The 270 kda splice variant of erythrocyte β-spectrin (βi∑2) segregates in vivo and in vitro to specific domains of cerebellar neurons

Many isoforms of α and β-spectrin are now recognized^* (for reviews see Coleman et al., 1989; Goodman et al., 1988; Winkelmann and Forget, 1993; S. P. Kennedy and J. S. Morrow, unpublished). In humans, these arise from two sets of α and β genes. With the exception of erythrocyte αspectrin (from chromosome 1), additional spectrin diversity is generated by alternative transcription from each of these genes. Thus, two transcripts arise from the β-spectrin gene on chromosome 14 (Winkelmann et al., 1990b); potentially four transcripts arise from the αspectrin gene on chromosome 9 (Leto et al., 1988; Moon and McMahon, 1990); and at least six transcripts could arise from the β spectrin gene on chromosome 2 via alternative splicing near both the 5′ and 3′ termini (Winkelmann and Forget, 1983), (reviewed by S. P. Kennedy and J. S. Morrow, unpuβ lished). Since there appears to be no βIochemical restriction on which αsubunit will associate with a given βsuβ unit (at least in vitro) (reviewed by Coleman et al., 1989), it is likely that a rich diversity of subtly distinct spectrin species exist within some cells. While the precise role of spectrin diversity is unknown, it is presumably required for regional compartmentalization and the generation of orgα nized cytoplasmic and membrane surface domains.

Nervous tissue displays many spectrin isoforms. Two developmentally regulated spectrins have been identified in chicken (Lazarides and Nelson, 1983a,b) and mouse brain (Riederer et al., 1987). The basic unit of both is an ab heterodimer. The most widely distributed form, also known as fodrin (Levine and Willard, 1981), is abundant during fetal development, and is largely confined to axons and presynaptic terminals in mature brain (Riederer et al., 1986; Zagon et al., 1986). A second form of spectrin also exists in brain and muscle tissue, based on cross-reactivity with antibodies to erythroid αβ-spectrin. This spectrin has been termed 240/235E in the mouse and (β′) in chickens, and is most concentrated in neuronal cell bodies, dendrites, postsynaptic terminals and some glial cells, particularly in the cerebellum (Lazarides and Nelson, 1983a; Riederer et al., 1987; Siman et al., 1987). This spectrin appears late during neuronal maturation. It is likely that other isoforms also exist, such as another spectrin with immunological cross-reactivity to erythroid β-spectrin and is apparently distinct from 235E, and is concentrated with the acetylcholine (ACh) receptor at the motor endplate of rat skeletal muscle (Bloch and Morrow, 1989). As informative as these studies have been, their interpretation is complicated by the heretofore unappreciated complexity of spectrin isoform diversity, and by the likelihood that antibodies broadly reactive to both subunits of a spectrin may probe more than one isoform. For example, a unique carboxyl terminus derived from differential processing of 3′ β-spectrin pre-mRNA distinguishes βIS1 spectrin (erythrocyte) from βIΣ2 spectrin (muscle) (Winkelmann et al., 1990b). This larger β-spectrin transcript is also present in brain tissue and might be identical to β-spectrin 235E, although this possiβIlity has never been confirmed. In addition, large portions of the new COOH-terminal region of βIΣ2 spectrin are nearly identical to COOH-terminal regions of the general β-spectrin (βIIΣ1) isoform derived from chromosome 14 (Hu et al., 1992). Such regions of shared sequence are potentially a source of immunocrossreactivity between these isoforms. Thus, to appreciate the distinctive distributions of specific spectrin isoforms, antibodies monospecific for each isoform are required.

