Functional Properties of CaV1.3 (α1D) L-type Ca2+ Channel Splice Variants Expressed by Rat Brain and Neuroendocrine GH3 Cells*

Ca2+ enters pituitary and pancreatic neuroendocrine cells through dihydropyridine-sensitive channels triggering hormone release. Inhibitory metabotropic receptors reduce Ca2+ entry through activation of pertussis toxin-sensitive G proteins leading to activation of K+channels and voltage-sensitive inhibition of L-type channel activity. Despite the cloning and functional expression of several Ca2+ channels, those involved in regulating hormone release remain unknown. Using reverse transcription-polymerase chain reaction we identified mRNAs encoding three α1(α1A, α1C, and α1D), four β, and one α2-δ subunit in rat pituitary GH3 cells; α1B and α1Stranscripts were absent. GH3 cells express multiple alternatively spliced α1D mRNAs. Many of the α1D transcript variants encode “short” α1D (α1D-S) subunits, which have a QXXER amino acid sequence at their C termini, a motif found in all other α1 subunits that couple to opioid receptors. The other splice variants identified terminate with a longer C terminus that lacks the QXXER motif (α1D-L). We cloned and expressed the predominant α1D-S transcript variants in rat brain and GH3 cells and their αlD-L counterpart in GH3 cells. Unlike α1A channels, α1D channels exhibited current-voltage relationships similar to those of native GH3 cell Ca2+channels, but lacked voltage-dependent G protein coupling. Our data demonstrate that alternatively spliced α1Dtranscripts form functional Ca2+ channels that exhibit voltage-dependent, G protein-independent facilitation. Furthermore, the QXXER motif, located on the C terminus of α1D-S subunit, is not sufficient to confer sensitivity to inhibitory G proteins.

All excitable cells express voltage-activated Ca 2ϩ channels fulfilling diverse cellular functions. Among other things, Ca 2ϩ influx can regulate gene expression and propagate action potentials, and it is a prerequisite for sustained neurotransmitter and hormone release (1,2). Specialized Ca 2ϩ channels have evolved to serve these specific purposes. Ca 2ϩ channels native to excitable cells can be classified into five categories (T, L, N, P/Q, and R) on the basis of their unique biophysical and pharmacological properties. The genetic basis for the heterogeneity of Ca 2ϩ channels is apparent now that genes have been identified that encode ten ␣ 1 (␣ 1A through ␣ 1I , and ␣ 1S ) subunits, four ␤ subunits and three ␣ 2 -␦ dimers (3). Ca 2ϩ channel ␣ 1 subunits have been grouped into three families (Ca V 1, 2, and 3) on the basis of their levels of amino acid sequence identity (4). Additional structural heterogeneity occurs through alternative splicing (5)(6)(7)(8)(9). Expression of recombinant homomeric ␣ 1 subunits in Xenopus oocytes or cell lines produces functional Ca 2ϩ channels with properties reminiscent of those of their naturally expressed counterparts. Although the ␣ 1 subunit is the principle component of the Ca 2ϩ channel, containing 24 membranespanning regions that form both the channel and the voltage sensor, coexpression of ␣ 1 subunits with ␤ and ␣ 2 -␦ subunits produces a greater current density and alters voltage dependence, current amplitude, activation, and inactivation kinetics (10 -12).
High voltage-activated P/Q-and N-type Ca 2ϩ channels have been extensively studied because of their central role in controlling Ca 2ϩ entry into, and thus neurotransmitter release from, neuronal synaptic terminals (13). The activity of these channels is inhibited by the activation of metabotropic receptors that couple to pertussis toxin-sensitive G proteins. This pathway is thought to be an important means of presynaptic inhibition, allowing the feedback control of neurotransmitter release. Following metabotropic receptor activation, ␤␥ subunits liberated from pertussis toxin-sensitive G i/o protein complexes bind directly to P/Q-and N-type Ca 2ϩ channels at their QXXER amino acid motifs making them less "willing" to open when stimulated by an action potential invading the synaptic terminal (14 -18). The inhibition is voltage-dependent; prolonged (19) or repetitive (20) depolarization can overcome ␤␥ blockade.
Inhibitory metabotropic receptors can also attenuate the activity of L-type Ca 2ϩ channels in several neuroendocrine cells and some neurons (21,22). In the growth hormone-and prolactin-secreting anterior pituitary GH 3 cell line, activation of either native somatostatin and muscarinic receptors or recombinant opioid receptors leads to inhibition of L-type Ca 2ϩ channel activity (23)(24)(25). This effect can be prevented by pertussis toxin pretreatment or reversed by strong depolarization, suggesting the involvement of ␤␥ subunits. Voltage-dependent coupling between metabotropic receptors and cloned P/Q-and N-type channels can be reconstituted in recombinant expression systems by expressing ␣ 1A or ␣ 1B subunits, respectively, with the inclusion of auxiliary ␤ and ␣ 2 -␦ subunits. For example, cloned opioid receptors couple to Ca 2ϩ channels containing ␣ 1A or ␣ 1B Ca 2ϩ channel subunits expressed in Xenopus oocytes (26). By contrast, expression of the ␣ 1C subunit forms L-type Ca 2ϩ channels that are insensitive to opioid receptor activation.
The prevalence of ␣ 1D subunit expression in neuroendocrine cells led to the suggestion that this may be the molecular identity of the G protein-coupled L-type channel (21,22). Fur-thermore, the existence of a QXXER amino acid motif in an ␣ 1D subunit splice variant led us to further hypothesize that this may be necessary for direct G protein-mediated L-type channel inhibition (27). In this study we used RT-PCR 1 to identify Ca 2ϩ channel transcripts in GH 3 cells. In addition to ␣ 1A and ␣ 1C transcripts, we found numerous alternatively spliced ␣ 1D transcripts. The predicted amino acid sequences of the ␣ 1D -splice variants terminated with either a long (␣ 1D-L ) or a short (␣ 1D-S ) C terminus, the latter of which contained a QXXER amino acid motif. We used long range RT-PCR to compare the relative prevalence of alternatively spliced ␣ 1D transcripts in GH 3 cells and rat brain and created cDNAs encoding ␣ 1D-S and corresponding ␣ 1D-L splice variants for electrophysiological analyses. Using these clones we examined the functional properties of several alternatively spliced ␣ 1D isoforms and determined whether the QXXER motif is sufficient to confer sensitivity of the ␣ 1D subunit to G proteins either activated directly by GTP␥S or through opioid receptor activation.

