Roscovitine Binds to Novel L-channel (CaV1.2) Sites That Separately Affect Activation and Inactivation*

L-type (CaV1.2) calcium channel antagonists play an important role in the treatment of cardiovascular disease. (R)-Roscovitine, a trisubstituted purine, has been shown to inhibit L-currents by slowing activation and enhancing inactivation. This study utilized molecular and pharmacological approaches to determine whether these effects result from (R)-roscovitine binding to a single site. Using the S enantiomer, we find that (S)-roscovitine enhances inactivation without affecting activation, which suggests multiple sites. This was further supported in studies using chimeric channels comprised of N- and L-channel domains. Those chimeras containing L-channel domains I and IV showed (R)-roscovitine-induced slowed activation like that of wild type L-channels, whereas chimeric channels containing L-channel domain I responded to (R)-roscovitine with enhanced inactivation. We conclude that (R)-roscovitine binds to distinct sites on L-type channels to slow activation and enhance inactivation. These sites appear to be unique from other calcium channel antagonist sites that reside within domains III and IV and are thus novel sites that could be exploited for future drug development. Trisubstituted purines could become a new class of drugs for the treatment of diseases related to hyperfunction of L-type channels, such as Torsades de Pointes.

Cardiac L-type (Ca V 1.2) channels are central to the regulation of a number of physiological processes (1,2). Activation of these channels in cardiac and smooth muscle generates the calcium influx that triggers calcium release from the sarcoplasmic reticulum (3) to induce contraction. In contrast, inactivation of these channels provides a negative feedback mechanism to limit calcium influx into the cardiac myocyte, which helps protect from excessive calcium influx that can lead to cardiac arrhythmias (3). L-channel antagonists are routinely used to treat cardiovascular diseases, such as hypertension and angina pectoris (4 -8). Therefore, drugs that inhibit L-channel function have high clinical relevance.
Roscovitine is a 2,6,9-trisubstituted purine that was originally developed as a selective blocker of cyclin-dependent kinases (9) and is currently undergoing phase II clinical trials as an anticancer drug (10). It has recently become apparent that roscovitine can affect voltage-dependent ion channels at clinically relevant concentrations (10 -50 M) (10 -13). We (14,15) and others (16,17) have shown that (R)-roscovitine differentially affects voltage-dependent calcium channels. (R)-Roscovitine has two effects on Ca V 2 channels (N-type, P/Q-type, and R-type), which are a rapid onset agonist effect and a more slowly developing antagonist effect (14,15). The agonist effect results from (R)-roscovitine specifically binding to activated Ca V 2 channels to slow channel closing (14,15), which results in a significant enhancement of action potential induced calcium influx (14). The antagonist effect appears to result from (R)roscovitine preferentially enhancing occupancy of a "resting" inactivated state to inhibit channel activity (18). Interestingly, the racemic variant (S)-roscovitine has been found to exhibit only an antagonist effect on N-channels, which is one result supporting unique binding sites for the agonist versus antagonist effects (14,15,18).
L-type channels are also inhibited by (R)-roscovitine, but by a unique mechanism relative to N-channels (19). L-channel inhibition results from slowed activation and enhanced open state voltage-dependent inactivation (VDI), 3 but the resting inactivated state is not affected. These two effects were characterized by approximately equal EC 50 values (ϳ30 M), which suggested a single binding site. However, the Hill coefficient for (R)-roscovitine-induced slowed activation was ϳ1, whereas that for enhanced inactivation was Ͼ2, which could result from multiple binding sites (19). Intracellularly applied (R)-roscovitine failed to affect L-channel activity, which supported an extracellularly exposed binding site(s) mediating both effects. The differential effect of (R)-roscovitine on N-type versus L-type channels provides an opportunity to localize the L-channel binding site(s) by making chimeric channels. In addition, the differential effect of racemic roscovitine variants on N-channels led us to investigate if the L-channel site(s) also showed chiral specificity.
We found that (S)-roscovitine enhanced L-channel VDI without slowing activation, which supports separate binding sites for these effects. In addition, the chimeric channels showed that the L-channel domain I (L-DI) is both necessary and sufficient for the (R)-roscovitine-induced enhancement of VDI, but slowed activation requires both L-DI and L-DIV. These results reveal novel sites on the L-type calcium channel that can be exploited for the development of drugs that can specifically target the activation or inactivation function.

EXPERIMENTAL PROCEDURES
Construction of Chimeric Channels-Chimeric calcium channels ( Fig. 1) were constructed using cDNAs encoding the rabbit cardiac L-channel (GenBank TM number X15539) and rat neuronal N-channel (GenBank TM number AF055477; generously provided by Dr. Lucie Parent) ␣1 subunits. Briefly, N-channel domains were amplified by PCR, subcloned into pCR-Blunt II-TOPO vectors (Invitrogen), excised by restric-tion enzyme digestion, and subcloned in frame into an engineered L-channel. The sequences of the wild type (WT) and mutant channels were aligned using Vector NTI (Invitrogen), and domain boundaries were placed in regions of high amino acid sequence homology. The engineered L-channel was generated by introducing unique AgeI and NotI sites into the intracellular linkers between the II/III (nucleotide 2751) and III/IV (nucleotide 3680) domains, respectively, using silent mutagenesis via the QuikChange II site-directed mutagenesis kit (Stratagene, La Jolla, CA). A unique XbaI site was introduced at nucleotide 4633 in the C terminus of the engineered channel using a similar strategy. The overall integrity of the engineered L-channel and each chimera was confirmed by qualitative restriction enzyme digests, DNA sequence analyses, Western blot analysis, and patch clamp electrophysiology .