In the present study, an antibody specific for a sequence unique to spectrin isoform βIΣ2 was prepared and used to examine this transcript in rat cerebellum and in cultured rat cerebellar granule cells. We also compared the distribution of this spectrin isoform with the staining pattern of commercial antibodies to erythrocyte spectrin (240/235E), the general form of αspectrin (αIIΣ*, fodrin) and with microtubule-associated protein-2 (MAP2) and synapsin I. These latter two proteins are known to be compartmentalized in dendrites and presynaptic terminals, respectively. Spectrin βIΣ2 is found to be compartmentalized in the somαdendritic domain of certain neurons, and is concentrated at the plasma membrane only in specialized structures at the postsynaptic densities. It is absent from the cell body of Purkinje cells and oligodendroglia, and overall its pattern of reactivity is a subset of that reported for 240/235E. Granule cells also display a somato-dendritic staining pattern in culture, indicating that underlying mechanisms of spectrin sorting may be preserved in these neurons in vitro. The extreme concentration of βIΣ2 spectrin at the PSD and in selected cytoplasmic compartments of some but not all neurons suggests that unique isoforms of spectrin may play a role in organizing topographically defined clusters of receptors or cytoplasmic protein complexes, and that their expression is cell-type specific.

Antibodies were prepared in New Zealand white rabβIts by repeated intradermal injection, following previously published protocols (Harris et al., 1985). To prepare sequence-specific antibodies against βIΣ2 spectrin, the synthetic peptide ‘βIΣ2-A’ (Pro-Gly-Gln-His-Lys-Asp-Gly-Gln-Lys-Ser-Thr-Gly-Asp-Glu-Arg-Pro-Thr) with Gly-Gly-Cys added at the carboxyl terminus, was prepared by the Yale Peptide and Protein Chemistry Facility (#BB120-MOR) by standard solid phase peptide synthesis, and was purified by reverse phase HPLC chromatography. Identity and purity were confirmed by amino acid analysis and by gas chro-matography and mass spectroscopy. This peptide was covalently conjugated via the cysteine residue to m-maleimidobenzoyl-N-hydroxylsuccinimide (MBS)-activated keyhole limpet hemo-cyanin (KLH) (Sigma Chemical Co.) by reacting 5 mg of peptide with 5 mg of KLH activated with 2 mg of MBS in the presence of 50 mM NaPO₄, pH 7.5 for 3 hours at 25°C. After gel filtrα tion in G-100 Sephadex in conjugation buffer, the KLH-conjugated peptide was used in 150 mg aliquots for immunization. Antibodies were purified on an affinity column prepared by coupling 2 mg of unconjugated peptide with 4 ml of CNBr-activated Sepharose CL4B (Pharmacia, Sweden). The Sepharose was activated by reacting 100 ml of a 50% slurry of CL4B at pH 11 with 150 mg of CNBr. Peptide was coupled overnight at pH 7.0 in 0.2 M NaPO₄, 0.1 M citric acid, 0.01 M EDTA. The coupling reaction was enhanced with 10 mg of 1-ethyl-3-(dimethyl aminopropylcarbodiimide)-N-hydroxylsulfosuccinimide (SPDP, Pierce Chemical Co.) (Staros et al., 1986). Bound antibodies were eluted with 1 M acetic acid, immediately neutralized with pH 8 Tris buffer, and dialyzed into PBS. Another polyclonal antibody against βIIΣ1 spectrin, prepared against recomβInant peptide representing approximately the carboxyl-third of this spectrin (Kennedy et al., 1991a), was also provided by Dr Scott P. Kennedy (Yale). Affinity-purified antibodies against bovine synapsin I and human and bovine αIIΣ* spectrin (fodrin) have been previously described (Di Stasi et al., 1991; Harris et al., 1985; Petrucci and Morrow, 1987). Mouse monoclonal antibody (AP-20) against MAP2 was from Sigma Chemical Co. Polyclonal anti-red cell spectrin antibody that reacts with 240/235E spectrin (Zagon et al., 1986) was from Chemicon Inc. (Temecula, Ca). βIotinylated secondary antibodies and avidin-peroxidase conjugate were purchased from Dako (Netherlands).

