C-terminal splice variants of P/Q-type Ca2+ channel CaV2.1 α1 subunits are differentially regulated by Rab3-interacting molecule proteins

Voltage-dependent Ca2+ channels (VDCCs) mediate neurotransmitter release controlled by presynaptic proteins such as the scaffolding proteins Rab3-interacting molecules (RIMs). RIMs confer sustained activity and anchoring of synaptic vesicles to the VDCCs. Multiple sites on the VDCC α1 and β subunits have been reported to mediate the RIMs-VDCC interaction, but their significance is unclear. Because alternative splicing of exons 44 and 47 in the P/Q-type VDCC α1 subunit CaV2.1 gene generates major variants of the CaV2.1 C-terminal region, known for associating with presynaptic proteins, we focused here on the protein regions encoded by these two exons. Co-immunoprecipitation experiments indicated that the C-terminal domain (CTD) encoded by CaV2.1 exons 40–47 interacts with the α-RIMs, RIM1α and RIM2α, and this interaction was abolished by alternative splicing that deletes the protein regions encoded by exons 44 and 47. Electrophysiological characterization of VDCC currents revealed that the suppressive effect of RIM2α on voltage-dependent inactivation (VDI) was stronger than that of RIM1α for the CaV2.1 variant containing the region encoded by exons 44 and 47. Importantly, in the CaV2.1 variant in which exons 44 and 47 were deleted, strong RIM2α-mediated VDI suppression was attenuated to a level comparable with that of RIM1α-mediated VDI suppression, which was unaffected by the exclusion of exons 44 and 47. Studies of deletion mutants of the exon 47 region identified 17 amino acid residues on the C-terminal side of a polyglutamine stretch as being essential for the potentiated VDI suppression characteristic of RIM2α. These results suggest that the interactions of the CaV2.1 CTD with RIMs enable CaV2.1 proteins to distinguish α-RIM isoforms in VDI suppression of P/Q-type VDCC currents.

Fine regulation of neurotransmitter release is integral to adaptive functions of the nervous system, including learning, memory, and cognition. Neurotransmitter release is triggered by depolarization-induced Ca 2ϩ influx via voltage-dependent Ca 2ϩ channels (VDCCs) 2 in presynaptic active zones (AZs), where synaptic vesicles (SVs) dock in close vicinity to VDCCs at the presynaptic membrane (1,2). Among different VDCC types, which are distinguished on the basis of their pharmacological and biophysical properties, L-, N-, R-, and P/Q-types have been reported to mediate Ca 2ϩ influx responsible for neurotransmitter release (3)(4)(5)(6). Different VDCC types show distinct tissue expression patterns, subcellular localizations, activity-dependent properties, and amounts of Ca 2ϩ influx, all of which contribute to the fine regulation of neurotransmitter release (7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18). In particular, the local Ca 2ϩ concentration ([Ca 2ϩ ] local ) and spacing between VDCCs and SVs are tightly regulated by the molecular organization of presynaptic AZs and influence the dynamic properties of neurotransmitter release (2, 9, 18 -27). It is also understood that the number of open VDCCs, which determines the [Ca 2ϩ ] local and release probability of SVs, depends on the efficiency of targeting and availability of VDCCs in the AZ (28). In response to membrane depolarization, VDCCs open to evoke [Ca 2ϩ ] local rises and simultaneously close via inactivation. This negative feedback reduces the number of VDCCs available and restricts the amplitude of Ca 2ϩ influx, which is important for the diversification of Ca 2ϩ signaling (29). Inactivation of VDCCs in the presynapse is largely dependent upon the inward Ca 2ϩ current magnitude and displays only a weak voltage dependence (13,30).
In the P/Q-type, VDCCs are composed of the pore-forming ␣ 1 subunit (Ca V 2.1) and accessory ␣ 2 ␦, ␤, and ␥ subunits. Ca V 2.1 is the most abundantly expressed VDCC ␣ 1 subunit in the mammalian brain (31), and mutations in the Ca V 2.1 gene, cacna1a, cause several autosomal-dominant neurological disorders, including familial hemiplegic migraine type 1, episodic ataxia type 2, and spinocerebellar ataxia type 6 (SCA6) (32)(33)(34). Multiple functional P/Q-type VDCC variants are generated by alternative splicing of subunit genes (31,35), different subunit compositions (36), post-translational processing (37), and association with interacting proteins (20,24,38,39). Several types of P/Q-type VDCC complexes can be co-localized in a single neuron and are believed to contribute to the fine-tuning of neuronal processes, such as neurotransmitter release, because formation of each type of Ca V 2.1 channel complex is regulated in a different manner (35, 40 -42).
Rab3-interacting molecules (RIMs) are multidomain scaffolding proteins expressed in secretory cells (43). Long isoform ␣-RIMs, including RIM1␣ and RIM2␣, contain an N-terminal zinc finger domain, a central PDZ domain, and two C-terminal domains, C 2 A and C 2 B. Physiological experiments have shown that ␣-RIMs are essential for docking and priming of SVs and for recruiting and tethering VDCCs to the presynaptic AZ, thereby regulating VDCC function and short-term plasticity of neurotransmitter release (20, 22, 44 -49). We have reported that ␣-RIMs increase neurotransmitter release by sustaining Ca 2ϩ influx through strong inhibition of voltage-dependent inactivation (VDI) of VDCCs and by anchoring vesicles in the vicinity of VDCCs via interaction with VDCC ␤ subunits ( Fig.  1A) (20). We have also revealed that the RIM C-terminal C 2 B domain is essential for RIM-␤ subunit interaction and inhibition of VDI of VDCCs (20,39). Mutations in the gene encoding RIMs associated with autism and cone-rod dystrophy, CORD7, modify this interaction and/or the regulation of VDCC currents (50,51). Functional coupling of RIM1␣ to ␤ subunits of VDCCs is also essential for insulin secretion in non-neuronal cells (52). In addition to the ␤ subunits, the PDZ domain of ␣-RIMs has been reported to interact with the PDZ-binding motif located at the end of the C terminus of ␣ 1 subunits to modulate localization of P/Q-and N-type VDCC complexes to presynaptic AZs ( Fig. 1A) (22). Thus, ␣-RIMs may interact with VDCC complexes through multiple sites of the constituent subunits. However, the significance of multipoint interactions among VDCC ␣ 1 subunits, ␤ subunits, and ␣-RIMs, as well as functional effects of interactions between ␣-RIMs and the Ca V 2.1 C terminus on VDCC currents, remains unclear.
To quantify the functional significance of multipoint interaction, it is interesting to focus on alternative splicing of exons 44 and 47, because this generates major Ca V 2.1 C-terminal splice variants expressed in the human cerebellum ( Fig. 1B and Table 1). Introns 42-44 are flanked by GT/AG splice-site sequences, and alternative splicing leads to either the inclusion or exclusion of exons 43 and 44 (referred to as (ϩ43 or Ϫ43) and (ϩ44 or Ϫ44) in Fig. 1B) (53). Insertion of a pentanucleotide GGCAG at the beginning of exon 47 allows in-frame translation of exon 47 to produce a long version of the C terminus (referred to as 47) in Fig. 1B). Otherwise, omission of the GGCAG in transcripts causes a frameshift, leading to stop codon termination near the beginning of exon 47 (referred to as (⌬47) in Fig. 1B) to generate the human homolog of the rabbit VDCC ␣ 1A subunit BI-I (Fig. 1B) (31,34,53). The 12-amino acid region encoded by exon 44 starts with the arginine residue, which is located 292 amino acids downstream from the transmembrane segment S6 of repeat IV. The exon 44-encoded region is thought to have an AT-hook domain, which is a tripartite DNA-binding motif specific for AT-rich sequences that is typically found in nuclear proteins and DNA-binding proteins (54,55). The 244-amino acid region encoded by exon 47 with the GGCAG insertion starts with the glycine residue, which is located 451 amino acids downstream from S6 of repeat IV. The exon 47-encoded region has Src homology 3 and PDZ domain-binding motifs, which are targets of synaptic proteins such as CASK, Mint1, RIM-binding protein (RIM-BP), and ␣-RIMs (19,22,56). It is also known that expansion of the polyglutamine tract (polyQ), encoded by CAG trinucleotide repeats in exon 47 of human Ca V 2.1, causes the neurological disease, SCA6 (34).
Here, we studied the interactions between ␣-RIMs and the Ca V 2.1 C-terminal regions encoded by exons 44 and 47. We revealed the functional impacts of ␣-RIM interaction with different Ca V 2.1 C-terminal regions on its VDI. The 17 amino acid residues on the C-terminal side of the polyQ stretch play an essential role in the pronounced VDI suppression characteristic of RIM2␣. In the Ca V 2.1 splice variant lacking exons 44 and 47, VDI suppression remained intact for RIM1␣, but for RIM2␣, it was reduced to a level comparable with that of RIM1␣. These results suggest that the CTD region plays an important role in the ␣-RIM isoform-dependent potentiation of VDI suppression. Also, our data reveal that the interaction of ␣-RIMs with the CTD regions encoded by exons 44 and 47 is not essential for the suppressive effects of ␣-RIMs on VDI, further raising the possibility that interactions of the VDCC ␤ subunits with ␣-RIMs underlie their strong suppressive effects on the VDI of VDCCs.

