Rapid Calcium Release from Cardiac Sarcoplasmic Reticulum Vesicles Is Dependent on Ca2+ and Is Modulated by Mg2+, Adenine Nucleotide, and Calmodulin*

A subpopulation of canine cardiac sarcoplasmic re- ticulum vesicles has been found to contain a “Ca2+ release channel’’ which mediates the release of intravesicular Ca2+ stores with rates sufficiently rapid to contribute to excitation-contraction coupling in cardiac muscle. 45Ca2+ release behavior of passively and actively loaded vesicles was determined by Millipore filtration and with the use of a rapid quench apparatus using the two Ca2+ channel inhibitors, Mg2+ and ruthenium red. At pH 7.0 and 5-20 PM external Ca2+, cardiac vesicles released half of their 45Ca2+ stores within 20 ms. Ca2+-induced Ca2+ release was inhibited by raising and lowering external Ca2+ concentration, by the addition of Mg2+, and by decreasing the pH. Calmodulin reduced the Ca2+-induced Ca2+ release rate 3-6-fold in a reaction that did not appear to involve a calmodulin- dependent protein kinase. Under various experimental conditions, ATP or the nonhydrolyzable ATP analog, adenosine 5’-(B,y-methylene)triphosphate (AMP-PCP), and caffeine stimulated 4SCa2+ release 2-500- fold. Maximal release rates (tU = 10 ms) were observed in media containing 10 PM Ca2+ and 5 mM AMP-PCP or 10 mM caffeine. An increased external Caz+ concentration (21 mM) was required to optimize the 45Ca2+ efflux rate in the presence of 8 mM Mg2+ and 5 mM AMP-PCP. These results suggest that cardiac sarco- plasmic reticulum contains a ligand-gated Ca2+ channel which is activated by Ca2+, adenine nucleotide, and caffeine, and inhibited by Mg2+, H’, and calmodulin.

Heavy rabbit skeletal muscle SR Ca2+ release vesicle fractions were recovered from the 36-45% region of a sucrose gradient that contained membranes sedimenting at 2,600-35,000 X g (13).
4aCa2+ Flux Measurements with Passively Loaded Vesi~les-'~Ca*+ efflux rates from vesicles passively loaded with 45Ca2+ were determined with the use of an Update System 1000 Chemical Quench apparatus (Madison, WI) and by Millipore filtration (13, 15). Unless otherwise indicated, vesicles (2-10 mg of protein/ml) were passively loaded for 60 min at 23 "C in a medium containing 1.1 mM 45CaC12, 100 p M EGTA, 100 mM KCl, 1 mM diisopropyl fluorophosphate, and 20 mM K/Pipes, pH 7.0. 45Ca2+ efflux behavior of the vesicles was measured by diluting vesicles 5-300-fold into iso-osmotic unlabeled release media containing varying concentrations of free CaZ+, adenine nucleotide, and M e . In the rapid quench experiments, &Ca2+ efflux was inhibited at time intervals ranging from 25 to 1000 ms by the addition of 10 mM M e , 5 mM EGTA, and 10 p~ ruthenium red (final concentrations). Extravesicular 45Ca2+ was separated by placing the vesicles on 0.45-p Millipore filters followed by rapid rinsing with 100 mM KC1, 10 mM M$+, 0.1 mM EGTA, 10 FM ruthenium red, 20 mM K/Pipes, pH 7.0. Radioactivity retained by the vesicles on the filters was determined by liquid scintillation counting. 4sCaz+ flux measurements were carried out at least in duplicate with two or more time points. For a given preparation the standard errors were +15% or less.
Incubation of Passively Loaded Vesicles with Calmodulin-Vesicles (1 mg of protein/ml) were initially incubated for 10 min at 23 "C in a medium containing 20 mM K/Pipes, pH 7.0, 0.1 M KCl, 100 p M Ca2+, 100 p~ EGTA, 1 mM M F , 10 mM D-glucose, and hexokinase (Sigma, 10 units/ml) in order to remove possible small amounts of contaminating ATP. After sedimentation and resuspension, vesicles (2 mg of protein/ml) were incubated at 23 "C with or without exogenously added calmodulin or TnI in a medium containing 20 mM K/ Pipes, pH 7.0, 0.12 M KCl, 0.1 mM EGTA, and 0.1 mM 45Ca2c. After 5 min, the Ca2+ concentration was raised to 1.1 mM and vesicles were incubated for an additional 60 min at 23 "C.
The amount of ATP remaining in vesicles after treatment with hexokinase was determined by high performance liquid chromatography. Vesicles were sedimented, resuspended, and extracted for 10 min on ice with 1 N perchloric acid. After removal of the membranes, the extract was neutralized with 5 N K2C0, and analyzed for ATP content on a SynChropak AX100 column (Linden, IN).
