A New β Subtype-specific Interaction in α1ASubunit Controls P/Q-type Ca2+ Channel Activation*

The cytoplasmic β subunit of voltage-dependent calcium channels modulates channel properties in a subtype-specific manner and is important in channel targeting. A high affinity interaction site between the α1 interaction domain (AID) in the I-II cytoplasmic loop of α1 and the β interaction domain (BID) of the β subunit is highly conserved among subunit subtypes. We describe a new subtype-specific interaction (Ss1) between the amino-terminal cytoplasmic domain of α1A (BI-2) and the carboxyl terminus of β4. Like the interaction identified previously (21) between the carboxyl termini of α1A and β4 (Ss2), the affinity of this interaction is lower than AID-BID, suggesting that these are secondary interactions. Ss1 and Ss2 involve overlapping sites on β4 and are competitive, but neither inhibits the interaction with AID. The interaction with the amino terminus of α1 is isoform-dependent, suggesting a role in the specificity of α1-β pairing. Coexpression of β4 in Xenopus oocytes produces a reduced hyperpolarizing shift in the I-V curve of the α1A channel compared with β3 (not exhibiting this interaction). Replacing the amino terminus of α1A with that of α1C abolishes this difference. Our data contribute to our understanding of the molecular organization of calcium channels, providing a functional basis for variation in subunit composition of native P/Q-type channels.

The cytoplasmic ␤ subunit of voltage-dependent calcium channels modulates channel properties in a subtype-specific manner and is important in channel targeting. A high affinity interaction site between the ␣ 1 interaction domain (AID) in the I-II cytoplasmic loop of ␣ 1 and the ␤ interaction domain (BID) of the ␤ subunit is highly conserved among subunit subtypes. We describe a new subtype-specific interaction (Ss1) between the amino-terminal cytoplasmic domain of ␣ 1A (BI-2) and the carboxyl terminus of ␤ 4 . Like the interaction identified previously (21) between the carboxyl termini of ␣ 1A and ␤ 4 (Ss2), the affinity of this interaction is lower than AID-BID, suggesting that these are secondary interactions. Ss1 and Ss2 involve overlapping sites on ␤ 4 and are competitive, but neither inhibits the interaction with AID. The interaction with the amino terminus of ␣ 1 is isoform-dependent, suggesting a role in the specificity of ␣ 1 -␤ pairing. Coexpression of ␤ 4 in Xenopus oocytes produces a reduced hyperpolarizing shift in the I-V curve of the ␣ 1A channel compared with ␤ 3 (not exhibiting this interaction). Replacing the amino terminus of ␣ 1A with that of ␣ 1C abolishes this difference. Our data contribute to our understanding of the molecular organization of calcium channels, providing a functional basis for variation in subunit composition of native P/Qtype channels.
Despite their functional diversity, high voltage-gated Ca 2ϩ channels have three subunit types in common (1,2). The ␣ 1 , pore-forming component of the channel is associated with a cytoplasmic ␤ subunit of 52-78 kDa and a largely extracellular ␣ 2 ␦ component, anchored by a single transmembrane domain. These subunits are encoded by at least 7 ␣ 1 , 4 ␤, and 1 ␣ 2 ␦ genes, respectively, of which numerous splice variants exist (3).
The ␤ subunit, when coexpressed with the ␣ 1 subunit, results in an increase in current density, alteration of the voltage dependence and kinetics of both inactivation and activation, and an increase in the number of recognition sites for channelspecific ligands (for review, see Refs. 4 and 5). These effects reflect not only conformational modulation but also an increase in the number of channels properly addressed to the cell sur-face, suggesting multiple roles for the ␤ subunit. Although the effects of ␤ are highly conserved, significant differences are seen depending on the combination of ␣ 1 and ␤ subunits studied. For example, the kinetics of inactivation shows a general trend of variation with ␤ subtype (6 -9), whereas a shift in the voltage dependence of inactivation has been reported only for non-L-type, A, B, and E (10 -12), and not L-type channels (13). ␤ subunits also seem to differ in the mechanism by which they become localized to the plasma membrane (14,15), perhaps suggesting that they are differentially targeted. Finally, ␣ 1 and ␤ subtypes differ in their potential (based on sequence predictions) to be phosphorylated by various protein kinases. These factors together point to a functional explanation for the growing evidence that the in vitro promiscuity of ␣ 1 -␤ interactions is reflected by a heterogeneity of combinations in native channels (N (16), P/Q (17), and L type (18)).
