Roles of molecular regions in determining differences between voltage dependence of activation of CaV3.1 and CaV1.2 calcium channels.

Voltage-dependent calcium channels are classified into low voltage-activated and high voltage-activated channels. We have investigated the molecular basis for this difference in voltage dependence of activation by constructing chimeras between a low voltage-activated channel (Ca(V)3.1) and a high voltage-activated channel (Ca(V)1.2), focusing on steady-state activation properties. Wild type and chimeras were expressed in oocytes, and two-electrode voltage clamp recordings were made of calcium channel currents. Replacement of domains I, III, or IV of the Ca 3.1 channel with the corresponding domain of Ca(V)1.2 led (V)to high voltage-activated channels; for these constructs the current/voltage (I/V) curves were similar to those for Ca(V)1.2 wild type. However, replacement of domain II gave only a small shift to the right of the I/V curve and modulation of the activation kinetics but did not lead to a high voltage-activating channel with an I/V curve like Ca 1.2. We also investigated the role of the voltage sensor (V)S4 by replacing the S4 segment of Ca(V)3.1 with that of Ca 1.2. For domain I, there was no shift in the I/V curve (V)as compared with Ca(V)3.1, and there were relatively small shifts to the right for domains III and IV. Taken together, these results suggest that domains I, III, and IV (rather than domain II) are apparently critical for channel opening and, therefore, contribute strongly to the difference in voltage dependence of activation between Ca 3.1 and Ca(V)1.2. However, the S4 segments in domains I, (V)III, and IV did not account for this difference in voltage dependence.

calcium channels, and the Ca V 3 family displays T-type currents and includes ␣ 1G , ␣ 1H , and ␣ 1I (3)(4)(5)(6)(7). Ca V 3 channels have markedly different biophysical characteristics from the other families; Ca V 3 channels are low voltage-activating and have fast inactivation and small single channel conductance. On the other hand, Ca V 1 channels, for instance, are high voltageactivating and show little inactivation (with Ba 2ϩ as the charge carrier) and large single channel conductance (2,8,9).
Voltage-dependent ion channels form a huge family comprising potassium and sodium as well as calcium channels. For this family, the S4 segments have a conserved amino acid sequence with four to eight positively charged residues (arginines or lysines), each separated by two hydrophobic residues. This segment plays a vital role in voltage-dependent activation (10). The suggested crystal structure for the K V AP potassium channel shows a possible "paddle-like" movement of the S4 segment during depolarization (11,12). Most research on the molecular basis of activation has previously been carried out with potassium channels and, to a lesser extent, sodium channels. For calcium channels there are very few studies on the molecular mechanisms of activation (13,14), and voltage gating in the calcium channels is not yet clear. However, because the four S4 segments in the calcium channel have similar conserved motifs of charged amino acids as in sodium and potassium channels, it is thought that the S4 segment in calcium channels has a similar function as in sodium and potassium channels (2,15). For calcium channels, it is not understood at a molecular level why Ca V 3 channels are low voltage-activating, whereas Ca V 1 and Ca V 2 channels are high voltage-activating. One might perhaps expect that these differences in activation are due to differences in S4 segments. In the present work we have investigated the molecular basis for these differences in voltage dependence of activation between Ca V 3 channels and Ca V 1 channels, where there are striking differences between the two types of wild type channels. For this, we have created chimeras between Ca V 3.1 and Ca V 1.2 channels, swapping the four domains in Ca V 3.1 separately with the corresponding region in Ca V 1.2 to study the effects of each domain on channel activation. Similarly, we have investigated the role of the S4 segments by swapping these segments between the two channels. In this way we have aimed to locate key molecular regions underlying the differences in activation properties, focusing on the striking differences (which make a chimeric study possible) in voltage dependence of steady-state activation between these high and low voltage-activated channels. diac/brain ␤ 2 , GenBank™ accession number M80545 (20); rabbit skeletal muscle ␣ 2 ␦, GenBank™ accession number M21948 (21) (all the clones were kindly donated by Professor F. Hofmann). Unless stated otherwise, restriction enzymes were purchased from New England Biolabs and Promega, and chemical reagents were obtained from Sigma.
Subcloning of Ca V 3.1 cDNA into pGEM-HEL Vector-To improve the expression of Ca V 3.1 in oocytes, the full-length cDNA of the channel was subcloned from pcDNA3 into "pGEM-HEL" vector (a modified version of pGEM-HE (22)). The pGEM-HEL vector was constructed from pGEM-HE by modifying its multiple cloning site as well as the 3Ј-linker downstream from the 3Ј end of the untranslated ␤-globin. For the multiple cloning site, two matched oligonucleotides were used containing XbaI, EcoRI, and NotI sites, with compatible BamHI and HindIII overhangs although with incomplete recognition sites for the latter two enzymes. The phosphorylated oligonucleotides were annealed and ligated into pGEM-HE with BamHI and HindIII, the resulting vector replacing the latter two sites with XbaI, EcoRI, and NotI. Also, using a similar procedure the 3Ј-linker of pGEM-HE was modified; an unwanted NotI site was deleted, and a unique MluI site was introduced for linearization purposes. The annealed oligonucleotides had a MluI site and PstI and SalI overhangs and were inserted into pGEM-HE using the latter two enzymes. Sequences were verified at the modified regions by automated sequencing.