The chimeric channels were constructed as described below. For NNLL, L-DI and L-DII (Met 1 -Asn 917 ) were replaced by N-DI and N-DII (Met 1 -Asn 1134 ). N-DI and N-DII were amplified from full-length Ca V 2.2 and subcloned into the HindIII/ AgeI sites of the engineered L-channel. For LLNN, L-DIII and L-DIV (Leu 895 -Leu 2171 ) were replaced by N-DIII and N-DIV (Asp 1112 -Cys 2333 ). The construction of this chimera did not utilize the engineered L-channel containing the introduced AgeI and NotI sites. Instead, a XbaI site at nucleotide 2678 was introduced into the L-channel, resulting in a missense mutation of a glutamate to arginine at position 894. For NLLL, L-DI (Met 1 -Cys 519 ) was replaced by N-DI (Met 1 -Pro 447 ). The insert for this chimera was constructed via overlap extension PCR. Briefly, N-DI and L-DII were amplified from full-length Ca V 2.2 and Ca V 1.2, respectively. The PCR products of these two domains contain a small region of overlap. They were then combined and used as templates in a second round of PCR. The final PCR product was subcloned into the HindIII/AgeI sites of the engineered L-channel. For LNLL, L-DII (Ala 524 -Val 781 ) was replaced by N-DII (Ser 452 -Val 909 ). This chimera was constructed using a strategy similar to that described for NLLL above. For LLNL, N-DIII (Arg 1137 -Ala 1443 ) was amplified from full-length Ca V 2.2 and ligated into the AgeI/NotI sites of the engineered L-channel, thereby replacing L-DIII (Arg 920 -Ala 1226 ). For LLLN, N-DIV (Pro 1445 -Ile 1741 ) was amplified from full-length Ca V 2.2 and ligated into the NotI/BstEII sites of the engineered L-channel, thereby replacing L-DIV (Pro 1228 -Ile 1542 ). For LNNL, N-DII and N-DIII were amplified from Ca V 2.2 with a 5Ј primer containing an AscI site and a 3Ј primer containing a NotI site, excised with AscI/NotI, and subcloned into the AscI/Not sites of LLNL. For LNNN, N-DIV was excised with NotI/XbaI from LLLN and subcloned into the NotI/XbaI sites of LNNL. For NNNL, N-DI was excised with HindIII/AscI from NLLL and subcloned into the HindIII/AscI sites of LNNL. For NNLN, N-DIV was excised with NotI/XbaI from LLLN and subcloned into the NotI/XbaI sites of NNLL. For NLLN, N-DI was excised with HindIII/AscI from NLLL and subcloned into the HindIII/AscI sites of LLLN. For L(l/n)LL, L-DII transmembrane segments 1-4 were amplified with a 5Ј primer containing an AscI site and a 3Ј primer (5Ј-ACC AGG TTC CTC AGG GAG TTC CAG TAC CTT GTA ATT TTG-3Ј) using NLLL as template, and N-DII transmembrane segments 5 and 6 with N-type intracellular linker between transmembrane segments 4 FIGURE 1. The calcium channel constructs used in this study. Domains I-IV are shown with each domain represented as a set of six transmembrane segments, 1-6. Loops between the segments and domains are shown as lines. N-channel structures are shown in black, whereas those from the L-channel are in gray. The numbers at the connections between N-type and L-type domains show the number of amino acid residues within that intracellular loop contributed by a given channel type. To highlight the connection, the N-channel section is shown as a thicker line. For LN*LL, LN*NN, and LN*NL chimeras, N* refers to an engineered N-DII, which is composed of N-type transmembrane segments 1-4 and L-type transmembrane segments 5 and 6. The loop between segments 4 and 5 is from the L-channel. and 5 were amplified with a 5Ј primer (5Ј-ATT ACA AGG TAC TGG AAC TCC CTG AGG AAC CTG GTT G-3Ј) and a 3Ј primer lying downstream of an existing BfrI site in the WT ␣ 1C plasmid (pCDNA3). The final overlap PCR products were amplified using the above 5Ј primer containing an AscI site and the above 3Ј primer lying downstream of an existing BfrI site. The resulting PCR products were excised with AscI/BfrI and subcloned into the AscI/BfrI sites of LNLL to make L(l/n)LL. For LN*LL (also known as L(n/l)LL), N-DII transmembrane segments 1-4 were amplified with a 5Ј primer containing an AscI site and a 3Ј primer (5Ј-CAG GTT GCT CAA GGA GTT CCA ATA CTT GGT GAC TTT GAA AAT CCT CAG-3Ј) using LNLL as template, and L-DII transmembrane segments 5 and 6 with an L-type intracellular linker between transmembrane segments 4 and 5 were amplified with a 5Ј primer (5Ј-GTC ACC AAG TAT TGG AAC TCC TTG AGC AAC CTG GTG GCC-3Ј) and a 3Ј primer lying downstream of an existing BfrI site in WT ␣ 1C plasmid (pCDNA3). The final overlap PCR products were amplified using the above 5Ј primer containing an AscI site and the above 3Ј primer lying downstream of an existing BfrI site. The resulting PCR products were excised with AscI/BfrI and subcloned into the AscI/BfrI sites of LNLL to make L(n/l)LL. For LN*NN (also known as L(n/l)NN), first L-DIV in LLNL was replaced by N-DIV from LLLN to make a different LLNN maintaining the engineered unique AgeI site between the II and III domains, NotI sites between the III and IV domains, and XbaI site just after domain IV. Then the chimera domain II n/l was excised with AscI/BfrI from L(n/l)LL and subcloned into the AscI/BfrI sites of LLNN to make L(n/ l)NN. For LN*NL (also known as L(n/l)NL), the fragment containing the chimera domain II n/l and N-DIII was excised with AscI/NotI from L(n/l)NN and ligated into the AscI/NotI sites of LNNL to make L(n/l)NL. The mutation and integrity of the mutant cDNAs was confirmed by qualitative restriction map analysis and directional DNA sequence analysis of the entire subcloned region. Functional expression of the mutant cDNAs was confirmed by Western blot analysis and patch clamp electrophysiology.