Characterization of Ca V 2.1 C-terminal splice variation in the human cerebellum
We have previously demonstrated that the RIM C-terminal region containing the C 2 B domain interacts with VDCC ␤ subunits (20,39,50). It has also been reported that PDZ domains of ␣-RIMs interact with the PDZ-binding motif located at the C-terminal end of Ca V 2.1 and Ca V 2.2 ( Fig. 1A) (22,57). The C-terminal region is highly divergent in VDCC ␣ 1 subunits because of multiple alternative splice sites (58). In particular, splicing out exon 47 generates Ca V 2.1 splice variants that lack the most C-terminal region, including the PDZ-binding motif. It is important to quantitatively assess the relative significance of ␣-RIM interactions with the C-terminal region of Ca V 2.1 and the VDCC ␤ subunit by comparing ␣-RIM actions on P/Q-type VDCCs containing different Ca V 2.1 splice variants carrying the C terminus with and without the ␣-RIM-interacting region. Previous studies have revealed that exons 43, 44, and 47 contribute to C-terminal splice variations in human Ca V 2.1 (53), but relative levels of splice variants with different combinations of these exons have not been quantified. We performed sequence analysis of PCR products from a cDNA library of the human cerebellum, in which abundant expressions of ␣-RIMs and Ca V 2.1 mRNAs were reported (39,59,60). A set of PCR oligonucleotide primers were located in exon 42 (forward) and ␣-RIMs diversify inactivation of Ca V 2.1 splice variants exon 47 (reverse). Agarose gel electrophoresis (1%) revealed a broad band of PCR products of ϳ1,000 bp consistent with the predicted sizes ranging from 834 to 989 bp (Fig. 1C). This DNA band was subcloned into a vector, and the relative levels of splice variants were determined by counting the number of clones containing each exon. The relative proportions of individual splice variants of exon ϩ43/Ϫ43, ϩ44/Ϫ44, and 47/⌬47 were 100/0, 86/14, and 66/34%, as reported previously (Table 1) (53). Ca V 2.1-containing regions encoded by exons 44 and 47 (ϩ44 ,47) were detected at the highest relative proportion (56%) (Fig. 1A and Table 1). Relative proportions of Ca V 2.1 (ϩ44,⌬47), Ca V 2.1 (Ϫ44,47), and Ca V 2.1 (Ϫ44,⌬47) were 30, 11, and 3%, respectively ( Fig. 1A and Table 1).

