Nucleotide Activation of the Ca-ATPase*

Background: FITC is a useful but underutilized covalent probe of the Ca-ATPase nucleotide-binding site. Results: We measured time-resolved emission, anisotropy, and quenching of FITC-labeled Ca-ATPase. We used enzyme reverse mode to synthesize FITC monophosphate as a tethered, fluorescent ATP analog. Conclusion: The Ca-ATPase active site exhibits increased dynamics when enclosed with bound ATP. Significance: Internal entropy contributes to long range coupling and catalysis in the Ca-ATPase. We have used fluorescence spectroscopy, molecular modeling, and limited proteolysis to examine structural dynamics of the sarcoplasmic reticulum Ca-ATPase (SERCA). The Ca-ATPase in sarcoplasmic reticulum vesicles from fast twitch muscle (SERCA1a isoform) was selectively labeled with fluorescein isothiocyanate (FITC), a probe that specifically reacts with Lys-515 in the nucleotide-binding site. Conformation-specific proteolysis demonstrated that FITC labeling does not induce closure of the cytoplasmic headpiece, thereby assigning FITC-SERCA as a nucleotide-free enzyme. We used enzyme reverse mode to synthesize FITC monophosphate (FMP) on SERCA, producing a phosphorylated pseudosubstrate tethered to the nucleotide-binding site of a Ca2+-free enzyme (E2 state to prevent FMP hydrolysis). Conformation-specific proteolysis demonstrated that FMP formation induces SERCA headpiece closure similar to ATP binding, presumably due to the high energy phosphoryl group on the fluorescent probe (ATP·E2 analog). Subnanosecond-resolved detection of fluorescence lifetime, anisotropy, and quenching was used to characterize FMP-SERCA (ATP·E2 state) versus FITC-SERCA in Ca2+-free, Ca2+-bound, and actively cycling phosphoenzyme states (E2, E1, and EP). Time-resolved spectroscopy revealed that FMP-SERCA exhibits increased probe dynamics but decreased probe accessibility compared with FITC-SERCA, indicating that ATP exhibits enhanced dynamics within a closed cytoplasmic headpiece. Molecular modeling was used to calculate the solvent-accessible surface area of FITC and FMP bound to SERCA crystal structures, revealing a positive correlation of solvent-accessible surface area with quenching but not anisotropy. Thus, headpiece closure is coupled to substrate binding but not active site dynamics. We propose that dynamics in the nucleotide-binding site of SERCA is important for Ca2+ binding (distal allostery) and phosphoenzyme formation (direct activation).

SERCA 3 is a 110-kDa membrane protein that relaxes muscle by transporting calcium from the cytoplasm into SR (1,2). SERCA comprises 10 transmembrane (TM) helices, plus a large cytoplasmic headpiece with three domains as follows: nucleotide binding (N), phosphorylation (P), and phosphatase actuator (A) (Fig. 1) (3). SERCA binds two Ca 2ϩ ions in the TM domain, which are pumped into the SR lumen using energy derived from ATP hydrolysis and proton exchange (4,5). The kinetic cycle of SERCA is a series of structural and chemical transitions, including intermediates with high Ca 2ϩ affinity (E1), low Ca 2ϩ affinity (E2), and phosphoenzyme formation at Asp-351 (EP) (Scheme 1) (6). SERCA is a member of the "P-type" ion motive ATPase family, forming a transient aspartyl phosphate intermediate during the transport cycle (7). The three cytoplasmic domains are collectively responsible for phosphoryl transfer and phosphoenzyme turnover, resulting in energy transduction to the TM domain for Ca 2ϩ transport (black arrow in Scheme 1) (2,8).
Catalysis by model enzymes is dependent on protein dynamics (domain, backbone, and sidechain), as demonstrated for dihydrofolate reductase, adenylate kinase, and cAMP-dependent protein kinase (9 -14). However, detailed connections between SERCA structure, dynamics, and mechanism remain largely unknown. Ligand-induced changes in SERCA have been successfully detected using spectroscopic probes attached to naturally reactive residues and genetically encoded sites (15,16). Fluorescein isothiocyanate (FITC) selectively reacts with Lys-515 (17)(18)(19)(20) in the nucleotide-binding pocket of the N domain ( Fig. 1). Fluorescence of FITC-SERCA decreases by 5% upon Ca 2ϩ binding, indicating long range coupling between Ca 2ϩ binding in the TM domain and FITC fluorescence in the N domain (21)(22)(23). Time-resolved phosphorescence anisotropy of erythrosin iodoacetamide at Cys-674 in the P domain detected increased microsecond dynamics upon ATP binding to a Ca 2ϩ -free enzyme (E2 to ATP⅐E2), revealing nucleotide-de-pendent coupling between N and P domain dynamics (24). Fluorescence resonance energy transfer (FRET) and molecular dynamics simulations have identified Ca 2ϩ -induced domain motions in the cytoplasmic headpiece that are critical for ATP hydrolysis by SERCA (25,26). Thus, biophysical analysis is useful for detecting SERCA structural dynamics, but more correlations are needed with enzyme kinetics.
SERCA is competent to synthesize ATP using enzyme reverse mode (Ca 2ϩ efflux) in SR vesicles preloaded with Ca 2ϩ (Scheme 1) (27). A unique low fluorescence state (LFS) of FITC-SERCA, in which fluorescence is decreased by ϳ50% compared with other states, was observed long ago following Ca 2ϩ chelation and efflux from preloaded vesicles (28). FITC-SERCA in LFS was initially assigned as a standard phosphoenzyme (phospho-Asp-351) in the "E1P-like" or E1P⅐apo state (28 -32). More recently, McIntosh et al. (33) have rigorously demonstrated that FITC is the phosphorylated species in LFS, becoming FITC monophosphate (FMP) with a stable phosphoester bond in the N domain, instead of a labile acylphosphate bond at Asp-351 in the P domain (34). Thus, for FITC-SERCA in preloaded vesicles, chelation of external Ca 2ϩ induces enzyme reverse mode (Ca 2ϩ efflux) using FITC as phosphoryl acceptor (green arrow in Scheme 1), thereby synthesizing FMP tethered in the active site of SERCA (33). For comparison, TNP-8N 3 -ATP is a similar covalent pseudosubstrate that serves to drive Ca 2ϩ transport into SR (black arrow in Scheme 1) when tethered in the active site of SERCA (35). Both FMP-SERCA and TNP-8N 3 -ATP-SERCA have been characterized by steady-state fluorescence and absorbance measurements (33,35,36).
