Unique phosphorylation site on the cardiac ryanodine receptor regulates calcium channel activity.

Ryanodine receptors have recently been shown to be the Ca2+ release channels of sarcoplasmic reticulum in both cardiac muscle and skeletal muscle. Several regulatory sites are postulated to exist on these receptors, but to date, none have been definitively identified. In the work described here, we localize one of these sites by showing that the cardiac isoform of the ryanodine receptor is a preferred substrate for multifunctional Ca2+/calmodulin-dependent protein kinase (CaM kinase). Phosphorylation by CaM kinase occurs at a single site encompassing serine 2809. Antibodies generated to this site react only with the cardiac isoform of the ryanodine receptor, and immunoprecipitate only cardiac [3H]ryanodine-binding sites. When cardiac junctional sarcoplasmic reticulum vesicles or partially purified ryanodine receptors are fused with planar bilayers, phosphorylation at this site activates the Ca2+ channel. In tissues expressing the cardiac isoform of the ryanodine receptor, such as heart and brain, phosphorylation of the Ca2+ release channel by CaM kinase may provide a unique mechanism for regulating intracellular Ca2+ release.

Ca2+ conductances (2-7), protease sensitivities (11, E ) , calmodulin-binding capabilities (ll), and modulation by allosteric regulators such as Ca2+, Mg2+, ATP, and calmodulin (13-15), they also exhibit several differences in protein structure and function. Quantitative differences have been noted on the effects of modulators on ryanodine binding to the two proteins (16-18), as well as on Ca2+ channel kinetics (13, 14). In addition, the cardiac ryanodine receptor exhibits a slightly smaller apparent molecular weight than the skeletal muscle receptor on SDS-PAGE ( l l ) , and monoclonal antibodies can be made which react with the cardiac receptor but not the skeletal receptor (16).
Recent work on characterization of the two ryanodine receptors has culminated in elucidation of the primary structures of the proteins by sequencing of their cDNAs (19-21). Consistent with the differences between the two protein isoforms noted above, the cardiac and skeletal muscle receptors have been found to be the products of different genes, with overall amino acid identities of 66% (21). Both protein isoforms are very large, containing approximately 5,000 amino acids and exhibiting predicted molecular weights of 564,711 for the cardiac protein (21) and 565,223 (19) or 563,584 (20) for the skeletal muscle protein. In the native state, ryanodine receptors are arranged as tetramers (1-7).
In an earlier study (22), we demonstrated that the canine cardiac high molecular weight protein (or ryanodine receptor; Ref. 3) was an excellent substrate for the multifunctional CaM kinase (23,24) endogenous to junctional SR membranes. In the work described here, we show that phosphorylation of the cardiac receptor by CaM kinase occurs a t a single site, which is not substantially phosphorylated in the skeletal muscle receptor, and that phosphorylation of the cardiac ryanodine receptor at this site activates the Ca2+ channel. Our data are the first to support the theoretical model of Otsu et al. (21), that the modulator-binding sites of the cardiac ryanodine receptor are contained within residues 2619-3016.

EXPERIMENTAL PROCEDURES
Materials-CaM kinase was isolated from rat brain as described previously (25). cAMP kinase (the catalytic subunit from bovine heart) was purchased from Sigma or prepared according to Beavo et al. (26). Identical results were obtained with either cAMP kinase preparation. [y-"PJATP and "'I-protein A were obtained from Du Pont-New England Nuclear. Bovine brain calmodulin, protein Aagarose, iminodiacetic acid-agarose, and Sephacryl S-500 were purchased from Sigma.
