INTRODUCTION

Ryanodine receptors are a family of intracellular Ca²⁺ release channels that were originally identified in the sarcoplasmic reticulum (SR)¹ of striated muscles. To date, three members of this family have been identified in mammalian tissues, namely the skeletal muscle (RyR1), the cardiac muscle (RyR2), and the brain (RyR3) ryanodine receptor. These proteins are the products of different genes and share 66-70% amino acid sequence identity (1-5). Earlier studies using RNA blot analysis revealed that the expression patterns of these isoforms were very different (6). RyR1 was predominantly expressed in skeletal muscle, whereas RyR2 was mainly expressed in heart and brain. The expression of RyR3 was detected in smooth muscle tissues and certain regions of the brain. However, results of recent ribonuclease protection assay demonstrate that all three RyR isoforms are widely and differentially expressed (7). These studies also indicate that most tissues express more than one RyR isoform. For example, skeletal muscles express both RyR1 and RyR3, although RyR3 is expressed at a much lower level than that of RyR1.

The function and regulation of RyR1 and RyR2 have been extensively studied. They function as Ca²⁺ release channels of the sarcoplasmic reticulum and play an essential role in excitation-contraction coupling in striated muscles (1-4). Both RyR1 and RyR2 have been purified and characterized. The single channel properties of RyR1 and RyR2 that have been incorporated into lipid bilayers are very similar. They both form a large conductance channel permeable to monovalent and divalent cations and can be activated by submicromolar Ca²⁺, ATP, and caffeine and inhibited by millimolar Ca²⁺, Mg²⁺, and ruthenium red. Ryanodine, a plant alkaloid, locks both channels in a subconductance state. However, differences in sensitivities to Ca²⁺ activation, Ca²⁺ inactivation, and Mg²⁺ inhibition have been reported (8-11). Differences in responses to cyclic ADP-ribose (12, 13), inorganic phosphate, and perchlorate (14) have also been observed.

On the other hand, little is known about the physiological role, regulation, and channel properties of the RyR3 isoform (5). Isolation and biochemical and biophysical characterization of RyR3 have been impeded by its low abundance and its coexpression with other RyR isoforms. Biochemical studies and single channel recordings of ryanodine-sensitive Ca²⁺ release channels isolated from brain and smooth muscle tissues have been reported (15-17). It is, however, uncertain whether the observed properties correspond to those of the RyR3 isoform, since these tissues are known to express RyR3 as well as RyR2 and/or RyR1. Recently, Murayama and Ogawa (18) were able to purify the rabbit brain RyR3 by using an RyR3-specific antibody. They demonstrated that the immunoprecipitated RyR3 was capable of binding [³H]ryanodine in a Ca²⁺-dependent and caffeine-sensitive manner (18). However, channel properties of the purified RyR3 have not been characterized.

Significant progress in understanding the function and regulation of RyR3 has recently been made through the generation of mice with a null mutation in the ryr1 gene (19). Skeletal muscles isolated from these mutant mice lack RyR1 and RyR2 but contain RyR3, thus providing a valuable system to study RyR3. Analysis of SR Ca²⁺ release in these mutant muscles suggests that RyR3 is modulated by Ca²⁺, adenine nucleotide, caffeine, and ryanodine (20). More recently, a mutant mouse lacking RyR3 has also been generated (21). Ca²⁺ release studies have shown that SR Ca²⁺ release in the RyR1-deficient muscle (or RyR3 containing muscle) was much less sensitive to Ca²⁺ activation than that in the RyR3-deficient muscle (or RyR1 containing muscle). Based on these data, it was suggested that RyR3 may function as a Ca²⁺-induced Ca²⁺ release channel with a high threshold in mammalian skeletal muscles (21). It remains, however, to be determined whether the intrinsic sensitivity of the RyR3 channel to Ca²⁺ activation differs from that of the RyR1 channel.

To gain insights into the Ca²⁺ sensitivity, ligand gating properties, and single channel characteristics of RyR3, we have cloned the RyR3 cDNA from rabbit uterus and expressed the cDNA in HEK293 cells. The CHAPS-solubilized and sucrose gradient-purified cloned RyR3 was characterized by using single channel recordings in planar lipid bilayers. Our results demonstrate that the cloned RyR3 single channel is modulated by caffeine, ryanodine, and other physiological and pharmacological ligands. The sensitivity of the cloned RyR3 to Ca²⁺ activation was found to be similar to that of RyR1. However, the cloned RyR3 differs from RyR1 in the gating behavior, the extent of maximal activation by Ca²⁺, and the sensitivity to Ca²⁺ inactivation.

EXPERIMENTAL PROCEDURES

Restriction endonucleases and DNA modifying enzymes were purchased from Boehringer Mannheim, Life Technologies Inc., and Pharmacia Biotech Inc. Ryanodine was obtained from Agri Systems International (Wind Gap, PA). Soybean phosphatidylcholine, brain phosphatidylethanolamine, and brain phosphatidylserine were from Avanti Polar Lipids. Calmodulin was obtained from Calbiochem. Horseradish peroxidase-conjugated anti-rabbit IgG antibody was purchased from Promega Biotech. Rhodamine-conjugated anti-rabbit IgG antibody was from Jackson ImmunoResearch Laboratories, Inc. Prestained protein standards were purchased from Bio-Rad. Cyclic ADP-ribose, CHAPS, and other chemicals were from Sigma.

Total RNA from rabbit uterus, isolated by the method of Chomczynski and Sacchi (22), was used to purify poly(A)⁺ RNA using the mRNA Purification Kit (Pharmacia) according to the manufacturer's instructions. First strand cDNA was prepared from mRNA (10 µg) using the SuperScript Preamplification System (Life Technologies Inc.) with random primers and was resuspended in 150 µl of H₂O. Polymerase chain reactions (PCR) were carried out in a 50-µl solution containing 20 mM Tris-HCl, pH 8.8, 10 mM KCl, 10 mM (NH₄)₂SO₄, 2.0 mM MgS0₄, 0.1% Triton X-100, 0.1 mg/ml bovine serum albumin, 50 ng of each DNA primer, 200 µM concentration of each dATP, dCTP, dGTP, and dTTP, 2.5 units of Pfu DNA polymerase, and 1.0-2.0 µl of first strand cDNA. The DNA was denatured for 3 min at 94 °C followed by 25-30 cycles of amplification. Each cycle consists of 45 s at 94 °C, 1 min at 46-53 °C, and 4 min at 72 °C. An additional extension for 5 min at 72 °C was performed after the final cycle. The annealing temperature and the number of cycles for each pair of primers were determined empirically.

