Role of Leucine 31 of Phospholamban in Structural and Functional Interactions with the Ca2+ Pump of Cardiac Sarcoplasmic Reticulum*

The ability of two loss-of-function mutants, L31A and L31C, of phospholamban (PLB) to bind to and inhibit the Ca2+ pump of cardiac sarcoplasmic reticulum (SERCA2a) was investigated using a molecular cross-linking approach. Leu31 of PLB, located at the cytoplasmic membrane boundary, is a critical amino acid shown previously to be essential for Ca2+-ATPase inhibition. We observed that L31A or L31C mutations of PLB prevented the inhibition of Ca2+-ATPase activity and disabled the cross-linking of N27C and N30C of PLB to Lys328 and Cys318 of SERCA2a. Although L31C-PLB failed to cross-link to any Cys or Lys residue of wild-type SERCA2a, L31C did cross-link with high efficiency to T317C of SERCA2a with use of the homobifunctional sulfhydryl cross-linking reagent, 1,6-bismaleimidohexane. This places Leu31 of PLB within 10 Å of Thr317 of SERCA2a in the M4 helix. Thus, contrary to previous suggestions, PLB with loss-of-function mutations at Leu31 retains the ability to bind to SERCA2a, despite losing inhibitory activity. Cross-linking of L31C-PLB to T317C-SERCA2a occurred only in the absence of Ca2+ and in the presence of nucleotide and was prevented by thapsigargin and by anti-PLB antibody, demonstrating for a fourth cross-linking pair that PLB interacts near M4 only when the Ca2+ pump is in the Ca2+-free, nucleotide-bound E2 conformation, but not in the E2 state inhibited by thapsigargin. L31I-PLB retained full functional and cross-linking activity, suggesting that a bulky hydrophobic residue at position 31 of PLB is essential for productive interaction with SERCA2a. A model for the three-dimensional structure of the interaction site is proposed.

Leu 31 is another critically important residue in domain IB of PLB that is required for PLB function. Changing Leu 31 of PLB to Ala (L31A-PLB) abolishes its inhibitory capacity (14). Kimura et al. (14) and Asahi et al. (15) postulate that the loss-of-function exhibited by a class of mutants including L31A-PLB is due to their inability to bind to SERCA1a in the membrane-spanning region. In the case of Leu 31 , however, other evidence is consistent with localizing this residue at the cytoplasmic membrane interface within domain IB (6,16,17). A subsequent structural model by Toyoshima et al. (18) suggests that Leu 31 of PLB interacts directly with Thr 805 and Phe 809 in M6 of SERCA1a. However, this model is inconsistent with the fact that L31C of PLB failed to cross-link either to L802C or to F809C of SERCA1a in M6 (15,18,19).
Here we report on our cross-linking approach to assess the physical and functional interactions between SERCA2a and Leu 31 of PLB in domain IB. Especially, we tested the hypothesis that loss-of-function by PLB with small side chain substitutions at Leu 31 is due to its inability to bind to SERCA2a (14,15). In fact, our data contradict this hypothesis by showing efficient cross-linking of the loss-of-function mutant, L31C-PLB, to T317C of SERCA2a. Moreover, this cross-linking also requires the nucleotide-bound E2 conformation of the cardiac Ca 2ϩ pump, consistent with our previous results on crosslinking of N27C-PLB and N30C-PLB to WT-SERCA2a (11,12). Like the other two domain IB residues, Asn 27 and Asn 30 , Leu 31 of PLB appears to interact with M4 on SERCA2a, not M6 as recently proposed by Toyoshima et al. (18). Possible mechanisms for the important role of Leu 31 in domain IB of PLB are discussed.

EXPERIMENTAL PROCEDURES
Materials-The homobifunctional sulfhydryl cross-linking agent, BMH, and the heterobifunctional amine-sulfhydryl cross-linking agents, EMCS and KMUS, were obtained from Pierce. Thapsigargin was purchased from Sigma. Iodogen was from Pierce.
