Beta subunit heterogeneity in neuronal L-type Ca2+ channels.

Heterologous expression studies have shown that the activity of voltage-gated Ca2+ channels is regulated by their beta subunits in a beta subunit isoform-specific manner. In this study we therefore investigated if one or several beta subunit isoforms associate with L-type Ca2+ channels in different regions of mammalian brain. All four beta subunit isoforms (beta1b, beta2, beta3, and beta4) are expressed in cerebral cortex as shown in immunoblots. Immunoprecipitation of (+)-[3H]isradipine-labeled L-type channels revealed that the majority of beta subunit-associated L-type channels was associated with beta3 (42 +/- 8%) and beta4 (42 +/- 7%) subunits, whereas beta1b and beta2 were present in a smaller fraction of channel complexes. beta3 and beta4 were also the major L-type channel beta subunits in hippocampus. In cerebellum beta1b, beta2, and beta3 but not beta4 subunits were expressed at lower levels than in cortex. Accordingly, beta4 was the most prominent beta subunit in cerebellar L-type channels. This beta subunit composition was very similar to the one determined for 125I-omega-conotoxin-GVIA-labeled N-type and 125I-omega-conotoxin-MVIIC-labeled P/Q-type channel complexes in cerebral cortex and cerebellum. Our data show that all four beta subunit isoforms associate with L-type Ca2+ channels in mammalian brain. This beta subunit heterogeneity may play an important role for the fine tuning of L-type channel function and modulation in neurons.

Voltage-gated Ca 2ϩ channels control the depolarization-induced influx of extracellular Ca 2ϩ into neurons and other electrically excitable cells. They exist as hetero-oligomeric complexes of different subunits (␣1, ␣2-␦, and ␤). Different types of neuronal Ca 2ϩ channels (termed L-, N-, P-, Q-, and R-type; 1) are discriminated by biophysical and pharmacological criteria (for reviews see Refs. [2][3][4][5]. N-and P/Q-type channels are blocked by peptide toxins (-CTx 1 -GVIA and -CTx-MVIIC or -agatoxin-IVA, respectively), whereas L-type channels are modulated by drugs, such as dihydropyridines (6). These channel types are differentially distributed in the brain and even within a neuron (7,8). Thereby they serve different physiolog-ical functions. N-and P/Q-type channels are abundant in nerve terminals and control Ca 2ϩ -dependent neurotransmitter release (3). L-type channels are localized mainly on neuronal cell somata and proximal dendrites where they may control Ca 2ϩdependent modulatory processes and excitation-transcription coupling (9).
The above Ca 2ϩ channel types consist of different ␣1 subunit isoforms (class A-E) that also form their drug or toxin binding domains and therefore determine their pharmacological properties (1). In contrast, important biophysical and modulatory properties, such as voltage-dependent gating (10,11) and channel modulation by G-proteins (12,13) and kinases (14), are determined not only by ␣1 but also by associated ␣2-␦ and ␤-subunits. Whereas only one ␣2-␦ isoform is known, four different ␤ subunit isoforms (␤1-␤4) are expressed in mammalian brain (15,16). Heterologous expression studies revealed that ␤ subunits can affect ␣1 function in a ␤ subunit isoform-specific manner. For example, Ca 2ϩ currents carried by ␣1A, ␣1E, and ␣1C inactivate faster with coexpressed ␤3 than with ␤2 (14,17,18) subunits. ␤1, ␤3, and ␤4, but not ␤2, are permissive for voltage-dependent facilitation of Ca 2ϩ channels formed by ␣1C (19). ␤3 and ␤1 subunits confer slightly different pharmacological properties to L-type channels (20). Therefore ␤ subunit heterogeneity could participate in the fine-tuning of channel function. However, it is unclear if only one or several ␤ subunit isoforms associate with these channels in mammalian brain. So far only the ␤ subunit composition of L-type Ca 2ϩ channels in skeletal muscle has been studied. In this tissue exclusively ␤1a subunits are associated with the channel complex (15,21).
Biochemical evidence for ␤ subunit heterogeneity in mammalian brain has recently been provided for N-type and P/Qtype channels (22,23). Multiple ␤ subunit isoforms were found to be associated to different extents with both channel types after extraction from whole rabbit brain.