Skeletal TnI was purified and stored as described by Potter (18).
dialyzed overnight a t 4 "C against 0.6 M KCl, 100 p M EGTA, 100 p M Prior to the flux studies, TnI (protein concentration 25 gM) was Ca2+, and 20 mM K/Pipes, pH 7. 45Ca2+ Flux Measurements with Actively Loaded Vesicles-Vesicles (25-35 pg of protein/ml) were incubated for 1 min at 23 "C in a medium containing 20 mM K/Pipes, pH 7.0,O.l M KC1,lOO p M '%a2+, 50-600 p~ EGTA, and 8 mM M$+. Uptake media also contained 5 mM NaN3 and 1.5 p~ carbamyl cyanide N-chlorophenylhydrazone in ordar to inhibit possible CaZ+ uptake by contaminating mitochondria. %a2+ uptake was initiated by the addition of 5 mM ATP and was determined at varying time intervals by placing an aliquot of 0.4 ml on a 0.45-p Millipore filter followed by a 10-s rinse with 20 mM K/ Pipes, pH 7.0, 0.1 M KC1 medium that either inhibited (plus 10 mM M e , 0.1 mM EGTA, and 10 p M ruthenium red) or promoted (plus 100 p~ EGTA and 106 p~ Ca2+, 10 p~ free Ca2+) "Ca2+ efflux from Ca2+ release vesicles. Radioactivity retained by actively loaded vesicles on the filters was determined by liquid scintillation counting.
Ca*+ loading rates were determined in media containing 0.1 mM 45Ca2+, 5 mM MgZ+, 5 mM ATP, and 5 mM oxalate. Vesicles were subsequently rinsed with the channel inhibiting medium after placement onto the filter. t'H/Ryanodine Binding-Cardiac vesicles (0.5 mg of protein in 0.25 ml) were incubated at 37 "C in a medium containing 20 mM K/ Pipes, pH 7.0, 0.1 M KC1, 1 mM diisopropyl fluorophosphate, 0.1 mM EGTA, 0.1 mM Ca", and [3H]ryanodine (specific activity, 60.8 c i / mmol). After 2 and 4 h, an aliquot of the vesicles (0.2 mg of protein) was sedimented by centrifugation for 30 min a t 90,000 X g in a Beckman Airfuge. Pellets were briefly washed at 4 "C and radioactivity associated with the vesicles and radioactivity remaining in the supernatant were determined by liquid scintillation counting to obtain bound and free [3H]ryanodine. ['HIRyanodine was added to a concentration of 5 nM, greater concentrations were prepared as admixtures of labeled and unlabeled ryanodine. BiochemicalAssays-Protein was determined by the Lowry method using bovine serum albumin as a standard. "Basic" and Me-dependent, Caz+-stimulated ATPase activities were determined in the presence of the ionophore A23187 (2 pg/ml) at 32 "C as previously described (13). Free Ca2+ and Mg2+ concentrations were calculated according to a computer program using binding constants published by Fabiato (19); free divalent cation concentrations in the presence of AMP-PCP (15) were estimated using the ATP binding constants published by Fabiato.

Isolation of Cardiac Ca2+
Release Vesicles-A crude microsomal fraction of canine cardiac muscle was separated into five fractions on a sucrose gradient ( Table I). The Ca2+ release activities of the five vesicle fractions were determined as outlined for Fraction IV in Fig. 1. Vesicles were passively loaded with 1 mM 45Ca2+ by incubation for 60 min a t 23 "C and then either diluted into a medium which inhibited or activated the Ca2+ release channel of cardiac SR. 45Ca2+ efflux was slow when the vesicles were placed into a medium containing 10 mM Mg2+ and 10 p M ruthenium red (Fig. L4). This allows one to determine the amounts of 45Ca2+ (22 nmol/mg protein) trapped by the vesicles in the incubation medium. When vesicles were diluted into a medium which contained 10 p~ free Ca", about half of this amount was released in less than 30 s. Most of the 45Ca2+ remaining with the vesicles in the 10 p~ Ca2+ release medium was present in a releasable form since less than 1 nmol of 45Ca2+/mg protein was retained when the Ca2+ ionophore A23187 (2 pg/ml) was present. Consequently, Fraction IV appeared to be made up of two subpopulations of vesicles, only one of which was capable of rapidly releasing its 45Ca2+ stores in the 10 ~L M Ca2+ medium. Among the five fractions, the 31-40% sucrose gradient fractions (Fractions IV and V) contained the highest percentage of vesicles displaying Ca2+ release activity (Table I). In comparison, the activity of the Ca2+-ATPase, measured in the   TABLE I Enzymatic and Ca2+ release properties of sucrose gradient membrane fractions of canine cardiac muscle Canine cardiac membranes were subfractionated into five membranous fractions on a sucrose gradient (cf. "Experimental Procedures"). Ca2+ release properties of vesicles passively loaded with 1 mM "Ca2+ were determined as described in the legend to Fig. 1. The total amount of "Ca2+ trapped by the vesicles was determined by diluting vesicles 100-fold into a release medium containing 10 mM M T plus 10 p M ruthenium red. The percent of 45Ca2+ release indicates the portion of trapped 45Ca2+ that was rapidly released in the 10 p M free ca2+ medium. Ca2+-ATPase activity was determined as described under "Experimental Procedures." The data are the average of five preparations f S.E.  Table I) were passively loaded with 1 mM "Caz+ as described under "Experimental Procedures." 