Preliminary studies (10,19) have identified a high affinity interaction between a highly conserved region in the cytoplasmic loop linking transmembrane regions I and II of ␣ 1 (AID, 1 or ␣ 1 interaction domain) and a 30-residue region in the second conserved domain of ␤ subunits (BID, or ␤ interaction domain). This interaction occurs with a stoichiometry of 1:1 (20) and (at least in vitro and in expression systems) occurs between all combinations of ␣ 1 and ␤ subtypes tested so far. We have since reported (21) the existence of a subunit-specific interaction between the carboxyl-terminal domain of ␣ 1A and the most carboxyl-terminal 109 residues of ␤ 4 , and a similar interaction has been reported (22,23) between ␣ 1E and ␤ 2a . The comparative high affinity of the AID-BID interaction (20,21), coupled with the abolition of all ␤ modulatory effects by mutation of residues critical to the interaction between AID and BID (10,19), suggests that this interaction represents a primary, anchoring interaction upon which further, secondary, interactions might depend. The specificity of such secondary interactions, or at least differences in affinity, represents a potential source for the variation seen for different ␣ 1 and ␤ combinations, in terms of both the electrophysiological properties of the channel and potential differences in control by other cellular factors, such as protein kinases and G proteins. We therefore set out to determine whether further secondary interaction sites exist. The present report describes the identification of an interaction between the amino-terminal cytoplasmic region of ␣ 1A and the ␤ 4 subunit of P/Q channels providing a refreshed understanding of the molecular organization of voltagedependent calcium channels. The interaction plays a critical role in the precise positioning of the channel activation process on the voltage axis. It constitutes yet another molecular determinant underlying functional differences among various ␤ subunits and, by extension, probably among various native P/Q channel subtypes.
Binding Assays-These were carried out using fusion proteins coupled to glutathione-agarose in Tris-buffered saline as described previously (21). Binding reactions were incubated for 5 h unless otherwise stated.

RESULTS
A GST fusion protein, GST-NT A , expressing the entire amino-terminal cytoplasmic region of ␣ 1A (splice variant BI-2) was assayed for in vitro binding to 35 S-␤ 4 . As Fig. 1A shows, the NT A region exhibits a significant and specific interaction with ␤ 4 which is comparable to the binding observed to a GST fusion protein carrying the AID A sequence. The binding of GST-NT A to 35 S-␤ 4 appears slightly stronger than the binding of GST-AID A , but the relative efficiency of binding of these fusion proteins varied slightly depending on the ␤-translation reactions used. The affinity of this interaction was determined by carrying out similar binding assays using a range of concentrations of GST-NT A fusion protein. Fig. 1B shows the resulting saturation curve, which is compared with that observed previously for the interaction of 35 S-␤ 4 with GST-AID A . The affinity of interaction of GST-NT A is 100-fold lower (k D ϭ 336 nM) than that for the AID interaction (close to 3 nM). These data are in favor of the idea that the AID-BID interaction represents a primary anchoring site of interaction between the two subunits which allows secondary interactions of lower affinity to occur. As already mentioned, it is interesting that in Fig demonstrates greater binding than GST-AID A to 35 S-␤ 4 . Given that both fusion proteins are at concentrations giving maximal binding (Fig. 1B), this demonstrates a difference in maximal binding which appears to reflect a difference in conformational requirements, coupled with conformational heterogeneity in the 35 S-␤ 4 preparation (permissive and nonpermissive binding states; data not shown).