Interaction between ␣-RIMs and Ca V 2.1 C-terminal splice variants
We next performed yeast two-hybrid screening of a human brain cDNA library using the C-terminal domain encoded by exons 40 -47 of human Ca V 2.1, Ca V 2.1 CTD (ϩ44,47), as bait, and we identified an interaction between Ca V 2.1 CTD and the amino acid residues 487-1349 of human RIM2␣ (GenBank TM accession number NM_001100117) ( Fig. 2A). We also performed co-immunoprecipitation (co-IP) experiments to confirm RIM-Ca V 2.1 CTD interactions (Fig. 2B). As a control, we chose the VDCC ␤ 4 subunit, because ␤ 4 is abundantly expressed in the brain and the spontaneous ␤ 4 mutant lethargic mouse (cacnb4 lh ) has clear neurological defects, supporting the physiological significance of ␤ 4 in the brain (61,62). YFP-tagged ␤ 4 was co-immunoprecipitated with FLAG-tagged ␣-RIMs, RIM1␣ and RIM2␣, in HEK293T cells (Fig. 2C), as reported previously (20,39). Next, we performed co-IP between YFPtagged ␣-RIMs and FLAG-tagged CTDs of Ca V 2.1 variants derived from alternative splicing of exons 44 and 47 in HEK293T cells. RIM1␣ and RIM2␣ were co-immunoprecipitated with the Ca V 2.1 CTD splice variants except for the variant lacking exons 44 and 47 (Fig. 2D). These results suggest that the two regions encoded by exons 44 and 47 contribute significantly to the interaction between the Ca V 2.1 CTD and ␣-RIMs.

␣-RIMs diversify inactivation of Ca V 2.1 splice variants Functional impacts of RIM-Ca V 2.1 C-terminal interaction on Ca V 2.1 channel properties
We have previously reported that RIMs strongly suppress VDI of neuronal VDCCs by interacting with the ␤ subunits (20,39,50). It has also been reported that the RIM-Ca V 2.1 C-terminal interaction modulates localization of VDCCs to presynaptic AZs (22). However, the effects of RIM-Ca V 2.1 C-terminal interaction on VDCC properties have not been examined. We characterized whole-cell Ba 2ϩ currents through recombinant P/Qtype VDCCs containing the Ca V 2.1 splice variants, ␤ 4 and ␣ 2 ␦-1 subunits, in HEK293 cells. We chose Ba 2ϩ as a charge carrier, because Ca V 2.1 splice variants show different Ca 2ϩ -dependent properties in HEK293 cells (63). Voltage dependence of inactivation at different voltages (inactivation curve) (Fig.  3A) was first examined in Ca V 2.1 (ϩ44,47)-expressing and Ca V 2.1 (Ϫ44,⌬47)-expressing cells. No significant difference was detected between inactivation curves of Ca V 2.1 (ϩ44,47) and Ca V 2.1 (Ϫ44,⌬47) channels ( Fig. 3B and Table 2), which is consistent with previous reports (63,64). In cells co-expressing ␣-RIMs and Ca V 2.1 (Ϫ44,⌬47), we detected remarkable inactivation curve shifts toward depolarizing potentials, as described previously (20,39). Co-expression of RIM2␣ induced a more significant inactivation curve shift toward depolarizing potentials compared with that for RIM1␣ in Ca V 2.1 (ϩ44,47)-expressing cells, whereas this difference between ␣-RIMs was not observed in Ca V 2.1 (-44,⌬47)-expressing cells (the half-in-  47) with amino acid residues 487-1349 of human RIM2␣ (GenBank TM accession number NM_001100117.1) as a function of time. The interactions are scored by ␤-galactosidase activity judged on a scale of one to five, with one meaning "low activity" and five meaning "high activity." B, domain structures of mouse ␣-RIMs. The arrows indicate molecules interacting with RIM proteins at the following domains: Zn 2ϩ finger-like domain (Zn 2ϩ ), PDZ domain (PDZ), first and second C 2 domains (C 2 A and C 2 B), and proline-rich region (PXXP). Primary ␤ subunit-binding site (RIM1␣(1079 -1257)) and ␤ subunit modulatory region (RIM1␣(1258 -1463)) are indicated (20). C, interactions of the YFP-tagged ␤ 4 subunit with FLAG-tagged ␣-RIMs in HEK293T cells. D, interactions of the FLAG-tagged Ca V 2.1 CTD splice variants with YFP-tagged ␣-RIMs in HEK293T cells. The interactions are evaluated by co-IP with monoclonal anti-FLAG antibody, followed by WB with polyclonal anti-YFP antibody. Input is 10% of the amount of cell lysate used for co-IP and is analyzed by WB using polyclonal anti-YFP antibody. Immunoprecipitation (IP) of FLAG-tagged Ca V 2.1 CTD splice variants or ␣-RIMs with monoclonal anti-FLAG antibody is analyzed by WB using polyclonal anti-FLAG antibody.
To further assess the importance of interactions via the regions encoded by exons 44 and 47 in this potentiation characteristic of RIM2␣ in suppression of VDI, we examined the effect of RIM2␣ on Ca V 2.1 (Ϫ44,47) and Ca V 2.1 (ϩ44,⌬47) inactivation curves. A significantly enhanced shift in the inactivation curve toward depolarizing potentials was detected in cells co-expressing RIM2␣ and Ca V 2.1 (Ϫ44,47), but not in cells co-expressing RIM2␣ and Ca V 2.1 (ϩ44,⌬47), compared with cells co-expressing RIM2␣ and Ca V 2.1 (Ϫ44,⌬47) (V 0.5 of Ca V 2.1 (Ϫ44,47) and Ca V 2.1 (ϩ44,⌬47) with RIM2␣ was Ϫ8.6 Ϯ 2.0 and Ϫ15.4 Ϯ 2.4 mV, respectively) ( Fig. 3B and Table 2). We could not observe a significant difference between Ca V 2.1 (ϩ44,47) and Ca V 2.1 (Ϫ44,⌬47) in current density-voltage (I-V) relationships, with or without RIM2␣ co-expression ( Fig. 4 and Table 3). These results suggest that the RIM-Ca V 2.1 C-terminal interaction via the region encoded by exon 47 is important for the potentiated suppressive effect of RIM2␣ on VDI.
Inactivation kinetics of P/Q-type VDCCs was characterized by analyzing the decay phase of Ba 2ϩ currents evoked by 1-s test pulses in HEK293 cells (Fig. 5A). The decay phase was well fitted by two exponential functions with a non-inactivating component (Fig. 5B) as reported previously (41). The two exponential time constants ( fast and slow ) and the ratio of fast, slow, and non-inactivating components were similar in Ca V 2.1 (ϩ44,47)-and Ca V 2.1 (Ϫ44,⌬47)-expressing cells at test poten-  Table 2 for statistical significance of the differences. Error bars, S.E.