4sCa'+ efflux was initiated by diluting vesicles 100-fold into release media containing either 10 mM M e , 1 mM EGTA plus 10 p M ruthenium red ( R R ) (0) or 1 mM EGTA plus 0.95 mM Ca2+ (10 p M free Ca" after the addition of the vesicles) (0). 45Ca2+ efflux was terminated by placing vesicles on 0.45-p Millipore filters followed by rapid rinsing with release medium to remove extravesicular 45Caz+. Amounts of '5CaZ+ trapped by all intact vesicles in the incubation medium (22 nmol/mg of protein) as well as amounts not readily released in the 10 p~ Ca2+ release medium (10 nmol/mg of protein) were obtained by back extrapolation to the time of vesicle dilution. B, rapid initial 4sCa2+ efflux rates were determined with the use of an Update System 1000 Chemical Quench apparatus (15). Vesicles (2 mg of protein/ml) passively loaded with 1 mM 4sCa2+ were diluted with 4 volumes of release medium containing 0.5 mM EGTA, 0.24 mM Ca2+ and 0 (0) or 0.875 mM M e (A). 4sCazf release in the presence of 5 mM Mg2+ and 5 mM nucleotide was measured by adding 4 volumes of 10.4 mM M e plus 11.25 mM AMP-PCP in a second mixing step a t 72 ms to 10 p M Ca", 0.7 mM M e release medium (0). Rapid 4sCa2+ efflux was inhibited at the indicated times (50 and 140 ms in A ;  ms in B ) by the addition of 4 volumes of a quench solution containing M e , EGTA, and ruthenium red (final concentrations 10 mM, 5 mM and 10 pM, respectively). Vesicles were subsequently placed on 0.45-p Millipore filters and rinsed with a medium containing 10 mM M g + , 0.1 mM EGTA, and 10pM ruthenium red. The amount of 4sCa" remaining in the vesicles was determined by back extrapolation to the time of addition of the quench solution.

Sucrose gradient fractions
presence of the ionophore A23187, was highest in the 19-28% sucrose gradient fractions (Fractions I and 11).
The time course of the rapid release phase in Fig. lA was determined by inhibiting 4sCa2+ efflux at short time intervals by the addition of Mg2+ and ruthenium red using a rapid quench apparatus (15). In preliminary experiments it was found that the cardiac channel was less sensitive than the skeletal channel to inhibition by Mg2' and ruthenium red.
Effective inhibition of 45Ca2+ release from cardiac vesicles required the presence of both M$+ and ruthenium red in the quench solution. Omission of either 10 p~ ruthenium red or 10 mM Mg2+ from the channel inhibiting medium resulted in incomplete inhibition of rapid 45Ca2+ efflux. Release of more than one-half of the 45Ca2+ stores in the Ca2+-permeable vesicle population occurred within 30 s, i.e. when the first time point was taken using the Millipore filtration technique. In the presence of adenine nucleotide, an additional requirement was that the free Ca2+ concentration was decreased to below 1O"j M Ca2+ during the quench step. This was achieved by including in the quench solution the Ca2+ chelating agent EGTA. Fig. 1A shows that an appreciable fraction of the rapidly releasable 45Ca2+ remained with the vesicles when 45Ca2+ efflux was inhibited at 50 and 140 ms by the addition of the two channel inhibitors, Mg2+ and ruthenium red. In Fig. lB, rapid 45Caz+ efflux was stopped at time intervals ranging from 25 to 500 ms. In the 10 p~ Ca2+ medium, the Ca2+ releasing vesicle fraction released its %az+ stores with a half-time of about 25 ms. 45Ca2+ efflux was slowed about 5-fold by the addition of 0.7 mM Mg2+ to the 10 p~ Ca2+ medium. In the presence of 10 p~ ca2+ and 5 mM Mg.AMP-PCP (0.7 mM free M e ) , 45Ca2+ was released with a rate comparable to that in the Mg+-and nucleotide-free release medium. Taken together, the data of Fig. 1 suggest that cardiac sarcoplasmic reticulum vesicles contain a Ca2+ release channel, which in the presence of micromolar external Ca2+, mediates the release of Ca2+ with rates sufficiently rapid to contribute to excitation-contraction coupling. This channel appears to be inhibited by M$+ and activated by adenine nucleotides.
In Fig. 2, vesicles were passively loaded for varying times with 1 mM 45Ca2+. After incubation of 1 min to 4 h at 23 "C, vesicles were diluted into Ca2+ release blocking medium to estimate the amounts of 45Ca'+ that were accumulated by all vesicles. Dilution into the Ca2+ release promoting medium was used to determine the fraction of vesicles that lacked the Ca2+ release channel. Fig. 2 shows that 45Ca2+ influx was a rapid process with a half-time of about 2 min. The difference  Table I)

ea2+ Release by Cardiac
Sarcoplasmic Reticulum in the two uptake curves, i.e. the amount of 45Ca2+ that was released in the 10 p~ Ca2+ medium, remained fairly constant with time, suggesting that a similar number of vesicles containing viable Ca2+ release channels were present during the 4-h incubation period. Comparison of the release activity of passively and actively loaded vesicles (see below) also suggested that the Ca2+ permeability of the vesicles remained largely unaltered during incubation for 1-2 h at 23 "C. In other experiments it was found that the Ca2+ pump mediated Ca2+ loading rate of the vesicles (0.5 pmol/mg of protein/min) was independent of the time of vesicle incubation (not shown).