Primary cultures of cerebellar granule cells were prepared from postnatal day-8 rat cerebellum (Di Stasi et al., 1991; Levi et al., 1984). Dissociated cells were plated in 60 mm plastic dishes precoated with 10 mg/ml poly-L-lysine at a density of 3 × 10⁵ cells/cm², and grown in Eagle’s basal medium containing 10% fetal calf serum, 2 mM glutamine, 100 mg/ml gentamicin and 25 mM KCl (final concentrα tion). After 16-20 hours, araβInocyl cytosine (10 mM) was added to the cultures to prevent the replication of non-neuronal cells. Cerebellar cultures containing over 95% granule cells were evaluated by immunostaining or extraction after 8-12 days in vitro.

Proteins were analyzed by SDS-PAGE (Laemmli, 1970) and trans ferred to nitrocellulose for western blotting (TowβIn et al., 1979). Western blots were visualized with ¹²⁵I-labeled staphylococcal protein A (SPA), followed by autoradiography. Alternatively, an enhanced chemiluminescence procedure was utilized (ECL), as described in the manufacturer’s kit (Amersham Corporation, 1993).

Synaptosomes were prepared from adult rat cerebellum by homogenization in 0.32 M sucrose followed by density gradient centrifugation (Carlin et al., 1983). Spectrin complexes were immunoprecipitated under high salt conditions from detergent treated synaptosomes, using a procedure communicated by Dr Velia Fowler (Scripts). Briefly, synaptosomes were soluβIlized in 1% Triton X-100, 0.5 M NaCl, 20 mM Tris-HCl, pH 8.8, 1 mM\ MgCl₂, 2.5 mM EGTA, 2 mM EDTA, 10 mg/ml leupeptin, 0.15 mM PMSF, 10 mg/ml aprotinin, and centrifuged at 100,000 g for 20 minutes at 4°C. Samples of supernatant were incubated overnight at 4°C with 10 ml of antibody and 20 ml of a 50% suspension of protein-G agarose in PBS. Protein-G agarose beads (Sigma) were collected by centrifugation, washed and soluβIlized in electrophoresis sample buffer and analyzed.

Frozen sections (10 mm) of rat cerebellum or granule cells on polylysine-coated glass coverslips were fixed at room temperature (RT) for 20 minutes in 3.7% paraformaldehyde, permeaβIlized for 5 minutes in 0.3% Triton X-100 in PBS, treated for 30 minutes with 3% H₂O₂ in PBS to inhiβIt endogenous peroxidase activity, and blocked with non-immune goat IgG (10% in PBS for 1 hour). The samples were then washed four times with 1% BSA in PBS. Primary antibodies were diluted into PBS with 1% BSA [rabβIt anti-βIΣ2-A (1:30); rabβIt anti-240/235E (1:30); mAb AP20 anti-MAP2 (1:500); rabβIt anti-αIIΣ* (1:50)], and then incubated with sections for 1 hour at RT. After several washes in PBS the cells were incubated for 30 minutes with goat anti-rabβIt or goat antimouse rhodamine-conjugated F(ab)₂ (Cappel) at a 1:300 dilution. Sections were examined using a Nikon fluorescent microscope equipped with epilumination and appropriate optics for observing fluorescein and rhodamine fluorescence.

Anesthetized adult rats (Wistar, Charles River) were perfused through the left ventricle with 4% phosphate-buffered formaldehyde (pH 7.2-7.4) freshly made from paraformaldehyde. The cerebellum was removed and fixed for an additional 2 hours in the same fixative. After exhaustive washing in phosphate buffer (PBS), the tissue was immersed in 20% glycerol in PBS, quick frozen in Freon^® and liquid nitrogen, and cryostat-sectioned at 20 mm. Sections were treated briefly with 3% H₂O₂ in PBS to inhiβIt endogenous peroxidase activity, and blocked with non-immune goat IgG. Primary antibodies were used at the following dilutions: anti-βIΣ2-A (1:3); anti-αIIΣ* (1:20); anti-240/235E (1:20); anti-MAP2 (1:500); anti-synapsin I (1:100). Sections were incubated in the primary antisera overnight at 4°C. After prolonged washes in PBS, sections were incubated in βIotinylated secondary antisera diluted 1:100 for 1 hour at room temperature, washed in PBS and treated for 1 hour with 1:100 avidin-peroxidase conjugate (Vector Laboratories). After several rinses in PBS, peroxidase was revealed by a solution of 0.2 mg/ml diaminobenzidine in 0.001% H₂O₂. Sections were postfixed in 1% phosphate-buffered OsO₄, dehydrated in ethanol and embedded in epoxy resin. Ultrathin sections were observed with an EM 102 Zeiss electron microscope. Alternatively, formaldehyde-fixed tissues were postfixed in 1% OsO₄, dehydrated and embedded in epoxy resin for conventional ultrastructural examination. Ultrathin sections were counterstained with lead citrate and uranyl acetate.