Inactivation parameters
The fast-inactivating component was significantly decreased at 0 and 10 mV, whereas the non-inactivating component as well as slow were significantly increased by ␣-RIMs at 0, 10, and 20 mV (Fig. 5, C and D, and Table 4). Thus, it is unlikely that these effects are mediated by interaction of ␣-RIMs with the Ca V 2.1 C-terminal region encoded by exons 44 and 47. In cells co-expressing RIM2␣ and Ca V 2.1 (ϩ44,47), increases in the noninactivating component at 0, 10, and 20 mV and fast at 0 and 10 mV were more pronounced compared with cells co-expressing RIM2␣ and Ca V 2.1 (Ϫ44,⌬47) ( Fig. 5, C and D, and Table 4). This effect can be mediated by the RIM2␣-Ca V 2.1 C-terminal interaction. Furthermore, currents evoked by trains (100 Hz) of action potential (AP)-like waveforms for 3 s, a more physiological voltage-clamp protocol used to determine closed-state inactivation (65), showed a more rapid decrease in amplitude in cells co-expressing RIM2␣ and Ca V 2.1 (Ϫ44,⌬47) compared with cells co-expressing RIM2␣ and Ca V 2.1 (ϩ44,47) (Fig. 5E). These results suggest that RIM2␣-Ca V 2.1 C-terminal interaction is not essential for but enhances the VDI suppression induced by ␣-RIMs.
It is important to note that, without the interaction of ␣-RIMs, these splice variants were indistinguishable in channel properties such as inactivation curve ( Fig. 3 and Table 2), inactivation kinetics ( Fig. 5 and Table 4), and I-V relationship ( Fig. 4 and Table 3). This underscores the significance of protein-protein interaction in differentiating functional properties of alternatively spliced Ca V 2.1 variants.
To clarify the RIM2␣-interacting regions responsible for the potentiation of VDI suppression by RIM2␣, additional deletion mutants of Ca V 2.1 were constructed (Fig. 7A). In cells  Table 3 for statistical significance of the differences. Error bars, S.E.

Table 3
Effect of ␣-RIMs on the I-V relationships of P/Q-type VDCCs in HEK293 cells expressing Ca V 2.1s, ␣ 2 /␦-1, and ␤ 4 The number of cells analyzed are indicated in parentheses.
These data suggest that 17 amino acid residues, 2328 -2344 (RPGRAATSGPRRYPGPT), on the C-terminal side of the polyQ stretch of Ca V 2.1 (ϩ44,47) is a RIM2␣-interacting region that potentiates the suppressive effect on VDI.
In co-IP experiments, however, YFP-tagged RIM2␣ showed a comparable level of co-IP with FLAG-tagged CTD of Ca V 2.1 (ϩ44,2344X) and FLAG-tagged CTD of Ca V 2.1 (ϩ44,2327X) but a decreased level of co-IP with CTD of Ca V 2.1 (ϩ44,2275X) (Fig. 7C). To eliminate possible contributions of exon 44-en- Figure 5. Effects of ␣-RIMs on VDI kinetics of P/Q-type Ca V 2.1 channels. A, effects of ␣-RIMs on inactivation of Ba 2ϩ currents mediated by P/Q-type Ca V 2.1 splice variants in HEK293 cells expressing ␤ 4 and ␣ 2 ␦-1 subunits. The peak amplitudes are normalized for Ba 2ϩ currents elicited by 1-s pulses to 10 mV from a V h of Ϫ90 mV. B, Ba 2ϩ current evoked by 1-s test pulse to 10 mV from a V h of Ϫ90 mV in HEK293 cells expressing ␤ 4 , ␣ 2 ␦-1 subunits, and RIM1␣. Current decay is fitted by a sum of two exponential functions with time constants of 55 and 995 ms, whose fraction of components is 0.19 and 0.61, respectively. The fraction of its non-inactivating component is 0.24. C, voltage dependence of the two inactivation time constants, fast and slow . The mean inactivation time constants are plotted as a function of test potential from 0 to 20 mV. D, voltage dependence of the fraction of the three components, fast-, slow-, and non-inactivating components. The fractions of the components are plotted against test potentials. †, p Ͻ 0.05; † †, p Ͻ 0.01; statistical significance of differences between cells co-expressing Ca V 2.1 (Ϫ44,⌬47) and RIM2␣ and cells co-expressing Ca V 2.1 (ϩ44,47) and RIM2␣. See Table 4 for statistical significance of the differences.