[3H]Ryan~dine binding and 45Ca2+ flux measurements have indicated that ryanodine specifically binds to the skeletal and cardiac SR Ca2+ release channels (20-22). Cardiac preparations have been reported by Pessah et al. (20) to bind 0.5 pmol of ryanodine/mg of protein with a KD of 36 nM and 1.7 pmol/ mg of protein with a KD of 340 nM. The Ca2+ releasing membrane fractions obtained in this study (Fractions IV and V of Table I) contained an appreciably higher concentration of specific high-affinity ryanodine binding sites (9 f 3 pmol/ mg of protein). Scatchard plot analysis indicated the presence of a single high-affinity receptor site with a KO of about 7 nM ( Fig. 3). Specific high-affinity [3H]ryanodine binding to membranes displaying a low Ca2+ release activity (Fraction I of Table I) amounted to 2-3 pmol/mg of protein (not shown).
Inhibition of 45Ca2+ Release-Ca2+-induced 45Ca2+ release is partially inhibited by 1 p~ ruthenium red, 1 pM neomycin, and 50 p~ tetracaine (Fig. 4). Preincubation of the vesicles with the inhibitors was required to observe the extent of inhibition seen in Fig. 4. Without prior exposure, a large fraction of the intravesicular 45Ca2+ was rapidly released before the three compounds were capable of exerting their inhibitory effects.
The effect of varying external M$+ concentration on 45Ca2+ release is shown in   an apparent n value of 1.5 for Mg2+ inhibition of 45Ca2+ release (not shown). A Hill coefficient of 1.5 is in accord with the existence of multiple interacting Mg2+ binding sites. Due to the complexity of the system, other explanations cannot, however, be ruled out at present. Fig. 6 shows that the Ca2+ release behavior of cardiac vesicles is affected by calmodulin. In order to establish that calmodulin acted directly on the channel as observed by us for skeletal SR (16) rather than via a calmodulin-dependent protein kinase as suggested by Tuana and MacLennan (23), vesicles were incubated with D-glucose, M$+, and hexokinase, sedimented, resuspended in release medium, and extracted with perchloric acid. High performance liquid chromatography analysis of the neutralized extract indicated that the amount of ATP remaining in the vesicles was less than the limit of detection (100 pmol/mg of protein). After treatment with hexokinase, vesicles were preincubated with or without 6 p~ calmodulin and diluted into media containing or lacking calmodulin. The inset of Fig. 6 shows that exposure of the vesicles to calmodulin increased the tlh of 45Ca2+ efflux from about 25 to 150 ms. On a time scale of 30-60 s, similar amounts of 45Ca2+ were released from vesicles exposed and not exposed to calmodulin. The similar extent of release in the presence  Table I) were treated with D-glUCOSe and hexokinase and passively loaded with 1 mM 45Ca2+ in the absence (0, 0 ) or presence of 6 p M calmodulin (0, . ) as described under "Experimental Procedures." After the addition of vesicles, release media contained 0 (0, 0 ) or 1.2 g M (0, . ) calmodulin and 5 p M free ca2+ (0, 0 ) or 10 mM Me, 1 mM EGTA plus 10 p~ ruthenium red (RR) (W, 0). 45Caz+ release rates from the Ca'+-permeable vesicle fraction in the 5 p~ Ca2+ release medium were determined as indicated in the legend to Fig. 1 Table I) were treated with D-glucose and hexokinase and passively loaded with 1 mM 45Ca2+ in the presence of the indicated concentrations of TnI and calmodulin as described under "Experimental Procedures." 4sCa2+ efflux rates from the Ca2'-permeahle vesicle population were determined by diluting vesicles &fold into a release medium (10 p M free Ca2+ after the addition of the vesicles) and were measured with the use of a rapid quench apparatus as indicated in the legend to Fig. 1 and absence of calmodulin suggested that the effect of calmodulin was to slow down the release of 45Ca2+ from all Ca2+ release vesicles rather than to inhibit a selected fraction of the vesicles. Variation of the calmodulin concentration in vesicle and release media indicated that the Ca2+ release rate was half-maximally reduced at 0.1-0.2 p~ calmodulin and maximally reduced a t a calmodulin concentration of 1-3 p~ (not shown). For comparison, the calmodulin concentration of heart has been estimated to be 2-3 p~ (24). Cardiac vesicles were incubated with the calmodulin-binding component of Tn12 to assess possible inhibition of 45Ca2+ release by an impurity present in the commercial preparation of calmodulin or by endogenous calmodulin. Table I1 shows J. Potter, personal communication.
that TnI was effective in neutralizing the inhibitory effect of 1 p~ calmodulin in that it maintained the release rate at a value seen in the absence of added calmodulin. Inhibition of the channel by another component present in the calmodulin preparation appears therefore to be unlikely. In the absence of added calmodulin, TnI increased the release rate of cardiac vesicles by about 15%. This result raises the possibility that isolated cardiac vesicles contain small amounts of calmodulin which inhibit the channel.
pH Dependence of 45Ca2+ Release and Exchange-The pH dependence of 45Ca2+ efflux from cardiac vesicles was measured in a medium containing 5 p~ or 1 mM free Ca2+ (Fig. 7).
