9) and several G-protein-coupled receptors, including the somatostatin receptor subtype 2 (5, 10) and the calcium-independent receptor for ␣-latrotoxin (12,13). Via synapse-associated protein-associated protein/guanylate kinaseassociated protein, the PDZ domain of the Shank proteins is linked to the N-methyl-D-aspartate receptor-PSD-95 complex, whereas one of their proline-rich regions serves as a docking site for the metabotropic glutamate receptor-binding protein Homer (11). In addition, the SH3 domain has been proposed to interact indirectly with the ␣-amino-3-hydroxy-5-methylisoxazole-4-propionate-type glutamate receptor-interacting protein (14). Taken together, these data suggest that the Shank proteins act as a master scaffold for glutamate receptor complexes in postsynaptic specialization (14). Firm contact of this scaffold with the cytoskeleton is then established via the interaction of another proline-rich region of Shanks and the SH3 domain of the F-actin-binding protein cortactin (8).
To date, little is known of potential protein candidates interacting with the N-terminally located ankyrin repeats of Shank1 and -3. First described by Breeden und Nasmyth (15) in two yeast cell cycle regulator proteins (Swi6p and CDC10p), ankyrin repeats were named after repetitive units present in the ubiquitous adapter protein ankyrin (16) and are known for their interaction with many diverse proteins (17). Here we report on a novel interaction of the ankyrin repeats of Shank1 and -3 with a component of the membrane-associated cytoskeleton ␣-fodrin (also known as brain ␣-spectrin), another multidomain protein that contains 22 spectrin repeats, one SH3 domain, and two EF-hand motifs proximal to the C terminus of the protein (18 -20). Our data indicate that the complexes consisting of ␣-fodrin and either Shank1 or Shank3 may represent a dynamic substructure of the synapse responsible for alterations of the functional architecture that occur during the organization and reorganization of spines and synapses in the central nervous system (21,22).

MATERIALS AND METHODS
Expression of Fusion Proteins-A cDNA fragment encompassing the ankyrin repeat region (six repeats) of human Shank1 was cloned into the bacterial expression vector pQE-30 (Qiagen, Hilden, Germany), which allows the expression of 6xHis-tagged fusion proteins. His 6 -Shank1 189 -399 was expressed in Escherichia coli strain TopF10Ј and purified on Ni-chelating Sepharose (Qiagen) following the manufacturer's instructions. cDNA fragments coding for the C-terminal two spectrin repeats and the EF-hands of rat ␣-fodrin as well as the EF-hands alone were cloned into the pGEX4T-2 glutathione S-transferase fusion protein expression vector (AP Biotech, Freiburg, Germany); proteins were expressed in TopF10Ј cells and purified on glutathione-Sepharose (AP Biotech).
Purification of Proteins Binding to Shank1-Affinity purification of His 6 -Shank1 189 -399 -binding proteins followed a protocol essentially as described by Firestein and Bredt (23). After dialysis into 0.1 M NaHCO 3 and 0.5 M NaCl, 10 mg of purified protein were coupled to 5 ml of N-hydroxy-succinimidyl-Sepharose (Amersham Pharmacia Biotech). Bovine serum albumin (BSA) was also coupled to N-hydroxy-succinimidyl-Sepharose (5 mg BSA/ml of Sepharose). Brains of 25 adult rats were homogenized in 200 ml of buffer A (20 mM HEPES, pH 7.4, 125 mM NaCl, 10% glycerol, 1 mM dithiothreitol, 1 mM EGTA, 1 mM EDTA, and four tablets of complete protease inhibitor mixture (Roche Molecular Biochemicals)). Triton X-100 was added to a final concentration of 1%, and the extract was incubated for 1 h at 4°C. Insoluble matter was removed by centrifugation at 18,000 ϫ g for 15 min. The supernatant fraction was applied to the BSA-Sepharose and incubated overnight. The BSA-resin was removed by centrifugation, and the supernatant fraction was applied to either fresh BSA-Sepharose (10% of the sample) or His 6 -Shank1 189 -399 -Sepharose (90% of the sample). After incubation for 3 h, unbound material was removed, and the Sepharose was washed extensively in buffer A containing 1% Triton X-100 and 0.5% Nonidet P-40, followed by buffer A containing 1% Triton X-100 only. Bound proteins were eluted with buffer A containing 1% Triton X-100 and 2.5 M urea, followed by a second elution step with 4 M urea in the same buffer. 20% trichloroacetic acid (0.66 volume) was added to precipitate proteins. After centrifugation, pellets were washed with ether/ethanol (1:1) and solubilized in 100 mM Tris-HCl, pH 8.8, including 1% SDS. Proteins were resolved by 10% SDS-polyacrylamide gel electrophoresis (PAGE).
