Spatial and Dynamic Interactions between Phospholamban and the Canine Cardiac Ca2+ Pump Revealed with Use of Heterobifunctional Cross-linking Agents*

Heterobifunctional thiol to amine cross-linking agents were used to gain new insights on the dynamics and conformational factors governing the interaction between the cardiac Ca2+ pump (SERCA2a) and phospholamban (PLB). PLB is a small protein inhibitor of SERCA2a that reduces enzyme affinity for Ca2+ and thereby regulates cardiac contractility. We found that the PLB monomer with Asn27 or Asn30 changed to Cys (N27C-PLB or N30C-PLB) cross-linked to lysine of SERCA2a within seconds with ≥80% efficiency. Optimal cross-linking occurred at spacer chain lengths of 10 and 15 Å for N27C and N30C, respectively. The rapid time course of cross-linking indicated that neither dissociation of PLB pentamers nor binding of PLB monomers to SERCA2a was rate-limiting. Cross-linking occurred only to the E2 (Ca2+-free) conformation of SERCA2a, was strongly favored by nucleotide binding to this state, and was completely inhibited by thapsigargin. Protein sequencing in combination with mutagenesis identified of Lys328 of SERCA2a as the target of cross-linking. A three-dimensional map of interacting residues indicated that the cross-linking distances were entirely compatible with the 10-Å distance recently determined between N30C of PLB and Cys318 of SERCA2a. In contrast, Lys3 of PLB did not cross-link to any Lys (or Cys) of SERCA2a, suggesting that previous three-dimensional models that constrain Lys3 near residues 397–400 of thapsigargin-inhibited SERCA2a should be viewed with caution. Furthermore, although earlier models of PLB·SERCA2a are based on thapsigargin-bound SERCA, our results suggest that the nucleotide-bound, E2 conformation is substantially different and represents the key conformational state for interacting with PLB.

PLB 1 is a small phosphoprotein modulator of the Ca 2ϩ -pump (SERCA2a) in cardiac SR, which is critically involved in regu-lating the strength and duration of the heartbeat (1,2). In the dephosphorylated state, PLB inhibits the Ca 2ϩ -ATPase by decreasing its apparent affinity for Ca 2ϩ (3). Phosphorylation of PLB at Ser 16 and Thr 17 during ␤-adrenergic stimulation of the heart reverses PLB inhibition and augments Ca 2ϩ loading of the SR (4), thus producing positive inotropic and lusitropic effects (2,5). PLB is a single-span membrane protein composed of 52 amino acids, which forms a homopentamer within the SR membrane (1,4). Recent work suggests that there is a dynamic equilibrium between PLB monomers and pentamers (6), and that the PLB monomer is responsible for binding to SERCA2a (7) and inhibiting it (8,9). In intact myocardium, ␤-adrenergic receptor stimulation disrupts the inhibitory interaction between PLB and SERCA2a rapidly; PLB phosphorylation, Ca 2ϩ transport, and contractility all increase within seconds after ␤-receptor activation (4,10). This suggests that PLB monomers must associate and dissociate quickly, over a time course of seconds or faster, to allow for dynamic regulation of SERCA2a and the strength of contractility.
The emerging picture for the mechanism of SERCA2a inhibition by PLB suggests mutually exclusive binding of Ca 2ϩ and PLB to SERCA2a. Chemical cross-linking (7,11) and immunoprecipitation (12) indicate that PLB binds preferentially to the Ca 2ϩ -free conformation of SERCA2a, dubbed E2 (3). The alternative conformation induced by Ca 2ϩ , E1, does not appear to be capable of binding PLB (7), but instead is primed for catalyzing the hydrolysis of ATP and subsequent steps leading to Ca 2ϩ transport across the SR membrane (13). By stabilizing SERCA2a in the E2 conformation, PLB slows the transition from E2 to E1 during the catalytic cycle (3), thus impeding overall Ca 2ϩ transport (Fig. 1). Thus, the apparent reduction of Ca 2ϩ affinity is a kinetic effect of PLB on conformational changes, not necessarily a direct effect on Ca 2ϩ binding sites (3). Recently, the crystal structure of SERCA1a, the skeletal muscle isoform of the Ca 2ϩ pump, was determined, both in the E1 state with bound Ca 2ϩ (14), and in the E2 state, bound with the irreversible inhibitor thapsigargin (15). These structures reveal a relatively complex molecule composed of ϳ1000 amino acids, ten transmembrane helices, and three distinct cytoplasmic domains (16). Although no crystal structure of PLB has yet been reported, numerous biophysical methods have been used to study its structure (17). Because of technical limitations of the various techniques, no consensus has yet been reached for the complete structure of PLB within biological membranes or, more particularly, in association with SERCA2a (1,17). Nev-ertheless, there is great interest in understanding the physical basis for Ca 2ϩ pump regulation (1,2,5). As a result, several models of the three-dimensional interactions between different residues of PLB and SERCA2a have recently been proposed (18 -20), using the E2 conformation of SERCA determined by crystallography (15) or by cryoelectronmicroscopy (21) as a structural template.