In the presence of a maximally activating Ca2+ concentration of 5 p~ (see below), vesicles released half their 45Ca2+ within 20 ms at pH 7.0. At pH 6.0, 45Ca2+ release was about 50-fold reduced, occurring with a half-time of 1 s. Cardiac vesicles were equilibrated with 1 mM 45Ca2+ by incubating for 1 h (Fig.  2). An increase in external Ca2+ concentration to 1 mM in the release medium therefore allowed measurement of the Ca2+ release behavior of the vesicles under 4sCa2+-40Ca2+ exchange conditions. Calcium exchange proceeded with about a &fold lower rate than 45Ca2+ release at pH values below 7. The calcium exchange rate steadily increased from pH 6 to 8, without appearing to reach a maximal value at pH 8.
Activation of 45Ca2' Release by ea2+, Nucleotides, and Caffeine-The 'Ta2+ release behavior of canine cardiac vesicles has been studied in the presence of Ca2+, ATP (AMP-PCP), and caffeine, and compared with the release behavior of rabbit skeletal muscle SR Ca2+ release vesicles (Table 111). At pH 7.0 in 10-~ M free Ca2+, 45Ca2+ release was slow, requiring a tJh of 25 and 8 s for cardiac and skeletal vesicles, respectively. An increase in free Ca2+ to 2 p~ and addition of the nonhydrolyzable ATP analog AMP-PCP increased the release rate by a factor of about 1000 in both preparations. In the presence of micromolar Ca2+ and millimolar nucleotide, nearly optimal release rates were observed since the cardiac and skeletal channels are essentially open all the time, as shown previously in single channel measurements (6, 17). Intermediate release rates were obtained when the two channels were activated at lo-' M Ca2+ by either ATP or caffeine or by M Ca2+ alone.  Table I) were passively loaded with 1 mM 45Ca2+ by incubating vesicles for 60 min at 23 "C in a medium containing 0.1 M KC1 and either 20 mM K/Pipes (pH 6.0-7.0) or 20 mM K/Hepes (pH 7.0-8.0). Vesicles were diluted into media containing 5 p~ (0) or 1 mM free (0) Ca2+, after the addition of the vesicles. *'CaZ+ efflux rates from the Ca2+-permeable vesicle fraction were determined as described in the legend to Fig. 1  Activation of "Ca" release from cardiac and skeletal muscle SR vesicles by Ca2+, adenine nucleotide, and caffeine "Ca2+ efflux rates were determined as described in the legend to Several important differences were noted to exist between cardiac and skeletal SR. At M external Ca2+, the tt,$ of 45Ca2+ release was 20 ms for cardiac vesicles, as compared to 600 ms for skeletal vesicles. Caffeine mimicked the effects of external Ca2+ in that it was more effective in stimulating 45Ca2+ release from cardiac than skeletal vesicles in the low Ca2+ medium. At micromolar Ca", 10 mM caffeine, like 5 mM nucleotide, optimally stimulated 45Ca2+ efflux from cardiac vesicles, whereas in skeletal vesicles it was without a significant effect. In contrast, adenine nucleotides were more effective in stimulating 45Ca2+ release from skeletal than cardiac vesicles. Addition of 5 mM ATP to the lo-' M Ca2+ medium increased 45Ca2+ efflux 100-fold from skeletal vesicles, as compared to an only %fold increase from cardiac vesicles.

Ca2+
Dependence of Ca2+ Release-We have attempted to mimic the ionic conditions in relaxed and contracted muscle by determining the Caz+ dependence of Ca2+ release from cardiac vesicles in the presence and absence of Mg2+ and the nonhydrolyzable ATP analog AMP-PCP. Canine myocardium contains about 5 mM ATP (28). The free Mg2+ concentration of cardiac muscle is not known but has been estimated to range from about 0.2 to 4 mM in skeletal muscle (29,30). In the absence of Mg2+ and nucleotide, a maximal release rate was measured in media containing about 5-20 p~ free Ca2+ (Fig. 8). 45Ca2+ release was slow at nanomolar Ca2+ and was severalfold reduced at millimolar concentrations of Ca2+. Effects of external Ca2+ on 45Ca2+ efflux in Fig. 8 can be most easily explained by assuming that the channel possesses activating and inhibitory Ca2+ binding sites.
Adenine nucleotide modified Ca2+-induced Ca2+ release from cardiac SR in two ways. One effect of AMP-PCP was to increase the maximally observable Ca2+ release rate. In Fig.  8, the t,,+ of 45Ca2+ release was decreased from about 25 to 15 ms by 5 mM AMP-PCP. A second effect of AMP-PCP was to shift the Ca2+ activating curve to the left by lowering the halfmaximally activating Ca2+ concentration from about 2 x 10"j M to 8 X M. Variation of AMP-PCP concentration a t 3 x M Ca2+ indicated that a high concentration of nucleotide (5-10 mM range) was required to optimally stimulate 45Caz+ release. At 3 X loF7 M Ca2+, the 45CaZ+ efflux rate was half-maximally accelerated by 1-2 mM AMP-PCP (not shown).
Mg2+ affected the channel in an opposite manner. Ca2+ release rates in the micromolar Ca2+ concentration range were greatly reduced by 3 mM Mg2+, and millimolar free Ca2+ concentrations were required to activate the channel (Fig. 8).