EXPERIMENTAL PROCEDURES
Cultures and Transfections-Control GH 3 cells (ATCC, Manassas, VA) and GH 3 cells were stably transfected withand ␦-opioid receptors (GH 3 MORDOR cells) as described previously (25). Both cell lines were grown in Dulbecco's modified Eagle's medium containing 10% v/v calf serum, penicillin (100 units/ml) and streptomycin (100 g/ml). GH 3 MORDOR cells were grown under positive selection conditions with hygromycin (200 g/ml) and Geneticin (400 g/ml). Cells were subcultured once each week and seeded into 75-cm 2 flasks, for maintenance, and into 35-mm diameter dishes for use in electrophysiological experiments. Human embryonic kidney (HEK) cells and Chinese hamster ovary (CHO) cells stably transfected with either ␦-(HEKDOR and CHODOR cells) or -(HEKMOR and CHOMOR cells) opioid receptors, a gift from Dr. C. Evans (Department of Psychiatry, UCLA, Los Angeles, CA), were grown in the medium described above for GH 3 cells under positive selection with Geneticin (400 g/ml). Cells were subcultured 24 h prior to transfection. cDNAs encoding ␣ 1D , ␣ 1A provided by Dr. Y. Mori (National Institute for Physiological Sciences, Okazaki, Japan), ␤ 2a provided by Dr. E. Perez-Reyes (University of Virginia, Charlottesville, VA), and ␣ 2 -␦ provided by Dr. L. Birnbaumer (UCLA) were transiently transfected in combination with green fluorescence protein cDNA into HEK or CHO cells by either the CaPO 4 precipitation or electroporation technique. In some experiments, cells were also transfected with cDNAs encoding G ␣oA or G ␣oB provided by Dr. M. Simon (California Institute of Technology, Pasadena, CA), ␤-ARK minigene provided by Dr. R. Lefkowitz (Duke University, Durham, NC), or syntaxin-1A provided by Dr. G. Zamponi (University of Calgary, Calgary, Canada). Cells were incubated in a humid atmosphere of 5% CO 2 and 95% air at 37°C.
Patch-clamp Recordings-Green fluorescence protein-expressing transiently transfected cells were located using fluorescence microscopy (Nikon Diaphot, Image Systems Inc., Columbia, MD). The whole-cell patch-clamp (Axopatch 200A amplifier, Axon Instruments Inc., Foster City, CA) technique was used to record Ca 2ϩ channel activity with Ba 2ϩ as the charge carrier. Cells were initially bathed in a solution containing (in mM) NaCl 140, KCl 4.7, MgCl 2 1.2, CaCl 2 2.5, and HEPES 10 (pH 7.2). After establishing the whole-cell configuration, the bath solution was replaced by one comprised of (in mM): NaCl 125, CsCl 5.4, BaCl 2 10.8, MgCl 2 1, and HEPES 10. In all cases electrodes contained (in mM): CsCl 120, EGTA 10, MgCl 2 1, Mg-ATP 3, and HEPES 10 (pH 7.2). The potential difference between the open electrode and the bath ground was zeroed prior to establishing a Ն1 G⍀ resistance seal. The liquid junction potential was negligible, and no compensation was made for its cancellation. Unless otherwise stated voltage-activated Ba 2ϩ currents were recorded from cells depolarized from the holding potential of Ϫ80 to 0 mV for 80 ms at 10-s intervals. The voltage dependence of current inhibitions induced by opioids or GTP␥S (300 M) was investigated using a double-pulse protocol: cells were depolarized from a holding potential of Ϫ80 mV to between Ϫ50 and 60 mV (10-mV increments, 95-ms duration). The pre-pulse was followed (after 10 ms at Ϫ80 mV) by a 10-ms pulse to 0 mV. The effect of a depolarizing prepulse on the current-voltage relationship was investigated by applying a 60-mV pulse for 26 ms between two depolarizing test pulses (from Ϫ80 mV to between Ϫ50 and 60 mV in 10-mV increments). Time required for recovery from voltage-dependent facilitation was examined by depolarizing cells from Ϫ80 to 60 mV followed by 10 -140 ms (in 10-ms increments) by a test pulse to 0 mV. Currents were low-pass-filtered at 2 kHz and digitized (Digidata, Axon Instruments Inc., CA) at 10 kHz for storage on the hard drive of a Pentium PC. Tail currents were recorded using Sylgard (Dow Corning Corp., Midland, MI)-coated borosilicate glass (Garner Glass Co., Claremont, CA). Tail currents generated by repolarization to Ϫ80 mV following 5.8-ms depolarizations to voltages between Ϫ60 and 60 mV were filtered at 10 kHz and digitized at 200 kHz for storage on a PC hard disc. Experiments were performed at room temperature (22-24°C).
Curve Fitting and Statistics-Tail currents were fitted by single exponentials using pCLAMP8 software (Axon Instruments Inc, CA). Tail current amplitudes were derived by extrapolating the fit to the point of hyperpolarization. There was no relationship between test pulse potential and the time constant () of the tail currents. The relationship between tail current amplitude and the voltage of depolarization was fitted with a Boltzmann equation of the form, where I tail is the tail current amplitude as a percentage of maximum, V is the command voltage, V1 ⁄2 is the voltage for 50% maximal activation, and k is the slope factor. Concentration-response curves were fitted using the logistic equation, where I is the Ba 2ϩ current amplitude in the presence of a specific concentration of nimodipine expressed as percent control, E max is the maximal percent inhibition of the current, IC 50 is the concentration of nimodipine that had a half-maximal effect, and n is the slope. Data obtained from the voltage-dependent reversal of Ba 2ϩ current inhibitions were fitted with the Boltzmann equation, where I is the Ba 2ϩ current amplitude as a percentage of control current amplitude, F max is the maximum percent facilitation of the current amplitude, x is the prepulse potential, F 50 is the prepulse potential required for half-maximal facilitation, and S is the slope factor. In these experiments control amplitude was the amplitude of the current activated by stepping from Ϫ80 to 0 mV after a prepulse to Ϫ50 mV. Data are expressed as means Ϯ S.E., and statistical significance was established using the Student's t test.
