Opposite Effects of a Single IIIS5 Mutation on Phenylalkylamine and Dihydropyridine Interaction with L-type Ca2+ Channels*

Replacement of L-type Ca2+ channel α1 subunit residue Thr-1066 in segment IIIS5 by a tyrosine residue conserved in the corresponding positions of non-L-type Ca2+ channels eliminates high dihydropyridine sensitivity through a steric mechanism. To determine the effects of this mutation on phenylalkylamine interaction, we exploited the availability of Cav1.2DHP–/– mice containing the T1066Y mutation. In contrast to dihydropyridines, increased protein-dependent binding of the phenylalkylamine (–)-[3H]devapamil occurred to Cav1.2DHP–/– mouse brain microsomes. This effect could be attributed to an at least 2-fold increase in affinity as determined by saturation analysis and binding inhibition experiments. The latter also revealed a higher affinity for (–)-verapamil but not for (–)-gallopamil. The mutation caused a pronounced slowing of (–)-[3H]devapamil dissociation, indicating a stabilization of the drug-channel complex. The increased affinity of mutant channels was also evident in functional studies after heterologous expression of wild type and T1066Y channels in Xenopus laevis oocytes. 100 μm (–)-verapamil inhibited a significantly larger fraction of Ba2+ inward current through mutant than through WT channels. Our results provide evidence that phenylalkylamines also interact with the IIIS5 helix and that the geometry of the IIIS5 helix affects the access and/or binding of different chemical classes of Ca2+ channel blockers to their overlapping binding domains. Mutation of Thr-1066 to a non-L-type tyrosine residue can be exploited to differentially affect phenylalkylamine and dihydropyridine binding to L-type Ca2+ channels.

Replacement of L-type Ca 2؉ channel ␣ 1 subunit residue Thr-1066 in segment IIIS5 by a tyrosine residue conserved in the corresponding positions of non-L-type Ca 2؉ channels eliminates high dihydropyridine sensitivity through a steric mechanism. To determine the effects of this mutation on phenylalkylamine interaction, we exploited the availability of Ca v 1.2DHP ؊/؊ mice containing the T1066Y mutation. In contrast to dihydropyridines, increased protein-dependent binding of the phenylalkylamine (؊)-[ 3 H]devapamil occurred to Ca v 1.2DHP ؊/؊ mouse brain microsomes. This effect could be attributed to an at least 2-fold increase in affinity as determined by saturation analysis and binding inhibition experiments. The latter also revealed a higher affinity for (؊)-verapamil but not for (؊)-gallopamil. The mutation caused a pronounced slowing of (؊)-[ 3 H]devapamil dissociation, indicating a stabilization of the drug-channel complex. The increased affinity of mutant channels was also evident in functional studies after heterologous expression of wild type and T1066Y channels in Xenopus laevis oocytes. 100 M (؊)verapamil inhibited a significantly larger fraction of Ba 2؉ inward current through mutant than through WT channels. Our results provide evidence that phenylalkylamines also interact with the IIIS5 helix and that the geometry of the IIIS5 helix affects the access and/or binding of different chemical classes of Ca 2؉ channel blockers to their overlapping binding domains. Mutation of Thr-1066 to a non-L-type tyrosine residue can be exploited to differentially affect phenylalkylamine and dihydropyridine binding to L-type Ca 2؉ channels.