Granule cells grown on polylysine-coated glass coverslips were fixed at RT with 4% phosphate-buffered formaldehyde, 0.12 M sucrose, for 2 hours. After rinsing in PBS, they were then incubated for 2 hours with primary antibody and visualized as above. After osmium postfixation, samples were dehydrated and flatembedded in epoxy resin.

The specificity of the antibody prepared to synthetic peptide ‘A’ was examined by western blotting of homogenates of rat cerebellum and rat erythrocyte ghosts (Fig. 1). Peptide ‘A’ represents a 17-residue sequence near the 5′ end of domain III of βIΣ2 spectrin (Winkelmann et al., 1990a). It was chosen on the basis of predicted antigenicity (Jameson and Wolf, 1988), and because it shares no homology with other known α or β-spectrins. A comparison of this sequence with βIS1 (erythrocyte β-spectrin; Winkelmann et al., 1990a) and βIIΣ1 (βfodrin or general b_G-spectrin) (Hu et al., 1992) is shown in Fig. 1. By western blotting, this antibody demonstrated specificity for a single 270 kDa protein in rat cerebellum and recognized no proteins in ghosts. This protein band was distinct from the position of MAP2, as marked by western blotting with mAb AP20 (Fig. 1), thereby ruling out cross-reactivity between this spectrin antibody and mAb AP20 (Davis and Bennett, 1982).

The overall distribution of βIΣ2 spectrin and αIIΣ* spectrin was compared in rat cerebellum by immunofluorescence microscopy (Fig. 2). The αspectrin antibodies used here were prepared to purified ‘brain spectrin’, and react primarily with the αsubunit of this protein. While these antibodies do not recognize aIS1 spectrin (red cell), they presumably cannot distinguish different alternative transcripts of αII spectrin. Hence, the designation αIIΣ* is used to reflect their general αII spectrin reactivity. Spectrin αIIΣ* was found in all layers of the cerebellum, although it was notably reduced or absent in the soma of granule cells (but not Purkinje cells). In cultured granule cells, αIIΣ* spectrin was most abundant in neurites (Di Stasi et al., 1991). In contrast, the distribution of βIΣ2 spectrin was highly concentrated in the soma of cells in the granular layer, and in scattered stellate cells within the molecular layer. Notably, the soma of Purkinje cells did not contain appreciable βIΣ2 spectrin. The distribution of βIΣ2 spectrin was also compared to MAP2 in cultured granule cells. MAP2 stains the soma of these cells in a pattern coincident to that of βIΣ2 spectrin, but unlike this spectrin it is also abundant along neurites.

To identify the nature of the αsubunit associated with βIΣ2 spectrin, and to determine whether spectrin complexes containing more than one type of β-spectrin were formed, immunoprecipitation experiments were performed with antibodies to either βIΣ2 spectrin or to βIIΣ1 spectrin (Fig. 3). Blotting of these precipitates with antibodies to βIIΣ1, βIΣ2 or αIIΣ* spectrin indicated that while αIIΣ* spectrin associated with both β-spectrins, presumably forming heterodimers and tetramers, there were no detectable complexes containing both βIΣ2 and βIIΣ1 spectrin.