The full-length Ca V 3.1 cDNA was excised in two subfragments from pcDNA3 by XbaI and EcoRI, then by EcoRI and NotI digestion. Both subfragments were inserted into pGEM-HEL using the same enzymes, re-assembling the full-length Ca V 3.1. The construct was verified by automated sequencing at the joins.
Construction of Ca 2ϩ Channel Chimeras-Referring to the domains I-IV of Ca V 3.1 as GGGG and that of Ca V 1.2 as CCCC, four chimeras were made by swapping domains I-IV as follows: CGGG, GCGG, GGCG, and GGGC. Also, four chimeras were made by replacing S4 segments in Ca V 3.1 with the corresponding segments in Ca V 1.2 for domain I (IS4C), domain II (IIS4C), domain III (IIIS4C), and domain IV (IVS4C). The schematic structure of each chimera is shown in Fig. 1.
All chimeras were constructed using standard PCR overlap extension methods (23). Briefly, in this method first round PCR products were made with wild type Ca V 3.1 and Ca V 1.2 cDNA as template using primers with appropriate overhangs at the ends of the regions to be replaced in the chimera, then a second round PCR product was made with primers spanning the whole region using the first round PCR products as template. Chimeric PCR products were then inserted back into the Ca V 3.1 wild type clones using appropriate restriction enzymes. Further details for each construct are as follows.
For the domain I chimera, CGGG, amino acids 81-398 of Ca V 3.1 were replaced by residues 154 -438 of Ca V 1.2. In the second round PCR product, the chimeric fragment included XbaI and HindIII restriction sites in the Ca V 1.3 sequences on either side of the Ca V 1.2 sequence. The PCR product was digested with these enzymes to yield a 1186-bp product and ligated into the similarly digested wild type Ca V 3.1-PGEM-HEL. For the domain II chimera, GCGG, residues 399 -967 of Ca V 3.1 were replaced by residues 439 -786 of Ca V 1.2. The final chimeric PCR fragment included Csp45I and PinAI sites in the Ca V 3.1 sequences on either side of the Ca V 1.2 sequence; after digestion (3008-bp product) it was ligated into Ca V 3.1-pGEM-HEL using these two enzymes. For the domain III chimera, GGCG, residues 968 -1541 of Ca V 3.1 were replaced by residues 787-1199 of Ca V 1.2. The final chimeric PCR product contained BamHI and XhoI sites; after digestion (2573-bp product), it was inserted with the same enzymes into Ca V 3.1-(HindIII-KpnI)-pUC18. The latter is a subclone of wild type Ca V 3.1, which had been made previously by inserting the HindIII-KpnI subfragment of Ca V 3.1 (4543 bp) into the vector pUC18. For the domain IV chimera, GGGC, residues 1542-1861 of Ca V 3.1 were replaced by residues 1200 -1508 of Ca V 1.2. The chimeric PCR product included EcoRI and KpnI sites (1499 bp after digestion); it was inserted into Ca V 3.1-pcDNA3 between the above sites. For the S4 chimera in domain I, IS4C, residues 175-199 of Ca V 3.1 were replaced by residues 262-286 of Ca V 1.2. The chimeric PCR product included KspI and Csp45I sites; the digested product (500 bp) was inserted into Ca V 3.1-PGEM-HEL using these two enzymes. For the S4 chimera in domain II, IIS4C, residues 830 -856 of Ca V 3.1 were replaced by residues 464 -672 in Ca V 1.2. The PCR fragment contained BamHI and BsaAI sites and was inserted, after digestion (741-bp product) into Ca V 3.1(HindIII-KpnI)-pUC18. For the S4 chimera in domain III, IIIS4C, residues 1377-1401 of Ca v 3.1 were replaced by residues 1023-1047 of Ca V 1.2. The PCR product contained StuI and PinAI sites and, after digestion (361-bp product), was ligated into Ca V 3.1(HindIII-KpnI)-pUC18. For the S4 chimera in domain IV, IVS4C, residues 1716 -1740 of Ca V 3.1 were replaced by residues 1355-1379 of Ca V 1.2. The PCR fragment included EcoRI and KpnI sites and, after digestion (1532 bp), was inserted into Ca V 3.1-pcDNA3 with these enzymes. Finally, the above inserts in pcDNA3 were subcloned into Ca V 3.1-PGEM-HEL using EcoRI and NotI digestion; the inserts in Ca V 3.1(HindIII-KpnI)-pUC18 were transferred to Ca V 3.1-PGEM-HEL by HindIII and XhoI digestion. The sequences of the chimeras were verified by DNA sequencing; the joins and the entire PCR inserts were sequenced. Ca V 3.1 and chimeric cDNAs (in pGEM-HEL) were linearized with MluI; Ca V 1.2 (in pcDNA3) was linearized with Asp718; ␤ 2 and ␣ 2 ␦ (both in pcDNA3) were linearized with NotI. Capped cRNAs were synthesized in vitro using T7 MEGAscript (Ambion).