Voltage-gated L-type Ca 2ϩ channels (LTCCs) 1 are characterized by their high sensitivity toward Ca 2ϩ channel blockers (CCBs; Ca 2ϩ antagonists). High CCB sensitivity is encoded in their pore-forming ␣ 1 -subunits (Ca v 1.1-Ca v 1.4 ␣ 1 ), which contain a number of unique amino acid residues critical for the formation of a binding pocket for different chemical classes of CCBs. Clinically relevant are 1,4-dihydropyridines (DHPs; e.g. isradipine and nifedipine), phenylalkylamines (PAAs; e.g. verapamil and gallopamil) and benzothiazepines (BTZs; e.g. (ϩ)-cisdiltiazem). These chemical classes affect each other's binding to the ␣ 1 -subunit through noncompetitive interactions (1). A detailed molecular analysis of the distinct drug binding domains was performed, which revealed that DHP, PAA, and BTZ binding residues are located within the same regions on ␣ 1 subunits and that some residues even participate in the binding of more than one class of CCBs (1)(2)(3). This suggested that all three binding domains are located in close proximity to each other and even overlap (Fig. 1). Studies with fluorescently labeled DHPs and BTZs supported this model and provided experimental evidence for a steric interference of bound DHPs with the accession pathway for BTZs (4). A multisubsite domain binding model was therefore proposed in which the apparent noncompetitive interactions observed between different classes of CCBs mainly result from steric interactions rather than druginduced conformational changes (1). This multisubsite binding domain is located in the pore-forming regions of repeats III and IV of the ␣ 1 -subunit. It allows drug binding to a domain interface, thus facilitating stabilization of closed channel conformations ("domain interface model") (2).
Detailed molecular analysis of amino acid residues involved in drug binding is available for all three chemical classes of CCBs. Altered sensitivity of currents through mutated channels was tested by electrophysiological means after expression of the channel complex in heterologous systems. However, this analysis is complicated by the fact that the apparent sensitivity of CCBs is critically affected by channel gating in an either voltage-or use-dependent manner (5,6). Mutation-induced changes in channel gating can therefore lead to pronounced alterations in drug sensitivity without necessarily changing binding affinity (6). Radioreceptor assays with recombinant channels directly allow us to measure changes in binding affinity (i.e. the dissociation constant, K D ) and are therefore a valuable addition to confirm electrophysiological findings. In contrast to the successful application of this approach for DHPs, PAA radioligand binding to recombinant LTCCs has not yet been accomplished. This is not only due to their lower affinity for LTCCs as compared with DHPs (7) but also to higher nonspecific binding of suitable radioligands ((Ϫ)-[ 3 H]devapamil or [ 3 H]verapamil) to low affinity binding sites on other proteins (8,9).
In this paper, we exploited a new mouse model in which Thr-1066 in transmembrane segment IIIS5 was mutated to a tyrosine residue (T1066Y) in Ca v 1.2 ␣ 1 , the major LTCC isoform expressed in heart and brain (10). The mutation serves as a molecular switch that eliminates high DHP sensitivity by steric hindrance of DHP binding (11). These mutant mice pro-vide us with the unique opportunity to directly quantify the consequences of this mutation by radioligand binding in native channels of mouse brain membranes. Here we report that Tyr-1066 unexpectedly increased PAA affinity by stabilization of the PAA-channel complex. This clearly shows that PAAs can also interact with amino acids in repeat IIIS5 and provides an unequivocal example for a residue that is able to control both DHP and PAA interaction. Since a Tyr residue is present in the corresponding position of non-LTCC ␣ 1 subunits, this residue must confer some of the known sensitivity of these channels for PAAs.
Animals-The generation and breeding of Ca v 1.2DHP Ϫ/Ϫ mice was described recently (10) and was approved by the Austrian Bundesministerium fü r Bildung, Wissenschaft, und Kultur. To minimize use and breeding of genetically modified animals, the number of experiments had to be kept to a minimum.
Membrane Preparation and Immunoblot Analysis-Brain membranes and membranes from tsA201 cells transfected with 4.5 g of ␣ 1 , 3.5 g of ␣ 2 -␦, 2.5 g of ␤, and 4.5 g of pUC carrier DNA were prepared as described (12). Membrane protein concentrations were determined according to Lowry (13), using bovine serum albumin as a standard. Expression of WT and mutant Ca v 1.2 ␣ 1 (␣ 1 C) subunits was quantified in immunoblots employing affinity-purified anti-␣ 1 C 818 -835 antibody as described previously (14). Prestained molecular mass standards (high range; Bio-Rad) separated on the same gel were used to calculate the apparent molecular masses of immunoreactive bands.