After staining with Coomassie Brilliant Blue, bands of interest were excised, cut into small pieces, and rinsed exhaustively in water until the pH of the supernatant was neutral. The gel pieces were lyophilized and subsequently rehydrated with 50 mM Tris-HCl, pH 8.5, containing trypsin (10% of the estimated protein content of the band). After digestion overnight at 37°C, peptides were extracted with 10% trifluoroacetic acid, purified on Sep-Pak C-18 cartridges (Waters, Eschborn, Germany), and lyophilized.
Mass Spectrometry-Samples were dissolved in water; 1 l was applied on a fast evaporation nitrocellulose/␣-cyano-4-hydroxycinnamic acid layer (24) and analyzed by matrix-assisted laser desorption ionization-time of flight mass spectrometry (Bruker Reflex mass spectrometer, reflector mode, pulsed-ion extraction). For protein identification, a data base search using the recorded peptide masses was performed with the program ProFound version 4.10.3 (www.proteometrics.com/). The following search parameters were used: data base NCBInr (2000/07/15), entries only from rat (Rattus norvegicus), full mass and pI range, maximum one missed cleavage, protease trypsin, monoisotopic masses MϩH, and mass tolerance Ϯ0.3 Da (external calibration only).
Yeast Two-hybrid Screening-A yeast-two-hybrid screen was performed using the Y190 yeast strain harboring the reporter genes HIS3 and ␤-galactosidase under the control of upstream GAL1-activating sequence. As a bait, the cDNA coding for the six ankyrin repeats of Shank3 240 -442 were fused to the GAL4 DNA binding domain in vector pAS2-1 (CLONTECH). A rat brain cDNA library cloned into the pACT vector (GAL4 activation domain; CLONTECH) was screened. Putative protein-protein interactions in yeast were tested by the ability to activate both HIS3 and ␤-galactosidase gene transcription. To eliminate false-positive putative interaction partners, the library plasmids were cotransformed with various bait constructs, and afterward candidates were sequenced.
To further determine the interaction domain of the initial ␣-fodrin prey clone, several partial cDNAs coding for one or a combination of different C-terminal ␣-fodrin domains were amplified by polymerase chain reaction and cloned into the pACT2 vector. Protein-protein interactions in yeast were subsequently tested by the ability to activate both HIS3 and ␤-galactosidase gene transcription after retransformation in yeast expressing the Shank3 240 -442 construct as a bait.
Pull-down Assays-Glutathione S-transferase fusion proteins of ␣-fodrin fragments were expressed and purified; proteins were not eluted but left on the glutathione-Sepharose. 35 g of His 6 -Shank1 189 -399 were added in 50 mM Tris-HCl, pH 7.4, and 0.1% Triton X-100 and incubated at 4°C for 2 h. After extensive washing in the same buffer, bound fusion proteins were eluted by boiling in SDS sample buffer, separated on 10% SDS-PAGE, and analyzed by Western blotting using an antiserum directed against the ankyrin repeat region of Shank1.