To characterize more precisely the region of ␣ 1A responsible for interaction with ␤ 4 , a series of GST fusion proteins carrying truncations of the region concerned (depicted in Fig. 2A) was constructed, and the proteins were assayed for their capacity to interact with 35 S-␤ 4 . As Fig. 2B shows, removal of the most carboxyl-terminal amino acids, or of the 42 most amino-terminal, does not abolish the capacity to interact with ␤ 4 . Concomitantly, fusion proteins corresponding to the most carboxylterminal region, which is most highly conserved among ␣ 1 subtypes, are incapable of binding. In addition, the interaction between GST-NT A and 35 S-␤ 4 was not inhibited by addition to the binding reaction of a peptide (500 M) corresponding to amino acids 76 -98. These data suggest that the ␤ 4 binding site concerns a region between residues 1 and 66 of ␣ 1A , maybe comprising, but not necessarily limited to, residues 42-52. The reduced binding to NT A ,2-52 and NT A ,42-77 compared with full-length NT A probably reflects instability and/or sequence reduction of the interaction site. The reduction in binding of smaller deleted derivatives meant that we were unable to pursue this approach further. A sequence alignment of this ␣ 1A binding domain with equivalent domains of other ␣ 1 subunits (␣ 1B , ␣ 1E , ␣ 1C , ␣ 1D , and ␣ 1S ), some used in this investigation, suggests a relatively low level of sequence conservation, although ␣ 1B and ␣ 1A show some similarity (Fig. 2C). This observation implies that the interaction may not be conserved, a prediction that we went on to test (see Fig. 5).
To identify the region of ␤ 4 which interacts with the aminoterminal region of ␣ 1A , we initially analyzed the binding capacity of several deleted derivatives of ␤ 4 , translated in vitro (Fig.  3, A and B). These derivatives lacked either the amino-terminal, carboxyl-terminal, or both regions, which shows a low level of conservation among ␤ subunit subtypes. As Fig. 3B shows, removal of the amino-terminal region had no effect, whereas removal of the carboxyl-terminal abolished binding completely, illustrating the importance of this region in the interaction. We also found that although ␤ 3 does not interact with GST-NT A (see Fig. 5), the opposite is true for a ␤ 3 -␤ 4 chimera, in which the nonconserved carboxyl terminus of ␤ 3 is replaced by the equivalent domain of ␤ 4 (Fig. 3B).
We have shown previously (21) that the carboxyl-terminal region of ␤ 4 also interacts with the carboxyl-terminal cytoplasmic domain of ␣ 1A (BI-2). We therefore wanted to map the two interaction sites more precisely, for which we constructed two additional derivatives of ␤ 4 , lacking a third (residues 483-519) and two-thirds (residues 447-519) of the carboxyl terminus (Fig. 3, A and B). As Fig. 3C shows, deletion of residues 483-519 of ␤ 4 had no effect on its capacity to bind to GST-NT A , whereas truncation of the carboxyl terminus of ␤ 4 up to residue 446 resulted in a total loss of binding capacity. This indicates that the NT A binding region is located between residues 446 and 482 of ␤ 4 . Analysis of the capacity of these truncates to bind to a GST fusion protein of the carboxyl-terminal region (residues 2090 -2424) of ␣ 1A (GST-CT A ) resulted in binding capacity being gradually lost with each further deletion. This suggests that the binding site of CT A spans a wider region than the NT A binding site, is dependent on secondary or tertiary structures that are disrupted by the deletions, or consists of a series of dispersed sites. It is noteworthy that the previously characterized ␣ 1A carboxyl-terminal binding site was also difficult to define, in that deleted derivatives over a long region retained binding capacity, giving support to the hypothesis that there are microdomains of interaction between these two sites (21). In contrast, the NT A site and corresponding domain on ␤ 4 are shorter and seem more easily delineated. In any case, the different patterns of interaction capacities seen for GST-NT A and GST-CT A suggest that these two regions of ␣ 1A occupy different but overlapping sites on ␤ 4 .
The involvement of overlapping regions of ␤ 4 in interactions with the amino-and carboxyl-terminal domains of ␣ 1A also raised the question as to whether these interactions could occur simultaneously or whether they were mutually exclusive. To investigate this as well as their relationship with the AID-BID interaction, we tested whether the binding of AID A (21-amino acid peptide) or GST-CT A to 35 S-␤ 4 could prevent its interaction with GST-NT A . The results, illustrated in Fig. 4, show that although the AID peptide was effective in preventing the interaction of ␤ 4 to GST-AID A , it did not prevent the concomitant interaction with either GST-NT A or GST-CT A,2070 -2275 (Fig.  4A). On the other hand, the association of MBP-CT A,2120 -2275 with ␤ 4 blocked the ability of ␤ 4 to interact with GST-CT A,2070 -2275 and also significantly reduced the binding of ␤ 4 to GST-NT A (Fig. 4B), suggesting that ␤ 4 is able to interact with AID and only one of the secondary interaction sites at a time.