The entire 15-kilobase pair coding region of the RyR3 mRNA from rabbit uterus was amplified by PCR using 16 pairs of primers (Table I and Fig. 1). PCR primers were synthesized based on the reported brain RyR3 cDNA sequence (6). A 5-flanking BamHI restriction site is included in primers CP-1F, CP-1R, CP-2F, CP-2R, CP-3F, CP-3R, CP-5F, CP-5R, CP-7F, CP-7R, FP-6F, and FP-8F, whereas a 5-flanking EcoRI site is included in primers CP-4F, CP-4R, CP-6F, CP-6R, and FP-6R. Primers CP-8F and CP-8R each contain a 5-flanking XbaI site. CP-1F also contains a 5-flanking NheI site downstream of the BamHI. Mutations that did not change amino acid sequence were introduced in overlapping PCR primers CP-1R and CP-2F, CP-2R and CP-3F, and CP-5R and CP-6F to generate three unique restriction sites XhoI, SalI, and SmaI in the RyR3 cDNA.

Table I. Sequences of PCR primers Name Sequence PCR product CP-1F 5-ACTGGATCCAGCTAGCAGCCATGGCCGAAGGCGGAGA-3 PCR1 CP-1R 5-ACTGGATCCTCGAGTCTGTCCAATTTACTGATAAGC-3 CP-2F 5-ACTGGATCCTCGAGTCTTCCTCAGGTATCCTGGA-3 PCR2a SP2-2R 5-TTCAGCCAGCCTGTCTC-3 SP2-2F 5-GCATCTCCTTCCGCATC-3 PCR2b CP-2R 5-ACTGGATCCATGTCAGGTCGACAGCCTGGCCT-3 CP-3F 5-ACTGGATCCAAGGCTGTCGACCTGACATTGAGCT-3 PCR3a SP3-2R 5-ACCGGGTTCTTCTCTTCAC-3 SP3-1F 5-GCAACGTGGACCTGGAG-3 PCR3b CP-3R 5-ACTGGATCCGCCAAACACAGAGGGATCGATGA-3 CP-4F 5-ACTGAATTCCCTCCTCATCGATCCCTCTGTGT-3 PCR4a SP4-2R 5-CTCCAGCAGGTAACTCAG-3 SP4-3F 5-ACCAGCATCCCAACCTC-3 PCR4b FP-6R 5-CAGGAATTCTCAGCCCATGTGCACAATCTCTTCT-3 FP-6F 5-CTGGGATCCAGGGTGCCATTAAAATCTCCG-3 PCR4c CP-4R 5-ACTGAATTCAGAATCGATCAGAGAGGTGTAGTGT-3 CP-5F 5-ACTGGATCCTGATCGATTCTACACTGCAGACAA-3 PCR5a SP5-2R 5-GAAGGCATCCAGCTCCAT-3 FP-8F 5-CTGGGATCCAGAGGACCAAAGAGGGTGAAGC-3 PCR5b CP-5R 5-ACTGGATCCCGGGGAGTTTTGGTGTTGAAGACT-3 CP-6F 5-ACTGAATTCCCGGGAAAGGTCTATTCTGGGGATG-3 PCR6a SP6-1R 5-CCGCTCCTGGTCCTGT-3 SP6-2F 5-ACCAACTCTTCCGCATG-3 PCR6b CP-6R 5-ACTGAATTCAGCCAAATCCTGTACTAGTTTCTCT-3 CP-7F 5-ACTGGATCCAAGAGAAACTAGTACAGGATTTGGCT-3 PCR7a SP7-1R 5-GTGAACTCATCATTCTGG-3 SP7-1F 5-GCCGAGATGGTCCTTCA-3 PCR7b SP7-4R 5-TCCATGGCCTTCTGGAATTC-3 SP7-4F 5-TTAACCAGCTCAGATAC-3 PCR7c CP-7R 5-ACTGGATCCTGCACTAGTTCAGTCGTGATGCCT-3 CP-8F 5-CAGTCTAGAAGGGGCAAAGAACATCAGAGTGACT-3 PCR8 CP-8R 5-GTCTCTAGACCAGTGTCGTGCTGTAGCTTTCAC-3 FP-8F 5-CTGGGATCCAGAGGACCAAAGAGGGTGAAGC-3 FP-8 FP-8R 5-CAGGAATTCTCATTCTCTGGAGAGCACAACGTTTG-3 FP-11F 5-GTCGGATCCAGGCAGCAGAGACGAAG-3 FP-11 FF-11R 5-CAGGAATTCTCATCGCTCCTCTTCTTTGGCT-3

These primers were used to generate 16 PCR fragments PCR1, PCR2a, PCR2b, PCR3a, PCR3b, PCR4a, PCR4b, PCR4c, PCR5a, PCR5b, PCR6a, PCR6b, PCR7a, PCR7b, PCR7c, and PCR8 (Table I and Fig. 1). These PCR fragments were subcloned into plasmid pBluescript. Individual clones of each PCR fragment were analyzed by restriction endonuclease digestion and DNA sequencing analysis by the Sanger dideoxy chain termination method (23) with -³⁵S-dATP (1000 Ci/mmol, Amersham Corp.) using the T7 Sequencing Kit from Pharmacia.

All recombinant DNA manipulations were carried out using standard procedures (24, 25). Subfragments PCR2a and PCR2b, PCR3a and PCR3b, PCR4a, PCR4b and PCR4c, PCR5a and PCR5b, PCR6a and PCR6b, PCR7a, PCR7b, and PCR7c were joined together to form PCR2, PCR3, PCR4, PCR5, PCR6, PCR7, respectively (Fig. 1). PCR1 and PCR2, PCR3 and PCR5, and PCR6 and PCR8 were then connected together to yield PCR1,2, PCR3,5, and PCR6,8, respectively. These three fragments were then ligated together sequentially to form PCR1-3,5,6,8 (Fig. 1). PCR4 was ligated with PCR1-3,5,6,8 to form PCR1-6,8 which was subsequently subcloned into another plasmid pBlueBac4 (Invitrogen) to yield PCR1-6,8 (pBB4). The final gap was filled by ligating PCR7 with PCR1-6,8 to form pRyR3 (pBB4) (Fig. 1). Finally, the full-length RyR3 cDNA from pRyR3 (pBB4) (Fig. 1) was subcloned into the mammalian expression vector pcDNA3 (Invitrogen) to form pRyR3.