Biochemical work has demonstrated that the PLB pentamer is stabilized by a network of interdigitating Leu/Ile residues along one face of the transmembrane helix between residues 37 and 51 of each monomer (17,22). Point mutations made at these Leu/Ile zipper residues destabilized the pentamer (22), and the concomitant enhancement of SERCA2a inhibition led to the idea that the PLB monomer is the species responsible for enzyme inhibition (8,9,24). This has now been verified directly by chemical cross-linking (7). Point mutations along the opposite face of the transmembrane helix attenuated SERCA2a inhibition, and residues here were proposed to constitute an interaction site between the two molecules (9), although certain residues in the zipper domain now also appear to interact directly with SERCA2a (24). Amino acid substitutions within the cytoplasmic region of PLB (residues 1-31) also affected SERCA2a activity (23,25), suggesting that this region of PLB is equally important for direct physical and functional interactions with the Ca 2ϩ pump. In particular, replacement of Asn 27 or Asn 30 of PLB with Ala had no major effect on the equilibrium between PLB pentamers and monomers, but significantly enhanced inhibition of SERCA2a, suggesting that these two PLB residues are directly involved in the interaction with SERCA2a (18,23,24,26). Finally, an interaction between the extreme N terminus of PLB and an exposed loop in the nucleotide domain of SERCA2a was suggested by chimeric constructs (27) and by early cross-linking studies (11).
To characterize further the interactions between the PLB monomer and SERCA2a at the molecular level, we initiated a series of studies using cross-linking agents as molecular rulers to measure distances between key amino acid residues of PLB and SERCA2a (7). Using the homobifunctional, thiol crosslinking agent BMH, we demonstrated a highly specific crosslink between Asn 30 of PLB changed to cysteine (N30C-PLB) and Cys 318 of SERCA2a at the cytoplasmic boundary of M4. The length of the cross-linking spacer chain suggested that these two residues are 10 Å apart. Cross-linking of PLB to SERCA2a occurred only in the absence of Ca 2ϩ and was strikingly potentiated by ATP or ADP, providing strong physical evidence that monomeric PLB binds preferentially to the nucleotide-stabilized, E2 conformation of SERCA2a (7). Significantly, the irreversible SERCA inhibitor thapsigargin completely prevented cross-linking of PLB to SERCA2a (7), supporting previous evidence that the E2-like conformation stabilized by thapsigargin is conformationally distinct from the native E2 conformation (28,29) (Fig. 1).
Here we report on the use of second generation cross-linking agents, which are heterobifunctional and provide additional constraints on the interaction and dynamics between PLB and SERCA2a. The new cross-linking agents contain a maleimide group at one end for coupling to Asn 27 or Asn 30 of PLB replaced with cysteine, and an NHS-ester group at the other end, for coupling to nearby lysine residues of WT-SERCA2a (30). We show that these heterobifunctional cross-linking agents work faster and even more efficiently than the homobifunctional agent, BMH, reported on previously (7). Nearly quantitative coupling between PLB and SERCA2a is achieved with heterobifunctional cross-linking agents. At the same time, the heterobifunctional cross-linkers remain highly specific, and react only with Lys 328 of SERCA2a when attached to residues 27 and 30 of PLB, with optimal linker lengths of 10 and 15 Å, respectively.
Mutagenesis and Baculovirus Production-Point mutations in the cDNA encoding canine PLB were made as previously described (8,22,24). N27C-PLB and N30C-PLB were expressed on Cys-less PLB background, which is canine PLB with Cys residues 36, 41, and 46 replaced by Ala (7). Wild-type canine SERCA2a cDNA was mutated directly in the transfection vector pVL1393 using the QuikChange TM XL-Gold system (Stratagene) (7). All cDNA mutations were confirmed by DNA sequencing. Baculoviruses encoding mutated proteins were generated by co-transfecting into Sf21 insect cells mutated cDNAs in pVL1393 with BaculoGold TM (Pharmingen) linearized baculovirus DNA (7,8,24).
Protein Co-expression in Insect Cells and Isolation of Microsomes-Canine SERCA2a and N27C-PLB or N30C-PLB were co-expressed in Sf21 insect cells as described (7,8,24). Microsomes were harvested 60 h after initiating baculovirus infections and stored frozen in small aliquots at Ϫ40°C at a protein concentration of 6 -10 mg/ml. Protein assay and Ca 2ϩ -ATPase activities of microsomes were determined as previously described (8).
Cross-linking-Cross-linking between single Cys residues of N27C-PLB or N30C-PLB and endogenous Lys residue(s) of WT-SERCA2a was conducted using the heterobifunctional cross-linking agents of increasing lengths listed under "Materials." All heterobifunctional agents tested contained a maleimide group at one end for coupling to Cys residues and an NHS-ester group at the other end for coupling to Lys residues (30).
Cross-linking between PLB and SERCA2a was conducted essentially as recently described (7). Reactions were conducted with 11 g of microsomal protein in 12 l of buffer A, which contained 40 mM MOPS (pH 7.0), 3.2 mM MgCl 2 , 75 mM KCl, 3 mM ATP, and 1 mM EGTA. Reactions were started by adding 0.75 l of cross-linking agent from a 1.6 mM stock solution in dimethyl sulfoxide (0.1 mM final cross-linker concentration) and stopped by adding 7.5 l of SDS-PAGE sampleloading buffer containing 15% SDS plus 100 mM dithiothreitol. Heterobifunctional cross-linking was conducted for 10 min at room temperature unless otherwise indicated. Homobifunctional cross-linking between Cys residues with 0.1 mM BMH was conducted for 1 h at room temperature (7). After terminating the reactions, samples were sub-FIG. 1. Simplified reaction scheme of cardiac Ca 2؉ pump. E1 and E2 represent the high Ca 2ϩ affinity and low Ca 2ϩ affinity conformations of the Ca 2ϩ -ATPase, respectively (3). ATP is hydrolyzed by E1 after binding Ca 2ϩ at the two high affinity sites. ATP also stimulates the conformational transition from E2 to E1 (39), as well as the binding of PLB to E2 (7). Thapsigargin (Tg) binding to E2 forms an irreversibly inhibited, dead-end complex, which prevents the binding of ATP to SERCA (28,29) and also cross-linking to PLB (7). jected to SDS-PAGE and immunoblotting.