Protein Co-expression in Insect Cells and Isolation of Microsomes-Co-expression of PLB mutants with WT-SERCA2a or T317C-SERCA2a (canine isoforms) was conducted in Sf21 insect cells as described previously (11)(12)(13)20). Microsomes (48,000 ϫ g) were isolated 60 h after initiating baculovirus infections and stored frozen in small aliquots at Ϫ40°C. Protein concentration was determined by the Lowry method (13).
Cross-linking-Cross-linking between Cys-substituted residues of PLB and residues of SERCA2a in insect cell microsomes was conducted at room temperature as described recently, using a final concentration of cross-linking reagent of 0.1 mM (11,12). Reactions were started by adding 0.75 l of concentrated cross-linking reagent in dimethyl sulfoxide to 11 g of microsomal protein in 12 l of buffer A, which consisted of 40 mM MOPS (pH 7.0), 3.2 mM MgCl 2 , 75 mM KCl, 3 mM ATP, and 1 mM EGTA. Reactions were terminated by adding 7.5 l of SDS-PAGE sample loading buffer containing 15% SDS and 100 mM dithiothreitol. The samples were then subjected to SDS-PAGE and immunoblotting. When the heterobifunctional (Cys to Lys) cross-linking reagents, EMCS and KMUS, were used, reactions were generally conducted for 10 min at room temperature. With use of the homobifunctional (Cys to Cys) cross-linker, BMH, reactions were conducted for 30 -60 min at room temperature (except in Fig. 6).
To assess Ca 2ϩ effects on cross-linking, ionized Ca 2ϩ was varied by adding CaCl 2 to buffer A (11,12). In some experiments, ATP in buffer A was omitted or was replaced by other nucleotides as indicated. To determine antibody effects on cross-linking, 5.5 g of affinity-purified anti-PLB monoclonal antibody, 2D12, or anti-SERCA2a monoclonal antibody, 2A7-A1, both predialyzed in 20 mM MOPS (pH 7.0) and 150 mM NaCl, were included with 11 g of microsomal protein in buffer A (11,12).
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 (11,12). Antibody-binding protein bands were visualized with 125 I-protein A, except for the experiment depicted in Fig. 5D, in which 125 I-2D12 was used directly for PLB visualization (11,12) to avoid interference from antibodies carried over from the cross-linking reactions. Antibody binding bands were quantified with Bio-Rad Personal Fx phosphorimaging.

L31A Mutation Prevents PLB Cross-linking to WT-
SERCA2a-Using thiol-and amine-selective cross-linking agents, we previously demonstrated that residues 27 and 30 in domain IB of PLB cross-link with high specificity to Lys 328 and Cys 318 of SERCA2a at the cytoplasmic extension of M4 (11,12). N27C-PLB cross-linked to Lys 328 of WT-SERCA2a most efficiently with the heterobifunctional cross-linking agent, EMCS, which has a length of 10 Å (Fig. 1A, left panel) (12), whereas N30C-PLB cross-linked most efficiently to Lys 328 of WT-SERCA2a with the heterobifunctional cross-linking agent, KMUS, with a length of 15 Å (Fig. 1C, left panel) (12). In addition, N30C-PLB cross-linked to Cys 318 of WT-SERCA2a with the homobifunctional cross-linking agent, BMH, with a length of 10 Å (Fig. 1C, left panel) (11). Like WT-PLB (13), N27C-PLB and N30C-PLB both inhibit the Ca 2ϩ -ATPase activity of WT-SERCA2a by shifting the Ca 2ϩ activation curve to the right, giving K Ca values of 0.22 and 0.27 M, respectively ( Fig. 1, B and D, open squares). Addition of the anti-PLB monoclonal antibody, 2D12, which mimics the effect of PLB phosphorylation (4), reverses this inhibition, giving K Ca values of 0.12 and 0.15 M, respectively (Fig. 1, B and D, closed squares).