The subcellular distribution of βIΣ2 spectrin in the cerebellum was examined by immunoelectron microscopy (Figs 4 to 8). In the granular layer, the granule cell cytoplasm (Fig. 4A) and the granule cell dendritic processes (Fig. 5A) were diffusely positive. All nuclei were negative, and only rare and irregular deposits of stain were seen abutting the nuclear membrane (Fig. 4B). Within the cell cytoplasm, stain accumulated in a punctate and linear pattern, suggesting an association of βIΣ2 spectrin with ER cisternae, microtubules and possibly other organelles or free ribosomes (Fig. 4B, also see Fig. 6αD). Axons and myelin sheaths were unstained (Fig. 4B). βIΣ2 spectrin was closely associated with the granular cell plasma membrane only at postsynaptic densities (PSDs), where it formed a thick plaque at synapses with both mossy fiber termini and with golgi cells (Fig. 5A). The mossy fiber terminals and the golgi axons themselves were negative (Fig. 5A).

In the Purkinje layer, cell bodies were negative but Purkinje cell dendrites showed a distinctive reactivity (Fig. 6A). In the dendritic stalk, βIΣ2 spectrin was localized both in stacked cisternae (Fig. 6C) and in filamentous structures resembling microtubules (Fig. 6D) in a pattern coincident with MAP2. As seen in granule cells, βIΣ2 spectrin was concentrated at the plasma membrane in Purkinje cells only at PSDs (Fig. 6B and 7A,C). Cross-sections of dendritic spines also revealed a punctate cytoplasmic distribution of βIΣ2 spectrin, again suggestive of its association with cytoplasmic microtubules (Fig. 7C). Stellate cells and the peripheral processes of astrocytes in the molecular layer also contained abundant cytoplasmic βIΣ2 spectrin (data not shown).

The mAb AP-20 recognizes epitopes on MAP2a and b, which are developmentally regulated, high molecular weight microtubule βInding proteins found mainly in neuronal dendrites and cell bodies (Matus et al., 1981; Tucker et al., 1988). Anti-MAP2 diffusely stained the cytoplasm of neurons, including that of granule cells (Fig. 4C). The dendrites of Purkinje cells were also positive, and the reaction product was localized along microtubules (Fig. 6E), in a pattern like that of βIΣ2 spectrin. Conversely, the synapsin I distribution did not overlap that of βIΣ2 spectrin. Synapsin I is a neuron-specific phosphoprotein primarily localized at presynaptic terminals, where it links small vesicles to cytoskeletal structures and modulates neurotransmitter release (Südhof and Jahn, 1991). It is synthesized in the neuronal cell body and conveyed to synaptic terminals by axonal transport (Petrucci et al., 1991). In the granular layer, synapsin I was present on vesicles in mossy fiber termini (Figs 4D and 5B). In the molecular layer, it marked presynaptic vesicles associated with synapses on the dendritic spines of Purkinje cells (Fig. 7B,D). In no instance did the distribution of synapsin I and βIΣ2 spectrin overlap.

Cerebellar primary cultures prepared from postnatal rats are enriched in granule neurons, which after some days in culture represent more that 95% of the total cells (Levi et al., 1984). After plating on poly-L-lysine-coated glass coverslips, they rapidly extend cell processes. By seven days in vitro, these cells display morphological, βIochemical and electrophysiological properties of excitatory interneurons (Cull-Candy et al., 1989; Kingsbury et al., 1985; Thangnipon et al., 1983), including the post-translational modification of αIIΣ* spectrin by physiological stimuli acting at the NMDA receptor-channel complex (Di Stasi et al., 1991). In these cell cultures after ten days growth in vitro, βIΣ2 spectrin was diffusely present in the cell body cytoplasm and along short cell processes (Fig. 8A) in a pattern indistinguishable from granular cells in vivo. All anti-βIΣ2 spectrin staining could be inhiβIted by pre-incubation of the antibody with 200 mg/ml peptide immunogen, confirming the specificity of the staining pattern (Fig. 8B). Significantly, more elongated neurites (Fig. 8A, arrowheads) did not contain detectable βIΣ2 spectrin, even though they did contain microtubules and MAP2 (Fig. 8C). MAP2 was also localized to the cytoplasm (Fig. 5C). αIIΣ* spectrin was present mainly in neurites (Fig. 8D), although faint, irregular staining was observed along the plasma membrane of the cell body.