Electrophysiological Recording and Data Analysis-DuPont stage VI oocytes were prepared from Xenopus laevis frogs using standard techniques (24,25). Each oocyte was injected with 10 -20 ng of cRNA in a volume of 50 nl. For co-injection, the ratio of ␣ 1 :␤ 2 :␣ 2 ␦ was about 3  For electrophysiological recording, oocytes were held in a 50-l recording chamber and perfused with barium solution (40 mM Ba(OH) 2 , 50 mM NaOH, 2 mM KOH, and 5 mM HEPES, adjusted to pH 7.4 with methanesulfonic acid). The calcium channel currents (i.e. with Ba 2ϩ as charge carrier) were measured at 22-25°C by the 2-electrode voltageclamp technique using a Geneclamp500 amplifier (Axon Instruments) as described previously (24,25). Currents were filtered at 2 kHz and sampled at 4 kHz using a CED1401Plus interface with CED data acquisition software. The membrane potential of oocytes was held at Ϫ80 mV. To create the current-voltage (I/V) relationship, Ba 2ϩ currents were elicited by a series of 500-ms pulses every 10 s, from Ϫ70 mV to ϩ70 mV in a 10-mV step. This was followed by twenty 200-ms hyperpolarizing pulses every 2 s to Ϫ90 mV for subsequent leak and capacity current subtractions. In cases where channel currents were particularly small (i.e. for small voltage steps with high voltage-activating channels/ chimeras), a low signal to background ratio did not always allow reliable determination of channel currents.
Electrophysiology data was acquired and analyzed by CED software, with Origin 5.0 used for further analysis and curve-fitting. The I/V curves were obtained for peak current amplitudes and were fitted with the Boltzmann equation, where I is the measured peak current, V is test potential, V rev is the reversal potential, G max is the maximum conductance, V 0.5 is the potential for half-maximal activation, and k is the slope parameter. The activation times were taken as the times from 20 to 80% of maximum current (t 20 -80 ). Inactivation time courses were fit with a single exponential of time constant inact . Data are given as the mean Ϯ S.E., and statistical significance was calculated according to Student's t test with p Ͻ 0.05 as the significant level.

Voltage Dependence of Activation for Wild Type Channels-
The properties of the wild type Ca V 1.2 and Ca V 3.1 were first compared when expressed in oocytes using two-electrode voltage-clamp recordings of calcium channel currents with Ba 2ϩ as charge carrier using identical laboratory conditions. As expected (7, 25), Ca V 3.1 wild-type currents were transient and activated at low voltages, whereas Ca V 1.2 wild-type currents were sustained and high voltage-activating (Fig. 2, A-C). The difference in voltage dependence of activation between these channels can also be seen for the normalized currents ( A and B, sample current traces are shown in this and subsequent figures for voltage steps from a holding potential of Ϫ80 mV to potentials corresponding to near the maxima of the I/V curves. These are shown for Ca V 3.1 with (Ϫ30 mV) and without (Ϫ30 mV) auxiliary subunits and for Ca V 1.2 with (ϩ10 mV) and without (ϩ20 mV) auxiliary subunits. C, I/V curves are shown for the mean values of peak current amplitudes of wild type Ca V 3.1 channel expressed in the absence (f, n ϭ 12) and presence (OE, n ϭ 7) of ␣ 2 ␦/␤ 2 auxiliary subunits. The mean I/V curves for wild type Ca V 1.2 with (q, n ϭ 8) and without (, n ϭ 7) ␣ 2 ␦/␤ 2 are also shown. The mean values of current were fitted by the Boltzmann curves shown. D, using the same experimental data as in C, I/V curves are again plotted with the same symbols but this time using mean values of normalized currents (I norm ). Normalization was to current values at Ϫ30 mV for Ca V 3.1 (with and without ␣ 2 ␦/␤ 2 ), at ϩ10 mV for Ca V 1.2 with ␣ 2 ␦/␤ 2 and ϩ20 mV for Ca V 1.2 without ␣ 2 ␦/␤ 2 . The mean values were again fitted by the Boltzmann curves shown. E, the Boltzmann parameters V 0.5 and k are shown. For this (in contrast to C and D), Boltzmann curves were first fitted to each of the I/V curves for each individual cell, and then mean values were obtained by averaging V 0.5 and k over the number of cells. The data are shown as the mean Ϯ S.E. in this and all subsequent figures; Boltzmann fitting and analysis was also carried out similarly in subsequent figures.
proach to the study of molecular regions that contribute to the differences in activation between these two channels.
