Redistribution of SERCA calcium pump conformers during intracellular calcium signaling

The conformational changes of a calcium transport ATPase were investigated with molecular dynamics (MD) simulations as well as fluorescence resonance energy transfer (FRET) measurements to determine the significance of a discrete structural element for regulation of the conformational dynamics of the transport cycle. Previous MD simulations indicated that a loop in the cytosolic domain of the SERCA calcium transporter facilitates an open-to-closed structural transition. To investigate the significance of this structural element, we performed additional MD simulations and new biophysical measurements of SERCA structure and function. Rationally designed in silico mutations of three acidic residues of the loop decreased SERCA domain–domain contacts and increased domain–domain separation distances. Principal component analysis of MD simulations suggested decreased sampling of compact conformations upon N-loop mutagenesis. Deficits in headpiece structural dynamics were also detected by measuring intramolecular FRET of a Cer–YFP–SERCA construct (2-color SERCA). Compared with WT, the mutated 2-color SERCA shows a partial FRET response to calcium, whereas retaining full responsiveness to the inhibitor thapsigargin. Functional measurements showed that the mutated transporter still hydrolyzes ATP and transports calcium, but that maximal enzyme activity is reduced while maintaining similar calcium affinity. In live cells, calcium elevations resulted in concomitant FRET changes as the population of WT 2-color SERCA molecules redistributed among intermediates of the transport cycle. Our results provide novel insights on how the population of SERCA pumps responds to dynamic changes in intracellular calcium.

The sarcoendoplasmic reticulum Ca 2ϩ -ATPase (SERCA) 2 is the ion transporter responsible for sequestering calcium in the sarcoplasmic reticulum (SR) and endoplasmic reticulum (ER). Mutations in the skeletal muscle isoform SERCA1a cause Brody myopathy, with impaired muscle relaxation (1). Mutations of the nonmuscle SERCA2b isoform are the basis for Darier disease, a disorder characterized by epidermal lesions (2). Alterations in cardiac SERCA2a expression, activity, and regulation have been linked to cardiovascular diseases such as heart failure, hypertrophy, and senescence (3)(4)(5). SERCA is therefore a high-value therapeutic target for many diseases and cell types (6,7), but enhancing calcium handling by SERCA gene delivery has proven challenging (8). Thus, there is a critical unmet need in the development of small-molecule therapies based on modulation of endogenous SERCA function.
SERCA is composed of four major domains. The cytosolic headpiece consists of an actuator (A) domain, a nucleotidebinding (N) domain, and an autophosphorylation (P) domain. The spatial arrangement of these domains changes during phosphoryl transferase steps in the catalytic cycle, thereby altering the orientation and affinity of the Ca 2ϩ -transport sites in the transmembrane (TM) domain (9 -11). SERCA transports two Ca 2ϩ ions per ATP molecule hydrolyzed, with the formation of a phosphoenzyme intermediate (phospho-Asp), and thus is classified as a P-type ATPase (12). The catalytic cycle of P-type pumps was first identified for the Na ϩ /K ϩ -ATPase and is referred to as the Post-Albers transport mechanism, where the Ca 2ϩ -transport sites show alternating access (cytosolic and luminal) and alternating affinity (high and low, respectively).
To investigate the rearrangement of SERCA cytosolic domains and identify key structural determinants we previously performed a computational study of SERCA headpiece motions (13). The analysis indicated that a specific structural feature of the N-domain, the N␤5-␤6 loop, facilitates the transition of the SERCA headpiece from an open arrangement of domains to a more compact architecture. This short ␤-loop is composed of residues 426 -436 ( 426 DYNEAKGVYEK 436 ) and contains three negatively-charged amino acids (Asp-426, Glu-429, and Glu-435) that form salt bridges and hydrogen bonds with basic and polar residues on the surface of the A-domain.
These interactions help initiate domain-domain contact and eventually induce SERCA headpiece closure. When these three negative charges were mutated in silico to Ala, contacts between the A-and N-domains were decreased, and headpiece dynamics were altered in short MD trajectories. This previous finding suggested that the three negatively-charged residues of the N␤5-␤6 loop are important for SERCA headpiece closure, and we predicted that these residues are key determinants of transport function.
In the present study, we tested this hypothesis with longer MD simulations and additional trajectory analyses, plus we performed physical experiments including functional assays of N-loop mutants. We also measured FRET as an index of overall headpiece conformation. These experiments exploited a "2-color SERCA" construct consisting of fluorescent protein tags fused to the N-and A-domains of cardiac SERCA2a (14,15). We previously used this 2-color construct as an intramolecular biosensor in high-throughput screening assays to identify small-molecule effectors that affect pump structure and activity (16). Here the 2-color construct was used to report the redistribution of the population of SERCA pumps among different conformational states and to directly test the proposed role of the N␤5-␤6 loop in the structural transitions of the Ca 2ϩ transport cycle.