HEK Cell Transfection-We utilized either the calcium phosphate precipitation method (19) or Lipofectamine 2000 (following the manufacturer's directions) to transfect HEK 293 cells with WT L-, N-, and chimeric channels, which provided highly reproducible expression 24 -72 h after transfection. HEK293 cells were maintained in standard Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and 1% antibiotic-antimycotic mixtures (regular medium) at 37°C in a 5% CO 2 incubator. HEK293 cells were transfected with cDNA plasmids using the following molar ratio: 1 ␣ 1 subunit (L-, N-, or chimeric channel):1 ␣ 2 ␦:1 ␤ 1b :1 TAG (to increase expression efficiency):0.2 green fluorescent protein (to visualize transfected cells). The transfected cells were split next day into 35-mm dishes that served as the recording chamber.
Measurement of Ionic Currents-Cells were voltage-clamped using the whole-cell configuration of the patch clamp technique. Pipettes were pulled from Schott 8250 glass (Garner Glass, Claremont, CA) on a Sutter P-97 puller (Sutter Instruments Co., Novato, CA). Currents were recorded using an Axopatch 200A amplifier (Molecular Devices, Sunnyvale, CA) and digitized with the ITC-18 data acquisition interface (Instrutech Corp., Port Washington, NY). Experiments were controlled by a Power Macintosh G3 computer (Apple Computer, Cupertino, CA) running S5 data acquisition software written by Dr. Stephen Ikeda (NIAAA, National Institutes of Health, Bethesda, MD). Leak current was subtracted online using a P/4 protocol. All recordings were carried out at room temperature, and the holding potential was Ϫ120 mV. Whole-cell currents were digitized depending on test step duration at 50 (25 ms), 10 (200 ms), and 4 (1000 ms) kHz after analog filtering at 1-10 kHz.
Data Analysis-Data were analyzed using IgorPro (WaveMetrics, Lake Oswego, OR) running on a Macintosh computer.
Step currents were measured as the average of 1 ms at the end of the voltage step. Activation ( Act ) was determined by fitting a single exponential function to the step current after a 0.3-ms delay (15). Inactivation ( Inact ) was determined by fitting a single exponential function from peak step current to the end of the step. The effect of roscovitine on inactivation of L-type or chimeric calcium channels was measured by using either 200or 1000-ms voltage steps. The magnitude of inactivation was measured from either the I End /I Peak ratio, where I End was measured at the end of the step and I Peak was peak current, or as the I Post /I Pre ratio from a triple pulse protocol consisting of identical 25-ms pre-and postpulse steps (to elicit peak current) bracketing a 200-ms test pulse to voltages ranging from Ϫ120 to ϩ80 mV. Group data were calculated as mean Ϯ S.D. throughout the paper. A paired t test was used for within-cell comparisons. One-way analysis of variance with Tukey's HSD post hoc test was used to test for differences among three or more independent groups.
Solutions-The internal pipette solution contained 104 mM NMG-Cl, 14 mM creatine-PO 4 , 6 mM MgCl 2 , 10 mM NMG-HEPES, 5 mM Tris-ATP, 0.3 mM Tris-GTP, and 10 mM NMG-EGTA with osmolarity of 280 mosM and pH 7.3. The external recording solution contained 30 mM BaCl 2 , 100 mM NMG-Cl, 10 mM NMG-HEPES, osmolarity of 300 mosM, and pH 7.3. We used Ba 2ϩ as the permeant ion to exclude Ca 2ϩ -dependent inactivation (25,26), since we have previously demonstrated that (R)-roscovitine enhances VDI but does not affect calciumdependent inactivation (19). Both (R)-and (S)-roscovitine were prepared as a 50 mM stock solution in DMSO and stored at Ϫ30°C. All external solutions contained the same DMSO concentration so that the roscovitine concentration was the sole variable when changing solutions. Test solutions were applied from a gravity-fed perfusion system with an exchange time of 1-2 s.