Radioreceptor Assay-Binding experiments with the phenylalkylamine (Ϫ)-[ 3 H]devapamil were performed in 50 mM Tris-HCl buffer, pH 7.4, supplemented with 0.1 mM phenylmethyl sulfonylfluoride and 0.2 mg/ml bovine serum albumin in a final assay volume of 0.5-1 ml, employing membrane protein and radioligand concentrations as indicated in the figure legends. Nonspecific binding was determined in the presence of 3 M unlabeled racemic devapamil and subtracted from total binding to yield specific binding. Binding inhibition (and stimulation) studies were performed in the absence (control) and presence of increasing concentrations of unlabeled drugs serially diluted in Me 2 SO. The maximal Me 2 SO concentration in the assay did not exceed 1% (v/v), which does not affect DHP and PAA binding. For dissociation kinetic experiments, dissociation of the preformed drug-channel complex was monitored by the addition of 3 M unlabeled devapamil at different time intervals before determining the concentration of bound complexes. The concentration of bound ligand was determined by rapid filtration of the assay mixture through GF/C Whatman glass fiber filters, pretreated with 0.25% (v/v) polyethyleneimine and 0.25 mg/ml bovine serum albumin for 30 min at 22°C (15). Filters were washed three times with ice-cold filtration buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl), and filter-bound radioactivity was determined by liquid scintillation counting (16). Binding inhibition (and stimulation) data were fitted to the general dose-response equation (17) to obtain IC 50 values and slope factors. Biphasic saturation data were subjected to Scatchard transformation. Mutant-induced changes of a high affinity binding site were estimated from the initial slopes of the Scatchard analysis by linear regression analysis of the five or six data points with the lowest free radioligand concentration (see "Results" and corresponding figure legends). Dissociation rate constants (k Ϫ1 ) were determined by nonlinear fitting of the data to equations describing mono-or biexponential dissociation reactions as appropriate. Binding parameters were calculated using GraphPad Prism 4.0®.
Electrophysiological Measurements in X. laevis Oocytes-Capped run-off poly(A ϩ ) cRNA transcripts from XbaI-linearized cDNA templates of WT and the T1066Y mutant were synthesized according to Krieg and Melton (20). ␣ 1 cRNA was coinjected with ␤ 1 a (21) and ␣ 2 ␦ (22) subunit cRNAs into stage V-VI oocytes from X. laevis. To avoid bias, the experimenter was blinded to the injected oocytes until data analysis was finished.
2-3 days after cRNA injection, inward barium currents (I Ba ) through voltage-gated Ca 2ϩ channels were measured at room temperature using the two-microelectrode voltage clamp technique as described previously (11). To quantify endogenous I Ba , X. laevis oocytes injected only with ␤ 1 a and ␣ 2 ␦ were analyzed in parallel. Only oocytes expressing peak I Ba through recombinant Ca 2ϩ channels at least 7 times larger than the highest endogenous currents were included in the analysis. Data analysis and acquisition was performed using the pClamp software package (version 6.0; Axon Instruments). Leakage correction was performed by adjusting the current traces by a factor calculated from the difference between the leak at Ϫ80 mV and Ϫ90 mV, respectively. The extracellular solution contained 40 mM Ba(OH) 2 , 50 mM NaOH, 2 mM CsOH, and 5 mM HEPES (pH adjusted to 7.4 with methanesulfonic acid). The voltage recording and current-injecting microelectrodes were filled with 2.8 M CsCl, 0.2 M CsOH, 10 mM HEPES, and 10 mM EGTA (adjusted to pH 7.4 with HCl) and had resistances of 0.7-5 megaohms.