Directly adjacent to Asn 30 of PLB, Leu 31 is at the cytoplasmic boundary of the membrane (6) and appears to be essential for PLB function (14,15). To clarify the role of Leu 31 of PLB, we first confirmed the original observation of Kimura et al. (14) that replacement with the smaller residue alanine (i.e. L31A) abolishes PLB inhibition of Ca 2ϩ -ATPase activity and Ca 2ϩ transport (data not shown). To explain this result, Kimura et al. hypothesized that L31A-PLB exhibits a decreased binding affinity for SERCA2a (14), an idea that was supported in a subsequent co-immunoprecipitation study (15). To further test this hypothesis, we co-expressed two double mutants of PLB, namely, N27C,L31A-PLB and N30C,L31A-PLB, with WT-SERCA2a and checked for loss of cross-linking function. As predicted, when the L31A mutation was introduced into PLB, neither N27C nor N30C of PLB retained the ability to crosslink to Lys 328 of WT-SERCA2a, using EMCS and KMUS, respectively ( Fig. 1, A and C). Likewise, the L31A mutation prevented the cross-linking of N30C of PLB to Cys 318 of WT-SERCA2a by BMH (Fig. 1C). Furthermore, neither of the two PLB double mutants, N27C,L31A-PLB nor N30C,L31A-PLB, was able to inhibit the Ca 2ϩ -ATPase activity of WT-SERCA2a (Fig. 1, B and D, open and closed circles). These results appear to confirm the idea that the L31A mutation impairs the binding of PLB to WT-SERCA2a and thus the ability of PLB to inhibit Ca 2ϩ pump activity.
Failure of L31C-PLB to Cross-link to WT-SERCA2a-To investigate further whether Leu 31 of PLB is essential for physical association with WT-SERCA2a and thereby inhibition of its ATPase activity, we made the L31C-PLB mutant and tested for  Fig. 2A shows that L31C-PLB did not cross-link to any endogenous Cys or Lys residue of WT-SERCA2a using the cross-linking agents BMH, EMCS, and KMUS, whereas formation of PLB homodimers (11,12) was readily apparent, especially with use of the homobifunctional thiol cross-linking agent, BMH ( Fig. 2A, PLB 2 ). Further testing of a series of homobifunctional Cys to Cys crosslinking agents and heterobifunctional Cys to Lys cross-linking agents spanning distances between 5 and 15 Å also failed to reveal any cross-linking between L31C-PLB and WT-SERCA2a (data not shown). Cross-linking of N27C and N30C of PLB to WT-SERCA2a was also completely prevented in the corresponding double mutants, N27C,L31C-PLB and N30C,L31C-PLB (Fig. 2, B and C). Consistent with these results, L31C-PLB was incapable of inhibiting WT-SERCA2a activity, and the L31C mutation, like L31A, also abolished inhibition of SERCA2a by the corresponding double mutants N27C,L31C-PLB and N30C,L31C-PLB (Fig. 2D).