Drugs Used-Pertussis toxin (from Sigma Chemical Co., St. Louis, MO) was applied to the culture medium at a final concentration of 100 ng/ml for at least 24 h before experiments. All other drugs were diluted into the extracellular solution that constantly perfused the recording chamber through gravity feed at a rate of ϳ5 ml/min. [D-Pen 2 ,D-Pen 5 ]enkephalin (DPDPE) was obtained from Peninsula Laboratories (Belmont, CA). GTP␥S (Life Technologies, Inc., Gaithersburg, MD) was loaded into the recording electrode at a final concentration of 300 M. Recordings were made ϳ3 min after achieving the whole-cell configuration. All tissue culture reagents, including antibiotics were obtained from Life Technologies, Inc. (Gaithersburg, MD).
RT-PCR-RNA was prepared from whole rat brain and GH 3 cells (passage 12) and reverse-transcribed using random hexamers or poly-dT. Specific oligonucleotides were designed for ␣ 1 subunits (␣ 1B-1E , ␣ 1S ), ␤ subunits (␤ 1-4 ), and the ␣ 2 -␦ subunit. GenBank accession numbers, forward and reverse primers, respectively, are as follows: The specific forward and reverse primers used to identify the ␣ 1A-b splice variant were 5Ј-CTCAACACCATCGTGCTAATGATG-3Ј and 5Ј-CAGGTTGATGAAGTTATTCGGATT-3Ј, respectively. We used the same sense and two different antisense oligonucleotides to examine whether the three known ␣ 1D subunit C-terminal splice variants (28) are expressed by GH 3 cells (Table I). Primers specific to exons of the rat ␣ 1D subunit gene are also listed in Table I. Amplification was carried out at 35 cycles with Pfu polymerase (Stratagene, La Jolla, CA). Each cycle was comprised of 30 s at 95°C, 1 min at 55°C, and 2 min at 72°C. For the last cycle, annealing and extension times were increased to 3 and 10 min, respectively. Long-range RT-PCR amplification was carried out at 40 cycles with a Taq/Pfu polymerase mix (Stratagene, La Jolla, CA). The extension cycles were increased to 9 min at 72°C, and the final extension time was 15 min. Q solution (Qiagen, Valencia, CA) was added to the PCR mix to obtain a maximal yield of product. For exon-specific PCR, individual clones of plasmid DNA were used as template, and amplification was carried out at 20 cycles with Taq polymerase (Stratagene). Each cycle consisted of 30 s at 95°C, 60 s at 55°C, and 45 s at 72°C.
Determination of Relative Abundance of Alternatively Spliced ␣ 1D -Subunit Transcripts-We used either RT-PCR or a hybridization strategy to examine the number and relative abundance of ␣ 1D subunit splice variants encoded by full-length ␣ 1D transcripts in GH 3 cells and rat brain. By using an oligonucleotide primer set designed to the 5Ј-and 3Ј-untranslated regions, respectively (Table I), full-length ␣ 1D-S transcripts encoding subunits with the short C terminus were amplified by RT-PCR generating a 5.1-kb product. For the full-length transcripts encoding ␣ 1D subunits with the long C terminus (␣ 1D-L ) we used the same sense primer to the 5Ј-untranslated region and an antisense oligonucleotide primer directed to the coding region 3Ј to the C-terminal splice locus, generating a 5.2-kb product. Due to the existence of additional splice sites, amplification products represented transcripts encoding multiple ␣ 1D-S and ␣ 1D-L isoforms. Subcloning PCR products en masse provided a means by which to determine the abundance of each particular variant. Upon transformation into an appropriate bacterial host, individual clones represented single isoforms and analysis of each provided the splice variation of the two upstream regions of known splicing activity. A PCR-based strategy was employed to analyze individual clones by designing exon-specific primers for each exon within the sites of splicing activity (Table I). The PCR products for each clone were visualized by gel electrophoresis and representative clones for each isoform were confirmed by sequencing. Therefore, by analyzing 100 individual clones, a percent abundance for each isoform could be obtained. Alternatively, we used dot blots of individual splice variant clones and tested the presence or absence of individual exons by using ␥-32 P-labeled exon-specific primers. Oligonucleotides were end-labeled with [␥-32 P]dATP using T4 polynucleotide kinase (Ambion, Austin, TX), and 1.0 ϫ 10 6 dpm were added to the hybridization solution. Hybridization was carried out at 55°C in 1 M NaCl, 50 mM Tris (pH 7.5), 5ϫ Denhardt's, 0.1% sodium pyrophosphate, 0.5% SDS, and 10% polyethylene glycol 8000 for a minimum of 4 h. The blots were washed three times in 2ϫ SSPE, 0.5% SDS at room temperature and autoradiography was carried out at Ϫ80°C.