Time-resolved spectroscopy and molecular modeling bridge the gap between enzyme kinetics and static crystal structures. FITC remains a highly useful fluorescent probe of SERCA structural dynamics. Previously, we used steady-state emission of FITC-SERCA to monitor Ca 2ϩ binding in the TM domain (37,38). We also used FITC as a FRET acceptor to monitor distances from FRET donors in the P and A domains (16,39). Here, we examined structural dynamics of FITC-SERCA and FMP-SERCA using time-correlated single photon counting (TCSPC), a technique in which a subnanosecond laser pulse excites the sample, followed by subnanosecond-resolved detection of fluorescence emission (40). This experiment detected directly excited-state lifetimes, thereby resolving structural states of SERCA. Time-resolved detection of fluorescence anisotropy and iodide quenching provided additional information on dynamics and accessibility in the active site. Results from time-resolved spectroscopy were correlated with molecular modeling and conformation-specific proteolysis, thereby identifying an ATP-induced order-to-disorder transition that precedes Ca 2ϩ binding and phosphoryl transfer. We propose that internal entropy contributes to long range coupling and catalysis in SERCA.
FITC Labeling of SERCA-SR vesicles were isolated from rabbit fast twitch muscle using differential centrifugation (41). SR vesicles were resuspended in 300 mM sucrose and 30 mM MOPS (pH 7.0), flash-frozen in liquid nitrogen, and stored at Ϫ80°C. Protein concentration of SR vesicles was determined by the Biuret method using bovine serum albumin (BSA) as standard. SERCA was labeled in SR vesicles (2 mg/ml) with 10 M FITC for 20 min at 25°C in 100 mM KCl, 5 mM MgCl 2 , and 30 mM Tris (pH 8.9) (16,39). The labeling reaction was terminated by 5-fold dilution in 300 mM sucrose and 50 mM MOPS (pH 7.0) with bovine serum albumin (1 mg/ml) as a scavenger of unreacted dye. Labeled SR vesicles were collected by centrifugation (62,000 ϫ g for 30 min at 4°C), washed once, and resuspended in 300 mM sucrose and 30 mM MOPS (pH 7.0). Stoichiometry of labeling was determined by absorbance of SR vesicles solubilized in 0.1% SDS and 1.0 N NaOH, using an extinction coefficient of 69,300 M Ϫ1 cm Ϫ1 at 494 nm for FITC conjugated to SERCA (16,39). Specificity of labeling was verified by in-gel fluorescence.
SDS-PAGE, In-gel Fluorescence, and Coomassie Densitometry-Electrophoresis was performed using Laemmli gels with 4 -15% FIGURE 1. Molecular models of FITC-SERCA and FMP-SERCA. Atomic resolution models were constructed using x-ray crystal structures. Structural state, probe identity, and PDB code are indicated at bottom (see also Scheme 1). Green, nucleotide-binding domain. Blue, phosphorylation domain. Red, actuator domain. Gray, transmembrane domain. Orange, FITC. Purple, FMP. SCHEME 1. SERCA kinetic cycle. Active Ca 2ϩ transport into SR utilizes cytosolic ATP (black arrows). Reverse mode is driven by Ca 2ϩ efflux and cytosolic phosphate (green arrow), using ADP or FITC as phosphate acceptor. FMP-SERCA is the fluorescent analog of SERCA in the ATP⅐E2 state (green box). E2 is the ground state (black box). ATP⅐E1⅐2Ca is the fully activated enzyme (red box), leading to phosphoryl transfer and Ca 2ϩ transport.
acrylamide. SR vesicles were solubilized for 10 min at 25°C in 2.5% SDS, 5% glycerol, and 62.5 mM Tris (pH 6.8). In-gel fluorescence of labeled SERCA was quantified using the Storm 860 Imaging System (GE Healthcare) in blue fluorescence mode (excitation ϭ 450 Ϯ 30 nm; emission Ͼ520 nm). SERCA content in SR vesicles was quantified by Coomassie densitometry using the Odyssey Imaging System (LI-COR, Lincoln NE).
ATPase Assay-SERCA activity was assayed at 25°C in 100 mM KCl, 5 mM MgCl 2 , 3 mM Na 2 ATP, and 50 mM MOPS (pH 7.0). ADP production by SERCA was coupled to NADH oxidation by an ATP-regenerating system (0.2 mM NADH, 0.6 mM phosphoenolpyruvate, 10 units/ml lactate dehydrogenase, and 10 units/ml pyruvate kinase) (16). The rate of ATP hydrolysis was calculated as the rate of NADH oxidation, measured by the decrease in NADH absorbance at 340 nm using an extinction coefficient of 6220 M Ϫ1 cm Ϫ1 . The Ca 2ϩ ionophore A23187 was added (3 g/ml) to eliminate the buildup of a Ca 2ϩ gradient inside SR vesicles (i.e. product inhibition) (42). Specific activity is expressed in international units (1 IU ϭ 1 mol mg Ϫ1 min Ϫ1 at 25°C), calculated using Ca 2ϩ -dependent ATPase activity and fractional SERCA content in SR vesicles.
Ligand Stabilization of SERCA Structural States-We carefully followed the protocol previously described for formation of LFS (29,30,32,33). FITC-SERCA was run through the series of five states as follows: 1) Ca 2ϩ -free E2; 2) Ca 2ϩ -bound E1; 3) actively cycling phosphoenzyme EP; 4) single-turnover enzyme reverse mode to produce the low fluorescence state FMP-SERCA, and 5) return to actively cycling phosphoenzyme EP final (Fig. 3, A and B). The standard solution contained 100 mM KCl, 3 mM MgCl 2 , and 50 mM MOPS (pH 7.0). Ionized Ca 2ϩ concentrations were calculated using the FREE1 program. At the beginning of the series, the Ca 2ϩ -free state (E2) was stabilized by adding 40 M EGTA to chelate contaminating Ca 2ϩ from water and solution chemicals. After 1 min in E2, the Ca 2ϩbound state (E1) was stabilized by adding 50 M CaCl 2 . After 1 min in E1, 10 mM acetyl phosphate (AcP) was added to initiate Ca 2ϩ transport and phosphoenzyme cycling (EP). After 3 min of Ca 2ϩ loading of SR vesicles, FMP was formed by chelating all extravesicular Ca 2ϩ with 2 mM EGTA ([Ca 2ϩ ] i Ͻ7.0 nM), thus inducing Ca 2ϩ efflux and single-turnover FMP synthesis, as described previously (29,30,32,33). After 2 min as FMP-SERCA (ATP⅐E2 analog), actively cycling phosphoenzyme was re-formed (EP final ) by adding 2 mM CaCl 2 . Ligand-stabilized states of FITC-SERCA and FMP-SERCA were analyzed by proteolytic cleavage and fluorescence spectroscopy, as described below.