␣-RIMs diversify inactivation of Ca V 2.1 splice variants
coded amino acid residues to the interaction of RIM2␣ with CTD mutants, we next constructed deletion mutants of the CTD of Ca V 2.1 (Ϫ44,47). The co-IP was nearly abolished for the CTD of Ca V 2.1 (Ϫ44,2263X), in which 2263X corresponds to 2275X in Ca V 2.1 (ϩ44,47) (Fig. 7D). This result suggests that the amino acid residues 2264 -2280 (GTSTPRRGRRQ-LPQTPS) of Ca V 2.1 (Ϫ44,47), which correspond to 2276 -2292 of Ca V 2.1 (ϩ44,47), are an important region for the RIM2␣-Ca V 2.1 CTD interaction. However, the co-IP experiments failed to unveil interaction between the 17 amino acid residues, 2328 -2344, of Ca V 2.1 (ϩ44,47) with RIM2␣, although this region is supposed to be essential for potentiated VDI suppression by RIM2␣.
It is interesting to note that deletion of the region encoded by exon 44 failed to elicit suppression of interaction of the Ca V 2.1 CTD containing the region encoded by exon 47 (compare the bands of Ca V 2.1 CTD (Ϫ44,47) and Ca V 2.1 CTD (ϩ44,47) in Fig. 2D). In contrast, deletion of the region encoded by exon 47 suppressed the interaction of the Ca V 2.1 CTD containing the region encoded by exon 44 (compare the bands of Ca V 2.1 CTD (ϩ44,⌬47) and Ca V 2.1 (ϩ44,47) in Fig. 7C). These results suggest that the region encoded by exon 47 binds more strongly to ␣-RIMs compared with the region encoded by exon 44. Thus, formation of RIM2␣-Ca V 2.1 complexes is mediated by multipoint interaction.  Fig. 3 and is shown for comparison. See Table 5 for statistical significance of the differences.

Effects of polyQ elongation in the Ca V 2.1 C-terminal region on regulation of VDI by RIM2␣
The polyQ stretch is an interesting characteristic of the Ca V 2.1 C-terminal primary structure. SCA6 is caused by expansion of the polyQ tract in the human Ca V 2.1 gene from a normal repeat size range of 4 -17 to a size range of 20 -33 (34,54,66). To confirm the effect of polyQ expansion on the interaction between the Ca V 2.1 C terminus and RIM2␣, association between YFP-tagged RIM2␣ and FLAG-tagged CTD of Ca V 2.1 (ϩ44,47) Gln-40 with an elongated polyQ stretch of 40 residues was tested by co-IP in HEK293T cells (Fig. 9A). The intensity of the co-IP band for RIM2␣ normalized to the IP band for the CTD was moderately but significantly decreased for Ca V 2.1 CTD (ϩ44,47) Gln-40 compared with that for Ca V 2.1 CTD (ϩ44,47) with a polyQ stretch of 11 residues (Fig. 9B). To examine the effect of polyQ elongation on suppression of VDI by RIM2␣, a recombinant P/Q-type VDCC was expressed as a complex of Ca V 2.1 (ϩ44,47) with polyQ expansion (Ca V 2.1 (ϩ44,47) Gln-40), ␤ 4 , and ␣ 2 ␦-1 subunits in HEK293 cells (Fig.  9C). Without co-expression of ␣-RIMs, inactivation curves of whole-cell Ba 2ϩ currents elicited by Ca V 2.1 (ϩ44,47) and Ca V 2.1 (ϩ44,47) Gln-40 were indistinguishable ( Fig. 9C and Table 2). This contradicts with our previous report that polyQ expansion itself causes a hyperpolarizing shift of the inactivation curve for rabbit Ca V 2.1 channels carrying the ␤ 1 subunit (67). It is possible that the effect of polyQ expansion on VDI depends on species and ␤ subtypes (68). When RIM2␣ was co-expressed, inactivation curves of Ca V 2.1 (ϩ44,47) and Ca V 2.1 (ϩ44,47) Gln-40 were indistinguishable (V 0.5 of Ca V 2.1 (ϩ44,47) Gln-40 and Ca V 2.1 (ϩ44,47) were Ϫ11.0 Ϯ 2.2 mV and Ϫ6.7 Ϯ 2.3 mV, respectively) ( Fig. 9C and Table 2). These data indicate that polyQ expansion reduced the binding affinity of Ca V 2.1 C terminus for RIM2␣ but not the suppressive effect of RIM2␣ on the VDI of Ca V 2.1 (ϩ44,47).

Functional impacts of ␣-RIMs on VDI of N-type Ca V 2.2 and R-type Ca V 2.3 channels
Ca V 2 VDCCs are a major source of presynaptic Ca 2ϩ influx (69 -71), and ␣-RIMs also interact with the C terminus of Ca V 2.2 (22,57). To explore the generality of our findings using Ca V 2.1, associations of the C-terminal region of Ca V 2.2  10B and Table 6), as reported previously (20). RIM2␣ induced significant reduction of low voltage-inactivated phases (the ratios of low voltage-inactivating phases of Ca V 2.2 co-transfected with vector, RIM1␣, and RIM2␣ were 0.64 Ϯ 0.11, 0.63 Ϯ 0.12, and 0.29 Ϯ 0.05, respectively) ( Fig. 10B and Table 6). Suppressive effects of ␣-RIMs on VDI were also observed for Ca V 2.3 (the ratios of inactivating components of Ca V 2.3 with co-transfection of vector, RIM1␣, and RIM2␣ were 0.90 Ϯ 0.03, 0.63 Ϯ 0.09, and 0.67 Ϯ 0.08, respectively). We failed to detect signifi-  Fig. 3 and is shown for comparison. *, p Ͻ 0.05; **, p Ͻ 0.01; statistical significance of differences versus cells co-expressing RIM2␣ and Ca V 2.1 (ϩ44,47). See Table 5 for statistical significance of the differences. Error bars, S.E. C, interactions of FLAG-tagged Ca V 2.1 CTD (ϩ44) deletion mutants with YFP-tagged RIM2␣ in HEK293T cells. D, interactions of FLAG-tagged Ca V 2.1 CTD (Ϫ44) deletion mutants with YFP-tagged RIM2␣ in HEK293T cells. E, interactions of FLAG-tagged Ca V 2.1 CTD (Ϫ44,⌬2264 -2280) deletion mutants with YFP-tagged RIM2␣ in HEK293T cells. The interactions are evaluated by co-IP with monoclonal anti-FLAG antibody, followed by WB with polyclonal anti-YFP antibody. Input is 10% of the amount of cell lysate used for co-IP and is analyzed by WB using polyclonal anti-YFP antibody. IP of FLAG-tagged Ca V 2.1 CTDs with monoclonal anti-FLAG antibody is analyzed by WB using polyclonal anti-FLAG antibody. Dashed lines divide mutants in terms of co-IP level (blue) and inactivation curve (red).