Anti-FP8 and anti-FP11 antibodies were raised in rabbits against short sequences of the rabbit RyR3 in the form of GST fusion proteins. Oligodeoxynucleotides with a 5-flanking BamHI restriction site for the forward primers (FP-8F and FP-11F) and 5-flanking EcoRI restriction site and a stop codon for the reverse primers (FP-8R and FP-11R) were synthesized (Table I). Primers, FP-8F and FP-8R, were used to synthesize the DNA fragment encoding amino acids 2690-2734 of the rabbit RyR3 by PCR using PCR5b as DNA template. Primers, FP-11F and FP-11R, were used to synthesize the DNA fragment encoding amino acids 4328-4363 of RyR3 by PCR using PCR8 fragment as DNA template. PCR products were digested with BamHI and EcoRI, purified, and ligated into the BamHI/EcoRI sites in the polylinker in pGEX-3X to form pFP8 and pFP11. The sequences of pFP8 and pFP11 were confirmed by DNA sequencing. The expression and isolation of glutathione S-transferase (GST) fusion protein FP8 and FP11 were carried out according to the standard protocols from Pharmacia. Rabbits were immunized subcutaneously with 200 µg of affinity purified FP8 or FP11 in complete Freund's adjuvant, and booster injections of the same amount in incomplete Freund's adjuvant were given at 17-19-day intervals. Antiserum was collected 2 weeks after each booster injection.

GST protein and FP8 and FP11 fusion proteins were immobilized to separate Affi-Gel 15 columns (Bio-Rad) according to the standard protocol from Bio-Rad. Antiserum was first absorbed on the GST column. Absorbed antiserum was then loaded onto the FP8 or FP11 column, washed, and eluted with 0.1 M glycine, pH 2.5, and neutralized with 0.2 volume of 1 M Tris-HCl, pH 8.0. Affinity purified antibodies were dialyzed against phosphate-buffered saline (PBS) and concentrated in a Centricon-10 concentrator (Amicon).

HEK293 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 0.1 mM minimum Eagle's medium nonessential amino acids, 4 mM L-glutamine, 100 units of penicillin/ml, 100 mg of streptomycin/ml, 4.5 g of glucose/liter, and 10% fetal calf serum, at 37 °C under 5% CO₂. DNA transfection was carried out using calcium phosphate (24). Cells were plated in 100-mm tissue culture dishes 20-22 h before transfection and were transfected with 8 µg of pRyR3 per dish. Control cells were treated in the same way either with no DNA or with only the expression vector DNA.

Coverslips were placed in a 100-mm tissue culture dish. Cell culture and DNA transfection were then carried out as described above. Twenty-four hours after transfection, the coverslips were washed three times with PBS, fixed with 4% formaldehyde in PBS for 15 min, and washed once with PBS and three times with PBS containing 0.1% saponin 5 min each time. The coverslips were blocked with buffer A (2% skim milk powder, 0.1% saponin in PBS) for 30 min before washing and incubating with anti-FP11 antibody in buffer A for 1 to 2 h. The coverslips were washed with buffer A and incubated with rhodamine-conjugated anti-rabbit IgG in buffer A for 30-60 min. The coverslips were then washed, mounted in 95% glycerol, and analyzed with the Leica DMRB photomicroscope using a 63 × objective.

Free cytosolic Ca²⁺ in transfected and non-transfected HEK293 cells was measured with the fluorescence Ca²⁺ indicator dye fluo 3 (26). Cells grown for 24 h after transfection were washed three times with KRH buffer without MgCl₂ and CaCl₂ (KRH buffer: 125 mM NaCl, 5 mM KCl, 1.2 mM KH₂PO₄, 6 mM glucose, 1.2 mM MgCl₂, 2 mM CaCl₂, and 25 mM Hepes, pH 7.4) and incubated in the same buffer at 37 °C for 30-60 min. After being detached from culture dishes by pipetting, cells were collected by centrifugation at 2,500 rpm for 2 min in a Beckman TH-4 rotor. Cell pellets were washed twice with KRH buffer and loaded with 10 µM fluo 3 in KRH buffer plus 0.1 mg/ml bovine serum albumin and 250 µM sulfinpyasone at room temperature for 30 min, followed by washing with KRH buffer three times and resuspended in KRH buffer plus 0.1 mg/ml bovine serum albumin and 250 µM sulfinpyasone. Aliquot of 100-150 µl of fluo 3 loaded cells was added to 2 ml (final volume) KRH buffer in a cuvette. Fluorescence intensity of fluo 3 at 530 nm was measured in an SLM-Aminco series 2 luminescence spectrometer with 480 nm excitation at 25 °C (SLM Instruments, Urbana, IL).

Heavy sarcoplasmic reticulum was isolated from bovine diaphragm or rabbit fast-twitch skeletal muscle according to the method described by Campbell and MacLennan (27).

Microsomal membranes were prepared from transfected and non-transfected HEK293 cells as described previously (28) with some modifications. Cells grown for 28 h after transfection on 100-mm tissue culture dishes were washed twice with 3 ml of PBS containing 5 mM EDTA and incubated in the same solution for 5 min at room temperature. Cells were detached from dishes by gentle pipetting and collected by centrifugation at 2,500 rpm for 2 min in a Beckman TH-4 rotor. Cell pellets were suspended in a solution containing 5 mM Tris-HCl, pH 7.0, 0.5 mM MgCl₂, and a protease inhibitor mix (0.5 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, 2 µg/ml leupeptin, 2 µg/ml pepstatin A, 1 µg/ml aprotinin, and 2.5 mM dithiothreitol) and incubated on ice for 20-30 min. Cells were homogenized in a glass Dounce homogenizer for 50 strokes with pestle A. An equal volume of a solution containing 15 mM Tris-HCl, pH 7.0, 0.5 M sucrose, 0.3 M KCl, 100 µM CaCl₂, and the protease inhibitor mix was added to the homogenate and homogenized for 35 strokes. The homogenate was centrifuged at 8,000 rpm for 20 min in a Beckman JA-20 rotor at 4 °C. The supernatant was centrifuged at 37,000 rpm in a Beckman Ti70 rotor for 60 min. The pellets were suspended in a buffer containing 25 mM Tris-Hepes, pH 7.4, 1 M NaCl, and the protease inhibitor mix and solubilized by 1.5% (final concentration) CHAPS in a final volume of 2.5 ml and a final membrane protein concentration of 2 mg/ml for 1 h on ice. The suspension was spun at 35,000 rpm in a Beckman Ti75 rotor for 1 h at 4 °C, and 2.5 ml of the supernatant was layered on top of a 10.5-ml (7.5-25%, w/w) linear sucrose gradient containing 25 mM Tris-Hepes, pH 7.4, 0.3 M NaCl, 0.1 mM CaCl₂, 0.25 mM phenylmethylsulfonyl fluoride, 4 µg/ml leupeptin, 5 mM dithiothreitol, 0.3% CHAPS, and 0.17% phosphatidylcholine. The tubes were centrifuged at 28,500 rpm in a Beckman SW-41 rotor for 17-18 h at 4 °C. Fractions of 0.7 ml each were collected and monitored for immunoreactivity by enzyme-linked immunosorbent assay as described below. The peak fractions of immunoreactivity were pooled, aliquoted, and stored at 80 °C.