Creating Clones of ␣ 1D Subunit Splice Variants-The long-range PCR products were digested by EcoRI generating multiple fragments with sizes of ϳ3.3 kb. These fragments were ligated into pBluescript SK(Ϫ) (Stratagene, La Jolla, CA) to generate pBlue Q.PCR (x) (where x represents a specific splice variant). A rat ␣ 1D subunit clone (Gen-Bank accession number D38101), kindly provided by Dr. S. Seino (Chiba University, Chiba, Japan), digested partially with BamHI and fully with XhoI, was subcloned into pcDNA3.1ϩ (Invitrogen, Carlsbad, CA) generating C1.0. To generate a nucleotide sequence encoding the ␣ 1D-S C terminus, RT-PCR was performed on GH 3 cell RNA using amplimers flanking DraIII and XhoI sites. The DraIII/XhoI-cut PCR product was ligated into C1.0, generating Q1.0. Both Q1.0 and C1.0 were digested with EcoRI, gel-purified, and re-ligated to remove the 3.3-kb EcoRI fragment, generating Q1.1 and C1.1. The constructs ␣ 1D-S(x) and ␣ 1D-L(x) were generated by ligating the EcoRI fragment of pBlue Q.PCR (x) into Q1.1 and C1.1, respectively. Sequencing revealed that cDNAs generated by long-range PCR often contained nucleotide sequence errors that would introduce incorrect amino acids into recombinant channels. Full-length ␣ 1D cDNAs were constructed from clones that had minimal sequence errors. Initial electrophysiological recordings from recombinant ␣ 1D-S channels made use of these clones. Subsequently, experiments examining current-voltage relationships, facilitation, and G protein coupling were repeated (n Ն 3) demonstrating similar functional properties for sequence error-free recombinant ␣ 1D-S subunits. A construct that encoded an error-free ␣ 1D-S(31b) subunit was generated by ligating its 2-kb NotI/AgeI fragment to the 8-kb NotI/AgeI fragment of an error-free ␣ 1D-S clone identical in sequence 3Ј to the AgeI restriction site. The 3.3-kb EcoRI fragment of the ␣ 1D-S(31b) was then ligated to C1.1 to generate ␣ 1D-L(31b) . The 3.2-kb BamHI fragment of ␣ 1D-S(31b) was ligated to a 6.8-kb BamHI fragment of an ␣ 1D-S(31a) clone to generate a construct that encoded an error-free ␣ 1D-S(31a) subunit.
DNA Sequencing-Appropriate length PCR products were excised from ethidium bromide-stained gels and purified with QIAEX beads (Qiagen, Valencia, CA) according to the manufacturer's instructions.
Purified DNA was subcloned into Bluescript SK(Ϫ) by blunt-end ligation and sequenced manually by 32 P-labeled dideoxynucleotide incorporation (Sequenase II, Amersham Pharmacia Biotech) and polyacrylamide gel electrophoresis. We used PCR-based cycle sequencing (FS chemistry; PE Biosystems, Foster City, CA) with custom internal primers (Life Technologies, Inc., Gaithersburg, MD) to sequence the ϳ3.3-kb PCR products subcloned into pBlue Q (i.e. pBlue Q.PCR (x) ). Sequences were obtained using an ABI Prism 377 genetic DNA analyzer, and data were analyzed using ABI Prism DNA sequencing software (PE Biosystems).

RESULTS
Opioid Receptors Couple to -Agatoxin IVa-resistant Ca 2ϩ Channels in GH 3 MORDOR Cells-Ca 2ϩ enters GH 3 cells through dihydropyridine-sensitive channels that can be inhibited by the activation of pertussis toxin-sensitive G proteins (24,25,29). Taken together with similar studies in pancreatic cell lines and neurons, these observations indicate the existence of G protein-sensitive L-type Ca 2ϩ channels (21). However, because G i /G o proteins inhibit P/Q-and N-type channels (18), we re-examined the possibility that GH 3 cells may express either of these Ca 2ϩ channel subtypes (Figs. 1 and 2). The P/Q-type channel antagonist -agatoxin IVa (10 nM) had no discernible effect on Ba 2ϩ currents activated by depolarizing GH 3 cells stably expressing and ␦ opioid receptors (GH 3 MORDOR cells) (Fig. 1A). We used RT-PCR to examine whether GH 3 cells express transcripts encoding the ␣ 1A subunit, the principle component of P/Q-type channels (Fig. 1B). Transcripts encoding ␣ 1A subunits can be alternatively spliced giving rise to ␣ 1A-a and ␣ 1A-b subunits that may represent Pand Q-type Ca 2ϩ channels, respectively (6). Low concentrations of -agatoxin IVa inhibit recombinant Ca 2ϩ channels containing ␣ 1A-a subunits, whereas higher concentrations are required to inhibit channels formed by ␣ 1A-b subunits. We identified mRNA encoding the ␣ 1A-b subunit in GH 3 cells using RT-PCR analysis with a primer (see "Experimental Procedures") selective for this splice variant (Fig. 1B). In view of the presence of mRNA encoding the ␣ 1A-b subunit, we examined a potential role for -agatoxin IVa-sensitive channels in G protein-mediated inhibition of Ca 2ϩ channel activity in GH 3 MORDOR cells. Despite an inhibition of Ca 2ϩ channel activity induced by -agatoxin IVa (500 nM), the ␦ opioid receptor-selective agonist DPDPE (100 nM) continued to inhibit Ba 2ϩ currents. Therefore, opioid receptors couple to -agatoxin-resistant Ca 2ϩ channels in GH 3 MORDOR cells (Fig. 1C). In later experiments we examined the properties of recombinant ␣ 1A channels providing further evidence for a lack of P/Q-type channels in GH 3 cells (Fig. 7).
Having ruled out a role for P/Q Ca 2ϩ channels in the G protein modulation of Ba 2ϩ currents recorded from GH 3 MORDOR cells, we turned our attention to N-type channels. Electrophysiological studies demonstrate that GH 3 cells lack functional N-type Ca 2ϩ channels (24). Furthermore, Lievano and colleagues (30) were unable to detect the ␣ 1B transcript in these cells using the ribonuclease protection assay. We confirmed a lack of mRNA encoding the ␣ 1B subunit in these cells using RT-PCR (Fig. 2). GH 3 Cells Contain Multiple ␣ 1 subunit Transcripts-We synthesized cDNAs from RNA extracted from GH 3 cells, whole rat brain and rat skeletal muscle, using PCR to determine the complement of transcripts encoding Ca 2ϩ channel ␣ 1 subunits and accessory ␤ and ␣ 2 -␦ subunits (Fig. 2). We did not examine whether GH 3 cells express transcripts encoding T-type Ca 2ϩ channels, because several studies have demonstrated their resistance to G protein modulation (31,32). Furthermore, Ba 2ϩ currents recorded from GH 3 cells lack a consistent T-type channel contribution (24). Specific forward and reverse oligonucleo-tide primers were designed for five cloned high voltage-activated ␣ 1 subunits (␣ 1B-1E , ␣ 1S ), ␤ subunits (␤ 1-4 ), and an ␣ 2 -␦ subunit (see "Experimental Procedures"). Sequencing of the bands excised from gels positively identified mRNAs encoding ␣ 1C and ␣ 1D subunits (confirming a previous report (30)) and ␤ 1-4 and ␣ 2 -␦ subunits in GH 3 cells (Fig. 2). Because ␣ 1B , ␣ 1E , or ␣ 1S transcripts were absent and ␣ 1A channels do not contribute to the effects of DPDPE (Fig. 1), G protein-coupled inhibition of Ca 2ϩ channels in GH 3 cells appears to occur through regulation of ␣ 1C and/or ␣ 1D channel activity. We considered the ␣ 1C subunit to be an unlikely candidate because of its inability to couple to opioid receptors (26). Therefore, we focused our attention on the ␣ 1D subunit believing this to be the most likely candidate for the G protein-sensitive L-type Ca 2ϩ channel expressed by GH 3 cells.