All-atom molecular dynamics simulations of SERCA mutants
Our previous MD simulations of SERCA structural dynamics predicted that the open-to-closed transition of the SERCA cytosolic headpiece is facilitated by electrostatic and hydrogen bond interactions between basic and polar residues in the A-domain and acidic residues in the N-domain N␤5-␤6 loop (13). The contacts help to pierce a structured and poorly diffusible water layer between the domains. These residue-residue interactions also support a stable, ordered arrangement of the N-and A-domains during the transition from an open to a closed headpiece conformation. In silico triple mutation of the three negatively-charged loop residues to Ala abolished these stabilizing interactions and decreased the likelihood of spontaneous transition of the SERCA headpiece from an open architecture to a closed, compact structure, as assessed by 40-ns MD simulations (13). Here we extended these simulations to 100 ns, ran new simulations of single-point mutants of the SERCA N␤5-␤6 loop, and performed additional structural analyses on all trajectories. Fig. 1A shows the starting X-ray crystal structure of SERCA used for all-atom MD simulations (PDB accession code 1SU4) (17), including the A-, N-, P-, and TM-domains, plus the N␤5-␤6 loop of the N-domain and its three acidic residues, highlighted in orange. For triple mutant AAA (D426A/E429A/E435A) or single substitutions of N␤5-␤6 loop residues, we observed similar root mean square deviation trajectories in N-domain motions and insignificant differences in residue-specific root mean square fluctuations, and the angular autocorrelation of the cytoplasmic domains was not detectably different (data not shown); thus, the overall structure and dynamics of SERCA are not disrupted by single or triple mutations. Covariance matrices were only modestly different for WT-and AAA-SERCA, although there was a decrease in negatively correlated motions of the N-and A-domains (Fig. 1B, dotted box outlines) for AAA. The most pronounced effect of N␤5-␤6 loop mutation observed in the present MD simulations is a reduction in contacts between the N-and A-domains ( Fig. 1, C and D). In particular, substitution of three N␤5-␤6 loop residues (Asp-426, Glu-429, and Glu-435) to Ala decreased domain-domain contacts (Fig. 1, C and D, AAA), resulting in greater separation of the domains over the course of the trajectories (Fig. 1E, AAA). Fig. 1F reveals the general trend of a negative dependence of domain separation distance on the number of domain-domain contacts, whereby WT shows a short distance between N-and A-domains stabilized by a large number of contacts, whereas AAA shows a large domaindomain separation and fewer domain-domain contacts. Single-point mutants yielded intermediate values or were comparable with WT ( Fig. 1, C-F). The data are consistent with the proposed role of the three acidic N-loop residues in the opento-closed structural transition (13).

Comparison between structural ensembles of WT-SERCA and AAA-SERCA
We performed principal component analysis (PCA) of the ensembles of WT-and AAA-SERCA trajectories (n ϭ 6 each) and analyzed how the ensemble trajectory of each construct sampled the first two principal components. The most dominant motion PC1 (48% of all motions) was SERCA headpiece opening/closing, and the second major component PC2 (15% of all motions) was twisting of the cytosolic domains. Fig. 1G summarizes the positive and negative deflections of the A-and N-domains along components 1 and 2, with the origin (0, 0) representing the starting equilibrated structure. With respect to PC1, WT-SERCA sampled values ranging from a minimum of Ϫ36 Å (closed) to a maximum of ϩ25 Å (open) (Fig. 1H, black). AAA-SERCA trajectories (Fig. 1H, red) showed a similar range along PC1 (minimum of Ϫ30 Å and maximum of ϩ31 Å), but with relatively more population of open structures at the expense of closed conformations. This result is consistent with the observed increased minimum N-to A-domain-domain distance for AAA compared with WT ( Fig. 1, E and F). WT showed several trajectories that populated the extremes of the PC2 axis (twisting of the N-and A-domains) (Fig. 1H), ranging from Ϫ27 to ϩ23 Å, which is 56% greater than the range of AAA-SERCA along PC2 (from Ϫ14 to ϩ18 Å). Thus, the AAA mutant showed smaller twisting motions of the N-and A-domains (Fig. 1G). Overall, PCA indicates that the AAA loop mutations decrease the range of motions of SERCA cytosolic domains (with respect to PC2), and shift the population of structures toward a more open architecture (with respect to PC1).

The role of the N␤5-␤6 loop in SERCA ATPase and Ca 2؉ -transport activities
To determine the functional significance of the N␤5-␤6 loop in Ca 2ϩ -activated ATP hydrolysis, we prepared ER microsomes from HEK-293 cells expressing 2-color WT-SERCA or AAA-SERCA (Fig. 2, A-C). Fig. 2A shows an example experiment measuring Ca 2ϩ -activated ATP hydrolysis by SERCA con-Structural and cellular determinants of SERCA conformation structs. The apparent Ca 2ϩ affinity (K Ca ) of AAA-SERCA was not significantly different from that of WT-SERCA, yielding K Ca values of 7.0 Ϯ 0.2 and 7.1 Ϯ 0.3, respectively (n ϭ 4, p ϭ 0.61) (Fig. 2B). However, the maximal activity of AAA-SERCA was decreased by 63 Ϯ 5% (n ϭ 4, p ϭ 0.015) (Fig. 2C). This result is compatible with the previous finding by the Inesi lab that the single-point mutation of Asp-426 to Ala decreases SERCA ATPase activity similarly by 61% (18). Overall, the data demonstrate that the SERCA cycling rate is decreased by mutation of acidic loop residues, yet the mutated transporter still retains the ability to perform Ca 2ϩ -activated ATP hydrolysis with the same Ca 2ϩ affinity as WT-SERCA.
To determine the rate of Ca 2ϩ transport by the AAA mutant, we quantified intracellular Ca 2ϩ uptake in the ER of HEK-293 cells permeabilized with saponin ( Fig. 2, D-F). For each cell, Cer-SERCA expression (WT or AAA) was determined from the emission intensity of Cer fusion tag. To provide pharmacological control over ER Ca 2ϩ content, cells were co-transfected with the cardiac ryanodine receptor (RyR), an SR Ca 2ϩ channel that opens in response to application of caffeine (19). ER Ca 2ϩ content of HEK-293 cells was measured with a geneticallyencoded low-affinity Ca 2ϩ sensor R-CEPIA1er (20). The combination of a high level of Ca 2ϩ -transport activity from exogenous Cer-SERCA together with the ER Ca 2ϩ -load depen-