Aliquots of 50 µl of the sucrose density gradient fractions were added to microtiter wells containing 150 µl of 50 mM sodium carbonate buffer, pH 9.6. The microtiter plate was incubated at 4 °C for 18 h and then blocked with 5% skim milk powder in 50 mM sodium carbonate buffer, pH 9.6, for 1 h at room temperature. The plate was washed 8-10 times with cold tap water and incubated with anti-FP11 in PBS solution containing 5% milk powder and 0.1% saponin for 2-4 h, washed again, and allowed to react with alkaline phosphatase-conjugated anti-rabbit IgG for 1 h. The samples were again washed, and bound antibodies were quantified by the alkaline phosphatase reaction in diethanolamine buffer containing 10% diethanolamine, pH 9.8, 5 mM MgCl₂, and one phosphatase substrate tablet (Sigma) per 5 ml of diethanolamine buffer.

Microsomal membranes and sarcoplasmic reticulum membranes were denatured in Laemmli sample buffer at 100 °C for 2 min and separated in 5% SDS-polyacrylamide minigels (29). The resolved proteins were transferred to nitrocellulose membranes at 20 V for 15-16 h at 4 °C in the presence of 0.01% SDS according to Towbin et al. (30). The nitrocellulose membranes were blocked for 1 h with PBS plus 0.5% Tween 20 and 5% skim milk powder, incubated 2-4 h with anti-FP8 antibody in the same solution, and then washed three times for 15 min each with the same buffer. The bound antibodies were visualized by using horseradish peroxidase-conjugated anti-rabbit IgG and the enhanced chemiluminescence Western blotting analysis system from Amersham Corp.

Single channel recordings were carried out using CHAPS-solubilized and sucrose density gradient-purified ryanodine receptors from either rabbit skeletal muscle heavy sarcoplasmic reticulum or from transfected HEK293 cells as described previously (28). Brain phosphatidylserine and brain phosphatidylethanolamine, dissolved in chloroform, were combined in a 3:5 ratio (w/w), dried under nitrogen gas, and suspended in 30 µl of n-decane at a concentration of 35 mg of lipid/ml. Bilayers were formed across a 250-µm hole in a Delrin partition separating two chambers. The trans chamber (500 µl) was connected to the head stage input of an Axopatch 200A amplifier (Axon Instruments Inc.). The cis chamber (1.2 ml) was held at virtual ground. A symmetrical solution containing 250 mM KCl and 25 mM Hepes, pH 7.4, was used for recordings. A 12-µl aliquot of the sucrose density gradient-purified cloned RyR3 or a 4-µl aliquot of the sucrose density gradient-purified native rabbit skeletal muscle ryanodine receptor was added to the cis chamber. Unless indicated otherwise, spontaneous channel activity was tested for sensitivity to EGTA and/or Ca²⁺, thereby providing information about Ca²⁺ sensitivity, orientation in the bilayer, and stability of the incorporated channel. Free Ca²⁺ concentrations were calculated using the computer program of Fabiato and Fabiato (31). All subsequent additions were made to that chamber in which the addition of EGTA inhibited the activity of the incorporated channel. This chamber presumably corresponds to the cytoplasmic side of the Ca²⁺ release channel. Recordings were filtered at 1500 Hz using a low-pass Bessel filter (Frequency Devices, Haverhill, MA), digitized at 20 kHz, and recorded on optical disk cartridges. The data were analyzed using pClamp 6.0.3 software (Axon Instruments Inc.).

RESULTS

To obtain the full-length RyR3 cDNA for expression and functional studies, we employed reverse transcriptase-polymerase chain reaction (RT-PCR) to amplify the entire coding sequence of the RyR3 cDNA from rabbit uterus. Sixteen overlapping cDNA fragments were obtained and sequenced (Fig. 1). The deduced amino acid sequence of the rabbit uterus RyR3 cDNA is identical to that of the rabbit brain RyR3 cDNA reported previously (6) except for the following differences. A 15-base sequence after A10020, encoding amino acids AMQVK, was found to be deleted in all the PCR6a clones (Fig. 1). Further RT-PCR analysis revealed that the deletion was tissue-specific. This 15-base sequence was detected in RyR3 mRNA isolated from brain but not in RyR3 mRNAs isolated from uterus, aorta, heart, and diaphragm skeletal muscle (data not shown). The corresponding region in the human (32, 33), rabbit (34), and mouse (35) RyR1 mRNA and in mink RyR3 mRNA (36) has been shown to be alternatively spliced. Thus the 15-base deletion most likely resulted from alternative splicing of the ryr3 gene. An insertion of G after G-3345 and a deletion of C-3423 were found in all the PCR3a clones. These changes resulted in a frameshift between amino acid 1116 and 1140. The previously reported amino acid sequence of this region (PMTKPLCLKAAGASVGTKVVGILGVP) shows no homology to the corresponding region of the mink, chicken, or frog RyR3, whereas the amino acid sequence deduced from the rabbit uterus RyR3 mRNA (ADDQAFVFEGSRGQRWHQGSGYFGRT) is identical or highly homologous to those of the mink, chicken, and frog RyR3 (36-38). We also found two amino acid substitutions, Thr-534 was changed to Asn and Val-3675 was changed to Glu. These substitutions may represent polymorphic changes of the rabbit ryr3 gene. However, it should be noted that the corresponding amino acids in the mink, chicken, and frog RyR3 cDNA are Asn and Glu, respectively.

To express the RyR3 cDNA in mammalian expression system, we constructed a full-length RyR3 cDNA from those overlapping PCR fragments (Fig. 1). The full-length cDNA was subcloned into the mammalian expression vector pcDNA3 (Invitrogen) and transiently expressed in HEK293 cells. The specific expression of RyR3 cDNA in HEK293 cells was examined by immunofluorescence staining and immunoblotting using antibodies raised against short amino acid sequences of RyR3. The binding of the site-directed anti-RyR3 antibody, anti-FP11, was detected only in HEK293 cells transiently transfected with the RyR3 cDNA (Fig. 2), but not in cells transfected with only the expression vector pcDNA3 (not shown), indicating that the expression was specific. The transfection efficiency was about 15-30%.

