Caldendrin, a Novel Neuronal Calcium-binding Protein Confined to the Somato-dendritic Compartment*

Using antibodies against synaptic protein preparations, we cloned the cDNA of a new Ca2+-binding protein. Its C-terminal portion displays significant similarity with calmodulin and contains two EF-hand motifs. The corresponding mRNA is highly expressed in rat brain, primarily in cerebral cortex, hippocampus, and cerebellum; its expression appears to be restricted to neurons. Transcript levels increase during postnatal development. A recombinant C-terminal protein fragment binds Ca2+ as indicated by a Ca2+-induced mobility shift in SDS-polyacrylamide gel electrophoresis. Antisera generated against the bacterial fusion protein recognize a brain-specific protein doublet with apparent molecular masses of 33 and 36 kDa. These data are confirmed by in vitro translation, which generates a single 36-kDa polypeptide, and by the heterologous expression in 293 cells, which yields a 33/36-kDa doublet comparable to that found in brain. On two-dimensional gels, the 33-kDa band separates into a chain of spots plausibly due to differential phosphorylation. This view is supported by in situ phosphorylation studies in hippocampal slices. Most of the immunoreactivity is detectable in cytoskeletal preparations with a further enrichment in the synapse-associated cytomatrix. These biochemical data, together with the ultra-structural localization in dendrites and the postsynaptic density, strongly suggest an association with the somato-dendritic cytoskeleton. Therefore, this novel Ca2+-binding protein was named caldendrin.

Neurons are structurally and functionally highly polarized cells consisting of an axonal sending and a somato-dendritc receptive compartment. Communication between neurons occurs at synapses, which are asymmetric cell-cell contact sites that typically consist of specialized membrane structures and the underlying cytoskeleton. These cellular specializations derive from the axon of the presynaptic neuron and a dendrite of the postsynaptic neuron. Normal synaptic transmission as well as synaptic plasticity, i.e. use-dependent modulation of synaptic strength, critically depend on intracellular signaling processes on either side of the synapse. Although many of these signaling pathways are mediated by few common messengers, including Ca 2ϩ or cAMP, local specificity is achieved by clustering protein components involved in these pathways at distinct subcellular sites. One such site is the postsynaptic density (PSD), 1 an electron-dense proteinaceous structure of the postsynaptic cytoskeleton at excitatory synapses, which is thought to cluster together neurotransmitter receptors with components of signaling pathways (1)(2)(3)(4).
Signaling by a great variety of external stimuli, including neurotransmitters growth factors and hormones converges at the level of the Ca 2ϩ ion and is further mediated by diverse intracellular Ca 2ϩ -binding proteins (CaBPs) that either can modulate enzymes or structural proteins of the cytoskeleton or are enzymes or cytoskeletal proteins themselves (5). Some of these CaBPs, like calmodulin, are ubiquitously expressed in all cells and mediate a plethora of intracellular responses to Ca 2ϩ signals. Others, e.g. members of the intracellular neuronal calcium sensor (NCS) family (6,7) are restricted to the central nervous system and even to specific neural cell types.
Several neuronal CaBPs have been localized to the postsynaptic compartment. These include ␣-actinin-2 (8) and fodrin (9) as structural components, as well as the regulatory molecule calmodulin (10), the protein phosphatase calcineurin (11), and the NCS protein VILIP (12). Recent studies revealed an important role for ␣-actinin and Ca 2ϩ /calmodulin in the Ca 2ϩ -dependent attachment of NMDA receptors to the synaptic cytomatrix (8,13,14). Moreover, calmodulin in its Ca 2ϩ -bound state is able to bind and activate calmodulin kinase II, a multifunctional kinase highly enriched in the PSD, which is engaged in several processes of neuronal plasticity including the induction of long term potentiation (15,16). Calcineurin has been shown to dephosphorylate the NMDA receptor in a Ca 2ϩdependent manner, thereby shortening the channel openings (17). Calcineurin overexpression affects the transition of short term to long term memory and reveals a novel intermediate phase of long term potentiation (18,19). VILIP appears to be involved in the cross-talk of cyclic nucleotide-and Ca 2ϩ -regulated signaling pathways (20).