Purification ~f f'hospho~lntcd ('nrdinr Rynnrdinr. Hccrptnr ~ 50 mg of cardiac junctional Sf< vesicles were phosphonlatetl for 5 min at 30 "C in 50 ml of huffer containing 50 mM hLIOI' S ( p H 7.4).  T n p t i r I'rotco/ysis and Isolntion of I'h~~splrr~~ptidr--liOO-HOO p g nf purified phosphorylated ryanodine receptors frnm two rnlumn runs weredialyzed against 4 liters of50 mM NH,HCO,. 1 nlM dithiothreitol for I6 h at room temperature. Tr.ypsin was ad(letl at a 1/50 weight ratio. and proteolysis contlucted  trifluoroacetic acid. Ahsorhance was measured at 214 nm and radioactivity of recovered fractions was monitored hy Cerenkov rounting. Only a single major peak o f radioactivity was detected. which eltrtrd at 24% acetonitrile and accounted for 7 7 ; o f the total radioactivity applied to the column. The sequence o f the recovered radioactive peptide was determined hv automated Edman tlegradation using an Applied Hiosvstems Model 470A gas-phase sequenatnr ( 2 1 I. I'hosphoamino acid analysis was performed as tlescrihetl (:I1 1. Antisrrum Production-Peptide 2805-2819 o f the rardiac ryanodine receptor (Fig. 5) was synthesized with an added rysteine nt the carhoxyl terminus and coupled t o thyroglohulin ( 3 2 ) . A rahbit was immunized suhcutaneouslv with :I00 pg of peptide con.jrrgate in romplete Freunds adjuvant, and hooster injections o f 200 p g trf peptide conjugate in incomplete adjuvant were given at monthly intervals. Antiserum was ohtainetl on day 7 after hooster injections.

lc/t panel (I'mtrin S t a i n )
shows the proteins transferred to the nitrocellulose sheet stained with Amitlo Hlack and the right pond (Immunoblnt) shows the autoradiogram of the same sheet after incuhation with a l/lOOO dilution of antipeptitle antihodv followed hv ""I-protein A. 100 pg of S R protein were electrophoresed per gel lane. I,onc, 1, canine cardiac SI3 vesicles; lonr 2, canine fast skeletal SI1 vesicles; lanr 3 , canine slow skeletal SR vesicles; lnnc 4. rahhit skeletal SR vesicles. Nli indicates the ryanodine receptor in cardiac (lnrrrr arrow) and skeletal (upprr arrow) samples.
S i n g k Chnnnrl Rrcordings-Single channel recordings were performed using the planar hilayer method as descrihed previously (34). Ultrapure phosphatidylserine and phosphatidvlethanolamine (Avanti Polar Lipids) were dried in a nitrogen stream and resuspended in decane at a 4/5 weight ratio, respectively. Rilavers were formed in a 150-pm hole separating two chamhers of a custom designed hath, and single channels were recorded using either a Dagan :K)O0 or 8900 patch clamp amplifier. Signals were digitized and stored on standard video tape for off-line analysis using an IRM compatihle personal computer fitted with custom software. IJsual hilaver conventions are followed; the chamher ( 4 ml) designated cis is the chamher to which D. R. Witcher and I,. R. .Jones. unpuhlished data.

RESIILTS ANI) 1)ISClISSION
Prcfcrcntial Phosphovlation of the Cardiac l?\.onodinr. Rcwpptor- Fig. 1 shows that the canine cardiac ryanodine receptor (arrou~hcnds) is phosphorylated in junctional SK vesicles by an endogenous calmodulin-requiring protein kinase and that this phosphorylation is stimulated severalfold when exogenous CaM kinase is added. In contrast, the ryanodine receptor in canine fast and slow skeletal muscle S R vesicles, which migrates with a slightly higher molecular weight on SDS-PAGE (2,11,16), is not significantly phosphorylated hy either endogenous or exogenous protein kinase (Fig. 1, small arrows). Similar results were ohtainetl with rabbit skeletal muscle SR vesicles.
The identity of the skeletal muscle ryanodine receptor in these studies (Fig. 1, small arrow) was confirmed by immunoblotting with a skeletal muscle isoform-specific antiborly (supplied by K. Campbell, University of Iowa). \i'e did detect a low level of phosphorvlation of a protein in slow skeletal muscle samples migrating slightly faster than the cardiac receptor ( Fig. 1, nstcrish), but this protein did not cross-react with skeletal muscle (or cardiac, see below) antibodies, suggesting that it is unrelated to the ryanodine receptor.