The expression of the cloned RyR3 was further characterized by Western blotting. Because of the high background binding of anti-FP11 on Western blot, we developed another site-directed anti-RyR3 antibody, anti-FP8, for Western blot analysis. As shown in Fig. 3, a major high molecular weight band was recognized by anti-FP8 in bovine diaphragm SR membranes (lane 2) and in transfected microsomal membranes (lane 3) but not in control microsomal membranes (lane 4). The bovine RyR1 was stained by monoclonal antibody 34C (lane 1). This monoclonal antibody recognizes both the chicken  and  RyR isoforms that are homologues of the mammalian RyR1 and RyR3, respectively (5, 39). The cloned RyR3 and bovine diaphragm RyR3 migrated similarly but slightly faster than the bovine diaphragm RyR1, similar to those reported previously (7). Thus, the full-length RyR3 cDNA encodes a protein with molecular size and immunoreactivity indistinguishable from those of the native mammalian RyR3. Minor bands of lower molecular weight were also detected in transfected membranes but not in control membranes. These minor bands presumably resulted from proteolytic degradation of the cloned RyR3 protein. The site-directed anti-FP8 antibody did not recognize the bovine RyR1 (lane 2) on Western blot, indicating that it is isoform-specific. Using anti-FP8 or monoclonal antibody 34C, we were unable to detect RyR3 proteins in microsomes from rabbit or bovine uterus by Western blotting. This observation may indicate that the level of RyR3 proteins in uterus microsomes is much lower than that in diaphragm SR.

We next examined the functional properties of the cloned RyR3 expressed in HEK293 cells by monitoring cytosolic Ca²⁺ concentrations of the transfected and control HEK293 cells using the fluorescence Ca²⁺ indicator dye fluo 3. As illustrated in Fig. 4A, addition of 50 µM ryanodine to suspended HEK293 cells transfected with the RyR3 cDNA caused a slow and steady increase in fluorescence signal. Subsequent addition of 5 mM caffeine resulted in a much faster and larger transient increase in fluorescence (n = 13). The ryanodine-induced increase in fluo 3 signal was more obvious if cells were preincubated with 0.5 mM caffeine (n = 2) (Fig. 4B). On the other hand, fluo 3 signals in control HEK293 cells never increased after addition of either ryanodine or caffeine but did respond to Ca²⁺ ionophore A23187 (n = 8) (data not shown). Thus, these results indicate that the cloned RyR3 functions as a ryanodine- and caffeine-sensitive Ca²⁺ release channel in HEK293 cells.

In the next several series of experiments, we investigated the single channel properties and modulation of the cloned RyR3. To study the cloned RyR3 at the single channel level, microsomal membranes prepared from transfected HEK293 cells were solubilized in CHAPS and subjected to sucrose density gradient centrifugation. Enzyme-linked immunosorbent assay was used to localize the cloned RyR3 in sucrose gradients (data not shown). Gradient fractions containing the peak of immunoreactivity were pooled and used directly for single channel recordings in planar lipid bilayers.

Fig. 5A shows single channel current fluctuations of a cloned RyR3 channel recorded in a symmetrical 250 mM KCl at voltages between +40 and 40 mV (Fig. 5A). When K⁺ was used as the current carrier, the cloned RyR3 channel activity gave a linear current-voltage relationship with a slope conductance of 777 ± 17 pS (mean ± S.D.) (n = 8) (Fig. 5C) (open circles). To examine the ion selectivity of the cloned RyR3, single channel currents were recorded under a bi-ionic condition (250 mM KCl on the cytoplasmic face and 250 mM CaCl₂ on the luminal face of the channel) (Fig. 5B). Under these conditions, the reversal potential was 37.8 ± 1.5 mV (n = 5), indicating that the cloned RyR3 channel is selective for divalent over monovalent cations. The permeation ratio of pCa²⁺/pK⁺ was calculated to be 6.3 (40), and the Ca²⁺ conductance of the cloned RyR3 channel, determined from the current-voltage relationship between 20 and 60 mV, was 137 ± 4.3 pS (n = 5) under these bi-ionic conditions (Fig. 5C) (solid circles). A conductance of 112 ± 2.0 pS (mean ± S.E.) (n = 2) was obtained in the presence of 50 mM luminal CaCl₂ and 250 mM cytoplasmic KCl (measured between 20 and 60 mV) (data not shown). Well resolved subconductance states were seldom observed, although some brief truncated openings could be detected in the current records. These truncated openings are likely events that are too brief to be completely resolved by the bilayer system.

Modulation of the cloned RyR3 channel by Ca²⁺, Mg²⁺, ATP, caffeine, ryanodine, and ruthenium red is shown in Fig. 6. We found that CHAPS-solubilized and sucrose gradient-purified RyRs could be incorporated into the bilayers in either orientation. The EGTA sensitivity of the spontaneous single channel activity stimulated by contaminating Ca²⁺ in the recording solution was used to establish the cis-trans orientation. The channel shown in Fig. 6 was completely blocked by the cis addition of 0.1 mM EGTA (data not shown), indicating that the cytoplasmic side of the channel faced the cis chamber. All subsequent additions were then made to the cis (cytoplasmic) chamber. An aliquot of 10 mM CaCl₂ solution was added to the cis chamber to raise the cytoplasmic free Ca²⁺ concentration to 104 µM. At this Ca²⁺ concentration, the channel was fully activated with open probability (P_o) close to 1 (Fig. 6A). After lowering the Ca²⁺ concentration to 202 nM by addition of EGTA, the channel activity was reduced markedly (Fig. 6B). Subsequent addition of 1.5 mM caffeine strongly activated the channel (Fig. 6C). The average P_o increased 27.5 ± 12.6-fold (n = 8). The kinetics of the channel was also changed considerably. In the presence of 100-200 nM Ca²⁺, the cloned RyR3 channel displayed an average open time constant of 1.16 ± 0.22 ms (n = 7) (Fig. 7M) and a closed time constant ranging from tens of milliseconds to several seconds. On the other hand, the gating of the caffeine-activated channel could be described by two open time constants of 1.31 ± 0.7 and 8.15 ± 5.3 ms and a closed time constant of 4.52 ± 3.5 ms (n = 8). Subsequent addition of 1 mM MgCl₂ deceased P_o significantly (Fig. 6D). The Mg²⁺-inhibited channel could be reactivated by subsequent addition of 1.5 mM ATP (Fig. 6E). Subsequent addition of 10 µM ryanodine shifted the channel to a long-lived open state with a reduced conductance of 440 pS, about 57% of the unmodified conductance (Fig. 6F). The ryanodine-modified channel was blocked by 30 µM ruthenium red (Fig. 6G).