Using a combined biochemical and molecular approach to identify new synapse-associated proteins (21,22), we cloned a cDNA encoding a novel neuron-specific polypeptide that was named caldendrin because it binds Ca 2ϩ and is primarily localized in dendrites and neuronal somata. The protein was formerly called calp (calmodulin-like protein), indicating its similarity to calmodulin (22). The present study describes the cloning of the full-length caldendrin cDNA, the tissue distribution of the corresponding transcript, and the biochemical characterization and subcellular localization of the encoded protein.
In Vitro Transcription/Translation-A coupled cell-free transcription/translation system (TNT T3, Promega) was used to generate the primary caldendrin translation product in vitro. Translation products were analyzed by SDS-PAGE and Western blotting.
Isolation of Subcellular Protein Fractions and Western Blotting-Tissue from adult rats (total brain, neocortex, heart, or liver) was homogenized in 20 mM Tris buffer, pH 7.4, containing either 2 mM CaCl 2 or 1 mM EDTA and protease inhibitor mixture (Boehringer, Mannheim, Germany). Soluble proteins were obtained as the supernatant after 100.000 ϫ g centrifugation. After detergent extraction of the remaining pellet with 1% Triton X-100, the detergent-insoluble pellet was extracted with 1% SDS to obtain a fraction of cytoskeletal proteins. Subcellular membrane fractions were prepared essentially as described in Ref. 25 with some modifications from Ref. 26. For isolation of the synaptic junction proteins (PSD fraction), the synaptosomal fraction of the first gradient was diluted with 320 mM sucrose (60 ml/10 g of wet tissue) and an equal volume of 1% Triton X-100, 320 mM sucrose, and 12 mM Tris-HCl, pH 8.1. The suspension was kept on ice for 15 min and centrifuged for 30 min at 32,800 ϫ g. The pellet was resuspended in 320 mM sucrose, 1 mM NaHCO 3 (2.5 ml/10 g pf wet tissue) and fractionated in a second sucrose step gradient as described (25). Synaptic junction proteins (PSD fraction) were diluted with 320 mM sucrose, 1 mM NaHCO 3 and an equal volume of 1% Triton X-100, 150 mM KCl solution, and pelleted for 20 min at 201,800 ϫ g. All steps were carried out at 4°C.
Extraction experiments of P2 pellets with various agents were performed as follows. P2 pellets were resuspended in homogenization buffer, aliquoted into six samples (200 mg of protein each) and centrifuged at 15,000 ϫ g for 20 min. Each pellet was then resuspended in 0.5 ml of one of the extraction buffers, incubated for 15 min at 4°C with gentle shaking, and centrifuged again for 15 min at 100,000 ϫ g. The resulting pellets were washed in homogenization buffer and dissolved in 80 l of gel loading buffer (27). The supernatants were precipitated with trichloroacetic acid, and the resulting pellets were dissolved in 80 l of loading buffer. Separation of proteins by SDS-polyacrylamide gel electrophoresis on 5-20% gels under fully reducing conditions and transfer onto nitrocellulose were performed as described previously (21). For two-dimensional separation, the small gel two-dimensional technique described in Ref. 28 was applied using the Hoefer SE 250 two-dimensional system. Western blots were immunodeveloped by overnight incubation with primary antiserum (29) and immunoreactivity visualized using the ECL detection system (Amersham Buchler, Braunschweig, Germany).
Transient Expression of Caldendrin in HEK293 Cells-A caldendrin cDNA fragment that comprises nucleotides 1-1025 containing the entire open reading frame (ORF, see Fig. 1) was subcloned into the pRC/CMV vector (Invitrogen). Human embryonic kidney 293 (HEK293) cells were transiently transfected using the DAC-30 transfection reagent (Eurogentech) according to the manufacturer's instructions.
Immunocytochemistry-Immunocytochemical staining was carried out using 7-m microtome sections from rat brains, which were fixed by immersion in Bouin's fluid for 48 h, dehydrated, and embedded in paraplast.