Multiple ␣ 1D Splice Variants in GH 3 Cells-Multiple splice variants of the ␣ 1D Ca 2ϩ channel subunits exist in various cell types (7,28,33). In rat tissue, splicing causes variations in amino acid sequences located at three distinct loci: 1) between domains I and II, 2) between the S2 and S4 regions of domain IV, and 3) in the C-terminal region (Fig. 3). We used RT-PCR with oligonucleotide primer sets ( Table I) spanning each of the three regions to investigate the presence or absence of splice variants in GH 3 cells (see "Experimental Procedures"). Gel extraction of candidate bands and subsequent sequencing confirmed multiple splice variants at all three splicing loci, resulting in numerous combinatorial ␣ 1D subtype possibilities (Fig.  3). Interestingly, analysis of the amino acid sequences in the C termini of ␣ 1D subunits indicated the presence of a truncated isoform containing the Gln-X-X-Glu-Arg (QXXER) motif, which has been observed previously in rat brain (33). We refer to ␣ 1D subunit variants that contain this truncated C terminus as ␣ 1D-S subunits. The QXXER motif is absent from the longer alternatively spliced ␣ 1D-L isoforms (Fig. 3). The fact that the QXXER sequence is present in ␣ 1A , ␣ 1B , and ␣ 1E subunits, which can all couple to inhibitory G proteins (15,16), led to our hypothesis that this motif is sufficient to confer G protein sensitivity to recombinant channels formed by expressing  (Table I) demonstrates the presence of the ␣ 1A-b transcript. C and D, the ␦ opioid receptor agonist DPDPE (100 nM) inhibited Ba 2ϩ currents evoked by depolarizing GH 3 MORDOR cells from Ϫ80 to 0 mV (every 10 s). The inhibition was unaffected by the application of -agatoxin IVa (500 nM) at a concentration sufficient to block ␣ 1A-b Ca 2ϩ channels (n ϭ 3).
Relative Abundance of Alternatively Spliced ␣ 1D Isoforms-We used RT-PCR with exon-specific primers (Table I) and a hybridization strategy to examine the number of ␣ 1D subunit mRNA variants caused by the combinatorial possibilities enabled by virtue of the existence of the three known splicing hotspots depicted in Fig. 3 (see "Experimental Procedures"). This process was performed first using RNA isolated from GH 3 cells in which ␣ 1D-S and ␣ 1D-L variants were examined and subsequently, for ␣ 1D-S variants specifically, using rat whole brain RNA (Fig. 5). Bands containing a mixture of all cDNAs encoding ␣ 1D-S or ␣ 1D-L subunits were excised from agarose gels (Fig. 4A). We used EcoRI to excise cDNA sequences encoding the two initial regions of splicing activity. These were then ligated into pBluescript allowing transformation of bacteria. We picked 100 colonies of bacteria transformed with the vector containing EcoRI fragments of either ␣ 1D-S from GH 3 cells or rat brain, or ␣ 1D-L from GH 3 cells. We then employed PCR or exon-specific hybridization, with specific oligonucleotides (Table I) to identify exons introduced by alternative splicing into cDNAs isolated from individual bacterial colonies (Fig. 4B).
A failure of Ba 2ϩ currents mediated by ␣ 1D channels to reverse (presumably due to low Cs ϩ permeability) prevented accurate determinations of voltages for half-maximal activation by fitting the current-voltage relationships in Fig. 6B. Therefore, we performed tail current analysis to determine the voltage required for half-maximal channel activation (V1 ⁄2 ) for currents mediated by ␣ 1D-S(31a) and ␣ 1D-S(31b) subunits. Tail currents were well fitted by single exponentials with time constants of 0.2 ms in both cases (Fig. 6C). The relationships between tail current amplitude and voltage for ␣ 1D-S(31a) and ␣ 1D-S(31b) subunits were similar. Fitting the data points with a Boltzmann function (see "Experimental Procedures") yielded

FIG. 3. Summary of the ␣ 1D subunit splice variants detected in GH 3 cells.
Three separate coding regions of the ␣ 1D Ca 2ϩ channel subunit were investigated for the presence of splice variants by RT-PCR. Exons 10 -13 correspond to the intracellular loop between domain I and II. Exons 30 -33 correspond to a region between S1 and S5 of domain IV. Exons 40 -49 correspond to the cytoplasmic tail. These three regions are indicated by a, b, and c, respectively. The identity of each PCR product was confirmed by sequencing. *, this splice fragment (c2) contains the putative G protein ␤␥ subunit-binding motif, whereas the other fragment, c1, does not.
302 a All starting positions of primers are based on rat ␣ 1D sequence (GenBank™ accession number D38101) except: b accession M57682 and c accession M57970. For exon-specific PCR, ␣ 1D1-5 and ␣ 1D7-11 forward primers were used with ␣ 1D6 and ␣ 1D12 reverse primers, respectively. For C-terminal splice analysis ␣ 1D13 forward primer was used with ␣ 1D14 and ␣ 1D15 . ␣ 1D16 together with ␣ 1D14 or ␣ 1D15 were used for amplification of full-length ␣ 1D clones (long-range PCR).
The dihydropyridine sensitivity of Ba 2ϩ currents recorded from cells expressing the ␣ 1D-S(31b) subunit was similar to that of Ba 2ϩ currents recorded from GH 3 cells (24), whereas cur-rents recorded from cells expressing the ␣ 1A subunit were significantly less sensitive to nimodipine (Fig. 7). A previous study demonstrated that nimodipine caused a more potent inhibition of Ca 2ϩ channels when GH 3 cells were depolarized from a holding potential of Ϫ40 mV rather than one of Ϫ80 mV (24).