We have shown previously (21) that ␤ subtypes differ in their capacity to interact with the carboxyl-terminal region of ␣ 1A , with ␤ 4 interacting with greatest affinity, ␤ 2A with a lesser affinity, and ␤ 1b and ␤ 3 showing no significant interaction. We therefore wished to determine whether the same was true for interaction with the amino-terminal domain. In addition, because the ␤ interaction site in the amino-terminal region of ␣ 1A shows a variable level of conservation among ␣ 1 subtypes, we wished to investigate whether ␤ interaction capacities were conserved among them. Both of these questions were addressed by constructing a series of GST fusion proteins carrying the amino-terminal cytoplasmic region of ␣ 1B , ␣ 1C , and ␣ 1S (Fig.  5A). These fusion proteins, along with GST alone, GST-AID A (for comparison purposes), and GST-NT A , were assayed for their ability to interact with four different ␤ subtypes, translated in vitro in presence of [ 35 S]methionine (Fig. 5B). Interestingly, interaction with GST-NT A showed a pattern similar to that observed for the carboxyl-terminal region of ␣ 1A (21) in that ␤ 4 exhibited the most significant interaction, ␤ 2a interacted to a lesser degree, and ␤ 1b and ␤ 3 showed no significant interaction. The amino-terminal domains of ␣ 1B showed no significant interaction despite its closer sequence relatedness to ␣ 1A . GST-NT S , on the other hand, showed significant interaction with all four ␤ subunits, whereas GST-NT C , another L-type channel member, showed no interaction with any of the ␤ subunits.
Because the amino-terminal sequences of ␣ 1A and ␣ 1S are very different, we checked whether binding of ␤ 4 to NT S involved the same interaction domain of ␤ 4 . Fig. 6 demonstrates that, as for NT A , the carboxyl terminus of ␤ 4 was required for binding to NT S , and the use of deleted derivatives of the carboxyl terminus of ␤ 4 also indicates an important role for residues 446 -482 of ␤ 4 in this interaction. These results suggest that the interaction site is defined more by the tertiary structure of the ␣ 1 amino-terminal region than by its primary sequence, also explaining why the NT A site could not be localized more precisely than to residues 1-66 (Fig. 2).
Finally, we questioned the relevance of the interaction between the amino terminus of ␣ 1A and the carboxyl terminus of ␤ 4 in terms of channel functioning. First, because ␤ 3 , in contrast to ␤ 4 , does not interact with the amino terminus of ␣ 1A , we investigated whether there were significant differences in terms of channel regulation by these two subunits. We found that in addition to triggering different inactivation kinetic behaviors (6), the two subunits differed in terms of their ability to shift the activation curve toward hyperpolarized potentials (Fig. 7A). Although both ␤ subunits shifted the activation curve along the voltage axis, the shift induced by ␤ 3 was significantly more pronounced than the one produced by ␤ 4 . The estimated half-activation potential shifted from 17 mV (␣ 1A -expressing oocytes) toward Ϫ13 mV (␣ 1A ␤ 3 oocytes) and 1.5 mV (␣ 1A ␤ 4 oocytes). There is thus an approximately 14 -15 mV difference in the shift induced by the ␤ 3 and ␤ 4 subunits. In addition, we found that depending on the ␤ subunit being expressed, the channels differed in their voltage dependence of inactivation with half-inactivation at Ϫ50 and Ϫ37 mV for ␣ 1A ␤ 3 and ␣ 1A ␤ 4 channels, respectively (data not shown). Because these differences in functional regulation by the various ␤ subunits may be the result of differences in interaction levels between ␣ 1A and the two ␤ subunits, we determined the role of the NT A site in ␤-induced channel regulation. We took advantage of the observation that