Structural and cellular determinants of SERCA conformation
dence of the exogenous RyR opening (21) results in spontaneous Ca 2ϩ release events, followed by refilling of the ER Ca 2ϩ stores by SERCA ( Fig. 2D: 8 release/uptake events between t ϭ 0 -15 s). Cells expressing AAA-SERCA showed similar average basal ER Ca 2ϩ content compared with WT, with few spontaneous Ca 2ϩ release events (Fig. 2D). Spontaneous Ca 2ϩ release was not observed in nontransfected cells (Ctrl) (Fig. 2D), as these cells express a low amount of native SERCA and lack endogenous RyR (22).
To determine the Ca 2ϩ uptake rate by SERCA pumps, caffeine (10 mM) was applied to empty the ER Ca 2ϩ stores (Fig. 2D, Caf). Once caffeine was washed out, the RyR inhibitor ruthenium red (10 M) was applied to stop Ca 2ϩ release and allow measurement of the rate of luminal [Ca 2ϩ ] ER recovery. At the end of each experiment, the R-CEPIA1er signal was calibrated by addition of the Ca 2ϩ ionophore ionomycin (Iono) (21). ER Ca 2ϩ recovery, monitored by [Ca 2ϩ ] ER accumulation, was analyzed to determine the maximum ER Ca 2ϩ uptake rate (i.e. SERCA transport rate) and maximum ER Ca 2ϩ load. ER Ca 2ϩ uptake in cells expressing WT-SERCA was determined to be 0.14 Ϯ 0.02 mM/s, a 47-fold increase over control cells (0.003 Ϯ 0.0005 mM/s) (Fig. 2E). AAA-SERCA Ca 2ϩ uptake rate was 0.05 Ϯ 0.007 mM/s, a 3-fold reduction in Ca 2ϩ transport rate versus WT. Furthermore, AAA-SERCA generated a ϳ25% lower maximal ER Ca 2ϩ load. The differences in Ca 2ϩ uptake rate and ER maximal load were not due to differential expression of WT-and AAA-SERCA, because the Cer fluorescence emission was similar for these two groups: 73 Ϯ 5.0 arbitrary units of WT and 86 Ϯ 11 arbitrary units in AAA-SERCA. The localization of 2-color SERCA was not significantly altered by the AAA mutation, as determined from confocal microscopy. Moreover, we observed similar levels of SERCA protein in ER microsomal preparations as evaluated by comparing exogenous Cer-labeled SERCA with endogenous SERCA by Western blotting. In HEK-293 cell microsomes, exogenous WT-SERCA expression was 68 Ϯ 7% of total SERCA, whereas AAA-SERCA expression was 78 Ϯ 7% of total SERCA. Ponceau staining of the blot also indicated similar amounts of WT-and AAA-SERCA, 6.8 Ϯ 0.6 and 6.5% Ϯ 0.7% of total protein, respectively. The data suggest that translation and localization of SERCA in the ER membrane was similar for WT and AAA. We conclude that AAA-SERCA exhibits lower Ca 2ϩ transport activity (65% less) than WT-SERCA (Fig. 2, D and E), which is consistent with the decreased ATPase rate by AAA-SERCA (63% inhibition) relative to WT (Fig. 2, A and C).

Quantification of 2-color SERCA FRET in ER microsomes from HEK-293 cells
We have previously used intermolecular FRET to quantify SERCA regulatory interactions with phospholamban and sarcolipin (23)(24)(25)(26)(27), and intramolecular FRET to detect SERCA structural transitions (14 -16). The latter experiments utilized a doubly labeled SERCA with fluorescent proteins fused to the Nand A-domains (2-color SERCA). Here we prepared microsomal membranes from cells expressing WT 2-color SERCA to quantify the FRET response to Ca 2ϩ binding using confocal fluorescence microscopy. Fig. 3A shows that FRET increased with Ca 2ϩ concentration, with an EC 50 of 1.25 Ϯ 0.22 M and a Hill coefficient (n) of 0.76. The apparent lack of cooperativity is compatible with previous studies that suggest the E2-E1 structural transition of the cytosolic headpiece is complete after binding of the first Ca 2ϩ to transport site I (28 -33). Overall, the data are consistent with our previous observation that ionophore treatment of HEK-293 cells expressing 2-color SERCA caused accumulation of 2-color SERCA in a high FRET state over the course of a few minutes (14).

Structural and cellular determinants of SERCA conformation
SERCA conformational states ("conformers") were stabilized with substrates to characterize the enzymatic intermediates of the Ca 2ϩ transport cycle (9). Importantly, whereas the state designations applied here are widely used in the field to describe the biochemical states stabilized by particular conditions, it is likely that significant structural heterogeneity exists for all ligand-stabilized biochemical states (34). For example, our previous time-resolved fluorescence measurements have shown that SERCA bound to the inhibitor TG can sample at least two major conformations (15), even though X-ray crystallography and EM have identified only one structural state of SERCA bound to TG (35,36). The present measurements capture the average FRET of the population ensemble.
We observed generally low FRET for 2-color SERCA in HEK-293 microsomes in solution conditions under which E2 conformers (low Ca 2ϩ affinity) are expected to predominate (Fig.  3B). For example, E2 (protonated) SERCA yielded ϳ10% FRET, as did E2-thapsigargin (E2-TG), a potent inhibitor that locks SERCA in the calcium-free E2 state (37,38). E2-AlF 4 Ϫ and E2-V i biochemical intermediates (analogs of the E2P phosphoenzyme intermediate) also showed low FRET (ϳ8%) (Fig.  3B). In contrast, we observed generally high FRET (ϳ15%) for 2-color SERCA in microsomes in solution conditions under which E1 conformers (high Ca 2ϩ affinity) are expected to predominate, such E1-ATP, E1-2Ca, 2Ca-AMP-PCP, and E1-2Ca-ADP-AlF 4 Ϫ (Fig. 3B). The FRET values for each state are summarized in Table 1, and these results provide the basis for quantitative analysis of the population distribution of SERCA in HEK-293 cells (see next section). Fluorescent protein separation distances (calculated from measured FRET efficiency) correlated well with distances measured from X-ray crystal structures (Fig. 3C), with the exception of the E1-2Ca crystal structure (PDB code 1SU4). We conclude that SERCA E2 conformers have more open headpiece structures (lower FRET), whereas E1 conformers have more closed headpiece structures (higher FRET).