The addition of 8 mM M$+ and 5 mM AMP-PCP (3 mM free Mg2+) to the Ca2+ release media shifted the Ca2+ activation curve to higher Ca" concentrations, similarly as observed for 3 mM Mg2+ alone (Fig. 8). An important difference was, however, that the channel could be nearly fully activated in the presence of nucleotide and Mgz+ by raising the external Ca2+ concentration to 1 mM. 45Ca2+ Release in Actively Loaded Vesicles-Dependence of the release behavior of cardiac vesicles on free Ca2+ concentration was also studied using actively loaded vesicles (Fig.  9). Vesicles were actively filled with 45Ca2+ in media containing 8 mM M$+, 5 mM ATP, and varying concentrations of free Ca'+. The amounts of 45Ca2+ sequestered in the loading medium by vesicles containing and lacking the Ca2+ release channel were determined using conditions similar to those of Fig. 1. Actively loaded vesicles were placed on Millipore filters and rinsed for 10 s with a medium which either inhibited (10 mM M8+, 0.1 M EGTA, 10 p~ ruthenium red) or activated (10 p~ free Caz+) the channel. The difference in 45Ca2+ retained by vesicles under the two washing conditions was defined as the amount of 45Ca2+ taken up by the Ca2+-permeable vesicle population.
Cardiac vesicles which were rinsed with the 10 mM M e , 10 p~ ruthenium red medium retained between 30 and 35 nmol of 45Ca2+/mg of protein after loading for 10-20 min at 5 X M free Ca2+ (Fig. 9A). In agreement with previous studies (31), a decrease in the free Ca2+ concentration of the loading medium to 0.3 and 0.1 p~ lowered the initial uptake rate and the amount of 45Ca2+ that was sequestered by the vesicles within 10-20 min. In contrast, Fig. 9B gives the fraction of sequestered 45Ca2+ that was released when vesicles were rinsed for 10 s with the channel activating rinse medium (10 p~ Ca2+). Although a change in the composition of the rinse medium was without appreciable effect when vesicles were actively loaded in the presence of 50 pM Ca2+, an increasing fraction of the sequestered 45Ca2+ was released when the free Ca2+ concentration in the 45Ca2+ uptake medium was  Table I) were actively loaded as described under "Experimental Procedures" in media containing 8 mM M$+, 5 mM ATP, and the indicated concentrations of free ea2+. After 0.5-20 min, vesicles were placed on 0.45-p Millipore filters and rinsed 10 s with medium containing 10 mM M P , 0.1 mM EGTA, and 10 pM ruthenium red. The amount of '%a2+ retained by the vesicles on the filters is given. B, conditions are the same as A except that vesicles were rinsed 10 s with 10 p~ Caz+ medium instead of with the Mg+/rutheniurn red medium. The fraction of %az+ that was released during rinsing with the 10 p~ Ca2+ medium is shown. lowered from about 1 to 0.1 pM. About one-half of the sequestered 45Ca2+ was released from vesicles that were actively loaded at 0.3 and 0.1 p~ Ca".
The specificity of 45Ca" uptake and release was assessed by actively loading vesicles in the presence of 0.1 M NaCl and by preincubation with 1 FM ryanodine. In the presence of 0.1 M NaC1, 45Ca2+ uptake by sarcolemmal vesicles is minimal due to the immediate release of any accumulated Ca2+ through the Na+/Ca2+ exchange system (32). In the experiments of Fig. 9, substitution of 0.1 M KC1 by 0.1 M NaCl did not significantly reduce the amount of 45Ca2' retained by vesicles loaded in the presence of 0.3 FM free Ca2+ and rinsed with the channel inhibiting medium, suggesting that 45Ca2+ uptake by contaminating sarcolemmal vesicles was negligible. Low concentrations of ryanodine "open" the cardiac Ca2+ release channel and thereby render passively loaded vesicles permeable to Ca2+ in Ca2+ release channel activating and inhibiting media (22). In actively loaded vesicles, preincubation with 1 PM ryanodine for 5 min at 37 "C reduced 1.5-%fold the amount of 4sCa2+ that was retained by vesicles loaded at 0.3 p~ free Ca2+ and rinsed with the channel inhibiting medium. Only a small fraction of the sequestered 45Ca2+ was released when ryanodine-treated vesicles were rinsed with channel activating medium, suggesting that ryanodine rendered Ca2+ release channel containing vesicles permeable to 45Ca2+. Possible Ca2+ uptake by contaminating mitochondria was blocked by actively loading vesicles in the presence of 5 mM NaN3, an inhibitor of the mitochondrial electron transport chain, and carbamyl cyanide N-chlorophenylhydrazine, a potent inhibitor of oxidative phosphorylation.
Data obtained with actively and passively loaded vesicles are in good general agreement. At free Ca2+ concentrations of 1 PM and more, Ca2+ release vesicles are incapable of sequestering significant amounts of 45Ca2+, because the cardiac Ca2+ release channel is sufficiently activated to cause the rapid release of the transported 45Ca2+. On the other hand, because of channel inactivation, the Ca2+ transport rate exceeds the release rate at submicromolar Ca2+ concentrations in the presence of Mg2+ and nucleotide, with the result that significant amounts of the transported 45Ca2+ are retained by Ca2+ release vesicles.