Cytoplasmic Ca²⁺ was capable of activating the cloned RyR3 channel as demonstrated in Fig. 6, A and B. To further examine the sensitivity to Ca²⁺ activation and Ca²⁺ inactivation, the response of the cloned RyR3 to a wide range of cytoplasmic Ca²⁺ concentrations was measured. At 2 nM Ca²⁺ concentration (pCa 8.7), no open events were seen (Fig. 7A). The channel remained closed until the Ca²⁺ concentration was raised to about pCa 7.19 where a few open events could be detected (Fig. 7B). The channel was further activated by increasing the Ca²⁺ concentration. P_o increased from 0.001 to 0.028, 0.400, 0.706, 0.947, and 0.981 when the Ca²⁺ concentrations were increased from pCa 7.19 to 6.73, 6.56, 6.38, 5.39, and 4.05, respectively (Fig. 7, B-G). At Ca²⁺ concentrations about 1 mM, the channel activity began to decline. P_o decreased from 0.981 to 0.878, 0.568, and 0.113 when Ca²⁺ concentrations were raised from pCa 4.05 to 3.00, 2.50, and 2.00, respectively (Fig. 7, G-J). The P_o values at different Ca²⁺ concentrations obtained from nine separate experiments are plotted in Fig. 7K. The data could be fitted by a bell-shaped curve, indicating that the cloned RyR3 channel is activated at low Ca²⁺ concentrations (about 100 nM) and inactivated at high Ca²⁺ concentrations (about 10 mM).

The relationship between the mean open time and the Ca²⁺ concentration is shown in Fig. 7L (solid circles). It could also be fitted by a bell-shaped curve. The mean open time increased from about 1-2 to about 10-30 ms when Ca²⁺ concentrations were raised from pCa 7.19 to about pCa 4.0. However, when Ca²⁺ concentrations were further increased from about pCa 4.0 to pCa 2.0, the mean open time decreased from 10-30 ms to less than 1 ms. In contrast, the mean closed time decreased from several seconds to about 1 ms when Ca²⁺ concentrations were raised from pCa 7.19 to 4.0 and increased from about 1 ms to about several milliseconds when Ca²⁺ concentrations were further raised from pCa 4.0 to 2.0 (Fig. 7L) (open circles).

To determine if ATP can activate the cloned RyR3 at very low levels of Ca²⁺, we first inhibited the cloned RyR3 channel by addition of 0.1 mM EGTA. Similar to that shown in Fig. 7A, the cloned RyR3 channel was inhibited completely by 0.1 mM EGTA (data not shown). Subsequent addition of 1.5 mM ATP did not activate the channel significantly (Fig. 8A). The average P_o in the presence of 0.1 mM EGTA and 1.5 mM ATP was 0.00031 ± 0.00020 (n = 6). Subsequent addition of 0.05 mM CaCl₂, however, activated the channel substantially (Fig. 8B). The average P_o in the presence of 0.1 mM EGTA, 0.05 mM CaCl₂ (64 nM free Ca²⁺), and 1.5 mM ATP was 0.039 ± 0.028 (n = 5). The gating kinetics of the ATP-activated channel could be described by two open time constants of 0.47 ± 0.15 and 6.37 ± 3.57 ms and a closed time constant of 50.6 ± 30 ms (n = 5). The extent of activation in the presence of 64 nM Ca²⁺ plus 1.5 mM ATP (Fig. 8B) was much greater than the sum of that in the presence of 1.5 mM ATP (Fig. 8A) and 64 nM Ca²⁺ (Fig. 7B) alone. Thus, these data indicate that the cloned RyR3 is activated by Ca²⁺ and ATP synergistically.

The effect of ATP on single channel activity of the cloned RyR3 in the presence of high Ca²⁺ concentrations is shown in Fig. 8C. In the presence of 1.5 mM ATP, 30 mM CaCl₂ failed to inactivate the cloned RyR3. The average P_o in the presence of 30 mM CaCl₂ plus 1.5 mM ATP was 0.78 ± 0.13 (n = 4). It should be noted that in the presence of 30 mM CaCl₂ the single channel conductance was reduced. These results suggest that the cloned RyR3 is insensitive to Ca²⁺ inactivation in the presence of ATP.

Calmodulin (CaM) has been shown to have dual effect on RyR1 channel activities. At low Ca²⁺ concentrations (<= 200 nM), CaM activates RyR1, but at micromolar to millimolar concentrations, CaM inhibits channel activity (41). To examine whether the cloned RyR3 is regulated by CaM in a similar manner, the effect of calmodulin on single channel activity of the cloned RyR3 was investigated at both low and high Ca²⁺ concentrations (Fig. 9). At 88 nM cis (cytoplasmic) Ca²⁺, the channel was activated slightly (Fig. 9A). Addition of 1 µM CaM to the cis (cytoplasmic) chamber resulted in a further activation of the channel (Fig. 9B). The average P_o increased 29.5 ± 15.9-fold and the average mean open time increased 3.0 ± 0.6-fold. The average mean closed time decreased to 16.0 ± 11.5% (n = 5). The channel could also be activated by CaM at low Ca²⁺ concentrations in the presence of 2.0 mM ATP (Fig. 9D). The average P_o increased 8.7 ± 3.5-fold and the mean open time increased 3.6 ± 1.0-fold. The average mean closed time decreased to 26.8 ± 8.5% (n = 3).

On the other hand, at 50 µM cytoplasmic Ca²⁺ concentration, addition of 1 µM CaM decreased the channel activity markedly (Fig. 10). The average P_o decreased to 28.4 ± 9.4% and the average mean open time decreased to 12.8 ± 8.4%. The average mean closed time increased 3.76 ± 0.86-fold (n = 4). Further addition of CaCl₂ did not reverse the inhibitory effect of CaM (data not shown).

Recent studies suggest that cyclic ADP-ribose (cADPr) is a second messenger capable of triggering intracellular Ca²⁺ release in a variety of cell types (42, 43). The cADPr-induced Ca²⁺ release is thought to be mediated by ryanodine receptors and requires CaM (44, 45). Moreover, cADPr appeared to activate RyR2 but not RyR1 in lipid bilayers (12, 13, 46, 47). To determine whether cADPr can activate the cloned RyR3, we examined the effect of cADPr on single channel activity of the cloned RyR3 in the presence of CaM. As illustrated in Fig. 11, addition of 10 µM cADPr to the cloned RyR3 channel stimulated by 88 nM Ca²⁺ plus 1 µM CaM (Fig. 9B) (n = 5) or stimulated by 50 µM Ca²⁺ plus 1 µM CaM (Fig. 10B) (n = 4) did not cause, in either case, significant changes in P_o, mean open time, or mean closed time (Fig. 11, A and B).