Before starting an experiment, the oocytes were held at Ϫ80 mV until I Ba stabilized. Use-dependent modulation of peak I Ba by 100 M (Ϫ)-verapamil was measured during 20 consecutive 100-ms pulses applied at a frequency of 0.66 Hz from a holding potential of Ϫ80 mV to a test pulse corresponding to the peak potential of the current-voltage relations. Control traces were recorded twice in the absence of drug separated by a 5-min interval to allow for complete channel recovery. The same protocol was then repeated after perfusion of the oocyte with drugcontaining solution at the holding potential. Use-dependent block was expressed as the percentage decrease of peak I Ba during the last test pulses of the train compared with I Ba during the first train. Only cells that showed peak current rundown between control test protocols less than 15% and current decay during trains of pulses in the absence of drug of less then 10% of the peak current were included in the analysis.
Statistics-Data are given as mean Ϯ S.E. for the indicated number of experiments. Statistical significance was determined by an unpaired Student's t test employing GraphPad Prism 4.0®.  Fig. 1 illustrates the location of residue Thr-1066 (numbering according to Ca v 1.2) (10) in the putative folding structure of LTCC ␣ 1 -subunits. Mutation to tyrosine eliminates high affinity for DHPs as assessed by radioligand binding and functional studies (3,10).
The observation of a higher specific (Ϫ)-[ 3 H]devapamil binding signal for Ca v 1.2DHP ϩ/Ϫ and Ca v 1.2DHP Ϫ/Ϫ mice raised the question of whether this effect was due to an increase in binding affinity or resulted from a mutation-induced up-regulation of Ca v 1.2 channels. We considered the latter possibility unlikely, because we found no evidence of increased immunoreactivity for mutated Ca v 1.2 ␣ 1 in brain and heart membranes of Ca v 1.2DHP Ϫ/Ϫ mice in an earlier study (10); nor did we detect increased expression of recombinant mutant T1066Y channel complexes after heterologous expression in tsA201 cells (Fig. 3A). Furthermore, the mutation did not alter the neuronal expression pattern of Ca v 1.2 mRNAs as determined by in situ hybridization experiments in whole brain sections (Fig. 3B). Thus, it appeared more likely that the mutation enhanced binding by increasing (Ϫ)-devapamil binding affinity. Fig. 4 illustrates the inhibition of specific (Ϫ)-[ 3 H]devapamil binding to WT and mutant brain LTCCs by the more potent (Ϫ)-enantiomers of devapamil, verapamil, and gallopamil. These PAAs only differ with respect to the number of methoxy substituents at their phenyl rings (7). For (Ϫ)-devapamil and (Ϫ)-verapamil a reproducible decrease of the IC 50 values was observed in three independent experiments (see the legend to Fig. 4). This increase in affinity was also observed in the presence of a physiologically relevant (1 mM) concentration of CaCl 2 (see the legend to Fig. 4), which affects PAA interaction with the extracellularly oriented selectivity filter glutamates (2). In contrast, the IC 50 was not increased for (Ϫ)gallopamil (Fig. 4). These data confirmed the higher binding signal observed for (Ϫ)-[ 3 H]devapamil and suggested that mutation T1066Y increases the affinity of LTCCs for (Ϫ)-devapamil and (Ϫ)-verapamil.
To further substantiate this finding, we attempted to determine the K D for (Ϫ)-[ 3 H]devapamil in saturation experiments, although the low slope factor in displacement studies (Fig. 4A) suggested that it labeled more than one population of sites in mouse brain membranes. As shown in Fig. 5, this was indeed the case. Such low affinity sites for (Ϫ)-[ 3 H]devapamil have previously been reported in mammalian brain (23) and in car- diac muscle (8). Since the affinity for the low affinity binding component was too low to be reliably quantified, we derived estimates for the change in the K D for the high affinity site from the initial slopes of the Scatchard-transformed binding data (see "Experimental Procedures" and the legend to Fig. 5). The apparent K D (1/slope factor) decreased from 1.93 Ϯ 0.39 in WT to 0.53 Ϯ 0.07 (n ϭ 3; p Ͻ 0.05) in mutant membranes. Although these data do not rule out changes in maximal binding capacity, they confirm the higher affinity of (Ϫ)-[ 3 H]devapamil for the T1066Y mutant.