(S)-Roscovitine Does Not Affect L-channel Activation-We
have shown that (R)-roscovitine significantly slows activation and enhances inactivation of L-type channels and hypothesized that a single extracellular binding site mediated both effects (19). We further investigated this idea by examining the effect of (S)-roscovitine on L-type channels, which was useful in differentiating multiple roscovitine binding sites on N-type channels (15,17,18). Using 25-ms voltage steps to ϩ20 mV, we verified that 100 M (R)-roscovitine slowed L-channel activation (15,19) (Fig. 2). However, the activation rate was not altered by the application of 100 M (S)-roscovitine ( Fig. 2A), which was estimated by fitting the ϩ20 mV step current using a single exponential function to generate Act . The average change in Act induced by (S)-roscovitine was 8.6 Ϯ 8.5% (ϮS.D., n ϭ 7), whereas the (R)-roscovitine-induced increase of Act was 148.6 Ϯ 24.8% (n ϭ 9, p Ͻ 0.001; Fig. 2C). Thus, the L-channel binding site involved with slowed activation shows stereo-selectivity for roscovitine. (S)-Roscovitine Enhances L-channel VDI-The inability of (S)-roscovitine to slow L-channel activation could be explained by low affinity of a single roscovitine binding site for the S enantiomer. If this were true, inactivation should also be insensitive to (S)-roscovitine. This hypothesis was tested using 1000-ms voltage steps to ϩ30 mV to measure VDI (30 mM Ba 2ϩ ). Fig. 3 shows typical L-current traces recorded from the same cell exposed to either 100 M (S)-or (R)-roscovitine. The speed of inactivation ( Inact ) was determined from single exponential fits to the inactivating portion of the current (peak to end). Counter to our hypothesis, (S)-roscovitine increased inactivation to a degree similar to that of (R)-roscovitine (Fig. 3). (S)-Roscovitine decreased Inact from 591 Ϯ 133 to 220 Ϯ 11 ms (p Ͻ 0.05, n ϭ 3), whereas Inact was decreased from 504 Ϯ 84 to 216 Ϯ 7 ms (p Ͻ 0.05, n ϭ 3) by (R)-roscovitine (Fig. 3B). There was no significant difference in Inact between (S)-roscovitine and (R)roscovitine. The L-channel binding site that mediates roscovitine-induced enhancement of inactivation does not show stereo-selectivity.
The effect of (S)-roscovitine on the voltage dependence of inactivation was determined using a three-pulse protocol where the ratio of the postpulse/prepulse (25 ms, ϩ30 mV) current (I Post /I Pre ) was used to monitor inactivation induced by 200-ms inactivating steps ranging in voltage from Ϫ120 to ϩ80 mV. Under control conditions (30 mM Ba 2ϩ ), the inactivation versus voltage relationship was U-shaped ( Fig. 3C), which was more prominent than we observed previously using 10 mM Ba 2ϩ (19). The reason for this difference is unknown, but as we did with our previous data, we fit the inactivation versus voltage data from Ϫ120 to ϩ50 mV (peak inactivation) using a single Boltzmann equation (Fig. 3C). (S)-Roscovitine increased the magnitude of inactivation so that the inactivation versus voltage relationship became less U-shaped. Maximal inactivation from the Boltzmann fit increased from 0.19 Ϯ 0.01 in control and 0.18 Ϯ 0.02 for recovery to 0.66 Ϯ 0.03 by (S)-roscovitine. Boltzmann fitting of the I Post /I Pre versus voltage relationship showed a small (ϳ5 mV) (S)-roscovitine-induced right shift in half-inactivation voltage (V1 ⁄ 2 ) and a small (e-fold/4 mV) decrease in slope (Fig. 3C) that was also observed with (R)roscovitine (see Fig. 6A). Maximal inactivation was increased from 0.23 in control to 0.67 in (R)-roscovitine, which is very

. (S)-Roscovitine does not slow activation of L-type channels.
A, L-current was activated during 25-ms voltage steps to ϩ20 mV. 100 M (S)-roscovitine (S-Rosc, black) slightly decreased current but did not alter activation relative to control (CNTL, gray) and washout (WO, gray). B, in the same cell, 100 M (R)-roscovitine (R-Rosc, black) slowed L-current activation. C, the percentage change in Act at ϩ20 mV induced by application of 100 M (R)roscovitine was significantly different (***, p Ͻ 0.001) from zero, whereas that for (S)-roscovitine was not. The data are shown as mean Ϯ S.D., and the number of cells examined is indicated. Steep Dose-Response Relationship for (S)-Roscovitine-The dose-response relationship of (R)-roscovitine showed similar EC 50 values for both slowed activation and enhanced inactivation (20 -30 M), which supported a single binding site (19). However, the Hill coefficient was close to 1 (1.2) for slowed activation, whereas that for enhanced inactivation was Ͼ2 (2.3), which suggested separate binding sites for these two effects. The differential sensitivity of slowed activation for (R)versus (S)-roscovitine supports the latter idea. We further investigated the enhanced inactivation by determining the EC 50 for (S)roscovitine. Inactivation was measured as the I End /I Peak ratio calculated from 200-ms steps to ϩ25 mV. (S)-Roscovitine increased inactivation in a dose-dependent manner from 10 to 100 M, but the response to 300 M was reproducibly reduced relative to that of 100 M. This effect was also observed in (R)roscovitine (19). The dose-response data were fit (0 -100 M) using the Hill equation to yield EC 50 ϭ 41.1 Ϯ 0.1 M and a Hill coefficient of 4.47 Ϯ 0.03 (Fig. 4). Both of these values are larger than that obtained from (R)-roscovitine (19). One thing that is clear from these data is that the binding site mediating enhanced inactivation shows a Hill coefficient of Ͼ2, suggesting positive cooperativity. One explanation for this cooperativity is that roscovitine selectively binds to inactivated channels to enhance VDI.