To assess Ca 2ϩ effects on cross-linking, ionized Ca 2ϩ was varied by adding CaCl 2 to buffer A (7). In some experiments, ATP in buffer A was omitted or was replaced by other nucleotides, as indicated. To determine antibody effects on cross-linking, affinity-purified anti-PLB monoclonal antibody, 2D12, or anti-SERCA2a monoclonal antibody, 2A7-A1, dialyzed in 20 mM MOPS (pH 7) and 150 mM NaCl, were used (7). 5.5 g of antibody were included with 11 g of microsomal protein in buffer A.
SDS-PAGE and Immunoblotting-SDS-PAGE was performed in 8% polyacrylamide, and immunoblots were probed with anti-PLB monoclonal antibody, 2D12, to detect free and cross-linked forms of PLB (7). In the experiment of Fig. 9, the immunoblot was also probed with anti-SERCA2a monoclonal antibody, 2A7-A1, to detect Ca 2ϩ pumps not cross-linked to PLB (8). Antibody-binding protein bands were visualized with 125 I-protein A, except for the experiment depicted in Fig. 7, in which 125 I-2D12 was used directly for PLB visualization (7), to avoid interference from antibodies carried over from the cross-linking reactions. Antibody-binding bands were quantified with a Bio-Rad Personal Fx phosphorimager.
Identification of Cross-linked Lys Residue of SERCA2a-SERCA2a cross-linked to N27C-PLB with EMCS, or to N30C-PLB with KMUS, was purified to homogeneity from insect cell microsomes in separate runs by sequential anti-SERCA2a (2A7-A1) and anti-PLB (2D12) monoclonal affinity chromatographies, as recently described in detail (7). Each type of purified, cross-linked Ca 2ϩ pump was then digested with endo-Asp-N and endo-Lys-C, and the limit SERCA2a fragment crosslinked to PLB was isolated by a second round of 2D12 chromatography and sequenced, as described (7). By this analysis, the cross-linked amino acid of SERCA2a was restricted to just three potential Lys residues (see Fig. 8). Replacement-mutagenesis of Lys with Ala was then used to determine which of these three Lys residues was the one cross-linked to N27C-PLB and to N30C-PLB.
Modeling-Coordinates for PLB residues 19 -52 (42) (Protein Data Bank accession code 1FJK) were manually docked to E2 coordinates of SERCA1a (15) (Protein Data Bank accession code 1IWO) using crosslinking constraints as a guide. The model was then subjected to a round of energy minimization in XPLOR, with 1IWO held rigid, to eliminate any steric clashes. Phospholamban residues 27 and 30 were mutated to Cys using the DeepView Swiss-PdbViewer program. A search for optimal rotomers for PLB residues, Cys 27 and Cys 30 , and SERCA residue, Lys 328 , was then performed using this program. The model was rendered using Pymol (W. L. DeLano, www.pymol.org).

Specific Cross-linking of N27C-PLB and N30C-PLB to
SERCA2a-We recently showed that N30C-PLB cross-links to Cys 318 of native SERCA2a near M4 using the homobifunctional thiol cross-linking agent BMH (7). In the work described here, we tested for cross-linking N27C-PLB and N30C-PLB to Lys residues of native SERCA2a using a series of heterobifunctional thiol to amine cross-linking agents of different lengths, spanning distances from 4 to 16 Å. N27C-PLB and N30C-PLB were expressed on the Cys-less PLB background, and were fully functional in their abilities to inhibit SERCA2a by lowering Ca 2ϩ affinity (7). Fig. 2A shows that N27C-PLB and N30C-PLB cross-linked strongly to wild-type SERCA2a in insect cell microsomes with use of heterobifunctional cross-linking agents. When immunoblots were probed with the anti-PLB antibody, 2D12, intense cross-linking signals were obtained at molecular masses of just over 100 kDa (PLB/SER), corresponding to heterodimers formed between PLB and SERCA2a. Some weaker homodimerization of PLB monomers was also detected (PLB 2 ). 2 Remarkably, the optimal lengths for heterobifunctional cross-linking were completely different for N27C-PLB and N30C-PLB ( Fig.  2A). Specifically, N27C-PLB cross-linked most efficiently to SERCA2a with the cross-linking agents EMCS, m-maleimido-benzoyl-N-hydroxysuccinimide ester, and succinimidyl 4-(Nmaleimidomethyl)cyclohexane-1-carboxylate, whereas N30C-PLB only cross-linked to SERCA2a with use of the two crosslinkers succinimidyl-6-(␤-maleimidopropionamido)hexanoate and KMUS ( Fig. 2A). The relative intensities of the crosslinking signals from three separate experiments were plotted versus the distances of cross-linker arms and fitted to a Gaussian distribution. Based on the Gaussian fit, N27C of PLB is about 10 Å away from a Lys residue of native SERCA2a (Fig. 2B), whereas N30C cross-links at a distance of 15-16 Å (Fig. 2B).