essential differences were found in ␤ subunit association with amino-terminal sequences of various ␣ 1 subtypes. We constructed a chimera ␣ 1A subunit (␣ 1A (NT) C ), in which we replaced the amino terminus of ␣ 1A (interacts with ␤ 4 but not ␤ 3 ) with the amino terminus of ␣ 1C (does not interact with either ␤ 4 or ␤ 3 ). Coexpression of this chimeric channel with ␤ 3 or ␤ 4 triggers high voltage-activated currents in Xenopus oocytes (Fig. 7B). The amplitude of the currents elicited by membrane depolarization are reduced slightly compared with those obtained for the wild-type ␣ 1A channel. Cells expressing ␣ 1A ␤ 3 , for instance, have a peak current amplitude of 1,001 Ϯ 651 nA (n ϭ 7, S.D.), whereas cells expressing ␣ 1A (NT) C ␤ 3 peak at 423 Ϯ 655 nA (n ϭ 12), which corresponds to a 2.37-fold  6. The carboxyl terminus of ␤ 4 is also involved in NT S binding. 35 S-␤ 4 and deleted derivatives were assayed for their capacity to interact with GST-NT S (5 M). Specific binding was calculated by subtraction of binding to GST (at the same concentration) and normalized by expression as a percentage of maximal binding to GST-AID A (500 nM). Error bars represent normalized S.D. reduction. A similar 2.06-fold reduction in current amplitude is seen when ␤ 4 is coexpressed with ␣ 1A (NT) C (peak current 542 Ϯ 240 nA, n ϭ 12) rather than ␣ 1A (peak current 1,118 Ϯ 871 nA, n ϭ 6). These results suggest that the amino terminus plays a role in channel expression levels at the plasma membrane but that ␤ subunits and the NT A interaction site have little influence on this process. Also, the amino-terminal substitution induced an important shift in the voltage dependence of inactivation with half-inactivation occurring at Ϫ52 mV for ␣ 1A (NT) C ␤ 4 channels compared with Ϫ37 mV for ␣ 1A ␤ 4 channels (data not shown). Because a similar shift is seen with ␤ 3 (not shown), this supports the idea that ␤ subunit interaction with the amino terminus plays a minor role in this modification. In contrast, we found that the difference in the shift of voltage dependence of activation of ␣ 1A (NT) C ␤ 3 and ␣ 1A (NT) C ␤ 4 channels was reduced significantly (Fig. 7B). The average halfactivation potential of ␣ 1A (NT) C ␤ 3 channels was Ϫ13 mV and thus remained identical to that of the ␣ 1A ␤ 3 channels, whereas the V 1/2 of ␣ 1A (NT) C ␤ 4 channels was Ϫ9 mV, a significant hyperpolarizing shift compared with the ␣ 1A ␤ 4 channels. These data suggest that in the absence of an NT A /␤ 4 interaction, the I-V shift induced by the ␤ 4 subunit resembles the shift induced by the ␤ 3 subunit. Finally, the substitution of the ␣ 1A NT A sequence by NT C produced a slowing of channel inactivation with ␤ 4 but not with ␤ 3 . The decay of ␣ 1A (NT) C ␤ 4 currents occurred along two components with time constants of 1 ϭ 80 Ϯ 4 ms and 2 ϭ 368 Ϯ 39 ms (at 10 mV, n ϭ 7) compared with 1 ϭ 51 Ϯ 9 ms and 2 ϭ 246 Ϯ 27 ms (n ϭ 6) for ␣ 1A ␤ 4 currents. In contrast, no significant differences were seen in inactivation kinetics of ␣ 1A ␤ 3 or ␣ 1A (NT) C ␤ 3 channels with time constants at 10 mV of 1 ϭ 65 Ϯ 15 ms and 2 ϭ 243 Ϯ 47 ms (n ϭ 10) for ␣ 1A (NT) C ␤ 3 currents and 1 ϭ 62 Ϯ 14 ms and 2 ϭ 222 Ϯ 13 ms (n ϭ 7) for ␣ 1A ␤ 3 currents. These data further confirm a functional role in inactivation kinetics of the carboxyl terminus of ␤ 4 by its interaction with the carboxyl terminus (21) and amino terminus of the ␣ 1A subunit. DISCUSSION We describe the identification of a specific interaction site between the amino-terminal cytoplasmic region of the calcium channel ␣ 1A subunit and the ␤ 4 subunit. The ␤ 4 subunit is widely expressed in the brain, especially in the cerebellum (30). )/k)), where g is the normalized conductance (g ϭ 0.032, no ␤; 0.018, ϩ␤ 3 ; and 0.026, ϩ␤ 4 ); V 1/2 is the half-activation potential (V 1/2 ϭ 17 mV, no ␤; Ϫ13 mV, ϩ␤ 3 ; and 1. 