Steady-state Fluorescence Spectroscopy-Steady-state measurements were recorded on a Varian Cary Eclipse fluorometer using a xenon lamp as excitation source (47). Fluorescence was measured at 25°C with 25 g/ml SR protein (ϳ0.16 M SERCA). Samples were preincubated for 3 min at 25°C and stirred continuously. The standard solution contained 100 mM KCl, 3 mM MgCl 2 , and 50 mM MOPS (pH 7.0). For FITC and FMP, the excitation and emission wavelengths were 480 Ϯ 5 and 520 Ϯ 5 nm, respectively. For tryptophan, the excitation and emission wavelengths were 295 Ϯ 5 and 340 Ϯ 5 nm.
Time-resolved Fluorescence Spectroscopy-Time-resolved measurements were recorded using TCSPC (40). Samples were excited with a subnanosecond pulsed diode laser at 485 Ϯ 10 nm (LDH 485 from Picoquant, Berlin, DE). The laser power was 0.6 milliwatt with a repetition rate of 10 MHz. The laser pulses are highly uniform in shape and intensity (full width at halfmaximum Ͻ100 ps; 6 nJ/pulse) (40). Emission was selected using a bandpass filter (519 Ϯ 5 nm) and detected using a single photon avalanche photomultiplier module (PMN-100 from Photonics Solutions, Edinburgh UK) with a photon-counting board (SPC-130-EM from Becker and Hickl, Berlin DE). To avoid anisotropy effects, the emission polarizer was set to the magic angle (54.7°) during lifetime measurements. The instrument response function (IRF) was acquired using scattered excitation light detected with emission polarizer set to vertical (0°) but without an emission filter.
Time-resolved fluorescence waveforms were analyzed by multiexponential decay simulation and nonlinear least squares minimization (24,40). The observed waveform, F obs (t), was fit by the decay simulation, F sim (t), which had been iteratively convolved with the measured instrument response function (IRF) shown in Equations 1 and 2, where F(0) is the initial fluorescence intensity; x i is the mole fraction, and i is the decay lifetime. The number of exponentials, n, was determined by minimizing the 2 value between F obs (t) and F sim (t) waveforms. The time-resolved fluorescence waveform of each biochemical state was independently fitted. Total emission was determined by integrating F obs (t).
Time-resolved Fluorescence Anisotropy-Time-resolved fluorescence anisotropy (TFA) experiments were performed with emission polarizer oriented vertically (0°, F V (t)), horizontally (90°, F H (t)), and at the magic angle (54.7°, F M (t)). TFA data analysis (48) was used to calculate anisotropy as shown in Equation 3, where g is a correction factor based on polarizer calibration. TFA curves were analyzed using a model-independent sum of exponentials plus a constant shown in Equation 4, where r i is the pre-exponential factor of each correlation time component; i is the correlation time, and r ∞ is the amplitude of immobilized component (residual anisotropy). We tested up to three correlation time components and found that two correlation time components are necessary and sufficient to fit r(t). Because all TFA decay curves fit best to two correlation times and had similar values for observed initial anisotropy (r(t) at t ϭ 0) and residual anisotropy (r ∞ ), we globally linked correlation times for anisotropy decay fitting of FITC-SERCA and FMP-SERCA.
Fluorescence Quenching-Solvent accessibility of FITC and FMP was assessed using iodide (I Ϫ ) quenching. Time-resolved detection demonstrated that iodide quenching was mostly independent of temperature (10, 25, and 37°C), indicating little to no static quenching (ϳ10%). Steady-state detection demonstrated that quenching was linear from 0 to 200 mM iodide, indicating that dynamic quenching is the predominant mechanism (collisional). Because iodide quenching was collisional, we used steady-state intensity to calculate K SV . The solvent accessibility of the probe was determined by plotting F 0 /F against iodide concentration and fitting the data to Equation 5 for collisional quenching, where ͗͘ is average fluorescence lifetime; F is steady-state emission intensity; [Q] is quencher concentration, and K SV is the Stern-Volmer collisional quenching constant, an indicator of solvent accessibility (49,50).
Molecular Modeling-FITC and FMP were modeled into atomic coordinates of SERCA using the DS Visualizer molecular modeling software (Accelrys, San Diego), as described previously for FITC linked to Lys-515 of CFP-SERCA (16). FITC models were built using x-ray crystal structures of E2⅐Tg (PDB code 1IWO) (51), ADP⅐E2⅐MgF Ϫ ⅐Tg (PDB code 1WPG) (52), E1⅐2Ca (PDB code 1SU4) (3), and ADP⅐.E1⅐2Ca⅐AlF 4 Ϫ (PDB code 2ZBD) (52). FMP and FITC were modeled into ATP⅐E2⅐Tg (PDB code 3AR4) (53). For models based on x-ray structures with bound nucleotide (ATP or ADP), the nucleotide was first removed, and then the fluorescent probe was linked to Lys-515 and manually docked in the nucleotide pocket of SERCA (Fig.  2B). The clean geometry function of DS Visualizer was used to energy-minimize the orientation of fluorescent probes and SERCA residues within 5 Å. The VMD program (54) was used to calculate solvent-accessible surface area (SASA) of FITC and FMP bound to SERCA.
Statistical Analysis-Experiments were performed in triplicate or greater (n Ն3). Data are presented as means Ϯ S.E.