Efficient Cross-linking of L31C-PLB to T317C-SERCA2a-The failure to obtain cross-linking between Leu 31 mutants of PLB and WT-SERCA2a appears to support the hypothesis that L31C (or A)-PLB is incapable of binding to SERCA2a. However, the possibility remained that these PLB mutants could still bind to the Ca 2ϩ pump but that residue 31 is sterically hindered from efficiently cross-linking to endogenous Cys or Lys residues of WT-SERCA2a. Indeed, of the cross-linked residues of SERCA2a identified in this region (11,12), Cys 318 appears to be buried on the wrong side of the M4 helix in x-ray crystallographic structures of SERCA (21,22), and Lys 328 is over 15 Å away from PLB residue 31 at the upper end of M4 (see Fig. 8). We reasoned that Leu 31 of PLB might be considerably closer to Thr 317 of SERCA2a, which is located one residue below Cys 318 and on the outside face of the M4 helix. To further test for a potential interaction between Leu 31 on PLB and M4 on SERCA2a, we made the T317C mutant and checked for crosslinking of T317C-SERCA2a to L31C-PLB. Indeed, strong crosslinking of L31C-PLB to T317C-SERCA2a was observed using the 10 Å-long, homobifunctional thiol-reactive cross-linking agent, BMH (Fig. 3A). At the same time, L31C-PLB failed to cross-link to Lys 328 of T317C-SERCA2a using the heterobifunctional cross-linking agents, EMCS and KMUS (Fig. 3A), confirming results obtained with WT-SERCA2a (Fig. 2) and supporting the notion of spatial constraints as an explanation for the lack of cross-linking between Leu 31 of PLB and Lys 328 (or Cys 318 ) of SERCA2a. Despite the implication of strong binding between L31C-PLB and T317C-SERCA2a as indicated by the immunoblotting, and blots probed with anti-PLB monoclonal antibody, 2D12. Control samples (CON) had no cross-linker added. PLB/SER, PLB cross-linked to WT-SERCA2a; PLB 1 and PLB 2 , PLB monomer and dimer. All PLB mutants were expressed on the Cys-less PLB background. strong PLB cross-linking signal at 110 kDa obtained with BMH ( Fig. 3A), L31C-PLB was incapable of inhibiting ATP hydrolysis by T317C-SERCA2a over the entire range of ionized Ca 2ϩ concentrations tested (Fig. 3B, open and closed squares), consistent with results using WT-SERCA2a (Fig. 2D). All of these results suggest that L31C (or A)-PLB binds strongly to SERCA2a near residue 317 but is nevertheless devoid of inhibitory function.
A variety of tests were made to verify that the Ca 2ϩ pump was not adversely affected by the T317C point mutation. First, T317C-SERCA2a and WT-SERCA2a exhibited similar Ca 2ϩ -ATPase activities and identical K Ca values for ATPase activation when expressed alone (data not shown). Second, T317C-SERCA2a was inhibited normally by PLB lacking the L31C mutation, and this inhibition was completely reversed by the anti-PLB antibody (Fig. 3B, open and closed circles). Third, we checked for cross-linking of T317C-SERCA2a to N27C-PLB and N30C-PLB, the two gain-of-function PLB mutants (Fig. 4). As expected, the heterobifunctional cross-linking agents, EMCS and KMUS, efficiently cross-linked N27C-PLB and N30C-PLB, respectively, to T317C-SERCA2a (Fig. 4B). Likewise, BMH efficiently cross-linked N30C-PLB (as well as N27C-PLB) to T317C-SERCA2a (Fig. 4B). In contrast, L31C-PLB was only capable of cross-linking to T317C-SERCA2a (not WT-SERCA2a) and only with BMH (Fig. 4, middle panels). Fourth, we also observed that N27C-PLB and N30C-PLB inhibited the Ca 2ϩ -ATPase activity of T317C-SERCA2a like WT-SERCA2a (data not shown). Therefore, the T317C-SERCA2a mutant ap-pears to be a useful surrogate of WT-SERCA2a for assessing PLB and SERCA2a physical and functional interactions.
We also noted that like cross-linking to WT-SERCA2a ( 4A, two bottom panels), cross-linking of the two PLB double mutants, N27C,L31C-PLB and N30C,L31C-PLB, to Cys 318 or Lys 328 of T317C-SERCA2a was virtually eliminated by the L31C mutation (Fig. 4B, two bottom panels). However, in this instance, the N27C or N30C PLB mutations reciprocated by also preventing the cross-linking of L31C of PLB to T317C of SERCA2a.