SERCA structural dynamics in HEK-293 cells
For initial investigation of SERCA structural changes in response to Ca 2ϩ elevations in live cells, we quantified intramolecular FRET of 2-color SERCA by widefield fluorescence microscopy. Compared with WT, 2-color AAA-SERCA showed decreased basal FRET (Fig. 4A) in live HEK-293 cells, indicating a more open headpiece architecture compared with WT. We regard this result as consistent with our MD simulations, which showed fewer contacts between N-and A-domains for AAA-SERCA (Fig. 1D). To a lesser degree, the single-point mutations of acidic N␤5-␤6 loop residues also reduced FRET in live HEK-293 cells (Fig. 4A). The more moderate phenotypes of D426A, E429A, and E435A are also compatible with MD simulations, which showed domain contacts of the point mutants are intermediate between WT and AAA, or close to WT (Fig. 1D).
2-Color SERCA FRET is markedly decreased with binding to TG (14), and here we found that all of the mutants, including AAA-SERCA, responded normally to this SERCA inhibitor (Fig. 4B). Increasing intracellular Ca 2ϩ by application of ionomycin to cells expressing WT 2-color SERCA resulted in a biphasic response (Fig. 4C, black trace), with a rapid decrease in FRET (phase 1), followed by a slower increase in FRET (phase 2), which recovers to around the initial high FRET value (pre-Iono). Interestingly, the point mutants all showed a normal two-phase FRET response to Ca 2ϩ (like WT) (Fig. 4C, blue, green, and magenta traces). However, mutation of all three acidic loop residues (AAA-SERCA) abolished the slow FRET increase in phase 2 (Fig. 4C, red trace). Proposed mechanistic origins of the two-phase response are detailed under "Discussion." Overall, the widefield fluorescence microscopy data are consistent with the MD simulation analyses, showing that mutation of one negative residue is moderately tolerated, but that  6 experiments). B, FRET of WT 2-color SERCA stabilized in key enzymatic states. *, p Յ 0.008 compared with H ϩ . C, calculated FRET distances compared with distances between the fluorescent protein fusion sites measured from X-ray crystal structures. Select structures are labeled for comparison, other data are identified in Table 1. Table 1 FRET values for SERCA E1 and E2 conformers Intramolecular FRET was measured in ER microsomes from HEK-293 cells expressing 2-color SERCA. Yellow highlights E1 states, while green highlights E2 states. Data represent at least 5 independent measurements. FRET distances are compared to distances measured between residues 1 and 510 (fusion sites for fluorescent protein tags) in the corresponding X-ray crystal structure.

Structural and cellular determinants of SERCA conformation
mutation of three acidic residues from the N␤5-␤6 loop impairs N-to A-domain contacts (Fig. 1D), thereby decreasing the likelihood of forming a more compact headpiece conformation. Importantly, the observation that AAA-SERCA can still respond to TG in live cells (Fig. 3B) indicates that the protein structure is intact and TG-induced changes in headpiece structural dynamics are preserved after loop mutation. Thus, we propose that the impaired second-phase of Ca 2ϩ response and the reduced activity of AAA-SERCA are not due to a gross structural defect (such as misfolding), but instead due to the lack of the N-domain structural determinant that is responsible for the stable interaction with the A-domain during the SERCA headpiece transition from an open-to-closed conformation. These FRET experiments serve as the foundation for assessing the redistribution of SERCA calcium pump conformers during dynamic intracellular Ca 2ϩ signaling.

SERCA structural dynamics in response to changes in intracellular Ca 2؉ concentration
We have previously quantified FRET in cardiac myocytes (15,39,40), but motion artifacts of actively contracting cells make quantification of dynamic changes in fluorescence challenging. As an alternative, we reconstituted aspects of muscle cell Ca 2ϩ handling with co-expression of RyR2 and SERCA2a in live HEK-293 cells, and subjected the proteins to confocal microscopy.
In these experiments, Ca 2ϩ in the cytosol or ER lumen was monitored with X-Rhod (X-Rhod-1/AM) or R-CEPIA1er, respectively. SERCA conformational changes were quantified by excitation of Cer at 458 nm and observed as anti-correlated changes in the fluorescence intensities of Cer and enhanced yellow fluorescent protein (YFP) (Fig. 5A). The ratio of YFP/Cer fluorescence was taken as a measure of relative intramolecular FRET in the SERCA headpiece (Fig. 5, B, C, and E-L). Based on microsomal membrane experiments that showed high FRET for E1 states (Fig. 3B), and the eventual accumulation of a high FRET state after ionomycin treatment of cells (Fig. 4C), we anticipated that spontaneous Ca 2ϩ release events would be accompanied by increases in intramolecular FRET. Instead, Ca 2ϩ elevations in the cytosol corresponded to reductions in FRET, as indicated by decreases in the ratio of YFP/Cer fluorescence (Fig. 5B). Likewise, depletions of ER Ca 2ϩ occurred simultaneously with decreases in SERCA intramolecular FRET (Fig. 5C). We noted that the recovery of basal SERCA FRET was complete before full restoration of ER Ca 2ϩ stores, as shown by SERCA returning to the low FRET state in the middle of the sawtooth profile in ER Ca 2ϩ content detected by R-CEPIA1er (Fig. 5C). In contrast, cytosolic Ca 2ϩ elevations were square steps that closely mirrored SERCA FRET depressions (Fig. 5B). The result suggests that the ER continues to fill even after cytosolic Ca 2ϩ is already back to baseline and the population of SERCA has returned to the basal high FRET conformation. This may be due to store-operated Ca 2ϩ entry mechanisms that largely bypass the bulk cytoplasm.
The FRET fluctuations were not due to changes in intermolecular FRET between different SERCA molecules (39). Although we detected intermolecular FRET between Cer-SERCA and YFP-SERCA under these experimental conditions (Fig. 5D), intermolecular FRET did not change in response to ionomycin (Fig. 5E) or spontaneous Ca 2ϩ release events (Fig. 5F).
Despite the surprising response to intracellular Ca 2ϩ release, WT-SERCA intramolecular FRET still showed a biphasic response (decrease, then increase) in these confocal microscopy experiments after addition of ionomycin. Fig. 5G compares these contrasting results in a single trace: transient elevations of Ca 2ϩ resulted in transiently decreased FRET, but a sustained increase in Ca 2ϩ after ionomycin addition caused a lagging increase in FRET. As an alternative, we activated the co-expressed RyR with caffeine, and observed transiently increased cytosolic Ca 2ϩ followed by rapid equilibration to a low concentration of cytosolic Ca 2ϩ (Fig. 5H). Remarkably, this event was mirrored by a FRET change (decrease, then increase) that had similar kinetics.
In light of the poor responsiveness of AAA-SERCA to ionomycin-induced Ca 2ϩ influx (Fig. 4C, red), we were also surprised to find AAA-SERCA responded to spontaneous cytosolic Ca 2ϩ elevations (Fig. 5I) with no apparent deficit compared with WT (Fig. 5B), but again, Ca 2ϩ influx after ionomycin addition yielded a sustained decrease in FRET (Fig. 5J) instead of the biphasic FRET response seen for WT (Fig. 5G). The FRET response of AAA-SERCA to caffeine (Fig. 5K) was similar to that of WT-SERCA (Fig. 3H); a transient decrease followed by a sustained increase in FRET.