FIG. 4. Isolation and identification of ␣ 1D transcripts expressed by GH 3 cells and rat brain.
A, amplification of full-length ␣ 1D-S transcripts containing the putative G ␤␥ binding domain (denoted by ϩQXXER) and partial-length ␣ 1D-L transcripts (denoted by ϪQXXER). B, PCR with exon-specific amplimers was used to analyze ␣ 1D-S clones in GH 3 cells and rat brain. Numbers above each lane denote exons recognized by each primer set. Shown are representatives from a GH 3 cell (top) and a rat brain (bottom) ␣ 1D-S clone.
FIG. 5. Percent abundance of ␣ 1D isoforms in rat brain and GH 3 cells. A total of 100 individual clones for rat brain ␣ 1D-S (bars with lines), GH 3 cell ␣ 1D-S (checkered bars), and ␣ 1D-L (solid bars) isoforms were analyzed by PCR or hybridization to determine the percent abundance of each splice variant. The ␣ 1D-S variants 3 (␣ 1D-S(31a) ) and 4 (␣ 1D-S(31b) ) are the most prevalent in GH 3 cells and brain, respectively. The numbering scheme for ␣ 1D subunit splice variants is provided in Table II. Likewise, the nimodipine inhibition of currents mediated by ␣ 1D-S(31b) subunits was potent in cells held at Ϫ40 mV (IC 50 ϭ 0.25 Ϯ 0.02 M). Nimodipine inhibited Ba 2ϩ currents recorded from cells held at Ϫ80 mV with an IC 50 of 1.0 Ϯ 0.2 M. Recombinant ␣ 1A channels were inhibited by nimodipine with an IC 50 of 49 Ϯ 1 M in cells voltage-clamped at Ϫ40 mV. Taken together, these data support our assertion that, despite the presence of detectable levels of mRNA encoding the ␣ 1A subunit in GH 3 cells (Fig. 1), their functional channels are predominantly L-type.
G Protein Sensitivity of Recombinant Ca 2ϩ Channels-Cloned and ␦ opioid receptors couple to L-type Ca 2ϩ channels in GH 3 MORDOR cells through activation of inhibitory G i/o proteins (25). We hypothesized that the G protein-sensitive Ca 2ϩ channel component is comprised of an ␣ 1D subunit. Fur-thermore, we anticipated that the QXXER motif within specific ␣ 1D-S subunits was necessary for ␤␥ binding and therefore channel inhibition. Here we examined whether the QXXER motif is sufficient to confer opioid receptor coupling or G protein sensitivity to recombinant channels formed by the ␣ 1D-S subunit coexpressed in HEKDOR cells with ␤ 2a and ␣ 2 -␦ subunits.
FIG. 7. Nimodipine sensitivity of ␣ 1A and ␣ 1D channels. Concentration-response relationship for the inhibition of ␣ 1A (OE, n ϭ 6) and ␣ 1D-S(31b) (q, n ϭ 5) Ca 2ϩ channels by nimodipine. Cells were held at Ϫ40 mV, and currents were recorded in the presence of 30 mM Ba 2ϩ to minimize voltage-dependent inactivation. The concentration-response relationship of ␣ 1D-S(31b) Ca 2ϩ channels for inhibition by nimodipine was shifted to the right when cells were held at Ϫ80 mV (छ, n ϭ 5). Curves were generated by a logistic equation (see "Experimental Procedures").
when ␣ 1D-S(31b) or ␣ 1D-L(31b) subunits were expressed in HEK-DOR cells (n ϭ 6 and 4, respectively). Because opioid receptor coupling to Ca 2ϩ channels is thought to be mediated by G ␣o (34), CHO, CHODOR, and HEKDOR cells were transiently transfected with G ␣oA or G ␣oB cDNAs to determine whether the lack of coupling is due to inadequate levels of these G proteins. Neither the overexpression of G ␣oA (n ϭ 5 and 4 for HEKDOR and CHODOR, respectively) nor G ␣oB (n ϭ 3 and 5 for HEK-DOR and CHO) resulted in G protein-mediated inhibition of ␣ 1D-S(31b) channels by GTP␥S. Syntaxin-1A may play a permissive role in G ␤␥ -mediated inhibition of N-type Ca 2ϩ channels (35). We examined whether this may also be the case for ␣ 1D-S channels. No Ba 2ϩ current inhibition was observed in the presence of GTP␥S (n ϭ 3) in CHO cells transiently transfected with cDNAs encoding syntaxin-1A and ␣ 1D-S(31b) , ␤ 2a , and ␣ 2 -␦. These data demonstrate that the QXXER motif is not sufficient to confer sensitivity of ␣ 1D-S(31a) or ␣ 1D-S(31b) subunits to activated G proteins.
Voltage-dependent Facilitation of ␣ 1D Channel Activity-Depolarizing pre-pulses caused voltage-dependent facilitation of Ba 2ϩ currents mediated by all three recombinant ␣ 1D channel variants (Fig. 9). We investigated the effect of pre-depolarization on the current-voltage relationship of ␣ 1D-S(31b) Ca 2ϩ channels. An increase in peak current and a transient leftward shift in the current-voltage relationship occurred after application of a 60-mV depolarizing pre-pulse (I peak ϭ 938 pA versus I peak ϭ 1097 pA for pre-and post-pulse, respectively; Fig. 10A). Current amplitude was unaffected at voltages greater than 30 mV. We also examined the time required for recovery from voltagedependent facilitation in cells expressing ␣ 1D-S(31b) channels (Fig. 10B). A single-exponential fit of the decline in current amplitude following a 60-mV pre-pulse yielded the time constant for recovery from facilitation ( ϭ 58.6 Ϯ 10.5 ms). The rapid time course of recovery indicates that this event is unlikely to involve protein phosphorylation. This time course is similar to those reported previously for voltage-dependent reversal of G protein inhibition of ␣ 1A and ␣ 1B channels (36). However, G protein activation through inclusion of GTP␥S in the recording electrode had no effect on the observed facilitation of ␣ 1D-S(31b) channels (Fig. 9A). We examined further whether the facilitation of ␣ 1D channel activity was caused by reversal of constitutive G protein coupling. The reversal of G i/o protein-mediated inhibition of Ca 2ϩ channel activity is known to depend on the concentration of activated G proteins (37). However, pretreatment of cells with pertussis toxin (100 ng/ml) 24 h prior to recording, a procedure that should significantly reduce G i/o activation, had no significant effect on the time course for the recovery from facilitation ( ϭ 51.5 Ϯ 5.3 ms).