In-gel fluorescence imaging demonstrated that virtually all of FITC is covalently bound to SERCA (Fig. 3C, right panel). The labeling stoichiometry was 1.2 Ϯ 0.1 FITC molecules bound per SERCA molecule. This slight excess labeling is not enough to affect significantly the interpretation of the results below. ATPase activity of FITC-SERCA was inhibited Ͼ95%, indicating that FITC labeling blocks catalytic binding of ATP.
To determine the effect of fluorescent labeling on SERCA, we measured steady-state Trp fluorescence of SERCA, FITC-SERCA, and FMP-SERCA (supplemental Fig. S2). Steady-state Trp fluorescence indicated that SERCA, FITC-SERCA, and FMP-SERCA show similar fluorescence changes in response to Ca 2ϩ binding and phosphoenzyme formation (Table 1 and supplemental Fig. S2). We conclude that fluorescently labeled SERCA shows the same conformational coupling as unlabeled SERCA.
In-gel Fluorescence-SDS-PAGE was used to verify FMP-SERCA formation (Fig. 3C). Here, we observed that E2 and E1 states of FITC-SERCA show the same fluorescence on SDS-PAGE (Fig. 3D), unlike the small but significant difference observed in the absence of SDS (compare Fig. 3, B with D). Thus, SDS denaturation abolishes Ca 2ϩ -induced fluorescence changes of FITC-SERCA. However, FMP-SERCA in LFS showed the same decrease in fluorescence both in the absence and presence of SDS (compare Fig. 3, B with D). Thus, our quantitative gel results for LFS/FMP fluorescence match the qualitative in-gel observations by McIntosh et al. (33), where LFS was shown to be FMP-SERCA. The fact that low fluorescence is preserved in the presence of SDS indicates that the fluorescence phenomenon for LFS is due to a chemical modification of FITC (i.e. FMP formation), instead of phosphoenzyme-induced structural changes in SERCA. Our in-gel fluorescence results further suggest that FMP-SERCA forms in low amounts in the EP state, under conditions where the forward and reverse cycles are in steady-state equilibrium but heavily favored toward the forward cycle. Thus, we provide the first quantitative correlation between solution and in-gel fluorescence, concluding that LFS is caused predominantly by FMP formation.
Proteolysis Identifies the Predominant Structural State of FMP-SERCA-Limited ProtK cleavage is an effective assay to identify the predominant structural state of SERCA in a variety of ligand-stabilized conditions (16,43,44,57). Unlabeled SERCA shows specific ProtK cleavage patterns for the Ca 2ϩfree (E2) and Ca 2ϩ -bound (E1) states, producing 96-and 83-kDa fragments, respectively (1st and 4th lanes for SERCA in Fig. 4A). ATP slightly but significantly protects SERCA from ProtK digestion, preserving the 110-kDa band (1st and 2nd   Fig. 4). ATP also slightly but significantly protects the primary proteolytic fragment of E2 (96 kDa) from secondary cleavage by ProtK (1st and 2nd lanes for SERCA in Fig. 4). Unlike ATP, ADP does not provide protection of SERCA in E2 (1st and 3rd lanes for SERCA in Fig. 4). Thus, proteolysis distinguishes the predominant headpiece structure of ATP⅐E2 from those of ADP⅐E2, E2, and E1. FITC-SERCA was assayed for ligand-induced changes in ProtK digestion. The same cut patterns are found for FITC-SERCA and SERCA in E2 and E1 states (1st and 4th lanes in Fig.  4, A and B), indicating the following: (i) FITC-SERCA undergoes similar Ca 2ϩ -dependent conformational changes as unlabeled SERCA and (ii) FITC labeling does not mimic ATP-induced protection from ProtK cleavage (16). We previously demonstrated that addition of AMPPCP to FITC-SERCA in the E1 state does not provide additional ProtK protection, indicating that FITC labeling blocks the catalytic nucleotide-binding site (16). Here, we further demonstrated that neither ATP nor ADP provides protection for FITC-SERCA in the E2 state (1st and 3rd lanes for FITC-SERCA in Fig. 4). Thus, proteolysis indicates that the structure of FITC-SERCA is similar to that of nucleotide-free SERCA.
ProtK digestion of FMP-SERCA in LFS, however, demonstrated that FMP formation gives similar protection as ATP binding to SERCA (1st to 3rd lanes for FMP-SERCA versus 1st and 2nd lanes for SERCA in Fig. 4, A and B). In particular, FMP-SERCA in the absence of Ca 2ϩ shows high amounts of 110-kDa (uncut) and 96-kDa (E2 cut) bands, similar to unlabeled SERCA in the ATP⅐E2 state (compare 1st lane of FMP-SERCA versus 2nd lane of SERCA in Fig. 4). Unlike unlabeled SERCA, FMP-SERCA shows no additional protection with ATP, indicating that FMP blocks the catalytic nucleotide-binding site, similar to FITC (compare 1st and 2nd lanes for SERCA, FITC-SERCA, and FMP-SERCA in Fig. 4). We propose that FMP mimics ATP bound to SERCA, presumably due to similar positioning of each respective high energy phosphoryl group utilized in enzyme catalysis.
Time-resolved Fluorescence Spectroscopy-The free dyes fluorescein monophosphate and fluorescein are reported to have a "very high" quantum yield (56). However, the absolute value for the quantum yield of fluorescein monophosphate has not been reported, and fluorescein monophosphate is not commercially available. To compare quantum yields, FMP was synthesized from FITC-SERCA by Ca-ATPase reverse mode, and fluorescence lifetimes were determined by TCSPC using excitation at 485 nm. Time-resolved fluorescence was analyzed by waveform integration (Fig. 5, A and B) and lifetime fitting (Fig. 5, C and D) to determine whether changes in emission or nonradiative relaxation (58) contribute to the change in absorbance (FITC-SERCA versus FMP-SERCA) to produce LFS.
The integrated intensities of fluorescence waveforms, which are proportional to initial fluorescence intensity F(0) (Equation 1) reveal the same conformation-dependent changes observed by steady-state fluorescence, including ϳ50% decrease in FMP-SERCA (compare Figs. 3, A and B, and 5, A and B, and also see Table 1). Normalized waveforms exhibit similar decay rates, revealing that the fluorescence lifetime of FMP-SERCA (green trace in Fig. 5C) is nearly identical to FITC-SERCA in E2, E1, and EP states. Quantitative lifetime fitting of fluorescence waveforms (Equations 1 and 2) determined that two exponential components were necessary and sufficient to generate an optimal fit for FMP-SERCA and FITC-SERCA (supplemental  Table S1), indicating that the decrease in steady-state fluorescence of LFS is not due is not due to a change in radiative relaxation processes of FMP-SERCA.