Chimeric N-L Channels-We were interested in localizing the L-type channel structures that mediate both roscovitineinduced effects to further investigate the hypothesis of multiple roscovitine binding sites. We focused on a chimeric strategy that had been previously exploited to determine DHP binding sites on L-type channels (20) but needed to establish that the (R)-roscovitine effect on L-channels was unique. We had previously demonstrated that (R)-roscovitine uniquely affected L-channel activation because N-channel activation was slowed only at voltages of Ͻ0 mV and L-channel activation was slowed at all voltages (14,19). However, our recordings of native N-current (bullfrog sympathetic neurons) showed enhanced inactivation during voltage steps (peak current) (14). Surprisingly, our data from N-type channels expressed in HEK293 cells failed to reproduce those results (Figs. 5 and 6). Inactivation during 200-ms steps to ϩ30 mV was not enhanced (Fig. 5B), and the mean I Post /I Pre ratio was not changed at any voltage by 100 M (R)-roscovitine (Fig. 6B). Single Boltzmann equation fits from Ϫ120 mV to peak inactivation (ϩ10 mV) yielded maximum inactivation of 0.48 Ϯ 0.05 in control versus 0.47 Ϯ 0.03 (ϮS.D., n ϭ 6, not significant) in (R)-roscovitine. Thus, (R)roscovitine does not enhance U-type inactivation of N-type calcium channels (21) expressed in HEK293 cells. These results demonstrate that N-L chimeric channels can be used to localize L-channel domain(s) transducing the roscovitine effects.  Initially, chimeric channels were generated by placing one or two N-channel domains into an L-channel backbone. Accordingly each domain is identified by a single-letter code corresponding to the contributing channel type (L for L-type and N for N-type). For example, the WT L-channel is LLLL, whereas the chimera with domain III of the N-type channel in the L-channel backbone is indicated by LLNL. The data were gathered from five L-channel-based chimeric channels (LLNN, NLLN, NLLL, LLNL, and LLLN) as well as the two WT channels (LLLL and NNNN). However, two other chimeras (NNLL and LNLL) failed to generate measurable current. Additional chimeras were generated by inserting a single L-channel domain into the N-channel backbone (e.g. LNNN, NNLN, and NNNL), but none of these generated measurable current. Our detailed investigation into these non-functional chimeric channels revealed N-DII as the problem. We further investigated this domain using hemidomain chimeras that separated N-DII into a V-region that encompassed transmembrane segments S1-S4 and a P-region with S5 and S6 (see "Experimental Procedures" and Fig. 1), which revealed that chimeric channels containing the N-DII P-region (S5 and S6) failed to generate functional channels (not shown). To overcome this problem, we engineered a domain II with an N-channel V-region (S1-S4) and L-channel P-region (S5 and S6) (see "Experimental Procedures" and Fig. 1), which we call N*. This engineered domain II allowed us to generate functional chimeric channels with a predominately N-type backbone, including LN*NN, and LN*NL.
Inactivation-VDI was examined using the triple pulse protocol described above (Fig. 6). The first chimeric channel tested was LLNN, which would potentially allow us to localize the roscovitine binding sites to half of the channel. 100 M (R)roscovitine enhanced VDI of the LLNN chimera (Figs. 5C and 6C) similar to that observed with the WT L-channel (LLLL) (Figs. 5A and 6A), which suggests that the relevant binding site is within domain I and/or II. This was further supported by (R)-roscovitine-enhanced VDI for both the LLNL (Figs. 5E and 6E) and LLLN (Figs. 5F and 6F) chimeras. The role of L-DI was examined by applying 100 M (R)-roscovitine to the NLLL chimera, which, like the WT N-channel (NNNN), failed to alter VDI (Figs. 5D and 6D). If L-DI is the relevant domain, we should be able to transfer enhanced inactivation to the N-channel by introducing replacing N-DI with the L-channel homolog. Unfortunately, the LNNN chimera contained N-DII, which caused our chimeric channels to fail to express current (see above and "Experimental Procedures"). This problem was overcome by introducing the N* domain discussed above. As a first step, we examined the effect of the N* domain on (R)-roscovitine-enhanced VDI by generating the LN*LL chimera, which showed enhanced inactivation like the WT L-channel. This result also supports L-DI as the relevant domain because introduction of the N-DII V-region into the L-channel failed to affect (R)-roscovitine-enhanced VDI. Placing L-DI into the N-channel to generate the LN*NN chimera made this channel respond to 100 M (R)-roscovitine with enhanced VDI that was significantly different from that of WT N-channel but similar to that of WT L-channels (see Fig. 9A).