FIG. 8. Opioid-receptors couple to ␣ 1A channels but not to either of the ␣ 1D channel variants tested. A, top, Ba 2ϩ currents through ␣ 1A channels before, during, and after application of the ␦ opioid receptor agonist DPDPE. Bottom, voltage-dependent reversal of inhibition. Current amplitudes were normalized to the value obtained after a prepulse to Ϫ50 mV (I 1 ) and plotted as a function of the pre-pulse potential. Mean data from five cells in the presence of DPDPE were fitted using the Boltzmann equation (see "Experimental Procedures"). The fit indicated that currents in the presence of DP-DPE were maximally facilitated (F max ) by 107%, and half-maximal facilitation (F 50 ) occurred at Ϫ6.5 mV. B, top, ␣ 1D-S(31b) and bottom, ␣ 1D-L(31b) Ba 2ϩ currents were not inhibited by DPDPE (100 nM). Furthermore, transfection of cells with a cDNA encoding the ␤-adrenergic receptor kinase (␤-ARK) minigene peptide thought to inhibit G ␤␥ from binding to its effectors (38) had no significant effect on the time constant of recovery from facilitation ( ϭ 69.8 Ϯ 19.5 ms). Our data therefore suggest that the observed voltage-dependent facilitation of ␣ 1D-S(31b) channels is independent of G protein activity. DISCUSSION In this study we identified mRNAs encoding three ␣ 1 (␣ 1A , ␣ 1C , and ␣ 1D ), four ␤, and ␣ 2 -␦ Ca 2ϩ channel subunits in GH 3 cells. The transcripts encoding ␣ 1B and ␣ 1S subunits were not present. Despite detectable levels of ␣ 1A-b subunit mRNA, which would be expected to encode Q-type Ca 2ϩ channels, Ba 2ϩ currents recorded from GH 3 MORDOR cells are almost abolished by the dihydropyridine nimodipine (24). Furthermore, -agatoxin IVa (500 nM) at a sufficient concentration to block ␣ 1A-b channels (6) did not affect the inhibition of Ba 2ϩ currents induced by ␦ opioid receptor activation. Together, these observations suggest that P/Q-type channels do not contribute to the opioid receptor-sensitive Ca 2ϩ channel population in GH 3 MORDOR cells. The highest concentration of -agatoxin IVa (500 nM) tested did cause an inhibition of Ba 2ϩ current amplitude. There are two possible explanations for this observation. Functional Q-type Ca 2ϩ channels that are unable to couple to opioid receptors may exist in GH 3 MORDOR cells. Alternatively -agatoxin IVa (500 nM) may inhibit a subset of L-type Ca 2ϩ channels (perhaps ␣ 1C ) that do not couple to opioid receptors. Further experiments will be required to distinguish between these two possibilities. Taken together these data support the hypothesis that opioid receptors can couple to dihydropyridine-sensitive L-type Ca 2ϩ channels in GH 3 MORDOR cells. However, the fact that and ␦ receptor activation inhibits Ca 2ϩ channel activity in GH 3 MORDOR cells by Ͻ20% on average suggests that not all of their L-type channels are G protein-regulated (25).
Both ␣ 1C and ␣ 1D subunits may participate in the L-type Ca 2ϩ channel activity of GH 3 cells. However, recombinant receptors failed to couple to ␣ 1C subunits when both were expressed in Xenopus oocytes (26). Therefore, we hypothesized that ␣ 1D subunits couple to opioid receptors in GH 3 MORDOR cells (25). RT-PCR analysis revealed three loci within the ␣ 1D gene at which alternative splicing occurs in GH 3 cells. Transcripts encoding ␣ 1D subunits are also alternatively spliced at these sites in human and rat brain, hamster insulin-secreting cells, and chicken cochlea (7,33,39,40). The observation that alternative splicing occurs at the C terminus of the gene producing transcripts of differing lengths was of particular interest; the shorter transcript (␣ 1D-S ) (33) ends with a sequence that encodes an amino acid motif (QXXER) found in all other Ca 2ϩ channel ␣ 1 subunits (␣ 1A , ␣ 1B , and ␣ 1E ) that couple to G proteins. This motif is also present in other effectors, including adenylyl cyclase 2, G protein-activated inward rectifying K ϩ channels, and ␤-adrenergic receptor kinase (␤-ARK), which all couple in a membrane-delimited fashion to inhibitory metabotropic receptors (16). Several lines of evidence support the hypothesis that ␤␥ subunits mediate the inhibitory actions of G protein-coupled receptors on Ca 2ϩ channels, and inhibition is thought to occur subsequent to the binding of the ␤␥ subunit to the QXXER domain (15,18,37).
In view of their putative role in G protein signaling, we used long range RT-PCR to identify the alternatively spliced transcripts encoding ␣ 1D subunits found in GH 3 cells and rat whole brain. The relative abundances of ␣ 1D-S and ␣ 1D-L variants in GH 3 cells were determined by analyzing 100 transformed bacterial clones for each. Similarly, the number of ␣ 1D-S variants was examined in whole rat brain. In retrospect, the large number of isoforms detected was not surprising considering the presence of multiple hotspots of splice variation. However, the number of observed variants was larger in GH 3 cells than in brain. The additional splicing may be a characteristic of GH 3 FIG. 10. Voltage-dependent facilitation of ␣ 1D channels. A, voltage-dependent facilitation of currents was investigated by comparing the current-voltage relationship before and after a depolarization to 60 mV. Left, superimposed Ba 2ϩ currents recorded from a CHO cell expressing ␣ 1D-S(31b) . Right, graph showing the relationship between current amplitude and voltage before (q) and after (E) a depolarization to 60 mV. B, left, superimposed currents mediated by ␣ 1D-S(31b) channels during a prepulse to 60 mV and in response to a test pulse to 0 mV 10 -140 ms following the prepulse. Right, a graph of the time required for recovery from voltage-dependent facilitation. The peak current amplitude was measured and expressed relative to the peak amplitude of the maximally facilitated current, i.e. the first current (I 1 ) after the depolarizing pulse. Curves are exponential fits to the data. Neither pre-treatment with pertussis toxin (100 ng/ml, E) nor co-expression of the ␤-ARK minigene (OE) caused a significant change in the time required for recovery from voltage-dependent facilitation when compared with control (q). Data points are averages of at least four determinations. cells. Alternatively, analyzing 100 clones in the whole brain might not be sufficient to detect the presence of specific variants, which may only occur in small subpopulations of neurons. Nonetheless, analysis of the transcripts in the whole brain confirmed that the majority of splicing activity was not exclusive to GH 3 cells.