Analysis of average lifetime ͗͘ determined from two-lifetime components and their mole fractions (supplemental Eq. S1) was used to compare quantum yield between states. All measured states exhibit an similar average lifetime ͗͘ Ϸ2.6 ns ( Fig. 5D and Table 1), demonstrating that both FITC-SERCA and FMP-SERCA have the same quantum yield. Thus, the decrease in fluorescence for FMP-SERCA was due to optical ground-state phenomena (for decrease in absorbance, see supplemental Fig. S1) and possibly due to the formation of an unobserved population of FMP-SERCA molecules (lifetime-independent static quenching), which is beyond the detection capability of our instrument. There is a correlation between decreased absorbance at 480 nm (supplemental Fig. S1) and decreased fluorescence at 520 nm (Figs. 3 and 5) for FMP-SERCA, indicating that the primary cause of the apparent low steady-state fluorescence is decreased absorbance (33). Here, for the first time we have measured the lifetime of FMP-SERCA in LFS, thereby eliminating changes in radiative and nonradiative relaxation rates as secondary causes for low fluorescence (supplemental Table S1), as compared with FITC-SERCA.
Fluorescence lifetimes of FITC-SERCA determined here by TCSPC are similar to lifetimes previously reported using TCSPC and phase domain spectroscopy (50, 59 -62). Most of these studies measured the average lifetime ͗͘ of FITC-SERCA in a single ligand-stabilized structural state (59 -62). Two of these previous studies measured time-resolved fluorescence of FITC-SERCA in two structural states (E1 versus E2), finding that Ca 2ϩ binding has no effect on ͗͘ (50, 60), similar to current results (Fig. 5, C and D). Here, we have extended previous work by comparing FMP-SERCA to FITC-SERCA in E2, E1, and EP states and by analyzing the distribution of lifetimes and amplitudes. We conclude that FMP-SERCA has the same two fluorescence lifetime components and associated mole fractions as FITC-SERCA in E2, E1, and EP.
Time-resolved Fluorescence Anisotropy-TFA, a technique sensitive to nanosecond dynamics (58), was used to examine the active site of SERCA. TFA is sensitive to probe motion and backbone dynamics but not uniaxial rotation of SERCA in the membrane (rotational correlation time Ͼ1000 ns) or tumbling of SR vesicles (63). TFA decay of FITC-SERCA has not been previously reported. Here, we used TCSPC with excitation at 485 nm to detect TFA of FMP-SERCA and FITC-SERCA. FMP-SERCA exhibits the fastest anisotropy decay (green trace in Fig.  6A), as compared with FITC-SERCA in E2, E1, and EP states, which show slower anisotropy decays that are nearly identical (Fig. 6A). FMP-SERCA and all measured states of FITC-SERCA have similar initial anisotropy (ϳ0.37) and similar residual anisotropy (ϳ0.14) (Fig. 6A), indicating restricted motion on the nanosecond time scale.
TFA data were analyzed by fitting to multiexponential decays. Two exponential components were necessary and sufficient to generate an optimal fit for FMP-SERCA and FITC-SERCA ( Fig. 6 and supplemental Fig. S4). Fitting of TFA data determined that both FMP-SERCA and FITC-SERCA have a fast correlation time ( fast ϭ 0.285 Ϯ 0.071 ns) and a slow correlation time ( slow ϭ 2.42 Ϯ 0.076 ns) (supplemental Fig. S4) but exhibit distinct distributions of fast and slow components (r fast and r slow ) (Fig. 6, B and C). The ratio r fast /r slow qualitatively indicates the dynamic disorder of the active site in each structural state (Fig. 6C). FMP-SERCA exhibits a higher ratio of r fast / r slow , indicating higher dynamic disorder for FMP-SERCA than all biochemical states of FITC-SERCA, which have the same r fast /r slow (Fig. 6C). Here, we observe changes in rotational motion on the nanosecond time scale, consistent with protein backbone motion, and changes on the subnanosecond time scale, consistent with probe motion (58). We propose that the active site of SERCA is dynamically disordered and that the ATP⅐E2 state has greater disorder than E2, E1, and EP states.  Fluorescence Quenching-To further examine structural changes in the cytoplasmic headpiece of SERCA, iodide quenching was used as an indicator of FITC and FMP accessibility in the active site ( Fig. 7A and Table 1). Ligand-stabilized structural states were assayed by adding 0 -279 mM KI, and a standard Stern-Volmer plot was used to quantitate fluorescence quenching. For FITC-SERCA, the K SV of the Ca 2ϩ -free ground state (E2) is 1.78 Ϯ 0.07, whereas the K SV of the Ca 2ϩactivated state (E1) is 2.79 Ϯ 0.05 (Fig. 7A, Table 1), indicating that Ca 2ϩ increases active site accessibility (16,51). Iodide quenching results for E2 and E1 here (Fig. 7A) are similar to those of Highsmith (50), who first identified a Ca 2ϩ -induced increase in K SV (greater accessibility) for FITC-SERCA. Thus, FITC-SERCA in E2 and E1 states show a large difference in active site accessibility (quenching) yet have similar fluorescence emission (lifetime) and dynamics (anisotropy) (summarized in Table 1). FMP-SERCA exhibits the lowest quenching (K SV ϭ 1.23 Ϯ 0.08) of all states tested, indicating that the active site of the ATP⅐E2 state has the most restricted solvent accessibility (Fig. 7A). Actively cycling phosphoenzyme (EP) has a K SV of 2.32 Ϯ 0.04 (Fig. 7A), an intermediate value consistent with EP comprising all structural states in the SERCA kinetic cycle (Scheme 1). We conclude that the cytoplasmic headpiece of SERCA is predominantly closed in the ATP⅐E2 state.