The present findings indicate that βIΣ2 spectrin, a specific alternative transcript of the β-spectrin gene on (human) chromosome 2 that was first identified in skeletal muscle, is: (1) expressed in most but not all cells of the rat cerebellar cortex, (2) is highly polarized in a somαdendritic pattern, (3) is concentrated on the plasma membrane only at the PSD, (4) may be associated in the cytoplasm with various organelles and cytoskeletal structures, including some (but not all) microtubules and (5) is expressed and polarized in cultured cerebellar granule cells. At least some βIΣ2 spectrin is associated with αIIΣ* spectrin (αfodrin); however the intracellular distribution of αIIΣ* spectrin only partially overlaps that of βIΣ2 spectrin, indicating that either βIΣ2 spectrin is unassembled with an αsubunit in the cytoplasm of granule cells, or that additional αspectrin isoforms exist that are not recognized by the antibodies to αIIΣ* used in these experiments. No mixed complexes containing both βIΣ2 and βIIΣ1 spectrin appear to exist.

The distribution of βIΣ2 spectrin reported here is similar in many respects to the distribution of 240/235E spectrin (Riederer et al., 1986; Zagon et al., 1986). Spectrin 240/235E has been identified only by its reactivity to polyclonal antibodies to erythrocyte aβ-spectrin, and its precise relationship to other spectrins has remained uncertain. Spectrin 240/235E was detected in neuronal and glial cell bodies, dendrites, and in association with cytoplasmic organelles and cytoskeletal structures in mouse cerebellum (Zagon et al., 1986), a pattern that we confirmed using the commercially available antibodies to 240/235E from Chemicon. On the basis of similar cellular distributions and similar apparent molecular weights, it is thus likely that 235E spectrin is largely βIΣ2. However, the distribution of 240/235E is somewhat broader than that noted for βIΣ2 spectrin, since βIΣ2 spectrin is absent in Purkinje cell bodies and has a more restricted distribution on intracellular organelles. While these differences may relate to differences in sensitivity, it is more likely that the polyclonal erythrocyte spectrin antibody used to first identify 240/235E (Zagon et al., 1986) also detects additional isoforms of spectrin beyond βIΣ2. For example, the erythrocyte β-spectrin related isoform described in muscle that is associated with the acetylcholine receptor also reacts with some erythrocyte spectrin antibodies (Bloch and Morrow, 1989; VyβIral et al., 1992). This muscle isoform is distinct from βIΣ2 spectrin (Winkelmann and Forget, 1993), and if present in cerebellum might contribute to the staining pattern of 240/235E spectrin. It is also possible that αspectrin transcripts were being detected in the 240/235E studies, although there is no direct evidence that aIS1 spectrin (red cell αspectrin) is expressed in cerebellum. Future studies with additional isoform-specific antibodies will be required to sort out the relα tionship of specific spectrin transcripts to the staining patterns observed with broadly reactive antibodies.

The confinement of βIΣ2 but not αIIΣ* spectrin to the soma in cultured cerebellar neurons is also interesting, since it indicates that sorting mechanisms underlying compartmentalization of protein components still operate in vitro. However, these cells are not as fully differentiated as granule cells in vivo. Although they display the electrophysiological properties of excitatory interneurons, they do not form recognizable synapses. Perhaps reflecting this intermediate level of differentiation, the synapsin I distribution in cultured granule cells is beadlike along neurites (Gallo et al., 1986), and as noted here βIΣ2 spectrin does not migrate into the neurite processes. Yet, the presence of MAP2, a dendritic marker (Matus, 1988), in neurites suggests that at least some neurite processes may be (nascent?) dendrites, and raises the interesting speculation that MAP2 assembly on microtubules precedes the sorting of βIΣ2 spectrin to PSDs during neuronal ontogeny. In future experiments, it will be interesting to identify the factor(s) that trigger βIΣ2 spectrin transport into MAP2 containing neurites, and whether this event is sufficient to induce the formation of PSDs.