Although we reported earlier that a rabbit skeletal muscle high molecular weight protein could he phosphorylated by an endogenous CaM kinase (11), it is probable that much of the 9 ' incorporation detected in this earlier study was due to phosphorylation of this unrelated protein. Pretreatment of SR membranes with acid phosphatase t o remove endogenous phosphate did not affect the results described presently; CaM kinase-catalyzed phosphorylation of the cardiac ryanodine receptor was always at least IO-fold greater than skeletal receptor phosphorylation. Chu et al. (3.5) recently reported that the skeletal muscle ryanodine receptor is phosphorylated by CaM kinase, but the present results demonstrate that this phosphorylation is insignificant compared to cardiac protein phosphorylation. Consistent with our results, Otsu ct al. (21 1 have recently shown that, the cardiac isoform of the ryanodine receptor is absent from fast and slow skeletal muscle. As reported earlier (ll), and later confirmed by Takasago ct af. ( X ) , cAMP kinase also phosphorylates the cardiac ryanodine receptor. However, maximal phosphnrylat inn by added cAMP kinase is no greater than that achieved with endogenous CaM kinase (Fig.  2). In contrast. 1/2Oth the amount of exogenous CaM kinase increases receptor phosphorylation 4-fold, to a maximal level of 26 pmol of P,/mg of SR protein (Fig. 2). Thus, efficient phosphorylation of the cardiac ryanodine receptor occurs only with CaM kinase. In agreement with recent results (3.5, X ) , we observed no significant phosphorylation of canine fast and slow or rabbit skeletal muscle ryanodine receptors by cAMJ' kinase, although the adventitious protein mentioned above was phosphorylated.
Maximal ["Hlryanodine binding ( 3 ) in these preparations ranged between 5 and 6 pmol/mg of protein, a value nearly identical to the level of receptor phosphorylation achieved with exogenous cAMP kinase (see also, Kef. 3 6 ) or endogenous CaM kinase, but one-fourth the value achieved with added CaM kinase. Since the functional unit o f the cardiac ('a"' release channel contains only one high affinity ryanorlinebinding site/tetramer (4), our results suggest that the endogenous CaM kinase is capable of phosphorylating only onefourth of the available sites, whereas the exogenous kinase can fully phosphorylate the receptor, i.c. all four subunits (see below).

CHANNEL CURRENT (PA)
Sequencing of the Cardiac Phosphorylation Site-In order to sequence the phosphorylation site of the cardiac ryanodine receptor, we phosphorylated junctional SR membranes on large scale with added CaM kinase and purified the phosphorylated denatured ryanodine receptor to homogeneity in one step using SDS-gel filtration chromatography (Fig. 3). The purified cardiac ryanodine receptor was digested with trypsin, and the radioactive peptides recovered using Fe3+ affinity chromatography (30,37). 90% of the loaded radioactivity was recovered in the pH 8.6 and 10 eluates from the Fe'+ column (Fig. 4). These fractions were then combined and subjected to reverse-phase chromatography, yielding a single major radioactive peptide peak eluting at approximately 24% acetonitrile (Fig. 4, inset). Gas-phase sequencing of the radioactive tryptic peptide gave a single sequence of 18 consecutive residues, which corresponded exactly to residues 2807-2824 reported for the rabbit cardiac ryanodine receptor from cDNA cloning (Fig. 5) (21). When CNBr and endoproteinase Lys-C were used to cleave the receptor, another "P-labeled peptide was isolated and sequenced, which matched with residues 2800-2811 of the rabbit cardiac ryanodine receptor (Fig. 5).
Serine 2809 within the phosphorylated tryptic peptide is situated on the carboxyl-terminal side of 2 arginine residues.
The fact that R-R-X-S and R-X-X-S/T are minimal consensus phosphorylation sequences (38,39) for CAMP kinase and CaM kinase, respectively, makes this residue the likely phosphorylation site. Consistent with this, the ratio of dithiothreitol-serine to phenylthiohydantoin-serine recovered during cycle 3 of sequencing of this peptide was 10 times greater than that recovered during cycles 6 and 9. It is known that dithiothreitol-serine is the predominant breakdown product of phosphoserine (40,41). Phosphoamino acid analysis revealed that this peptide contained only phosphoserine; moreover, >90% of the 3'Pi was released from the peptide by cycle 10 (40, 42), demonstrating that no serine residue downstream of this region was significantly labeled.' Based on these results, we conclude that serine 2809 is the amino acid phosphorylated by CaM kinase.