Caldendrin was detected with a purified Ig fraction of the rabbit anti caldendrin polyclonal antibody diluted 1:3,000 using the peroxidaseantiperoxidase method. Antibody binding was visualized by incubating sections first with porcine anti-rabbit IgG (DAKO, Hamburg, Germany) and subsequently with rabbit peroxidase-antiperoxidase complex (DAKO). Subsequently, the detection reagent, 3,3-diaminobenzidine (DAB, Sigma, Munich, Germany), was applied. Some sections were counterstained with hematoxylin for morphological orientation. Controls were performed as follows: (i) preabsorption of the antibody with the antigen and (ii) omitting the first or secondary antibody.
Electron microscopy was carried out using vibratome sections (50 m) from rat brains fixed by perfusion with 3% acrolein, 2% paraformaldehyde in 0. First, sections were incubated for 10 min at 60°C in a premade mixed solution (5 ml (A) ϩ 250 l (B) ϩ 500 l (C)) and afterward washed in distilled water (dH 2 O) (3 ϫ 3 min). After these washing steps, sections were incubated in solution D at room temperature for 3 min, washed with dH 2 O (3 ϫ 3 min), incubated in solution E for 3 min, and washed again for 3 ϫ 3 min in dH 2 O. Subsequently, the sections were postfixed in 1% OsO 4 , dehydrated in ethanol, and embedded in epon. Parallel semithin sections were stained with toluidine blue for morphological orientation; ultrathin sections were contrasted with uranyl acetate/lead citrate prior to analysis with a Philips electron microscope.
Preparation of Hippocampal Slices, in Situ Phosphorylation, and Immunoprecipitation-For in situ phosphorylation studies, hippocampal slices were prepared and incubated in 100 Ci of 33 P (final concentration in the incubation medium 50 Ci/ml) as described previously (30). After 90 min of labeling time, slices were rinsed with homogenization buffer A (10 mM Tris/HCl, 0.5 M NaCl, 1% Triton X-100, pH 7.4), homogenized, and spun at 100,000 ϫ g for 1 h. The supernatant was immunoprecipitated with 10 l of polyclonal rabbit antiserum preabsorbed to GammaBind Plus Sepharose (Pharmacia Biotech, Uppsala, Sweden). The remaining pellet was extracted with 1% SDS. For immunoprecipitation of caldendrin, the resulting supernatant was diluted with 10 mM Tris/HCl to give a final SDS concentration of 0.2%. Immunoprecipitation was performed as described previously (23). Non-binding proteins and immunoprecipitates were separated by SDS-PAGE and transferred onto nitrocellulose (0.45 m pore size). Western blots were dried, and 33 P incorporation was visualized using a Fujix BAS1000 Bioimager. Detection of the immunoprecipitate on Western blots was performed with a polyclonal mouse caldendrin antiserum.

RESULTS
Identification and Cloning of Caldendrin cDNAs-From a collection of cDNA clones isolated by expression screening with antisera against a rat brain synaptic protein preparation (22), one cDNA clone, termed sap8i, encoded a protein fragment that displayed distinct sequence similarity to rat calmodulin. The clone was used to isolate a set of seven overlapping cDNAs ranging in size from 1 to 1.4 kilobases. They harbor an open reading frame of 298 amino acids (Fig. 1A). The translation initiation site was assigned to nucleotides 86 -88 encoding the first in-frame methionine. This region fits well with the consensus pattern for eukaryotic translation initiation ((A/G)CC AUG (G/A); Ref. 31), including a 5Ј GC-rich region with only one mismatch at the Ϫ1 position (G instead of C). In vitro translation and heterologous expression studies confirmed the use of this initiation site (see below). A putative polyadenyl-ation signal is formed by nucleotides 1385-1390 (bold in Fig.  1A). The deduced protein caldendrin has a calculated M r of 33,071 and a theoretical isoelectric point of 7.42. Whereas primary structure analysis revealed no identifiable structural features in the N-terminal half of the protein, the C-terminal part displays a high degree of similarity to the entire reading frame of rat calmodulin. As shown in an alignment of both proteins (Fig. 1B), two of the four Ca 2ϩ -binding EF-hand structures found in calmodulin are conserved in the novel protein.
However, the remaining two miss several essential residues in the consensus pattern. Within the aligned region sequence, identity and similarity between caldendrin and calmodulin are 42% and 57%, respectively. The N-terminal half of caldendrin contains seven putative protein kinase C phosphorylation sites (arrows in Fig. 1A) and an unusually high proline content (13%).