In early experiments, we tested the possibility that the nat- 2 We should point out that Cys-less PLB is predominately a pentamer in insect cell microsomal membranes, even though it migrates primarily as a monomer under standard SDS-PAGE conditions (7). Cys-less PLB is an unstable pentamer, which totally dissociates at SDS concentrations greater than 1% (7). ural Lys 3 of PLB might cross-link to native Cys or Lys residues of SERCA2a. However, we found no evidence for this. Lys 3 of PLB was not cross-linked to Cys residues of SERCA2a with any of the heterobifunctional agents tested above. Likewise, when a similar series of homobifunctional NHS-ester agents were tested, no cross-linking of Lys 3 of PLB to any Lys residue of SERCA2a was detected (data not shown). Moreover, after mutation of Lys 3 of PLB to Cys (K3C-PLB), we detected no crosslinking of K3C-PLB to SERCA2a with any of the heterobifunctional agents depicted in Fig. 2. We therefore conclude that Lys 3 of PLB is not in proximity to any endogenous Lys or Cys residue of wild-type SERCA2a. These results contradict the earlier conclusion of James et al. (11), who reported crosslinking of Lys 3 of PLB at low efficiency to Lys 397 or Lys 400 of SERCA after detergent solubilization of the purified proteins.
For the remaining experiments, we used EMCS and KMUS for cross-linking SERCA2a to N27C-PLB and N30C-PLB, respectively, because these two compounds gave maximal crosslinking at these two PLB residues (Fig. 2).
Micromolar Ca 2ϩ Effect on Cross-linking-Recent work suggests that micromolar Ca 2ϩ disrupts the physical interaction between PLB and SERCA2a (7,12), by occupying the two high affinity Ca 2ϩ binding sites and driving the pump into the E1 conformation (Fig. 1). Consistent with this hypothesis, crosslinking of N27C-PLB to SERCA2a with EMCS and of N30C-PLB to SERCA2a with KMUS was abolished by Ca 2ϩ at micromolar concentration (Fig. 3A). The K i value for Ca 2ϩ inhibition of cross-linking to SERCA2a was 0.4 M for both N27C-PLB (black) and N30C-PLB (red) using heterobifunctional agents (Fig. 3B). Interestingly, the K i value for Ca 2ϩ inhibition of cross-linking of N30C-PLB to Cys 318 of SERCA2a by BMH (blue) was approximately 3 times lower (0.15 M) (inset), which confirms earlier results with BMH (7). This suggests that different cross-linking sites between PLB and SERCA2a respond differentially to conformational changes in SERCA2a associated with Ca 2ϩ binding. Nevertheless, crosslinking results obtained with all three agents, EMCS, KMUS, and BMH, agree with the consensus that PLB binds only to SERCA2a in the Ca 2ϩ -free or E2 conformation. Fig. 3A also illustrates that heterobifunctional cross-linking agents (sulfhydryl to amine) produce more efficient cross-linking of PLB to SERCA2a than the homobifunctional cross-linking agent, BMH (sulfhydryl to sulfhydryl). In multiple experiments, EMCS and KMUS cross-linked ϳ3 times more Ca 2ϩ pump molecules to N27C-PLB and N30C-PLB, respectively, than did BMH to N30C-PLB. Essentially, virtually all of the Ca 2ϩ pump molecules were cross-linked to PLB with use of EMCS and KMUS (see below).
Time Course of Cross-linking Using Heterobifunctional Agents-Previously, we demonstrated that cross-linking of N30C-PLB to Cys 318 of SERCA2a is relatively slow, with a t1 ⁄2 of 15 min at room temperature (7). In contrast, cross-linking with heterobifunctional agents is much faster, occurring over a time course of seconds (Fig. 4). Specifically, when measured at 37°C in Ca 2ϩ -free buffer, the t1 ⁄2 for cross-linking N27C-PLB to SERCA2a with EMCS was 42 s, and the t1 ⁄2 for cross-linking N30C-PLB to SERCA2a with KMUS was 13 s. In the results Reactions were terminated by adding sample-loading buffer at the times indicated. The insets show the autoradiographs of anti-PLB immunoblots obtained after cross-linking PLB to SERCA2a (PLB/SER), and the graph shows the results plotted. Data points were fitted with a monoexponential function using Origin. Open symbols designate control-sample results, obtained by adding cross-linkers, but no EGTA, at the start of the reactions. depicted in Fig. 4, microsomes had been preincubated in Ca 2ϩcontaining buffer to place SERCA2a in the E1 conformation prior to initiation of cross-linking reactions. Cross-linking was then initiated by simultaneously shifting SERCA2a to E2 with EGTA (arrow) at the time of addition of cross-linking agents. Identical t1 ⁄2 values were obtained by preincubating membranes in the absence of Ca 2ϩ (data not shown), however, indicating that heterobifunctional cross-linking to monitor PLB/SERCA physical interactions is not fast enough to resolve either the conformational transitions of SERCA2a from E1 to E2 (Fig. 1), or the physical association of PLB with the E2 conformation. Although the former has been shown to occur within milliseconds (31), the latter has not been previously characterized.