When only endogenous CaM kinase was used to phosphorylate the cardiac ryanodine receptor, the same labeled tryptic peptide was recovered and sequenced in four separate runs. Thus, although exogenously added kinase gives a 4-fold stimulation of receptor phosphorylation (Fig. 2), no new sites are phosphorylated. The reason for the low level of phosphorylation obtained with endogenous CaM kinase remains undefined, but could be due to membrane-limited structural constraints, or to endogenous phosphatase activity in the membranes (43) overwhelming a low level of kinase. It should be pointed out that the heart contains both soluble (44) and particulate (22) forms of CaM kinase. The sequence surrounding residue 2809 of the rabbit and human skeletal muscle receptors is quite different from that of the cardiac receptor (Fig. 5). Out of 25 residues, 14 or 15 residues, respectively, are altered. These differences may account for the insignificant phosphorylation of the skeletal muscle isoforms (see last section).
Antibody to the Cardiac Phosphorylation Site-In order to confirm that the phosphorylation site sequenced originated from the cardiac isoform of the ryanodine receptor, we raised a rabbit antiserum to a synthetic peptide (residues 2805-2819) encompassing serine 2809.
Immunoblot analysis revealed that this antiserum reacted only with the cardiac ryanodine receptor (Fig. 6). Ryanodine receptors in canine fast and slow and rabbit skeletal muscle membranes were unreactive. Moreover, the antiserum immunoprecipitated [3H]ryan~dine-binding sites solubilized from cardiac membranes but did not sediment binding sites solubilized from either fast or slow skeletal muscle membranes ( Table I). The antiserum also recognized the human cardiac ryanodine receptor, but not the human skeletal muscle ryanodine receptor.' In preliminary experiments, we have used this same antiserum to immunoprecipitate >90% of the ryanodine receptors solubilized from neuronal membranes, which like cardiac receptors, are substrates for CaM kinase (45). Thus our results confirm that the cardiac isoform of the ryanodine receptor is expressed in brain (21), and demonstrate that the unique phosphorylation site is conserved.
Phosphorylation Effect on Ca'+ Channel Actiuity-Previous studies have shown that ATP activates, whereas calmodulin inhibits the ryanodine receptor/Ca'+ release channel (13-15). Consistent with previous results (3, 13-15), we observed that cardiac junctional SR vesicles fused with the planar bilayer exhibited a characteristic 75 picosiemens divalent cation conductance, which required micromolar Ca2+ on the cis side of the bilayer and was regulated by ATP and calmodulin. With 11 PM Ca2+ and 3 mM Mg'+ on the cis side of the bilayer (Fig.  7A), addition of 1 mM ATP to the cis side produced prolonged openings of the channel and increased the open state probability p(open) from 0.26 to 0.81 (Fig. 7B). Cis addition of 3 p M calmodulin then inhibited channel activity, reducing the p(open), shortening the mean open time, and producing prolonged closures (Fig. 7C).
Subsequent addition of CaM kinase to the cis side of the bilayer, under conditions known to give phosphorylation of the cardiac ryanodine receptor, reversed the inhibitory effect of calmodulin on p(open) (Fig. 8, A and B ) , and restored the prolonged openings of the channel (Fig. 8, analog tracing) observed in the presence of ATP (Fig. 7 B ) . No effect on the unitary current of the channel was seen. In multiple experiments, we observed that the net effect of CaM kinase was to reverse the inhibitory effect of calmodulin. No effect of the kinase on channel activity was observed when it was added to the trans side of the bilayer, nor was an effect seen in the absence of ATP or calmodulin. We also observed no effect of added CaM kinase when skeletal muscle SR vesicles were used.
To rule out the possibility that phosphorylation of an accessory protein by CaM kinase was activating the Ca2+ channel in cardiac SR vesicles, we prepared a partially purified preparation of the ryanodine receptor by solubilization of cardiac membranes in CHAPS followed by sucrose density gradient centrifugation (Fig. 9). The ryanodine receptor in this partially purified preparation was visualized by Coomassie Blue staining (left panel) and, importantly, was the only protein phosphorylated by exogenously added CaM kinase (right panel). Phosphorylation by CaM kinase was blocked by the affinity purified antibody (data not shown), demonstrating that the same region of the receptor was phosphorylated in the partially purified preparation as in native SR vesicles." The partially purified receptor was fused with the planar bilayer, and single channel activity was measured. As shown in Fig. 10, addition of CaM kinase to the bilayer bath (I?) increased channel activity substantially over that observed in the presence of ATP and calmodulin ( A ) , similar to results obtained with native S R vesicles. Thus, phosphorylation of the ryanodine receptor itself was sufficient to activate the channel.