DISCUSSION
Results presented here show that a subpopulation of canine cardiac SR vesicles contains a ligand-gated Ca2+ release channel that mediates the rapid release of intravesicular Ca2+ stores from passively loaded vesicles. At least two lines of evidence suggest that the channel is localized in sarcoplasmic reticulum as opposed to another cellular structure. First, passively loaded "release" vesicles from cardiac membranes demonstrated 45Ca2+ release behavior similar to vesicles actively loaded via the SR Ca2+ pump. Second, Ca2+ release from cardiac vesicles qualitatively resembled Ca2+ release from heavy skeletal muscle SR vesicles. Both channels were activated by external Caz+ and adenine nucleotide and inhibited by Mg" and calmodulin. Another argument strongly favoring localization of the Ca2+ channel in the SR membrane is that Ca2+ release from skeletal and cardiac SR vesicle fractions used in this study was similarly affected by the plant alkaloid ryanodine (22).
Ca2+-induced calcium release from passively and actively loaded cardiac vesicles has been previously observed. 45Ca2+ efflux was stimulated a t micromolar Ca2+ and was sensitive to inhibition by ruthenium red (33-35). However, the release rates reported were several orders of magnitude slower than those observed in the present study. In a preliminary report, Ca2+ release peaked at 0.2 and 1 p M external Ca2+, suggesting the possible involvement of two components in the regulation of the cardiac Ca2+ release channel (36). The presence of a specific Ca2+ permeability mechanism in a subpopulation of cardiac SR vesicles has also been suggested in studies with the plant alkaloid ryanodine (20-22, 35, 37-40). Th' 1s compound profoundly alters the Ca2+ uptake and release properties of isolated cardiac and skeletal SR vesicle fractions by specifically binding to the Ca" release channels.
Cardiac SR Ca2+ release vesicle fractions used in this study were prepared according to a protocol similar to the one used for the isolation of heavy SR junctional derived Ca2+ release vesicles from rabbit skeletal muscle (13). Canine ventricular tissue was homogenized in buffered 0.3 M sucrose solution and a Ca2+ release vesicle fraction was isolated in the presence of the protease inhibitor diisopropyl fluorophosphate by differential and sucrose gradient centrifugation. Between 40 and 65% of the vesicular 45Ca2+ could be rapidly released from the 30-40% sucrose gradient fractions, as compared to 70-90% from heavy skeletal muscle SR vesicles. ['HH]Ryanodine binding measurements indicated a similar number of specific highaffinity binding sites (7-14 pmol/mg protein) as reported for skeletal SR junctional derived vesicles (4-20 pmol/mg protein; Refs. 20 and 21). In contrast, Pessah et al. (20) obtained only 0.5 pmol of high-affinity binding sites/mg of protein using a purified cardiac SR preparation.
Ca2+ release behavior of skeletal muscle SR vesicles has been extensively studied using rapid flow and rapid quench techniques (11,12,(14)(15)(16), while comparable detailed studies with cardiac SR vesicle fractions have not been previously reported to our knowledge. The nonlinear 45Ca2+ efflux curves on the semilog plots of this study indicate that Ca2+ release from cardiac vesicles is more difficult to evaluate by kinetic analysis than that from skeletal Ca2+ release vesicle fractions. The latter display a higher Ca2+ release activity and more importantly, appear to be made up of a fairly homogenous population of vesicles (41), with the result that 45Ca2+ efflux can be reasonably well fitted by a single exponential function and thus approximated by first-order kinetics. Data of the present study nevertheless show that, in a qualitative manner, the cardiac and skeletal channels are similarly affected by Ca2+, Mg", H ' , adenine nucleotide, and calmodulin. In both systems, regulation of the channel can be summarized by the equation shown below. calrnodulin] The cardiac and skeletal channels are opened for intermittent times in the presence of micromolar calcium and millimolar adenine nucleotide. The presence of both ligands is required to keep the two channels open essentially all the time (6, 17). The equation also depicts M$+ and H+ as inhibitors of the channel.
In muscle, regulation of the channel is more complex than indicated by the above equation in that channel behavior is modulated by Mg-ATP and calmodulin (enclosed in box). Most of the ATP in muscle is present in the form of a Mg. ATP complex which modulates the Ca2+ dependence of 45Ca2+ release by skeletal (15) and cardiac (Fig. 8) SR vesicles. Furthermore, the rate of Ca2+ release from the Ca2+-permeable subpopulation of skeletal (16) and cardiac (Fig. 6) SR vesicles is decreased by a factor of 2-6 by exogeneously added calmodulin. Inhibition occurred in the absence of added ATP and therefore does not appear to involve a calmodulin-dependent phosphorylation reaction. Instead experiments with rabbit skeletal muscle SR vesicles have suggested that calmodulin slows down 45Ca2+ release by directly interacting in a rapid, reversible, Ca2+-dependent manner with the channel (16). The physiological function of calmodulin in excitation-contraction is difficult to define at present. One proposal is that calmodulin, by partially inhibiting release of Ca2+ from SR, compensates for possible increases in the resting Ca2+ level during increased muscle activity.