To further investigate the dynamics of PLB binding to SERCA2a, we measured cross-linking of the two proteins at lower temperatures. When reactions were carried out at room temperature, t1 ⁄2 values for cross-linking N27C-PLB and N30C-PLB to SERCA2a were 110 and 36 s, respectively. Cross-linking reactions were further slowed down when conducted at 4°C (Table I). These temperature-dependent effects on cross-linking suggest that PLB binding to SERCA2a may be membrane diffusion-limited at lower temperatures (32,33).
Thapsigargin Effect on Heterobifunctional Cross-linking-To assess the effects of SERCA2a conformation on heterobifunctional cross-linking in more detail, we checked for inhibition by thapsigargin. Previously, we demonstrated that thapsigargin completely prevents cross-linking of N30C-PLB to Cys 318 of SERCA2a by BMH (7). Similarly, 2 M thapsigargin abolished cross-linking of N27C-PLB and N30C-PLB to SERCA2a with EMCS and KMUS, respectively (Fig. 5A); K i values for thapsigargin were 0.19 M with N27C-PLB and 0.32 M with N30C-PLB (Fig. 5B).
To confirm the specificity of this thapsigargin effect, we co-expressed N27C-PLB and N30C-PLB with F256V-SERCA2a, a Ca 2ϩ pump mutant that is resistant to thapsigargin (34). As expected, cross-linking of PLB to F256V-SERCA2a by heterobifunctional agents was resistant to thapsigargin (Fig. 5A), with K i values of 8.6 and 23 M obtained for N27C-PLB and N30C-PLB, respectively (Fig. 5B). The overall reduction in thapsigargin sensitivity was greater than 45-fold, consistent with earlier results (34). Fig. 5A also indicates that the cross-linking efficiency of F256V-SERCA2a is equivalent to WT-SERCA2a, suggesting that the F256V mutation only affects binding of thapsigargin, but not of PLB, to SERCA2a. In the x-ray structure, Phe 256 contributes directly to the thapsigargin binding site (15); thus, the unimpaired cross-linking by the F256V mutant suggests that PLB binds at a different site, and that thapsigargin effects on PLB interactions are therefore allosteric.
Nucleotide Effects-In addition to low Ca 2ϩ levels, crosslinking of N30C-PLB to Cys 318 of SERCA2a with BMH required the presence of the nucleotide ATP or ADP (7) (Fig. 6A,  lower panel). Qualitatively identical results were obtained when EMCS and KMUS were used to couple N27C-PLB and N30C-PLB to SERCA2a, respectively (Fig. 6A). For both PLB mutants, cross-linking was increased 3-4-fold when ATP or ADP was present, whereas AMP was much less effective (Fig.  6B). These results provide additional evidence that PLB interacts preferentially with the Ca 2ϩ pump when it is in the E2, nucleotide-bound conformation, but not in the E2, thapsigargin-inhibited conformation (7).
Anti-PLB Antibody Effect on Cross-linking-Anti-PLB monoclonal antibody, 2D12, which recognizes residues 7-13 of PLB, reverses PLB inhibition of SERCA2a at low ionized Ca 2ϩ concentration (3,8,24), and at the same time prevents the crosslinking of N30C-PLB to Cys 318 of SERCA2a with BMH (7) (Fig.  7, lower panel). This suggests that the antibody blocks the physical interaction between the two proteins. This conclusion is further supported by cross-linking results with heterobifunctional agents (Fig. 7). Cross-linking of N27C-PLB to SERCA2a with EMCS and of N30C-PLB to SERCA2a with KMUS was virtually eliminated by the anti-PLB antibody. However, the anti-SERCA2a antibody, 2A7-A1, which recognizes residues 386 -396 of SERCA2a (7), had no effect on cross-linking of PLB to SERCA2a with any of the cross-linking agents tested (Fig.  7), nor did it affect the functional activity of SERCA2a expressed by itself or co-expressed with PLB (data not shown). The region of SERCA2a recognized by the 2A7-A1 antibody (residues 386 -396) is directly adjacent to the region previously proposed to be crucial for the physical interaction between Lys 3 of PLB and SERCA2a (residues 397-402) (11,27).  SERCA2a Cross-linking Site Determination-To determine the Lys residue(s) of SERCA2a cross-linked to N27C-PLB and N30C-PLB by EMCS and KMUS, respectively, we followed the strategy developed in our previous work (7). SERCA2a was cross-linked to each PLB mutant on a large scale, then individually purified from insect cell microsomes by sequential 2A7-A1 and 2D12 monoclonal antibody affinity chromatographies, to yield the exclusively cross-linked complexes (7). During the purifications, we determined that Ն80% of the SERCA2a molecules in insect cell microsomes were crosslinked to N27C-PLB and N30C-PLB by EMCS and KMUS, respectively. Comparison of this exceptionally high cross-linking efficiency to the 20 -40% efficiency previously reported for BMH by the same method (7) corroborates results shown in Figs. 3, 6, and 7. After purification, the cross-linked complexes were subjected to endo-Asp-N and endo-Lys-C digestions and the limit dipeptides were isolated and sequenced. The same 16-kDa limit dipeptide was obtained after EMCS and KMUS cross-linking, as was previously obtained after BMH crosslinking (7) (Fig. 8). Specifically, the dipeptides began at Val 4 of the PLB sequence and at Asp 254 of the SERCA2a sequence. Neither peptide extended beyond Lys 352 (Fig. 8). Lys 262 at the cytoplasmic face of M3 was excluded as the amino acid crosslinked to N27C-PLB or N30C-PLB because readable SERCA2a sequence ran past this residue (see Fig. 8 legend). Lys 297 near the lumenal border of M4 of SERCA2a could be excluded, because it is located on the opposite side of the membrane from residues 27 and 30 of PLB, which project into the cytoplasm. This left Lys residues 328, 329, and 352 of SERCA2a as the candidate amino acids cross-linked to N27C-PLB and N30C-PLB.