Coimmunoprecipitation Experiments-Membrane fractions from brain were solubilized in radioimmunoprecipitation assay (RIPA) buffer (0.1% SDS, 0.5% sodium deoxycholate, 1% Nonidet P-40, 50 mM Tris-HCl, pH 8.0, and 150 mM NaCl) at 37°C for 30 min, followed by incubation on ice for 30 min (12). Similarly, transfected HEK cells were solubilized in RIPA buffer (5). Insoluble matter was removed by centrifugation at 20,800 ϫ g for 30 min, and the supernatant fractions were incubated with the anti-␣-fodrin antibody (5 g) for 1 h at 4°C. Immune complexes were collected by incubation (overnight at 4°C) with Protein A/G-Sepharose (Santa Cruz Biotechnology, Santa Cruz, CA). For precipitation with a Shank3 antibody (directed against residues 980 -1786 of rat Shank3), solubilization was performed with 1% deoxycholate, 50 mM Tris-HCl, pH 9.0, and 50 mM EDTA. Extracts were incubated with 10 l of antiserum overnight, followed by protein A/G-Sepharose treatment for 2 h. After washing the Sepharose beads extensively in lysis buffer, the immune complexes were eluted from the beads with Laemmli sample buffer, separated on 7% polyacrylamide gels, and electroblotted onto nitrocellulose. The filters were probed with antibodies directed against the Shank1 PDZ domain (5, 10), the C-terminal 800 amino acids of Shank3 or ␣-fodrin. Bands were visualized by horseradish peroxidase-conjugated secondary antibodies and ECL (Amersham Pharmacia Biotech) (in the case of HEK cells) or by alkaline phosphatase-conjugated secondary antibodies (in the case of brain membranes).
Immunohistochemistry of Hippocampal Neurons and Rat Brain Sections-Hippocampal neuronal cultures from 18-day-old embryonic rats were prepared and grown on coverslips as described by Goslin and Banker (25), washed in phosphate-buffered saline (PBS), fixed in 4% paraformaldehyde at room temperature for 15 min, and blocked with 10% horse serum in PBS (0.01% Triton X-100) for 30 min. For double immunofluorescence, the primary cultures were incubated overnight at 4°C with rabbit anti-Shank3 (1:400) and mouse anti-␣-fodrin antibodies (Chemicon International Inc., Temecula, CA; 1:50) in 10% horse serum in PBS (0.91% Triton X-100). After three washes in PBS, coverslips were incubated for 3 h with goat anti-mouse Cy2 and goat antirabbit Cy3 (Jackson ImmunoResearch, West Grove, PA). Secondary FIG. 1. Yeast two-hybrid assay with Shank3 240-442 . A yeast twohybrid (YTH) assay with the ankyrin repeat region of Shank3 (Shank3 240 -442 ) as a bait yielded a partial cDNA clone of ␣-fodrin as an interacting partner (Orig-Prey); mapping of the interaction site by cotransformation of different prey constructs with the Shank3 bait revealed that spectrin repeat 21 (Fod1) is sufficient to induce bait-prey interaction. Interestingly, spectrin repeats 21 and 22 (Fod2) show only weak interaction that is probably due to misfolding of the prey protein.
Constructs that do not contain spectrin repeat 21 (Fod4 -Fod9) were negative in the YTH assay. Spec, spectrin repeat (gray ovals); EF, common motif of a superfamily of calcium-binding proteins (dark rectangles).

FIG. 2.
Interaction of ␣-fodrin with His-tagged Ank 6 -Shank1 189 -399 . A, affinity chromatography of Shank1-binding proteins using His-tagged Ank 6 -Shank1 189 -399 coupled to N-hydroxy-succinimidyl-Sepharose. Aliquots of rat brain extracts were purified as described under "Materials and Methods." SDS-PAGE (Coomassie Brilliant Blue staining of protein bands): lane 1, His-tagged Ank 6 -Shank1 189 -399 protein with the expected molecular mass of 25 kDa eluted from the matrix under denaturing conditions in the absence of rat brain proteins; lane 2, aliquot (40 l) of rat brain lysate (200 ml) preadsorbed to BSA-Sepharose; lane 3, 4 l of the affinity-purified fraction eluted from Ank 6 -Shank1 189 -399 -Sepharose beads; lane 4, control, 20 l of the rat brain lysate as used for the experiment in lane 3 but bound to and eluted from BSA-Sepharose beads. The positions of the major bands excised for mass spectroscopy analysis are indicated by filled arrowheads. Western blot: lane 5, 1 l of the affinity-purified material was used in the blotting experiment; the membrane was treated with a monoclonal anti-␣-fodrin antibody followed by alkaline phosphate detection. Open arrowheads, positions of ␣-fodrin (at 260 kDa) and some of its known breakdown products at 150 and 120 kDa. B, Pull-down assay with recombinant fusion proteins. His-tagged Ank 6 -Shank1 189 -399 protein was incubated with glutathione-Sepharose loaded with glutathione Stransferase (GST) fusion proteins containing the C terminus of the ␣-fodrin molecule (spectrin repeats 21 and 22 and the EF-hands) or the EF-hands alone. After washing, bound proteins were detected by Western blotting using an antiserum directed against the ankyrin repeats of Shank1 (upper panel). Aliquots of the glutathione-Sepharose beads were analyzed by SDS-PAGE, followed by Coomassie Brilliant Blue staining, to verify that equal amounts of protein were loaded on the beads (lower panel).