Discussion
In the present study we aimed to quantify SERCA dynamics during calcium signaling in live cells to examine the role of a

Structural and cellular determinants of SERCA conformation
discrete structural element that we previously hypothesized was an important determinant of SERCA headpiece structural transitions. Our new experiments confirm and extend our previous results (13). In particular, new MD experiments further defined the structural role of the acidic residues in the N␤5-␤6 loop, which is to establish inter-domain contacts that determine the range of domain movements and facilitate headpiece closure. The functional significance of this structural element is indicated by in vitro ATPase assays and live cell Ca 2ϩ uptake assays, which showed that mutations of key loop residues reduced ATPase activity and Ca 2ϩ transport kinetics. In addition, simultaneous FRET and Ca 2ϩ measurements provided insight into the dynamic redistribution of SERCA conformers in the physiologically relevant context of the cell (Fig. 5). To interpret the results of Ca 2ϩ dynamics experiments, we also quantified the average headpiece conformation of ligandstabilized enzymatic intermediates (Fig. 3B), which was useful for gauging the average conformation of the population of SERCA molecules for each biochemically defined intermediate state. Fig. 6A shows a simplified Post-Albers reaction cycle, with high Ca 2ϩ -affinity states highlighted in yellow and low Ca 2ϩaffinity states highlighted in green. The relative distribution of SERCA molecules among the conformational states depends on whether ligands such as Ca 2ϩ are abundant or limiting. Importantly, when ligands are not limiting, the distribution of states depends on the relative kinetics of partial reactions (41), with an increased population of states preceding slow steps. Major physiological states and transitions are shown in black (Fig. 6A), plus an additional nonphysiological state (TG-inhibited) and an alternate pathway (Ca 2ϩ binding prior to ATP) are shown in gray. States with similar headpiece conformations are grouped in blue boxes annotated with the FRET efficiency observed for those conformations in steady-state experiments (Fig. 3). In this context, we may interpret the observed changes in FRET in live cells.

Intracellular Ca 2؉ signaling and SERCA conformational changes
Redistribution of SERCA conformers in response to changes in the concentration of cytosolic Ca 2ϩ during signaling is summarized in the schematic diagram of Fig. 6B. When cells are at rest (e.g. nonstimulated HEK-293 cells or noncontracting myocytes), Ca 2ϩ is low (ϳ100 -200 nM) and ATP is saturating (3-5 mM), so the major populated conformation of SERCA is E1-ATP. Pre-bound ATP shifts the Ca 2ϩ binding E2-E1 (apo) . L, addition of ionomycin causes an increase in ER Ca 2ϩ content, but this increase occurs more slowly than the second phase of the observed FRET response of 2-color SERCA. We conclude that the phase 2 FRET increase of 2-color SERCA is not due to saturation of SERCA luminal Ca 2ϩ -binding sites (i.e. low affinity E2 orientation).

Structural and cellular determinants of SERCA conformation
population toward the high affinity E1 (apo)-ATP conformer ready to bind Ca 2ϩ (42)(43)(44). This state exhibited high FRET in vitro, and indeed we observed high FRET for WT 2-color SERCA in HEK-293 cells in the basal state. Then, when Ca 2ϩ is released from intracellular stores into the cytosol, FRET is transiently decreased (Fig. 6B). This FRET change occurs as a consequence of redistribution of the population of transporters to more open headpiece conformations. Specifically, at high Ca 2ϩ the pumps are continuously cycling, and all states of catalytic cycle are populated, with a relative build-up of E1P-2Ca and E2 conformers before the slow interconversion step of the transport sites, plus the E2P conformer before the slow phosphoenzyme hydrolysis and release (Fig. 6A) (32, 45-48). For SERCA, these are isomerization steps, the E2 to E1 transition and the rate-limiting transition from E1P to E2P (41). Therefore, when Ca 2ϩ increases to micromolar concentrations the population shifts from being predominantly E1-ATP (high FRET) to a predominant mixture of E1P-2Ca (high FRET) and E2 (low FRET) (Fig. 6A) (32, 45-48). The result of accumulation of the low FRET E2 population decreased overall FRET. Such conditions occur during contractions of cardiac or skeletal muscle cells, or in nonmuscle cells during intracellular Ca 2ϩ waves (as in Figs. 2D and 5B). Termination of Ca 2ϩ release and restoration of basal Ca 2ϩ returns the pumps to the resting condition, in which the majority conform to E1-ATP. Thus, the redistribution of SERCA conformers during Ca 2ϩ signaling is somewhat counterintuitive: the E1-ATP state prevails during periods of low Ca 2ϩ , and the E2 state population increases at high Ca 2ϩ .
Likewise, bottlenecks in catalytic cycle account for the rapid decrease in FRET immediately after ionomycin addition to HEK-293 cells (Fig. 6B, phase 1) as the population of transporters redistributes to a mixture of E1P/E2. This is the point where the WT and mutant SERCA diverge: Fig. 6B shows that WT-SERCA FRET rebounds (phase 2), whereas AAA-SERCA remains in predominantly low FRET conformations (Fig. 6B,  dotted line). The phase 2 increase in FRET for WT-SERCA was remarkable because cytosolic Ca 2ϩ remained elevated at millimolar concentrations for the remainder of the experiment. Several possible mechanisms could account for this result. We considered the possibility of alkalinization of the cell through Ca 2ϩ /H ϩ exchange by ionomycin (49). If this is the origin of the apparent second phase of the FRET change it must be due specifically to a change in SERCA conformation as opposed to a direct effect of pH on the fluorescent protein tags, because control experiments with Cer-SERCA and YFP-SERCA showed no change in intermolecular FRET with ionomycin addition (Fig. 5E). We do not attribute the phase 2 FRET change to saturation of SERCA luminal Ca 2ϩ binding and accumulation of SERCA in the E2-2Ca state, because accumulation of Ca 2ϩ in the ER after ionomycin treatment was much slower than the phase 2 FRET change (Fig. 5L). Another possible mechanism for the phase 2 FRET change is that full activation of SERCA in the ionomycin-treated cells may deplete ATP and increase ADP over the course of several minutes, with consequent accumulation of the population of 2-color SERCA in ATP-free, high FRET conformations. In vitro measurements revealed E1-2Ca and E1-2Ca-ADP-AlF 4 Ϫ to be a high (15%) FRET conformations (Fig. 3B). Finally, SERCA cycling is inhibited at very high (mM) Ca 2ϩ concentrations (50, 51), a condition that stabilizes the physiologically rare E1-2Ca state (52). Thus, we propose that the high FRET observed upon ionomycin-induced Ca 2ϩ release (Figs. 5G and 6A) is due to stabilization of E1-2Ca (branched gray pathway in Fig. 6A). Interestingly, a recent molecular dynamics study of SERCA at supraphysiologic Ca 2ϩ concentration (10 mM) indicated possible Ca 2ϩ binding to the N␤5-␤6 loop (53).