Molecular Modeling of the Active Site-Molecular modeling was used to examine FMP and FITC binding to SERCA (Fig. 1). FITC was docked in four crystal structures of SERCA (E2⅐Tg, E1⅐2Ca, ADP⅐E1P⅐2Ca, and ADP⅐E2P⅐Tg), with conjugation of isothiocyanate to Lys-515 (thiourea linkage). FMP and FITC were also docked into the ATP⅐E2⅐Tg crystal structure. Because of overlapping binding sites, ATP and ADP were removed from the ATP⅐E2⅐Tg and ADP⅐E1P⅐2Ca structures prior to FMP and FITC docking. Fluorescent probes and surrounding SERCA side chains were energy-minimized to attain more accurate positioning of the probe in the nucleotide-binding pocket. Models of FMP-SERCA and FITC-SERCA revealed major differences in the interactions of FMP and FITC with residues in the nucleotide-binding pocket, presumably due to large differences in the SERCA headpiece structure in different crystal structures (Fig. 1).
Our energy-minimized model of FMP-SERCA in the ATP⅐E2⅐Tg crystal structure predicts that FMP and ATP exhibit multiple, overlapping structural motifs within the nucleotidebinding site (Fig. 2B). For example, the thiourea linkage of FMP overlaps with the amide group of adenine; the benzoate ring of FMP shows similar stacking interactions with Phe-487 as the adenine ring, the benzoate oxygens of FMP match location with ribose oxygens, and the phosphate group of FMP is in the same location as the ␥-phosphate of ATP. Thus, our molecular modeling results predict that that FMP utilizes the same nucleotidebinding motif as ATP and suggests mechanisms for synthesis and hydrolysis of FMP, where the 3-O oxygen of FITC is in optimal position to accept or donate the phosphoryl group from Asp-351 (Fig. 2B).
The first molecular model of FITC-SERCA in ADP⅐E1P⅐2Ca (with ADP removed and FITC hand-docked) placed the phenolic 3-O of the xanthenone ring system adjacent to AlF 4 Ϫ at the normally occupied ␤-phosphate position of ADP (32); this model is based on a hybrid structure of PDB 1T5T (64) and EM 1KJU (65). When we tried to align AlF 4 Ϫ with the same 3-O of FITC-SERCA in ADP⅐E1P⅐2Ca based on PDB 2ZBD (66) (with ADP removed and FITC hand-docked and energy-minimized), it resulted in severe ring distortion of FITC due to torsional strain. However, these seemingly contradictory results are not necessarily inconsistent. Because of the high dynamic disorder of the nucleotide-binding site preceding ATP hydrolysis, it is not surprising that molecular models of SERCA show variability in probe orientation following ATP hydrolysis. We suggest that the ring structure of our model of FITC-SERCA in ADP⅐E1P⅐2Ca represents a pre-ADP release state, where the xanthenone ring system has rotated away from Asp-351 (supplemental Fig. S5), whereas the first molecular model of FITC-SERCA in ADP⅐E1P⅐2Ca represents the transition state between phosphoryl hydrolysis and transfer, prior to rotation of the ring system (32).
Solvent Accessibility in the Active Site-As a quantitative measure of probe-protein interactions and headpiece closure, we calculated SASA of FMP and FITC in five models ( Fig. 7B and Table 1). The model of FITC-SERCA in E1⅐2Ca had the highest SASA (282 Å 2 ) of all models, illustrating the openness of the cytoplasmic headpiece in the Ca 2ϩ -bound crystal structure (Fig. 7B). The ADP⅐E1P⅐2Ca model had the lowest SASA (29 Å 2 ), indicating tight headpiece closure in the nucleotidebound structure. The ADP⅐E2P and E2⅐Tg models of FITC-SERCA showed moderate headpiece opening, with SASA of 160 and 147 Å 2 , respectively (Fig. 7B). For the model of FMP-SERCA, we calculated a SASA of 171 Å 2 , indicating that the headpiece is partially closed in ATP⅐E2⅐Tg but that further headpiece closure is required following Ca 2ϩ binding to initiate ATP hydrolysis (ATP⅐E1⅐2Ca in Scheme 1). For comparison, SASA of ATP in the E2⅐Tg crystal structure is 172 Å 2 , very similar to FMP-SERCA (171 Å 2 ). It is apparent that FMP-SERCA shows low accessibility (iodide quenching and predicted SASA) but high mobility (time-resolved anisotropy), indicating that headpiece closure is not coupled to active site dynamics. We propose that ATP increases active site dynamics within a closed cytoplasmic headpiece of SERCA.

DISCUSSION
ATP in the Active Site of SERCA-Nearly 20 residues have been identified that participate in nucleotide binding by SERCA in various structural states, including residues from N, P, and A domains, as determined by mutagenesis (1,57,(67)(68)(69)(70) and crystallography (53,64,(71)(72)(73)(74). The nucleotide-binding site of SERCA is malleable, with alternate modes of binding, including catalytic and regulatory (2,53,72,74). The commonality of binding is stabilization by an intricate network of H-bonds, hydrophobic interactions, and charge-charge interactions between nucleotides and SERCA. For catalytic binding of ATP, Phe-487 (N domain) and Lys-515 (N domain) interact with adenine, whereas Arg-489 (N domain) and Arg-560 (N domain) interact with the polyphosphate tail (64,71). In one proposed mode for regulatory binding of ADP, adenine is bound between "pinchers" formed by Arg-489 (N) and Arg-678 (P domain) (74); this mode of adenine binding is also observed in crystal structures of SERCA with bound TNP-ATP and TNP-ADP (53). In another proposed mode of regulatory binding of ATP, adenine interacts with Phe-487 (N) and Lys-515 (N), but the polyphosphate tail interacts with Arg-678 (P) and Lys-202 (A domain) (2,72). Thus, ATP binding by SERCA is adaptable, with a range of nucleotide orientations and interactions in the active site.
Crystal structures show stable nucleotide binding but provide no information on differences in active site dynamics for substrate and product complexes (ATP⅐E2, ATP⅐E1⅐Ca, ADP⅐E1P⅐2Ca, and ADP⅐E2P). Prior to the availability of crystal structures of SERCA with bound nucleotide, NMR spectroscopy identified a handful of residues in an N-domain fragment of SERCA that interacts with ATP (75). NMR further identified coupling of nucleotide binding to internal dynamics of the N domain, with six residues showing increased backbone mobility and four residues showing decreased backbone mobility (76). Thus, ATP binding induces changes in SERCA internal dynamics.