When Ca V 3.1 was expressed in the presence of ␣ 2 ␦/␤ 2 auxiliary subunits, there was no change in the voltage dependence of the normalized I/V curves compared with expression of Ca V 3.1 alone (Fig. 2). For the Ca V 1.2 channel, co-expression with auxiliary subunits increased the current by about 2-fold and shifted the I/V curves to the left (Fig. 2). For a clean approach, we would have preferred to carry out experiments throughout in the absence of auxiliary subunits. For the S4 chimeras, we indeed carried out comparisons with wild type in the absence of auxiliary subunits both for test and control. However, for chimeras CGGG, GGCG, and GGGC, comparisons with wild type were mainly made in the presence of subunits because currents were rather small in the absence of subunits, although a very limited number of experiments were also carried out in the absence of subunits. For GCGG, where currents are larger, we carried out detailed comparisons with and without subunits, because the I/II linker (where the ␤ subunit binds) is swapped in this chimera.
Effects of Domain I and Its S4 Segment on Voltage Dependence of Activation-To investigate the role of domain I in determining the differences in voltage dependence of activation between Ca V 1.2 and Ca V 3.1 we have replaced the S1 to S6 region of this domain in Ca V 3.1 with the corresponding domain in Ca V 1.2 to form chimera CGGG (Fig. 1). Upon expression in oocytes (with ␣ 2 ␦/␤ 2 ), the calcium channel currents for this chimera were like Ca V 1.2, and the I/V curves were shifted to higher voltages, similar to the I/V curves for Ca V 1.2 (Fig. 3A), and indeed the Boltzmann parameters V 0.5 and k were not significantly different from those for Ca V 1.2 ( Fig. 3C) but significantly different from those for Ca V 3.1. Therefore, domain I contributes to differences in voltage dependence of activation between these two channels. By analogy with potassium channels (10), the S4 segment in calcium channels may be the most important part of the voltage sensor, so that one might perhaps expect this region to be important in determining differences in voltage dependence in calcium channels. Therefore, we have replaced the S4 segment of domain I in Ca V 3.1 by the corresponding segment in Ca V 1.2. Surprisingly, this substitution did not affect the voltage dependence of activation; the I/V curves remained essentially as for Ca V 3.1 (Fig. 3B), and the Boltzmann parameters were also not significantly different from those for Ca V 3.1. Thus, in summary, the IS4 segment does not contribute to differences in voltage dependence of activation between the two channels; other regions in domain I must, therefore, contribute to these differences.
Effect of Domain II and Its S4 Segment on Voltage Dependence of Activation-The contribution of domain II to determining differences in voltage dependence of activation was studied A sample current trace for IS4C is shown in the inset (voltage step to Ϫ30 mV). C, the Boltzmann parameters V 0.5 and k are shown for CGGG and for IS4C in comparison with appropriate wild type controls. *, significant difference (p Ͻ 0.05) compared with wild type Ca V 3.1. ϩ, significant difference from wild type Ca V 1.2.
using chimera GCGG. More precisely, this chimera has the second domain and the I/II linker in Ca V 3.1 replaced by the corresponding sequence for Ca V 1.2 (Fig. 1). GCGG and wild type channels were first compared in the presence of auxiliary subunits. As compared with Ca V 3.1 wild type channel, the I/V curve was not shifted for the chimera (Fig. 4A); the Boltzmann parameters for the chimera remained as for Ca V 3.1 (Fig. 4D). However, because the auxiliary ␤ subunit binds to the I/II linker in the Ca V 1.2 wild type channel, giving a leftward shift in the I/V curve, in the above experiments the ␤ subunit may have also caused an underlying shift for the chimera (removal of the auxiliary subunits would, therefore, be expected to uncover a shift to the right). Indeed, when experiments were carried out in the absence of auxiliary subunits, there was a shift to the right in the I/V curve for the chimera as compared with Ca V 3.1 but not so far to the right as to be similar to Ca V 1.2 (Fig. 4B, Boltzmann parameters Fig. 4D). Therefore, taken together, the data for this chimera indicate that domain II only contributes to a small extent to the differences in voltage dependence of activation between the two channels. To determine whether this effect is due to the S4 segment of domain II, we replaced this segment in Ca V 3.1 with that for Ca V 1.2. We found that this replacement gave a similar shift in the I/V curves (Fig.  4C) as for chimera GCGG without auxiliary subunits. Also, the Boltzmann parameters were not significantly different from those for chimera GCGG without subunits. This suggests that the (smaller) effect of domain II on voltage differences of activation is due largely if not entirely to the S4 segment.