Here we report that ␤ subunit heterogeneity also exists within neuronal L-type channels. We found that regional differences in the ␤ subunit expression pattern affect ␤ subunit composition in different regions of mammalian brain.
A preliminary report of our findings has appeared previously (24).
Sequence-directed Antibodies-For antibody production in rabbits peptides were coupled to bovine serum albumin with glutaraldehyde (25) or synthesized on a lysine branch (octavalent NovaSyn PA resin, * This work was supported by the Fonds zur Förderung der Wissenschaftlichen Forschung S6602 (to J. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Recipient of a Fullbright Fellowship. ¶ Supported by a grant from the Austrian Academy of Sciences. Work performed in partial fulfillment of a thesis.
Membrane Preparation-Membranes were prepared from guinea pig or rabbit cortex, hippocampus, cerebellum, and heart muscle as described (29). Brain regions were rapidly removed from rabbit or guinea pig brains and immediately placed in ice-cold homogenization buffer containing 0.02 M NaHCO 3 and a protease inhibitor mixture (2 mM EDTA, 0.2 mM PMSF, 0.5 mM benzamidine, 2 mM iodoacetamide, 1 M pepstatin A, 1 g/ml leupeptin, 1 g/ml aprotinin, 20 g/ml calpain inhibitor I and II, 0.1 mg/ml trypsin inhibitor). The tissues were then homogenized by 10 -20 strokes in a Dounce homogenizer, and microsomes were collected by centrifugation at 45,000 ϫ g (10 min, 4°C). Microsomes were then washed three times with 50 mM Tris-HCl, pH 7.4 (37°C), containing the same protease inhibitor mix. Membranes were resuspended in the same buffer at a protein concentration about 5 mg/ml and stored at Ϫ80°C until use.
Affinity Purification of ␤ Subunits and Immunoblotting-Glutathione S-transferase (GST) and a GST fusion protein with the ␣1 subunit interaction domain of the ␣1A subunit (AIDA) were prepared as described (21). All further steps were carried out on ice or at 4°C. Typically 20 mg of microsomal protein isolated from rabbit or guinea pig brain regions were solubilized in 9 ml of buffer A (50 mM Tris-HCl, pH 7.4, containing the protease inhibitors used for membrane preparation) supplemented with 1% (w/v) CHAPS and 1 M NaCl according to Ref. 21. 30-l aliquots of glutathione-Sepharose equilibrated in buffer B (buffer A containing 0.1% (w/v) CHAPS, 0.1 M NaCl) were coupled with 10 g of GST or GST-AIDA and washed three times with the above buffer. Solubilized membranes were diluted 10-fold in buffer A, and 4 ml were mixed with the coupled glutathione-Sepharose beads for 4 h or overnight. The beads were washed three times with 1.5 ml of buffer B, mixed with SDS-polyacrylamide gel electrophoresis sample buffer (15 min, 56°C or 3 min, 95°C), and the eluted protein separated on 10% polyacrylamide gels.
Immunoblot experiments were carried out as described (30). Prestained molecular weight markers (Bio-Rad) were run on the same gels. The apparent molecular masses of each batch were provided by the supplier.
High affinity binding of (ϩ)-[ 3 H]isradipine to solubilized L-type Ca 2ϩ channels was determined using a filtration assay as described (29). This assay underestimated the total specific (ϩ)-[ 3 H]isradipine binding activity by about 20%. This was taken into account to calculate the binding activity employed for immunoprecipitation assays.
Statistics-Data are given as means Ϯ S.D. for the indicated number of experiments.

Region-specific Expression of ␤ Subunit Isoforms in Mamma-
lian Brain-To investigate the association of all known subunit isoforms with neuronal voltage-gated L-type Ca 2ϩ channels in mammalian brain, we raised anti-peptide antibodies against unique sequences of ␤1b, ␤2, ␤3, and ␤4 subunits. In addition, an antibody against an epitope highly conserved in all ␤ subunit isoforms (anti-␤com) was generated. We used these antibodies to determine their association with neuronal Ca 2ϩ channels solubilized from rabbit or guinea pig cerebral cortex, hippocampus, and cerebellum membranes in immunoprecipitation experiments. Their expression in these brain regions was analyzed in Western blots.