The effect of N␤5-␤6 loop mutations
The failure of AAA-SERCA to undergo a phase 2 redistribution to high FRET states (Fig. 6B, dotted line) illustrates the consequences of mutation of acidic residues in the N␤5-␤6 loop for SERCA dynamics. We observed that the triple mutation AAA impairs SERCA cytosolic headpiece closure in silico and we attribute the lack of a phase 2 FRET change to the decreased kinetics of structural transition from low FRET states (open headpiece) to high FRET states (closed headpiece). Thus, loss of N-to A-domain contacts results in a new rate-limiting step in the SERCA catalytic cycle, the E2 to E1-ATP transition (Fig. 6A). Although the mutant pump is functional (Fig. 2), there is much greater accumulation of the population of cycling transporters in E2 when there is a sustained elevation of Ca 2ϩ , so FRET remains low, with no second phase increase.

Structural and cellular determinants of SERCA conformation
Importantly, impaired headpiece dynamics did not prevent apparently normal AAA-SERCA FRET changes during spontaneous Ca 2ϩ release events and subsequent Ca 2ϩ re-uptake into the ER lumen (Fig. 5I). In particular, we did not detect any apparent delay in the upstroke of the FRET signal that occurs concomitantly with the decrease in cytosolic Ca 2ϩ . This surprising phenomenon may now be understood as another indication that SERCA function is decreased, but not abolished, by loop mutations. Termination of Ca 2ϩ release allows resequestration of Ca 2ϩ by the combined efforts of AAA-SERCA and endogenous SERCA. After withdrawal of Ca 2ϩ , flux through the transport cycle ceases, permitting return of AAA-SERCA back to the high FRET basal conformation, E1-ATP. The kinetic delay due to disruption of structural determinants in the loop is not resolvable as a slow FRET transition on this timescale; it is only detectable as a bottleneck step in actively cycling AAA-SERCA.

Comparison with other structural studies
The present results are in harmony with X-ray crystallography of SERCA stabilized by ligands in different biochemical states, with relatively open crystal structures (PDB 1IWO (54), 3W5C (55), 5A3Q (56), and 2O9J (57)) corresponding to low FRET states observed here and compact crystal structures (PDB 1VFP (58), 1T5S (11), and 2Z9R (60)) corresponding to high FRET states (Table 1). A notable exception is the crystal structure of E1-2Ca (PDB 1SU4) (17), which showed a widely open headpiece conformation. We measured high FRET values for this state in vitro and under conditions that should stabilize this state in cells, consistent with a closed headpiece. Fig. 3C shows the close correlation of the calculated FRET distance and the distance between fluorescent protein fusion sites, as measured from X-ray structures. The regression of these data (excluding outlier 1SU4) yielded a y intercept of 36 Å. We attribute this offset to the additional distance conferred by the difference between the fusion site and the chromaphore at the center of the fluorescent protein ␤-barrel.
A recent single-molecule FRET study of a related ATPase revealed a similar trend of high FRET between A-and P-domains for E1 conformers and low FRET for A-and P-domains for E2 conformers (61). Although it is difficult to compare FRET measurements taken from different labeling sites, overall the studies are in harmony. Both suggest E1 conformations have a compact, ordered headpiece, whereas E2 states are characterized by an open, disordered architecture.

Summary
The goals of this study were 2-fold. The first was to quantify how the overall conformation of the SERCA cytosolic headpiece changes as the transporter steps through the structural transitions of the catalytic cycle. Second, we sought to test directly the hypothesis that residues in a loop of the SERCA N-domain are key determinants of transport function (13). The present results are compatible with this proposed mechanism, as mutations of the loop residues resulted in altered headpiece dynamics, and functional measurements revealed a consequent decrease in ATP hydrolysis rate and Ca 2ϩ transport. The results support the proposed role of the loop in facilitating SERCA headpiece closure during functional enzymatic cycling. As a discrete structural element, the N-loop may be a worthwhile target for development of small molecules to enhance (16) or inhibit (62)(63)(64) SERCA function in vivo.