To identify the cross-linked lysine, we individually mutated Lys 328 , Lys 329 , and Lys 352 of WT-SERCA2a to Ala, and tested for loss of cross-linking function. K328A-SERCA2a and K329A-SERCA2a retained Ca-ATPase activity, whereas K352A-SERCA2a was catalytically inactive, as predicted from earlier mutagenesis work with SERCA1a (35). Fig. 9A shows that K329A-SERCA2a and K352A-SERCA2a retained the ability to cross-link efficiently to N27C-PLB and N30C-PLB, whereas K328A-SERCA2a lost all cross-linking ability. The inability of K328A-SERCA2a to cross-link to PLB was also reflected by its absence of an upward mobility shift, when the identical crosslinked samples were probed with the anti-SERCA2a antibody, 2A7-A1 (Fig. 9B). These results thus identify Lys 328 of SERCA2a as the sole lysine residue (of 56 residues) crosslinked to N27C-PLB and N30C-PLB with EMCS and KMUS, respectively.

DISCUSSION
Recently, using the homobifunctional thiol-cross-linking agent BMH as a molecular ruler, we reported the 10-Å distance between residue 30 of PLB and Cys 318 of SERCA2a, which is located near the cytoplasmic border of M4 (7). The current data obtained with heterobifunctional agents complements our previous results by showing cross-linking of N27C and N30C of  red C over a black N, and Lys 328 of SERCA2a cross-linked to each PLB peptide (verified in Fig. 9) is highlighted in red. Structures of EMCS and KMUS are shown between the cross-linked peptides. Transmembrane regions are highlighted in yellow. The epitope recognized by 2D12 is highlighted in blue. Residues 1-3 of PLB cleaved off by proteases, including Lys 3 which was not cross-linked, are indicated.
PLB to Lys 328 of SERCA2a, at distances of 10 and 15 Å, respectively (Fig. 10). Remarkably, the cross-linking reported here with the heterobifunctional couplers EMCS and KMUS is stronger and faster than that previously achieved with the homobifunctional agent, BMH (7). Whereas BMH cross-linked ϳ20 -40% of SERCA2a molecules to N30C-PLB in insect cell microsomes with t1 ⁄2 of 15 min, the heterobifunctional agents cross-linked 80% or greater of SERCA2a molecules in insect cell microsomes to PLB with a t1 ⁄2 of 13-42 s. Like cross-linking at Cys 318 of SERCA2a (7), PLB cross-linking at Lys 328 of SERCA was inhibited by micromolar Ca 2ϩ , by the anti-PLB monoclonal antibody 2D12, and by the Ca 2ϩ pump inhibitor, thapsigargin. Furthermore, cross-linking at both Lys 328 and Cys 318 was strongly potentiated by the nucleotides ATP or ADP, but not by AMP (7). Taken together, these new crosslinking data support our previous conclusion that PLB interacts only with SERCA2a in the Ca 2ϩ -free, E2 state, with a strong preference for the nucleotide-bound, E2 state.
Dynamics of PLB⅐SERCA2a Interaction-Cross-linking agents used as molecular rulers have proven to be reliable reagents for deciphering not only accurate distances between key residues of interacting protein molecules, but also for monitoring the dynamic changes between protein molecules that affect the protein-protein interactions (7,36,37). In intact myocardium, modulation of contractility (and Ca 2ϩ -ATPase activity) by the ␤-adrenergic signaling pathway is a relatively fast process, occurring over a time frame of seconds (10). This The NMR structure for PLB (42) (accession code 1FJK) was docked next to the x-ray structure of SERCA1a in the thapsigargin-inhibited E2 state (15) (accession code1IWO). A, PLB residues 19 -25 (green) are juxtaposed with the M4/S4 helix of SERCA1a (cyan) with side chains shown for N27C and N30C of PLB and Cys 318 and Lys 328 of SERCA1a. A large portion of the SERCA phosphorylation domain is also visible (gray), and part of bound thapsigargin is shown as a space-filling model on the lower right. In addition, the salt bridge between SERCA residues Lys 329 and Glu 748 is shown (numbering corresponds to SERCA2a). Distances are indicated between sulfur atoms of Cys residues and the terminal nitrogen of Lys 328 . B, orthogonal view of the structures looking toward the membrane from the cytoplasmic domain. Only the transmembrane region of SERCA (helices M1-M9) are rendered for purposes of clarity. PLB and thapsigargin have distinct binding sites on opposite sides of SERCA. We note that, although the distances shown on this model are consistent with those documented by the highly specific and efficient cross-linking reported here and in Ref. 7, the conformation of SERCA in the thapsigargin-inhibited state shown in this figure fails to cross-link with PLB. Instead, the preference of PLB for the nucleotide-bound, E2 state suggests that nucleotide binding produces a substantial conformational change, which is as yet uncharacterized (see "Discussion").