␣-RIMs diversify inactivation of Ca V 2.1 splice variants
cant differences between the effect of RIM1␣ and RIM2␣ on VDI of Ca V 2.3 channels (Fig. 10B and Table 6). Thus, N-type Ca V 2.2 channels but not R-type Ca V 2.3 channels are susceptible to RIM2␣-mediated potentiation of VDI suppression.

Discussion
In presynaptic AZs, where SVs dock in close vicinity to VDCCs at the presynaptic membrane, depolarization-induced Ca 2ϩ influx via VDCCs triggers neurotransmitter release (1, 2). Previous proteomic analysis has shown that P/Q-, N-, and R-type VDCCs are embedded into protein networks assembled from a pool of 200 proteins (72). Unveiling the manner of protein-protein interactions in protein networks and their functional consequences is important for understanding the process of synapse formation and the modulation of synaptic transmission (73). Previously, we reported that ␣-RIMs increase neurotransmitter release by sustaining Ca 2ϩ influx through strong inhibition of VDI of VDCCs and by anchoring vesicles in the vicinity of VDCCs via the RIM-␤ subunit interaction (20,39). It has also been reported that ␣-RIMs interact with the Ca V 2.1 C-terminal region and modulate VDCCs targeting to presynaptic AZs (22). These previous studies have shown that RIM-VDCC interactions are key to the protein assembly responsible for stimulus-secretion coupling in presynaptic AZs.
Multipoint interaction plays important roles in the regulation of properties such as stabilization of protein complexes. In RIM-VDCC complexes, the significance of multipoint interaction has not yet been resolved. To approach this question, the contributions of each interaction should be quantitatively assessed. Our data strongly indicate that RIM-␤ subunit interaction is sufficient and necessary for ␣-RIMs to exert prominent suppressive effects of VDI of VDCCs, because in the Ca V 2.1 splice variant with the deletion of exons 44 and 47 (Ca V 2.1 (Ϫ44,⌬47)), strong VDI suppression remains intact for RIM1␣, but for RIM2␣ it is attenuated to a level comparable with that of RIM1␣. In the presynapse, VDCCs display a weak VDI (13,30). Moreover, in the calyx of Held nerve terminals of RIM1 and RIM2 conditional double knock-out mice, depolarization pre-pulses induced stronger inactivation of VDCCs compared with wild-type mice (48).
Ca V 2.1 is known to be inactivated through at least two voltage-dependent mechanisms (fast and slow inactivation) (74). The mechanism underlying inactivation is not completely understood but may involve "hinged lid" or pore block-type  Table 5 for statistical significance of the differences. Error bars, S.E. D, interactions of the deletion mutants of FLAG-tagged Ca V 2.1 (Ϫ44,⌬2264 -2280) with YFP-tagged RIM2␣(1183-1572) in HEK293T cells. The interactions are evaluated by co-IP with monoclonal anti-FLAG antibody, followed by WB with polyclonal anti-YFP antibody. Input is 10% of the amount of cell lysate used for co-IP and is analyzed by WB using polyclonal anti-YFP antibody. IP of FLAG-tagged Ca V 2.1 CTDs with monoclonal anti-FLAG antibody is analyzed by WB using polyclonal anti-FLAG antibody.

␣-RIMs diversify inactivation of Ca V 2.1 splice variants
mechanisms (75)(76)(77). It has also been reported that fast and slow inactivation represents structurally independent conformational changes (78). Our kinetic analyses of current decay showed that both RIM1␣ and RIM2␣ decreased the fast inactivation component and increased slow regardless of the presence of the exon 44 and 47 regions in Ca V 2.1 (Fig. 5, C and D), although only RIM2␣ (but not RIM1␣) increased fast in the presence of the exon 44 and 47 regions in Ca V 2.1 (Fig. 5C). These findings may suggest that the interaction between Ca V 2.1 CTD and RIM2␣ induces conformational changes of the VDCC Ca V 2.1 ␣ 1 subunit in addition to those induced by the interaction between ␤ subunit and ␣-RIMs.
By focusing on the splice variants of the Ca V 2.1 C-terminal region, we have deepened our understanding of the interaction between VDCCs and ␣-RIMs. ␣-RIMs interact with the Ca V 2.1 C-terminal region via the regions encoded by exons 44 and 47, the alternative splicing of which generates major C-terminal variants in the human cerebellum. Our experiments revealed at least four regions in the Ca V 2.1 CTD involved in interaction with RIM2␣: a site encoded by exon 44 and three sites in the region encoded by exon 47. Analysis of relative mRNA levels of these Ca V 2.1 C-terminal splice variants indicates that 56, 86, or 67% of the total Ca V 2.1 mRNA carries either both exons 44 and 47, exon 44 alone, or exon 47 alone, respectively (Table 1). These spliced mRNAs are capable of encoding the Ca V 2.1 variants that interact with ␣-RIMs. The effect of RIM2␣-Ca V 2.1 C-terminal interaction on VDI can be generated from 67% of the total Ca V 2.1 mRNA. Interestingly, there is a wide range of fundamental properties for individual synapses, including release probability, unitary response, and effects of previous stimulation on subsequent response (79). This suggests that Ca V 2.1 C-terminal splice variants may contribute to the heterogeneous molecular composition and function of VDCCs in  Table 2 for statistical significance of the differences. Error bars, S.E.  Table 6 for statistical significance of the differences. *, p Ͻ 0.05; **, p Ͻ 0.01; statistical significance of differences between RIM1␣-and RIM2␣-expressing cells. Error bars, S.E.