Spatial and Temporal Expression Pattern of the Caldendrin Transcript-Northern blot analysis revealed a single hybridizing band of 1.8 kilobases that is detectable in various brain regions with strongest expression in the cerebral cortex. No hybridization is detected in heart, muscle, liver, and C6 glioma cells ( Fig. 2A). During brain development, hybridization signals were first observed in the second postnatal week, i.e. during period of synapse formation and terminal differentiation of the brain. After the initial period of increase caldendrin, mRNA levels remain essentially unchanged throughout later ontogenetic stages (Fig. 2B).
To examine the transcript distribution at the cellular level, we performed in situ hybridization studies in horizontal brain sections (Fig. 2, C and D). With this more sensitive method the transcript can be clearly localized first at day P10 to distinct brain regions including cerebral cortex, developing cerebellar cortex, and hippocampal formation (Fig. 2C). In adult rats, the hybridization pattern is qualitatively the same, while the labeling intensity has increased (Fig. 2D). Silver grains were mainly found in cortically organized parts of the brain. Inter-estingly, the hippocampal subfields are not uniformly labeled. In contrast to a very prominent hybridization in the CA3 pyramidal cell layer, labeling of the CA1 region and the dentate gyrus is much less intense (Fig. 2E). Thus, caldendrin expression is restricted to certain brain regions and the transcript seems to be differentially expressed, at least in the hippocampus, in different subsets of cells.
In control experiments, the use of a sense oligonucleotide, competition with 100-fold excess of unlabeled oligonucleotide as well as washing steps above the calculated melting temperature of the hybrid yielded no significant labeling. Specificity of the hybridization was further confirmed by use of two independent oligonucleotide probes that produced virtually identical results (data not shown).
Calcium Binding Activity of Recombinant Caldendrin Fragments-A caldendrin cDNA fragment comprising nucleotides 500 -979 (see Fig. 1) was generated via polymerase chain reaction and subcloned into the pQE30 vector to generate a bacterially expressed fusion protein including amino acids 139 -298 and six histidine residues at the N terminus. This fusion protein has a calculated M r of 19.386 kDa. In the presence of 1 mM Ca 2ϩ , a clear mobility shift in the PAGE gel can be observed, indicating the ability of the caldendrin C terminus to bind Ca 2ϩ (Fig. 3). This shift is most plausibly based on conformational changes that occur in caldendrin upon Ca 2ϩ binding as has been shown for other CaBP, e.g. calmodulin (32,33) or VILIP (34).
Characterization of Caldendrin-The fusion protein was used to raise polyclonal caldendrin antisera in mice and rabbits. Three independent antisera specifically detect a protein doublet of 33/36 kDa in brain protein preparations (Fig. 4). This doublet is not recognized by either of the pre-immune sera (data not shown). To get a first clue to the biochemical nature of the two polypeptides, we tested their solubility in low salt buffer and detergent. The 33-kDa protein isoform is partly found in soluble protein fractions (SP in Fig. 4). The Triton X-100 extract of the remaining pellet does not contain significant amounts of either of the immunoreactive bands. The 36-kDa caldendrin polypeptide is almost exclusively found in the aqueous and detergent-insoluble cytoskeletal pellet (CP in Fig.  4). During postnatal brain development, the amounts of caldendrin immunoreactivity increase from the second to the eighth postnatal week both in soluble and cytoskeletal protein fractions. Heart and kidney protein preparations obtained from 8-week-old rats are devoid of caldendrin immunoreactivity (H and K in Fig. 4).
To verify the proposed ORF, a coupled cell-free transcription/ translation system was used to generate the primary calden- drin translation product in vitro. Translation products were analyzed by SDS-PAGE and Western blotting. A single immunoreactive product with an apparent molecular mass of 36 kDa was observed that co-migrates with the upper caldendrin band detected on brain protein blots (Fig. 5A). Control reactions without added caldendrin cDNA do not synthesize this polypeptide (lane 1 in Fig. 5A).