5 mV, ϩ␤ 4 ); E is the reversal potential (E ϭ 67 mV, no ␤; 63 mV, ϩ␤ 3 ; and 58 mV, ϩ␤ 4 ); and k is the range of potential for an e-fold change around V 1/2 (k ϭ 7. 9 mV, no ␤; 4. 2 mV, ϩ␤ 3 ; and 5. 8 mV, ϩ␤ 4 ). B, change in difference in the ␤-induced I-V shift by ␣ 1A amino-terminal sequence substitution. Left and center, currents elicited by various membrane depolarizations (Ϫ30, Ϫ20, Ϫ10, and 0 mV) showing the absence of a difference in channel activation for ␣ 1A (NT) C ␤ 3 and ␣ 1A (NT) C ␤ 4 channels. Right, corresponding average I-V curves for ␣ 1A (NT) C ␤ 3 (n ϭ 13) and ␣ 1A (NT) C ␤ 4 channels (n ϭ 12). The fit of the experimental data yields V 1/2 ϭ Ϫ13 (ϩ␤ 3 ) and Ϫ9 mV (ϩ␤ 4 ); k ϭ 4. 2 (ϩ␤ 3 ) and 4. 4 mV (ϩ␤ 4 ); g ϭ 0. 017 (ϩ␤ 3 ) and 0. 018 (ϩ␤ 4 ); and E ϭ 60 (ϩ␤ 3 ) and 63 mV (ϩ␤ 4 ).
On the basis of their colocalization in many tissue types, ␤ 4 appears largely to be associated with the ␣ 1A subunit in native channels (17,31). However, coimmunoprecipitation studies demonstrate that ␣ 1A is also found associated with ␤ 1b , ␤ 2 , and ␤ 3 (17) and that ␤ 4 is also found associated with ␣ 1B (16). The importance of the ␤ 4 subunit is illustrated by the recent demonstration that a lethargic phenotype in mice results from a deletion of approximately 60% of the ␤ 4 coding sequence (32). This truncated ␤ subunit would lack all three of the interactions described with ␣ 1A (AID A (19) and NT A and CT A (21)), although such a deletion is also likely to result in severe conformational perturbation and probably degradation of the protein. This mutation is not entirely lethal, however, which is reminiscent of a growing number of experiments in which knockout of proteins of central importance does not turn out to be lethal. This is probably explained by a partial compensation by a related protein, in this case suggesting that other ␤ subunits are expressed in parallel or that their expression is switched on to compensate for this deficiency (33). In fact, ␤ 3 is known to be a normal constituent of about one-third of P/Q-type channels (17). Because ␤ 3 expression is high in brain and parallels that of ␤ 4 (34), it would be the most likely candidate for ␤ 4 substitution in the lethargic mice. Since ␤ 3 lacks both secondary interaction sites described so far in ␤ 4 , such a substitution would not be functionally equivalent, perhaps explaining some of the neurological defects encountered in these mice.
The NT A interaction identified is of relatively low affinity, supporting the idea that this is one of several secondary interactions between the two subunits that rely on the initial, high affinity interaction between the AID and BID sites identified previously. This idea is supported by the observation that mutagenesis of AID or BID to disrupt interaction between the two sites also disrupts the ability of the ␤ subunit to modify channel properties (10). It also stems from the fact that this is the third interaction site mapped between ␣ 1A and ␤ 4 and that binding of multiple ␤ subunits to ␣ 1 does not seem very plausible. The new interaction site that we describe involves the amino terminus of ␣ 1A (residues 1-66) and carboxyl terminus (residues 446 -482) of ␤ 4 . This is particularly interesting given the rather low level of sequence conservation in the two regions identified. With regard to ␣ 1A splice variants, the sequence of the amino-terminal cytoplasmic region is identical in BI-1 and BI-2 subtypes, indicating conservation of this interaction (24). This is in contrast to the ␤ 4 interaction site that we have identified previously in the carboxyl-terminal region of the BI-2 splice variant.