All-atom molecular dynamics simulations
All-atom MD simulations were performed as described (13). Briefly, the GROMACS software package (65,66) with CHARMM 27 forcefield (67) and TIP3P water model (68) were used to carry out MD simulations. The reference Ca 2ϩ -bound crystal structure of SERCA (17) was used to run WT simulations and to introduce N␤5-␤6 loop mutations. For all constructs, energy minimization was performed using the steepest descent method for 1000 steps. Then the structures were embedded into a POPC lipid bilayer and solvated in a rectangular water box with dimension sizes 130 ϫ 130 ϫ 160 Å. Na ϩ and Cl Ϫ ions were added to the solution to a concentration of 150 mM. The Berendesen method (69) with relaxation time of 0.1 ps was used to increase the temperature of the system to 300 K and reach the pressure of 1 bar. After 1-ns equilibration, the production run was performed in the NPT assemble using the Nose-Hoover thermostat (70,71) and the Parrinello-Rahman barostat (72) with a relaxation time of 1 ps. Six independent production runs for each WT or mutant construct were started with a different set of assigned velocities at 300 K. The integration time was 2 fs, and atom coordinates were saved every 1 ps. Production runs were carried out for 100 ns (n ϭ 6 for WT-and AAA-SERCA).

Structural analysis and visualization
The VMD program (73) was used for visualization and rendering snapshots. The GROMACS program (65,66) was used for the computational analysis of MD production runs. The N-to A-domain-domain distance was defined as the minimum distance between any atom of the A-domain (residues number 1-40 and 128 -241) and any atom of the N-domain (residues 360 -603). The number of contacts was calculated with a 4-Å cutoff between atoms of the N-and A-domains. One contact of a N-domain atom with multiple atoms of A-domain was counted as one contact (i.e. instead of multiple contacts). The first 10 ns of MD simulations were considered as equilibration time and not included in structural analyses.

Principal component analysis
To identify the major motions of the SERCA headpiece during MD trajectories, we aligned SERCA structures using the 10-helix TM domain as a reference and used PCA (74 -76). To compare structural ensembles with respect to the same eigenvectors, we combined three SERCA reference X-ray crystal structures (PDB 1SU4 (17), 3W5B (55) and 1VFP (58)), six MD trajectories of WT-SERCA, and six MD trajectories of AAA-SERCA into a single trajectory. To obtain sets of eigenvectors and eigenvalues corresponding to principal components, we Structural and cellular determinants of SERCA conformation built covariance matrixes of the atomic fluctuations in GRO-MACS (65,66). The diagonalization of matrixes yielded the eigenvectors (which are principal components) and their associated eigenvalues.

Molecular biology and cell culture
The engineering and functional characterization of 2-color SERCA was previously described (14 -16). We used a canine SERCA2a construct labeled with Cer on the N terminus and an YFP intrasequence tag inserted before residue 509 in the N- HEK-293 cells were cultured in Dulbecco's modified Eagle's medium cell culture medium supplemented with 10% fetal bovine serum (ThermoScientific, Waltham, MA) and transiently transfected using MBS mammalian transfection kit (Agilent Technologies, Stratagene), as described previously (14). The transfected cells were trypsinized (Thermo-Scientific) and replated onto poly-D-lysine-coated glass bottom chambers and allowed to adhere for 1-2 h prior to imaging.

Wide-field acceptor sensitization fluorescence microscopy to measure SERCA intramolecular FRET
Wide-field fluorescent microscopy was done as described previously (26). Briefly, cells were imaged with an inverted microscope (Nikon Eclipse TE2000-U) equipped with a metal halide lamp and a back-thinned CCD camera (iXon 887: Andor Technology, Belfast, Northern Ireland). For each sample, acquisition of field was performed with a ϫ60 1.49 N.A. objective with 100 or 150 ms exposure for each channel: Cer, YFP, and FRET. Fluorescence intensity was automatically quantified with a multiwavelength cell scoring application in MetaMorph software (Molecular Devices, Sunnyvale, CA). FRET efficiency was calculated according to (26,40), where F FRET , F YFP , and F Cer are the matching fluorescence intensities from FRET, YFP, and Cer images, respectively, and G represents FRET intensity corrected for the bleedthrough of the channels. The parameters a and d are bleedthrough constants calculated as a ϭ F FRET / F Cer for a control sample transfected with only YFP-SERCA and d ϭ F FRET /F Cer for a control sample transfected with only Cer-SERCA. These values were determined to be a ϭ 0.074 and b ϭ 0.70. Apparent probe separation distance (R) was calculated from FRET efficiency (E) according to the relationship (78), with a Förster distance R 0 of 49.8 Å for the mCer and EYFP FRET pair (77). The error of the distance measurement was estimated from the standard deviation of repeated FRET measurements.

Ratiometric confocal fluorescence microscopy to measure SERCA intramolecular FRET, HEK-293 cytosolic Ca 2؉ , and HEK-293 ER luminal Ca 2؉
HEK-293 cells were transiently co-transfected with expression plasmids containing cDNA of GFP-human ryanodine receptor-2 fusion protein (RyR), R-CEPIA1er, and 2-color canine SERCA2a wildtype (WT) or triple loop mutant AAA. Transfected cells were cultured for 24 h and seeded into poly-D-lysine-coated glass-bottom chamber slides in Dulbecco's modified Eagle's medium plus 10% fetal bovine serum. 24 h after seeding, cell culture medium was changed with PBS with Ca 2ϩ /Mg 2ϩ , and experiments were conducted with a Leica SP5 laser scanning confocal microscope equipped with a ϫ63 water objective. R-CEPIA1er was excited with the 543-nm line of a He-Ne laser, and emitted fluorescence was measured at wavelength Ն580 nm. 2-color SERCA fluorophores Cer and YFP were excited with the 430 and 514 nm lines of an argon laser, respectively, and emitted fluorescence was measured at wavelengths 485 Ϯ 15 and 537 Ϯ 15 nm, respectively. Images were acquired in line-scan mode for up to 8 -12 min with addition of 10 mM caffeine or 100 M ionomycin at time (t) indicated in the figures. Ionomycin powder (Sigma) was dissolved in DMSO to make a 13.3 mM stock solution, which was used to prepare 2ϫ Iono solution (200 M) in PBS. The final concentration of DMSO applied to cells was 0.75%. Fluorescence image analysis was performed with ImageJ software (79).