To reveal the mechanism by which the mutation decreased the K D for (Ϫ)-[ 3 H]devapamil, we investigated the dissociation kinetics of (Ϫ)-[ 3 H]devapamil (Fig. 6). Experiments were performed at low radioligand concentrations to preferentially label LTCCs and minimize occupancy of the low affinity sites detected in saturation studies. After equilibrium binding had been reached, we followed the dissociation of the radioligand by the addition of 3 M unlabeled devapamil to the reaction mixture. In five independent experiments, we reproducibly observed a significantly slower dissociation from mutant brain membranes as compared with WT (Fig. 6). After 20 min, almost complete dissociation of radioligand occurred from WT channel complexes, whereas more than 50% of specific binding to the mutant channel remained (Fig. 6A). Whereas monophasic dissociation of (Ϫ)-[ 3 H]devapamil from mutant channels occurred with a dissociation rate constant (k Ϫ1 ) of 0.0209 Ϯ 0.0023 min Ϫ1 (n ϭ 5), k Ϫ1 was 5-10-fold higher for WT (see legend to Fig. 6B). This suggests that the mutation stabilizes (Ϫ)-[ 3 H]devapamil binding to LTCCs in accordance with the observed decrease in K D .
To prove the increased PAA affinity also in functional studies, we expressed WT and mutant Ca v 1.2 ␣ 1 subunits in X. laevis oocytes together with ␤ 1 and ␣ 2 ␦ subunits and measured the inhibitory effect of 100 M (Ϫ)-verapamil on channel currents using Ba 2ϩ as a charge carrier. Verapamil was used, because our binding experiments indicated a slightly higher increase in binding affinity of the (Ϫ)-enantiomer for the mutant channel (Fig. 4), and only small amounts of (Ϫ)-devapamil were available for our studies. PAAs are use-dependent blockers of LTCCs. PAA sensitivity depends on depolarization frequency and pulse duration (24), making it difficult to assess small differences in affinity by generating concentration-response curves. We therefore selected a pulse protocol in which I Ba decay during pulse trains was minimal (Ͻ7%; see legend to Fig. 8), and thus even a small increase in use-dependent block by (Ϫ)-verapamil should be de-
tectable. This was achieved by applying trains of 20 consecutive 100-ms pulses at a frequency of 0.66 Hz from a holding potential of Ϫ80 mV to a test pulse corresponding to the peak potential of the current-voltage relations. Under these conditions, 100 M verapamil inhibited 23.7 Ϯ 2.56% (n ϭ 5) of I Ba in WT channels (Fig. 8). A significantly (p Ͻ 0.01) larger inhibition (42.94 Ϯ 3.70%; n ϭ 7) was induced in mutant channels. Although the complexity of the use-dependent verapamil inhibition does not allow us to quantitate the apparent sensitivity difference, these functional data nicely confirm our results obtained by radioligand binding studies. DISCUSSION By demonstrating that a tyrosine residue in position 1066 increases PAA affinity, we provide novel insight into the organization of the LTCC multisubsite drug binding domain. So far, no interactions of PAAs with segment IIIS5 have been demonstrated. We show that the bound PAA molecule is well within reach of IIIS5 amino acid side chains and that a tyrosine side chain (as present in non-L-type channel ␣ 1 -subunits) is able to interact both with PAAs and DHPs exerting opposite effects on binding affinity.