FITC in the Active Site-FITC and ATP bind competitively in the active site; FITC labeling precludes ATP hydrolysis (55). There is no crystal structure for FITC-SERCA, so the precise location and orientation of FITC are not known. Proteolysis studies indicate that FITC labeling does not induce cytoplasmic headpiece closure like ATP binding (16,32,33), presumably because FITC is missing high energy phosphoryl group(s) necessary for nucleotide-mediated structural changes. FITC-SERCA is competent for phosphoenzyme formation and Ca 2ϩ transport using small molecules with high energy phosphate bonds, including AcP, para-nitrophenyl phosphate, and 3-methoxyfluorescein phosphate (22,28,55). Kinetic studies demonstrate that SERCA follows a similar reaction pathway using ATP or AcP to drive Ca 2ϩ transport (black arrow in Scheme 1) (77). Thus, FITC labeling blocks catalytic ATP binding while maintaining Ca-ATPase function, without inducing nucleotide-dependent structural changes.
FITC-SERCA shows distinct changes in steady-state fluorescence due to Ca 2ϩ binding (Fig. 3) (21-23, 28, 29, 38, 39, 55, 62). These changes in steady-state fluorescence between different structural states of FITC-SERCA are well documented, but their photochemical and structural mechanisms are unclear. Our time-resolved fluorescence data reveal that lifetime components for FITC-SERCA are the same for key structural states (Fig. 5, C and D), indicating that FITC emission is insensitive to active site environmental changes and that changes in steadystate intensity are due to conformation-specific changes in absorbance. Anisotropy and quenching data demonstrate that FITC-SERCA in E2 and E1⅐2Ca (nucleotide-free analogs) have the same active site dynamics but different solvent accessibility (Figs. 6 and 7A), emphasizing the need for complementary high resolution fluorescence assays to detect structural changes between states.
FMP in the Active Site-Strong biochemical evidence, including absorbance, fluorescence, trypsinolysis, and 32 P localization, indicate that FITC on SERCA in LFS is phosphorylated, thereby forming FMP-SERCA (33). FMP-SERCA has approximately the same quantum yield as FITC-SERCA, but a much lower absorption at 500 nm (supplemental Fig. S1), which provides an explanation for the phenomenon of low fluorescence detected at 520 nm (Figs. 3 and 5) (33,56). Here, we performed in-gel fluorescence and confirmed that LFS is preserved in the presence of SDS (Fig. 3), strongly suggesting that a chemical modification occurs to FITC to account for low fluorescence. We performed time-resolved fluorescence spectroscopy to measure the lifetime of FITC-SERCA in different biochemical states. Here, we observed that the average lifetime of FITC-SERCA, which is proportional to quantum yield, was the same for all biochemical states tested, including FMP-SERCA in LFS (Fig. 5). Our in-gel fluorescence of EP (Fig. 3) demonstrates reversibility of the Ca-ATPase kinetic cycle (i.e. low amount of FMP-SERCA synthesis) even under conditions that heavily favor the forward reaction mechanism (i.e. Ca 2ϩ transport by FITC-SERCA). Results obtained here are consistent with previous biochemical evidence that formation of FMP-SERCA is responsible for LFS (33).
Both SERCA and the plasma membrane Ca-ATPase are able to utilize the commercially available 3-methoxy-FMP (free dye) as a substrate for Ca 2ϩ transport (33, 78 -80). Both SERCA and plasma membrane Ca-ATPase are also able to synthesize 3-methoxy-FMP using enzyme reverse mode (green arrow in Scheme 1) (33, 78 -80). Thus, it is likely that FMP (which is not commercially available) can also substitute for ATP as when tethered to SERCA as a pseudosubstrate in the active site (Fig. 2). To test this hypothesis, we performed proteolysis assays to assess the protective effects of nucleotides and fluorophores on ligand-stabilized structural states. We found that for unlabeled SERCA, the nonhydrolyzable ATP analog AMPPCP protects SERCA in the E2 and E1 states from proteolysis ( Fig. 4) (16). Likewise, we found that FMP-SERCA is protected from proteolysis, indicating that FMP formation induces the same structural change as ATP binding (Fig. 4). However, FITC labeling does not provide protection from proteolysis for the E2 or E1 states, verifying the assignment of the active site of FITC-SERCA as nucleotide-free (Fig. 4). Therefore, we propose that FMP acts as an ATP analog for SERCA, thereby serving as a useful fluorescent reporter of ATP dynamics in the nucleotidebinding site.
Anisotropy measurements revealed that FMP-SERCA (ATP⅐E2) has a faster anisotropy decay than FITC-SERCA in the E2 state, suggesting that ATP induces an order-to-disorder transition in the active site of the Ca 2ϩ -free enzyme (Fig. 6). Iodide quenching revealed that FMP-SERCA has the lowest solvent accessibility, whereas FITC-SERCA in the E1 state has the highest solvent accessibility (Fig. 7). These quenching results were supported by molecular modeling, which was used to calculate SASA ( Figs. 1 and 7). We propose that the active site of SERCA in the ATP⅐E2 state has low accessibility but high disorder.
Active Site Dynamics Mediates ATP Activation of Ca 2ϩ Binding and Phosphoryl Transfer-Fluorescence spectroscopy and molecular modeling have provided new insights into the mechanism of nucleotide activation of SERCA. Classic kinetics studies have measured most transition rates in the catalytic cycle. However, these same kinetics studies provide no structural information about SERCA. Conversely, new crystal structures have provided atomic details of many key intermediates but provide no information on the role of thermodynamics in catalysis. Here, we have measured time-resolved fluorescence anisotropy of FITC and FMP in the nucleotide-binding site and interpret our results in light of recent articles that propose active site conformational dynamics mediates substrate binding, enzyme catalysis, and product release (9 -14). An inherent advantage of our study is that FITC and FMP are tethered to the nucleotide-binding site of SERCA, which prevents probe release. Thus, our high resolution fluorescence assays bridge outstanding questions between SERCA kinetics and structure.