For immunohistochemistry (26), rat brains were perfusion-fixed with Bouin's fluid and embedded in Paraplast (Sigma). Subsequently, 15-m frontal and sagittal brain section were deparaffinized in xylene, rehydrated through a graded ethanol series, and equilibrated in 0.1 M Tris-HCl buffer, pH 7.6, for 10 min. After preincubation with 5% normal swine serum in 0.1 M Tris-HCl, pH 7.6, and 0.2% Triton X-100 for 30 min, the primary antibodies (Shank3, 1:1000; and ␣-fodrin, 1:100) were applied in preincubation buffer for 22 h at room temperature. Subsequently, sections were incubated with swine anti-rabbit (mouse) IgG (Dako, Hamburg, Germany) diluted 1:50 for 30 min and then with rabbit (mouse) peroxidase-antiperoxidase complex (Dako) diluted 1:100 for 30 min. Antibody binding was visualized by the application of 3,3-diaminobenzidine (0.05%) and H 2 O 2 (0.001%; Sigma) for 6 min. After completion of the staining procedure, sections were dehydrated and mounted in DePeX (Serva, Heidelberg, Germany). brid experiment, the ankyrin repeats of Shank3 240 -442 were used as bait to screen for potentially interacting proteins in a rat brain cDNA library. Of 1 ϫ 10 6 clones, four positive candidates were obtained: three clones coded for rat amyloid precursor-like protein 1 and one clone for the C-terminal part of ␣-fodrin (amino acids 2019 -2472; Fig. 1). The sequence starts in the middle of the 20th spectrin repeat, followed by spectrin repeats 21 and 22 and two EF-hand regions known to be common motifs of a superfamily of calcium-binding proteins; thus the clone lacks most of the spectrin repeats as well as the SH3 domain (20).
To further map the region of ␣-fodrin interacting with the ankyrin repeats of Shank3, several ␣-fodrin prey constructs were designed, coding for one or a combination of the abovementioned domains (named Fod1-Fod9; see Fig. 1). The yeasttwo-hybrid assay revealed that spectrin repeat 21 alone (Fod1) displayed a strong interaction with the ankyrin repeats of Shank3, whereas the construct containing spectrin repeats 21 and 22 (Fod2) showed only a weak interaction, suggesting that the latter construct may not fold properly in the yeast. Alternatively, intramolecular interactions between different repeats may prevent accessibility of spectrin repeat 21 to the ankyrin repeats of Shank3. A strong interaction was also observed with Fod3 coding for spectrin 21 and 22 plus the first EF-hand motif ( Fig. 1), whereas all other prey constructs lacked any interaction with the ankyrin repeats.
In a parallel affinity chromatography experiment, a His 6tagged recombinant Shank1 protein containing the six N-terminally located ankyrin repeats (His 6 -Shank1 189 -399 ) was used as an affinity matrix to search for binding partners of the ankyrin domains of Shank1 in the rat brain ( Fig. 2A, lane 1). After eluting the bound proteins from the His 6 -Shank1 189 -399 -Sepharose beads, three major bands were visible on SDS-PAGE ( Fig. 2A, lane 3). Two bands at 45 and 260 kDa were clearly enriched when compared with the supernatant fraction after preadsorption with BSA-Sepharose (lane 2); a third band at 50 -55 kDa was also very abundant in the affinity-purified preparation but had about the same intensity as in the initial supernatant fraction (lane 2). These bands were excised from the gel, digested with trypsin, and identified by matrix-assisted laser desorption ionization-time of flight mass spectrometry using the mass fingerprinting technique as ␤-actin (45-kDa band), tubulin (50 -55-kDa band), and ␣-fodrin (260-kDa band). A Western blot of the affinity-purified eluate incubated with an anti-␣-fodrin antibody confirmed the presence of ␣-fodrin at 260 kDa and some of its major proteolytic breakdown products at 120 and 150 kDa (lane 5).