To date, there have been relatively few studies demonstrating functional expression of ␣ 1D Ca 2ϩ channel variants (41)(42)(43)(44). Analysis of the biophysical properties of the three ␣ 1D variants expressed in our study revealed that they have a similar current-voltage relationship to native ␣ 1D currents of the cochlear inner hair cells, activation threshold occurred at around Ϫ45 mV, and currents peaked near 0 mV (45). By contrast, ␣ 1C channels expressed with ␤ 2a and ␣ 2 -␦ subunits had a higher threshold of activation and peaked at 20 mV (46).
Voltage-dependent inhibition by dihydropyridine antagonists distinguishes L-type channels from all other Ca 2ϩ channels (47). Nimodipine blocked ␣ 1D-S(31b) Ca 2ϩ channels with an approximately 4-fold lower IC 50 at Ϫ40 mV holding potential than at Ϫ80 mV. The nimodipine sensitivity of Ba 2ϩ currents through ␣ 1D-S(31b) channels was similar to that of native GH 3 cell Ca 2ϩ channels (24). Intriguingly, Bell and colleagues (44) did not observe a voltage-dependent block of expressed ␣ 1D channels by dihydropyridine antagonists. This divergence may be due to the ␣ 1D isoforms used. We characterized the dihydropyridine sensitivity of the short isoform, ␣ 1D-S(31b) , terminating with exon 41a. Bell and colleagues (44) employed the long isoform that also lacked exon 11 and contained exon 31a. However, this explanation seems unlikely, because the IIIS5, IIIS6, and IVS6 regions of the L-type Ca 2ϩ channels, thought to be involved in dihydropyridine sensitivity (48), are the same in the two expressed ␣ 1D variants.
We chose the predominant ␣ 1D-S transcripts in GH 3 cells and rat brain, ␣ 1D-S(31a) and ␣ 1D-S(31b) , respectively, to assess whether the QXXER motif is sufficient to confer G protein coupling to these channels. We compared their sensitivity to activated G proteins to that of ␣ 1D-L(31b) subunits that lack the QXXER motif. Activation of G proteins by either opioid receptors or GTP␥S failed to inhibit Ba 2ϩ currents mediated by any of the ␣ 1D variants. Experiments in NG108-15 cells identified G ␣o as the inhibitory G protein required for the inhibition of Ca 2ϩ channels by opioids (34). Functional analysis of ␣ 1D-S(31b) and ␣ 1D-L(31b) in CHODOR cells, whose expression of endogenous G ␣o is well established (49), revealed a lack of G proteinmediated inhibition of ␣ 1D Ca 2ϩ channels. Moreover, increasing the levels of either G ␣oA or G ␣oB in HEKDOR and CHO cells by transient transfection with the appropriate cDNAs had no effect on the lack of G protein coupling to ␣ 1D channels.
Recently, Jarvis and colleagues (35) demonstrated a role for syntaxin-1A in the augmentation of G protein-mediated inhibition of ␣ 1A channels. Furthermore, syntaxin-1A was shown to colocalize in the plasma membrane and co-immunoprecipitated with ␣ 1D Ca 2ϩ channels in pancreatic beta cells (50). We postulated that syntaxin-1A might be required for G protein-mediated modulation of ␣ 1D Ca 2ϩ channels. However, transient expression of syntaxin-1A into CHO cells did not result in G protein-mediated inhibition of ␣ 1D channels.
All ␣ 1D splice variants examined exhibited a potentiation or facilitation evoked by depolarizing pre-pulses. Facilitation was independent of G protein activation, and the rapid rate of recovery ( Ͻ 70 ms) suggests that it is unlike the phosphorylation-dependent facilitation seen in recordings of L-type Ca 2ϩ channel activity in bovine chromaffin cells where reversal of facilitation is slow (ϳ60 s) (51). Immunohistochemical analysis of the distribution of ␣ 1D subunits in neurons reveals their distribution on cell bodies and proximal dendrites in the hip-pocampus and many other brain regions (52,53), where they may participate in the activity-dependent initiation of gene expression that is inhibited by dihydropyridines (54,55). Furthermore, their voltage-dependent facilitation may also enable activity-dependent enhancement of Ca 2ϩ influx contributing to long-term potentiation and depression (56 -58).
Our results suggest that the QXXER motif is not sufficient to confer G protein sensitivity to ␣ 1D channels. Recent reports suggest that, in addition to the two proposed G ␤␥ sites in the intracellular loop between domains I and II, domain I and Nand C-terminal regions of ␣ 1B , and ␣ 1E Ca 2ϩ channels may be important for G protein coupling (36,59,60). Similarly, multiple regions may be required for G protein-mediated inhibition of ␣ 1D channels to occur. Thus other ␣ 1D splice variants may be responsible for the G protein-sensitive L-type Ca 2ϩ channels in GH 3 cells. Alternatively, the required signaling molecules for G protein-mediated inhibition of L-type Ca 2ϩ channels present in GH 3 cells may not be present in HEK or CHO cells.
There are few studies of functional ␣ 1D channels, and this is among the first to examine multiple alternatively spliced isoforms. We have shown that alternatively spliced ␣ 1D subunits form functional L-type channels that resemble those of GH 3 cells in their voltage dependence of activation and nimodipine sensitivity.