Perchlorate anions have been shown to enhance excitation-contraction coupling in skeletal muscle but not in cardiac muscle (14, 48-51). At the single channel level, perchlorate anions activate the RyR1 channel (51). It is not known, however, whether perchlorate anions affect RyR3. We tested the effect of perchlorate on the cloned RyR3 channel activity using a protocol similar to that developed by Ma et al. (51). Since micromolar Ca²⁺ is required for the activation of RyR1 by perchlorate (14), a cloned RyR3 single channel was first activated by 100 µM Ca²⁺. P_o was then lowered by addition of 6 mM Mg²⁺ to detect the stimulatory effect of perchlorate (data not shown). Under these conditions, P_o, mean open time, and mean closed time were 0.006, 0.49, and 34 ms, respectively. Addition of 30 mM sodium perchlorate to the cytoplasmic side of the Mg²⁺-inhibited channel resulted in a significant increase in channel activity. P_o increased to 0.143 and mean open time increased to 0.73 ms. Mean closed time decreased to 3.22 ms. On average, P_o increased 15.4 ± 8.1-fold and mean open time increased 2.1 ± 0.9-fold and mean closed time decreased to 13.5 ± 5.3% (n = 7). At low Ca²⁺ concentrations, however, the cloned RyR3 channels were not affected significantly by perchlorate (data not shown).

DISCUSSION

The mammalian RyR3 protein has been detected in mammalian brain and skeletal muscle (7, 52). However, the RyR3 protein consists of only 1-2% or less of the total amount of ryanodine receptor isoforms present in these tissues (18, 53). The low abundance of RyR3 protein and coexpression of RyR3 with other ryanodine receptor isoforms have hampered the biochemical isolation and functional characterization of RyR3. As a result, very little information about the channel properties of the native mammalian RyR3 is available.

While isolation and characterization of the native RyR3 have been difficult, molecular cloning and expression of cDNA in heterologous systems, followed by characterization of the expressed protein, may provide an alternative approach to study the channel properties of RyR3. This approach has been successfully used to study the structure and function of RyR1 (28, 54-57). Recently, single channel properties of the recombinant RyR1 have been extensively characterized. The RyR1 channels expressed in HEK293 cells exhibit conductance, gating kinetics, Ca²⁺ permeability, Ca²⁺ activation, Ca²⁺ inactivation, and ligand gating properties indistinguishable from those of the native rabbit skeletal muscle Ca²⁺ release channel (58).

In the present study, we expressed the RyR3 cDNA in HEK293 cells and characterized the cloned RyR3 by using single channel recordings in planar lipid bilayers. Results of these studies have provided initial insights into the single channel properties of RyR3. The cloned RyR3 forms a cation channel with a large conductance (777 pS in 250 mM KCl and 137 pS in 250 mM CaCl₂), which is selective for divalent over monovalent cations with a permeation ratio (pCa²⁺/pK⁺) of 6.3. The cloned RyR3 can be activated by submicromolar Ca²⁺, millimolar ATP, and millimolar caffeine, is inhibited by millimolar Ca²⁺, millimolar Mg²⁺, and micromolar ruthenium red, and is modified by ryanodine. Furthermore, it is activated by CaM at low Ca²⁺ concentrations but inhibited by CaM at high Ca²⁺ concentrations. Perchlorate is able to activate the cloned RyR3. These single channel properties of the cloned RyR3 are similar to those of the recombinant RyR1 (58) and the native RyR1 (1-5, 41).

Cyclic ADP-ribose which is capable of activating the RyR2 but not RyR1 channel appears to have no significant effect on single channel activity of the cloned RyR3 in the presence of CaM in lipid bilayers (Fig. 11). In addition to CaM, however, other factors such as CaM-dependent protein kinase II (59) have also been shown to mediate cADPr-induced Ca²⁺ release. The lack of effect of cADPr on single channel activity of the cloned RyR3 might be due to the absence or loss of some of these factors in our detergent-solubilized and purified recombinant RyR3 preparations.

While the overall single channel properties of the cloned RyR3 are similar to those of RyR1, several important differences have been identified. First, the cloned RyR3 differs from RyR1 in gating kinetics. The cloned RyR3 channels have longer mean open time than RyR1. The fitted open time constant of the cloned RyR3 determined at 100-200 nM activating Ca²⁺ is about 1.16 ms, while that of the recombinant or native RyR1 is about 0.22 ms (58). Furthermore, the mean open time of the recombinant RyR1 changed about 2-fold with varying Ca²⁺ concentrations (58), whereas the mean open time of the cloned RyR3 varied more than 10-fold when Ca²⁺ was raised from pCa 7.19 to 4 (Fig. 7L). The second major difference of the cloned RyR3 from RyR1 is in its sensitivity to inactivation by Ca²⁺. The cloned RyR3 channel is about 10 times less sensitive to inactivation by high Ca²⁺ concentrations. P_o of the recombinant or native RyR1 channel was reduced to close to zero at 1-3 mM Ca²⁺ concentrations (5, 58), whereas significant RyR3 channel activity could still be detected at Ca²⁺ concentrations as high as 20 mM (Fig. 7K). Third, the cloned RyR3 differs markedly from RyR1 in the extent of maximal activation by Ca²⁺. The maximal P_o of the recombinant or native RyR1 channel activated by Ca²⁺ alone was about 0.2-0.6 (5, 58), whereas the cloned RyR3 could be fully activated by Ca²⁺ alone (Fig. 7K). It is apparent that the level of maximal activation of RyR by Ca²⁺ depends on both the extent of Ca²⁺ activation and the extent of Ca²⁺ inactivation. In the case of RyR1, Ca²⁺ inactivation is prominent and may develop before the channel is maximally activated by Ca²⁺. Therefore, the modest level of maximal activation of RyR1 by Ca²⁺ may be partially due to Ca²⁺ inactivation of the channel.

The Ca²⁺ concentration required to achieve half-maximal activation has often been used to describe the sensitivity of RyR to activation by Ca²⁺. However, like the level of maximal activation, the apparent half-maximal activation level by Ca²⁺ could be affected by Ca²⁺ inactivation. Therefore, the sensitivity of RyR1 to activation by Ca²⁺ based on the half-maximal activation might have been overestimated due to the presence of a large extent of Ca²⁺ inactivation of RyR1.