To determine whether the binding between Shank1 and ␣-fo- drin was direct or indirect via bound tubulin or actin, pulldown experiments were performed using recombinant fusion proteins of the interacting domains. When a glutathione Stransferase fusion protein containing spectrin repeats 21 and 22 including the EF-hand was coupled to glutathione-Sepharose beads, the His 6 -Shank1 189 -399 protein bound efficiently to these beads (Fig. 2B), clearly demonstrating the direct interaction between the ankyrin repeats of Shank1 with ␣-fodrin. No binding was observed when only the EF-hands were used, which is in agreement with the mapping analysis in the yeast two-hybrid system with Shank3 (Fig. 1). In addition, a specific interaction of the ␣-fodrin (amino acids 2019 -2472) cDNA clone with the Shank1 189 -399 cDNA construct was also observed in the yeast two-hybrid system (data not shown), indicating that both Shank1 and -3 are equivalent in their interactions with ␣-fodrin.
Apparently the domains involved in the formation of the Shank1-or 3-␣-fodrin complexes are different from those involved in the well characterized spectrin-ankyrin complex observed, for example, in red blood cells. In this case, the 15th spectrin repeat in the ␤-subunit of spectrin (27) interacts with a spectrin binding domain (which does not contain ankyrin repeats) of the ankyrin molecule (28).
Colocalization of Shank3 or Shank1 with ␣-Fodrin in Transfected Cells-To verify the interaction between Shank3 and ␣-fodrin in a mammalian system, COS7 cells were transfected either alone or in combination with cDNAs encoding the Flagtagged ␣-fodrin 2011-2472 and EGFP-tagged Shank3. Two Shank3 constructs were analyzed, one carrying the ankyrin repeats (Shank3 240 -442 ) and the other carrying the SH3 domain (Shank3 536 -610 ). As shown in Fig. 3, in cells transfected only with Shank3 240 -442 carrying the ankyrin repeats, the tagged protein was specifically recruited to distinct structures within the cell nuclei (Fig. 3A). This is presumably due to a cryptic nuclear localization signal in the construct (Lys-Arg-Arg-Arg, amino acids 429 -432), because full-length Shank3 was not observed to appear in nuclei (data not shown). In contrast, when cells were transfected only with the Shank3 536 -610 construct that encodes the SH3 domain (Fig. 3B) or with ␣-fodrin 2011-2472 , the tagged proteins were diffusely distributed in the cell cytoplasm (Fig. 3C). Although it is unclear at present whether nuclear localization of the Shank3-ankyrin construct is of physiological relevance, it provided us with a unique opportunity to assay for strong protein interactions by looking for changes in the cellular localization of interacting proteins. Thus, when cotransfected in COS7 cells with the Shank3 construct carrying the ankyrin repeats, ␣-fodrin 2011-2472 undergoes a dramatic redistribution into the same small punctae of the cell nuclei as seen for the tagged Shank3 240 -442 alone (Fig.  3, D-F). Similarly, a construct containing only spectrin repeat 21 (␣-fodrin 2091-2198 ) is targeted to the nucleus by interaction with the Shank3 ankyrin repeats (Fig. 3, G-I) underscoring the strong interaction of Shank3 and ␣-fodrin in mammalian cells. When full-length ␣-fodrin was coexpressed with the Shank3 ankyrin construct, nuclear redistribution was also observed, but only in ϳ60% of analyzed cells; an example is shown in Fig.  3, J-L. The lower efficiency of this redistribution with the full-length construct may be due to the numerous other interactions of the ␣-fodrin molecule, which may require localization in the cytoplasm rather than in the nucleus. This is in agreement with colocalization experiments of full-length ␣-fodrin and human Shank1 proteins on transfection in HEK cells; here both proteins were mainly colocalized in the cytoplasm (Fig.  4A). The interaction between both proteins in these cells was verified by coprecipitating both proteins using an antibody against ␣-fodrin (Fig. 4B). Taken together with the mapping analysis performed in the yeast two-hybrid system and the protein purification experiments by the affinity matrix, the in vivo colocalization data demonstrate that the ankyrin repeats of the Shank protein family interact specifically and strongly with spectrin repeat 21 of ␣-fodrin.