In order to elucidate the nature of the protein doublet found in brain protein compared with the single in vitro translation product, the caldendrin cDNA was transiently expressed in human embryonic kidney cells (HEK293). Whereas untransfected HEK cells do not express any caldendrin immunoreactivity (Fig. 5B, lane 1), transfected cells produce two caldendrin isoforms of 36 and 33 kDa comparable to that detected in brain (Fig. 5B). Furthermore, in both cases, only the 33-kDa isoform appears in the soluble fraction. These findings clearly indicate that both isoforms originate from a single transcript and that the difference in migration behavior should result from posttranslational modification, which occurs both in brain and in the heterologous 293 cell expression system, but not in the in vitro reticulocyte system. Subcellular Localization of Caldendrin-Subcellular frac-tionation by differential centrifugation of brain proteins demonstrated that caldendrin immunoreactivity is associated with nearly all particulate fractions including light and heavy membrane fractions and several synaptic fractions, like synaptosomes, synaptic membranes and the PSD fraction (Fig. 6A). The immunoreactivity in particular of the 36-kDa band is clearly enriched in the PSD fraction after the second washing step with Triton X-100 (T2 in Fig. 6A), indicating a tight association with the synaptic cytomatrix. However, caldendrin appears not to be an exclusively synaptic protein.
To investigate the association of caldendrin with the cytoskeleton, several attempts were made to solubilize the 36-kDa isoform from the crude membrane fraction (14,000 ϫ g pellet P2) as shown in Fig. 6B. A complete extraction was successful with a combination of high salt and detergent (1 M NaCl, 2.5% Chaps), with chaotropic agents (3 M KSCN) and under fully denaturing conditions (8 M urea). Treatment with 1 M Tris, even in combination with 1% Triton X-100 as well as alkaline conditions (pH 11.5), did not lead to a release of substantial amounts of 36-kDa caldendrin. Interestingly, alkaline treatment results in a complete extraction of 33-kDa caldendrin, thereby fully separating both isoforms. These solubilization characteristics of the protein indicate an association, particularly of the larger caldendrin isoform, with the neuronal cytomatrix.  3). Lanes 4 and 5, insoluble protein preparations from rat brain and transfected HEK cells, respectively; lanes 6 and 7, soluble protein preparations from rat brain and transfectants, respectively.
FIG. 6. Distribution of caldendrin in subcellular fractions from rat brain. Panel A shows an immunoblot of subcellular fractions (50 g/lane) obtained by differential centrifugation from brains of 30 days-old rats. Ho, homogenate; S2, 13,000 ϫ g supernatant after removal of cell debris and nuclei; P2, corresponding pellet; My, myelin fraction (floating on top of the gradient); lMb, light membranes (upper interface of the gradient); Syn, synaptosomes (lower interface of the gradient); hMb, heavy membrane fraction (pellet of the gradient); T1, pellet obtained after Triton X-100 extraction of synaptosomes; PSD, postsynaptic density fraction obtained from T1 in a second gradient (25); T2, pellet obtained after Triton X-100 extraction of T1 (26). Apparent molecular masses of the immunoreactive bands are indicated. Panel B shows the extraction of caldendrin from the crude brain membrane pellet P2. S, extractable supernatant; P, remaining pellet after centrifugation. Membranes were extracted with the solutions indicated. Protein contents of S and P lanes add up to 50 g.
This assumption is supported by light and electron microscopic localization of caldendrin immunoreactivity in the brain. A representative light micrograph (Fig. 7A) shows several immunopositive neurons of the cerebral cortex. Within the neurons, caldendrin is found only in the somato-dendritic compartment. No immunostaining is found associated with axonal processes of neurons or in glial cells. Staining is most intense underneath the membranes of the neuronal cell bodies as well as within the dendrites as depicted at higher magnification in Fig. 7B. In the CA3 region of the hippocampal formation, the pyramidal neurons as well as some hilus neurons are intensely stained (Fig. 7C). The picture also reveals immunoreactivity in pyramidal cell primary dendrites. Ultrastructural localization studies in the hippocampal CA3 region confirm the dendritic and postsynaptic localization (Fig. 7, D and E). Interestingly, accumulation of the reaction product is not evenly distributed inside the dendrites but concentrates along intracellular membranes and cytoskeletal structures like tubular bundles as depicted in Fig. 7D. This is in good agreement with the biochemical data. We never observed any immunolabeling of glial cells, axons or axon terminals at the ultrastructural level (AT in Fig.  7E). Consistent with the subcellular fractionation studies caldendrin appears highly enriched at PSDs (Fig. 7E).