Statistical comparisons of all of the channels tested are shown in Fig. 9A. For this comparison, we measured the enhancement of fractional inactivation induced by (R)-roscovitine (⌬Rosc Inactivation in Fig. 9), which was calculated as the difference in the I Post /I Pre ratio with or without 100 M (R)roscovitine (see the legend to Fig. 9A). The analysis of variance revealed that the response of L-DI-containing chimeric channels to (R)-roscovitine was not significantly different from that of WT L-channels (LLLL) but was significantly different from response of NLLL, NLLN, and WT N-channels (NNNN). Together, these results demonstrate that L-DI is critical for the (R)-roscovitine enhancement of VDI.
As demonstrated in Fig. 4, the dose-response relationship for roscovitine-enhanced inactivation of WT L-channels shows positive cooperativity with a Hill coefficient of Ͼ2 (19), whereas WT N-channels respond to roscovitine with a Hill coefficient of 1 (14,15). Our data support localization of the roscovitine binding site within L-DI, which predicts that LN*NN will show positive cooperativity in response to increasing (R)-roscovitine concentrations. This idea was tested by measuring the (R)roscovitine dose response of the LN*NN chimera. As expected, the (R)-roscovitine-enhanced VDI of the LN*NN chimera (Fig.  7A) showed a steep dose-response relationship with the Hill coefficient of 2.3 (Fig. 7C). On the other hand, the NLLL chimera responded to (R)-roscovitine with a shallow dose-response relationship with the Hill coefficient of 0.6. The comparisons showed that the responses of LLLL and LN*NN to 100 M (R)-roscovitine were significantly larger relative to either NLLL and NNNN (see also Fig. 9A, different data set). Thus, the LN*NN chimera responds to (R)-roscovitine like WT L-channels.
Interestingly, the converse is true for N-DI. The (R)-roscovitine-induced inhibition of WT N-channels results in roughly equal reduction of peak and end current during long (200-or 1000-ms) voltage steps (Fig. 5B). This effect is mimicked by NLLN (not shown) and NLLL (Fig. 7B) but not by LN*NN (Fig.  7A). We quantified this effect by measuring the difference in end current versus peak current inhibition (⌬ Inh End-Peak) and compared that difference between the LLLL, LN*NN, NLLL, NLLN, and NNNN channels (Fig. 9C). Consistent with our analysis of VDI, LLLL and LN*NN showed larger (R)-roscovitine-induced inhibition of end versus peak current, which resulted in a significantly larger ⌬ Inh value relative to NLLL, NLLN, or NNNN. The WT L-channel response to (S)-roscovitine was also statistically similar to that of (R)-roscovitine, which supports our conclusion that these drugs enhance VDI by binding to a common site (Fig. 4). On the other hand, the response to (R)-roscovitine of both N-DI-containing chimeras (NLLL and NLLN) was statistically similar to that of WT FIGURE 7. Binding site properties for (R)-roscovitine-enhanced VDI follows L-DI. A, the LN*NN chimera responds to 100 M (R)-roscovitine (Rosc, black trace) with enhanced inactivation during 1-s steps to the voltage-generating maximal current (Test V ϭ ϩ30 mV). Note that the peak current early in the voltage step was little affected by roscovitine. B, the NLLL chimera responds to (R)-roscovitine (black trace) with a roughly equal inhibition of both peak and end current during 1-s steps to the voltage generating peak current (Test V ϭ ϩ10 mV). Note that inactivation during the voltage step was only weakly affected by (R)-roscovitine. C, the enhancement of inactivation induced by (R)-roscovitine was calculated as the difference in the I Post /I Pre ratio between control and Rosc (⌬Rosc Inactivation). This difference in inactivation is plotted versus (R)-roscovitine concentration for WT L-channels (LLLL) and WT N-channels (NNNN) as well as the LN*NN and NLLL chimeras. N-channels, but the response of all three channels was statistically different from that of either LN*NN or LLLL. Thus, the roscovitine binding site mediating N-channel inhibition appears to be localized to N-DI.