Structural and cellular determinants of SERCA conformation
itor mixture); and 4) passed through a 27-gauge needle 10 times. To prepare microsomal membranes, 1) cell homogenates were centrifuged at 1,000 ϫ g for 10 min at 4°C; 2) the low-speed supernatants were centrifuged at 126,000 ϫ g for 30 min at 4°C; 3) the high-speed membrane pellets were resuspended in a 1:1 mixture of homogenizing and sucrose solutions; and 4) the membrane suspensions were passed through a 27-gauge needle 10 times. The protein concentration of microsomal membranes were determined with a Pierce BCA assay kit (ThermoScientific). ATPase assay was performed the same day as membrane preparation; otherwise, microsomal membranes were aliquoted, snap-frozen in liquid nitrogen, and stored at Ϫ80°C.

Measuring Ca 2؉ -ATPase activity in ER microsomes from HEK-293 cells
2-Color SERCA ATPase activity was measured in ER microsomes from HEK-293 cells by spectrophotometric determination of the rate of NADH consumption as a function of Ca 2ϩ concentration using the enzyme-coupled activity assay in 96-well plate (82,83). The time dependence of the absorbance decrease was measured at 340 nm at 25°C in a PHERAstar FSX microplate reader (BMG Labtech, Cary, NC). Each well contained 3-4 g of microsomal membranes in 200 l of assay solution containing 50 mM MOPS, pH 7.0, 100 mM KCl, 5.0 mM MgCl 2 , 1.0 mM EGTA, 2.5 mM ATP, 0.2 mM NADH, 1 U of pyruvate kinase, 1 U of lactate dehydrogenase, 0.5 mM phosphoenol pyruvate, and 0.7 g of Ca 2ϩ ionophore (A23187). Chemicals were obtained from Sigma. Free Ca 2ϩ concentrations were calculated using a Ca/Mg/ATP/EGTA calculator from Theo Schoenmakers' Chelator (59). Data were fitted using the Hill function, V ϭ V max /͓1 ϩ 10 Ϫn(pKCaϪpCa) ͔ (Eq. 2) where V is the ATPase rate at a specific Ca 2ϩ concentration (pCa), n is the Hill coefficient, pK Ca is the apparent Ca 2ϩ dissociation constant, and V max was obtained from the fit of the Hill equation at saturating Ca 2ϩ concentrations.

Measuring ER Ca 2؉ uptake in permeabilized HEK-293 cells
Changes in [Ca 2ϩ ] ER were measured with laser scanning confocal microscopy (Radiance 2000 MP, Bio-Rad) equipped with a ϫ40 oil-immersion objective lens (NA ϭ 1.3). R-CEPIA1er was excited with the 543 nm line of a He-Ne laser and fluorescence was measured at wavelengths Ͼ600 nm. Fluorescence measurements were acquired in line-scan mode (20 ms per scan; pixel size 0.12 m). HEK-293 cells co-transfected with GFP-RyR, R-CEPIA1er, and Cer-SERCA were washed in Ca 2ϩ -free solution containing 150 mM K-aspartate, 0.25 mM MgCl 2 , 0.1 mM EGTA, 10 mM HEPES, pH 7.2. The plasma membrane of HEK-293 cells was permeabilized with 0.005% saponin to control cytosolic environment replacement with a saponin-free solution containing 100 mM K-aspartate, 15 mM KCl, 5 mM KH 2 PO 4 , 5 mM MgATP, 0.35 mM EGTA, 0.12 CaCl 2 , 0.75 mM MgCl 2 , 10 mM phosphocreatine, 2% (w/v) dextran (M r 40,000), and 10 mM HEPES, pH 7.2 (KOH). Free [Ca 2ϩ ] and [Mg 2ϩ ] of this solution were calculated to be 250 nM and 1 mM, respec-tively. Activation of RyR with caffeine (10 mM) was used to completely deplete [Ca 2ϩ ] ER . Once caffeine was removed, the RyR inhibitor ruthenium red (RR ϭ 10 M) was applied to measure the rate of ER Ca 2ϩ uptake. Changes in [Ca 2ϩ ] ER were calculated by the formula: [Ca 2ϩ ] ER ϭ K d ϫ [(F Ϫ F min )/(F max Ϫ F)], where F is the R-CEPIA1er fluorescence; F max and F min are the fluorescence level at 10 mM Ca 2ϩ /Iono before and after depletion of ER Ca 2ϩ with caffeine (10 mM), respectively. The R-CEPIA1er Ca 2ϩ dissociation constant (K d ) is 390 M based on in situ calibrations (21). At the end of each experiment, the R-CEPIA1er signal (F max ) was calibrated with addition of Iono (100 M). ER Ca 2ϩ uptake (i.e. SERCA transport rate) was calculated as the time-dependent change of [Ca 2ϩ ] ER after RyR inhibition on a cell-to-cell basis (mM Ca 2ϩ /s), and the maximal ER Ca 2ϩ load was determined for each individual cell (mM Ca 2ϩ ). The reported Ca 2ϩ uptake rate and maximal ER Ca 2ϩ load were calculated as mean Ϯ S.D.

Intramolecular FRET measurements of 2-color SERCA expressed in ER microsomes from HEK-293 cells
To stabilize SERCA in ligand-stabilized biochemical intermediates, various solutions were prepared by addition of corresponding substrates to the calcium-free base solution, which includes 100 mM KCl, 5 mM MgCl 2 , 2 mM EGTA, and 10 mM imidazole, pH 7.0. The following ligands were used to prepare specific solutions corresponding to SERCA biochemical state (in parentheses): 100 M thapsigargin (E2-TG), 3 mM ATP (E1-ATP), 2. Ϫ ). Chemicals were obtained from Sigma.
To measure SERCA intramolecular FRET in ligand-stabilized biochemical intermediates, 1 l of membrane preparations (7-10 g of total protein) was mixed with 9 l of ligand solution on a coverslip and immediately imaged using confocal fluorescent microscopy, as described above.

Statistical analyses
Data are presented as the mean Ϯ S.D. of n Ն 3 experiments. All statistical tests were performed using OriginPro 9.1 (Origin-Lab Corporation, Northampton, MA). Student's t test was used to compare differences between two groups, and one-way analysis of variance was used to compare the difference between three or more groups. One-way analysis of variance was followed by Tukey's post hoc test. A probability (p) value of Ͻ0.