Effect of Domains III and IV and Their S4 Segments on Voltage Dependence of Activation-The results for domains III and IV were similar and, therefore, are collected together here. For chimeras GGCG and GGGC (in the presence of ␣ 2 ␦/␤ 2 ), the I/V curves were shifted to the right as compared with Ca V 3.1 and in fact were similar to the I/V curves for Ca V 1.2 (Figs. 5A and 6A), with Boltzmann parameters not significantly different from those for Ca V 1.2 (Figs. 5C and 6C) (except for the k value for GGCG, which was somewhat larger than Ca V 1.2). Thus, both domains III and IV contribute strongly to the differences in voltage dependence of activation, as for domain I. Regarding the contribution of the S4 segments in domains III and IV, for the chimeras with S4 segments of Ca V 3.1 replaced by Ca V 1.2, the I/V curves were shifted to the right (as compared with A sample current trace for GCGG (with ␣ 2 ␦/␤ 2 ) with voltage step to Ϫ30 mV is shown in the inset. B, the normalized I/V curve for GCGG in the absence of ␣ 2 ␦/␤ 2 subunits is shown (n ϭ 11 (q)), normalized at Ϫ10 mV (Ϫ0.30 Ϯ 0.03 A). Curves for wild type channels (in the absence of ␣ 2 ␦/␤ 2 ) are again as in Fig. 2D. A sample current trace for GCGG without ␣ 2 ␦/␤ 2 (step to Ϫ10mV) is shown in the inset. C, the normalized I/V curve is shown for IIS4C (n ϭ 7 (), no auxiliary subunits), normalized at 0 mV (Ϫ0.32 Ϯ 0.04 A). Wild type I/V curves (no auxiliary subunits) are shown as before (dashed line, Ca V 3.1; dotted line, Ca V 1.2). A sample current trace for IIS4C (voltage step to 0 mV) is shown in the inset. D, the Boltzmann parameters V 0.5 and k are shown for GCGG with and without ␣ 2 ␦/␤ 2 and for IIS4C in comparison with appropriate wild type controls. ϩ, significant difference (p Ͻ 0.05) from appropriate wild type Ca V 1.2. *, significant difference (p Ͻ 0.05) from appropriate wild type Ca V 3.1. Ca V 3.1 wild type) but not so far to the right as Ca V 1.2 (Figs. 5B and 6B). The Boltzmann parameters indicated only relatively small but significant differences from values for Ca V 3.1 wild type (Figs. 5C and 6C). Taken together, the data suggest that domains III and IV contribute strongly to the difference in voltage dependence of activation but that the S4 segments in these domains do not contribute strongly.
When we carried out limited experiments for the GGCG and GGGC chimeras (as well as for CGGG) in the absence of subunits, we found mean values for V 0.5 of 7 mV (n ϭ 2), 12 mV (n ϭ 2), and Ϫ3 mV (n ϭ 3), respectively, and k values of 15, 16, and 11 mV. These are qualitatively similar to the values found in the presence of subunits. Because the auxiliary subunits had little effect on these chimeras and as little effect was expected because the I/II linker was G rather than C, we did not consider it useful to pursue detailed experiments further with CGGG, GGCG, and GGGC without subunits, particularly given the small size of currents.
Time Courses of Activation and Inactivation-Activation time courses were analyzed, and sample time courses are shown in Figs. 2-6. In some cases, measurements could not be carried out because of interfering capacitative spikes that could not always be reliably subtracted from the relatively small calcium channel currents. Fig. 7A shows that activation times were somewhat smaller for Ca V 3.1 than for Ca V 1.2. For chimera GCGG, activation times were larger than for Ca V 3.1 (Fig.   7A). This suggests that for the kinetics of activation, domain II is also important, in contrast to the results for steady-state activation obtained above. However, for kinetics the data are more difficult to interpret given the small difference in activation time between Ca V 3.1 and Ca V 1.2, and the differences in activation thresholds between the constructs. For the chimera with S4 swapped in domain II, activation times were somewhat larger than for Ca V 3.1 but still smaller than for GCGG. This indicates that IIS4 makes some contribution to activation kinetics but does not explain the whole effect. For the remaining S4 chimeras (Fig. 7B), the time courses of activation were not slower than for Ca V 3.1 (for IS4C, the time course was even faster than for Ca V 3.1 at low voltages). Thus, the data indicate that the S4 regions in domains I, III, and IV do not contribute to slowing the activation kinetics of wild type Ca V 1.2.
As shown in the sample currents in Figs. 3A, 4A and B, 5A, and 6A, currents for chimeras CGGG, GGCG, and GGGC, were non-inactivating, whereas for GCGG there was no fast inactivation. These observations were also found for every single recording that we made for these chimeras. Therefore, all four domains must contribute to the fast inactivation seen with Ca V 3.1. On the other hand, for the S4 chimeras there was always fast inactivation, similar to Ca V 3.1, although there was some variability in inactivation time constants (Fig. 7C), which were always vastly different from the non-inactivating Ca V 1.2. Thus, taken together, the data indicate that the S4 regions in FIG. 5. The role of domain III and its S4 segment in determining differences in voltage dependence of activation between Ca V 3.1 and Ca V 1.2. A, the normalized I/V curve is shown for chimera GGCG (with ␣ 2 ␦/␤ 2 subunits) (n ϭ 6 (OE)) with normalization to the current at ϩ10 mV (Ϫ0.31 Ϯ 0.05 A). The curves for wild type channels (with ␣ 2 ␦/␤ 2 ) are also shown as before (dashed line, Ca V 3.1; dotted line, Ca V 1.2). A sample current trace for the GGCG chimera with a voltage step to ϩ10 mV is shown in the inset. B, the normalized I/V curve is shown for chimera IIIS4C (n ϭ 7 (), no auxiliary subunits) with normalization to the value at Ϫ20 mV (Ϫ0.41 Ϯ 0.06 A). The curves for wild type channels (no auxiliary subunits) are again taken from Fig. 2D (dashed line, Ca V 3.1; dotted line, Ca V 1.2). A sample current trace for IIIS4C is shown in the inset (voltage step to Ϫ20 mV). C, the Boltzmann parameters V 0.5 and k are shown for GGCG and for IIIS4C in comparison with appropriate wild type controls. *, significant difference (p Ͻ 0.05) compared with wild type Ca V 3.1; ϩ, significant difference from wild type Ca V 1.2. all four domains do not contribute to the difference in inactivation kinetics between Ca V 1.2 and Ca V 3.1.