␣-RIMs diversify inactivation of Ca V 2.1 splice variants
presynaptic AZs. Importantly, a recent study has suggested that tissue-regulated alternatively spliced exons are significantly enriched in flexible regions of proteins that form conserved interaction surfaces to establish tissue-dependent protein-protein interaction networks (80). It is therefore possible that arrangements of presynaptic proteins that include VDCC subunits may be precisely coordinated via protein-protein interactions through the regions encoded by alternatively spliced exons 44 and 47. In terms of the effect of ␣-RIMs on VDCC channel properties, the interaction between RIM2␣ and the 17 amino acid residues, 2328 -2344, of Ca V 2.1 (ϩ44,47) in the region encoded by exon 47 of Ca V 2.1 is responsible for potentiation of VDI suppression (Fig. 7, B and E). In contrast, the interaction between ␣-RIMs and the region encoded by exon 44 of Ca V 2.1 failed to exert any significant effect on VDCC channel properties (Fig. 3B). Alternative splicing of exons 44 and 47 may contribute to the fine adjustment of RIM-VDCC complexes that can affect neurotransmitter release from AZs of different types of presynapses.
The interaction between the Ca V 2.1 C terminus and RIM-BPs, which bind to ␣-RIMs, is important for the coupling of SVs, VDCCs, and the SV fusion machinery (19,81). The AZ protein Bassoon interacts with both ␤ subunits and RIM-BPs to regulate Ca V 2.1 targeting to the AZ (9,(82)(83)(84). The ␤ subunits are also known to directly interact with the Ca V 2.1 C terminus (85). These findings may inform the physiological significance of interactions among the Ca V 2.1 C-terminal region, ␤ subunits, and ␣-RIMs. Interestingly, the RIM-BP-binding region in the Ca V 2.1 C-terminal region is the proline-rich region (PXXP) in amino acid residues 2276 -2292 of Ca V 2.1 (ϩ44,47), which is a RIM2␣-binding region in the region encoded by exon 47 (19). Previous studies have revealed that the amino acid residues 2502-2505 of Ca V 2.1 (ϩ44,47) bind to ␣-RIMs (22) and are essential for interaction with Mint1 (56). These results raise the possibility that the Ca V 2.1 C-terminal region is the target of AZ scaffolding proteins and that their competitive binding underlies dynamic properties of AZ protein networks in the molecular processes of neurotransmitter release.
Notably, both RIM1␣ and RIM2␣ bind to the C-terminal regions of Ca V 2.1, Ca V 2.2, and Ca V 2.3 (Figs. 2D and 10A), but potentiation of the suppressive effect on VDI was only observed for Ca V 2.1 or Ca V 2.2 combined with RIM2␣ (Figs. 3B and 10B). As already mentioned above, among four regions in the Ca V 2.1 CTD involved in interaction with RIM2␣, the two regions of Ca V 2.1 CTD (the amino acid residues 2276 -2292 and 2502-2505 of Ca V 2.1 (ϩ44,47)) are highly conserved among the three Ca V 2 isoforms (supplemental Fig. 1). It has been shown that Bassoon localizes Ca V 2.1 but not Ca V 2.2 to AZs via molecular interaction with RIM-BPs (9), despite the fact that RIM-BPs can also interact with the PXXP motif of Ca V 2.2 (19). Furthermore, mutation of the PDZ-binding motif of EGFP-tagged Ca V 2.1 failed to affect the localization pattern of Ca V 2.1 in cultured mouse hippocampal neurons (86). These results indicate that the functions of some protein-protein interactions are dependent on other components and that use of appropriate assay systems and neuronal types is necessary to reveal their functionality.
From a pathological point of view, it is interesting that mutations associated with genetic diseases in the genes encoding RIMs modify their function in regulating VDCC currents (50,51). In SCA6 patients, the relative mRNA level of the Ca V 2.1 splice variant, which possesses exon 47, is increased in cerebellar Purkinje cells but not in granule cells (87). In several episodic ataxia type 2 patients, mutations in Ca V 2.1 result in the loss of the regions encoded by exons 44 and 47 (32). These two diseases have similar symptoms, such as ataxic gait and loss of limb coordination. Although our results showed that the polyQ elongation itself does not affect functional regulation of VDCCs by RIM2␣, increase in the relative proportion of the Ca V 2.1 splice variants that possess the region encoded by exon 47 may cause excessive RIM2␣ functional regulation. Dysregulation of the molecular organization of presynaptic AZs containing VDCC complexes may lead to abnormalities in different functional hierarchies of nervous system.

Cell culture and cDNA expression in HEK293 or HEK293T cells
HEK293 and HEK293T cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, 30 units/ml penicillin, and 30 g/ml streptomycin at 37°C under 5% CO 2 . Transfection of cDNA plasmids was carried out using SuperFect Transfection Reagent (Qiagen). For electrophysiological measurements, recombinant plasmids were cotransfected with pIRES2-EGFP (Clontech), and HEK293 cells with green fluorescence were analyzed. Transfected cells were grown for 36 -48 h before electrophysiological measurements and co-IP assay.

Characterization of splice variation of Ca V 2.1 C terminus in the human cerebellum by sequence analysis
We designed PCR oligonucleotide primers as follows: forward in the exon 42, 5Ј-GCTGGTCACACCTCACAAG-3Ј, and reverse in the exon 47, 5Ј-GCTGGGCTTCCACTTACG-3Ј. Temperature cycles were as follows. Temperature initially 98°C for 2 min was followed by 25 cycles at 98°C for 10 s, 64°C for 30 s, and 68°C for 1 min. The PCR products from human cerebellum cDNA (Takara, 9523) were electrophoresed on an 1% agarose gel, cut out, purified using the QiaexII gel extraction kit (Qiagen), and ligated into the EcoRV-digested pBluescript II SK(Ϫ) (Stratagene). The ligation products were transformed into competent Escherichia coli DH5␣ cells and screened on Luria-Bertani plates containing ampicillin, X-Gal, and isopropyl ␤-D-1-thiogalactopyranoside for 15-20 h at 37°C. White colonies were picked randomly and subjected to sequencing analyses using T7 and M13-reverse universal sequencing primers. The sequencing results were analyzed by using BioEdit software version 7.2.5.