Our data also provide some indirect evidence concerning the orientation of the Tyr-1066 (numbering according to Ca v 1.2) side chain with respect to the drug binding pocket. One possibility is that it sterically blocks DHP binding by protruding deep into the multisubsite domain. In this case, we would also expect a steric interference with PAA binding based on the facts that (i) the PAAs examined here are larger than DHPs (such as isradipine) and (ii) the DHP and PAA binding domains must considerably overlap (1,2). Our data therefore suggest a model ( Fig. 9) in which the tyrosine side chain is oriented more toward the access pathway of DHPs and/or those regions of the multisubsite domain that are preferentially occupied by the DHP molecule. Opposite effects of this residue on DHP and PAA binding are also in agreement with previous electrophysiological studies, which used permanently charged drugs to demonstrate that DHPs and PAAs approach their binding domains from opposite sides of the channel. Whereas DHPs rely on an extracellular access pathway (25,26), PAAs block the channel through a cytoplasmic approach (24,27). In the proposed model, such separate accession pathways would allow the (also more flexible) PAA molecule to still reach its site inside the mutant channel and simultaneously provide a close interaction with the Tyr-1066 side chain, thereby stabilizing binding and decreasing K D . A stabilization of the PAA-channel complex was evident from our kinetic studies, which revealed a pronounced slowing of (Ϫ)-[ 3 H]devapamil dissociation from mutant brain membranes. The mutation may not only slow dissociation, but it seems to also slow association of (Ϫ)-devapamil, because the slowing of dissociation (5-10-fold) appeared larger than the increase in binding affinity.
Because we do not yet know how the S5 and pore-forming S6 helices are packed against each other, it is possible that the IIIS5 mutation affects preferentially the geometry of the binding site for DHPs. However, due to the strong overlap of the two binding domains, we consider this rather unlikely. S6 helices also control the gating and pore properties of the channel, and thus conformational changes may also alter the biophysical channel properties. However, we and others (see Ref. 10 and references cited therein) did not find evidence for gating changes of the mutant channel.
Note that our detailed analysis of the PAA interaction of In a second experiment, which allowed resolution of the fast dissociating component in WT membranes, a k Ϫ1 value of 0.107 min Ϫ1 was calculated. Monoexponential fits did not account for 4 -10% of the binding signal, indicating the possibility of slower dissociation from a small fraction of binding sites. Since this would not affect the interpretation of our data, this finding was not further investigated. The mean k Ϫ1 for mutant channels was 0.0209 Ϯ 0.0023 min Ϫ1 (n ϭ 5). For the illustrated experiment, total and nonspecific binding under equilibrium conditions were as follows (in dpm/0.5 ml): for WT, total ϭ 3486, nonspecific ϭ 1007; for mutant, total ϭ 5888, nonspecific ϭ 1119.
Our data also provide a molecular basis for the finding that N-and P/Q-type Ca 2ϩ channels exhibit some affinity for PAAs, such as devapamil or verapamil (29 -31), but not for DHPs. It is known that some drug binding residues of the PAA binding domain are conserved between LTCC and non-LTCC (e.g. Ca v 2.1 and Ca v 2.2) ␣ 1 subunits. Examples are Tyr-1151, Phe-1172, and Val-1173 in IIIS6. Like our mutant channels, non-LTCCs contain a tyrosine residue in the position corresponding to L-type Thr-1066. Our data provide evidence that this tyrosine together with the other conserved PAA binding residues mediates the weak PAA sensitivity of non-LTCCs.
Three-dimensional models of LTCC drug binding domains exist (12, 32) but are not yet suitable for rational drug design because they are based on the crystal structure of structurally related bacterial K ϩ -channels. Based on the rapid progress 2 I. Huber and J. Striessnig, unpublished observations.  (25,26) and intracellular side (24,27), respectively (left and middle), DHP binding and/or access is inhibited by the phenyl ring of tyrosine in position 1066 (right). At the same time, the tyrosine side chain increases PAA binding affinity and stability (right). made in elucidating the three-dimensional structure of large ion channels, it is likely that a higher resolution structure will soon also become available for voltage-gated Ca 2ϩ channels (33,34). In combination with detailed maps of the interaction sites of LTCC modulators with individual amino acids, this will not only allow the elucidation of the exact molecular mechanism for LTCC block by these drugs but eventually may also facilitate the rational development of selective blockers of other Ca 2ϩ channel ␣ 1 subunits or perhaps of other structurally related members of the cation channel family.