sets an upper limit on the time course of association and dissociation of PLB monomers and the Ca 2ϩ pump, which are major targets in this signaling cascade (1,2,4). In our previous work with the homobifunctional cross-linker BMH, we observed rapid kinetics for cross-linking of PLB monomers to each other, which occurred within seconds or faster and was independent of Ca 2ϩ concentration (7). This was consistent with homodimer formation within preformed pentamers. On the other hand, cross-linking of PLB monomers to SERCA2a in the same study was much slower, occurring with a t1 ⁄2 of 15 min at room temperature (7). This time scale for coupling between PLB and SERCA2a was too slow to account for the dynamic regulation of contraction in intact myocardium (10). The current cross-linking results with heterobifunctional agents have resolved this discrepancy and confirm that the two molecules associate over a time course of seconds or less, the current limitation being the time resolution of the experiments. Crosslinking of residue 30 of PLB to Lys 328 of SERCA2a, for example, occurred with a t1 ⁄2 of 13 s at 37°C. Most importantly, we obtained the same short t1 ⁄2 for cross-linking of N30C-PLB to SERCA2a, whether membranes were preincubated in Ca 2ϩcontaining buffer, in which PLB is presumed to be dissociated from SERCA2a prior to cross-linking, or in Ca 2ϩ -free buffer, in which PLB is already bound to SERCA2a at the start of the cross-linking reaction (7,12). Identical t1 ⁄2 values imply that neither dissociation of PLB monomers from pentamers nor diffusion of monomers in the membrane and binding to SERCA2a are rate-limiting. Instead, the measured rate must be intrinsic to the cross-linking mechanism, perhaps reflecting the time required for proper alignment of Lys 328 on the pump with residue 30 of PLB. Whereas Lys 328 is well exposed at the top of the M4/S4 helix, Cys 318 is buried in the interface between M4, M5, and M6 in the structure of the thapsigargin-inhibited, E2 state (15) (Fig. 10). Thus, although the ability of BMH to cross-link Cys 318 (7) indicates a dynamism in the structure of M4, the slower time scale most likely reflects constraints in exposing this particular Cys residue.
We observed that the rates of cross-linking N27C and N30C of PLB to SERCA2a were strongly temperature dependent, decreasing by 20 -40-fold when measured at 4°C (Table I). One explanation for rate reduction is that diffusion of PLB monomers within the plane of the membrane becomes limiting (32,33). This is currently under investigation. On the other hand, local domain motions within SERCA2a may be drastically slowed by temperature, as is well known from kinetic studies (3). It is clear that the cross-linking reaction itself is not ratelimiting under any of the conditions or with any cross-linkers tested. For example, all reagents tested to date cross-linked PLB monomers within pentamers (7) within a few seconds or less, even at 4°C (data not shown). This implies that the PLB pentamer is an abundant, stable structure in insect cell microsomes, with relevant residues favorably aligned to allow very fast cross-linking. In contrast, the PLB⅐SERCA2a heterodimer appears to be transient, with local motions of SERCA2a structural elements required for cross-linking, and presumably also for SERCA2a inhibition by PLB.
Ca 2ϩ Dependence of Cross-linking-Consistent with previous results (7), heterobifunctional cross-linking of PLB to Lys 328 of SERCA2a was completely inhibited by micromolar Ca 2ϩ , which provides additional strong evidence that binding of Ca 2ϩ to SERCA2a causes PLB to dissociate from the Ca 2ϩ pump. Similar conclusions have been made by others (11,12). X-ray crystal structures show that M4 and M2 undergo large movements during the Ca 2ϩ -induced conformational change from E2 to E1 and the movement of M2 has been specifically postulated to displace PLB from its binding site (15,20). The importance of M4 in the PLB⅐SERCA2a binding interaction was first suggested by successful cross-linking of PLB to Cys 318 of SERCA2a (7), which has now been reaffirmed by crosslinking of Lys 328 . Both Cys 318 and Lys 328 are on the cytoplasmic extension of M4 (sometimes referred to as S4) (Fig. 10), movements of which are likely controlled by Ca 2ϩ binding to Glu 309 . Interestingly, we found different K Ca values for inhibition of PLB cross-linking to Cys 318 versus Lys 328 of SERCA2a, indicating that there are Ca 2ϩ concentrations at which Lys 328 can still be cross-linked efficiently but Cys 318 cannot. This suggests that there are movements of M4/S4 during Ca 2ϩ binding that differentially affect interactions with residues 27 and 30 of PLB. These movements must precede the more global conformational change to E1, which would involve M2 displacements and complete disruption of the PLB binding site.
Conformational Dependence of Cross-linking-Our use of heterobifunctional cross-linking agents confirms the binding preference of PLB for the E2 conformation of SERCA2a stabilized by nucleotides (7,28,38). The efficiency of cross-linking of N27C and N30C of PLB to Lys 328 of SERCA2a in Ca 2ϩ -free medium was increased 3-4-fold by inclusion of nucleotides; no significant cross-linking at all of N30C-PLB to Cys 318 of SERCA2a was detected unless nucleotides were included (7). This suggests a substantial conformational effect of ATP or ADP on the E2 state (28,38), which is consistent with the well characterized effects of nucleotides on the E2 to E1 transition (39). In contrast to the effects of nucleotide, the specific inhibitor thapsigargin was found to completely prevent cross-linking of N27C and N30C to either Lys 328 or Cys 318 (7) of SERCA2a. This is unlikely to be a direct effect of this irreversible inhibitor at the PLB binding site, given the large distance between the thapsigargin-binding site and the S4 helix containing the crosslinked PLB residues (Fig. 10B), but instead appears to be related to the inhibitory effect of thapsigargin on ATP binding to E2 (28,29), preventing formation of the state that preferentially binds PLB (7). Given this and other evidence for global effects of both nucleotides and thapsigargin on SERCA conformation (21,29), it may be ill advised to use the x-ray structure of nucleotide-free, thapsigargin-inhibited SERCA as a basis for detailed modeling the interaction between PLB and SERCA2a (20).