Functional Considerations-As our work was being completed, Otsu et al. (21) reported the deduced primary structure of the rabbit cardiac ryanodine receptor, which contains the phosphorylation site we have presently sequenced. Our results suggest that phosphorylation a t a single site is sufficient to " T h e protein appeared as a doublet in this experiment due to partial proteolysis which occurred during the gradient run.
Hoth bands of the douhlet reacted with the site-specific antibody. The mohility form of the partially proteolyzed cardiac receptor (12) is apparent as a faint band in Figs. 1 and 6. change channel function. Since serine 2809 is remote in sequence from the putative pore-forming carboxyl-terminal transmembrane segments (19)(20)(21). it is apparent that some well-orchestrated mechanism must exist to transduce the phosphorylation signal into a change in Ca" conductance. Although our results do not address this mechanism, they do show that at least one regulatory site of the molecule is located within the region recently postulated to contain the modulator-binding sites (residues 2619-3016) (21). As a corollary, our results are the first to definitively localize this region of the molecule to the cytoplasmic side of the SR membrane.
At present, it is not possible to conclude whether phosphorylation a t one subunit or all four subunits is required for activation of the channel; in this regard it would be helpful to know whether the endogenous CaM kinase of native SR vesicles survives fusion with the planar bilayer, and remains capable of phosphorylating the ryanodine receptor to a low stoichiometry. In multiple experiments, we ohserved significant activation of the channel only when exogenous CaM kinase was added.
Distributed throughout the cardiac ryanodine receptor (21) are 16 candidate phosphorylation sites for multifunctional CaM kinase, even when the consensus site of the kinase is narrowly defined as R-X-X-S/T (38,39). There are also five candidate phosphorylation sites for cAMP kinase based on a consensus of R-R-X-S/T, where R represents a hasic residue, either Arg or Lys (38). Remarkably, only serine 2809 of the cardiac ryanodine receptor is phosphorylated by CaM kinase.
In particular, although CaM kinase phosphorylates pyruvate kinase as well as synthetic peptides at a sequence containing R-R-A-S, with a motif similar to R-R-I-S in the cardiac isoform of the ryanodine receptor (39, 46, 47), it does less well when a lysine replaces an arginine residue (191, as occurs in the rabbit skeletal muscle receptor (Fig. 5 ) . For example, a synthetic peptide containing the sequence K-R-I-S is not phosphorylated a t all (39). This same region of the skeletal muscle ryanodine receptor (Fig. 5) has been postulated to bind calmodulin (20), and it is also possible that phosphorylation of this site could be prevented by the presence of bound calmodulin. Such protection of a phosphorylation site by calmodulin is seen when myosin light chain kinase in its calmodulin-bound state is used as a substrate for CAMP kinase (48). The cardiac ryanodine receptor would be expected to bind calmodulin less well at this site since it differs markedly in the number of basic residues on the amino-terminal side of serine 2809 (Fig. 5).
Interestingly, the serine 2809 phosphorylation site is at the carboxyl-terminal end of the third of four large repeated sequences found in the cardiac ryanodine receptor (21). While the function of these repeats is unknown, it is possible that phosphorylation may modulate the receptor by altering the function of these repeated sequences.
Phosphorylation of the ryanodine receptor/Ca'+ release channel offers an attractive mechanism for increasing myoplasmic Ca'+ concentration during &adrenergic stimulation of the heart. Work with another cardiac SR protein. phospholamban, already demonstrates that during &adrenergic stimulation both CaM kinase and cAMP kinase are activated and phosphorylate S R proteins (42). In the case of phospholamhan phosphorylation, Ca2+ uptake into the SR is stimulated, resulting in an increased rate of myocardial relaxation (49,50). Simultaneous phosphorylation of the ryanodine receptor could increase the amount of Ca2+ released from the SR, contributing to the increased force of contraction. In neurons, phosphorylation of the cardiac isoform of the ryanodine receptor could contribute to the increase in cytoplasmic Ca" con- centration occurring during membrane depolarization and/or neurotransmitter release. In either case, the phosphorylation effect on Ca2+ channel activity seems to be modulatory, relieving an inhibitory effect already caused by calmodulin.