Several important differences exist in the extent to which the cardiac and skeletal channels are activated by Ca2+ and adenine nucleotide and inhibited by Mg". Micromolar concentrations of Ca2+ stimulate cardiac 45Ca2+ release in the absence of Mg2+ and nucleotide to as much as 50% of the optimal rate, whereas only 1-3% of the optimal rate could be achieved in skeletal vesicles in the absence of nucleotide and Mg2+ (15). A similar discrepancy in the rates of 45Ca2+ release was observed for vesicles diluted into low Ca2+ media containing 10 mM caffeine (Table 111). In contrast, adenine nucleotides were more effective in stimulating 45Ca2+ release from skeletal than from cardiac vesicles. This difference in effectiveness was particularly apparent a t low Ca2+ concentrations. At IOu9 M Ca", ATP was minimally effective in activating the cardiac channel. By comparison, in skeletal muscle, 5 mM ATP was effective in stimulating 45Ca2+ release at lo-' M Ca2+ to as much as 15% of the optimal rate (Table 111). Another significant difference is that 45Ca2+ efflux from cardiac vesicles is less sensitive to Mg2+ inhibition than that from skeletal vesicles. Thus, in the presence of nucleotide, it was possible to observe at high concentrations of inhibitory Mg2+ (3 mM), vesicular 45Ca2+ release rates which would appear to be significant to excitation-contraction coupling (Fig. 8). In order to observe similarly rapid release rates from skeletal vesicles, it was necessary to decrease the free Mg2+ concentration in the presence of nucleotide to 5 X M or lower (15). A Ca2+ concentration of 1 mM and higher was required to optimally activate the cardiac Ca2+ release channel in the presence of adenine nucleotide and Mg2+ (Fig. 8). This Ca2+ concentration is 2-3 orders of magnitude higher than the free Ca2+ concentration measured in contracting muscle (42), thus raising the question of whether cardiac SR in muscle is capable of rapid Ca2+-induced Ca2+ release in the presence of physiological concentrations of nucleotide and M P . It can be argued, however, that Ca2+ concentrations in excess of measured values may exist in the junctional area during surface membrane Ca2+ channel activation, considering that extracellular Ca" is 1-2 mM. In this regard it is of interest that recent sucrose gradient centrifugation studies with vesicle fractions have suggested co-migration of the dihydropyridine labeled cardiac surface membrane channel and the SR junctional feet (43). This would suggest that, as in skeletal muscle (3,21,35,41), surface membrane and SR Ca2+ channels are in close proximity.
It has been proposed that physiological release of Ca2+ from sarcoplasmic reticulum is induced by a second messenger such as Ca2+ (3,5) or inositol 1,4,5-triphosphate (25)(26)(27), "depolarization" of the SR membrane (3), changes in membrane surface charge (44, 45), or a pH gradient (46). Among these various putative effectors we have found Ca2+ to cause the most dramatic changes in the Ca2+ release behavior of cardiac SR vesicle fractions. In our bilayer (17) and vesicle flux experiments, we have been unable to demonstrate activation of the cardiac SR Ca2+ channel by inositol 1,4,5-triphosphate.
Also, the cardiac channel displays only a weak voltage dependence in planar lipid bilayers (17). Furthermore, changes in membrane surface charge, brought about by the addition to nanomolar Ca2+ release medium of 20 PM tetraphenylboronor tetraphenylarsonium+ (45), did not stimulate 45Ca2+ release by more than a factor of five.3 A similar result was seen upon the creation of a pH gradient, brought about by diluting vesicles from pH 6.5 to 7.5 and from pH 7.5 to 6.5.
On the other hand, a raise of external Ca2+ from lo-' to M Ca2+ increased the rate of 45Ca2+ release from cardiac SR vesicles by a factor of 1000 (Table 111, Fig. 8).
The cardiac and skeletal Ca2+ channels display a similar dependence on Ca", peaking a t micromolar external free Ca2+ in the absence of Mg2+ and adenine nucleotide. Addition of adenine nucleotide and Mg2+ shifts the Ca2+ concentration required for maximal Ca2+ release from about 2-20 WM Ca2+ to about 100-1000 PM (Fig. 8, Ref. 15). Also, in the skeletal system, the Hill plot n value of CaZ+ activation increases from 1 to 2 in the presence of adenine nucleotide and Mg2+. A consequence of this increase in cooperativity is that the skeletal channel is rendered sensitive to external Ca2+ in a quite narrow concentration range of about to loe4 M (15).
Similarly, 45Ca2+ release from cardiac vesicles could be nearly fully inhibited or activated in the presence of M$+ and adenine nucleotide by varying the external Ca2+ concentration from about 10"j to M (Fig. 8). One important conclusion that can be drawn from these data is that under ionic conditions resembling those in muscle, Le. in the presence of M$+ and nucleotide, the cardiac channel, like the skeletal channel, displays a strict dependence on Ca2+. In comparison, ATP and M$+ may play a more limited role in regulating Ca2+ release; our data suggest that ATP and Mg2+ act as allosteric effectors of the channel by modulating the channel's sensitivity to external Ca2+. This interpretation is supported by the fact that, in contrast to Ca", the concentrations of M e and nucleotide are thought to remain fairly constant during all phases of muscle activity.