To determine their relative expression densities the four ␤ subunit isoforms were extracted with CHAPS from microsomes prepared from brain (␤1b, ␤2, ␤3, ␤4) or, for control purposes, from skeletal muscle (␤1a). Extracts were affinity-purified on GST-AIDA-Sepharose (21) in the presence of protease inhibitors as described under "Experimental Procedures." As shown in Fig. 1 for skeletal muscle ␤1a (anti-␤com staining) and neuronal ␤3 and ␤4 subunits, the enrichment of ␤ subunit immunoreactivity was specific and absent when only GST was used as the affinity matrix (Fig. 1, lanes 4 -6). In ␤ subunit preparations from rabbit cerebral cortex anti-␤com specifically recognized a 63 Ϯ 3/67 Ϯ 3-kDa doublet and a 88 Ϯ 4-kDa band ( Fig. 2) (n Ն 4). A ϳ33-kDa band was also stained to a variable extent by ␤com as well as all the other ␤ antibodies and corresponded to the GST-AIDA polypeptide present at relatively high amounts (10 g) in the ␤ subunit preparations.
To assign the ␤com-stained bands to individual ␤ subunit isoforms, samples separated on the same gel were stained with isoform-specific antibodies. The 88-kDa band was composed of  4) and the supernatant after incubation with the resins (lanes 2 and 5) and resin-bound protein (lanes 3 and 6) were separated by SDS-polyacrylamide gel electrophoresis and analyzed for ␤ immunoreactivity in immunoblots employing anti-␤com. 52-56-kDa bands were stained as expected for ␤1a staining in skeletal muscle membranes (21). The same bands were stained in purified channel preparations (Ն90% pure, not shown). B, enrichment of ␤3 (left panel) and ␤4 (right panel) immunoreactivity from guinea pig cortex membranes. 20 mg of membrane protein were solubilized in a total volume of 9 ml and diluted 10-fold as described under "Experimental Procedures." 4-ml aliquots were subjected to affinity chromatography on GST-AIDA (10 g) coupled to glutathione-Sepharose. Immunostaining was with affinity-purified anti-␤3 (lanes 1-3, left) or anti-␤4 (lanes 1-3, right). Lanes 1 and 4, starting material (40-l aliquots); lanes 2 and 5, supernatants (40 l aliquots); lanes 3 and 6, GST-AIDA resin-bound protein. No other bands were specifically enriched. The electrophoretic mobilities of the stained bands were indistinguishable from those in Fig. 2. One of two experiments yielding similar results is shown.
␤ subunit isoform staining was specific. It was completely suppressed in the presence of 1 M of the respective antigenic peptides (not shown). As expected, only ␤com but not the isoform-selective antibodies specifically recognized ␤1a extracted from partially purified skeletal muscle T-tubule membranes ( Fig. 2A). ␤com staining in rabbit heart represented ␤2 immunoreactivity (Fig. 2B) suggesting that other isoforms are absent or expressed at much lower levels in this tissue.
The same bands were also present in hippocampus (Fig. 2) and cerebellum extracts (Fig. 3). The relative abundance of the 88-kDa band was lowest in cerebellum because ␤1b and ␤2 expression density was lower in this region as compared with cerebral cortex (Fig. 2). When similar amounts of solubilized membrane protein from cerebral cortex and cerebellum were subjected to ␤ subunit isolation and Western blotting (Fig. 3) similar ␤com staining intensity was found for the 63/67-kDa doublet. ␤3-specific immunoreactivity was less abundant in cerebellum, whereas ␤4 was expressed at similar densities as in cerebral cortex (Fig. 3).
Taken together, the ␤com staining pattern in mammalian brain can be explained by the presence of all four ␤ subunit isoforms which are expressed in a region-specific pattern.