Allosteric Regulation of Cross-linking of L31C-PLB to T317C-SERCA2a-Allosteric agents markedly affect the crosslinking of PLB to SERCA2a. We previously demonstrated that N27C-PLB and N30C-PLB cross-link most efficiently to WT-SERCA2a when the enzyme is in the E2, nucleotide-bound state. Moreover, specific cross-linking was prevented by the irreversible SERCA inhibitor, thapsigargin, as well as by the anti-PLB monoclonal antibody, 2D12, which reverses SERCA2a inhibition like PLB phosphorylation (11)(12)(13)20). To confirm the conformational requirement for covalent coupling of L31C-PLB to T317C-SERCA2a, effects of allosteric agents on cross-linking at this site were tested. Micromolar Ca 2ϩ , which drives SERCA2a to the E1 conformation, totally inhibited cross-linking of L31C-PLB to T317C-SERCA2a by BMH (Fig. 5A). A K i value of 0.38 Ϯ 0.03 M Ca 2ϩ was obtained, which is consistent with K i values for Ca 2ϩ inhibition of cross-linking at other sites reported previously (11,12) and suggests that PLB dissociates from SERCA2a at this site once the enzyme has bound Ca 2ϩ . Thapsigargin, the irreversible inhibitor of SERCA, abolished the cross-linking of L31C-PLB to T317C-SERCA2a with K i of 0.13 Ϯ 0.02 M (Fig.  5B), supporting our conclusion that PLB is incapable of binding to E2⅐thapsigargin (11,12). In the same experiments, the concentration of SERCA2a in the cross-linking assay was estimated to be 0.2-0.3 M by quantitative immunoblotting (11), confirming the earlier observation that thapsigargin titrates the Ca 2ϩ -ATPase essentially stoichiometrically (23). The nucleotides ATP or ADP, but not AMP, were required for crosslinking of L31C-PLB to T317C-SERCA2a (Fig. 5C), again consistent with our previous finding that domain IB residues of PLB interact most productively with SERCA2a when it is in the nucleotide-bound, E2 conformation (11,12). Half-maximal stimulation of cross-linking occurred at 29.0 Ϯ 3.0 M for ATP and at 50.3 Ϯ 8.7 M for ADP (mean Ϯ S.D. from three determinations), confirming the nucleotide affinities of E2 previously estimated from cross-linking N30C of PLB to Cys 318 of SERCA2a with BMH (11). It should be noted that the affinity of ATP for the E2 state of SERCA is reduced by two orders of magnitude by thapsigargin, as reported by DeJesus et al. (24). Finally, the anti-PLB monoclonal antibody, 2D12, completely eliminated cross-linking of L31C-PLB to T317C-SERCA2a (Fig. 5D), whereas the anti-SERCA2a monoclonal antibody, 2A7-A1, which has no effect on PLB function or SERCA2a activity (12), also had no effect on cross-linking of L31C-PLB to T317C-SERCA2a (Fig. 5D). All of these results demonstrate that defined conformations of SERCA2a and PLB are essential for cross-linking at multiple sites throughout both molecules (11,12).
Time Course of Cross-linking of L31C-PLB to T317C-SERCA2a-Cross-linking of N30C-PLB to Cys 318 of WT-SERCA2a in the presence of BMH is a slow process occurring with a t1 ⁄2 of ϳ15 min at room temperature (11) (Fig. 6). The slowness of the reaction was attributed to poor accessibility of Cys 318 of SERCA2a, which appears buried in the interface between M4, M5, and M6, when SERCA2a is in the E2 conformation (see Fig. 8). In contrast, cross-linking of the adjacent residue, L31C of PLB, to the adjacent residue, T317C of SERCA2a, by BMH is a much faster process, occurring with a t1 ⁄2 of 110 s at room temperature (Fig. 6) and with a t1 ⁄2 of 55 s at 37°C (data not shown). Thus, in comparison to residue 318 of SERCA2a, residue 317 appears more spatially accessible and better oriented to allow rapid cross-linking to PLB.