DISCUSSION
This study has investigated the contribution of molecular regions to the difference in voltage dependence of activation between a low and a high voltage-activating channel (Ca V 3.1 and Ca V 1.2, respectively), and we have focused mainly on the steady-state voltage-dependent differences in the I/V curves, where there are striking differences between the two types of channel. Our data showed that, for these steady-state data, replacing domains I, III, and IV of the Ca V 3.1 channel with the corresponding domain for Ca V 1.2 led to high voltage-activated channels (40 -45-mV shifts); for these constructs the I/V curves were similar to those for Ca V 1.2 wild type. However, replacement of domain II of Ca V 3.1 with the corresponding domain of Ca V 1.2 did not give a high voltage-activating channel like Ca V 1.2, although the I/V curve for this chimera was shifted to the right somewhat (around 15 mV) as compared with Ca V 3.1. These results suggest that domains in the calcium channel play different roles in the channel activation process; domains I, III, and IV are apparently critical for channel opening and, therefore, contribute strongly to the difference in voltage dependence of activation between Ca V 3.1 and Ca V 1.2, whereas domain II is less important in regulating the voltage dependence of activation between Ca V 3.1 and Ca V 1.2.
How might these differences in voltage dependence of activation be explained? Current models of ion channel function involve movement of the voltage sensor followed by channel opening (10). Therefore, for the Ca V 3.1 channel, voltage sensors would move at low voltages, whereas for the Ca V 1.2 channel they would move at high voltages. The simplest model would be that the voltage sensors in all four domains should move before the channel opens. In that case, replacing any one of the domains in Ca V 3.1 with Ca V 1.2 should produce a high voltageactivating channel (because the voltage sensor in a single Ca V 1.2 domain in the chimera would not move at low voltages). Our results for chimeras CGGG, GGCG, and GGGC (i.e. domains I, III, and IV) are consistent with this. However, surprisingly, the domain II chimera was not activated at high voltages. The simplest explanation of our results would be that channel opening is produced by the necessary prior movement of the voltage sensors in domains I, III, and IV; if any of these are in the resting position, the channel cannot be opened. The role of the voltage sensor in domain II is less critical, although movement of the voltage sensor in domain II does seem to modulate the voltage dependence of steady-state activation FIG. 6. The role of domain IV and its S4 segment in determining differences in voltage dependence of activation between Ca V 3.1 and Ca V 1.2. A, the normalized I/V curve is shown for chimera GGGC (with ␣ 2 ␦/␤ 2 subunits) (n ϭ 6 (OE)) with normalization to the current at ϩ10 mV (Ϫ0.36 Ϯ 0.05 A). The curves for wild type channels (with ␣ 2 ␦/␤ 2 ) are also shown as before (dashed line, Ca V 3.1; dotted line, Ca V 1.2). A sample current trace for the GGGC chimera with a voltage step to ϩ10 mV is shown in the inset. B, the normalized I/V curve is shown for chimera IVS4C (n ϭ 9 (), no auxiliary subunits), with normalization to the value at Ϫ10 mV (Ϫ0.40 Ϯ 0.09 A). The curves for wild type channels (no auxiliary subunits) are again taken from Fig. 2D (dashed line, Ca V 3.1; dotted line, Ca V 1.2). A sample current trace for IVS4C is shown in the inset (voltage step to Ϫ10 mV). C, the Boltzmann parameters V 0.5 and k are shown for GGGC and for IVS4C in comparison with appropriate wild type controls. *, significant difference (p Ͻ 0.05) compared with wild type Ca V 3.1. ϩ, significant difference from wild type Ca V 1.2. and, indeed, also affects activation kinetics (see the discussion below). To test these ideas we also constructed a chimera with domains II and IV replaced (GCGC), where we might expect high voltage activation because of the domain IV substitution; this indeed was found to be the case (data not shown).