Caldendrin as a Putative Substrate for Protein Kinases-The primary structure of caldendrin contains seven putative protein kinase C phosphorylation sites (see Fig. 1). Therefore, we tested whether caldendrin is a possible substrate for protein kinases. Two-dimensional gel electrophoresis of immunoprecipitated caldendrin demonstrates that both isoforms differ in their migration behavior along the pH gradient (Fig. 8). Whereas the 36-kDa isoform appears as a single spot at a pI of 6.8 -6.9 (calculated pI of 7.42), the 33-kDa band separates into a chain of at least three spots within the range of pH 5.9 -6.8. This could be a consequence of phosphorylation, because a phosphate group added to a protein shifts the pI for 0.1 to 0.3 units (35).
To test this hypothesis more directly, in situ phosphorylation experiments were performed in acute hippocampal slices. Phosphorylation was carried out for 90 min. Afterward, slices were homogenized and fractionated and the protein preparations used for caldendrin immunoprecipitation. The autoradiograph in Fig. 9 shows immunoprecipitation of radioactivity with polyclonal rabbit caldendrin antiserum from soluble proteins and clearly demonstrates phosphorylation of the 33-kDa isoform under in situ conditions. Interestingly, the 36-kDa isoform precipitated from a SDS-soluble fraction is unphosphorylated (data not shown).

DISCUSSION
In this study, we describe the cloning and first characterization of rat caldendrin, a novel EF-hand Ca 2ϩ -binding protein selectively localized to the somato-dendritic compartment of neurons and highly enriched at postsynaptic structures. Sequence alignment shows that ϳ150 amino acids at the C terminus of caldendrin exhibit high similarity with the ubiquitous calcium mediator protein calmodulin. Two of the EF-hand domains of caldendrin, i.e. the second and fourth ones, deviate from the consensus pattern defined for functional EF-hands and therefore may have lost their Ca 2ϩ -binding activity.
The predicted M r of 33,071 for caldendrin is in reasonable agreement with the apparent molecular mass of 33/36 kDa obtained from Western blots of rat brain proteins and confirms the assumed translation start site. The in vitro translation product generated from the full-length caldendrin cDNA comigrates with the upper band at 36 kDa. Therefore, we assume that this band represents the primary translation product. The biochemical modification or nature of the second isoform which migrates as a 33-kDa protein band is still unknown. Crossreactivity with a different protein as well as origin from a differentially spliced transcript may be excluded, because HEK293 cells transfected with the caldendrin cDNA express both isoforms. This clearly indicates that the modification of the primary translation product is post-translational and occurs in heterologous expression systems as in neurons.
The modification is accompanied by an increased solubility of the protein. Both the mechanism of the cytoskeletal association of caldendrin and the partial release into the cytosol are currently unknown. A Ca 2ϩ -dependent translocation, as shown for the neuronal EF-hand Ca 2ϩ -sensor protein VILIP (34), can be excluded because the subcellular distribution of caldendrin is not altered by changes in the Ca 2ϩ concentration during sample preparation. 2 The post-translational modification of the primary polypeptide not only leads to an increased electrophoretic mobility in SDS-PAGE, it also modifies migration in pH gradients as observed on two-dimensional gels. The appearance of a chain of spots at 33 kDa may indicate the occurrence of different states of phosphorylation; however, it could also result from an artificial carbamylation during the sample preparation (35). Analysis of the caldendrin primary structure revealed several potential phosphorylation sites in the N-terminal portion, and incorporation of radioactively labeled phosphorus into caldendrin was shown under in situ conditions in hippocampal slice preparations. Interestingly, only the 33-kDa isoform incorporated 33 P. This could point to an important role for phosphorylation in modulating biochemical properties of caldendrin. It is possible that an initial phosphorylation step of the 36-kDa isoform induces conformational alterations, which result in the observed change in migration behavior and solubility of the protein.