Spatial Considerations-The lengths of spacer chains for heterobifunctional cross-linking agents used here suggest atomic distances of 10 and 15 Å between N27C and N30C of PLB and Lys 328 of SERCA2a, respectively (Fig. 10A). Fig. 2B indicates that the radius of cross-linking of N27C of PLB to Lys 328 of SERCA2a ranges from 4 to 15 Å, albeit with lower efficiency if more or less than 10 Å. Surprisingly, Lys 329 was not cross-linked in the presence or absence of Ca 2ϩ , despite being well within range of most cross-linking agents used. This may be the result of a salt bridge formed between Lys 329 and Glu 748 , which could serve to hold Lys 329 away from the crosslinking agents (Fig. 10A). Nevertheless, Lys 329 and Cys 318 are on the same side of the S4 helix in the thapsigargin inhibited state (15), and the dynamics that allow cross-linking of N30C on PLB to Cys 318 on SERCA2a might also be expected to expose Lys 329 . The failure to do so may be another indication that the structure of E2 in the thapsigargin-inhibited state (15) may not be appropriate for assessing its interaction with PLB (20). It is intriguing that the optimal cross-linking distances for residues 27 and 30 of PLB differed by 5 Å, because this is the difference one would expect if they existed on a continuous helix (Fig.  10A). Furthermore, their separation by three residues would place them on the same side of this helix, which would be consistent with their interaction with the same Lys of SERCA2a. This idea contradicts recent models of PLB inter-acting with SERCA2a (18 -20), which assumed an unstructured region between PLB residues 21 and 30, the so-called domain IB (40). Although NMR structures suggest that the transmembrane helix continues without interruption into a helical domain IB (41,42), the collision of this structure with S4 was recently cited as a rationale for unwinding domain IB in a model for the inhibitory interaction with SERCA (20). This rationale should not be considered absolute, given the use of the thapsigargin-inhibited, nucleotide-free structure of SERCA (15) for this modeling exercise. In addition, this unwinding was necessary (20) to allow the N terminus of PLB to reach the loop on SERCA2a containing Lys 397 and Lys 400 , residues which have been previously implicated in interacting with PLB (11,27), but were not successfully cross-linked to PLB here.
In particular, an earlier study by James et al. (11) reported that Lys 3 of PLB cross-links to Lys 397 or Lys 400 of SERCA at a distance of 15 Å using purified, detergent-solubilized proteins and the Denny-Jaffe reagent (11). In the current study, we found no evidence for cross-linking of Lys 3 of PLB to any SERCA2a residue using a variety of different cross-linking reagents. We did observe that Lys 3 of PLB was able to crosslink to N27C or N30C of another PLB monomer causing homodimer formation (Fig. 2), confirming the reactivity of Lys 3 with the cross-linking reagents and suggesting a large freedom of motion in the N terminus of PLB. Furthermore, our experimental system maintains well characterized functional coupling between PLB and SERCA2a in non-perturbed microsomal membranes (7,8,24,26), and allows at least 10-fold higher cross-linking efficiency at several alternative sites from that suggested by James et al. (11). We therefore believe that the conclusions of James et al. (11) on cross-linking of Lys 3 of PLB to Lys 397 or Lys 400 of SERCA2a should be viewed with caution. Moreover, it does not seem justified at this time to constrain three-dimensional models of the PLB⅐SERCA2a inhibitory complex (18 -20, 40) with the physical proximity of these particular residues. Although residues 397-402 of SERCA2a are apparently required for functional coupling to PLB (27), this by no means demonstrates that these SERCA residues interact physically with Lys 3 of PLB, as has been often suggested (14,20,40). Likewise, the data provided here and previously (7) provide no evidence for the proposed interaction of residues 27 and 30 of PLB with the loop between M6 and M7 of SERCA (40), an idea that was recently withdrawn (20), and point out the questionable utility of immunoprecipitation for identifying direct residue interactions between PLB and SERCA2a (40) as well as in characterizing thapsigargin effects (12).
Finally, it is interesting to note that our anti-SERCA2a monoclonal antibody, 2A7-A1, serendipitously bound to residues 386 -396 of SERCA2a, amino acids that are directly contiguous to the 397-400 loop previously proposed to bind to Lys 3 of PLB (11,27). Nevertheless, the 2A7-A1 antibody had no effect on Ca 2ϩ -ATPase activity, on functional coupling to PLB, or on cross-linking of PLB to SERCA2a at different sites. In contrast, the anti-PLB monoclonal antibody, 2D12, which recognizes residues 9 -13 of PLB and disrupts the functional interaction between PLB and SERCA2a by mimicking PLB phosphorylation (8), prevented the cross-linking of residues 27 and 30 of PLB to residues 318 (7) and 328 of SERCA2a. Conformational changes in the vicinity of this epitope are likely to result from phosphorylation of Ser 16 during ␤-adrenergic stimulation (43), and it will therefore be important in future experiments to establish the sites of interaction between this important region of PLB and the Ca 2ϩ pump to understand the mechanism of regulation in a physiological context.