Neuronal L-type Ca 2ϩ Channels Are Associated with Different ␤ Subunit Isoforms in Mammalian Brain-After having established the specificity of our antibodies, we investigated if L-type Ca 2ϩ channels are associated with only one or several ␤ subunit isoforms and if ␤ subunit association varies in different brain regions. We reversibly labeled neuronal L-type Ca 2ϩ channels complexes in cerebral cortex, hippocampus, and cerebellum membranes with the L-type Ca 2ϩ channel-selective ligand (ϩ)-[ 3 H]isradipine and solubilized them in buffer containing 1% (w/v) digitonin. In cerebral cortex and hippocampus 74 Ϯ 9% (n ϭ 4) and 91 Ϯ 22% (n ϭ 4) of the solubilized (ϩ)-[ 3 H]isradipine labeling was immunoprecipitated with saturating concentrations of an antibody directed against ␣1C indicating that binding was associated with L-type channel complexes. 61 Ϯ 18% (n ϭ 5) and 80 Ϯ 31% (n ϭ 4) of the labeled L-type channels were immunoprecipitated by ␤com. Therefore, most of the L-type channel complexes are associated with a ␤ subunit which is accessible for ␤com under nondenaturating conditions. The results of Western blot analysis are shown. ␤ subunits were extracted from equal amounts of membrane protein prepared from cerebral cortex (CTX) or cerebellum (CER) as described in Fig. 2 and separated on adjacent lanes. Immunostaining was carried out with the indicated affinity-purified antibodies. Molecular weight markers are as in Fig. 2.   FIG. 2. Expression of ␤1b, ␤3, and ␤4 subunits in different brain regions. The results of Western blot analysis are shown. ␤ subunits were purified from the indicated brain regions (A and B), skeletal muscle (A) or heart microsomes (B) by GST-AIDA affinity chromatography and separated by SDS-polyacrylamide gel electrophoresis together with prestained marker proteins and subjected to immunoblot analysis with ␤ subunit isoformselective, affinity-purified antibodies. Skeletal muscle affinity chromatography on GST-AIDA-Sepharose was as in Fig. 1. For all neuronal tissues and cardiac muscle comparable amounts of membrane protein (0.45 ml of solubilization buffer per mg of protein) were used for solubilization and subsequent analysis (4 ml of diluted extract). Samples from one brain region were always separated on the same gel. In some experiments individual lanes were cut in half longitudinally and probed with different antibodies to allow an exact comparison of the relative migration of immunostained bands. The numbers denote ␤ subunit antibody selectivities. C, ␤com antibody recognizing an epitope common to all four ␤ subunit isoforms. The arrows indicate specifically stained ␤ subunits as discussed in the text. The migration of prestained molecular weight markers (105,000; 82,000; 49,000; 33,300; 28,600) is indicated on the left. CTX, cerebral cortex; CER, cerebellum; HIP, hippocampus; SKM, skeletal muscle. One of at least three independent experiments yielding similar results is shown.
Immunoprecipitation experiments with the isoform-selective antibodies revealed the association of more than one ␤ isoform with the channel complex. Affinity-purified anti-␤3 and anti-␤4 antibodies each immunoprecipitated 42% of the radioactivity recognized by anti-␤com (Fig. 4A). Smaller fractions were bound by anti-␤1b and anti-␤2 (Fig. 4A). Together our subunitspecific antibodies accounted for all (118%) ␤com immunoprecipitable radioactivity in cerebral cortex.
Immunoprecipitation by these antibodies was saturable (see Fig. 5C). The nonspecific background signal observed with the same concentrations of control rabbit immunoglobulin was less than 10% (n Ͼ 3) of the radioactivity recognized by anti-bcom.
A similar ␤ subunit composition was observed in hippocampus (Fig. 4C). In the cerebellum only ␤4 accounted for a large portion of L-type channel-associated ␤ subunits (Fig. 4D). Immunoprecipitation by ␤3 antibodies was less pronounced than in cerebral cortex (Fig. 4D). This is in good agreement with the lower relative abundance of ␤3 in this region (Fig. 3). Immunoprecipitation by ␤1b and ␤2 was difficult to detect in cerebellum (Fig. 4D) representing less than 10% of the channels immunoprecipitated by anti-␤com.
Together the isoform-selective antibodies accounted for most but not all of the ␤ subunit-associated radioactivity in hippocampus (70%) and cerebellum (66%).