L31I-PLB Retains Cross-linking Function and Ability to
Inhibit Ca 2ϩ -ATPase-Replacement of Leu 31 of PLB with either of the 2 amino acids containing short hydrocarbon side chains (L31A or L31C) completely abrogated the ability of PLB to cross-link to WT-SERCA2a and to inhibit Ca 2ϩ -ATPase activity ( Figs. 1 and 2). We wondered whether substitution of the comparably hydrophobic isomer, Ile, for Leu 31 would preserve PLB cross-linking function and inhibition of ATPase activity. Indeed, N27C,L31I-PLB cross-linked to WT-SERCA2a as strongly as the single point mutant, N27C-PLB, with EMCS being the most efficient cross-linking reagent for both PLB mutants (Fig. 7A). Likewise, N30C,L31I-PLB cross-linked to WT-SERCA2a as strongly as the single point mutant, N30C-PLB; in this case, BMH and KMUS were the most efficient cross-linkers for both PLB mutants (Fig. 7B). Both double mutants also inhibited Ca 2ϩ -ATPase activity by decreasing the apparent affinity of the enzyme for Ca 2ϩ , and inhibition was reversed by the anti-PLB antibody (Fig. 7, C and D). In other experiments, we observed that the single point mutant, L31I-PLB, was equally effective as WT-PLB in inhibiting Ca 2ϩ - ATPase activity (data not shown). Thus, a long hydrocarbon side chain at residue 31 of PLB appears to be sufficient for retention of PLB structural integrity and inhibition of SERCA2a enzymatic activity. DISCUSSION To gain a better understanding of the structural and functional interactions that govern inhibition of SERCA2a by PLB, we initiated a series of cross-linking studies to localize functionally important amino acid residues that participate in SERCA2a regulation (11,12). Using homobifunctional and heterobifunctional cross-linking agents as molecular rulers, we previously mapped the distances between N27C and N30C in domain IB of PLB and Cys 318 and Lys 328 along M4 of SERCA2a and determined the key allosteric factors that affect the interaction of these residues during the catalytic cycle (11,12). Asn 27 and Asn 30 of PLB, when changed to Ala, produce gainof-function effects characterized by enhanced inhibition of SERCA2a activity (8). Here we have shown that Leu 31 , which produces loss-of-function when mutated to Cys or Ala (14), can also be mapped relative to M4 on SERCA2a. In particular, we were successful in cross-linking L31C-PLB to T317C-SERCA2a using the homobifunctional thiol cross-linking agent, BMH, but not to WT-SERCA2a, using a variety of homobifunctional and heterobifunctional reagents. Furthermore, like previous results observed at Asn 27 and Asn 30 of PLB, the cross-linking at L31C of PLB occurred only in the absence of Ca 2ϩ and in the presence of nucleotide and was completely inhibited by thapsigargin or by the anti-PLB monoclonal antibody that mimics the physiological effects of phosphorylation (25). Taken together, these data encompassing two gain-of-function PLB mutants and one loss-of-function PLB mutant strongly support our hy-pothesis that domain IB of PLB interacts with SERCA2a exclusively in the Ca 2ϩ -free, E2 conformation, preferentially with bound nucleotide, and that PLB does not interact with the thapsigargin-inhibited, E2 state (11,12).
The highly efficient cross-linking of L31C-PLB to T317C-SERCA2a demonstrated for the first time that L31C-PLB, though devoid of inhibitory function, still actively binds to SERCA2a. Using alanine scanning mutagenesis, Kimura et al. first proposed that Leu 31 of PLB interacts directly with SERCA2a and that substitution with alanine at this position disrupts PLB binding to the Ca 2ϩ pump (8,14). This conclusion was supported by results from a subsequent co-immunoprecipitation study by Asahi et al. (15). However, the results presented here directly contradict the idea that Leu 31 mutants of PLB dissociate from SERCA2a and that inhibitory activity necessarily correlates with binding affinity. In particular, BMH cross-linked L31C-PLB to T317C-SERCA2a eight times more rapidly and with similar yield compared with its cross-linking of N30C-PLB to WT-SERCA2a, where up to 40% of the SERCA2a molecules were coupled (11). Thus, if inhibitory capacity of PLB could be explained simply by its ability to bind to the Ca 2ϩ pump, one would expect L31C, L31A, N30C, and N30A mutations all to be functional or to cause gain-of-function. Based on the loss-of-function actually observed for the Leu 31 mutations, we instead suggest that although the L31C (and presumably L31A) PLB mutants are able to bind to SERCA2a, the resulting complexes cannot inhibit Ca 2ϩ -ATPase activity.