The largely intracellular localization of βIΣ2 spectrin, its apparent association with specific organelles and with microtubules, and its polarized distribution, challenge many aspects of the spectrin membrane skeleton paradigm that has been derived from studies of the red cell (for reviews, see Coleman et al., 1989; S. P. Kennedy and J. S. Morrow, unpublished). Only patchy and unconvincing accumulations of reaction product were found along the nuclear and mitochondrial membranes and over most of the plasma membrane. Conversely, significant and consistent staining for βIΣ2 was present on the membranes of the ER cisternae, on some cytoplasmic vesicles, along some microtubules, and at the PSD. Occasional filamentous structures bridging two microtubules were also stained by reaction product, although the resolution of these studies is insufficient to be certain that these bridges are real. It is possible that the apparent association of βIΣ2 spectrin with microtubules and cytoplasmic organelles is an artifact caused by the diffusion of peroxidase staining (De Camilli et al., 1983). However, we do not believe that this can account for all of the microtubule and organelle staining observed, since other organelles (eg. mitochondria), the nuclear membrane and the adjacent plasma membrane were not stained in regions where ER cisternae or microtubules were, and also because some microtubules, such as those in myelinated axons, remain unstained even though they abut intensely positive βIΣ2 stained areas. Also, in synapsin I-stained sections, peroxidase-labeled material was never observed along microtubules, even though microtubules are prominent in axons and synaptic terminae. Nevertheless, in future studies, it will be important to confirm this organelle and microtubule staining by immunogold microscopy.

So, if βIΣ2 spectrin is not simply forming a generalized erythrocyte-like infrastructure supporting the plasma membrane, what is its role in the ecology of the neuron? Its prominent association with the cellular machinery of membrane assembly and vectorial transport suggests that βIΣ2 spectrin plays a role in organizing and/or staβIlizing preassembled membrane and cytoplasmic protein complexes destined for the PSD. This hypothesis is similar to that proposed for the assembly of basolateral proteins in epithelial cells (Rodriguez-Boulan and Nelson, 1989), and as observed in sea urchin embryos (Fishkind et al., 1990). Spectrin may also directly link other soluble and membrane proteins to microtubules (Fach et al., 1985) to facilitate their transport to synapses along soma and dendrites. In this model, the multifunctionality of spectrin allows it to participate as a key organizing center about which macromolecular complexes containing both cytoplasmic and integral membrane proteins can form, and links them to other cytoskeletal elements such as microfilaments or microtubules (S. P. Kennedy and J. S. Morrow, unpublished data). Candidate proteins that might directly or indirectly βInd to such a spectrin-centered complex would include neurotransmitter receptors and calcium channels that have been found in postsynaptic density preparations (Siekevitz, 1991), proteins such as the a1 isoform of Na,K-ATPase (Brines et al., 1991; Morrow et al., 1989), and others from over 30 proteins found to be associated with the PSD (Walsh and Kuruc, 1992). Yet another component of these complexes may be dystrophin. In muscle, βIΣ2 spectrin has recently been co-localized with dystrophin in distinct suβ sarcolemmal domains (Porter et al., 1992), and dystrophin has been found in postsynaptic regions of cortical neurons (Lidov et al., 1990). Supporting this hypothesis is a recent observation of altered calcium fluxes in cerebellar granule cells of mdx mice (Woodward and Steinhardt, 1992). If this model of spectrin’s role proves true, spectrin will need to be considered a key component in the signal conduction cascade of neurons.

We thank Gianfranco Macchia and Ms Carol Cianci for expert technical assistance, Lamberto Camilli for printing photographs and Dr Elsa Traina for use of the culture facilities. This work was partially supported by the Italian Ministry of Health ‘Project on AIDS’ grant 620/6/67 (T.C.P.), by NIH grant NS 29611 (J.S.M.), and by a Basic Research Grant (FY91-0240) from the March of Dimes Foundation (J.S.M.). The financial support of Telethon-Italy for the project ‘Role of spectrin isoforms in the central nervous system of normal and dystrophic mdx mice’ (T.C.P.) is also gratefully acknowledged.