To minimize the influence of Ca²⁺ inactivation, a threshold of activation rather than a half-maximal activation of RyR was used in our analyses to describe the sensitivity of the cloned RyR3 or RyR1 to activation by Ca²⁺. As shown in Fig. 7, the cloned RyR3 remained closed until the Ca²⁺ concentration was raised to about 100 nM which we denote as the threshold of activation of the cloned RyR3. This value is similar to that observed for the recombinant RyR1 (58). The estimation of the threshold of activation may, however, vary with different conditions of purification, reconstitution, and channel recordings. To compare directly the sensitivities of the cloned RyR3 and RyR1 to Ca²⁺ activation, we determined the response of purified rabbit skeletal muscle RyR1 to various Ca²⁺ concentrations under the same conditions used for the cloned RyR3 in our recording system. We found that the threshold of activation of the purified native RyR1 channel was about 100 nM. Thus the sensitivities of RyR1 and the cloned RyR3 to Ca²⁺ activation in the lipid bilayer system are similar. We also found that the kinetics, the effect of Ca²⁺ concentrations on mean open time, the extent of maximal activation by Ca²⁺, and the sensitivity to Ca²⁺ inactivation of the native RyR1 were essentially the same as those of the recombinant RyR1 observed previously (58).

The sensitivity to Ca²⁺ activation of RyR1 and RyR3 in muscle fibers has recently been investigated using muscles isolated from mice lacking RyR3 or RyR1. In Ca²⁺ release experiments using permeabilized muscle fibers (21), it was found that the amount of Ca²⁺ release in 30 s in the RyR1-deficient muscle (or RyR3-containing muscle) at low Ca²⁺ concentrations (pCa 7-5.5) was much less than that in the RyR3-deficient muscle (or RyR1-containing muscle). On the basis of these results, it was suggested that the RyR3 channel may be less sensitive to activation by Ca²⁺ than the RyR1 channel in mammalian skeletal muscles.

These results appear to be different from those obtained from the bilayer experiments. The molecular basis for this apparent discrepancy is not clear. It is possible that the sensitivities of the RyR1 and RyR3 channels to Ca²⁺ activation may be modulated differently by various factors and conditions present in muscle fibers (5). It is also possible that mammalian skeletal muscles may express an alternatively spliced variant of RyR3 with lower sensitivity to Ca²⁺ activation than the cloned RyR3 that we expressed in HEK293 cells. Another possibility is that variations in the amounts of Ca²⁺ release observed in the RyR1- and RyR3-deficient muscle at low Ca²⁺ concentrations may be due to different levels of the RyR3 and RyR1 protein present in RyR1- and RyR3-deficient muscle. It is known that the RyR3 protein is present at a much lower level in mammalian skeletal muscles as compared with RyR1 (52, 53). Accordingly, the lower amount of Ca²⁺ release in the RyR1-deficient muscle at low Ca²⁺ concentrations may be explained by the lower level of the RyR3 channel protein.

The non-mammalian homologue of RyR3 has been isolated and extensively studied in vitro (5). Non-mammalian skeletal muscles express  and  RyR isoforms, corresponding to the mammalian RyR1 and RyR3, respectively. Unlike mammalian skeletal muscles in which RyR3 exists as a minor component, non-mammalian skeletal muscles express equivalent levels of  and  RyR isoforms. It was not surprising that the single channel properties of the cloned RyR3 were found to be very similar to those of the avian  RyR, considering that the amino acid sequence identity between rabbit and avian RyR3 is more than 86% (37). Both the cloned RyR3 and the avian  RyR exhibit similar gating kinetics with longer open time constants and are less sensitive to Ca²⁺ inactivation as compared with RyR1 and  RyR. The lack of inactivation by high Ca²⁺ concentrations has also been observed with the fish  RyR (60). Furthermore, the cloned RyR3 was not inactivated by Ca²⁺ concentrations as high as 30 mM in the presence of ATP, similar to that observed with the avian  RyR (61). The sensitivity of the avian  RyR to activation by Ca²⁺ is similar to or slightly higher than that of the  RyR (61). However, the cloned RyR3 appeared to be different from the avian  RyR in the response to perchlorate anions. The cloned RyR3 but not the  RyR was activated by perchlorate (61). The structural basis and functional significance of this difference are unclear.

Despite these similarities between the cloned RyR3 and native non-mammalian RyR3, it is uncertain whether single channel properties of the cloned RyR3 reflect those of the native mammalian RyR3. It is possible that HEK293 cells may lack some factors that are required for proper expression and function of RyR3, even though these cells appear to be able to express recombinant RyR1 proteins with channel properties similar to those of the native RyR1. Consequently, channel properties of the cloned RyR3 expressed in HEK293 cells may differ from those of the native RyR3. Therefore, the native mammalian RyR3 has yet to be isolated and characterized at the single channel level.

Functional studies have suggested that in some cells RyR3 may serve as a Ca²⁺ release channel with properties different from those of RyR1 and RyR2. For example, fura-2 and ⁴⁵Ca²⁺ efflux measurements revealed that Ca²⁺ release from intracellular stores of human uterine smooth muscle cells was ryanodine-sensitive but caffeine-insensitive (62, 63). In addition, ryanodine caused an increase in intracellular Ca²⁺ concentrations in human Jurkat T-lymphocytes that express RyR3, but caffeine did not (64). Mink lung epithelial cells where RyR3 is expressed also exhibited ryanodine-sensitive, caffeine-insensitive Ca²⁺ release activity (65). Based on these results, it appeared that RyR3 may be a caffeine-insensitive Ca²⁺ release channel. However, caffeine- and ryanodine-sensitive Ca²⁺-induced Ca²⁺ release was observed in RyR1-deficient (or RyR3-containing) muscles (20). Results from the present study also indicate that the cloned RyR3 is sensitive to both caffeine and ryanodine. The molecular basis for these differences is not known. It has been suggested that lack of caffeine response in some cell types may result from low levels of RyR isoforms expressed in these cells (7, 20). Another possibility is that different tissues may express different alternatively spliced variants of RyR3 with different caffeine sensitivities. In favor of this possibility, tissue-specific alternatively spliced transcripts of the mink and mouse ryr3 gene have recently been reported (36, 66). In our amplification of the rabbit uterus RyR3 mRNA, we also detected several deletions in the RyR3 mRNA (67). These deletions presumably resulted from alternative splicing of the rabbit ryr3 gene. Alternative splicing might also generate RyR3 variants with different conductances or different response to ryanodine, like those observed in aortic tissues (15). Expression and functional characterization of these putative alternatively spliced variants of RyR3 should provide some insights into the functional heterogeneity and structure and function relationships of the RyR3 Ca²⁺ release channel.