Similar ␤ Subunit Composition of L-, N-, and P/Q-type Ca 2ϩ Channels-Next we tested if the ␤ subunit composition of Ltype Ca 2ϩ channels resembles the subunit composition of Nand P/Q-type Ca 2ϩ channels in these regions (22,23). For Nand P/Q-type channels it has been investigated before in digitonin extracts of whole brain membranes, but data on individual brain regions are unavailable. We therefore also subjected 125 I--CTx-GVIA-and 125 I--CTx-MVIIC-labeled channel complexes extracted from cerebral cortex and cerebellum to immunoprecipitation with our antibodies. We have previously shown that under our experimental conditions saturable high affinity 125 I--CTx-GVIA and 125 I--CTx-MVIIC binding occurs selectively to N-type and P/Q-type Ca 2ϩ channels, respectively, with dissociation constants in the subpicomolar range (31).
In cerebral cortex and cerebellum saturating concentrations of anti-␤com recognized 85 Ϯ 23% (n ϭ 5) and 84 Ϯ 13 (n ϭ 4) of channels associated with 125 I--CTx-GVIA binding activity, respectively. The immunoprecipitation profile was very similar to L-type channels (Fig. 5, A and B). ␤3 and ␤4 subunits together immunoprecipitated Ͼ80% of ␤com immunoprecipitable 125 I--CTx-GVIA binding in a saturable manner (Fig.  5C). As with L-type channels, a smaller fraction of N-type channel binding was associated with ␤1b and ␤2. In cerebellum again only ␤4 antibodies recognized substantial portions of N-type channel activity (Fig. 5B). Similar results as described for N-type and L-type channels were also obtained for 125 I--CTx-MVIIC-labeled P/Q-type channels in cerebral cortex (not shown). In cerebellum only ␤4 antibodies recognized significant FIG. 5. Specific immunoprecipitation of solubilized 125 I--CTx-GVIA-labeled N-type Ca 2؉ channels and 125 I--CTx-MVIIC-labeled P/Q-type Ca 2؉ channels from different brain regions. A-B, immunoprecipitation was as described for L-type Ca 2ϩ channels. Reversible labeling of channels was carried out at low picomolar concentrations of 125 I--CTx-GVIA (Ͻ10 pM). Data are shown for n ϭ 3-8. Numbers denote the ␤ isoform to which antibodies were generated; C, ␤com. C, the concentration-dependent immunoprecipitation of 125 I--CTx-GVIA binding activity from solubilized rabbit cerebral cortex membranes by antibodies against ␤com (f, upper abscissa), ␤1b (Ⅺ, upper abscissa), ␤3 (E, lower abscissa), and ␤4 (q, lower abscissa) is shown. One of at least two experiments yielding similar results is shown.
FIG. 4. Specific immunoprecipitation of solubilized (؉)-[ 3 H]isradipinelabeled L-type Ca 2؉ channels from different brain regions and skeletal muscle. Immunoprecipitation was carried out as described under "Experimental Procedures." A-C, the dpm immunoprecipitated by saturating concentrations of the respective isoform-selective antibody were normalized with respect to the dpm immunoprecipitated by saturating concentrations of anti-␤com (Ͼ1500 dpm in cerebral cortex; Ͼ2600 dpm in hippocampus; Ͼ270 dpm in cerebellum). Data are shown for n ϭ 3 with the exception of ␤1b and ␤2 immunoprecipitation (n ϭ 2) in cerebellum. D, skeletal muscle Ca 2ϩ channels were partially purified by affinity chromatography on WGA-Sepharose as described (37) and labeled with 2 nM (ϩ)-[ 3 H]isradipine. Aliquots of the labeled channel preparation were diluted with RIA buffer to a final volume of 0.3 ml and subjected to immunoprecipitation as described for neuronal channels. One of two typical experiments is shown. Numbers denote the ␤ isoform to which antibodies were generated; C, ␤com.
Isoform-selective antibodies completely accounted for the Ntype (105%, Fig. 5A) and P/Q-type (Ͼ85%, not shown) channel binding recognized by anti-␤com in cerebral cortex but only for 40 -50% in cerebellum. As for L-type channels this difference cannot be attributed to differences in membrane preparation because it was also found when the respective brain regions were isolated from the same animals in the same buffer and carried through the whole solubilization and immunoprecipitation procedure in parallel. It is therefore possible that in hippocampus and cerebellum immunoprecipitation by one or several of our antibodies was underestimated. At present we do not know if this is due to the expression of a yet uncharacterized ␤ subunit isoform, which is immunoprecipitated by ␤com but none of the other antibodies, or due to region-specific differences in proteolysis. C-terminal proteolysis could remove the C-terminal epitopes of our isoform-specific antibodies. However, we have obtained no evidence for extensive proteolytic breakdown of ␤ subunits in immunoblots with our ␤com antibody, which recognizes an epitope located near the N terminus of the ␤ subunits.