The low degree of sequence conservation observed for the respective interaction sites identified in the amino-terminal region of ␣ 1A and the carboxyl-terminal of ␤ 4 is reflected by the high degree of subtype specificity exhibited by this interaction with respect to both ␣ 1 and ␤ isoforms. Our results indicate that the equivalent amino-terminal regions of ␣ 1B and ␣ 1C did not interact with any of the ␤ subunits tested. Because other ␤ subtypes exist, we cannot rule out that this may reflect the use of an inappropriate ␣ 1 -␤ combination. Interestingly, we found that a fusion protein expressing the entire amino terminus of ␣ 1S could interact with all four different ␤ subunits tested. This is in contrast to the NT A binding, which occurs only on ␤ 4 and to a lesser extent on ␤ 2a . These results are indicative of a potential interaction of ␣ 1S with ␤ subunits other than ␤ 1a , the major ␤ subunit of skeletal muscle, and parallel recent findings that ␤ 3 (7, 32) and a ␤ 1 splice variant other than ␤ 1a (2) are also expressed, albeit at low levels, in skeletal muscle. Overall, our results are indicative of evolution to provide for ␣ 1 -␤ interaction specificity both within the ␣ 1 amino-terminal and the ␤ carboxyl-terminal sequences. Fig. 8 summarizes what is now known about ␣ 1A -␤ 4 interactions in terms of structure. One interesting aspect is that the ␤ 4 subunit can interact simultaneously with AID and, via its carboxyl-terminal region, with either the amino-or carboxylterminal regions of ␣ 1A , thereby defining two patterns of interactions. These interactions probably impose conformational constraints on the molecule which appear to affect channel function. It is also tempting to speculate that the conformational constraints are different depending on the patterns of interaction in use by the channel. The importance of the amino and carboxyl termini of ␣ 1A are underlined by the observation that truncations of equivalent domains in ␣ 1C result in enhanced current levels of the channel (35,36). These enhanced current levels occur either by a greater membrane incorporation (amino terminus) or enhanced open probability (carboxyl terminus). Because ␤ subunits also increase channel expression, and this effect varies in amplitude depending on the ␣ 1 and ␤ subtype studied, it is tempting to speculate that the secondary interaction sites described so far also intervene in ␣ 1A channel expression by one of the two mechanisms described for ␣ 1C . We did indeed find that substitution of the amino terminus of ␣ 1A by the equivalent sequence of ␣ 1C resulted in an important reduction in current density. This effect was, however, ␤ subtype-independent, and it is therefore unlikely that the NT A interaction site described here plays a role in ␤-induced enhancement of current amplitude. Despite this, secondary interactions appear to play other roles in several aspects of control of channel activity. We have shown previously (21) the importance of the carboxyl-terminal region of ␣ 1A in the control of channel inactivation kinetics. Here, we demonstrate that the amino-terminal interaction site of ␣ 1A is required for fine tuning the voltage dependence of activation. The NT A interaction with ␤ 4 appears to limit the amplitude of the hyperpolarizing ␤-induced shift of channel activation. By this unique mechanism, it can be predicted that the ␤ 3 -containing P/Q channel subtype is activated more easily than the ␤ 4 -containing P/Q channel subtype. In addition, secondary interaction sites may serve to protect or uncover phosphorylation sites in the ␣ 1A subunit, thereby altering the regulatory input of these. Another obvious possibility is that they play a role in the antagonistic relationship between the ␤ subunit and G␤␥ complex. In this respect, it is interesting that Qin et al. (23) have recently shown that, in addition to interacting with a region overlapping with the AID site (37,38), G␤␥ also interacts with the carboxyl-terminal domain of ␣ 1A , ␣ 1B , and ␣ 1E and that the amino terminus has recently been recognized as another determinant for G␤␥ regulation in ␣ 1E subunits (39). Finally, the existence of secondary interactions in addition to the AID-BID interaction could serve to favor certain combinations of subunits in cells where several subtypes are expressed. Given that ␤ subunits also play a role in the surface targeting of ␣ 1 and ␣ 2 ␦ (14), an interesting possibility is that specific ␤ subunits serve to target ␣ 1 subunits to specific regions of the cell surface.