Slowed Activation-We investigated the effect of (R)-roscovitine on activation using short steps of either 10-or 25-ms duration (determined by activation kinetics) to voltages ranging from ϩ10 to ϩ70 mV (Fig. 8). Current traces elicited by voltage steps to ϩ20 mV show the effect of 100 M (R)-roscovitine on activation and that effect was further quantified by fitting activation using a single exponential equation to determine Act , which was plotted versus step voltage to show the effect of (R)-roscovitine over a range of voltages. Using the approach that worked well for inactivation, we measured activation of the LLNN chimera, but (R)-roscovitine failed to slow activation (Figs. 8E and 9B). This suggests that L-DIII and/or L-DIV are important for the effect of (R)-roscovitine on activa-tion. The critical domain appeared to be L-DIV, since (R)roscovitine slowed activation of the LLNL chimera (Figs. 8C and 9B). The importance of L-DIV was further supported because activation of the LLLN chimera was not affected by (R)-roscovitine (Figs. 8F and 9B). Surprisingly, (R)-roscovitine also failed to affect activation of the NLLL chimera, which suggested that L-DI was also required for (R)-roscovitine-induced slowed activation. Consistent with this idea, LN*LL also responded to (R)-roscovitine with slowed activation, but LN*NN and NLLN did not (Figs. 8 and 9B). A critical test of our dual domain hypothesis was the LN*NL chimera, which responded to 100 M (R)-roscovitine with slowed activation (Figs. 8I and 9B). Thus, both L-DI and L-DIV are both necessary and sufficient to mediate (R)-roscovitine-induced slowed activation.

DISCUSSION
Roscovitine, a promising drug with anticancer action (10,22,23), has recently been shown to affect voltage-dependent calcium channels (14 -16, 19) and potassium channels (15,24). From our previous findings (19), we generated the hypothesis that (R)-roscovitine-induced slowed activation and enhanced inactivation were mediated by a common binding site. However, this hypothesis was not supported by the present results. Our pharmacological data showed that the roscovitine receptor site mediating slowed activation (RR SA ) was stereo-selective, whereas the site mediating enhanced inactivation (RR EI ) was not. It appears that L-type channels contain two extracellularly exposed roscovitine binding sites. Our chimera data support localization of RR EI to L-DI, whereas RR SA appears to be composed of amino acids from both L-DI and L-DIV.
Pharmacological Separation of Roscovitine Effects on L-type Channels-We have previously shown that intracellularly applied (R)-roscovitine failed to affect L-channel kinetics and failed to abrogate the effects induced by externally applied (R)roscovitine, which led us to conclude that (R)-roscovitine was interacting with an extracellularly exposed site to affect L-channel kinetics (19). Our results using (S)-roscovitine have enabled us to refine our original hypothesis to propose that there are two extracellularly exposed roscovitine binding sites. The stereo-selectivity of RR SA permits enhanced inactivation to be examined in isolation by using (S)-roscovitine, which appears to exclusively activate the chiral-insensitive RR EI . (S)-Roscovitine accelerated inactivation with the same potency as (R)roscovitine, which shows that RR EI is truly insensitive to the chiral carbon in the C2 side chain.
N-type channels also show two roscovitine-induced effects that are differentially affected by (S)-roscovitine (17,18). The agonist effect of (R)-roscovitine (slowed deactivation) required at least a 20-fold higher (S)-roscovitine concentration, whereas the roscovitine-induced antagonist effect showed a similar IC 50 for the two enantiomers (18). The two N-channel binding sites mediate effects that differ from those observed with L-channels. The N-channel agonist site appears to be revealed only upon channel activation (14,15), whereas the antagonist site is linked to closed state (resting) inactivation (18). Thus, both types of voltage-dependent calcium channels (N-type and L-type) appear to have two roscovitine binding sites, one of which is stereo-selective. The stereo-specific sites affect the activation-deactivation process (slowed activation for L-channels and slowed deactivation for N-channels), whereas the stereo-insensitive sites are involved with inactivation (open state voltage-dependent inactivation for L-channels and closed state inactivation for N-channels).
Localization of the Roscovitine Binding Sites-With the apparent identification of multiple binding sites from the pharmacological experiments, we wanted to determine if these sites could be physically differentiated. Upon verification that the (R)-roscovitine-induced effects on N-type and L-type channels were unique, we utilized N-L chimeras to determine the domains containing the roscovitine binding sites. The chimeras were generated by "domain swapping" in an attempt to localize the binding sites using relatively few manipulations, and the technique was successful. However, a major problem was that chimeric channels containing N-DII failed to express measurable ionic currents, and sequencing showed that this did not result from unintended mutations in these chimeras. Our investigation demonstrated that the problem was within transmembrane segments S5 and S6 (the P-region) of N-DII, because L-channels containing just the N-DII P-region failed to generate measurable current, whereas current was reproducibly found in L-channels containing just the N-DII V-region (S1-S4). Thus, the engineered N-channel-like domain II (N*), which had transmembrane segments S1-S4 from N-channel and S5 and S6 from L-channel, allowed us to test the (R)-roscovitine effect on chimeric "N-channels" containing single L-channel domains.