DISCUSSION
␤ Subunit Heterogeneity within L-type Ca 2ϩ Channels-The major findings of our study are as follows. 1) All known ␤ subunit isoforms participate in the formation of neuronal Ltype Ca 2ϩ channels in mammalian brain. 2) ␤3 and ␤4 subunits are most often found as part of the neuronal L-type channel complexes.
3) The fractional contribution of a particular ␤ subunit isoform for channel formation varies among different brain regions. 4) The ␤ subunit composition and regional differences are very similar to N-and P/Q-type channels in cerebral cortex and cerebellum.
This similarity of the ␤ subunit composition between L-type channels and N-as well as P/Q-type channels is interesting because the subcellular distribution of L-type channels in neurons differs significantly from the distribution of N-and P/Qtype channels. L-type ␣1C and ␣1D subunits are predominantly found on the cell soma and proximal dendrites, whereas N-type ␣1B and P/Q-type ␣1A are also found along the length of dendrites and in presynaptic terminals (8,32). Despite these differences in neuronal targeting, these channel types do not show major differences with respect to their ␤ subunit composition. Obviously different ␤ subunit isoforms can be targeted to different regions of a neuron.
Both ␣1C and ␣1D subunits participate in the formation of L-type Ca 2ϩ channels in mammalian brain. We have made no attempts to determine if differences exist between the two L-type channels with respect to their ␤ subunit composition. The fraction of channels associated with ␣1D is small (not more than 9 -26% in hippocampus and cerebral cortex as revealed by our immunoprecipitation experiments with ␣1C; see also Ref,33) and therefore complicates such an analysis. We cannot exclude the possibility that ␤1b and ␤2, which are found only in a minor fraction of channels, are selectively associated only with ␣1D. However, based on our finding that ␣1C is associated with the majority of labeled channels in cerebral cortex and hippocampus, ␤-subunit heterogeneity must exist within class C L-type channels in these regions.
Implications for Neuronal L-type Ca 2ϩ Channel Function-␤ subunits strongly affect the functional properties of the poreforming ␣1 subunits of L-type (and non L-type) channels. As shown by heterologous coexpression in Xenopus oocytes and mammalian cells, ␤ subunits affect channel gating (10, 11), modulation by G-proteins (12,13), and phosphorylation (14) as well as Ca 2ϩ and drug (34,35) interaction with L-type ␣1 subunits. Such studies also revealed that different ␤ isoforms are able to confer different channel properties. For example, ␤3 confers a more rapid inactivation to currents mediated by ␣1C (17) than does ␤2. ␤ isoform-specific effects on channel inactivation were also observed for non-L-type Ca 2ϩ channel ␣1 subunits (11,14,18). Only ␤1, ␤3, and ␤4, but not ␤2, support voltage-dependent facilitation of ␣1C-mediated Ca 2ϩ currents (19). Similarly, small differences in the sensitivity of the channel to the Ca 2ϩ antagonist mibefradil and the modulation by protein kinase C were observed when different ␤ subunits form part of the channel complex (14,20). We now provide direct biochemical evidence that indeed different ␤ subunits participate in the formation of neuronal L-type channels suggesting that these isoform-selective effects contribute to L-type Ca 2ϩ channel plasticity in mammalian brain. ␤ subunits could be involved in the fine tuning of L-type channel function in a region-specific manner. Based on our findings future coexpression studies should preferentially focus on the comparison of the properties of L-type channels containing ␤3 or ␤4 subunits, because these isoforms seem to be present in the majority of dihydropyridine-sensitive L-type channels in cortex, hippocampus, and cerebellum.
Taken together our data demonstrate that, like in other neuronal Ca 2ϩ channel types, several ␤ subunit isoforms contribute to the formation of neuronal L-type channels. Further studies must focus on the physiological and pathophysiological consequences of this heterogeneity and investigate if changes in ␤ subunit expression could account for changes in L-type Ca 2ϩ channel function also under pathophysiological conditions, such as neurodegeneration, cerebral ischemia, or aging (36).