Our results with the PLB double mutants, N27C,L31A (or C) and N30C,L31A (or C), suggest that Leu 31 plays a deterministic role in productive associations between PLB and SERCA2a that overrides the interactions at Asn 27 and Asn 30 . We note that L31C-PLB, the single point mutant, cross-linked strongly to T317C-SERCA2a even though all inhibitory function was lost. When coupled with mutations at Asn 27 and Asn 30 , L31C or L31A disrupted the ability of PLB residues 27 or 30 to crosslink to either Cys 318 or Lys 328 of WT-SERCA2a and also abolished all inhibition of the Ca 2ϩ pump normally associated with these gain-of-function Asn mutations (Figs. 1, 2, and 4). Thus, Leu 31 appears to provide the more dominant interaction with SERCA2a, although Asn 27 and Asn 30 also contribute, because both of the double mutants, N27C,L31C-PLB and N30C,L31C-PLB, also failed to cross-link to T317C-SERCA. A productive inhibitory interaction was shown to require a bulky hydrophobic residue at position 31, because either Leu or Ile are tolerated but cannot be substituted by either Ala or Cys.
Modeling of the structural interactions between PLB and SERCA2a at this region also supports the involvement of Leu 31 of PLB in a critical interaction with SERCA2a that is responsible for both inhibitory effect and for facilitating cross-linking to nearby residues. We have built a model using an ␣-helical PLB molecule derived from an NMR structure (26) juxtaposed with the x-ray structure of SERCA1a in the presence of thapsigargin and EGTA (22) to try to account for the cross-linking results obtained to date (Fig. 8). Although there is reason to believe that neither of these structures corresponds strictly to the physiological conformation of the inhibitory complex (see below), it is nevertheless instructive for potential interactions.
Inspection of the SERCA1a structure shows a hydrophobic patch at the cytoplasmic membrane surface formed by Gly 808 and Phe 809 at the top of M6 and Leu 321 , Cys 318 , and the methyl group from Thr 317 on M4. This patch would be suitable for interaction with Leu 31 and Leu 28 (8) of PLB. To maximize these associations, we rotated the PLB helix by ϳ60°relative to our earlier model (12). In addition to explaining the mutational sensitivity of Leu 31 and Leu 28 (8,14), this orientation is con- sistent with cross-linking both of L31C-PLB with T317C-SERCA2a and of V49C-PLB with V89C-SERCA2a (18). However, this orientation is inconsistent with recent threedimensional models of Toyoshima et al. (18) and Hutter et al. (27) in which residue 31 of PLB points in the opposite direction, toward M9. Asn 27 and Asn 30 of PLB are also quite close to Leu 321 of SERCA, as suggested by disulfide formation between N27C-PLB and N30C-PLB and L321C of SERCA2a (18). However, Asn 27 and Asn 30 have moved farther away from Cys 318 of SERCA2a, which is consistent with the slower rate of crosslinking of N30C-PLB to Cys 318 of SERCA2a (11), compared with cross-linking these Asn mutants to Lys 328 (12) and L321C (18). This lower rate of reactivity at Cys 318 also reflects steric constraints due to the location of Cys 318 on the buried face of M4.