In Vivo Coimmunoprecipitation of ␣-Fodrin with Shank1 or Shank3-To determine whether the interaction between the ankyrin repeats of Shank1 and ␣-fodrin can be observed in vivo, coimmunoprecipitation experiments were performed from a brain membrane fraction (Fig. 5). In these experiments we used the anti-Shank1 PDZ domain antibody that recognizes both the 260-kDa (containing the ankyrin repeats; lanes 1-4, asterisks) and the 180-kDa (lacking the ankyrin repeats; lanes  1 and 3, arrowheads) forms of Shank1. The assignment that the 180-kDa protein lacks the ankyrin region was supported by a parallel experiment using an antiserum directed against the ankyrin repeats (Fig. 5, lane 2, and Ref. 5). Both forms, the 260and 180-kDa proteins, could be partially solubilized in RIPA buffer as shown in Fig. 5, lane 3; precipitation from this lysate using an anti-␣-fodrin antibody revealed an immunoprecipitate that when analyzed by Western blotting with the PDZ domainspecific antibody depicted only the larger 260-kDa form of Shank1 (lane 4). The 180-kDa protein was missing in the precipitate, consistent with the notion that ␣-fodrin specifically interacts with the ankyrin repeats of Shank1 (or Shank3).  1 and 3), only the larger band (asterisks) contains the ankyrin repeats and thus coimmunoprecipitates with the anti-␣-fodrin antibody (lane 4), whereas the smaller band (arrowheads) does not. Brain membranes were similarly solubilized in deoxycholate buffer (input/ Shank3-IP, lanes 5 and 10) and precipitated with a nonrelated control antibody (lanes 6 and 11) or an anti-Shank3 antibody (lanes 7 and 12). Detection on Western blots was then performed with the anti-Shank3antibody (lanes 5-7) or the anti-␣-fodrin antibody (lanes 10 -12). Shank3-specific bands are labeled by triangles in lanes 5 and 7. In lanes 3 and 8, 50 l of a total volume of 350 l lysate were applied; in lanes 5 and 10, 50 l of a total volume of 4 ml of lysate were loaded. The remainder was in each case used for the IP experiments.
␣-Fodrin itself could be abundantly detected in the lysate as well as in the precipitate (lanes 8 and 9). When precipitating Shank3 with an antiserum directed against the C-terminal half of the molecule, three bands specific for Shank3 could be detected in the lysate (lane 5) and strongly enriched in the precipitate (lane 7). ␣-Fodrin was detected in the lysate (lane 10) as well as in the Shank3 precipitate (lane 12), whereas neither Shank3 nor ␣-fodrin was precipitated with a control antiserum (lanes 6 and 11). Thus our data indicate that ankyrin-containing forms of Shank1 and Shank3 are associated with ␣-fodrin in the rodent brain in vivo. It should be noted that with both the anti-␣-fodrin and the anti-Shank3 antibody, it was not possible to quantitatively precipitate ␣-fodrin, simply because this cytoskeletal protein is very abundant in brain and is present not only in postsynaptic areas but also in axons, dendrites, and cell bodies (see below). Shank3 and Shank1 proteins instead appear to be largely confined to the postsynaptic regions of glutamatergic synapses, and only there may they be complexed with ␣-fodrin molecules, explaining the weak bands in our Western blot analysis.
Localization of ␣-Fodrin and Shank3 by in Situ Hybridization and Immunohistochemistry-To compare the expression pattern of Shank3 and ␣-fodrin, we performed in situ hybridization and immunohistochemical experiments of cultured neurons and of rat brain sections. The double immunostaining of 3-week-old hippocampal culture neurons with the Shank3 and ␣-fodrin antibodies shows the colocalization of both proteins at a subcellular level (Fig. 6). Although Shank3 is found solely in the postsynaptic densities of these neuronal cells, ␣-fodrin is more widely distributed in dendrites and PSDs but is clearly enriched toward the postsynapse. At a higher magnification (insets), the arrows point to overlapping signals at PSDs after staining with Shank3 and ␣-fodrin antibodies.