At first glance, it seems paradoxical that the ATP⅐E2 state has both a closed headpiece and disordered nucleotide-binding site. The simplest explanation for the apparent paradox is that the key factor for catalysis of ATP is the dynamic disorder of the active site. We propose that active-site disorder enhances ATP hydrolysis by increasing the entropy of the transient intermediate, thereby decreasing the activation energy and increasing the forward reaction rate for ATP hydrolysis and phosphoryl transfer (Fig. 8) (9,10,13,81). Other studies support our interpretation of active site dynamic disorder. For comparison, SERCA shows conformation-specific site-directed labeling by ATP-pyridoxal, which reacts with Lys-684 in the ATP⅐E2 state, but Lys-492 in the ATP⅐E1 state, indicating movement of the ␥-phosphate group of ATP upon subsequent enzyme activation by Ca 2ϩ (82). Time-resolved phosphorescence anisotropy shows that microsecond dynamics in the SERCA headpiece is increased by ATP binding to the Ca 2ϩ -free enzyme (ATP⅐E2) (24). It has been proposed that the ␥-phosphate of ATP and Asp-351 of the P domain exhibit electrostatic repulsion during phosphoryl transfer and that mutation of Asp-351 to neutral or electropositive residues reduces the effects of these repulsive forces (45,67,71). Thus, it is likely that electrostatic repulsion increases dynamic disorder for ATP in the active site, helping to catalyze ATP hydrolysis and phosphoenzyme formation within a closed cytoplasmic headpiece.
Active site dynamics probably plays an important role on pre-catalysis arrangement of ATP for phosphoryl transfer. SERCA utilizes bimolecular nucleophilic substitution (S n 2) for phosphoryl transfer (67,83,84), a reaction mechanism that relies heavily on both the collisional frequency of the two reacting molecules and their orientation (83,85). SERCA hydrolyzes ATP at an extremely slow rate in the absence of Ca 2ϩ , suggesting that SERCA employs a mechanism to prevent nonproductive phosphoryl transfer in the Ca 2ϩ -free ATPase (6,45,86). We propose that the ATP⅐E2 state in the kinetic cycle has high dynamics but that the orientation and location of the phosphate tail is nonoptimal for phosphoryl transfer. Ca 2ϩ binding further closes the cytoplasmic headpiece (25,26) while maintaining high active site disorder, aiding in the orientation of the two chemical groups (␥-phosphate and Asp-351) needed for proper transition-state geometry, thus enabling the ATP hydrolysis step to proceed rapidly in ATP⅐E1⅐2Ca (Fig. 8) (6,45,85,86). It is possible that dynamic disorder in the active site is even further increased following Ca 2ϩ binding, because the two electronegative groups of Asp-351 and ␥-phosphate of ATP are brought closer together. Thus, we propose that active site disorder in the Ca-ATPase increases the probability of productive collision and phosphoryl transfer between ATP and Asp-351. Further spectroscopic and kinetics studies, preferably using new technology that resolves structural and kinetics states simultaneously (87), will be needed to test this hypothesis more rigorously.
A Branched Kinetic Pathway for Initial Activation of SERCA by ATP or Ca 2ϩ -Traditional kinetic schemes often show SERCA binding Ca 2ϩ before ATP (1,15,66), but in muscle cells SERCA probably binds ATP before Ca 2ϩ (2,6,88), indicating that there is a branched pathway of sequential ligand activation of SERCA (supplemental Scheme S1). We previously examined the effects of Ca 2ϩ binding to nucleotide-free ATPase (bottom pathway in supplemental Scheme S1) using all-atom molecular dynamics simulations, which demonstrated that Ca 2ϩ induces an activated, but empty, nucleotide-binding site in a closed cytoplasmic headpiece (25). Here, we used FMP as a fluorescent ATP-like pseudosubstrate to examine nucleotide activation of SERCA (top pathway in supplemental Scheme S1). For a twoligand enzyme, binding of the first ligand is able to allosterically regulate binding of the second ligand binding, where an increase in entropy at the primary active site decreases the activation energy for ligand binding at the secondary active site (89). Thus, we propose that increased disorder (entropy) upon ATP binding in the nucleotide site facilitates Ca 2ϩ binding in the TM domain of SERCA (Fig. 8).
It is likely that these two branched pathways are not mutually exclusive, rather, both pathways are utilized in muscle (supplemental Scheme S1). Kinetic studies demonstrate that ATP accelerates Ca 2ϩ binding to the ATPase (6,86,90), indicating that the rate of Ca 2ϩ binding to ATP-bound SERCA (ATP⅐E2) (step 2 in top pathway of supplemental Scheme S1) is greater than the rate of Ca 2ϩ binding to ATP-free SERCA (E2) (step 1Ј in the bottom pathway). Because of distinct kinetic and thermodynamic properties of SERCA, there are probably physiological differences in muscle that result from changes in ligand concentration and therefore the relative flux through one pathway or the other. Thus, the key determinant for pathway selection by SERCA is the concentration of Ca 2ϩ and ATP in muscle (91).
"Saturating ATP" Hypothesis Suggests ATP Activation of Ca 2ϩ Release and Phosphoenzyme Decay-The ATP binding affinity of SERCA at the catalytic site is 5-10 M (45), whereas the concentration of ATP in muscle cells is 5-8 mM (2). The importance of saturating ATP in muscle has recently been emphasized, proposing that ATP is bound to SERCA through most of the kinetic cycle (2,72,92). Not only does ATP increase Ca 2ϩ binding and phosphoryl transfer (Fig. 8), but ATP also accelerates Ca 2ϩ release from E2P⅐2Ca and E2P decomposition (Scheme 1, bottom row) (2,6,93). Thus, we propose that active site dynamics is also increased by rebinding of ATP immediately following ADP release, thereby accelerating subsequent Ca 2ϩ release and phosphoenzyme decay.
Conclusions-We used high resolution fluorescence assays to characterize FMP-SERCA synthesized using FITC-SERCA and enzyme reverse mode. Using conformation-specific proteolysis, we provide evidence that FMP-SERCA is a structural analog of SERCA in the ATP⅐E2 state. Quenching and anisotropy measurements of FMP-SERCA suggest that the ATP⅐E2 state of SERCA has a closed headpiece but disordered active site. These data reveal new insights into structural transitions required for coupling ATP activation to Ca 2ϩ transport by SERCA. These results, together with our recently published molecular dynamics simulations of Ca 2ϩ activation (25), provide a compelling mechanistic model for ligand activation of the Ca-ATPase.