Yeast two-hybrid screening and ␤-galactosidase assay
We subcloned CTD of human Ca V 2.1 (ϩ44,47) into pGBK-T7 and used it as a bait to screen a human brain pACT2 library (Clontech) in the yeast strain AH109 (Clontech). We plated transformants (2.5 ϫ 10 6 ) on synthetic medium lacking adenine, histidine, leucine, and tryptophan and assayed His ϩ colonies for ␤-galactosidase activity with a filter assay. Of the transformants, 56 were His ϩ , and 6 of these were also LacZ ϩ . We isolated prey clone encoding amino acid residues 487-1349 of RIM2␣ (GenBank TM accession number NM_001100117).

co-IP assay in HEK293T cells
36 -48 h after transfection, HEK293T cells were solubilized in Nonidet P-40 buffer (150 mM NaCl, 50 mM Tris, 1% Nonidet P-40, and 1 mM PMSF) and then centrifuged at 17,400 ϫ g for 20 min. The cell lysate was incubated with anti-FLAG M2 monoclonal antibody (Sigma, F3165), and then the immunocomplexes were incubated with protein A-agarose beads (Santa Cruz Biotechnology, SC-2001), and the beads were washed with Nonidet P-40 buffer. The co-immunoprecipitated and immunoprecipitated proteins were characterized by Western blotting (WB) with anti-YFP antibody (Clontech, 632592) and anti-FLAG antibody (Sigma, F7425), respectively. Input was 10% of the amount of cell lysate used for co-IP and was characterized by WB with anti-YFP antibody. The chemiluminescence intensities of the bands were measured by Multigauge version 3.0 (Fuji film).

Current recordings
Whole-cell mode of the patch-clamp technique was carried out at 22-25°C with an EPC-10 (HEKA Elektronik) patchclamp amplifier as described previously (39,41,90). Patch pipettes were made from borosilicate glass capillaries (1.5-mm outer diameter, 0.87-mm inner diameter; Hilgenberg) using a model P-97 Flaming-Brown micropipette puller (Sutter Instrument Co.). The patch electrodes were fire-polished. Pipette resistance ranged from 2 to 4 megohms when filled with the pipette solutions described below. The series resistance was electronically compensated to Ͼ60%, and both the leakage and the remaining capacitance were subtracted by the ϪP/4 method. Currents were sampled at 10 kHz after low-pass filtering at 3.0 kHz (3 db) in the experiments of inactivation kinetics and AP-like trains, otherwise sampled at 20 kHz after low-pass filtering at 3.0 kHz (3 db). Data were collected and analyzed using Patchmaster (HEKA Elektronik) software. An external solution contained 5 mM BaCl 2 , 148 mM tetraethylammonium chloride, 10 mM HEPES, and 10 mM glucose (pH 7.4-adjusted with tetraethylammonium-OH). The pipette solution contained 95 mM CsOH, 95 mM aspartate, 40 mM CsCl, 4 mM MgCl 2 , 5 mM EGTA, 2 mM disodium ATP, 5 mM HEPES, and 8 mM creatine phosphate (pH 7.2-adjusted with CsOH).

Voltage dependence of inactivation
To determine the inactivation curve of VDCCs, Ba 2ϩ currents were evoked by 40-ms test pulse to 10 mV after the 10-ms repolarization to Ϫ90 mV following 1-s (300 ms for Ca V 2.3) prepulse voltage (V pre ) displacement (conditioning pulse) from Ϫ80 to 20 mV with 10-mV increments. Amplitudes of currents elicited by the test pulses were normalized to those elicited by the test pulse after a 1-s V pre displacement to Ϫ80 mV. The mean values were plotted against potentials of the 1-s V pre displacement. When the inactivation curve was monophasic, the mean values were fitted to the single Boltzmann equation, h(V pre ) ϭ (1 Ϫ a) ϩ a/(1 ϩ exp((V 0.5 Ϫ V pre )/k)), where a is the rate of inactivating component; V 0.5 is the potential to give a half-value of inactivation; and k is the slope factor. Otherwise, the mean values were fitted to the sum of two Boltzmann equations: h(V pre ) ϭ (1 Ϫ a Ϫ b) ϩ a/(1 ϩ exp((V 0.5 low Ϫ V pre )/k low )) ϩ b/(1 ϩ exp((V 0.5 high Ϫ V pre )/k high )), where a, b, and (1 Ϫ a Ϫ b) are the ratios of a low voltage-induced phase, a high voltageinduced phase, and a non-inactivating phase; V 0.5 low and V 0.5 high are the potentials that give a half-value of components susceptible to inactivation at low voltages in inactivation curves and at high voltages; and k low and k high are the slope factors. The decay phase of Ba 2ϩ currents evoked by 1-s test pulses was fitted by two (fast and slow) exponential functions with a non-inactivating component: ϪI(t) ϭ a ϩ bexp(Ϫ 1 t) ϩ cexp(Ϫ 2 t), where I(t) is the inactivating current as a function of time; a is the current amplitude at t ϭ ∞; b and c are the amplitudes of the time-dependent components; and 1 and 2 are the reciprocals of the fast ( fast ) and the slow ( slow ) time constants of inactivation, respectively (41,91). APs began at Ϫ80 mV and peaked at 33 mV. Rising and falling slopes were 283 and Ϫ103 V/s, respectively (20,65). Leaks and capacitive transients were subtracted by a ϪP/4 protocol.

I-V relationships
The individual activation data were fitted to standard Boltzmann equation in the form I Ba ϭ G max (V m Ϫ V rev )/(1 ϩ exp(Ϫ(V m Ϫ V 0.5 )/k)), where G max is the maximal conductance; V m is the membrane voltage; V rev is the I Ba reversal potential; V 0.5 is the half-activation potential, and k is the slope factor.

Statistical analysis
All data are expressed as the means Ϯ S.E. The statistical analyses were performed using Student's t test. A value of p Ͻ 0.05 was considered significant.