To investigate the expression pattern of Shank3 and ␣-fodrin in rat brain, we used in situ hybridization (Fig. 7A) and immunohistochemistry (Fig. 7B). Shank3 mRNA as well as ␣-fodrin mRNA are widely expressed throughout the rat brain from day 1 postnatally onward. The expression patterns of both transcripts match at all stages of development. However, although ␣-fodrin is steadily expressed at a high level during postnatal development, Shank3 expression slightly increases toward days 8 -16 and is significantly reduced at adult stages. In contrast to ␣-fodrin transcripts, which are almost evenly distributed throughout the brain, Shank3 mRNA shows higher expression levels in the outer cortical layers (layers 1 and 2) and the cerebellum compared with other brain regions (Fig.  7A). A similar widespread transcript distribution in rat brain has been observed previously using probes specific for Shank1 (5).
␣-Fodrin and Shank3 proteins display a virtually identical distribution pattern throughout the rat brain. Although the cytoplasm of most neurons is devoid of staining, antibodies against both proteins detect their antigen in neurites and as small punctae in the neuropil. ␣-Fodrin as well as Shank3 is most intensively expressed in the hippocampal formation as well as in the cerebellum (Fig. 7B). In the hippocampal cornu ammonis 1 region, cell bodies in the stratum pyramidale are not stained, but an identical pattern of immunoreactivity is found along the neurites in the stratum radiatum and in the stratum oriens. In the cerebellum, the cytoplasm of the Purkinje cells is nearly devoid of staining, whereas in the molecular layer, dense labeling of the neuropil is observed. In the granular cell layer, the glomeruli, i.e. large synaptic structures connecting mossy fibers from deep cerebellar nuclei with the cerebellar granule cells, are intensely stained (arrows). The cytoplasm of the granule cells is not labeled (Fig. 7B). In conclusion, the widespread colocalization of Shank3 and ␣-fodrin supports a role for the interaction of these two proteins in vivo.

DISCUSSION
In this study we have shown that fodrin interacts with two members of the Shank protein family connecting these scaffold proteins to the brain spectrin family of cytoskeletal proteins. Fodrin consists of ␣and ␤-subunits, which form antiparallel dimers. Only ␣-fodrin has been found to interact with the N-terminal ankyrin repeats of Shank1 and -3, as has been documented by various means, including affinity chromatography, yeast two-hybrid screening, colocalization analysis, and coimmunoprecipitation experiments from transfected cells and brain extracts. The initial yeast two-hybrid screen indicated that within ␣-fodrin the interaction domain is located near the C terminus and includes spectrin repeats 21 and 22 and the two Ca 2ϩ binding EF-hand motifs. Further mapping experiments revealed that spectrin repeat 21 alone represents the minimal domain structure essential for the protein-protein interaction with the ankyrin repeats of Shank1 and -3. This is particularly remarkable because the amino acid sequence of the spectrin repeat 21 region is not only highly conserved within the ␣-fodrin molecule itself but also among species (100% identity to human, mouse, and chicken, 81% identity to Drosophila melanogaster, and 79% identity to Caenorhabditis elegans); in contrast, the sequence of the remaining spectrin repeats within the ␣-fodrin molecule are rather diverse (ϳ20 -25% identity). On the other hand, ankyrin repeats exhibit FIG. 6. Immunostaining of primary cultured hippocampal neurons. Staining of 21-day-old primary hippocampal cultures with Shank3 (A) and ␣-fodrin (B) antibodies revealed the punctate staining pattern of Shank3-labeled postsynaptic densities at hippocampal dendrites, in agreement with previous observations (11). ␣-Fodrin is also localized in hippocampal dendrites and clearly colocalizes with Shank3 at hippocampal synapses (C, arrows). Scale bars: main panels, 25 m; insets, 10 m. secondary structures in the form of pairs of antiparallel ␣-helices that are connected by a series of ␤-hairpin motifs; yet they do not prefer particular motifs, nor do they recognize consensus sequences of the target molecules (17). Thus the