The auxiliary subunits did not give a shift in the voltage dependence of activation for Ca V 3.1 wild type channels, but there was a leftward shift for the wild type Ca V 1.2 channel (see also Refs. 2, 26, and 27). In our present experiments, to disentangle the effect of the auxiliary subunits we always compared test and control under the same conditions, i.e. with or without subunits. This was a particular concern for chimera GCGG where the I/II linker was swapped, because this linker has been implicated (2,15) in ␤ subunit binding for Ca V 1.2, although not for Ca V 3.1. For GCGG, we observed a relatively small shift in the I/V curve by the auxiliary subunits (as compared with the absence of auxiliary subunits), and this was, as expected, to the left rather than in the direction of high voltages, where it would have complicated the interpretation of results. For the other chimeras (CGGG, GGCG, and GGGC), although currents were smaller in the absence of auxiliary subunits, qualitatively similar voltage dependences were observed in the absence of auxiliary subunits as compared with the voltage dependence in the presence of auxiliary subunits. Thus, taken together, the data indicate that the results we have obtained are qualitatively not dependent on the role of the auxiliary subunits.
Because the S4 segment is considered to be the voltage sensor (10,13,14), one might have expected that the S4 regions play an important role in determining voltage dependence of activation. However, when the S4 segment in domain I of Ca V 3.1 was replaced by that of Ca V 1.2, surprisingly, there was no shift in the I/V curves as compared with Ca V 3.1 wild type. Because we have shown that replacement of the whole of domain I of Ca V 3.1 by Ca V 1.2 led to a high voltage-activating channel, the data for S4 replacement suggests that other regions in domain I contribute to the difference in voltage dependence of activation. On the other hand, for domain II, replacement of the S4 region in Ca V 3.1 with Ca V 1.2 gave similar (though not identical) shifts in I/V curves as for replacement of the whole of domain II. This would suggest that the relatively small shifts observed for domain II are entirely due to the S4 segment. Finally, for domains III and IV, the S4 makes a contribution of some 8 or 15 mV to the shift in the I/V curves, but this does not explain the much larger shifts (40 -45 mV) observed for whole domain swapping of domains III and IV. This suggests that other regions of domains III and IV must also contribute to determining differences between the two channels.
Although the time courses of activation of Ca V 1.2 and Ca V 3.1 are different, the magnitude of the difference in activation kinetics is not so striking as the differences in voltage dependence of steady-state activation, so the time course did not lend itself so readily to a detailed chimeric analysis. However, our results suggested that, in contrast to the steady-state results outlined above, domain II might make a contribution to activation kinetics because the GCGG chimera had slower activation than for Ca V 3.1 wild type. There was a partial contribution to this effect from the S4 region in domain II (which also produced some slowing of activation kinetics), but this did not explain the whole effect of the GCGG chimera. The apparent effect of the GCGG chimera in producing a slowing of activation could arise from other parts of domain II, such as the pore region. We might also perhaps expect slowing of activation kinetics for the chimeras of the other domains, but the small size of the chimeric currents did not allow detailed analysis. However, we did observe that, for the S4 chimeras of domains I, III, and IV, there was no contribution to slowing of activation of Ca V 3.1 (IS4 was even faster), again indicating that the S4 regions in these domains do not primarily determine the differences in activation kinetics between the two wild type channels.
Although the S4 segment is of primary importance in gating, other regions must also contribute in determining the voltage dependence. For instance, the S2 and S3 regions also form the voltage sensor, at least for potassium channels (28 -31), so it is possible that these regions may also be involved in voltagesensing in calcium channels and, hence, may determine differences in voltage dependence of activation between high voltage-activating and low voltage-activating calcium channels. Fur- FIG. 7. Activation and inactivation times for chimeras and wild type channels. A, activation time, t 20 -80 , is plotted against test potential, V, for wild type Ca V 3.1 channels with (ࡗ, n ϭ 7) and without (f, n ϭ 6) auxiliary subunits for wild type Ca V 1.2 with auxiliary subunits (q, n ϭ 4), chimera GCGG with (OE, n ϭ 7) and without (, n ϭ 4) auxiliary subunits, and for chimera IIS4C without auxiliary subunits (ϫ, n ϭ 5). B, activation time, t 20 -80 , is plotted against test potential, V, for wild type Ca V 3.1 (f, n ϭ 7) and chimeras IS4C (q, n ϭ 6), IIIS4C (, n ϭ 5), and IVS4C (ࡗ, n ϭ 5), all in the absence of auxiliary subunits. C, inactivation time, inact , is plotted against test potential, V, for wild type Ca V 3.1 (f, n ϭ 7) and chimeras IS4C (q, n ϭ 6), IIS4C (OE, n ϭ 5), IIIS4C (, n ϭ 5), and IVS4C (ࡗ, n ϭ 5). thermore, point mutations in the pore (P) region of Ca V 3.1 showed that shifts in I/V curves can occur (32,33). Thus, it will be interesting in future studies to investigate the possible role of the pore helices S5-P-S6 as well as the voltage sensor domain S2/ S3/S4 in determining differences in voltage dependence of activation. Other regions that might contribute are the intracellular linkers between the domains; our chimeras GCGG, GGCG, and GGGC also involved swapping linkers (I/II, II/III, III/IV respectively). However, isoforms of Ca V 3.1 that are alternatively spliced in the II/III and III/IV linker do not show shifts in IV curves that are greater than some 5 mV, suggesting that these linkers do not contribute to determining whether the channel is low or high voltage-activating (34). Also, by studying other splice variants of Ca V 3.3 (35), it similarly appears that the C terminus does not contribute to the voltage dependence of activation.