The potential of caldendrin to bind Ca 2ϩ was shown in vitro with a bacterially expressed fusion protein. The electrophoretic mobility shift observed with this 18-kDa caldendrin fragment upon Ca 2ϩ binding is not detectable with the full-length brain protein. A possible explanation may be that the conformational changes are compensated in the much larger full-length pro- FIG. 8. Separation of caldendrin-immunoreactive polypeptides in two-dimensional gels. Western blot loaded with caldendrin immunoprecipitated with polyclonal rabbit antiserum from a cytoskeletal protein preparation solubilized in 1% SDS and afterward diluted to a final SDS concentration of 0.2%. Detection was performed with a polyclonal mouse antiserum. Apparent molecular masses and isoelectric points as deduced from marker proteins are indicated. FIG. 9. In situ phosphorylation of caldendrin in hippocampal slices. Caldendrin was immunoprecipitated from a soluble protein preparation obtained from acute hippocampal slices incubated with 33 P. Autoradiograph shows the radioactivity in the supernatant after precipitation (S) and in the immunoprecipitate (IP). Apparent molecular mass of the labeled band is indicated. tein which exhibits a clear bipartite structure. The two parts are predicted to differ significantly in their physico-chemical features. Whereas the N-terminal half is highly basic (pI ϭ 11.9; 13 positively charged amino acid residues), the Ca 2ϩbinding C-terminal half is acidic (calculated pI ϭ 4.5; 13 negatively charged residues). Therefore, intramolecular interactions are likely, which may be subject to regulation by Ca 2ϩ binding and phosphorylation.
On the basis of their postulated function, the EF-hand CaBPs have usually been classified as "trigger" or "buffer" proteins. Whereas trigger proteins like calmodulin (for review, see Ref. 36) or the members of the NCS family change their conformation upon Ca 2ϩ binding and then modulate a variety of channels or enzymes, buffer proteins like parvalbumin (reviewed in Ref. 37) are thought to play a more passive scavenger role to limit the rise in intracellular Ca 2ϩ concentration. The modular organization of caldendrin supports a postulated trigger function. Analysis of the caldendrin N terminus, which appears not to be structurally related to any known protein, may help to understand the functional significance of this bipartite molecular organization.
Caldendrin seems to be preferentially expressed in brain regions with a laminar organization (i.e. cortex, hippocampus, and cerebellum) and in neuronal cells with a broad dendritic tree. So far, we have found no evidence for expression of the protein in neuroglia. Its distribution within the neuron appears to be restricted to the somato-dendritic compartment, a feature that it shares for example with MAP2, a cytoskeletal protein associated with microtubules (38). Both the biochemical solubilization characteristics and the subcellular fractionation studies lead us to conclude that caldendrin is associated with the cortical cytoskeleton. Caldendrin seems to be localized to the subplasmalemmal cortex of cell soma and dendrites, and both biochemical fractionation and electron microscopic investigation clearly demonstrate a synaptic localization of caldendrin. Although caldendrin is not exclusively localized to synapses we found that, in contrast to many other proteins of the dendritic cytoskeleton, immunoreactivity is enriched in the postsynaptic density fraction.
Several other synaptic proteins display similar intramolecular dichotomies and therefore are good candidates to confer Ca 2ϩ -dependent regulation to subcellular structures or for the coupling to different signaling pathways. Examples include the cytoskeletal proteins ␣-spectrin, non-muscle ␣-actinins (39), and myosin. In the latter case, the calmodulinlike domain represents the regulatory light chain (40). The Ca 2ϩ -dependent Cys endopeptidase calpain also contains EFhand domains at the C terminus of each subunit that confers Ca 2ϩ -regulation to the proteolytic action of the holoenzyme (41). Another example is calcineurin, a Ca 2ϩ -calmodulin-dependent protein phosphatase that is composed of a catalytic subunit and a calmodulin-like regulatory subunit (42). Based on gene structure analyses, it has been suggested that these modular molecules, which are composed of one or more Ca 2ϩbinding EF-hand domains combined with other functional domains, have evolved from a single genetically mobile ancestral EF-hand motif (43).
The relatively late onset of transcription of the caldendrin gene during development is shared with several other neuronal CaBPs (44). This suggests that caldendrin expression is not crucial for synaptogenesis. However, in contrast to some other CaBPs, like hippocalcin (45) or calbindin-D 28K (46), no decrease in protein levels in mature neurons was observed. The ontogenetic expression profile, therefore, suggests a functional role for caldendrin in the dendritic tree of terminally differentiated neurons.