finding that the domain structure of ankyrin repeats of Shank1 and -3 recognizes only, preferentially and specifically, spectrin repeat FIG. 7. Localization of Shank3 and ␣-fodrin in rat brain. A, in situ hybridization of a horizontal section from brain with 35 S-labeled Shank3 and ␣-fodrin antisense oligonucleotides shows the overall expression of both transcripts at all developmental stages investigated. Intense labeling is observed in cerebral cortex, cerebellum (granule cells), and hippocampus. In contrast to ␣-fodrin mRNA, Shank3 expression is higher in the outer cerebral layers (layers 1 and 2) and the overall expression diminishes during postnatal development (d, days; w, weeks; m, months). B, labeling of rat cerebellum and hippocampus with Shank3-and ␣-fodrin-specific antibodies illustrates the identical distribution patterns of both proteins. Neuropilic punctate staining is observed especially in the stratum radiatum (StR) and stratum oriens (StO) of the hippocampus. In the cerebellum the molecular cell layer (MCL) and glomeruli in the granular cell layer (GCL, arrows) exhibit strong immunoreactivity. The stratum pyramidale (StP) of the hippocampus and the cytoplasm of cerebellar Purkinje cells (PCL) are free of staining. Scale bar, 40 m. 21 may underline the functional importance of the conserved sequence motif of the cytoskeletal protein ␣-fodrin.
The structural interaction of Shank3 with ␣-fodrin is complemented by in situ hybridization and immunocytochemistry experiments demonstrating the coexpression and colocalization of Shank3 with ␣-fodrin in rat brain. Shank1 displays a similar widespread distribution (3,5). In situ hybridization at different developmental stages documents that the mRNAs share overlapping expression patterns in all brain areas at all time points investigated. At the protein level, the virtually identical spatial expression of Shank3 and ␣-fodrin is especially obvious in the hippocampus and cerebellum. Colocalization of both proteins at a subcellular level has been documented by double fluorescence immunohistochemistry of hippocampal neurons. ␣-Fodrin is expressed in the neuronal dendritic compartment but shows strong enrichment toward dendritic spines and postsynaptic densities clearly overlapping with the PSD proteins Shank1 and -3. In fact, fodrin has been shown to be a major constituent of postsynaptic density (19), in which it interacts with actin (␤-subunit) as well as calmodulin (␣-subunit; Refs. 19,29).
To date, however, it has been unknown how ␣-fodrin is connected to membrane-bound receptors at the postsynaptic density. The finding that ␣-fodrin interacts with the multidomain proteins Shank1 and -3 now provides a physical link between the receptive apparatus such as glutamate receptors, i.e. N-methyl-D-aspartate, metabotrobic, and possibly also ␣amino-3-hydroxy-5-methylisoxazole-4-propionate-type receptors (14), and a cytoskeletal protein.
An interesting feature of ␣-fodrin is its selective calmodulindependent processing in response to the elevation of calcium levels (30), synaptic activity, or both; in fact, the occurrence of proteolytic cleavage products of ␣-fodrin has been used as a marker for recent synaptic activity (31). Elevated calcium activates the endogenous calpain protease, which subsequently degrades ␣-fodrin and thus induces an increase in the glutamate binding sites, an effect that can be prevented by the application of calpain inhibitors or by anti-␣-fodrin antibodies (32). Because calpain releases the C-terminal (Shank1-and Shank3-ankyrin binding) part of ␣-fodrin, calcium (or activity)dependent proteolysis of ␣-fodrin would lead to a a major structural rearrangement of the synapse in which the N-methyl-Daspartate receptor-metabotropic glutamate receptor complex remains attached to the Shank proteins but is released from the fodrin-containing part of the cytoskeleton, accompanied by dissociation of fodrin and actin (30). In cooperation with other activity-dependent modifications such as protein phosphorylation and local translation of mRNAs, proteolysis may thus represent another mechanism to contribute to the morpholog-ical alterations observed during synaptic plasticity, allowing a rapid remodeling of synapses after stimulation.