Another possible reason why the S4 regions did not contribute much to determine voltage-dependent differences in activation could be because of the marked homology of the S4 regions that were swapped in our chimeras. The alignments are shown in Fig. 8 and compared with the S4 region of the Shaker potassium channel. For the latter channel, the key region, which contributes to gating charge movement and is involved in movement across the membrane upon gating (10), involves charged residues 362, 365, 368, and 371 (Shaker residue numbering, Fig. 8). Over this latter region there is even more homology between S4 regions of the different domains of the two calcium channels studied here. In particular, there is the most similarity between the S4 of domains I and III and between domains II and IV. This mirrors our results where S4 chimeras in domains I and III gave no shift or a small shift in I/V curves, whereas S4 chimeras in domains II and IV gave larger shifts.
For calcium channels, unlike potassium and sodium channels, there are few papers that study the molecular mechanism for channel activation. Garcia et al. (13) made point mutations of charged residues in S4 segments of domains I-IV of a high voltage-activated L-type calcium channel. They showed that significant changes in activation properties were obtained for S4 regions of domains I and III rather than domains II and IV, indicating a differential role for these domains in activation. Furthermore, S4 regions of domains I and III each possess a proline residue (Fig. 8), and point mutations at these residues also show greater effects on activation than for mutations of the corresponding residues of domains II and IV (14). Again, this led to the conclusion that S4 regions in domains I and III contribute more to channel opening than domains IIS4 and IVS4. From our present results, we have also shown qualitative differences between the effects of S4 chimeras in domains I and III from the effects in IIS4 and IVS4. However, our results are quantitatively different from the above papers (we found smaller shifts for IS4 and IIIS4), which probably reflects the fact that we did not change either charged or proline residues in our S4 chimeras.
There have been more previous studies on sodium channel activation than for the calcium channel, and it is interesting to compare our results with those for the sodium channel, which has a similar molecular structure formed from four homologous domains I-IV. As for the calcium channel, mutation of charged residues of the S4 segments in the four domains of the sodium channel shows that each S4 segment does not contribute equally to the voltage-dependent properties of the channel (16 -18). Further studies of gating currents and of S4 movement using fluorescent tags indicate that S4 segments of domains I, II, and III of the sodium channel move first upon depolarization, leading to an open state; the IVS4 moves later to reveal a further open state. The IVS4 segment is then strongly involved in inactivation (36,37). Our results in the present study suggest that the calcium channel may have some similarities with the activation of the sodium channel. Because we have shown that regions in domains I, III, and IV of the calcium channel are critical for channel opening, we suggest that the S2-S4 voltagesensor regions of domains I, III, and IV may activate first in parallel, then the voltage sensor in domain II may be triggered to activate. Thus, domain II is less important in determining the voltage dependence of steady-state calcium channel activation, although it can be important in modulating the kinetics of activation. It will be very interesting to test this hypothesis by further experiments using fluorescence techniques.
Finally, as regards inactivation, we have shown that replacing any one of the domains I-IV of Ca V 3.1 (fast inactivating) by Ca V 1.2 (non-inactivating) led to a chimeric channel that was not fast-inactivating. Therefore, all four domains I-IV of the calcium channel seem to participate in the mechanism for fast inactivation of the Ca V 3.1 calcium channel. This is consonant with previous work for many calcium channel types showing that regions from all four domains contribute to inactivation (38 -43). Indeed, regions at the C terminus and segment IIIS6 of the Ca V 3.1 channel have already been shown to be involved in inactivation (42,43). We have also shown that replacement of each of the S4 segments in domains I-IV of Ca V 3.1 by Ca V 1.2 does not remove the fast inactivation, although there was some modulation of the inactivation time courses. The fact that the chimeric S4 substitutions did not contribute markedly to the inactivation could again be because of their marked homology, or more likely, the data suggest that the S4 segments have no obvious role in inactivation of the calcium channel.
In summary, by domain-swapping experiments between the low voltage-activating channel Ca V 3.1 and the high voltageactivating channel Ca V 1.2 channel, we have shown that domains I, III, and IV rather than domain II are of key importance in determining the difference in voltage dependence of activation between these two channels. Chimeras with the S4 segments swapped did not explain the whole effects, suggesting roles for segments S1-S3 and/or S5-S6. FIG. 8. Sequence alignment of the S4 segments in domains I-IV of Ca V 3.1 and Ca V 1.2. The sequence alignment of the S4 segments is shown for Ca V 3.1 (labeled IS4-G to IVS4-G) and for Ca V 1.2 (labeled IS4-C to IVS4-C) together with the S4 region of the Shaker potassium channel. The charged residues that are important for gating in Shaker are indicated with the residue numbers corresponding to the Shaker sequence.