It appears likely that the native structures of both PLB and SERCA vary to some degree from those depicted (Fig. 8). In the case of SERCA, crystallization required the presence of thapsigargin and occurred in the absence of ATP, which makes the structure incompatible on two counts with the cross-links to T317C (this study), Cys 318 (11), and Lys 328 (12). Thapsigargin is well known to trap SERCA in a rigid, dead-end complex that is derived from the E2 conformation (28), which may involve locking Cys 318 in the buried location seen in this structure. Thus, the particular configuration of residues forming the hydrophobic patch between M4 and M6 may be different in the E2 state not inhibited by thapsigargin or may benefit from flexibility not allowed by thapsigargin. Indeed, comparison of this structure with that formed in the presence of Ca 2ϩ (21) reveals major changes of both M4 and M6, thereby illustrating the dynamic nature of this region of the pump and suggesting that the accessibility of Cys 318 to crosslinking may reflect either differences in structure or dynamics of M4 in the absence of thapsigargin. Furthermore, we can conclude that Ca 2ϩ binding to SERCA will replace the hydrophobic patch at the cytoplasmic membrane interface of M4 and M6 with a surface of different character, very likely contributing to the displacement of PLB from its binding site and explaining the cross-linking requirement for the Ca 2ϩfree state of SERCA2a.
In the case of PLB, the model does not explain why L31C cannot cross-link to Cys 318 and, given the unphysiological con- ditions used for the NMR measurements (nonpolar solvent and lack of SERCA2a interaction) (26), changes to the cytoplasmic domain of PLB in this model can also be contemplated. One possibility is that Leu 31 is constrained by its participation in the transmembrane helix but that this helix becomes unwound beyond Leu 31 , which indeed was suggested in earlier models (6,16,18). The model of Tatulian et al. (16) includes a short anti-parallel, two-strand ␤-sheet that runs parallel to the membrane surface and that includes both Asn 27 and Asn 30 . Unlike the highly extended chain in the model of Toyoshima et al. (18), this short sheet would keep Asn 30 rather near Cys 318 and potentially maintain the 5 Å differential observed in crosslinking these Asn residues to Lys 328 of SERCA. This ␤-sheet would likely exclude interaction between Lys 3 of PLB and the KDDK 400 loop of SERCA2a, but the early cross-linking suggested at these sites (29) was not reproducible in later studies (12).
The results of cross-linking and mutagenesis make clear that the region of PLB at the membrane surface is crucial in defining the functional interaction with SERCA2a. This is likely to reflect a particular surface that is presented by SERCA2a in the E2, Ca 2ϩ -free conformation and that is recognized by the portion of PLB that includes residues 27-31. The idea of a hydrophobic surface on SERCA2a binding PLB is consistent with mutagenesis results showing that mutation of the Asn residues of PLB to either Ala or Cys at the interaction site enhances the inhibition. The mechanism of PLB inhibition has been attributed to a slowing of the conformational change during binding of Ca 2ϩ to SERCA2a (4). From a structural point of view, this could be explained by PLB binding to the E2 conformation of the pump and physically resisting the movements of M4 and M6 that are required to bind Ca 2ϩ , as well as the dramatic tilting of M2 that accompanies the transition to the E1 conformation (21,22).
In summary, our data on cross-linking the loss-of-function PLB mutant, L31C, to T317C-SERCA2a further demonstrate the power of cross-linking agents as molecular rulers to reveal important information on structural interactions between integral membrane proteins at a high level of resolution. Crosslinking agents also allow functional issues to be addressed, in this case by providing an opportunity to assess effects of allosteric regulators required for SERCA2a activity on protein-protein interactions occurring in the membrane. After applying N27C and N30C mutations to PLB, an initial model was created by manual docking using distance constraints determined by our cross-linking studies (11,12). Local energy minimization was applied to the interface between the two molecules using XPLOR. Key PLB residues involved in cross-linking are N27C, N30C, L31C, and V49C (18), whereas Leu 28 is in a position to interact with M4 of SERCA. Cross-linked residues on SERCA are V89C (18), T317C, Cys 318 (11), L321C (18), and Lys 328 (12); also shown are Phe 809 , which contributes to a hydrophobic patch that could represent a PLB binding surface, and several residues involved in Ca 2ϩ binding (namely, Glu 771 , Thr 796 , and Asp 800 in green) (21). Thapsigargin (TG) is shown in Corey-Pauling-Koltun representation, and the C␣ positions of the 10 transmembrane helices of the Ca 2ϩ -ATPase are traced by a wire. The figure was prepared using PyMOL.