The LLNN chimera was the first to be generated and demonstrated that RR EI was localized to L-DI and/or L-DII, but the impaired slowed activation response of this chimera suggested that RR SA could be within L-DIII or L-DIV. The strong (R)roscovitine-induced slowed activation of the LLNL chimera along with the loss of that effect with the LLLN chimera supported L-DIV for RR SA . However, the slowed activation effect was unexpectedly lost in the NLLL chimera, which along with the responsiveness of LLNL suggested that both L-DI and L-DIV were involved. This hypothesis was further supported by showing that (R)-roscovitine failed to alter activation of the NLLN chimera, whereas activation was slowed in the LN*LL chimera. The critical test was that (R)-roscovitine-induced slowed activation was transferred to the N-channel by inserting both L-DI and L-DIV to generate the LN*NL chimera. Thus, the evidence strongly supports the conclusion that RR SA is formed by amino acid residues from both L-DI and L-DIV. It seems very likely that the site resides at the interface between L-DI and L-DIV, which suggests that (R)-roscovitine can be used as a probe to investigate structures within these domains that interact to form functional channels. Although the RR SA site appears to span two L-channel domains, we conclude that RR EI is contained within L-DI. (R)-Roscovitine-enhanced VDI was observed in all chimeric channels containing L-DI, including LLNN, LLNL, LLLN, and LN*LL. The two critical tests of the L-DI hypothesis were the absence of enhanced inactivation of the NLLL chimera and the transfer of (R)-roscovitine-enhanced VDI to the N-channel by inserting L-DI to generate the LN*NN chimera. The dose-response for (R)-roscovitine-induced enhanced VDI was almost identical between LN*NN and WT L-channels, which strongly suggests that RR EI is transferred to the N-channel by the insertion of L-DI. Thus, our results are consistent with the localization of RR EI to L-DI and support the hypothesis that the two (R)-roscovitine-induced effects are mediated by physically distinct structures.
The Mechanism of Roscovitine-induced Inhibition-We previously concluded that both slowed activation and enhanced inactivation resulted from closed channel binding of (R)-roscovitine to a single site (19). This mechanism is still valid for (R)roscovitine-induced slowed activation, but the discovery of a separate site for enhanced inactivation dictates a reevaluation of the mechanism for this effect. One clue comes from the doseresponse relationship, where the Hill coefficient for the roscovitine-induced enhancement of inactivation is Ͼ2, which indicates positive cooperativity at RR EI . A likely possibility is that the cooperativity results from transient binding site availability that would occur if (R)-roscovitine specifically interacted with the inactivated (VDI) state. The positive cooperativity occurs because at low roscovitine concentrations, the on-rate is much lower than the mean dwell time of the channel in the inactivated state, so the effective binding is small relative to a site that is constantly available. As the concentration increases, the effective on-rate becomes high enough for roscovitine to bind during the brief sojourns into the inactivated state.
One problem with this putative mechanism is that roscovitine can still enhance inactivation of L-D1-containing channels that show little or no inactivation under control conditions (e.g. LLLN and LN*LL). This phenomenon has also been observed in L-channel containing the Timothy syndrome mutation that abrogates VDI, but roscovitine can still enhance inactivation (40). This suggests that the development of the inactivated state is not required to reveal RR EI . A likely alternative hypothesis is that a voltage-dependent conformational change in L-DI is required for roscovitine binding. One scenario is that the outward movement of L-DI S4 is the trigger that reveals RR EI . Further work is required to more precisely determine the mechanism for roscovitine-enhanced VDI, but the available evidence suggests that roscovitine binding moves the inactivation gate, and it is not the movement of the inactivation gate that permits roscovitine binding.
Our hypothesis of L-DI containing RR EI fits with the current idea that the L-channel I-II linker and domain IS6 form part of a "hinged lid" that mediates VDI (27). It seems that roscovitine binding to RR EI in L-DI induces a conformational change that speeds movement of the hinged lid so that VDI becomes faster. In addition, this conformational shift may impede the reopening of the inactivation gate, which would explain the slowed recovery from inactivation that we observed previously (19).
Unique Binding Sites for 2,6,9-Trisubstituted Purine Antagonists-Calcium channel antagonists play an important role in treatment of neurological and cardiovascular diseases (6). L-type channel antagonists, such as dihydropyridines, phenylalkylamines, and benzothiazepines, are used for treatment of hypertension and angina pectoris (4,5,7,8,28,29). The dihydropyridine, phenylalkylamine, and benzothiazepine binding sites appear to encompass separate but partially overlapping regions located within domains III and IV (30 -32). The effect of (R)-roscovitine on the LLNN chimera demonstrates that RR EI is a unique binding site on L-type calcium channels. Therefore, roscovitine (R and S) represents a new class of L-channel antagonists that induce L-channel inhibition by enhancing VDI without affecting closed state inactivation (19).
Activation of RR SA is unique among calcium channel antagonists, and such drugs could have an important clinical use in preventing reactivation of L-type channels that generate proarrhythmic events, such as early afterdepolarizations (33)(34)(35)(36). In addition, activation of RR EI could have important therapeutic effects for cardiac arrhythmias resulting from this "rebound" L-channel activity on the falling phase of the cardiac afterdepolarization (e.g. Torsades de Pointes) (35)(36)(37). Recordings of L-current stimulated by a cardiac afterdepolarization waveform show that (R)-roscovitine preferentially inhibits current late in the afterdepolarization as expected from enhanced VDI (19). In addition, activation of RR EI could normalize L-channel activity in patients suffering from genetic disorders that slow VDI, such as Timothy syndrome (38 -40). Indeed, we recently demonstrated that (R)-roscovitine could normalize inactivation of L-channels carrying the Timothy syndrome mutation (40). The further characterization of both RR SA and RR EI could lead to the development of a new class of drugs that specifically target L-type channels to control arrhythmic behavior by slowing activation and enhancing VDI.